Index: projects/em_mq/Makefile.inc1 =================================================================== --- projects/em_mq/Makefile.inc1 (revision 283943) +++ projects/em_mq/Makefile.inc1 (revision 283944) @@ -1,2237 +1,2238 @@ # # $FreeBSD$ # # Make command line options: # -DNO_CLEANDIR run ${MAKE} clean, instead of ${MAKE} cleandir # -DNO_CLEAN do not clean at all # -DDB_FROM_SRC use the user/group databases in src/etc instead of # the system database when installing. # -DNO_SHARE do not go into share subdir # -DKERNFAST define NO_KERNEL{CONFIG,CLEAN,DEPEND,OBJ} # -DNO_KERNELCONFIG do not run config in ${MAKE} buildkernel # -DNO_KERNELCLEAN do not run ${MAKE} clean in ${MAKE} buildkernel # -DNO_KERNELDEPEND do not run ${MAKE} depend in ${MAKE} buildkernel # -DNO_KERNELOBJ do not run ${MAKE} obj in ${MAKE} buildkernel # -DNO_PORTSUPDATE do not update ports in ${MAKE} update # -DNO_ROOT install without using root privilege # -DNO_DOCUPDATE do not update doc in ${MAKE} update # -DWITHOUT_CTF do not run the DTrace CTF conversion tools on built objects # LOCAL_DIRS="list of dirs" to add additional dirs to the SUBDIR list # LOCAL_ITOOLS="list of tools" to add additional tools to the ITOOLS list # LOCAL_LIB_DIRS="list of dirs" to add additional dirs to libraries target # LOCAL_MTREE="list of mtree files" to process to allow local directories # to be created before files are installed # LOCAL_TOOL_DIRS="list of dirs" to add additional dirs to the build-tools # list # METALOG="path to metadata log" to write permission and ownership # when NO_ROOT is set. (default: ${DESTDIR}/METALOG) # TARGET="machine" to crossbuild world for a different machine type # TARGET_ARCH= may be required when a TARGET supports multiple endians # BUILDENV_SHELL= shell to launch for the buildenv target (def:/bin/sh) # WORLD_FLAGS= additional flags to pass to make(1) during buildworld # KERNEL_FLAGS= additional flags to pass to make(1) during buildkernel # # The intended user-driven targets are: # buildworld - rebuild *everything*, including glue to help do upgrades # installworld- install everything built by "buildworld" # doxygen - build API documentation of the kernel # update - convenient way to update your source tree (eg: svn/svnup) # # Standard targets (not defined here) are documented in the makefiles in # /usr/share/mk. These include: # obj depend all install clean cleandepend cleanobj .if !defined(TARGET) || !defined(TARGET_ARCH) .error "Both TARGET and TARGET_ARCH must be defined." .endif .include "share/mk/src.opts.mk" .include .include # We must do lib/ and libexec/ before bin/, because if installworld # installs a new /bin/sh, the 'make' command will *immediately* # use that new version. And the new (dynamically-linked) /bin/sh # will expect to find appropriate libraries in /lib and /libexec. # SRCDIR?= ${.CURDIR} .if defined(SUBDIR_OVERRIDE) SUBDIR= ${SUBDIR_OVERRIDE} .else SUBDIR= lib libexec SUBDIR+=bin .if ${MK_GAMES} != "no" SUBDIR+=games .endif .if ${MK_CDDL} != "no" SUBDIR+=cddl .endif SUBDIR+=gnu include .if ${MK_KERBEROS} != "no" SUBDIR+=kerberos5 .endif .if ${MK_RESCUE} != "no" SUBDIR+=rescue .endif SUBDIR+=sbin .if ${MK_CRYPT} != "no" SUBDIR+=secure .endif .if !defined(NO_SHARE) SUBDIR+=share .endif SUBDIR+=sys usr.bin usr.sbin .if ${MK_TESTS} != "no" SUBDIR+= tests .endif .if ${MK_OFED} != "no" SUBDIR+=contrib/ofed .endif # # We must do etc/ last for install/distribute to work. # SUBDIR+=etc # Local directories are last, since it is nice to at least get the base # system rebuilt before you do them. .for _DIR in ${LOCAL_DIRS} .if exists(${.CURDIR}/${_DIR}/Makefile) SUBDIR+= ${_DIR} .endif .endfor # Add LOCAL_LIB_DIRS, but only if they will not be picked up as a SUBDIR # of a LOCAL_DIRS directory. This allows LOCAL_DIRS=foo and # LOCAL_LIB_DIRS=foo/lib to behave as expected. .for _DIR in ${LOCAL_DIRS:M*/} ${LOCAL_DIRS:N*/:S|$|/|} _REDUNDENT_LIB_DIRS+= ${LOCAL_LIB_DIRS:M${_DIR}*} .endfor .for _DIR in ${LOCAL_LIB_DIRS} .if empty(_REDUNDENT_LIB_DIRS:M${_DIR}) && exists(${.CURDIR}/${_DIR}/Makefile) SUBDIR+= ${_DIR} .else .warning ${_DIR} not added to SUBDIR list. See UPDATING 20141121. .endif .endfor .endif .if defined(NOCLEAN) NO_CLEAN= ${NOCLEAN} .endif .if defined(NO_CLEANDIR) CLEANDIR= clean cleandepend .else CLEANDIR= cleandir .endif LOCAL_TOOL_DIRS?= BUILDENV_SHELL?=/bin/sh SVN?= /usr/local/bin/svn SVNFLAGS?= -r HEAD MAKEOBJDIRPREFIX?= /usr/obj .if !defined(OSRELDATE) .if exists(/usr/include/osreldate.h) OSRELDATE!= awk '/^\#define[[:space:]]*__FreeBSD_version/ { print $$3 }' \ /usr/include/osreldate.h .else OSRELDATE= 0 .endif .endif .if !defined(VERSION) REVISION!= ${MAKE} -C ${SRCDIR}/release -V REVISION BRANCH!= ${MAKE} -C ${SRCDIR}/release -V BRANCH SRCRELDATE!= awk '/^\#define[[:space:]]*__FreeBSD_version/ { print $$3 }' \ ${SRCDIR}/sys/sys/param.h VERSION= FreeBSD ${REVISION}-${BRANCH:C/-p[0-9]+$//} ${TARGET_ARCH} ${SRCRELDATE} .endif KNOWN_ARCHES?= aarch64/arm64 amd64 arm armeb/arm armv6/arm armv6hf/arm i386 i386/pc98 mips mipsel/mips mips64el/mips mips64/mips mipsn32el/mips mipsn32/mips powerpc powerpc64/powerpc sparc64 .if ${TARGET} == ${TARGET_ARCH} _t= ${TARGET} .else _t= ${TARGET_ARCH}/${TARGET} .endif .for _t in ${_t} .if empty(KNOWN_ARCHES:M${_t}) .error Unknown target ${TARGET_ARCH}:${TARGET}. .endif .endfor .if ${TARGET} == ${MACHINE} TARGET_CPUTYPE?=${CPUTYPE} .else TARGET_CPUTYPE?= .endif .if !empty(TARGET_CPUTYPE) _TARGET_CPUTYPE=${TARGET_CPUTYPE} .else _TARGET_CPUTYPE=dummy .endif _CPUTYPE!= MAKEFLAGS= CPUTYPE=${_TARGET_CPUTYPE} ${MAKE} \ -f /dev/null -m ${.CURDIR}/share/mk -V CPUTYPE .if ${_CPUTYPE} != ${_TARGET_CPUTYPE} .error CPUTYPE global should be set with ?=. .endif .if make(buildworld) BUILD_ARCH!= uname -p .if ${MACHINE_ARCH} != ${BUILD_ARCH} .error To cross-build, set TARGET_ARCH. .endif .endif .if ${MACHINE} == ${TARGET} && ${MACHINE_ARCH} == ${TARGET_ARCH} && !defined(CROSS_BUILD_TESTING) OBJTREE= ${MAKEOBJDIRPREFIX} .else OBJTREE= ${MAKEOBJDIRPREFIX}/${TARGET}.${TARGET_ARCH} .endif WORLDTMP= ${OBJTREE}${.CURDIR}/tmp BPATH= ${WORLDTMP}/legacy/usr/sbin:${WORLDTMP}/legacy/usr/bin:${WORLDTMP}/legacy/bin XPATH= ${WORLDTMP}/usr/sbin:${WORLDTMP}/usr/bin STRICTTMPPATH= ${BPATH}:${XPATH} TMPPATH= ${STRICTTMPPATH}:${PATH} # # Avoid running mktemp(1) unless actually needed. # It may not be functional, e.g., due to new ABI # when in the middle of installing over this system. # .if make(distributeworld) || make(installworld) INSTALLTMP!= /usr/bin/mktemp -d -u -t install .endif # # Building a world goes through the following stages # # 1. legacy stage [BMAKE] # This stage is responsible for creating compatibility # shims that are needed by the bootstrap-tools, # build-tools and cross-tools stages. These are generally # APIs that tools from one of those three stages need to # build that aren't present on the host. # 1. bootstrap-tools stage [BMAKE] # This stage is responsible for creating programs that # are needed for backward compatibility reasons. They # are not built as cross-tools. # 2. build-tools stage [TMAKE] # This stage is responsible for creating the object # tree and building any tools that are needed during # the build process. # 3. cross-tools stage [XMAKE] # This stage is responsible for creating any tools that # are needed for building the system. A cross-compiler is one # of them. # 4. world stage [WMAKE] # This stage actually builds the world. # 5. install stage (optional) [IMAKE] # This stage installs a previously built world. # BOOTSTRAPPING?= 0 # Common environment for world related stages CROSSENV= MAKEOBJDIRPREFIX=${OBJTREE} \ MACHINE_ARCH=${TARGET_ARCH} \ MACHINE=${TARGET} \ CPUTYPE=${TARGET_CPUTYPE} .if ${MK_GROFF} != "no" CROSSENV+= GROFF_BIN_PATH=${WORLDTMP}/legacy/usr/bin \ GROFF_FONT_PATH=${WORLDTMP}/legacy/usr/share/groff_font \ GROFF_TMAC_PATH=${WORLDTMP}/legacy/usr/share/tmac .endif .if defined(TARGET_CFLAGS) CROSSENV+= ${TARGET_CFLAGS} .endif # bootstrap-tools stage BMAKEENV= INSTALL="sh ${.CURDIR}/tools/install.sh" \ PATH=${BPATH}:${PATH} \ WORLDTMP=${WORLDTMP} \ VERSION="${VERSION}" \ MAKEFLAGS="-m ${.CURDIR}/tools/build/mk ${.MAKEFLAGS}" BMAKE= MAKEOBJDIRPREFIX=${WORLDTMP} \ ${BMAKEENV} ${MAKE} ${WORLD_FLAGS} -f Makefile.inc1 \ DESTDIR= \ BOOTSTRAPPING=${OSRELDATE} \ SSP_CFLAGS= \ MK_HTML=no NO_LINT=yes MK_MAN=no \ -DNO_PIC MK_PROFILE=no -DNO_SHARED \ -DNO_CPU_CFLAGS MK_WARNS=no MK_CTF=no \ MK_CLANG_EXTRAS=no MK_CLANG_FULL=no \ MK_LLDB=no MK_TESTS=no \ MK_INCLUDES=yes # build-tools stage TMAKE= MAKEOBJDIRPREFIX=${OBJTREE} \ ${BMAKEENV} ${MAKE} ${WORLD_FLAGS} -f Makefile.inc1 \ TARGET=${TARGET} TARGET_ARCH=${TARGET_ARCH} \ DESTDIR= \ BOOTSTRAPPING=${OSRELDATE} \ SSP_CFLAGS= \ -DNO_LINT \ -DNO_CPU_CFLAGS MK_WARNS=no MK_CTF=no \ MK_CLANG_EXTRAS=no MK_CLANG_FULL=no \ MK_LLDB=no MK_TESTS=no # cross-tools stage XMAKE= TOOLS_PREFIX=${WORLDTMP} ${BMAKE} \ TARGET=${TARGET} TARGET_ARCH=${TARGET_ARCH} \ MK_GDB=no MK_TESTS=no # kernel-tools stage KTMAKEENV= INSTALL="sh ${.CURDIR}/tools/install.sh" \ PATH=${BPATH}:${PATH} \ WORLDTMP=${WORLDTMP} \ VERSION="${VERSION}" KTMAKE= TOOLS_PREFIX=${WORLDTMP} MAKEOBJDIRPREFIX=${WORLDTMP} \ ${KTMAKEENV} ${MAKE} ${WORLD_FLAGS} -f Makefile.inc1 \ DESTDIR= \ BOOTSTRAPPING=${OSRELDATE} \ SSP_CFLAGS= \ MK_HTML=no -DNO_LINT MK_MAN=no \ -DNO_PIC MK_PROFILE=no -DNO_SHARED \ -DNO_CPU_CFLAGS MK_WARNS=no MK_CTF=no # world stage WMAKEENV= ${CROSSENV} \ _SHLIBDIRPREFIX=${WORLDTMP} \ _LDSCRIPTROOT= \ VERSION="${VERSION}" \ INSTALL="sh ${.CURDIR}/tools/install.sh" \ PATH=${TMPPATH} # make hierarchy HMAKE= PATH=${TMPPATH} ${MAKE} LOCAL_MTREE=${LOCAL_MTREE:Q} .if defined(NO_ROOT) HMAKE+= PATH=${TMPPATH} METALOG=${METALOG} -DNO_ROOT .endif .if ${MK_CDDL} == "no" WMAKEENV+= MK_CTF=no .endif .if defined(CROSS_TOOLCHAIN) LOCALBASE?= /usr/local .include "${LOCALBASE}/share/toolchains/${CROSS_TOOLCHAIN}.mk" .endif .if defined(CROSS_TOOLCHAIN_PREFIX) CROSS_COMPILER_PREFIX?=${CROSS_TOOLCHAIN_PREFIX} CROSS_BINUTILS_PREFIX?=${CROSS_TOOLCHAIN_PREFIX} .endif # If we do not have a bootstrap binutils (because the in-tree one does not # support the target architecture), provide a default cross-binutils prefix. # This allows aarch64 builds, for example, to automatically use the # aarch64-binutils port or package. .if !empty(BROKEN_OPTIONS:MBINUTILS_BOOTSTRAP) && \ !defined(CROSS_BINUTILS_PREFIX) CROSS_BINUTILS_PREFIX=/usr/local/${TARGET_ARCH}-freebsd/bin/ .if !exists(${CROSS_BINUTILS_PREFIX}) .error In-tree binutils does not support the ${TARGET_ARCH} architecture. Install the ${TARGET_ARCH}-binutils port or package or set CROSS_BINUTILS_PREFIX. .endif .endif XCOMPILERS= CC CXX CPP .for COMPILER in ${XCOMPILERS} .if defined(CROSS_COMPILER_PREFIX) X${COMPILER}?= ${CROSS_COMPILER_PREFIX}${${COMPILER}} .else X${COMPILER}?= ${${COMPILER}} .endif .endfor XBINUTILS= AS AR LD NM OBJCOPY OBJDUMP RANLIB SIZE STRINGS .for BINUTIL in ${XBINUTILS} .if defined(CROSS_BINUTILS_PREFIX) && \ exists(${CROSS_BINUTILS_PREFIX}${${BINUTIL}}) X${BINUTIL}?= ${CROSS_BINUTILS_PREFIX}${${BINUTIL}} .else X${BINUTIL}?= ${${BINUTIL}} .endif .endfor WMAKEENV+= CC="${XCC} ${XCFLAGS}" CXX="${XCXX} ${XCFLAGS} ${XCXXFLAGS}" \ DEPFLAGS="${DEPFLAGS}" \ CPP="${XCPP} ${XCFLAGS}" \ AS="${XAS}" AR="${XAR}" LD="${XLD}" NM=${XNM} \ OBJDUMP=${XOBJDUMP} OBJCOPY="${XOBJCOPY}" \ RANLIB=${XRANLIB} STRINGS=${XSTRINGS} \ SIZE="${XSIZE}" .if ${XCC:M/*} .if defined(CROSS_BINUTILS_PREFIX) # In the case of xdev-build tools, CROSS_BINUTILS_PREFIX won't be a # directory, but the compiler will look in the right place for it's # tools so we don't need to tell it where to look. .if exists(${CROSS_BINUTILS_PREFIX}) BFLAGS+= -B${CROSS_BINUTILS_PREFIX} .endif .else BFLAGS+= -B${WORLDTMP}/usr/bin .endif .if ${TARGET} == "arm" .if ${TARGET_ARCH:M*hf*} != "" TARGET_ABI= gnueabihf .else TARGET_ABI= gnueabi .endif .endif .if defined(X_COMPILER_TYPE) && ${X_COMPILER_TYPE} == gcc XCFLAGS+= -isystem ${WORLDTMP}/usr/include -L${WORLDTMP}/usr/lib XCXXFLAGS+= -I${WORLDTMP}/usr/include/c++/v1 -std=gnu++11 -L${WORLDTMP}/../lib/libc++ DEPFLAGS+= -I${WORLDTMP}/usr/include/c++/v1 .else TARGET_ABI?= unknown TARGET_TRIPLE?= ${TARGET_ARCH:C/amd64/x86_64/}-${TARGET_ABI}-freebsd11.0 XCFLAGS+= -target ${TARGET_TRIPLE} .endif XCFLAGS+= --sysroot=${WORLDTMP} ${BFLAGS} XCXXFLAGS+= --sysroot=${WORLDTMP} ${BFLAGS} .else .if defined(CROSS_BINUTILS_PREFIX) && exists(${CROSS_BINUTILS_PREFIX}) BFLAGS+= -B${CROSS_BINUTILS_PREFIX} XCFLAGS+= ${BFLAGS} XCXXFLAGS+= ${BFLAGS} .endif .endif # ${XCC:M/*} WMAKE= ${WMAKEENV} ${MAKE} ${WORLD_FLAGS} -f Makefile.inc1 DESTDIR=${WORLDTMP} .if ${TARGET_ARCH} == "amd64" || ${TARGET_ARCH} == "powerpc64" # 32 bit world LIB32_OBJTREE= ${OBJTREE}${.CURDIR}/world32 LIB32TMP= ${OBJTREE}${.CURDIR}/lib32 .if ${TARGET_ARCH} == "amd64" .if empty(TARGET_CPUTYPE) LIB32CPUFLAGS= -march=i686 -mmmx -msse -msse2 .else LIB32CPUFLAGS= -march=${TARGET_CPUTYPE} .endif LIB32WMAKEENV= MACHINE=i386 MACHINE_ARCH=i386 \ MACHINE_CPU="i686 mmx sse sse2" LIB32WMAKEFLAGS= \ AS="${XAS} --32" \ LD="${XLD} -m elf_i386_fbsd -Y P,${LIB32TMP}/usr/lib32" \ OBJCOPY="${XOBJCOPY}" .elif ${TARGET_ARCH} == "powerpc64" .if empty(TARGET_CPUTYPE) LIB32CPUFLAGS= -mcpu=powerpc .else LIB32CPUFLAGS= -mcpu=${TARGET_CPUTYPE} .endif LIB32WMAKEENV= MACHINE=powerpc MACHINE_ARCH=powerpc LIB32WMAKEFLAGS= \ LD="${XLD} -m elf32ppc_fbsd" \ OBJCOPY="${XOBJCOPY}" .endif LIB32FLAGS= -m32 ${LIB32CPUFLAGS} -DCOMPAT_32BIT \ -isystem ${LIB32TMP}/usr/include/ \ -L${LIB32TMP}/usr/lib32 \ -B${LIB32TMP}/usr/lib32 .if ${XCC:M/*} LIB32FLAGS+= --sysroot=${WORLDTMP} .endif # Yes, the flags are redundant. LIB32WMAKEENV+= MAKEOBJDIRPREFIX=${LIB32_OBJTREE} \ _SHLIBDIRPREFIX=${LIB32TMP} \ _LDSCRIPTROOT=${LIB32TMP} \ VERSION="${VERSION}" \ INSTALL="sh ${.CURDIR}/tools/install.sh" \ PATH=${TMPPATH} \ LIBDIR=/usr/lib32 \ SHLIBDIR=/usr/lib32 \ LIBPRIVATEDIR=/usr/lib32/private \ DTRACE="${DTRACE} -32" LIB32WMAKEFLAGS+= CC="${XCC} ${LIB32FLAGS}" \ CXX="${XCXX} ${LIB32FLAGS}" \ DESTDIR=${LIB32TMP} \ -DCOMPAT_32BIT \ -DLIBRARIES_ONLY \ -DNO_CPU_CFLAGS \ MK_CTF=no \ -DNO_LINT \ MK_TESTS=no LIB32WMAKE= ${LIB32WMAKEENV} ${MAKE} ${LIB32WMAKEFLAGS} \ MK_MAN=no MK_HTML=no LIB32IMAKE= ${LIB32WMAKE:NINSTALL=*:NDESTDIR=*:N_LDSCRIPTROOT=*} \ MK_TOOLCHAIN=no ${IMAKE_INSTALL} .endif IMAKEENV= ${CROSSENV:N_LDSCRIPTROOT=*} IMAKE= ${IMAKEENV} ${MAKE} -f Makefile.inc1 \ ${IMAKE_INSTALL} ${IMAKE_MTREE} .if empty(.MAKEFLAGS:M-n) IMAKEENV+= PATH=${STRICTTMPPATH}:${INSTALLTMP} \ LD_LIBRARY_PATH=${INSTALLTMP} \ PATH_LOCALE=${INSTALLTMP}/locale IMAKE+= __MAKE_SHELL=${INSTALLTMP}/sh .else IMAKEENV+= PATH=${TMPPATH}:${INSTALLTMP} .endif .if defined(DB_FROM_SRC) INSTALLFLAGS+= -N ${.CURDIR}/etc MTREEFLAGS+= -N ${.CURDIR}/etc .endif _INSTALL_DDIR= ${DESTDIR}/${DISTDIR} INSTALL_DDIR= ${_INSTALL_DDIR:S://:/:g:C:/$::} .if defined(NO_ROOT) METALOG?= ${DESTDIR}/${DISTDIR}/METALOG IMAKE+= -DNO_ROOT METALOG=${METALOG} INSTALLFLAGS+= -U -M ${METALOG} -D ${INSTALL_DDIR} MTREEFLAGS+= -W .endif .if defined(DB_FROM_SRC) || defined(NO_ROOT) IMAKE_INSTALL= INSTALL="install ${INSTALLFLAGS}" IMAKE_MTREE= MTREE_CMD="mtree ${MTREEFLAGS}" .endif # kernel stage KMAKEENV= ${WMAKEENV} KMAKE= ${KMAKEENV} ${MAKE} ${.MAKEFLAGS} ${KERNEL_FLAGS} KERNEL=${INSTKERNNAME} # # buildworld # # Attempt to rebuild the entire system, with reasonable chance of # success, regardless of how old your existing system is. # _worldtmp: .if ${.CURDIR:C/[^,]//g} != "" # The m4 build of sendmail files doesn't like it if ',' is used # anywhere in the path of it's files. @echo @echo "*** Error: path to source tree contains a comma ','" @echo false .endif @echo @echo "--------------------------------------------------------------" @echo ">>> Rebuilding the temporary build tree" @echo "--------------------------------------------------------------" .if !defined(NO_CLEAN) rm -rf ${WORLDTMP} .if defined(LIB32TMP) rm -rf ${LIB32TMP} .endif .else rm -rf ${WORLDTMP}/legacy/usr/include # XXX - These three can depend on any header file. rm -f ${OBJTREE}${.CURDIR}/usr.bin/kdump/ioctl.c rm -f ${OBJTREE}${.CURDIR}/usr.bin/kdump/kdump_subr.c rm -f ${OBJTREE}${.CURDIR}/usr.bin/truss/ioctl.c .endif .for _dir in \ lib usr legacy/bin legacy/usr mkdir -p ${WORLDTMP}/${_dir} .endfor mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${WORLDTMP}/legacy/usr >/dev/null .if ${MK_GROFF} != "no" mtree -deU -f ${.CURDIR}/etc/mtree/BSD.groff.dist \ -p ${WORLDTMP}/legacy/usr >/dev/null .endif mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${WORLDTMP}/usr >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.include.dist \ -p ${WORLDTMP}/usr/include >/dev/null ln -sf ${.CURDIR}/sys ${WORLDTMP} .if ${MK_DEBUG_FILES} != "no" # We could instead disable debug files for these build stages mtree -deU -f ${.CURDIR}/etc/mtree/BSD.debug.dist \ -p ${WORLDTMP}/legacy/usr/lib >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.debug.dist \ -p ${WORLDTMP}/usr/lib >/dev/null .endif .if ${MK_TESTS} != "no" mkdir -p ${WORLDTMP}${TESTSBASE} mtree -deU -f ${.CURDIR}/etc/mtree/BSD.tests.dist \ -p ${WORLDTMP}${TESTSBASE} >/dev/null .endif .for _mtree in ${LOCAL_MTREE} mtree -deU -f ${.CURDIR}/${_mtree} -p ${WORLDTMP} > /dev/null .endfor _legacy: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 1.1: legacy release compatibility shims" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${BMAKE} legacy _bootstrap-tools: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 1.2: bootstrap tools" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${BMAKE} bootstrap-tools _cleanobj: .if !defined(NO_CLEAN) @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.1: cleaning up the object tree" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${WMAKE} ${CLEANDIR:S/^/par-/} .if defined(LIB32TMP) ${_+_}cd ${.CURDIR}; ${LIB32WMAKE} -f Makefile.inc1 ${CLEANDIR:S/^/par-/} .endif .endif _obj: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.2: rebuilding the object tree" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${WMAKE} par-obj _build-tools: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.3: build tools" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${TMAKE} build-tools _cross-tools: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 3: cross tools" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${XMAKE} cross-tools ${_+_}cd ${.CURDIR}; ${XMAKE} kernel-tools _includes: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 4.1: building includes" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${WMAKE} SHARED=symlinks par-includes _libraries: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 4.2: building libraries" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; \ ${WMAKE} -DNO_FSCHG MK_HTML=no -DNO_LINT MK_MAN=no \ MK_PROFILE=no MK_TESTS=no MK_TESTS_SUPPORT=${MK_TESTS} libraries _depend: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 4.3: make dependencies" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${WMAKE} par-depend everything: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 4.4: building everything" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${WMAKE} par-all .if defined(LIB32TMP) build32: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 5.1: building 32 bit shim libraries" @echo "--------------------------------------------------------------" mkdir -p ${LIB32TMP}/usr/include mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${LIB32TMP}/usr >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.include.dist \ -p ${LIB32TMP}/usr/include >/dev/null .if ${MK_DEBUG_FILES} != "no" mtree -deU -f ${.CURDIR}/etc/mtree/BSD.debug.dist \ -p ${LIB32TMP}/usr/lib >/dev/null .endif mkdir -p ${WORLDTMP} ln -sf ${.CURDIR}/sys ${WORLDTMP} .for _t in obj includes cd ${.CURDIR}/include; ${LIB32WMAKE} DIRPRFX=include/ ${_t} cd ${.CURDIR}/lib; ${LIB32WMAKE} DIRPRFX=lib/ ${_t} .if ${MK_CDDL} != "no" cd ${.CURDIR}/cddl/lib; ${LIB32WMAKE} DIRPRFX=cddl/lib/ ${_t} .endif cd ${.CURDIR}/gnu/lib; ${LIB32WMAKE} DIRPRFX=gnu/lib/ ${_t} .if ${MK_CRYPT} != "no" cd ${.CURDIR}/secure/lib; ${LIB32WMAKE} DIRPRFX=secure/lib/ ${_t} .endif .if ${MK_KERBEROS} != "no" cd ${.CURDIR}/kerberos5/lib; ${LIB32WMAKE} DIRPRFX=kerberos5/lib ${_t} .endif .endfor .for _dir in usr.bin/lex/lib cd ${.CURDIR}/${_dir}; ${LIB32WMAKE} DIRPRFX=${_dir}/ obj .endfor .for _dir in lib/ncurses/ncurses lib/ncurses/ncursesw lib/libmagic cd ${.CURDIR}/${_dir}; \ WORLDTMP=${WORLDTMP} \ MAKEFLAGS="-m ${.CURDIR}/tools/build/mk ${.MAKEFLAGS}" \ MAKEOBJDIRPREFIX=${LIB32_OBJTREE} ${MAKE} SSP_CFLAGS= DESTDIR= \ DIRPRFX=${_dir}/ -DNO_LINT -DNO_CPU_CFLAGS MK_WARNS=no MK_CTF=no \ build-tools .endfor cd ${.CURDIR}; \ ${LIB32WMAKE} -f Makefile.inc1 libraries .for _t in obj depend all cd ${.CURDIR}/libexec/rtld-elf; PROG=ld-elf32.so.1 ${LIB32WMAKE} \ DIRPRFX=libexec/rtld-elf/ ${_t} cd ${.CURDIR}/usr.bin/ldd; PROG=ldd32 ${LIB32WMAKE} \ DIRPRFX=usr.bin/ldd ${_t} .endfor distribute32 install32: cd ${.CURDIR}/lib; ${LIB32IMAKE} ${.TARGET:S/32$//} .if ${MK_CDDL} != "no" cd ${.CURDIR}/cddl/lib; ${LIB32IMAKE} ${.TARGET:S/32$//} .endif cd ${.CURDIR}/gnu/lib; ${LIB32IMAKE} ${.TARGET:S/32$//} .if ${MK_CRYPT} != "no" cd ${.CURDIR}/secure/lib; ${LIB32IMAKE} ${.TARGET:S/32$//} .endif .if ${MK_KERBEROS} != "no" cd ${.CURDIR}/kerberos5/lib; ${LIB32IMAKE} ${.TARGET:S/32$//} .endif cd ${.CURDIR}/libexec/rtld-elf; \ PROG=ld-elf32.so.1 ${LIB32IMAKE} ${.TARGET:S/32$//} cd ${.CURDIR}/usr.bin/ldd; PROG=ldd32 ${LIB32IMAKE} ${.TARGET:S/32$//} .endif WMAKE_TGTS= .if !defined(SUBDIR_OVERRIDE) WMAKE_TGTS+= _worldtmp _legacy _bootstrap-tools .endif WMAKE_TGTS+= _cleanobj _obj _build-tools .if !defined(SUBDIR_OVERRIDE) WMAKE_TGTS+= _cross-tools .endif WMAKE_TGTS+= _includes _libraries _depend everything .if defined(LIB32TMP) && ${MK_LIB32} != "no" WMAKE_TGTS+= build32 .endif buildworld: buildworld_prologue ${WMAKE_TGTS} buildworld_epilogue .ORDER: buildworld_prologue ${WMAKE_TGTS} buildworld_epilogue buildworld_prologue: @echo "--------------------------------------------------------------" @echo ">>> World build started on `LC_ALL=C date`" @echo "--------------------------------------------------------------" buildworld_epilogue: @echo @echo "--------------------------------------------------------------" @echo ">>> World build completed on `LC_ALL=C date`" @echo "--------------------------------------------------------------" # # We need to have this as a target because the indirection between Makefile # and Makefile.inc1 causes the correct PATH to be used, rather than a # modification of the current environment's PATH. In addition, we need # to quote multiword values. # buildenvvars: @echo ${WMAKEENV:Q} .if ${.TARGETS:Mbuildenv} .if ${.MAKEFLAGS:M-j} .error The buildenv target is incompatible with -j .endif .endif buildenv: @echo Entering world for ${TARGET_ARCH}:${TARGET} @cd ${.CURDIR} && env ${WMAKEENV} ${BUILDENV_SHELL} || true TOOLCHAIN_TGTS= ${WMAKE_TGTS:N_depend:Neverything:Nbuild32} toolchain: ${TOOLCHAIN_TGTS} kernel-toolchain: ${TOOLCHAIN_TGTS:N_includes:N_libraries} # # installcheck # # Checks to be sure system is ready for installworld/installkernel. # installcheck: _installcheck_world _installcheck_kernel _installcheck_world: _installcheck_kernel: # # Require DESTDIR to be set if installing for a different architecture or # using the user/group database in the source tree. # .if ${TARGET_ARCH} != ${MACHINE_ARCH} || ${TARGET} != ${MACHINE} || \ defined(DB_FROM_SRC) .if !make(distributeworld) _installcheck_world: __installcheck_DESTDIR _installcheck_kernel: __installcheck_DESTDIR __installcheck_DESTDIR: .if !defined(DESTDIR) || empty(DESTDIR) @echo "ERROR: Please set DESTDIR!"; \ false .endif .endif .endif .if !defined(DB_FROM_SRC) # # Check for missing UIDs/GIDs. # CHECK_UIDS= auditdistd CHECK_GIDS= audit .if ${MK_SENDMAIL} != "no" CHECK_UIDS+= smmsp CHECK_GIDS+= smmsp .endif .if ${MK_PF} != "no" CHECK_UIDS+= proxy CHECK_GIDS+= proxy authpf .endif .if ${MK_UNBOUND} != "no" CHECK_UIDS+= unbound CHECK_GIDS+= unbound .endif _installcheck_world: __installcheck_UGID __installcheck_UGID: .for uid in ${CHECK_UIDS} @if ! `id -u ${uid} >/dev/null 2>&1`; then \ echo "ERROR: Required ${uid} user is missing, see /usr/src/UPDATING."; \ false; \ fi .endfor .for gid in ${CHECK_GIDS} @if ! `find / -prune -group ${gid} >/dev/null 2>&1`; then \ echo "ERROR: Required ${gid} group is missing, see /usr/src/UPDATING."; \ false; \ fi .endfor .endif # # Required install tools to be saved in a scratch dir for safety. # .if ${MK_ZONEINFO} != "no" _zoneinfo= zic tzsetup .endif ITOOLS= [ awk cap_mkdb cat chflags chmod chown \ date echo egrep find grep id install ${_install-info} \ ln lockf make mkdir mtree mv pwd_mkdb \ rm sed services_mkdb sh strip sysctl test true uname wc ${_zoneinfo} \ ${LOCAL_ITOOLS} # Needed for share/man .if ${MK_MAN} != "no" ITOOLS+=makewhatis .endif # # distributeworld # # Distributes everything compiled by a `buildworld'. # # installworld # # Installs everything compiled by a 'buildworld'. # # Non-base distributions produced by the base system EXTRA_DISTRIBUTIONS= doc .if defined(LIB32TMP) && ${MK_LIB32} != "no" EXTRA_DISTRIBUTIONS+= lib32 .endif .if ${MK_TESTS} != "no" EXTRA_DISTRIBUTIONS+= tests .endif DEBUG_DISTRIBUTIONS= .if ${MK_DEBUG_FILES} != "no" DEBUG_DISTRIBUTIONS+= base ${EXTRA_DISTRIBUTIONS:S,doc,,:S,tests,,} .endif MTREE_MAGIC?= mtree 2.0 distributeworld installworld: _installcheck_world mkdir -p ${INSTALLTMP} progs=$$(for prog in ${ITOOLS}; do \ if progpath=`which $$prog`; then \ echo $$progpath; \ else \ echo "Required tool $$prog not found in PATH." >&2; \ exit 1; \ fi; \ done); \ libs=$$(ldd -f "%o %p\n" -f "%o %p\n" $$progs 2>/dev/null | sort -u | \ while read line; do \ set -- $$line; \ if [ "$$2 $$3" != "not found" ]; then \ echo $$2; \ else \ echo "Required library $$1 not found." >&2; \ exit 1; \ fi; \ done); \ cp $$libs $$progs ${INSTALLTMP} cp -R $${PATH_LOCALE:-"/usr/share/locale"} ${INSTALLTMP}/locale .if defined(NO_ROOT) echo "#${MTREE_MAGIC}" > ${METALOG} .endif .if make(distributeworld) .for dist in ${EXTRA_DISTRIBUTIONS} -mkdir ${DESTDIR}/${DISTDIR}/${dist} mtree -deU -f ${.CURDIR}/etc/mtree/BSD.root.dist \ -p ${DESTDIR}/${DISTDIR}/${dist} >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${DESTDIR}/${DISTDIR}/${dist}/usr >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.include.dist \ -p ${DESTDIR}/${DISTDIR}/${dist}/usr/include >/dev/null .if ${MK_DEBUG_FILES} != "no" mtree -deU -f ${.CURDIR}/etc/mtree/BSD.debug.dist \ -p ${DESTDIR}/${DISTDIR}/${dist}/usr/lib >/dev/null .endif .if ${MK_TESTS} != "no" && ${dist} == "tests" -mkdir -p ${DESTDIR}/${DISTDIR}/${dist}${TESTSBASE} mtree -deU -f ${.CURDIR}/etc/mtree/BSD.tests.dist \ -p ${DESTDIR}/${DISTDIR}/${dist}${TESTSBASE} >/dev/null .endif .if defined(NO_ROOT) ${IMAKEENV} mtree -C -f ${.CURDIR}/etc/mtree/BSD.root.dist | \ sed -e 's#^\./#./${dist}/#' >> ${METALOG} ${IMAKEENV} mtree -C -f ${.CURDIR}/etc/mtree/BSD.usr.dist | \ sed -e 's#^\./#./${dist}/usr/#' >> ${METALOG} ${IMAKEENV} mtree -C -f ${.CURDIR}/etc/mtree/BSD.include.dist | \ sed -e 's#^\./#./${dist}/usr/include/#' >> ${METALOG} .endif .endfor -mkdir ${DESTDIR}/${DISTDIR}/base cd ${.CURDIR}/etc; ${CROSSENV} PATH=${TMPPATH} ${MAKE} \ METALOG=${METALOG} ${IMAKE_INSTALL} ${IMAKE_MTREE} \ DISTBASE=/base DESTDIR=${DESTDIR}/${DISTDIR}/base \ LOCAL_MTREE=${LOCAL_MTREE:Q} distrib-dirs .endif ${_+_}cd ${.CURDIR}; ${IMAKE} re${.TARGET:S/world$//}; \ ${IMAKEENV} rm -rf ${INSTALLTMP} .if make(distributeworld) .for dist in ${EXTRA_DISTRIBUTIONS} find ${DESTDIR}/${DISTDIR}/${dist} -mindepth 1 -empty -delete .endfor .if defined(NO_ROOT) .for dist in base ${EXTRA_DISTRIBUTIONS} @# For each file that exists in this dist, print the corresponding @# line from the METALOG. This relies on the fact that @# a line containing only the filename will sort immediatly before @# the relevant mtree line. cd ${DESTDIR}/${DISTDIR}; \ find ./${dist} | sort -u ${METALOG} - | \ awk 'BEGIN { print "#${MTREE_MAGIC}" } !/ type=/ { file = $$1 } / type=/ { if ($$1 == file) { sub(/^\.\/${dist}\//, "./"); print } }' > \ ${DESTDIR}/${DISTDIR}/${dist}.meta .endfor .for dist in ${DEBUG_DISTRIBUTIONS} @# For each file that exists in this dist, print the corresponding @# line from the METALOG. This relies on the fact that @# a line containing only the filename will sort immediatly before @# the relevant mtree line. cd ${DESTDIR}/${DISTDIR}; \ find ./${dist}/usr/lib/debug | sort -u ${METALOG} - | \ awk 'BEGIN { print "#${MTREE_MAGIC}" } !/ type=/ { file = $$1 } / type=/ { if ($$1 == file) { sub(/^\.\/${dist}\//, "./"); print } }' > \ ${DESTDIR}/${DISTDIR}/${dist}.debug.meta .endfor .endif .endif packageworld: .for dist in base ${EXTRA_DISTRIBUTIONS} .if defined(NO_ROOT) ${_+_}cd ${DESTDIR}/${DISTDIR}/${dist}; \ tar cvf - --exclude usr/lib/debug \ @${DESTDIR}/${DISTDIR}/${dist}.meta | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/${dist}.txz .else ${_+_}cd ${DESTDIR}/${DISTDIR}/${dist}; \ tar cvf - --exclude usr/lib/debug . | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/${dist}.txz .endif .endfor .for dist in ${DEBUG_DISTRIBUTIONS} . if defined(NO_ROOT) ${_+_}cd ${DESTDIR}/${DISTDIR}/${dist}; \ tar cvf - @${DESTDIR}/${DISTDIR}/${dist}.debug.meta | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/${dist}-dbg.txz . else ${_+_}cd ${DESTDIR}/${DISTDIR}/${dist}; \ tar cvLf - usr/lib/debug | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/${dist}-dbg.txz . endif .endfor # # reinstall # # If you have a build server, you can NFS mount the source and obj directories # and do a 'make reinstall' on the *client* to install new binaries from the # most recent server build. # reinstall: .MAKE @echo "--------------------------------------------------------------" @echo ">>> Making hierarchy" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${MAKE} -f Makefile.inc1 \ LOCAL_MTREE=${LOCAL_MTREE:Q} hierarchy @echo @echo "--------------------------------------------------------------" @echo ">>> Installing everything" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${MAKE} -f Makefile.inc1 install .if defined(LIB32TMP) && ${MK_LIB32} != "no" ${_+_}cd ${.CURDIR}; ${MAKE} -f Makefile.inc1 install32 .endif redistribute: .MAKE @echo "--------------------------------------------------------------" @echo ">>> Distributing everything" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${MAKE} -f Makefile.inc1 distribute .if defined(LIB32TMP) && ${MK_LIB32} != "no" ${_+_}cd ${.CURDIR}; ${MAKE} -f Makefile.inc1 distribute32 \ DISTRIBUTION=lib32 .endif distrib-dirs distribution: .MAKE cd ${.CURDIR}/etc; ${CROSSENV} PATH=${TMPPATH} ${MAKE} \ ${IMAKE_INSTALL} ${IMAKE_MTREE} METALOG=${METALOG} ${.TARGET} # # buildkernel and installkernel # # Which kernels to build and/or install is specified by setting # KERNCONF. If not defined a GENERIC kernel is built/installed. # Only the existing (depending TARGET) config files are used # for building kernels and only the first of these is designated # as the one being installed. # # Note that we have to use TARGET instead of TARGET_ARCH when # we're in kernel-land. Since only TARGET_ARCH is (expected) to # be set to cross-build, we have to make sure TARGET is set # properly. .if defined(KERNFAST) NO_KERNELCLEAN= t NO_KERNELCONFIG= t NO_KERNELDEPEND= t NO_KERNELOBJ= t # Shortcut for KERNCONF=Blah -DKERNFAST is now KERNFAST=Blah .if !defined(KERNCONF) && ${KERNFAST} != "1" KERNCONF=${KERNFAST} .endif .endif .if ${TARGET_ARCH} == "powerpc64" KERNCONF?= GENERIC64 .else KERNCONF?= GENERIC .endif INSTKERNNAME?= kernel KERNSRCDIR?= ${.CURDIR}/sys KRNLCONFDIR= ${KERNSRCDIR}/${TARGET}/conf KRNLOBJDIR= ${OBJTREE}${KERNSRCDIR} KERNCONFDIR?= ${KRNLCONFDIR} BUILDKERNELS= INSTALLKERNEL= .for _kernel in ${KERNCONF} .if exists(${KERNCONFDIR}/${_kernel}) BUILDKERNELS+= ${_kernel} .if empty(INSTALLKERNEL) INSTALLKERNEL= ${_kernel} .endif .endif .endfor buildkernel ${WMAKE_TGTS} ${.ALLTARGETS:M_*}: .MAKE # # buildkernel # # Builds all kernels defined by BUILDKERNELS. # buildkernel: .if empty(BUILDKERNELS) @echo "ERROR: Missing kernel configuration file(s) (${KERNCONF})."; \ false .endif @echo .for _kernel in ${BUILDKERNELS} @echo "--------------------------------------------------------------" @echo ">>> Kernel build for ${_kernel} started on `LC_ALL=C date`" @echo "--------------------------------------------------------------" @echo "===> ${_kernel}" mkdir -p ${KRNLOBJDIR} .if !defined(NO_KERNELCONFIG) @echo @echo "--------------------------------------------------------------" @echo ">>> stage 1: configuring the kernel" @echo "--------------------------------------------------------------" cd ${KRNLCONFDIR}; \ PATH=${TMPPATH} \ config ${CONFIGARGS} -d ${KRNLOBJDIR}/${_kernel} \ -I '${KERNCONFDIR}' '${KERNCONFDIR}/${_kernel}' .endif .if !defined(NO_CLEAN) && !defined(NO_KERNELCLEAN) @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.1: cleaning up the object tree" @echo "--------------------------------------------------------------" cd ${KRNLOBJDIR}/${_kernel}; ${KMAKE} ${CLEANDIR} .endif .if !defined(NO_KERNELOBJ) @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.2: rebuilding the object tree" @echo "--------------------------------------------------------------" cd ${KRNLOBJDIR}/${_kernel}; ${KMAKE} obj .endif @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.3: build tools" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${KTMAKE} kernel-tools .if !defined(NO_KERNELDEPEND) @echo @echo "--------------------------------------------------------------" @echo ">>> stage 3.1: making dependencies" @echo "--------------------------------------------------------------" cd ${KRNLOBJDIR}/${_kernel}; ${KMAKE} depend -DNO_MODULES_OBJ .endif @echo @echo "--------------------------------------------------------------" @echo ">>> stage 3.2: building everything" @echo "--------------------------------------------------------------" cd ${KRNLOBJDIR}/${_kernel}; ${KMAKE} all -DNO_MODULES_OBJ @echo "--------------------------------------------------------------" @echo ">>> Kernel build for ${_kernel} completed on `LC_ALL=C date`" @echo "--------------------------------------------------------------" .endfor # # installkernel, etc. # # Install the kernel defined by INSTALLKERNEL # installkernel installkernel.debug \ reinstallkernel reinstallkernel.debug: _installcheck_kernel .if empty(INSTALLKERNEL) @echo "ERROR: No kernel \"${KERNCONF}\" to install."; \ false .endif @echo "--------------------------------------------------------------" @echo ">>> Installing kernel ${INSTALLKERNEL}" @echo "--------------------------------------------------------------" cd ${KRNLOBJDIR}/${INSTALLKERNEL}; \ ${CROSSENV} PATH=${TMPPATH} \ ${MAKE} ${IMAKE_INSTALL} KERNEL=${INSTKERNNAME} ${.TARGET:S/kernel//} distributekernel distributekernel.debug: .if empty(INSTALLKERNEL) @echo "ERROR: No kernel \"${KERNCONF}\" to install."; \ false .endif mkdir -p ${DESTDIR}/${DISTDIR} .if defined(NO_ROOT) echo "#${MTREE_MAGIC}" > ${DESTDIR}/${DISTDIR}/kernel.premeta .endif cd ${KRNLOBJDIR}/${INSTALLKERNEL}; \ ${IMAKEENV} ${IMAKE_INSTALL:S/METALOG/kernel.premeta/} \ ${IMAKE_MTREE} PATH=${TMPPATH} ${MAKE} KERNEL=${INSTKERNNAME} \ DESTDIR=${INSTALL_DDIR}/kernel \ ${.TARGET:S/distributekernel/install/} .if defined(NO_ROOT) sed -e 's|^./kernel|.|' ${DESTDIR}/${DISTDIR}/kernel.premeta > \ ${DESTDIR}/${DISTDIR}/kernel.meta .endif .for _kernel in ${BUILDKERNELS:S/${INSTALLKERNEL}//} .if defined(NO_ROOT) echo "#${MTREE_MAGIC}" > ${DESTDIR}/${DISTDIR}/kernel.${_kernel}.premeta .endif cd ${KRNLOBJDIR}/${_kernel}; \ ${IMAKEENV} ${IMAKE_INSTALL:S/METALOG/kernel.${_kernel}.premeta/} \ ${IMAKE_MTREE} PATH=${TMPPATH} ${MAKE} \ KERNEL=${INSTKERNNAME}.${_kernel} \ DESTDIR=${INSTALL_DDIR}/kernel.${_kernel} \ ${.TARGET:S/distributekernel/install/} .if defined(NO_ROOT) sed -e 's|^./kernel|.|' \ ${DESTDIR}/${DISTDIR}/kernel.${_kernel}.premeta > \ ${DESTDIR}/${DISTDIR}/kernel.${_kernel}.meta .endif .endfor packagekernel: .if defined(NO_ROOT) cd ${DESTDIR}/${DISTDIR}/kernel; \ tar cvf - @${DESTDIR}/${DISTDIR}/kernel.meta | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/kernel.txz .for _kernel in ${BUILDKERNELS:S/${INSTALLKERNEL}//} cd ${DESTDIR}/${DISTDIR}/kernel.${_kernel}; \ tar cvf - @${DESTDIR}/${DISTDIR}/kernel.${_kernel}.meta | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/kernel.${_kernel}.txz .endfor .else cd ${DESTDIR}/${DISTDIR}/kernel; \ tar cvf - . | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/kernel.txz .for _kernel in ${BUILDKERNELS:S/${INSTALLKERNEL}//} cd ${DESTDIR}/${DISTDIR}/kernel.${_kernel}; \ tar cvf - . | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/kernel.${_kernel}.txz .endfor .endif # # doxygen # # Build the API documentation with doxygen # doxygen: @if [ ! -x `/usr/bin/which doxygen` ]; then \ echo "You need doxygen (devel/doxygen) to generate the API documentation of the kernel." | /usr/bin/fmt; \ exit 1; \ fi cd ${.CURDIR}/tools/kerneldoc/subsys && ${MAKE} obj all # # update # # Update the source tree(s), by running svn/svnup to update to the # latest copy. # update: .if (defined(CVS_UPDATE) || defined(SUP_UPDATE)) && !defined(SVN_UPDATE) @echo "--------------------------------------------------------------" @echo "CVS_UPDATE and SUP_UPDATE are no longer supported." @echo "Please see: https://wiki.freebsd.org/CvsIsDeprecated" @echo "--------------------------------------------------------------" @exit 1 .endif .if defined(SVN_UPDATE) @echo "--------------------------------------------------------------" @echo ">>> Updating ${.CURDIR} using Subversion" @echo "--------------------------------------------------------------" @(cd ${.CURDIR} && ${SVN} update ${SVNFLAGS}) .endif # # ------------------------------------------------------------------------ # # From here onwards are utility targets used by the 'make world' and # related targets. If your 'world' breaks, you may like to try to fix # the problem and manually run the following targets to attempt to # complete the build. Beware, this is *not* guaranteed to work, you # need to have a pretty good grip on the current state of the system # to attempt to manually finish it. If in doubt, 'make world' again. # # # legacy: Build compatibility shims for the next three targets. This is a minimal # set of tools and shims necessary to compensate for older systems which don't have # the APIs that the targets built in bootstrap-tools, build-tools or cross-tools. # legacy: .if ${BOOTSTRAPPING} < 800107 && ${BOOTSTRAPPING} != 0 @echo "ERROR: Source upgrades from versions prior to 8.0 not supported."; \ false .endif .for _tool in tools/build ${_+_}@${ECHODIR} "===> ${_tool} (obj,includes,depend,all,install)"; \ cd ${.CURDIR}/${_tool} && \ ${MAKE} DIRPRFX=${_tool}/ obj && \ ${MAKE} DIRPRFX=${_tool}/ DESTDIR=${MAKEOBJDIRPREFIX}/legacy includes && \ ${MAKE} DIRPRFX=${_tool}/ depend && \ ${MAKE} DIRPRFX=${_tool}/ all && \ ${MAKE} DIRPRFX=${_tool}/ DESTDIR=${MAKEOBJDIRPREFIX}/legacy install .endfor # # bootstrap-tools: Build tools needed for compatibility. These are binaries that # are built to build other binaries in the system. However, the focus of these # binaries is usually quite narrow. Bootstrap tools use the host's compiler and # libraries, augmented by -legacy. # _bt= _bootstrap-tools .if ${MK_GAMES} != "no" _strfile= games/fortune/strfile .endif .if ${MK_CXX} != "no" _gperf= gnu/usr.bin/gperf .endif .if ${MK_GROFF} != "no" _groff= gnu/usr.bin/groff \ usr.bin/soelim .endif .if ${MK_VT} != "no" _vtfontcvt= usr.bin/vtfontcvt .endif .if ${BOOTSTRAPPING} < 900002 _sed= usr.bin/sed .endif .if ${BOOTSTRAPPING} < 1000002 _libohash= lib/libohash _m4= usr.bin/m4 ${_bt}-usr.bin/m4: ${_bt}-lib/libohash .endif .if ${BOOTSTRAPPING} < 1000026 _nmtree= lib/libnetbsd \ usr.sbin/nmtree ${_bt}-usr.sbin/nmtree: ${_bt}-lib/libnetbsd .endif .if ${BOOTSTRAPPING} < 1000027 _cat= bin/cat .endif .if ${BOOTSTRAPPING} < 1000033 _lex= usr.bin/lex .endif # r277259 crunchide: Correct 64-bit section header offset # r281674 crunchide: always include both 32- and 64-bit ELF support .if ${BOOTSTRAPPING} < 1100071 _crunch= usr.sbin/crunch .endif .if ${BOOTSTRAPPING} >= 900040 && ${BOOTSTRAPPING} < 900041 _awk= usr.bin/awk .endif _yacc= lib/liby \ usr.bin/yacc ${_bt}-usr.bin/yacc: ${_bt}-lib/liby .if ${MK_BSNMP} != "no" _gensnmptree= usr.sbin/bsnmpd/gensnmptree .endif # We need to build tblgen when we're building clang either as # the bootstrap compiler, or as the part of the normal build. .if ${MK_CLANG_BOOTSTRAP} != "no" || ${MK_CLANG} != "no" _clang_tblgen= \ lib/clang/libllvmsupport \ lib/clang/libllvmtablegen \ usr.bin/clang/tblgen \ usr.bin/clang/clang-tblgen ${_bt}-usr.bin/clang/clang-tblgen: ${_bt}-lib/clang/libllvmtablegen ${_bt}-lib/clang/libllvmsupport ${_bt}-usr.bin/clang/tblgen: ${_bt}-lib/clang/libllvmtablegen ${_bt}-lib/clang/libllvmsupport .endif # ELF Tool Chain libraries are needed for ELF tools and dtrace tools. # dtrace tools are required for older bootstrap env and cross-build # pre libdwarf .if ${BOOTSTRAPPING} < 1100006 || (${MACHINE} != ${TARGET} || \ ${MACHINE_ARCH} != ${TARGET_ARCH}) _elftoolchain_libs= lib/libelf lib/libdwarf .if ${MK_CDDL} != "no" _dtrace_tools= cddl/usr.bin/sgsmsg cddl/lib/libctf cddl/usr.bin/ctfconvert \ cddl/usr.bin/ctfmerge ${_bt}-cddl/usr.bin/ctfconvert: ${_bt}-lib/libelf ${_bt}-lib/libdwarf ${_bt}-cddl/lib/libctf ${_bt}-cddl/usr.bin/ctfmerge: ${_bt}-lib/libelf ${_bt}-lib/libdwarf ${_bt}-cddl/lib/libctf .endif .endif # Default to building the GPL DTC, but build the BSDL one if users explicitly # request it. _dtc= usr.bin/dtc .if ${MK_GPL_DTC} != "no" _dtc= gnu/usr.bin/dtc .endif .if ${MK_KERBEROS} != "no" _kerberos5_bootstrap_tools= \ kerberos5/tools/make-roken \ kerberos5/lib/libroken \ kerberos5/lib/libvers \ kerberos5/tools/asn1_compile \ kerberos5/tools/slc \ usr.bin/compile_et .ORDER: ${_kerberos5_bootstrap_tools:C/^/${_bt}-/g} .endif .if ${MK_MANDOCDB} != "no" _libohash?= lib/libohash _makewhatis= lib/libsqlite3 \ - usr.bin/mandoc + usr.bin/mandoc \ + usr.bin/gzip ${_bt}-usr.bin/mandoc: ${_bt}-lib/libohash ${_bt}-lib/libsqlite3 .else _makewhatis=usr.bin/makewhatis .endif # Rebuild up-to-date libmd for xinstall ${_bt}-usr.bin/xinstall: ${_bt}-lib/libmd bootstrap-tools: .PHONY # Please document (add comment) why something is in 'bootstrap-tools'. # Try to bound the building of the bootstrap-tool to just the # FreeBSD versions that need the tool built at this stage of the build. .for _tool in \ ${_clang_tblgen} \ ${_kerberos5_bootstrap_tools} \ ${_elftoolchain_libs} \ ${_dtrace_tools} \ ${_strfile} \ ${_gperf} \ ${_groff} \ ${_dtc} \ ${_awk} \ ${_cat} \ usr.bin/lorder \ ${_libohash} \ ${_makewhatis} \ usr.bin/rpcgen \ ${_sed} \ ${_yacc} \ ${_m4} \ ${_lex} \ lib/libmd \ usr.bin/xinstall \ ${_gensnmptree} \ usr.sbin/config \ ${_crunch} \ ${_nmtree} \ ${_vtfontcvt} ${_bt}-${_tool}: .PHONY .MAKE ${_+_}@${ECHODIR} "===> ${_tool} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_tool} && \ ${MAKE} DIRPRFX=${_tool}/ obj && \ ${MAKE} DIRPRFX=${_tool}/ depend && \ ${MAKE} DIRPRFX=${_tool}/ all && \ ${MAKE} DIRPRFX=${_tool}/ DESTDIR=${MAKEOBJDIRPREFIX}/legacy install bootstrap-tools: ${_bt}-${_tool} .endfor # # build-tools: Build special purpose build tools # .if !defined(NO_SHARE) _share= share/syscons/scrnmaps .endif .if ${MK_GCC} != "no" _gcc_tools= gnu/usr.bin/cc/cc_tools .endif .if ${MK_RESCUE} != "no" _rescue= rescue/rescue .endif build-tools: .MAKE .for _tool in \ bin/csh \ bin/sh \ ${_rescue} \ ${LOCAL_TOOL_DIRS} \ lib/ncurses/ncurses \ lib/ncurses/ncursesw \ ${_share} \ usr.bin/awk \ lib/libmagic \ usr.bin/mkesdb_static \ usr.bin/mkcsmapper_static \ usr.bin/vi/catalog ${_+_}@${ECHODIR} "===> ${_tool} (obj,build-tools)"; \ cd ${.CURDIR}/${_tool} && \ ${MAKE} DIRPRFX=${_tool}/ obj && \ ${MAKE} DIRPRFX=${_tool}/ build-tools .endfor .for _tool in \ ${_gcc_tools} ${_+_}@${ECHODIR} "===> ${_tool} (obj,depend,all)"; \ cd ${.CURDIR}/${_tool} && \ ${MAKE} DIRPRFX=${_tool}/ obj && \ ${MAKE} DIRPRFX=${_tool}/ depend && \ ${MAKE} DIRPRFX=${_tool}/ all .endfor # # kernel-tools: Build kernel-building tools # kernel-tools: .MAKE mkdir -p ${MAKEOBJDIRPREFIX}/usr mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${MAKEOBJDIRPREFIX}/usr >/dev/null # # cross-tools: All the tools needed to build the rest of the system after # we get done with the earlier stages. It is the last set of tools needed # to begin building the target binaries. # .if ${TARGET_ARCH} != ${MACHINE_ARCH} .if ${TARGET_ARCH} == "amd64" || ${TARGET_ARCH} == "i386" _btxld= usr.sbin/btxld .endif .endif # If we're given an XAS, don't build binutils. .if ${XAS:M/*} == "" .if ${MK_BINUTILS_BOOTSTRAP} != "no" _binutils= gnu/usr.bin/binutils .endif .if ${MK_ELFTOOLCHAIN_TOOLS} != "no" _elftctools= lib/libelftc \ usr.bin/elfcopy \ usr.bin/nm \ usr.bin/size \ usr.bin/strings # These are not required by the build, but can be useful for developers who # cross-build on a FreeBSD 10 host: _elftctools+= usr.bin/addr2line .endif .elif ${TARGET_ARCH} != ${MACHINE_ARCH} && ${MK_ELFTOOLCHAIN_TOOLS} != "no" # If cross-building with an external binutils we still need to build strip for # the target (for at least crunchide). _elftctools= lib/libelftc \ usr.bin/elfcopy .endif # If an full path to an external cross compiler is given, don't build # a cross compiler. .if ${XCC:M/*} == "" && ${MK_CROSS_COMPILER} != "no" .if ${MK_CLANG_BOOTSTRAP} != "no" _clang= usr.bin/clang _clang_libs= lib/clang .endif .if ${MK_GCC_BOOTSTRAP} != "no" _cc= gnu/usr.bin/cc .endif .endif .if ${MK_USB} != "no" _usb_tools= sys/boot/usb/tools .endif cross-tools: .MAKE .for _tool in \ ${_clang_libs} \ ${_clang} \ ${_binutils} \ ${_elftctools} \ ${_cc} \ ${_btxld} \ ${_crunchide} \ ${_usb_tools} ${_+_}@${ECHODIR} "===> ${_tool} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_tool} && \ ${MAKE} DIRPRFX=${_tool}/ obj && \ ${MAKE} DIRPRFX=${_tool}/ depend && \ ${MAKE} DIRPRFX=${_tool}/ all && \ ${MAKE} DIRPRFX=${_tool}/ DESTDIR=${MAKEOBJDIRPREFIX} install .endfor NXBENV= MAKEOBJDIRPREFIX=${OBJTREE}/nxb \ INSTALL="sh ${.CURDIR}/tools/install.sh" \ VERSION="${VERSION}" \ PATH=${PATH}:${OBJTREE}/gperf_for_gcc/usr/bin NXBMAKE= ${NXBENV} ${MAKE} \ TBLGEN=${OBJTREE}/nxb-bin/usr/bin/tblgen \ CLANG_TBLGEN=${OBJTREE}/nxb-bin/usr/bin/clang-tblgen \ MACHINE=${TARGET} MACHINE_ARCH=${TARGET_ARCH} \ MK_GDB=no MK_TESTS=no \ SSP_CFLAGS= \ MK_HTML=no NO_LINT=yes MK_MAN=no \ -DNO_PIC MK_PROFILE=no -DNO_SHARED \ -DNO_CPU_CFLAGS MK_WARNS=no MK_CTF=no \ MK_CLANG_EXTRAS=no MK_CLANG_FULL=no \ MK_LLDB=no # native-xtools is the current target for qemu-user cross builds of ports # via poudriere and the imgact_binmisc kernel module. # For non-clang enabled targets that are still using the in tree gcc # we must build a gperf binary for one instance of its Makefiles. On # clang-enabled systems, the gperf binary is obsolete. native-xtools: .MAKE .if ${MK_GCC_BOOTSTRAP} != "no" mkdir -p ${OBJTREE}/gperf_for_gcc/usr/bin ${_+_}@${ECHODIR} "===> ${_gperf} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_gperf} && \ ${NXBMAKE} DIRPRFX=${_gperf}/ obj && \ ${NXBMAKE} DIRPRFX=${_gperf}/ depend && \ ${NXBMAKE} DIRPRFX=${_gperf}/ all && \ ${NXBMAKE} DIRPRFX=${_gperf}/ DESTDIR=${OBJTREE}/gperf_for_gcc install .endif mkdir -p ${OBJTREE}/nxb-bin/bin mkdir -p ${OBJTREE}/nxb-bin/sbin mkdir -p ${OBJTREE}/nxb-bin/usr mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${OBJTREE}/nxb-bin/usr >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.include.dist \ -p ${OBJTREE}/nxb-bin/usr/include >/dev/null .for _tool in \ bin/cat \ bin/chmod \ bin/cp \ bin/csh \ bin/echo \ bin/expr \ bin/hostname \ bin/ln \ bin/ls \ bin/mkdir \ bin/mv \ bin/ps \ bin/realpath \ bin/rm \ bin/rmdir \ bin/sh \ bin/sleep \ ${_clang_tblgen} \ usr.bin/ar \ ${_binutils} \ ${_elftctools} \ ${_cc} \ ${_gcc_tools} \ ${_clang_libs} \ ${_clang} \ sbin/md5 \ sbin/sysctl \ gnu/usr.bin/diff \ usr.bin/awk \ usr.bin/basename \ usr.bin/bmake \ usr.bin/bzip2 \ usr.bin/cmp \ usr.bin/dirname \ usr.bin/env \ usr.bin/fetch \ usr.bin/find \ usr.bin/grep \ usr.bin/gzip \ usr.bin/id \ usr.bin/lex \ usr.bin/lorder \ usr.bin/mktemp \ usr.bin/mt \ usr.bin/patch \ usr.bin/sed \ usr.bin/sort \ usr.bin/tar \ usr.bin/touch \ usr.bin/tr \ usr.bin/true \ usr.bin/uniq \ usr.bin/unzip \ usr.bin/xargs \ usr.bin/xinstall \ usr.bin/xz \ usr.bin/yacc \ usr.sbin/chown ${_+_}@${ECHODIR} "===> ${_tool} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_tool} && \ ${NXBMAKE} DIRPRFX=${_tool}/ obj && \ ${NXBMAKE} DIRPRFX=${_tool}/ depend && \ ${NXBMAKE} DIRPRFX=${_tool}/ all && \ ${NXBMAKE} DIRPRFX=${_tool}/ DESTDIR=${OBJTREE}/nxb-bin install .endfor # # hierarchy - ensure that all the needed directories are present # hierarchy hier: .MAKE cd ${.CURDIR}/etc && ${HMAKE} distrib-dirs # # libraries - build all libraries, and install them under ${DESTDIR}. # # The list of libraries with dependents (${_prebuild_libs}) and their # interdependencies (__L) are built automatically by the # ${.CURDIR}/tools/make_libdeps.sh script. # libraries: .MAKE cd ${.CURDIR} && \ ${MAKE} -f Makefile.inc1 _prereq_libs && \ ${MAKE} -f Makefile.inc1 _startup_libs && \ ${MAKE} -f Makefile.inc1 _prebuild_libs && \ ${MAKE} -f Makefile.inc1 _generic_libs # # static libgcc.a prerequisite for shared libc # _prereq_libs= gnu/lib/libssp/libssp_nonshared gnu/lib/libgcc lib/libcompiler_rt # These dependencies are not automatically generated: # # gnu/lib/csu, gnu/lib/libgcc, lib/csu and lib/libc must be built before # all shared libraries for ELF. # _startup_libs= gnu/lib/csu _startup_libs+= lib/csu _startup_libs+= gnu/lib/libgcc _startup_libs+= lib/libcompiler_rt _startup_libs+= lib/libc _startup_libs+= lib/libc_nonshared .if ${MK_LIBCPLUSPLUS} != "no" _startup_libs+= lib/libcxxrt .endif gnu/lib/libgcc__L: lib/libc__L gnu/lib/libgcc__L: lib/libc_nonshared__L .if ${MK_LIBCPLUSPLUS} != "no" lib/libcxxrt__L: gnu/lib/libgcc__L .endif _prebuild_libs= ${_kerberos5_lib_libasn1} \ ${_kerberos5_lib_libhdb} \ ${_kerberos5_lib_libheimbase} \ ${_kerberos5_lib_libheimntlm} \ ${_libsqlite3} \ ${_kerberos5_lib_libheimipcc} \ ${_kerberos5_lib_libhx509} ${_kerberos5_lib_libkrb5} \ ${_kerberos5_lib_libroken} \ ${_kerberos5_lib_libwind} \ lib/libbz2 ${_libcom_err} lib/libcrypt \ lib/libelf lib/libexpat \ lib/libfigpar \ ${_lib_libgssapi} \ lib/libkiconv lib/libkvm lib/liblzma lib/libmd lib/libnv \ ${_lib_libcapsicum} \ lib/ncurses/ncurses lib/ncurses/ncursesw \ lib/libopie lib/libpam ${_lib_libthr} \ ${_lib_libradius} lib/libsbuf lib/libtacplus \ lib/libgeom \ ${_cddl_lib_libumem} ${_cddl_lib_libnvpair} \ ${_cddl_lib_libuutil} \ ${_cddl_lib_libavl} \ ${_cddl_lib_libzfs_core} \ ${_cddl_lib_libctf} \ lib/libutil lib/libpjdlog ${_lib_libypclnt} lib/libz lib/msun \ ${_secure_lib_libcrypto} ${_lib_libldns} \ ${_secure_lib_libssh} ${_secure_lib_libssl} \ gnu/lib/libdialog .if ${MK_GNUCXX} != "no" _prebuild_libs+= gnu/lib/libstdc++ gnu/lib/libsupc++ gnu/lib/libstdc++__L: lib/msun__L gnu/lib/libsupc++__L: gnu/lib/libstdc++__L .endif .if ${MK_LIBCPLUSPLUS} != "no" _prebuild_libs+= lib/libc++ .endif lib/libgeom__L: lib/libexpat__L .if ${MK_LIBTHR} != "no" _lib_libthr= lib/libthr .endif .if ${MK_RADIUS_SUPPORT} != "no" _lib_libradius= lib/libradius .endif .if ${MK_OFED} != "no" _ofed_lib= contrib/ofed/usr.lib/ .endif .if ${MK_CASPER} != "no" _lib_libcapsicum=lib/libcapsicum .endif lib/libcapsicum__L: lib/libnv__L lib/libpjdlog__L: lib/libutil__L lib/liblzma__L: lib/libthr__L _generic_libs= ${_cddl_lib} gnu/lib ${_kerberos5_lib} lib ${_secure_lib} usr.bin/lex/lib ${_ofed_lib} .for _DIR in ${LOCAL_LIB_DIRS} .if exists(${.CURDIR}/${_DIR}/Makefile) _generic_libs+= ${_DIR} .endif .endfor lib/libopie__L lib/libtacplus__L: lib/libmd__L .if ${MK_CDDL} != "no" _cddl_lib_libumem= cddl/lib/libumem _cddl_lib_libnvpair= cddl/lib/libnvpair _cddl_lib_libavl= cddl/lib/libavl _cddl_lib_libuutil= cddl/lib/libuutil _cddl_lib_libzfs_core= cddl/lib/libzfs_core _cddl_lib_libctf= cddl/lib/libctf _cddl_lib= cddl/lib cddl/lib/libzfs_core__L: cddl/lib/libnvpair__L cddl/lib/libzfs__L: lib/libgeom__L cddl/lib/libctf__L: lib/libz__L .endif # cddl/lib/libdtrace requires lib/libproc and lib/librtld_db; it's only built # on select architectures though (see cddl/lib/Makefile) .if ${MACHINE_ARCH} == "amd64" || ${MACHINE_ARCH} == "i386" || \ ${MACHINE_CPUARCH} == "mips" || ${MACHINE_CPUARCH} == "powerpc" || \ ${MACHINE_CPUARCH} == "arm" _prebuild_libs+= lib/libproc lib/librtld_db .endif .if ${MK_CRYPT} != "no" .if ${MK_OPENSSL} != "no" _secure_lib_libcrypto= secure/lib/libcrypto _secure_lib_libssl= secure/lib/libssl lib/libradius__L secure/lib/libssl__L: secure/lib/libcrypto__L .if ${MK_LDNS} != "no" _lib_libldns= lib/libldns lib/libldns__L: secure/lib/libcrypto__L .endif .if ${MK_OPENSSH} != "no" _secure_lib_libssh= secure/lib/libssh secure/lib/libssh__L: lib/libz__L secure/lib/libcrypto__L lib/libcrypt__L .if ${MK_LDNS} != "no" secure/lib/libssh__L: lib/libldns__L .endif .if ${MK_KERBEROS_SUPPORT} != "no" secure/lib/libssh__L: lib/libgssapi__L kerberos5/lib/libkrb5__L \ kerberos5/lib/libhx509__L kerberos5/lib/libasn1__L lib/libcom_err__L \ lib/libmd__L kerberos5/lib/libroken__L .endif .endif .endif _secure_lib= secure/lib .endif .if ${MK_KERBEROS} != "no" kerberos5/lib/libasn1__L: lib/libcom_err__L kerberos5/lib/libroken__L kerberos5/lib/libhdb__L: kerberos5/lib/libasn1__L lib/libcom_err__L \ kerberos5/lib/libkrb5__L kerberos5/lib/libroken__L \ kerberos5/lib/libwind__L lib/libsqlite3__L kerberos5/lib/libheimntlm__L: secure/lib/libcrypto__L kerberos5/lib/libkrb5__L \ kerberos5/lib/libroken__L lib/libcom_err__L kerberos5/lib/libhx509__L: kerberos5/lib/libasn1__L lib/libcom_err__L \ secure/lib/libcrypto__L kerberos5/lib/libroken__L kerberos5/lib/libwind__L kerberos5/lib/libkrb5__L: kerberos5/lib/libasn1__L lib/libcom_err__L \ lib/libcrypt__L secure/lib/libcrypto__L kerberos5/lib/libhx509__L \ kerberos5/lib/libroken__L kerberos5/lib/libwind__L \ kerberos5/lib/libheimbase__L kerberos5/lib/libheimipcc__L kerberos5/lib/libroken__L: lib/libcrypt__L kerberos5/lib/libwind__L: kerberos5/lib/libroken__L lib/libcom_err__L kerberos5/lib/libheimbase__L: lib/libthr__L kerberos5/lib/libheimipcc__L: kerberos5/lib/libroken__L kerberos5/lib/libheimbase__L lib/libthr__L .endif lib/libsqlite3__L: lib/libthr__L .if ${MK_GSSAPI} != "no" _lib_libgssapi= lib/libgssapi .endif .if ${MK_KERBEROS} != "no" _kerberos5_lib= kerberos5/lib _kerberos5_lib_libasn1= kerberos5/lib/libasn1 _kerberos5_lib_libhdb= kerberos5/lib/libhdb _kerberos5_lib_libheimbase= kerberos5/lib/libheimbase _kerberos5_lib_libkrb5= kerberos5/lib/libkrb5 _kerberos5_lib_libhx509= kerberos5/lib/libhx509 _kerberos5_lib_libroken= kerberos5/lib/libroken _kerberos5_lib_libheimntlm= kerberos5/lib/libheimntlm _libsqlite3= lib/libsqlite3 _kerberos5_lib_libheimipcc= kerberos5/lib/libheimipcc _kerberos5_lib_libwind= kerberos5/lib/libwind _libcom_err= lib/libcom_err .endif .if ${MK_NIS} != "no" _lib_libypclnt= lib/libypclnt .endif .if ${MK_OPENSSL} == "no" lib/libradius__L: lib/libmd__L .endif lib/libproc__L: \ ${_cddl_lib_libctf:D${_cddl_lib_libctf}__L} lib/libelf__L lib/librtld_db__L lib/libutil__L .if ${MK_CXX} != "no" .if ${MK_LIBCPLUSPLUS} != "no" lib/libproc__L: lib/libcxxrt__L .else # This implies MK_GNUCXX != "no"; see lib/libproc lib/libproc__L: gnu/lib/libsupc++__L .endif .endif gnu/lib/libdialog__L: lib/msun__L lib/ncurses/ncursesw__L .for _lib in ${_prereq_libs} ${_lib}__PL: .PHONY .MAKE .if exists(${.CURDIR}/${_lib}) ${_+_}@${ECHODIR} "===> ${_lib} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_lib} && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ obj && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ depend && \ ${MAKE} MK_TESTS=no MK_PROFILE=no -DNO_PIC \ DIRPRFX=${_lib}/ all && \ ${MAKE} MK_TESTS=no MK_PROFILE=no -DNO_PIC \ DIRPRFX=${_lib}/ install .endif .endfor .for _lib in ${_startup_libs} ${_prebuild_libs:Nlib/libpam} ${_generic_libs} ${_lib}__L: .PHONY .MAKE .if exists(${.CURDIR}/${_lib}) ${_+_}@${ECHODIR} "===> ${_lib} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_lib} && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ obj && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ depend && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ all && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ install .endif .endfor # libpam is special: we need to build static PAM modules before # static PAM library, and dynamic PAM library before dynamic PAM # modules. lib/libpam__L: .PHONY .MAKE ${_+_}@${ECHODIR} "===> lib/libpam (obj,depend,all,install)"; \ cd ${.CURDIR}/lib/libpam && \ ${MAKE} MK_TESTS=no DIRPRFX=lib/libpam/ obj && \ ${MAKE} MK_TESTS=no DIRPRFX=lib/libpam/ depend && \ ${MAKE} MK_TESTS=no DIRPRFX=lib/libpam/ \ -D_NO_LIBPAM_SO_YET all && \ ${MAKE} MK_TESTS=no DIRPRFX=lib/libpam/ \ -D_NO_LIBPAM_SO_YET install _prereq_libs: ${_prereq_libs:S/$/__PL/} _startup_libs: ${_startup_libs:S/$/__L/} _prebuild_libs: ${_prebuild_libs:S/$/__L/} _generic_libs: ${_generic_libs:S/$/__L/} .for __target in all clean cleandepend cleandir depend includes obj .for entry in ${SUBDIR} ${entry}.${__target}__D: .PHONY .MAKE ${_+_}@set -e; if test -d ${.CURDIR}/${entry}.${MACHINE_ARCH}; then \ ${ECHODIR} "===> ${DIRPRFX}${entry}.${MACHINE_ARCH} (${__target})"; \ edir=${entry}.${MACHINE_ARCH}; \ cd ${.CURDIR}/$${edir}; \ else \ ${ECHODIR} "===> ${DIRPRFX}${entry} (${__target})"; \ edir=${entry}; \ cd ${.CURDIR}/$${edir}; \ fi; \ ${MAKE} ${__target} DIRPRFX=${DIRPRFX}$${edir}/ .endfor par-${__target}: ${SUBDIR:S/$/.${__target}__D/} .endfor .include .if make(check-old) || make(check-old-dirs) || \ make(check-old-files) || make(check-old-libs) || \ make(delete-old) || make(delete-old-dirs) || \ make(delete-old-files) || make(delete-old-libs) # # check for / delete old files section # .include "ObsoleteFiles.inc" OLD_LIBS_MESSAGE="Please be sure no application still uses those libraries, \ else you can not start such an application. Consult UPDATING for more \ information regarding how to cope with the removal/revision bump of a \ specific library." .if !defined(BATCH_DELETE_OLD_FILES) RM_I=-i .else RM_I=-v .endif delete-old-files: @echo ">>> Removing old files (only deletes safe to delete libs)" # Ask for every old file if the user really wants to remove it. # It's annoying, but better safe than sorry. # NB: We cannot pass the list of OLD_FILES as a parameter because the # argument list will get too long. Using .for/.endfor make "loops" will make # the Makefile parser segfault. @exec 3<&0; \ cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_FILES -V "OLD_FILES:Musr/share/*.gz:R" | xargs -n1 | \ while read file; do \ if [ -f "${DESTDIR}/$${file}" -o -L "${DESTDIR}/$${file}" ]; then \ chflags noschg "${DESTDIR}/$${file}" 2>/dev/null || true; \ rm ${RM_I} "${DESTDIR}/$${file}" <&3; \ fi; \ done # Remove catpages without corresponding manpages. @exec 3<&0; \ find ${DESTDIR}/usr/share/man/cat* ! -type d | \ sed -ep -e's:${DESTDIR}/usr/share/man/cat:${DESTDIR}/usr/share/man/man:' | \ while read catpage; do \ read manpage; \ if [ ! -e "$${manpage}" ]; then \ rm ${RM_I} $${catpage} <&3; \ fi; \ done @echo ">>> Old files removed" check-old-files: @echo ">>> Checking for old files" @cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_FILES -V "OLD_FILES:Musr/share/*.gz:R" | xargs -n1 | \ while read file; do \ if [ -f "${DESTDIR}/$${file}" -o -L "${DESTDIR}/$${file}" ]; then \ echo "${DESTDIR}/$${file}"; \ fi; \ done # Check for catpages without corresponding manpages. @find ${DESTDIR}/usr/share/man/cat* ! -type d | \ sed -ep -e's:${DESTDIR}/usr/share/man/cat:${DESTDIR}/usr/share/man/man:' | \ while read catpage; do \ read manpage; \ if [ ! -e "$${manpage}" ]; then \ echo $${catpage}; \ fi; \ done delete-old-libs: @echo ">>> Removing old libraries" @echo "${OLD_LIBS_MESSAGE}" | fmt @exec 3<&0; \ cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_LIBS | xargs -n1 | \ while read file; do \ if [ -f "${DESTDIR}/$${file}" -o -L "${DESTDIR}/$${file}" ]; then \ chflags noschg "${DESTDIR}/$${file}" 2>/dev/null || true; \ rm ${RM_I} "${DESTDIR}/$${file}" <&3; \ fi; \ for ext in debug symbols; do \ if ! [ -e "${DESTDIR}/$${file}" ] && [ -f \ "${DESTDIR}${DEBUGDIR}/$${file}.$${ext}" ]; then \ rm ${RM_I} "${DESTDIR}${DEBUGDIR}/$${file}.$${ext}" \ <&3; \ fi; \ done; \ done @echo ">>> Old libraries removed" check-old-libs: @echo ">>> Checking for old libraries" @cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_LIBS | xargs -n1 | \ while read file; do \ if [ -f "${DESTDIR}/$${file}" -o -L "${DESTDIR}/$${file}" ]; then \ echo "${DESTDIR}/$${file}"; \ fi; \ for ext in debug symbols; do \ if [ -f "${DESTDIR}${DEBUGDIR}/$${file}.$${ext}" ]; then \ echo "${DESTDIR}${DEBUGDIR}/$${file}.$${ext}"; \ fi; \ done; \ done delete-old-dirs: @echo ">>> Removing old directories" @cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_DIRS | xargs -n1 | sort -r | \ while read dir; do \ if [ -d "${DESTDIR}/$${dir}" ]; then \ rmdir -v "${DESTDIR}/$${dir}" || true; \ elif [ -L "${DESTDIR}/$${dir}" ]; then \ echo "${DESTDIR}/$${dir} is a link, please remove everything manually."; \ fi; \ done @echo ">>> Old directories removed" check-old-dirs: @echo ">>> Checking for old directories" @cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_DIRS | xargs -n1 | \ while read dir; do \ if [ -d "${DESTDIR}/$${dir}" ]; then \ echo "${DESTDIR}/$${dir}"; \ elif [ -L "${DESTDIR}/$${dir}" ]; then \ echo "${DESTDIR}/$${dir} is a link, please remove everything manually."; \ fi; \ done delete-old: delete-old-files delete-old-dirs @echo "To remove old libraries run '${MAKE} delete-old-libs'." check-old: check-old-files check-old-libs check-old-dirs @echo "To remove old files and directories run '${MAKE} delete-old'." @echo "To remove old libraries run '${MAKE} delete-old-libs'." .endif # # showconfig - show build configuration. # showconfig: @(${MAKE} -n -f ${.CURDIR}/sys/conf/kern.opts.mk -V dummy -dg1; \ ${MAKE} -n -f ${.CURDIR}/share/mk/src.opts.mk -V dummy -dg1) 2>&1 | grep ^MK_ | sort -u .if !empty(KRNLOBJDIR) && !empty(KERNCONF) DTBOUTPUTPATH= ${KRNLOBJDIR}/${KERNCONF}/ .if !defined(FDT_DTS_FILE) || empty(FDT_DTS_FILE) .if exists(${KERNCONFDIR}/${KERNCONF}) FDT_DTS_FILE!= awk 'BEGIN {FS="="} /^makeoptions[[:space:]]+FDT_DTS_FILE/ {print $$2}' \ '${KERNCONFDIR}/${KERNCONF}' ; echo .endif .endif .endif .if !defined(DTBOUTPUTPATH) || !exists(${DTBOUTPUTPATH}) DTBOUTPUTPATH= ${.CURDIR} .endif # # Build 'standalone' Device Tree Blob # builddtb: @PATH=${TMPPATH} MACHINE=${TARGET} \ ${.CURDIR}/sys/tools/fdt/make_dtb.sh ${.CURDIR}/sys \ "${FDT_DTS_FILE}" ${DTBOUTPUTPATH} ############### .if defined(TARGET) && defined(TARGET_ARCH) .if ${TARGET} == ${MACHINE} && ${TARGET_ARCH} == ${MACHINE_ARCH} XDEV_CPUTYPE?=${CPUTYPE} .else XDEV_CPUTYPE?=${TARGET_CPUTYPE} .endif NOFUN=-DNO_FSCHG MK_HTML=no -DNO_LINT \ MK_MAN=no MK_NLS=no MK_PROFILE=no \ MK_KERBEROS=no MK_RESCUE=no MK_TESTS=no MK_WARNS=no \ TARGET=${TARGET} TARGET_ARCH=${TARGET_ARCH} \ CPUTYPE=${XDEV_CPUTYPE} XDDIR=${TARGET_ARCH}-freebsd XDTP?=/usr/${XDDIR} .if ${XDTP:N/*} .error XDTP variable should be an absolute path .endif CDBENV=MAKEOBJDIRPREFIX=${MAKEOBJDIRPREFIX}/${XDDIR} \ INSTALL="sh ${.CURDIR}/tools/install.sh" CDENV= ${CDBENV} \ _SHLIBDIRPREFIX=${XDDESTDIR} \ TOOLS_PREFIX=${XDTP} CD2CFLAGS=-isystem ${XDDESTDIR}/usr/include -L${XDDESTDIR}/usr/lib \ --sysroot=${XDDESTDIR}/ -B${XDDESTDIR}/usr/libexec \ -B${XDDESTDIR}/usr/bin -B${XDDESTDIR}/usr/lib CD2ENV=${CDENV} CC="${CC} ${CD2CFLAGS}" CXX="${CXX} ${CD2CFLAGS}" \ CPP="${CPP} ${CD2CFLAGS}" \ MACHINE=${TARGET} MACHINE_ARCH=${TARGET_ARCH} CDTMP= ${MAKEOBJDIRPREFIX}/${XDDIR}/${.CURDIR}/tmp CDMAKE=${CDENV} PATH=${CDTMP}/usr/bin:${PATH} ${MAKE} ${NOFUN} CD2MAKE=${CD2ENV} PATH=${CDTMP}/usr/bin:${XDDESTDIR}/usr/bin:${PATH} ${MAKE} ${NOFUN} XDDESTDIR=${DESTDIR}/${XDTP} .if !defined(OSREL) OSREL!= uname -r | sed -e 's/[-(].*//' .endif .ORDER: xdev-build xdev-install xdev-links xdev: xdev-build xdev-install .ORDER: _xb-worldtmp _xb-bootstrap-tools _xb-build-tools _xb-cross-tools xdev-build: _xb-worldtmp _xb-bootstrap-tools _xb-build-tools _xb-cross-tools _xb-worldtmp: mkdir -p ${CDTMP}/usr mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${CDTMP}/usr >/dev/null _xb-bootstrap-tools: .for _tool in \ ${_clang_tblgen} \ ${_gperf} ${_+_}@${ECHODIR} "===> ${_tool} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_tool} && \ ${CDMAKE} DIRPRFX=${_tool}/ obj && \ ${CDMAKE} DIRPRFX=${_tool}/ depend && \ ${CDMAKE} DIRPRFX=${_tool}/ all && \ ${CDMAKE} DIRPRFX=${_tool}/ DESTDIR=${CDTMP} install .endfor _xb-build-tools: ${_+_}@cd ${.CURDIR}; \ ${CDBENV} ${MAKE} -f Makefile.inc1 ${NOFUN} build-tools _xb-cross-tools: .for _tool in \ ${_binutils} \ ${_elftctools} \ usr.bin/ar \ ${_clang_libs} \ ${_clang} \ ${_cc} ${_+_}@${ECHODIR} "===> xdev ${_tool} (obj,depend,all)"; \ cd ${.CURDIR}/${_tool} && \ ${CDMAKE} DIRPRFX=${_tool}/ obj && \ ${CDMAKE} DIRPRFX=${_tool}/ depend && \ ${CDMAKE} DIRPRFX=${_tool}/ all .endfor _xi-mtree: ${_+_}@${ECHODIR} "mtree populating ${XDDESTDIR}" mkdir -p ${XDDESTDIR} mtree -deU -f ${.CURDIR}/etc/mtree/BSD.root.dist \ -p ${XDDESTDIR} >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${XDDESTDIR}/usr >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.include.dist \ -p ${XDDESTDIR}/usr/include >/dev/null .if ${MK_TESTS} != "no" mkdir -p ${XDDESTDIR}${TESTSBASE} mtree -deU -f ${.CURDIR}/etc/mtree/BSD.tests.dist \ -p ${XDDESTDIR}${TESTSBASE} >/dev/null .endif .ORDER: xdev-build _xi-mtree _xi-cross-tools _xi-includes _xi-libraries xdev-install: xdev-build _xi-mtree _xi-cross-tools _xi-includes _xi-libraries _xi-cross-tools: @echo "_xi-cross-tools" .for _tool in \ ${_binutils} \ ${_elftctools} \ usr.bin/ar \ ${_clang_libs} \ ${_clang} \ ${_cc} ${_+_}@${ECHODIR} "===> xdev ${_tool} (install)"; \ cd ${.CURDIR}/${_tool}; \ ${CDMAKE} DIRPRFX=${_tool}/ install DESTDIR=${XDDESTDIR} .endfor _xi-includes: ${_+_}cd ${.CURDIR}; ${CD2MAKE} -f Makefile.inc1 par-includes \ DESTDIR=${XDDESTDIR} _xi-libraries: ${_+_}cd ${.CURDIR}; ${CD2MAKE} -f Makefile.inc1 libraries \ DESTDIR=${XDDESTDIR} xdev-links: ${_+_}cd ${XDDESTDIR}/usr/bin; \ mkdir -p ../../../../usr/bin; \ for i in *; do \ ln -sf ../../${XDTP}/usr/bin/$$i \ ../../../../usr/bin/${XDDIR}-$$i; \ ln -sf ../../${XDTP}/usr/bin/$$i \ ../../../../usr/bin/${XDDIR}${OSREL}-$$i; \ done .else xdev xdev-build xdev-install xdev-links: @echo "*** Error: Both TARGET and TARGET_ARCH must be defined for \"${.TARGET}\" target" .endif Index: projects/em_mq/games/grdc/grdc.c =================================================================== --- projects/em_mq/games/grdc/grdc.c (revision 283943) +++ projects/em_mq/games/grdc/grdc.c (revision 283944) @@ -1,273 +1,273 @@ /* * Grand digital clock for curses compatible terminals * Usage: grdc [-st] [n] -- run for n seconds (default infinity) * Flags: -s: scroll * -t: output time in 12-hour format * * * modified 10-18-89 for curses (jrl) * 10-18-89 added signal handling * * modified 03-25-03 for 12 hour option * - Samy Al Bahra * * $FreeBSD$ */ #include #include #include #include #include #include #define YBASE 10 #define XBASE 10 #define XLENGTH 58 #define YDEPTH 7 static struct timespec now; static struct tm *tm; static short disp[11] = { 075557, 011111, 071747, 071717, 055711, 074717, 074757, 071111, 075757, 075717, 002020 }; static long old[6], next[6], new[6], mask; static volatile sig_atomic_t sigtermed; static int hascolor = 0; static void set(int, int); static void standt(int); static void movto(int, int); static void sighndl(int); static void usage(void); static void sighndl(int signo) { sigtermed = signo; } int main(int argc, char *argv[]) { struct timespec delay; time_t prev_sec; long t, a; int i, j, s, k; int n; int ch; int scrol; int t12; t12 = scrol = 0; while ((ch = getopt(argc, argv, "ts")) != -1) switch (ch) { case 's': scrol = 1; break; case 't': t12 = 1; break; case '?': default: usage(); /* NOTREACHED */ } argc -= optind; argv += optind; if (argc > 1) { usage(); /* NOTREACHED */ } if (argc > 0) { n = atoi(*argv) + 1; if (n < 1) { warnx("number of seconds is out of range"); usage(); /* NOTREACHED */ } } else n = 0; initscr(); signal(SIGINT,sighndl); signal(SIGTERM,sighndl); signal(SIGHUP,sighndl); cbreak(); noecho(); curs_set(0); hascolor = has_colors(); if(hascolor) { start_color(); init_pair(1, COLOR_BLACK, COLOR_RED); init_pair(2, COLOR_RED, COLOR_BLACK); init_pair(3, COLOR_WHITE, COLOR_BLACK); attrset(COLOR_PAIR(2)); } clear(); refresh(); if(hascolor) { attrset(COLOR_PAIR(3)); mvaddch(YBASE - 2, XBASE - 3, ACS_ULCORNER); hline(ACS_HLINE, XLENGTH); mvaddch(YBASE - 2, XBASE - 2 + XLENGTH, ACS_URCORNER); mvaddch(YBASE + YDEPTH - 1, XBASE - 3, ACS_LLCORNER); hline(ACS_HLINE, XLENGTH); mvaddch(YBASE + YDEPTH - 1, XBASE - 2 + XLENGTH, ACS_LRCORNER); move(YBASE - 1, XBASE - 3); vline(ACS_VLINE, YDEPTH); move(YBASE - 1, XBASE - 2 + XLENGTH); vline(ACS_VLINE, YDEPTH); attrset(COLOR_PAIR(2)); } clock_gettime(CLOCK_REALTIME_FAST, &now); prev_sec = now.tv_sec; do { mask = 0; tm = localtime(&now.tv_sec); set(tm->tm_sec%10, 0); set(tm->tm_sec/10, 4); set(tm->tm_min%10, 10); set(tm->tm_min/10, 14); if (t12) { - if (tm->tm_hour > 12) { - tm->tm_hour -= 12; - mvaddstr(YBASE + 5, XBASE + 52, "PM"); - } else { + if (tm->tm_hour < 12) { if (tm->tm_hour == 0) tm->tm_hour = 12; - mvaddstr(YBASE + 5, XBASE + 52, "AM"); + } else { + if (tm->tm_hour > 12) + tm->tm_hour -= 12; + mvaddstr(YBASE + 5, XBASE + 52, "PM"); } } set(tm->tm_hour%10, 20); set(tm->tm_hour/10, 24); set(10, 7); set(10, 17); for(k=0; k<6; k++) { if(scrol) { for(i=0; i<5; i++) new[i] = (new[i]&~mask) | (new[i+1]&mask); new[5] = (new[5]&~mask) | (next[k]&mask); } else new[k] = (new[k]&~mask) | (next[k]&mask); next[k] = 0; for(s=1; s>=0; s--) { standt(s); for(i=0; i<6; i++) { if((a = (new[i]^old[i])&(s ? new : old)[i]) != 0) { for(j=0,t=1<<26; t; t>>=1,j++) { if(a&t) { if(!(a&(t<<1))) { movto(YBASE + i, XBASE + 2*j); } addstr(" "); } } } if(!s) { old[i] = new[i]; } } if(!s) { refresh(); } } } movto(6, 0); refresh(); clock_gettime(CLOCK_REALTIME_FAST, &now); if (now.tv_sec == prev_sec) { if (delay.tv_nsec > 0) { delay.tv_sec = 0; delay.tv_nsec = 1000000000 - now.tv_nsec; } else { delay.tv_sec = 1; delay.tv_nsec = 0; } nanosleep(&delay, NULL); clock_gettime(CLOCK_REALTIME_FAST, &now); } n -= now.tv_sec - prev_sec; prev_sec = now.tv_sec; if (sigtermed) { standend(); clear(); refresh(); endwin(); errx(1, "terminated by signal %d", (int)sigtermed); } } while (n); standend(); clear(); refresh(); endwin(); return(0); } static void set(int t, int n) { int i, m; m = 7<>(4-i)*3)&07)< .if defined(LIB_CXX) LIB= ${LIB_CXX} _LD= ${CXX} .else _LD= ${CC} .endif LIB_PRIVATE= ${PRIVATELIB:Dprivate} # Set up the variables controlling shared libraries. After this section, # SHLIB_NAME will be defined only if we are to create a shared library. # SHLIB_LINK will be defined only if we are to create a link to it. # INSTALL_PIC_ARCHIVE will be defined only if we are to create a PIC archive. .if defined(NO_PIC) .undef SHLIB_NAME .undef INSTALL_PIC_ARCHIVE .else .if !defined(SHLIB) && defined(LIB) SHLIB= ${LIB} .endif .if !defined(SHLIB_NAME) && defined(SHLIB) && defined(SHLIB_MAJOR) SHLIB_NAME= lib${LIB_PRIVATE}${SHLIB}.so.${SHLIB_MAJOR} .endif .if defined(SHLIB_NAME) && !empty(SHLIB_NAME:M*.so.*) SHLIB_LINK?= ${SHLIB_NAME:R} .endif SONAME?= ${SHLIB_NAME} .endif .if defined(CRUNCH_CFLAGS) CFLAGS+= ${CRUNCH_CFLAGS} .endif .if ${MK_ASSERT_DEBUG} == "no" CFLAGS+= -DNDEBUG NO_WERROR= .endif .if defined(DEBUG_FLAGS) CFLAGS+= ${DEBUG_FLAGS} .if ${MK_CTF} != "no" && ${DEBUG_FLAGS:M-g} != "" CTFFLAGS+= -g .endif .else STRIP?= -s .endif .if ${MK_DEBUG_FILES} != "no" && empty(DEBUG_FLAGS:M-g) && \ empty(DEBUG_FLAGS:M-gdwarf*) SHARED_CFLAGS+= -g SHARED_CXXFLAGS+= -g CTFFLAGS+= -g .endif .include # prefer .s to a .c, add .po, remove stuff not used in the BSD libraries # .So used for PIC object files .SUFFIXES: .SUFFIXES: .out .o .po .So .S .asm .s .c .cc .cpp .cxx .C .f .y .l .ln .if !defined(PICFLAG) .if ${MACHINE_CPUARCH} == "sparc64" PICFLAG=-fPIC .else PICFLAG=-fpic .endif .endif PO_FLAG=-pg .c.o: ${CC} ${STATIC_CFLAGS} ${CFLAGS} -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .c.po: ${CC} ${PO_FLAG} ${STATIC_CFLAGS} ${PO_CFLAGS} -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .c.So: ${CC} ${PICFLAG} -DPIC ${SHARED_CFLAGS} ${CFLAGS} -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .cc.o .C.o .cpp.o .cxx.o: ${CXX} ${STATIC_CXXFLAGS} ${CXXFLAGS} -c ${.IMPSRC} -o ${.TARGET} .cc.po .C.po .cpp.po .cxx.po: ${CXX} ${PO_FLAG} ${STATIC_CXXFLAGS} ${PO_CXXFLAGS} -c ${.IMPSRC} -o ${.TARGET} .cc.So .C.So .cpp.So .cxx.So: ${CXX} ${PICFLAG} -DPIC ${SHARED_CXXFLAGS} ${CXXFLAGS} -c ${.IMPSRC} -o ${.TARGET} .f.po: ${FC} -pg ${FFLAGS} -o ${.TARGET} -c ${.IMPSRC} ${CTFCONVERT_CMD} .f.So: ${FC} ${PICFLAG} -DPIC ${FFLAGS} -o ${.TARGET} -c ${.IMPSRC} ${CTFCONVERT_CMD} .s.po .s.So: ${AS} ${AFLAGS} -o ${.TARGET} ${.IMPSRC} ${CTFCONVERT_CMD} .asm.po: ${CC} -x assembler-with-cpp -DPROF ${PO_CFLAGS} ${ACFLAGS} \ -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .asm.So: ${CC} -x assembler-with-cpp ${PICFLAG} -DPIC ${CFLAGS} ${ACFLAGS} \ -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .S.po: ${CC} -DPROF ${PO_CFLAGS} ${ACFLAGS} -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .S.So: ${CC} ${PICFLAG} -DPIC ${CFLAGS} ${ACFLAGS} -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} all: beforebuild .WAIT beforebuild: objwarn _LIBDIR:=${LIBDIR} _SHLIBDIR:=${SHLIBDIR} .if defined(SHLIB_NAME) .if ${MK_DEBUG_FILES} != "no" SHLIB_NAME_FULL=${SHLIB_NAME}.full # Use ${DEBUGDIR} for base system debug files, else .debug subdirectory .if ${_SHLIBDIR} == "/boot" ||\ ${SHLIBDIR:C%/lib(/.*)?$%/lib%} == "/lib" ||\ ${SHLIBDIR:C%/usr/lib(32)?(/.*)?%/usr/lib%} == "/usr/lib" DEBUGFILEDIR=${DEBUGDIR}${_SHLIBDIR} .else DEBUGFILEDIR=${_SHLIBDIR}/.debug DEBUGMKDIR= .endif .else SHLIB_NAME_FULL=${SHLIB_NAME} .endif .endif .include # Allow libraries to specify their own version map or have it # automatically generated (see bsd.symver.mk above). .if ${MK_SYMVER} == "yes" && !empty(VERSION_MAP) ${SHLIB_NAME_FULL}: ${VERSION_MAP} LDFLAGS+= -Wl,--version-script=${VERSION_MAP} .endif .if defined(LIB) && !empty(LIB) || defined(SHLIB_NAME) OBJS+= ${SRCS:N*.h:R:S/$/.o/} NOPATH_FILES+= ${OBJS} .endif .if defined(LIB) && !empty(LIB) _LIBS= lib${LIB_PRIVATE}${LIB}.a lib${LIB_PRIVATE}${LIB}.a: ${OBJS} ${STATICOBJS} @${ECHO} building static ${LIB} library @rm -f ${.TARGET} - @${AR} ${ARFLAGS} ${.TARGET} `NM='${NM}' lorder ${OBJS} ${STATICOBJS} | tsort -q` ${ARADD} + ${AR} ${ARFLAGS} ${.TARGET} `NM='${NM}' lorder ${OBJS} ${STATICOBJS} | tsort -q` ${ARADD} ${RANLIB} ${RANLIBFLAGS} ${.TARGET} .endif .if !defined(INTERNALLIB) .if ${MK_PROFILE} != "no" && defined(LIB) && !empty(LIB) _LIBS+= lib${LIB_PRIVATE}${LIB}_p.a POBJS+= ${OBJS:.o=.po} ${STATICOBJS:.o=.po} NOPATH_FILES+= ${POBJS} lib${LIB_PRIVATE}${LIB}_p.a: ${POBJS} @${ECHO} building profiled ${LIB} library @rm -f ${.TARGET} - @${AR} ${ARFLAGS} ${.TARGET} `NM='${NM}' lorder ${POBJS} | tsort -q` ${ARADD} + ${AR} ${ARFLAGS} ${.TARGET} `NM='${NM}' lorder ${POBJS} | tsort -q` ${ARADD} ${RANLIB} ${RANLIBFLAGS} ${.TARGET} .endif .if defined(SHLIB_NAME) || \ defined(INSTALL_PIC_ARCHIVE) && defined(LIB) && !empty(LIB) SOBJS+= ${OBJS:.o=.So} NOPATH_FILES+= ${SOBJS} .endif .if defined(SHLIB_NAME) _LIBS+= ${SHLIB_NAME} SOLINKOPTS= -shared -Wl,-x .if !defined(ALLOW_SHARED_TEXTREL) .if defined(LD_FATAL_WARNINGS) && ${LD_FATAL_WARNINGS} == "no" SOLINKOPTS+= -Wl,--no-fatal-warnings .else SOLINKOPTS+= -Wl,--fatal-warnings .endif SOLINKOPTS+= -Wl,--warn-shared-textrel .endif .if target(beforelinking) beforelinking: ${SOBJS} ${SHLIB_NAME_FULL}: beforelinking .endif ${SHLIB_NAME_FULL}: ${SOBJS} @${ECHO} building shared library ${SHLIB_NAME} @rm -f ${SHLIB_NAME} ${SHLIB_LINK} .if defined(SHLIB_LINK) @${INSTALL_SYMLINK} ${SHLIB_NAME} ${SHLIB_LINK} .endif ${_LD} ${LDFLAGS} ${SSP_CFLAGS} ${SOLINKOPTS} \ -o ${.TARGET} -Wl,-soname,${SONAME} \ `NM='${NM}' lorder ${SOBJS} | tsort -q` ${LDADD} .if ${MK_CTF} != "no" ${CTFMERGE} ${CTFFLAGS} -o ${.TARGET} ${SOBJS} .endif .if ${MK_DEBUG_FILES} != "no" CLEANFILES+= ${SHLIB_NAME_FULL} ${SHLIB_NAME}.debug ${SHLIB_NAME}: ${SHLIB_NAME_FULL} ${SHLIB_NAME}.debug ${OBJCOPY} --strip-debug --add-gnu-debuglink=${SHLIB_NAME}.debug \ ${SHLIB_NAME_FULL} ${.TARGET} ${SHLIB_NAME}.debug: ${SHLIB_NAME_FULL} ${OBJCOPY} --only-keep-debug ${SHLIB_NAME_FULL} ${.TARGET} .endif .endif #defined(SHLIB_NAME) .if defined(INSTALL_PIC_ARCHIVE) && defined(LIB) && !empty(LIB) && ${MK_TOOLCHAIN} != "no" _LIBS+= lib${LIB_PRIVATE}${LIB}_pic.a lib${LIB_PRIVATE}${LIB}_pic.a: ${SOBJS} @${ECHO} building special pic ${LIB} library @rm -f ${.TARGET} - @${AR} ${ARFLAGS} ${.TARGET} ${SOBJS} ${ARADD} + ${AR} ${ARFLAGS} ${.TARGET} ${SOBJS} ${ARADD} ${RANLIB} ${RANLIBFLAGS} ${.TARGET} .endif .if defined(WANT_LINT) && !defined(NO_LINT) && defined(LIB) && !empty(LIB) LINTLIB= llib-l${LIB}.ln _LIBS+= ${LINTLIB} LINTOBJS+= ${SRCS:M*.c:.c=.ln} NOPATH_FILES+= ${LINTOBJS} ${LINTLIB}: ${LINTOBJS} @${ECHO} building lint library ${.TARGET} @rm -f ${.TARGET} ${LINT} ${LINTLIBFLAGS} ${CFLAGS:M-[DIU]*} ${.ALLSRC} .endif .endif # !defined(INTERNALLIB) all: ${_LIBS} .if ${MK_MAN} != "no" all: _manpages .endif _EXTRADEPEND: @TMP=_depend$$$$; \ sed -e 's/^\([^\.]*\).o[ ]*:/\1.o \1.po \1.So:/' < ${DEPENDFILE} \ > $$TMP; \ mv $$TMP ${DEPENDFILE} .if !defined(NO_EXTRADEPEND) && defined(SHLIB_NAME) .if defined(DPADD) && !empty(DPADD) echo ${SHLIB_NAME_FULL}: ${DPADD} >> ${DEPENDFILE} .endif .endif .if !target(install) .if defined(PRECIOUSLIB) .if !defined(NO_FSCHG) SHLINSTALLFLAGS+= -fschg .endif SHLINSTALLFLAGS+= -S .endif _INSTALLFLAGS:= ${INSTALLFLAGS} .for ie in ${INSTALLFLAGS_EDIT} _INSTALLFLAGS:= ${_INSTALLFLAGS${ie}} .endfor _SHLINSTALLFLAGS:= ${SHLINSTALLFLAGS} .for ie in ${INSTALLFLAGS_EDIT} _SHLINSTALLFLAGS:= ${_SHLINSTALLFLAGS${ie}} .endfor .if !defined(INTERNALLIB) realinstall: _libinstall .ORDER: beforeinstall _libinstall _libinstall: .if defined(LIB) && !empty(LIB) && ${MK_INSTALLLIB} != "no" ${INSTALL} -C -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} lib${LIB_PRIVATE}${LIB}.a ${DESTDIR}${_LIBDIR} .endif .if ${MK_PROFILE} != "no" && defined(LIB) && !empty(LIB) ${INSTALL} -C -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} lib${LIB_PRIVATE}${LIB}_p.a ${DESTDIR}${_LIBDIR} .endif .if defined(SHLIB_NAME) ${INSTALL} ${STRIP} -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} ${_SHLINSTALLFLAGS} \ ${SHLIB_NAME} ${DESTDIR}${_SHLIBDIR} .if ${MK_DEBUG_FILES} != "no" .if defined(DEBUGMKDIR) ${INSTALL} -T debug -d ${DESTDIR}${DEBUGFILEDIR} .endif ${INSTALL} -T debug -o ${LIBOWN} -g ${LIBGRP} -m ${DEBUGMODE} \ ${_INSTALLFLAGS} \ ${SHLIB_NAME}.debug ${DESTDIR}${DEBUGFILEDIR} .endif .if defined(SHLIB_LINK) # ${_SHLIBDIRPREFIX} and ${_LDSCRIPTROOT} are both needed when cross-building # and when building 32 bits library shims. ${_SHLIBDIRPREFIX} is the directory # prefix where shared objects will be installed by the install target. # # ${_LDSCRIPTROOT} is the directory prefix that will be used when generating # ld(1) scripts. The crosstools' ld is configured to lookup libraries in an # alternative directory which is called "sysroot", so during buildworld binaries # won't be linked against the running system libraries but against the ones of # the current source tree. ${_LDSCRIPTROOT} behavior is twisted because of # the location where we store them: # - 64 bits libs are located under sysroot, so ${_LDSCRIPTROOT} must be empty # because ld(1) will manage to find them from sysroot; # - 32 bits shims are not, so ${_LDSCRIPTROOT} is used to specify their full # path, outside of sysroot. # Note that ld(1) scripts are generated both during buildworld and # installworld; in the later case ${_LDSCRIPTROOT} must be obviously empty # because on the target system, libraries are meant to be looked up from /. .if defined(SHLIB_LDSCRIPT) && !empty(SHLIB_LDSCRIPT) && exists(${.CURDIR}/${SHLIB_LDSCRIPT}) sed -e 's,@@SHLIB@@,${_LDSCRIPTROOT}${_SHLIBDIR}/${SHLIB_NAME},g' \ -e 's,@@LIBDIR@@,${_LDSCRIPTROOT}${_LIBDIR},g' \ ${.CURDIR}/${SHLIB_LDSCRIPT} > ${DESTDIR}${_LIBDIR}/${SHLIB_LINK:R}.ld ${INSTALL} -S -C -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} ${DESTDIR}${_LIBDIR}/${SHLIB_LINK:R}.ld \ ${DESTDIR}${_LIBDIR}/${SHLIB_LINK} rm -f ${DESTDIR}${_LIBDIR}/${SHLIB_LINK:R}.ld .else .if ${_SHLIBDIR} == ${_LIBDIR} ${INSTALL_SYMLINK} ${SHLIB_NAME} ${DESTDIR}${_LIBDIR}/${SHLIB_LINK} .else ${INSTALL_SYMLINK} ${_SHLIBDIRPREFIX}${_SHLIBDIR}/${SHLIB_NAME} \ ${DESTDIR}${_LIBDIR}/${SHLIB_LINK} .if exists(${DESTDIR}${_LIBDIR}/${SHLIB_NAME}) -chflags noschg ${DESTDIR}${_LIBDIR}/${SHLIB_NAME} rm -f ${DESTDIR}${_LIBDIR}/${SHLIB_NAME} .endif .endif .endif # SHLIB_LDSCRIPT .endif # SHLIB_LINK .endif # SHIB_NAME .if defined(INSTALL_PIC_ARCHIVE) && defined(LIB) && !empty(LIB) && ${MK_TOOLCHAIN} != "no" ${INSTALL} -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} lib${LIB}_pic.a ${DESTDIR}${_LIBDIR} .endif .if defined(WANT_LINT) && !defined(NO_LINT) && defined(LIB) && !empty(LIB) ${INSTALL} -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} ${LINTLIB} ${DESTDIR}${LINTLIBDIR} .endif .endif # !defined(INTERNALLIB) .if !defined(LIBRARIES_ONLY) .include .include .include .endif .include .if ${MK_MAN} != "no" && !defined(LIBRARIES_ONLY) realinstall: _maninstall .ORDER: beforeinstall _maninstall .endif .endif .if !target(lint) lint: ${SRCS:M*.c} ${LINT} ${LINTFLAGS} ${CFLAGS:M-[DIU]*} ${.ALLSRC} .endif .if ${MK_MAN} != "no" && !defined(LIBRARIES_ONLY) .include .endif .include .if !exists(${.OBJDIR}/${DEPENDFILE}) .if defined(LIB) && !empty(LIB) ${OBJS} ${STATICOBJS} ${POBJS}: ${SRCS:M*.h} .for _S in ${SRCS:N*.[hly]} ${_S:R}.po: ${_S} .endfor .endif .if defined(SHLIB_NAME) || \ defined(INSTALL_PIC_ARCHIVE) && defined(LIB) && !empty(LIB) ${SOBJS}: ${SRCS:M*.h} .for _S in ${SRCS:N*.[hly]} ${_S:R}.So: ${_S} .endfor .endif .endif .if !target(clean) clean: .if defined(CLEANFILES) && !empty(CLEANFILES) rm -f ${CLEANFILES} .endif .if defined(LIB) && !empty(LIB) rm -f a.out ${OBJS} ${OBJS:S/$/.tmp/} ${STATICOBJS} .endif .if !defined(INTERNALLIB) .if ${MK_PROFILE} != "no" && defined(LIB) && !empty(LIB) rm -f ${POBJS} ${POBJS:S/$/.tmp/} .endif .if defined(SHLIB_NAME) || \ defined(INSTALL_PIC_ARCHIVE) && defined(LIB) && !empty(LIB) rm -f ${SOBJS} ${SOBJS:.So=.so} ${SOBJS:S/$/.tmp/} .endif .if defined(SHLIB_NAME) .if defined(SHLIB_LINK) .if defined(SHLIB_LDSCRIPT) && exists(${.CURDIR}/${SHLIB_LDSCRIPT}) rm -f lib${LIB}.ld .endif rm -f ${SHLIB_LINK} .endif .endif # defined(SHLIB_NAME) .if defined(WANT_LINT) && defined(LIB) && !empty(LIB) rm -f ${LINTOBJS} .endif .endif # !defined(INTERNALLIB) .if defined(_LIBS) && !empty(_LIBS) rm -f ${_LIBS} .endif .if defined(CLEANDIRS) && !empty(CLEANDIRS) rm -rf ${CLEANDIRS} .endif .if !empty(VERSION_DEF) && !empty(SYMBOL_MAPS) rm -f ${VERSION_MAP} .endif .endif .if !empty(_LIBS) NOPATH_FILES+= ${_LIBS} .endif .include .include Index: projects/em_mq/share =================================================================== --- projects/em_mq/share (revision 283943) +++ projects/em_mq/share (revision 283944) Property changes on: projects/em_mq/share ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/share:r283923-283943 Index: projects/em_mq/sys/boot/forth/support.4th =================================================================== --- projects/em_mq/sys/boot/forth/support.4th (revision 283943) +++ projects/em_mq/sys/boot/forth/support.4th (revision 283944) @@ -1,1583 +1,1583 @@ \ Copyright (c) 1999 Daniel C. Sobral \ All rights reserved. \ \ Redistribution and use in source and binary forms, with or without \ modification, are permitted provided that the following conditions \ are met: \ 1. Redistributions of source code must retain the above copyright \ notice, this list of conditions and the following disclaimer. \ 2. Redistributions in binary form must reproduce the above copyright \ notice, this list of conditions and the following disclaimer in the \ documentation and/or other materials provided with the distribution. \ \ THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND \ ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE \ IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE \ ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE \ FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL \ DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS \ OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) \ HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT \ LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY \ OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF \ SUCH DAMAGE. \ \ $FreeBSD$ \ Loader.rc support functions: \ \ initialize ( addr len -- ) as above, plus load_conf_files \ load_conf ( addr len -- ) load conf file given \ include_conf_files ( -- ) load all conf files in load_conf_files \ print_syntax_error ( -- ) print line and marker of where a syntax \ error was detected \ print_line ( -- ) print last line processed \ load_kernel ( -- ) load kernel \ load_modules ( -- ) load modules flagged \ \ Exported structures: \ \ string counted string structure \ cell .addr string address \ cell .len string length \ module module loading information structure \ cell module.flag should we load it? \ string module.name module's name \ string module.loadname name to be used in loading the module \ string module.type module's type \ string module.args flags to be passed during load \ string module.beforeload command to be executed before load \ string module.afterload command to be executed after load \ string module.loaderror command to be executed if load fails \ cell module.next list chain \ \ Exported global variables; \ \ string conf_files configuration files to be loaded \ cell modules_options pointer to first module information \ value verbose? indicates if user wants a verbose loading \ value any_conf_read? indicates if a conf file was succesfully read \ \ Other exported words: \ note, strlen is internal \ strdup ( addr len -- addr' len) similar to strdup(3) \ strcat ( addr len addr' len' -- addr len+len' ) similar to strcat(3) \ s' ( | string' -- addr len | ) similar to s" \ rudimentary structure support \ Exception values 1 constant ESYNTAX 2 constant ENOMEM 3 constant EFREE 4 constant ESETERROR \ error setting environment variable 5 constant EREAD \ error reading 6 constant EOPEN 7 constant EEXEC \ XXX never catched 8 constant EBEFORELOAD 9 constant EAFTERLOAD \ I/O constants 0 constant SEEK_SET 1 constant SEEK_CUR 2 constant SEEK_END 0 constant O_RDONLY 1 constant O_WRONLY 2 constant O_RDWR \ Crude structure support : structure: create here 0 , ['] drop , 0 does> create here swap dup @ allot cell+ @ execute ; : member: create dup , over , + does> cell+ @ + ; : ;structure swap ! ; : constructor! >body cell+ ! ; : constructor: over :noname ; : ;constructor postpone ; swap cell+ ! ; immediate : sizeof ' >body @ state @ if postpone literal then ; immediate : offsetof ' >body cell+ @ state @ if postpone literal then ; immediate : ptr 1 cells member: ; : int 1 cells member: ; \ String structure structure: string ptr .addr int .len constructor: 0 over .addr ! 0 swap .len ! ;constructor ;structure \ Module options linked list structure: module int module.flag sizeof string member: module.name sizeof string member: module.loadname sizeof string member: module.type sizeof string member: module.args sizeof string member: module.beforeload sizeof string member: module.afterload sizeof string member: module.loaderror ptr module.next ;structure \ Internal loader structures (preloaded_file, kernel_module, file_metadata) \ must be in sync with the C struct in sys/boot/common/bootstrap.h structure: preloaded_file ptr pf.name ptr pf.type ptr pf.args ptr pf.metadata \ file_metadata int pf.loader int pf.addr int pf.size ptr pf.modules \ kernel_module ptr pf.next \ preloaded_file ;structure structure: kernel_module ptr km.name \ ptr km.args ptr km.fp \ preloaded_file ptr km.next \ kernel_module ;structure structure: file_metadata int md.size 2 member: md.type \ this is not ANS Forth compatible (XXX) ptr md.next \ file_metadata 0 member: md.data \ variable size ;structure \ end of structures \ Global variables string conf_files string nextboot_conf_file create module_options sizeof module.next allot 0 module_options ! create last_module_option sizeof module.next allot 0 last_module_option ! 0 value verbose? 0 value nextboot? \ Support string functions : strdup { addr len -- addr' len' } len allocate if ENOMEM throw then addr over len move len ; : strcat { addr len addr' len' -- addr len+len' } addr' addr len + len' move addr len len' + ; : strchr { addr len c -- addr' len' } begin len while addr c@ c = if addr len exit then addr 1 + to addr len 1 - to len repeat 0 0 ; : s' \ same as s", allows " in the string [char] ' parse state @ if postpone sliteral then ; immediate : 2>r postpone >r postpone >r ; immediate : 2r> postpone r> postpone r> ; immediate : 2r@ postpone 2r> postpone 2dup postpone 2>r ; immediate : getenv? getenv -1 = if false else drop true then ; \ determine if a word appears in a string, case-insensitive : contains? ( addr1 len1 addr2 len2 -- 0 | -1 ) 2 pick 0= if 2drop 2drop true exit then dup 0= if 2drop 2drop false exit then begin begin swap dup c@ dup 32 = over 9 = or over 10 = or over 13 = or over 44 = or swap drop while 1+ swap 1- repeat swap 2 pick 1- over < while 2over 2over drop over compare-insensitive 0= if 2 pick over = if 2drop 2drop true exit then 2 pick tuck - -rot + swap over c@ dup 32 = over 9 = or over 10 = or over 13 = or over 44 = or swap drop if 2drop 2drop true exit then then begin swap dup c@ dup 32 = over 9 = or over 10 = or over 13 = or over 44 = or swap drop if false else true then 2 pick 0> and while 1+ swap 1- repeat swap repeat 2drop 2drop false ; : boot_serial? ( -- 0 | -1 ) s" console" getenv dup -1 <> if s" comconsole" 2swap contains? else drop false then s" boot_serial" getenv dup -1 <> if swap drop 0> else drop false then or \ console contains comconsole ( or ) boot_serial s" boot_multicons" getenv dup -1 <> if swap drop 0> else drop false then or \ previous boolean ( or ) boot_multicons ; \ Private definitions vocabulary support-functions only forth also support-functions definitions \ Some control characters constants 7 constant bell 8 constant backspace 9 constant tab 10 constant lf 13 constant \ Read buffer size 80 constant read_buffer_size \ Standard suffixes : load_module_suffix s" _load" ; : module_loadname_suffix s" _name" ; : module_type_suffix s" _type" ; : module_args_suffix s" _flags" ; : module_beforeload_suffix s" _before" ; : module_afterload_suffix s" _after" ; : module_loaderror_suffix s" _error" ; \ Support operators : >= < 0= ; : <= > 0= ; \ Assorted support functions : free-memory free if EFREE throw then ; : strget { var -- addr len } var .addr @ var .len @ ; \ assign addr len to variable. : strset { addr len var -- } addr var .addr ! len var .len ! ; \ free memory and reset fields : strfree { var -- } var .addr @ ?dup if free-memory 0 0 var strset then ; \ free old content, make a copy of the string and assign to variable : string= { addr len var -- } var strfree addr len strdup var strset ; : strtype ( str -- ) strget type ; \ assign a reference to what is on the stack : strref { addr len var -- addr len } addr var .addr ! len var .len ! addr len ; \ unquote a string : unquote ( addr len -- addr len ) over c@ [char] " = if 2 chars - swap char+ swap then ; \ Assignment data temporary storage string name_buffer string value_buffer \ Line by line file reading functions \ \ exported: \ line_buffer \ end_of_file? \ fd \ read_line \ reset_line_reading vocabulary line-reading also line-reading definitions \ File data temporary storage string read_buffer 0 value read_buffer_ptr \ File's line reading function get-current ( -- wid ) previous definitions string line_buffer 0 value end_of_file? variable fd >search ( wid -- ) definitions : skip_newlines begin read_buffer .len @ read_buffer_ptr > while read_buffer .addr @ read_buffer_ptr + c@ lf = if read_buffer_ptr char+ to read_buffer_ptr else exit then repeat ; : scan_buffer ( -- addr len ) read_buffer_ptr >r begin read_buffer .len @ r@ > while read_buffer .addr @ r@ + c@ lf = if read_buffer .addr @ read_buffer_ptr + ( -- addr ) r@ read_buffer_ptr - ( -- len ) r> to read_buffer_ptr exit then r> char+ >r repeat read_buffer .addr @ read_buffer_ptr + ( -- addr ) r@ read_buffer_ptr - ( -- len ) r> to read_buffer_ptr ; : line_buffer_resize ( len -- len ) >r line_buffer .len @ if line_buffer .addr @ line_buffer .len @ r@ + resize if ENOMEM throw then else r@ allocate if ENOMEM throw then then line_buffer .addr ! r> ; : append_to_line_buffer ( addr len -- ) line_buffer strget 2swap strcat line_buffer .len ! drop ; : read_from_buffer scan_buffer ( -- addr len ) line_buffer_resize ( len -- len ) append_to_line_buffer ( addr len -- ) ; : refill_required? read_buffer .len @ read_buffer_ptr = end_of_file? 0= and ; : refill_buffer 0 to read_buffer_ptr read_buffer .addr @ 0= if read_buffer_size allocate if ENOMEM throw then read_buffer .addr ! then fd @ read_buffer .addr @ read_buffer_size fread dup -1 = if EREAD throw then dup 0= if true to end_of_file? then read_buffer .len ! ; get-current ( -- wid ) previous definitions >search ( wid -- ) : reset_line_reading 0 to read_buffer_ptr ; : read_line line_buffer strfree skip_newlines begin read_from_buffer refill_required? while refill_buffer repeat ; only forth also support-functions definitions \ Conf file line parser: \ ::= '='[] | \ [] \ ::= {||'_'} \ ::= '"'{|'\'}'"' | \ ::= ASCII 32 to 126, except '\' and '"' \ ::= '#'{} \ \ exported: \ line_pointer \ process_conf 0 value line_pointer vocabulary file-processing also file-processing definitions \ parser functions \ \ exported: \ get_assignment vocabulary parser also parser definitions 0 value parsing_function 0 value end_of_line : end_of_line? line_pointer end_of_line = ; \ classifiers for various character classes in the input line : letter? line_pointer c@ >r r@ [char] A >= r@ [char] Z <= and r@ [char] a >= r> [char] z <= and or ; : digit? line_pointer c@ >r r@ [char] - = r@ [char] 0 >= r> [char] 9 <= and or ; : quote? line_pointer c@ [char] " = ; : assignment_sign? line_pointer c@ [char] = = ; : comment? line_pointer c@ [char] # = ; : space? line_pointer c@ bl = line_pointer c@ tab = or ; : backslash? line_pointer c@ [char] \ = ; : underscore? line_pointer c@ [char] _ = ; : dot? line_pointer c@ [char] . = ; \ manipulation of input line : skip_character line_pointer char+ to line_pointer ; : skip_to_end_of_line end_of_line to line_pointer ; : eat_space begin end_of_line? if 0 else space? then while skip_character repeat ; : parse_name ( -- addr len ) line_pointer begin end_of_line? if 0 else letter? digit? underscore? dot? or or or then while skip_character repeat line_pointer over - strdup ; : remove_backslashes { addr len | addr' len' -- addr' len' } len allocate if ENOMEM throw then to addr' addr >r begin addr c@ [char] \ <> if addr c@ addr' len' + c! len' char+ to len' then addr char+ to addr r@ len + addr = until r> drop addr' len' ; : parse_quote ( -- addr len ) line_pointer skip_character end_of_line? if ESYNTAX throw then begin quote? 0= while backslash? if skip_character end_of_line? if ESYNTAX throw then then skip_character end_of_line? if ESYNTAX throw then repeat skip_character line_pointer over - remove_backslashes ; : read_name parse_name ( -- addr len ) name_buffer strset ; : read_value quote? if parse_quote ( -- addr len ) else parse_name ( -- addr len ) then value_buffer strset ; : comment skip_to_end_of_line ; : white_space_4 eat_space comment? if ['] comment to parsing_function exit then end_of_line? 0= if ESYNTAX throw then ; : variable_value read_value ['] white_space_4 to parsing_function ; : white_space_3 eat_space letter? digit? quote? or or if ['] variable_value to parsing_function exit then ESYNTAX throw ; : assignment_sign skip_character ['] white_space_3 to parsing_function ; : white_space_2 eat_space assignment_sign? if ['] assignment_sign to parsing_function exit then ESYNTAX throw ; : variable_name read_name ['] white_space_2 to parsing_function ; : white_space_1 eat_space letter? if ['] variable_name to parsing_function exit then comment? if ['] comment to parsing_function exit then end_of_line? 0= if ESYNTAX throw then ; get-current ( -- wid ) previous definitions >search ( wid -- ) : get_assignment line_buffer strget + to end_of_line line_buffer .addr @ to line_pointer ['] white_space_1 to parsing_function begin end_of_line? 0= while parsing_function execute repeat parsing_function ['] comment = parsing_function ['] white_space_1 = parsing_function ['] white_space_4 = or or 0= if ESYNTAX throw then ; only forth also support-functions also file-processing definitions \ Process line : assignment_type? ( addr len -- flag ) name_buffer strget compare 0= ; : suffix_type? ( addr len -- flag ) name_buffer .len @ over <= if 2drop false exit then name_buffer .len @ over - name_buffer .addr @ + over compare 0= ; : loader_conf_files? s" loader_conf_files" assignment_type? ; : nextboot_flag? s" nextboot_enable" assignment_type? ; : nextboot_conf? s" nextboot_conf" assignment_type? ; : verbose_flag? s" verbose_loading" assignment_type? ; : execute? s" exec" assignment_type? ; : module_load? load_module_suffix suffix_type? ; : module_loadname? module_loadname_suffix suffix_type? ; : module_type? module_type_suffix suffix_type? ; : module_args? module_args_suffix suffix_type? ; : module_beforeload? module_beforeload_suffix suffix_type? ; : module_afterload? module_afterload_suffix suffix_type? ; : module_loaderror? module_loaderror_suffix suffix_type? ; \ build a 'set' statement and execute it : set_environment_variable name_buffer .len @ value_buffer .len @ + 5 chars + \ size of result string allocate if ENOMEM throw then dup 0 \ start with an empty string and append the pieces s" set " strcat name_buffer strget strcat s" =" strcat value_buffer strget strcat ['] evaluate catch if 2drop free drop ESETERROR throw else free-memory then ; : set_conf_files set_environment_variable s" loader_conf_files" getenv conf_files string= ; : set_nextboot_conf \ XXX maybe do as set_conf_files ? value_buffer strget unquote nextboot_conf_file string= ; : append_to_module_options_list ( addr -- ) module_options @ 0= if dup module_options ! last_module_option ! else dup last_module_option @ module.next ! last_module_option ! then ; : set_module_name { addr -- } \ check leaks name_buffer strget addr module.name string= ; : yes_value? value_buffer strget \ XXX could use unquote 2dup s' "YES"' compare >r 2dup s' "yes"' compare >r 2dup s" YES" compare >r s" yes" compare r> r> r> and and and 0= ; : find_module_option ( -- addr | 0 ) \ return ptr to entry matching name_buffer module_options @ begin dup while dup module.name strget name_buffer strget compare 0= if exit then module.next @ repeat ; : new_module_option ( -- addr ) sizeof module allocate if ENOMEM throw then dup sizeof module erase dup append_to_module_options_list dup set_module_name ; : get_module_option ( -- addr ) find_module_option ?dup 0= if new_module_option then ; : set_module_flag name_buffer .len @ load_module_suffix nip - name_buffer .len ! yes_value? get_module_option module.flag ! ; : set_module_args name_buffer .len @ module_args_suffix nip - name_buffer .len ! value_buffer strget unquote get_module_option module.args string= ; : set_module_loadname name_buffer .len @ module_loadname_suffix nip - name_buffer .len ! value_buffer strget unquote get_module_option module.loadname string= ; : set_module_type name_buffer .len @ module_type_suffix nip - name_buffer .len ! value_buffer strget unquote get_module_option module.type string= ; : set_module_beforeload name_buffer .len @ module_beforeload_suffix nip - name_buffer .len ! value_buffer strget unquote get_module_option module.beforeload string= ; : set_module_afterload name_buffer .len @ module_afterload_suffix nip - name_buffer .len ! value_buffer strget unquote get_module_option module.afterload string= ; : set_module_loaderror name_buffer .len @ module_loaderror_suffix nip - name_buffer .len ! value_buffer strget unquote get_module_option module.loaderror string= ; : set_nextboot_flag yes_value? to nextboot? ; : set_verbose yes_value? to verbose? ; : execute_command value_buffer strget unquote ['] evaluate catch if EEXEC throw then ; : process_assignment name_buffer .len @ 0= if exit then loader_conf_files? if set_conf_files exit then nextboot_flag? if set_nextboot_flag exit then nextboot_conf? if set_nextboot_conf exit then verbose_flag? if set_verbose exit then execute? if execute_command exit then module_load? if set_module_flag exit then module_loadname? if set_module_loadname exit then module_type? if set_module_type exit then module_args? if set_module_args exit then module_beforeload? if set_module_beforeload exit then module_afterload? if set_module_afterload exit then module_loaderror? if set_module_loaderror exit then set_environment_variable ; \ free_buffer ( -- ) \ \ Free some pointers if needed. The code then tests for errors \ in freeing, and throws an exception if needed. If a pointer is \ not allocated, it's value (0) is used as flag. : free_buffers name_buffer strfree value_buffer strfree ; \ Higher level file processing get-current ( -- wid ) previous definitions >search ( wid -- ) : process_conf begin end_of_file? 0= while free_buffers read_line get_assignment ['] process_assignment catch ['] free_buffers catch swap throw throw repeat ; : peek_file 0 to end_of_file? reset_line_reading O_RDONLY fopen fd ! fd @ -1 = if EOPEN throw then free_buffers read_line get_assignment ['] process_assignment catch ['] free_buffers catch fd @ fclose ; only forth also support-functions definitions \ Interface to loading conf files : load_conf ( addr len -- ) \ ." ----- Trying conf " 2dup type cr \ debugging 0 to end_of_file? reset_line_reading O_RDONLY fopen fd ! fd @ -1 = if EOPEN throw then ['] process_conf catch fd @ fclose throw ; : print_line line_buffer strtype cr ; : print_syntax_error line_buffer strtype cr line_buffer .addr @ begin line_pointer over <> while bl emit char+ repeat drop ." ^" cr ; \ Debugging support functions only forth definitions also support-functions : test-file ['] load_conf catch dup . ESYNTAX = if cr print_syntax_error then ; \ find a module name, leave addr on the stack (0 if not found) : find-module ( -- ptr | 0 ) bl parse ( addr len ) module_options @ >r ( store current pointer ) begin r@ while 2dup ( addr len addr len ) r@ module.name strget compare 0= if drop drop r> exit then ( found it ) r> module.next @ >r repeat type ." was not found" cr r> ; : show-nonempty ( addr len mod -- ) strget dup verbose? or if 2swap type type cr else drop drop drop drop then ; : show-one-module { addr -- addr } ." Name: " addr module.name strtype cr s" Path: " addr module.loadname show-nonempty s" Type: " addr module.type show-nonempty s" Flags: " addr module.args show-nonempty s" Before load: " addr module.beforeload show-nonempty s" After load: " addr module.afterload show-nonempty s" Error: " addr module.loaderror show-nonempty ." Status: " addr module.flag @ if ." Load" else ." Don't load" then cr cr addr ; : show-module-options module_options @ begin ?dup while show-one-module module.next @ repeat ; only forth also support-functions definitions \ Variables used for processing multiple conf files string current_file_name_ref \ used to print the file name \ Indicates if any conf file was succesfully read 0 value any_conf_read? \ loader_conf_files processing support functions : get_conf_files ( -- addr len ) \ put addr/len on stack, reset var \ ." -- starting on <" conf_files strtype ." >" cr \ debugging conf_files strget 0 0 conf_files strset ; : skip_leading_spaces { addr len pos -- addr len pos' } begin pos len = if 0 else addr pos + c@ bl = then while pos char+ to pos repeat addr len pos ; \ return the file name at pos, or free the string if nothing left : get_file_name { addr len pos -- addr len pos' addr' len' || 0 } pos len = if addr free abort" Fatal error freeing memory" 0 exit then pos >r begin \ stay in the loop until have chars and they are not blank pos len = if 0 else addr pos + c@ bl <> then while pos char+ to pos repeat addr len pos addr r@ + pos r> - \ 2dup ." get_file_name has " type cr \ debugging ; : get_next_file ( addr len ptr -- addr len ptr' addr' len' | 0 ) skip_leading_spaces get_file_name ; : print_current_file current_file_name_ref strtype ; : process_conf_errors dup 0= if true to any_conf_read? drop exit then >r 2drop r> dup ESYNTAX = if ." Warning: syntax error on file " print_current_file cr print_syntax_error drop exit then dup ESETERROR = if ." Warning: bad definition on file " print_current_file cr print_line drop exit then dup EREAD = if ." Warning: error reading file " print_current_file cr drop exit then dup EOPEN = if verbose? if ." Warning: unable to open file " print_current_file cr then drop exit then dup EFREE = abort" Fatal error freeing memory" dup ENOMEM = abort" Out of memory" throw \ Unknown error -- pass ahead ; \ Process loader_conf_files recursively \ Interface to loader_conf_files processing : include_conf_files get_conf_files 0 ( addr len offset ) begin get_next_file ?dup ( addr len 1 | 0 ) while current_file_name_ref strref ['] load_conf catch process_conf_errors conf_files .addr @ if recurse then repeat ; : get_nextboot_conf_file ( -- addr len ) nextboot_conf_file strget strdup \ XXX is the strdup a leak ? ; : rewrite_nextboot_file ( -- ) get_nextboot_conf_file O_WRONLY fopen fd ! fd @ -1 = if EOPEN throw then fd @ s' nextboot_enable="NO" ' fwrite fd @ fclose ; : include_nextboot_file get_nextboot_conf_file ['] peek_file catch nextboot? if get_nextboot_conf_file ['] load_conf catch process_conf_errors ['] rewrite_nextboot_file catch then ; \ Module loading functions : load_parameters { addr -- addr addrN lenN ... addr1 len1 N } addr addr module.args strget addr module.loadname .len @ if addr module.loadname strget else addr module.name strget then addr module.type .len @ if addr module.type strget s" -t " 4 ( -t type name flags ) else 2 ( name flags ) then ; : before_load ( addr -- addr ) dup module.beforeload .len @ if dup module.beforeload strget ['] evaluate catch if EBEFORELOAD throw then then ; : after_load ( addr -- addr ) dup module.afterload .len @ if dup module.afterload strget ['] evaluate catch if EAFTERLOAD throw then then ; : load_error ( addr -- addr ) dup module.loaderror .len @ if dup module.loaderror strget evaluate \ This we do not intercept so it can throw errors then ; : pre_load_message ( addr -- addr ) verbose? if dup module.name strtype ." ..." then ; : load_error_message verbose? if ." failed!" cr then ; : load_succesful_message verbose? if ." ok" cr then ; : load_module load_parameters load ; : process_module ( addr -- addr ) pre_load_message before_load begin ['] load_module catch if dup module.loaderror .len @ if load_error \ Command should return a flag! else load_error_message true \ Do not retry then else after_load load_succesful_message true \ Succesful, do not retry then until ; : process_module_errors ( addr ior -- ) dup EBEFORELOAD = if drop ." Module " dup module.name strtype dup module.loadname .len @ if ." (" dup module.loadname strtype ." )" then cr ." Error executing " dup module.beforeload strtype cr \ XXX there was a typo here abort then dup EAFTERLOAD = if drop ." Module " dup module.name .addr @ over module.name .len @ type dup module.loadname .len @ if ." (" dup module.loadname strtype ." )" then cr ." Error executing " dup module.afterload strtype cr abort then throw \ Don't know what it is all about -- pass ahead ; \ Module loading interface \ scan the list of modules, load enabled ones. : load_modules ( -- ) ( throws: abort & user-defined ) module_options @ ( list_head ) begin ?dup while dup module.flag @ if ['] process_module catch process_module_errors then module.next @ repeat ; \ h00h00 magic used to try loading either a kernel with a given name, \ or a kernel with the default name in a directory of a given name \ (the pain!) : bootpath s" /boot/" ; : modulepath s" module_path" ; \ Functions used to save and restore module_path's value. : saveenv ( addr len | -1 -- addr' len | 0 -1 ) dup -1 = if 0 swap exit then strdup ; : freeenv ( addr len | 0 -1 ) -1 = if drop else free abort" Freeing error" then ; : restoreenv ( addr len | 0 -1 -- ) dup -1 = if ( it wasn't set ) 2drop modulepath unsetenv else over >r modulepath setenv r> free abort" Freeing error" then ; : clip_args \ Drop second string if only one argument is passed 1 = if 2swap 2drop 1 else 2 then ; also builtins \ Parse filename from a semicolon-separated list \ replacement, not working yet : newparse-; { addr len | a1 -- a' len-x addr x } addr len [char] ; strchr dup if ( a1 len1 ) swap to a1 ( store address ) 1 - a1 @ 1 + swap ( remove match ) addr a1 addr - else 0 0 addr len then ; : parse-; ( addr len -- addr' len-x addr x ) over 0 2swap ( addr 0 addr len ) begin dup 0 <> ( addr 0 addr len ) while over c@ [char] ; <> ( addr 0 addr len flag ) while 1- swap 1+ swap 2swap 1+ 2swap repeat then dup 0 <> if 1- swap 1+ swap then 2swap ; \ Try loading one of multiple kernels specified : try_multiple_kernels ( addr len addr' len' args -- flag ) >r begin parse-; 2>r 2over 2r> r@ clip_args s" DEBUG" getenv? if s" echo Module_path: ${module_path}" evaluate ." Kernel : " >r 2dup type r> cr dup 2 = if ." Flags : " >r 2over type r> cr then then 1 load while dup 0= until 1 >r \ Failure else 0 >r \ Success then 2drop 2drop r> r> drop ; \ Try to load a kernel; the kernel name is taken from one of \ the following lists, as ordered: \ \ 1. The "bootfile" environment variable \ 2. The "kernel" environment variable \ \ Flags are passed, if available. If not, dummy values must be given. \ \ The kernel gets loaded from the current module_path. : load_a_kernel ( flags len 1 | x x 0 -- flag ) local args 2local flags 0 0 2local kernel end-locals \ Check if a default kernel name exists at all, exits if not s" bootfile" getenv dup -1 <> if to kernel flags kernel args 1+ try_multiple_kernels dup 0= if exit then then drop s" kernel" getenv dup -1 <> if to kernel else drop 1 exit \ Failure then \ Try all default kernel names flags kernel args 1+ try_multiple_kernels ; \ Try to load a kernel; the kernel name is taken from one of \ the following lists, as ordered: \ \ 1. The "bootfile" environment variable \ 2. The "kernel" environment variable \ \ Flags are passed, if provided. \ \ The kernel will be loaded from a directory computed from the \ path given. Two directories will be tried in the following order: \ \ 1. /boot/path \ 2. path \ \ The module_path variable is overridden if load is succesful, by \ prepending the successful path. : load_from_directory ( path len 1 | flags len' path len 2 -- flag ) local args 2local path args 1 = if 0 0 then 2local flags 0 0 2local oldmodulepath \ like a string 0 0 2local newmodulepath \ like a string end-locals \ Set the environment variable module_path, and try loading \ the kernel again. modulepath getenv saveenv to oldmodulepath \ Try prepending /boot/ first bootpath nip path nip + \ total length oldmodulepath nip dup -1 = if drop else 1+ + \ add oldpath -- XXX why the 1+ ? then allocate if ( out of memory ) 1 exit then \ XXX throw ? 0 bootpath strcat path strcat 2dup to newmodulepath modulepath setenv \ Try all default kernel names flags args 1- load_a_kernel 0= if ( success ) oldmodulepath nip -1 <> if newmodulepath s" ;" strcat oldmodulepath strcat modulepath setenv newmodulepath drop free-memory oldmodulepath drop free-memory then 0 exit then \ Well, try without the prepended /boot/ path newmodulepath drop swap move newmodulepath drop path nip 2dup to newmodulepath modulepath setenv \ Try all default kernel names flags args 1- load_a_kernel if ( failed once more ) oldmodulepath restoreenv newmodulepath drop free-memory 1 else oldmodulepath nip -1 <> if newmodulepath s" ;" strcat oldmodulepath strcat modulepath setenv newmodulepath drop free-memory oldmodulepath drop free-memory then 0 then ; \ Try to load a kernel; the kernel name is taken from one of \ the following lists, as ordered: \ \ 1. The "bootfile" environment variable \ 2. The "kernel" environment variable \ 3. The "path" argument \ \ Flags are passed, if provided. \ \ The kernel will be loaded from a directory computed from the \ path given. Two directories will be tried in the following order: \ \ 1. /boot/path \ 2. path \ \ Unless "path" is meant to be kernel name itself. In that case, it \ will first be tried as a full path, and, next, search on the \ directories pointed by module_path. \ \ The module_path variable is overridden if load is succesful, by \ prepending the successful path. : load_directory_or_file ( path len 1 | flags len' path len 2 -- flag ) local args 2local path args 1 = if 0 0 then 2local flags end-locals \ First, assume path is an absolute path to a directory flags path args clip_args load_from_directory dup 0= if exit else drop then \ Next, assume path points to the kernel flags path args try_multiple_kernels ; : initialize ( addr len -- ) strdup conf_files strset ; : kernel_options ( -- addr len 1 | 0 ) s" kernel_options" getenv dup -1 = if drop 0 else 1 then ; : standard_kernel_search ( flags 1 | 0 -- flag ) local args args 0= if 0 0 then 2local flags s" kernel" getenv dup -1 = if 0 swap then 2local path end-locals path nip -1 = if ( there isn't a "kernel" environment variable ) flags args load_a_kernel else flags path args 1+ clip_args load_directory_or_file then ; : load_kernel ( -- ) ( throws: abort ) kernel_options standard_kernel_search abort" Unable to load a kernel!" ; -: load_xen ( -- ) +: load_xen ( -- flag ) s" xen_kernel" getenv dup -1 <> if - 1 1 load + 1 1 load ( c-addr/u flag N -- flag ) else drop - 0 + 0 ( -1 -- flag ) then ; : load_xen_throw ( -- ) ( throws: abort ) load_xen abort" Unable to load Xen!" ; : set_defaultoptions ( -- ) s" kernel_options" getenv dup -1 = if drop else s" temp_options" setenv then ; \ pick the i-th argument, i starts at 0 : argv[] ( aN uN ... a1 u1 N i -- aN uN ... a1 u1 N ai+1 ui+1 ) 2dup = if 0 0 exit then \ out of range dup >r 1+ 2* ( skip N and ui ) pick r> 1+ 2* ( skip N and ai ) pick ; : drop_args ( aN uN ... a1 u1 N -- ) 0 ?do 2drop loop ; : argc dup ; : queue_argv ( aN uN ... a1 u1 N a u -- a u aN uN ... a1 u1 N+1 ) >r over 2* 1+ -roll r> over 2* 1+ -roll 1+ ; : unqueue_argv ( aN uN ... a1 u1 N -- aN uN ... a2 u2 N-1 a1 u1 ) 1- -rot ; \ compute the length of the buffer including the spaces between words : strlen(argv) ( aN uN .. a1 u1 N -- aN uN .. a1 u1 N len ) dup 0= if 0 exit then 0 >r \ Size 0 >r \ Index begin argc r@ <> while r@ argv[] nip r> r> rot + 1+ >r 1+ >r repeat r> drop r> ; : concat_argv ( aN uN ... a1 u1 N -- a u ) strlen(argv) allocate if ENOMEM throw then 0 2>r ( save addr 0 on return stack ) begin dup while unqueue_argv ( ... N a1 u1 ) 2r> 2swap ( old a1 u1 ) strcat s" " strcat ( append one space ) \ XXX this gives a trailing space 2>r ( store string on the result stack ) repeat drop_args 2r> ; : set_tempoptions ( addrN lenN ... addr1 len1 N -- addr len 1 | 0 ) \ Save the first argument, if it exists and is not a flag argc if 0 argv[] drop c@ [char] - <> if unqueue_argv 2>r \ Filename 1 >r \ Filename present else 0 >r \ Filename not present then else 0 >r \ Filename not present then \ If there are other arguments, assume they are flags ?dup if concat_argv 2dup s" temp_options" setenv drop free if EFREE throw then else set_defaultoptions then \ Bring back the filename, if one was provided r> if 2r> 1 else 0 then ; : get_arguments ( -- addrN lenN ... addr1 len1 N ) 0 begin \ Get next word on the command line parse-word ?dup while queue_argv repeat drop ( empty string ) ; : load_kernel_and_modules ( args -- flag ) set_tempoptions argc >r s" temp_options" getenv dup -1 <> if queue_argv else drop then load_xen ?dup 0= if ( success ) r> if ( a path was passed ) load_directory_or_file else standard_kernel_search then ?dup 0= if ['] load_modules catch then then ; only forth definitions Index: projects/em_mq/sys/boot/userboot/userboot/conf.c =================================================================== --- projects/em_mq/sys/boot/userboot/userboot/conf.c (revision 283943) +++ projects/em_mq/sys/boot/userboot/userboot/conf.c (revision 283944) @@ -1,106 +1,107 @@ /*- * Copyright (c) 1997 * Matthias Drochner. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed for the NetBSD Project * by Matthias Drochner. * 4. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. * * $NetBSD: conf.c,v 1.2 1997/03/22 09:03:29 thorpej Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include "libuserboot.h" #if defined(USERBOOT_ZFS_SUPPORT) #include "../zfs/libzfs.h" #endif /* * We could use linker sets for some or all of these, but * then we would have to control what ended up linked into * the bootstrap. So it's easier to conditionalise things * here. * * XXX rename these arrays to be consistent and less namespace-hostile */ /* Exported for libstand */ struct devsw *devsw[] = { &host_dev, &userboot_disk, #if defined(USERBOOT_ZFS_SUPPORT) &zfs_dev, #endif NULL }; struct fs_ops *file_system[] = { &host_fsops, &ufs_fsops, &cd9660_fsops, - &gzipfs_fsops, #if defined(USERBOOT_ZFS_SUPPORT) &zfs_fsops, #endif + &gzipfs_fsops, + &bzipfs_fsops, NULL }; /* Exported for i386 only */ /* * Sort formats so that those that can detect based on arguments * rather than reading the file go first. */ extern struct file_format i386_elf; extern struct file_format i386_elf_obj; extern struct file_format amd64_elf; extern struct file_format amd64_elf_obj; struct file_format *file_formats[] = { &i386_elf, &i386_elf_obj, &amd64_elf, &amd64_elf_obj, NULL }; /* * Consoles * * We don't prototype these in libuserboot.h because they require * data structures from bootstrap.h as well. */ extern struct console userboot_console; extern struct console userboot_comconsole; struct console *consoles[] = { &userboot_console, &userboot_comconsole, NULL }; Index: projects/em_mq/sys/boot =================================================================== --- projects/em_mq/sys/boot (revision 283943) +++ projects/em_mq/sys/boot (revision 283944) Property changes on: projects/em_mq/sys/boot ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/sys/boot:r283923-283943 Index: projects/em_mq/sys/compat/linprocfs/linprocfs.c =================================================================== --- projects/em_mq/sys/compat/linprocfs/linprocfs.c (revision 283943) +++ projects/em_mq/sys/compat/linprocfs/linprocfs.c (revision 283944) @@ -1,1545 +1,1541 @@ /*- * Copyright (c) 2000 Dag-Erling Coïdan Smørgrav * Copyright (c) 1999 Pierre Beyssac * Copyright (c) 1993 Jan-Simon Pendry * Copyright (c) 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Jan-Simon Pendry. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)procfs_status.c 8.4 (Berkeley) 6/15/94 */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if defined(__i386__) || defined(__amd64__) #include #include #endif /* __i386__ || __amd64__ */ #include #include #include #include #include #include /* * Various conversion macros */ #define T2J(x) ((long)(((x) * 100ULL) / (stathz ? stathz : hz))) /* ticks to jiffies */ #define T2CS(x) ((unsigned long)(((x) * 100ULL) / (stathz ? stathz : hz))) /* ticks to centiseconds */ #define T2S(x) ((x) / (stathz ? stathz : hz)) /* ticks to seconds */ #define B2K(x) ((x) >> 10) /* bytes to kbytes */ #define B2P(x) ((x) >> PAGE_SHIFT) /* bytes to pages */ #define P2B(x) ((x) << PAGE_SHIFT) /* pages to bytes */ #define P2K(x) ((x) << (PAGE_SHIFT - 10)) /* pages to kbytes */ #define TV2J(x) ((x)->tv_sec * 100UL + (x)->tv_usec / 10000) /** * @brief Mapping of ki_stat in struct kinfo_proc to the linux state * * The linux procfs state field displays one of the characters RSDZTW to * denote running, sleeping in an interruptible wait, waiting in an * uninterruptible disk sleep, a zombie process, process is being traced * or stopped, or process is paging respectively. * * Our struct kinfo_proc contains the variable ki_stat which contains a * value out of SIDL, SRUN, SSLEEP, SSTOP, SZOMB, SWAIT and SLOCK. * * This character array is used with ki_stati-1 as an index and tries to * map our states to suitable linux states. */ static char linux_state[] = "RRSTZDD"; /* * Filler function for proc/meminfo */ static int linprocfs_domeminfo(PFS_FILL_ARGS) { unsigned long memtotal; /* total memory in bytes */ unsigned long memused; /* used memory in bytes */ unsigned long memfree; /* free memory in bytes */ unsigned long memshared; /* shared memory ??? */ unsigned long buffers, cached; /* buffer / cache memory ??? */ unsigned long long swaptotal; /* total swap space in bytes */ unsigned long long swapused; /* used swap space in bytes */ unsigned long long swapfree; /* free swap space in bytes */ vm_object_t object; int i, j; memtotal = physmem * PAGE_SIZE; /* * The correct thing here would be: * memfree = vm_cnt.v_free_count * PAGE_SIZE; memused = memtotal - memfree; * * but it might mislead linux binaries into thinking there * is very little memory left, so we cheat and tell them that * all memory that isn't wired down is free. */ memused = vm_cnt.v_wire_count * PAGE_SIZE; memfree = memtotal - memused; swap_pager_status(&i, &j); swaptotal = (unsigned long long)i * PAGE_SIZE; swapused = (unsigned long long)j * PAGE_SIZE; swapfree = swaptotal - swapused; memshared = 0; mtx_lock(&vm_object_list_mtx); TAILQ_FOREACH(object, &vm_object_list, object_list) if (object->shadow_count > 1) memshared += object->resident_page_count; mtx_unlock(&vm_object_list_mtx); memshared *= PAGE_SIZE; /* * We'd love to be able to write: * buffers = bufspace; * * but bufspace is internal to vfs_bio.c and we don't feel * like unstaticizing it just for linprocfs's sake. */ buffers = 0; cached = vm_cnt.v_cache_count * PAGE_SIZE; sbuf_printf(sb, " total: used: free: shared: buffers: cached:\n" "Mem: %lu %lu %lu %lu %lu %lu\n" "Swap: %llu %llu %llu\n" "MemTotal: %9lu kB\n" "MemFree: %9lu kB\n" "MemShared:%9lu kB\n" "Buffers: %9lu kB\n" "Cached: %9lu kB\n" "SwapTotal:%9llu kB\n" "SwapFree: %9llu kB\n", memtotal, memused, memfree, memshared, buffers, cached, swaptotal, swapused, swapfree, B2K(memtotal), B2K(memfree), B2K(memshared), B2K(buffers), B2K(cached), B2K(swaptotal), B2K(swapfree)); return (0); } #if defined(__i386__) || defined(__amd64__) /* * Filler function for proc/cpuinfo (i386 & amd64 version) */ static int linprocfs_docpuinfo(PFS_FILL_ARGS) { int hw_model[2]; char model[128]; uint64_t freq; size_t size; int class, fqmhz, fqkhz; int i; /* * We default the flags to include all non-conflicting flags, * and the Intel versions of conflicting flags. */ static char *flags[] = { "fpu", "vme", "de", "pse", "tsc", "msr", "pae", "mce", "cx8", "apic", "sep", "sep", "mtrr", "pge", "mca", "cmov", "pat", "pse36", "pn", "b19", "b20", "b21", "mmxext", "mmx", "fxsr", "xmm", "sse2", "b27", "b28", "b29", "3dnowext", "3dnow" }; switch (cpu_class) { #ifdef __i386__ case CPUCLASS_286: class = 2; break; case CPUCLASS_386: class = 3; break; case CPUCLASS_486: class = 4; break; case CPUCLASS_586: class = 5; break; case CPUCLASS_686: class = 6; break; default: class = 0; break; #else /* __amd64__ */ default: class = 15; break; #endif } hw_model[0] = CTL_HW; hw_model[1] = HW_MODEL; model[0] = '\0'; size = sizeof(model); if (kernel_sysctl(td, hw_model, 2, &model, &size, 0, 0, 0, 0) != 0) strcpy(model, "unknown"); for (i = 0; i < mp_ncpus; ++i) { sbuf_printf(sb, "processor\t: %d\n" "vendor_id\t: %.20s\n" "cpu family\t: %u\n" "model\t\t: %u\n" "model name\t: %s\n" "stepping\t: %u\n\n", i, cpu_vendor, CPUID_TO_FAMILY(cpu_id), CPUID_TO_MODEL(cpu_id), model, cpu_id & CPUID_STEPPING); /* XXX per-cpu vendor / class / model / id? */ } sbuf_cat(sb, "flags\t\t:"); #ifdef __i386__ switch (cpu_vendor_id) { case CPU_VENDOR_AMD: if (class < 6) flags[16] = "fcmov"; break; case CPU_VENDOR_CYRIX: flags[24] = "cxmmx"; break; } #endif for (i = 0; i < 32; i++) if (cpu_feature & (1 << i)) sbuf_printf(sb, " %s", flags[i]); sbuf_cat(sb, "\n"); freq = atomic_load_acq_64(&tsc_freq); if (freq != 0) { fqmhz = (freq + 4999) / 1000000; fqkhz = ((freq + 4999) / 10000) % 100; sbuf_printf(sb, "cpu MHz\t\t: %d.%02d\n" "bogomips\t: %d.%02d\n", fqmhz, fqkhz, fqmhz, fqkhz); } return (0); } #endif /* __i386__ || __amd64__ */ /* * Filler function for proc/mtab * * This file doesn't exist in Linux' procfs, but is included here so * users can symlink /compat/linux/etc/mtab to /proc/mtab */ static int linprocfs_domtab(PFS_FILL_ARGS) { struct nameidata nd; const char *lep; char *dlep, *flep, *mntto, *mntfrom, *fstype; size_t lep_len; int error; struct statfs *buf, *sp; size_t count; /* resolve symlinks etc. in the emulation tree prefix */ NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, linux_emul_path, td); flep = NULL; error = namei(&nd); lep = linux_emul_path; if (error == 0) { if (vn_fullpath(td, nd.ni_vp, &dlep, &flep) == 0) lep = dlep; vrele(nd.ni_vp); } lep_len = strlen(lep); buf = NULL; error = kern_getfsstat(td, &buf, SIZE_T_MAX, &count, UIO_SYSSPACE, MNT_WAIT); if (error != 0) { free(buf, M_TEMP); free(flep, M_TEMP); return (error); } for (sp = buf; count > 0; sp++, count--) { /* determine device name */ mntfrom = sp->f_mntfromname; /* determine mount point */ mntto = sp->f_mntonname; if (strncmp(mntto, lep, lep_len) == 0 && mntto[lep_len] == '/') mntto += lep_len; /* determine fs type */ fstype = sp->f_fstypename; if (strcmp(fstype, pn->pn_info->pi_name) == 0) mntfrom = fstype = "proc"; else if (strcmp(fstype, "procfs") == 0) continue; if (strcmp(fstype, "linsysfs") == 0) { sbuf_printf(sb, "/sys %s sysfs %s", mntto, sp->f_flags & MNT_RDONLY ? "ro" : "rw"); } else { /* For Linux msdosfs is called vfat */ if (strcmp(fstype, "msdosfs") == 0) fstype = "vfat"; sbuf_printf(sb, "%s %s %s %s", mntfrom, mntto, fstype, sp->f_flags & MNT_RDONLY ? "ro" : "rw"); } #define ADD_OPTION(opt, name) \ if (sp->f_flags & (opt)) sbuf_printf(sb, "," name); ADD_OPTION(MNT_SYNCHRONOUS, "sync"); ADD_OPTION(MNT_NOEXEC, "noexec"); ADD_OPTION(MNT_NOSUID, "nosuid"); ADD_OPTION(MNT_UNION, "union"); ADD_OPTION(MNT_ASYNC, "async"); ADD_OPTION(MNT_SUIDDIR, "suiddir"); ADD_OPTION(MNT_NOSYMFOLLOW, "nosymfollow"); ADD_OPTION(MNT_NOATIME, "noatime"); #undef ADD_OPTION /* a real Linux mtab will also show NFS options */ sbuf_printf(sb, " 0 0\n"); } free(buf, M_TEMP); free(flep, M_TEMP); return (error); } /* * Filler function for proc/partitions */ static int linprocfs_dopartitions(PFS_FILL_ARGS) { struct g_class *cp; struct g_geom *gp; struct g_provider *pp; int major, minor; g_topology_lock(); sbuf_printf(sb, "major minor #blocks name rio rmerge rsect " "ruse wio wmerge wsect wuse running use aveq\n"); LIST_FOREACH(cp, &g_classes, class) { if (strcmp(cp->name, "DISK") == 0 || strcmp(cp->name, "PART") == 0) LIST_FOREACH(gp, &cp->geom, geom) { LIST_FOREACH(pp, &gp->provider, provider) { if (linux_driver_get_major_minor( pp->name, &major, &minor) != 0) { major = 0; minor = 0; } sbuf_printf(sb, "%d %d %lld %s " "%d %d %d %d %d " "%d %d %d %d %d %d\n", major, minor, (long long)pp->mediasize, pp->name, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0); } } } g_topology_unlock(); return (0); } /* * Filler function for proc/stat */ static int linprocfs_dostat(PFS_FILL_ARGS) { struct pcpu *pcpu; long cp_time[CPUSTATES]; long *cp; int i; read_cpu_time(cp_time); sbuf_printf(sb, "cpu %ld %ld %ld %ld\n", T2J(cp_time[CP_USER]), T2J(cp_time[CP_NICE]), T2J(cp_time[CP_SYS] /*+ cp_time[CP_INTR]*/), T2J(cp_time[CP_IDLE])); CPU_FOREACH(i) { pcpu = pcpu_find(i); cp = pcpu->pc_cp_time; sbuf_printf(sb, "cpu%d %ld %ld %ld %ld\n", i, T2J(cp[CP_USER]), T2J(cp[CP_NICE]), T2J(cp[CP_SYS] /*+ cp[CP_INTR]*/), T2J(cp[CP_IDLE])); } sbuf_printf(sb, "disk 0 0 0 0\n" "page %u %u\n" "swap %u %u\n" "intr %u\n" "ctxt %u\n" "btime %lld\n", vm_cnt.v_vnodepgsin, vm_cnt.v_vnodepgsout, vm_cnt.v_swappgsin, vm_cnt.v_swappgsout, vm_cnt.v_intr, vm_cnt.v_swtch, (long long)boottime.tv_sec); return (0); } static int linprocfs_doswaps(PFS_FILL_ARGS) { struct xswdev xsw; uintmax_t total, used; int n; char devname[SPECNAMELEN + 1]; sbuf_printf(sb, "Filename\t\t\t\tType\t\tSize\tUsed\tPriority\n"); mtx_lock(&Giant); for (n = 0; ; n++) { if (swap_dev_info(n, &xsw, devname, sizeof(devname)) != 0) break; total = (uintmax_t)xsw.xsw_nblks * PAGE_SIZE / 1024; used = (uintmax_t)xsw.xsw_used * PAGE_SIZE / 1024; /* * The space and not tab after the device name is on * purpose. Linux does so. */ sbuf_printf(sb, "/dev/%-34s unknown\t\t%jd\t%jd\t-1\n", devname, total, used); } mtx_unlock(&Giant); return (0); } /* * Filler function for proc/uptime */ static int linprocfs_douptime(PFS_FILL_ARGS) { long cp_time[CPUSTATES]; struct timeval tv; getmicrouptime(&tv); read_cpu_time(cp_time); sbuf_printf(sb, "%lld.%02ld %ld.%02lu\n", (long long)tv.tv_sec, tv.tv_usec / 10000, T2S(cp_time[CP_IDLE] / mp_ncpus), T2CS(cp_time[CP_IDLE] / mp_ncpus) % 100); return (0); } /* * Get OS build date */ static void linprocfs_osbuild(struct thread *td, struct sbuf *sb) { #if 0 char osbuild[256]; char *cp1, *cp2; strncpy(osbuild, version, 256); osbuild[255] = '\0'; cp1 = strstr(osbuild, "\n"); cp2 = strstr(osbuild, ":"); if (cp1 && cp2) { *cp1 = *cp2 = '\0'; cp1 = strstr(osbuild, "#"); } else cp1 = NULL; if (cp1) sbuf_printf(sb, "%s%s", cp1, cp2 + 1); else #endif sbuf_cat(sb, "#4 Sun Dec 18 04:30:00 CET 1977"); } /* * Get OS builder */ static void linprocfs_osbuilder(struct thread *td, struct sbuf *sb) { #if 0 char builder[256]; char *cp; cp = strstr(version, "\n "); if (cp) { strncpy(builder, cp + 5, 256); builder[255] = '\0'; cp = strstr(builder, ":"); if (cp) *cp = '\0'; } if (cp) sbuf_cat(sb, builder); else #endif sbuf_cat(sb, "des@freebsd.org"); } /* * Filler function for proc/version */ static int linprocfs_doversion(PFS_FILL_ARGS) { char osname[LINUX_MAX_UTSNAME]; char osrelease[LINUX_MAX_UTSNAME]; linux_get_osname(td, osname); linux_get_osrelease(td, osrelease); sbuf_printf(sb, "%s version %s (", osname, osrelease); linprocfs_osbuilder(td, sb); sbuf_cat(sb, ") (gcc version " __VERSION__ ") "); linprocfs_osbuild(td, sb); sbuf_cat(sb, "\n"); return (0); } /* * Filler function for proc/loadavg */ static int linprocfs_doloadavg(PFS_FILL_ARGS) { sbuf_printf(sb, "%d.%02d %d.%02d %d.%02d %d/%d %d\n", (int)(averunnable.ldavg[0] / averunnable.fscale), (int)(averunnable.ldavg[0] * 100 / averunnable.fscale % 100), (int)(averunnable.ldavg[1] / averunnable.fscale), (int)(averunnable.ldavg[1] * 100 / averunnable.fscale % 100), (int)(averunnable.ldavg[2] / averunnable.fscale), (int)(averunnable.ldavg[2] * 100 / averunnable.fscale % 100), 1, /* number of running tasks */ nprocs, /* number of tasks */ lastpid /* the last pid */ ); return (0); } /* * Filler function for proc/pid/stat */ static int linprocfs_doprocstat(PFS_FILL_ARGS) { struct kinfo_proc kp; char state; static int ratelimit = 0; vm_offset_t startcode, startdata; sx_slock(&proctree_lock); PROC_LOCK(p); fill_kinfo_proc(p, &kp); sx_sunlock(&proctree_lock); if (p->p_vmspace) { startcode = (vm_offset_t)p->p_vmspace->vm_taddr; startdata = (vm_offset_t)p->p_vmspace->vm_daddr; } else { startcode = 0; startdata = 0; }; sbuf_printf(sb, "%d", p->p_pid); #define PS_ADD(name, fmt, arg) sbuf_printf(sb, " " fmt, arg) PS_ADD("comm", "(%s)", p->p_comm); if (kp.ki_stat > sizeof(linux_state)) { state = 'R'; if (ratelimit == 0) { printf("linprocfs: don't know how to handle unknown FreeBSD state %d/%zd, mapping to R\n", kp.ki_stat, sizeof(linux_state)); ++ratelimit; } } else state = linux_state[kp.ki_stat - 1]; PS_ADD("state", "%c", state); PS_ADD("ppid", "%d", p->p_pptr ? p->p_pptr->p_pid : 0); PS_ADD("pgrp", "%d", p->p_pgid); PS_ADD("session", "%d", p->p_session->s_sid); PROC_UNLOCK(p); PS_ADD("tty", "%ju", (uintmax_t)kp.ki_tdev); PS_ADD("tpgid", "%d", kp.ki_tpgid); PS_ADD("flags", "%u", 0); /* XXX */ PS_ADD("minflt", "%lu", kp.ki_rusage.ru_minflt); PS_ADD("cminflt", "%lu", kp.ki_rusage_ch.ru_minflt); PS_ADD("majflt", "%lu", kp.ki_rusage.ru_majflt); PS_ADD("cmajflt", "%lu", kp.ki_rusage_ch.ru_majflt); PS_ADD("utime", "%ld", TV2J(&kp.ki_rusage.ru_utime)); PS_ADD("stime", "%ld", TV2J(&kp.ki_rusage.ru_stime)); PS_ADD("cutime", "%ld", TV2J(&kp.ki_rusage_ch.ru_utime)); PS_ADD("cstime", "%ld", TV2J(&kp.ki_rusage_ch.ru_stime)); PS_ADD("priority", "%d", kp.ki_pri.pri_user); PS_ADD("nice", "%d", kp.ki_nice); /* 19 (nicest) to -19 */ PS_ADD("0", "%d", 0); /* removed field */ PS_ADD("itrealvalue", "%d", 0); /* XXX */ PS_ADD("starttime", "%lu", TV2J(&kp.ki_start) - TV2J(&boottime)); PS_ADD("vsize", "%ju", P2K((uintmax_t)kp.ki_size)); PS_ADD("rss", "%ju", (uintmax_t)kp.ki_rssize); PS_ADD("rlim", "%lu", kp.ki_rusage.ru_maxrss); PS_ADD("startcode", "%ju", (uintmax_t)startcode); PS_ADD("endcode", "%ju", (uintmax_t)startdata); PS_ADD("startstack", "%u", 0); /* XXX */ PS_ADD("kstkesp", "%u", 0); /* XXX */ PS_ADD("kstkeip", "%u", 0); /* XXX */ PS_ADD("signal", "%u", 0); /* XXX */ PS_ADD("blocked", "%u", 0); /* XXX */ PS_ADD("sigignore", "%u", 0); /* XXX */ PS_ADD("sigcatch", "%u", 0); /* XXX */ PS_ADD("wchan", "%u", 0); /* XXX */ PS_ADD("nswap", "%lu", kp.ki_rusage.ru_nswap); PS_ADD("cnswap", "%lu", kp.ki_rusage_ch.ru_nswap); PS_ADD("exitsignal", "%d", 0); /* XXX */ PS_ADD("processor", "%u", kp.ki_lastcpu); PS_ADD("rt_priority", "%u", 0); /* XXX */ /* >= 2.5.19 */ PS_ADD("policy", "%u", kp.ki_pri.pri_class); /* >= 2.5.19 */ #undef PS_ADD sbuf_putc(sb, '\n'); return (0); } /* * Filler function for proc/pid/statm */ static int linprocfs_doprocstatm(PFS_FILL_ARGS) { struct kinfo_proc kp; segsz_t lsize; sx_slock(&proctree_lock); PROC_LOCK(p); fill_kinfo_proc(p, &kp); PROC_UNLOCK(p); sx_sunlock(&proctree_lock); /* * See comments in linprocfs_doprocstatus() regarding the * computation of lsize. */ /* size resident share trs drs lrs dt */ sbuf_printf(sb, "%ju ", B2P((uintmax_t)kp.ki_size)); sbuf_printf(sb, "%ju ", (uintmax_t)kp.ki_rssize); sbuf_printf(sb, "%ju ", (uintmax_t)0); /* XXX */ sbuf_printf(sb, "%ju ", (uintmax_t)kp.ki_tsize); sbuf_printf(sb, "%ju ", (uintmax_t)(kp.ki_dsize + kp.ki_ssize)); lsize = B2P(kp.ki_size) - kp.ki_dsize - kp.ki_ssize - kp.ki_tsize - 1; sbuf_printf(sb, "%ju ", (uintmax_t)lsize); sbuf_printf(sb, "%ju\n", (uintmax_t)0); /* XXX */ return (0); } /* * Filler function for proc/pid/status */ static int linprocfs_doprocstatus(PFS_FILL_ARGS) { struct kinfo_proc kp; char *state; segsz_t lsize; struct thread *td2; struct sigacts *ps; l_sigset_t siglist, sigignore, sigcatch; int i; sx_slock(&proctree_lock); PROC_LOCK(p); td2 = FIRST_THREAD_IN_PROC(p); /* XXXKSE pretend only one thread */ if (P_SHOULDSTOP(p)) { state = "T (stopped)"; } else { switch(p->p_state) { case PRS_NEW: state = "I (idle)"; break; case PRS_NORMAL: if (p->p_flag & P_WEXIT) { state = "X (exiting)"; break; } switch(td2->td_state) { case TDS_INHIBITED: state = "S (sleeping)"; break; case TDS_RUNQ: case TDS_RUNNING: state = "R (running)"; break; default: state = "? (unknown)"; break; } break; case PRS_ZOMBIE: state = "Z (zombie)"; break; default: state = "? (unknown)"; break; } } fill_kinfo_proc(p, &kp); sx_sunlock(&proctree_lock); sbuf_printf(sb, "Name:\t%s\n", p->p_comm); /* XXX escape */ sbuf_printf(sb, "State:\t%s\n", state); /* * Credentials */ sbuf_printf(sb, "Pid:\t%d\n", p->p_pid); sbuf_printf(sb, "PPid:\t%d\n", p->p_pptr ? p->p_pptr->p_pid : 0); sbuf_printf(sb, "Uid:\t%d %d %d %d\n", p->p_ucred->cr_ruid, p->p_ucred->cr_uid, p->p_ucred->cr_svuid, /* FreeBSD doesn't have fsuid */ p->p_ucred->cr_uid); sbuf_printf(sb, "Gid:\t%d %d %d %d\n", p->p_ucred->cr_rgid, p->p_ucred->cr_gid, p->p_ucred->cr_svgid, /* FreeBSD doesn't have fsgid */ p->p_ucred->cr_gid); sbuf_cat(sb, "Groups:\t"); for (i = 0; i < p->p_ucred->cr_ngroups; i++) sbuf_printf(sb, "%d ", p->p_ucred->cr_groups[i]); PROC_UNLOCK(p); sbuf_putc(sb, '\n'); /* * Memory * * While our approximation of VmLib may not be accurate (I * don't know of a simple way to verify it, and I'm not sure * it has much meaning anyway), I believe it's good enough. * * The same code that could (I think) accurately compute VmLib * could also compute VmLck, but I don't really care enough to * implement it. Submissions are welcome. */ sbuf_printf(sb, "VmSize:\t%8ju kB\n", B2K((uintmax_t)kp.ki_size)); sbuf_printf(sb, "VmLck:\t%8u kB\n", P2K(0)); /* XXX */ sbuf_printf(sb, "VmRSS:\t%8ju kB\n", P2K((uintmax_t)kp.ki_rssize)); sbuf_printf(sb, "VmData:\t%8ju kB\n", P2K((uintmax_t)kp.ki_dsize)); sbuf_printf(sb, "VmStk:\t%8ju kB\n", P2K((uintmax_t)kp.ki_ssize)); sbuf_printf(sb, "VmExe:\t%8ju kB\n", P2K((uintmax_t)kp.ki_tsize)); lsize = B2P(kp.ki_size) - kp.ki_dsize - kp.ki_ssize - kp.ki_tsize - 1; sbuf_printf(sb, "VmLib:\t%8ju kB\n", P2K((uintmax_t)lsize)); /* * Signal masks */ PROC_LOCK(p); bsd_to_linux_sigset(&p->p_siglist, &siglist); ps = p->p_sigacts; mtx_lock(&ps->ps_mtx); bsd_to_linux_sigset(&ps->ps_sigignore, &sigignore); bsd_to_linux_sigset(&ps->ps_sigcatch, &sigcatch); mtx_unlock(&ps->ps_mtx); PROC_UNLOCK(p); sbuf_printf(sb, "SigPnd:\t%016jx\n", siglist.__mask); /* * XXX. SigBlk - target thread's signal mask, td_sigmask. * To implement SigBlk pseudofs should support proc/tid dir entries. */ sbuf_printf(sb, "SigBlk:\t%016x\n", 0); sbuf_printf(sb, "SigIgn:\t%016jx\n", sigignore.__mask); sbuf_printf(sb, "SigCgt:\t%016jx\n", sigcatch.__mask); /* * Linux also prints the capability masks, but we don't have * capabilities yet, and when we do get them they're likely to * be meaningless to Linux programs, so we lie. XXX */ sbuf_printf(sb, "CapInh:\t%016x\n", 0); sbuf_printf(sb, "CapPrm:\t%016x\n", 0); sbuf_printf(sb, "CapEff:\t%016x\n", 0); return (0); } /* * Filler function for proc/pid/cwd */ static int linprocfs_doproccwd(PFS_FILL_ARGS) { char *fullpath = "unknown"; char *freepath = NULL; vn_fullpath(td, p->p_fd->fd_cdir, &fullpath, &freepath); sbuf_printf(sb, "%s", fullpath); if (freepath) free(freepath, M_TEMP); return (0); } /* * Filler function for proc/pid/root */ static int linprocfs_doprocroot(PFS_FILL_ARGS) { struct vnode *rvp; char *fullpath = "unknown"; char *freepath = NULL; rvp = jailed(p->p_ucred) ? p->p_fd->fd_jdir : p->p_fd->fd_rdir; vn_fullpath(td, rvp, &fullpath, &freepath); sbuf_printf(sb, "%s", fullpath); if (freepath) free(freepath, M_TEMP); return (0); } /* * Filler function for proc/pid/cmdline */ static int linprocfs_doproccmdline(PFS_FILL_ARGS) { int ret; PROC_LOCK(p); if ((ret = p_cansee(td, p)) != 0) { PROC_UNLOCK(p); return (ret); } /* * Mimic linux behavior and pass only processes with usermode * address space as valid. Return zero silently otherwize. */ if (p->p_vmspace == &vmspace0) { PROC_UNLOCK(p); return (0); } if (p->p_args != NULL) { sbuf_bcpy(sb, p->p_args->ar_args, p->p_args->ar_length); PROC_UNLOCK(p); return (0); } if ((p->p_flag & P_SYSTEM) != 0) { PROC_UNLOCK(p); return (0); } PROC_UNLOCK(p); ret = proc_getargv(td, p, sb); return (ret); } /* * Filler function for proc/pid/environ */ static int linprocfs_doprocenviron(PFS_FILL_ARGS) { /* * Mimic linux behavior and pass only processes with usermode * address space as valid. Return zero silently otherwize. */ if (p->p_vmspace == &vmspace0) return (0); return (proc_getenvv(td, p, sb)); } static char l32_map_str[] = "%08lx-%08lx %s%s%s%s %08lx %02x:%02x %lu%s%s\n"; static char l64_map_str[] = "%016lx-%016lx %s%s%s%s %08lx %02x:%02x %lu%s%s\n"; static char vdso_str[] = " [vdso]"; static char stack_str[] = " [stack]"; /* * Filler function for proc/pid/maps */ static int linprocfs_doprocmaps(PFS_FILL_ARGS) { struct vmspace *vm; vm_map_t map; vm_map_entry_t entry, tmp_entry; vm_object_t obj, tobj, lobj; vm_offset_t e_start, e_end; vm_ooffset_t off = 0; vm_prot_t e_prot; unsigned int last_timestamp; char *name = "", *freename = NULL; const char *l_map_str; ino_t ino; int ref_count, shadow_count, flags; int error; struct vnode *vp; struct vattr vat; PROC_LOCK(p); error = p_candebug(td, p); PROC_UNLOCK(p); if (error) return (error); if (uio->uio_rw != UIO_READ) return (EOPNOTSUPP); error = 0; vm = vmspace_acquire_ref(p); if (vm == NULL) return (ESRCH); if (SV_CURPROC_FLAG(SV_LP64)) l_map_str = l64_map_str; else l_map_str = l32_map_str; map = &vm->vm_map; vm_map_lock_read(map); for (entry = map->header.next; entry != &map->header; entry = entry->next) { name = ""; freename = NULL; if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) continue; e_prot = entry->protection; e_start = entry->start; e_end = entry->end; obj = entry->object.vm_object; for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { VM_OBJECT_RLOCK(tobj); if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); lobj = tobj; } last_timestamp = map->timestamp; vm_map_unlock_read(map); ino = 0; if (lobj) { off = IDX_TO_OFF(lobj->size); - if (lobj->type == OBJT_VNODE) { - vp = lobj->handle; - if (vp) - vref(vp); - } - else - vp = NULL; + vp = vm_object_vnode(lobj); + if (vp != NULL) + vref(vp); if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); flags = obj->flags; ref_count = obj->ref_count; shadow_count = obj->shadow_count; VM_OBJECT_RUNLOCK(obj); - if (vp) { + if (vp != NULL) { vn_fullpath(td, vp, &name, &freename); vn_lock(vp, LK_SHARED | LK_RETRY); VOP_GETATTR(vp, &vat, td->td_ucred); ino = vat.va_fileid; vput(vp); } else if (SV_PROC_ABI(p) == SV_ABI_LINUX) { if (e_start == p->p_sysent->sv_shared_page_base) name = vdso_str; if (e_end == p->p_sysent->sv_usrstack) name = stack_str; } } else { flags = 0; ref_count = 0; shadow_count = 0; } /* * format: * start, end, access, offset, major, minor, inode, name. */ error = sbuf_printf(sb, l_map_str, (u_long)e_start, (u_long)e_end, (e_prot & VM_PROT_READ)?"r":"-", (e_prot & VM_PROT_WRITE)?"w":"-", (e_prot & VM_PROT_EXECUTE)?"x":"-", "p", (u_long)off, 0, 0, (u_long)ino, *name ? " " : "", name ); if (freename) free(freename, M_TEMP); vm_map_lock_read(map); if (error == -1) { error = 0; break; } if (last_timestamp != map->timestamp) { /* * Look again for the entry because the map was * modified while it was unlocked. Specifically, * the entry may have been clipped, merged, or deleted. */ vm_map_lookup_entry(map, e_end - 1, &tmp_entry); entry = tmp_entry; } } vm_map_unlock_read(map); vmspace_free(vm); return (error); } /* * Criteria for interface name translation */ #define IFP_IS_ETH(ifp) (ifp->if_type == IFT_ETHER) static int linux_ifname(struct ifnet *ifp, char *buffer, size_t buflen) { struct ifnet *ifscan; int ethno; IFNET_RLOCK_ASSERT(); /* Short-circuit non ethernet interfaces */ if (!IFP_IS_ETH(ifp)) return (strlcpy(buffer, ifp->if_xname, buflen)); /* Determine the (relative) unit number for ethernet interfaces */ ethno = 0; TAILQ_FOREACH(ifscan, &V_ifnet, if_link) { if (ifscan == ifp) return (snprintf(buffer, buflen, "eth%d", ethno)); if (IFP_IS_ETH(ifscan)) ethno++; } return (0); } /* * Filler function for proc/net/dev */ static int linprocfs_donetdev(PFS_FILL_ARGS) { char ifname[16]; /* XXX LINUX_IFNAMSIZ */ struct ifnet *ifp; sbuf_printf(sb, "%6s|%58s|%s\n" "%6s|%58s|%58s\n", "Inter-", " Receive", " Transmit", " face", "bytes packets errs drop fifo frame compressed multicast", "bytes packets errs drop fifo colls carrier compressed"); CURVNET_SET(TD_TO_VNET(curthread)); IFNET_RLOCK(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { linux_ifname(ifp, ifname, sizeof ifname); sbuf_printf(sb, "%6.6s: ", ifname); sbuf_printf(sb, "%7ju %7ju %4ju %4ju %4lu %5lu %10lu %9ju ", (uintmax_t )ifp->if_get_counter(ifp, IFCOUNTER_IBYTES), (uintmax_t )ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS), (uintmax_t )ifp->if_get_counter(ifp, IFCOUNTER_IERRORS), (uintmax_t )ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS), /* rx_missed_errors */ 0UL, /* rx_fifo_errors */ 0UL, /* rx_length_errors + * rx_over_errors + * rx_crc_errors + * rx_frame_errors */ 0UL, /* rx_compressed */ (uintmax_t )ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS)); /* XXX-BZ rx only? */ sbuf_printf(sb, "%8ju %7ju %4ju %4ju %4lu %5ju %7lu %10lu\n", (uintmax_t )ifp->if_get_counter(ifp, IFCOUNTER_OBYTES), (uintmax_t )ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS), (uintmax_t )ifp->if_get_counter(ifp, IFCOUNTER_OERRORS), (uintmax_t )ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS), 0UL, /* tx_fifo_errors */ (uintmax_t )ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS), 0UL, /* tx_carrier_errors + * tx_aborted_errors + * tx_window_errors + * tx_heartbeat_errors*/ 0UL); /* tx_compressed */ } IFNET_RUNLOCK(); CURVNET_RESTORE(); return (0); } /* * Filler function for proc/sys/kernel/osrelease */ static int linprocfs_doosrelease(PFS_FILL_ARGS) { char osrelease[LINUX_MAX_UTSNAME]; linux_get_osrelease(td, osrelease); sbuf_printf(sb, "%s\n", osrelease); return (0); } /* * Filler function for proc/sys/kernel/ostype */ static int linprocfs_doostype(PFS_FILL_ARGS) { char osname[LINUX_MAX_UTSNAME]; linux_get_osname(td, osname); sbuf_printf(sb, "%s\n", osname); return (0); } /* * Filler function for proc/sys/kernel/version */ static int linprocfs_doosbuild(PFS_FILL_ARGS) { linprocfs_osbuild(td, sb); sbuf_cat(sb, "\n"); return (0); } /* * Filler function for proc/sys/kernel/msgmni */ static int linprocfs_domsgmni(PFS_FILL_ARGS) { sbuf_printf(sb, "%d\n", msginfo.msgmni); return (0); } /* * Filler function for proc/sys/kernel/pid_max */ static int linprocfs_dopid_max(PFS_FILL_ARGS) { sbuf_printf(sb, "%i\n", PID_MAX); return (0); } /* * Filler function for proc/sys/kernel/sem */ static int linprocfs_dosem(PFS_FILL_ARGS) { sbuf_printf(sb, "%d %d %d %d\n", seminfo.semmsl, seminfo.semmns, seminfo.semopm, seminfo.semmni); return (0); } /* * Filler function for proc/scsi/device_info */ static int linprocfs_doscsidevinfo(PFS_FILL_ARGS) { return (0); } /* * Filler function for proc/scsi/scsi */ static int linprocfs_doscsiscsi(PFS_FILL_ARGS) { return (0); } /* * Filler function for proc/devices */ static int linprocfs_dodevices(PFS_FILL_ARGS) { char *char_devices; sbuf_printf(sb, "Character devices:\n"); char_devices = linux_get_char_devices(); sbuf_printf(sb, "%s", char_devices); linux_free_get_char_devices(char_devices); sbuf_printf(sb, "\nBlock devices:\n"); return (0); } /* * Filler function for proc/cmdline */ static int linprocfs_docmdline(PFS_FILL_ARGS) { sbuf_printf(sb, "BOOT_IMAGE=%s", kernelname); sbuf_printf(sb, " ro root=302\n"); return (0); } /* * Filler function for proc/filesystems */ static int linprocfs_dofilesystems(PFS_FILL_ARGS) { struct vfsconf *vfsp; mtx_lock(&Giant); TAILQ_FOREACH(vfsp, &vfsconf, vfc_list) { if (vfsp->vfc_flags & VFCF_SYNTHETIC) sbuf_printf(sb, "nodev"); sbuf_printf(sb, "\t%s\n", vfsp->vfc_name); } mtx_unlock(&Giant); return(0); } #if 0 /* * Filler function for proc/modules */ static int linprocfs_domodules(PFS_FILL_ARGS) { struct linker_file *lf; TAILQ_FOREACH(lf, &linker_files, link) { sbuf_printf(sb, "%-20s%8lu%4d\n", lf->filename, (unsigned long)lf->size, lf->refs); } return (0); } #endif /* * Filler function for proc/pid/fd */ static int linprocfs_dofdescfs(PFS_FILL_ARGS) { if (p == curproc) sbuf_printf(sb, "/dev/fd"); else sbuf_printf(sb, "unknown"); return (0); } /* * Filler function for proc/sys/kernel/random/uuid */ static int linprocfs_douuid(PFS_FILL_ARGS) { struct uuid uuid; kern_uuidgen(&uuid, 1); sbuf_printf_uuid(sb, &uuid); sbuf_printf(sb, "\n"); return(0); } /* * Filler function for proc/pid/auxv */ static int linprocfs_doauxv(PFS_FILL_ARGS) { struct sbuf *asb; off_t buflen, resid; int error; /* * Mimic linux behavior and pass only processes with usermode * address space as valid. Return zero silently otherwise. */ if (p->p_vmspace == &vmspace0) return (0); if (uio->uio_resid == 0) return (0); if (uio->uio_offset < 0 || uio->uio_resid < 0) return (EINVAL); asb = sbuf_new_auto(); if (asb == NULL) return (ENOMEM); error = proc_getauxv(td, p, asb); if (error == 0) error = sbuf_finish(asb); resid = sbuf_len(asb) - uio->uio_offset; if (resid > uio->uio_resid) buflen = uio->uio_resid; else buflen = resid; if (buflen > IOSIZE_MAX) return (EINVAL); if (buflen > MAXPHYS) buflen = MAXPHYS; if (resid <= 0) return (0); if (error == 0) error = uiomove(sbuf_data(asb) + uio->uio_offset, buflen, uio); sbuf_delete(asb); return (error); } /* * Constructor */ static int linprocfs_init(PFS_INIT_ARGS) { struct pfs_node *root; struct pfs_node *dir; root = pi->pi_root; /* /proc/... */ pfs_create_file(root, "cmdline", &linprocfs_docmdline, NULL, NULL, NULL, PFS_RD); pfs_create_file(root, "cpuinfo", &linprocfs_docpuinfo, NULL, NULL, NULL, PFS_RD); pfs_create_file(root, "devices", &linprocfs_dodevices, NULL, NULL, NULL, PFS_RD); pfs_create_file(root, "filesystems", &linprocfs_dofilesystems, NULL, NULL, NULL, PFS_RD); pfs_create_file(root, "loadavg", &linprocfs_doloadavg, NULL, NULL, NULL, PFS_RD); pfs_create_file(root, "meminfo", &linprocfs_domeminfo, NULL, NULL, NULL, PFS_RD); #if 0 pfs_create_file(root, "modules", &linprocfs_domodules, NULL, NULL, NULL, PFS_RD); #endif pfs_create_file(root, "mounts", &linprocfs_domtab, NULL, NULL, NULL, PFS_RD); pfs_create_file(root, "mtab", &linprocfs_domtab, NULL, NULL, NULL, PFS_RD); pfs_create_file(root, "partitions", &linprocfs_dopartitions, NULL, NULL, NULL, PFS_RD); pfs_create_link(root, "self", &procfs_docurproc, NULL, NULL, NULL, 0); pfs_create_file(root, "stat", &linprocfs_dostat, NULL, NULL, NULL, PFS_RD); pfs_create_file(root, "swaps", &linprocfs_doswaps, NULL, NULL, NULL, PFS_RD); pfs_create_file(root, "uptime", &linprocfs_douptime, NULL, NULL, NULL, PFS_RD); pfs_create_file(root, "version", &linprocfs_doversion, NULL, NULL, NULL, PFS_RD); /* /proc/net/... */ dir = pfs_create_dir(root, "net", NULL, NULL, NULL, 0); pfs_create_file(dir, "dev", &linprocfs_donetdev, NULL, NULL, NULL, PFS_RD); /* /proc//... */ dir = pfs_create_dir(root, "pid", NULL, NULL, NULL, PFS_PROCDEP); pfs_create_file(dir, "cmdline", &linprocfs_doproccmdline, NULL, NULL, NULL, PFS_RD); pfs_create_link(dir, "cwd", &linprocfs_doproccwd, NULL, NULL, NULL, 0); pfs_create_file(dir, "environ", &linprocfs_doprocenviron, NULL, &procfs_candebug, NULL, PFS_RD); pfs_create_link(dir, "exe", &procfs_doprocfile, NULL, &procfs_notsystem, NULL, 0); pfs_create_file(dir, "maps", &linprocfs_doprocmaps, NULL, NULL, NULL, PFS_RD); pfs_create_file(dir, "mem", &procfs_doprocmem, &procfs_attr, &procfs_candebug, NULL, PFS_RDWR|PFS_RAW); pfs_create_link(dir, "root", &linprocfs_doprocroot, NULL, NULL, NULL, 0); pfs_create_file(dir, "stat", &linprocfs_doprocstat, NULL, NULL, NULL, PFS_RD); pfs_create_file(dir, "statm", &linprocfs_doprocstatm, NULL, NULL, NULL, PFS_RD); pfs_create_file(dir, "status", &linprocfs_doprocstatus, NULL, NULL, NULL, PFS_RD); pfs_create_link(dir, "fd", &linprocfs_dofdescfs, NULL, NULL, NULL, 0); pfs_create_file(dir, "auxv", &linprocfs_doauxv, NULL, &procfs_candebug, NULL, PFS_RD|PFS_RAWRD); /* /proc/scsi/... */ dir = pfs_create_dir(root, "scsi", NULL, NULL, NULL, 0); pfs_create_file(dir, "device_info", &linprocfs_doscsidevinfo, NULL, NULL, NULL, PFS_RD); pfs_create_file(dir, "scsi", &linprocfs_doscsiscsi, NULL, NULL, NULL, PFS_RD); /* /proc/sys/... */ dir = pfs_create_dir(root, "sys", NULL, NULL, NULL, 0); /* /proc/sys/kernel/... */ dir = pfs_create_dir(dir, "kernel", NULL, NULL, NULL, 0); pfs_create_file(dir, "osrelease", &linprocfs_doosrelease, NULL, NULL, NULL, PFS_RD); pfs_create_file(dir, "ostype", &linprocfs_doostype, NULL, NULL, NULL, PFS_RD); pfs_create_file(dir, "version", &linprocfs_doosbuild, NULL, NULL, NULL, PFS_RD); pfs_create_file(dir, "msgmni", &linprocfs_domsgmni, NULL, NULL, NULL, PFS_RD); pfs_create_file(dir, "pid_max", &linprocfs_dopid_max, NULL, NULL, NULL, PFS_RD); pfs_create_file(dir, "sem", &linprocfs_dosem, NULL, NULL, NULL, PFS_RD); /* /proc/sys/kernel/random/... */ dir = pfs_create_dir(dir, "random", NULL, NULL, NULL, 0); pfs_create_file(dir, "uuid", &linprocfs_douuid, NULL, NULL, NULL, PFS_RD); return (0); } /* * Destructor */ static int linprocfs_uninit(PFS_INIT_ARGS) { /* nothing to do, pseudofs will GC */ return (0); } PSEUDOFS(linprocfs, 1, 0); #if defined(__amd64__) MODULE_DEPEND(linprocfs, linux_common, 1, 1, 1); #else MODULE_DEPEND(linprocfs, linux, 1, 1, 1); #endif MODULE_DEPEND(linprocfs, procfs, 1, 1, 1); MODULE_DEPEND(linprocfs, sysvmsg, 1, 1, 1); MODULE_DEPEND(linprocfs, sysvsem, 1, 1, 1); Index: projects/em_mq/sys/dev/ahci/ahci.h =================================================================== --- projects/em_mq/sys/dev/ahci/ahci.h (revision 283943) +++ projects/em_mq/sys/dev/ahci/ahci.h (revision 283944) @@ -1,620 +1,620 @@ /*- * Copyright (c) 1998 - 2008 Søren Schmidt * Copyright (c) 2009-2012 Alexander Motin * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer, * without modification, immediately at the beginning of the file. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. * * $FreeBSD$ */ /* ATA register defines */ #define ATA_DATA 0 /* (RW) data */ #define ATA_FEATURE 1 /* (W) feature */ #define ATA_F_DMA 0x01 /* enable DMA */ #define ATA_F_OVL 0x02 /* enable overlap */ #define ATA_COUNT 2 /* (W) sector count */ #define ATA_SECTOR 3 /* (RW) sector # */ #define ATA_CYL_LSB 4 /* (RW) cylinder# LSB */ #define ATA_CYL_MSB 5 /* (RW) cylinder# MSB */ #define ATA_DRIVE 6 /* (W) Sector/Drive/Head */ #define ATA_D_LBA 0x40 /* use LBA addressing */ #define ATA_D_IBM 0xa0 /* 512 byte sectors, ECC */ #define ATA_COMMAND 7 /* (W) command */ #define ATA_ERROR 8 /* (R) error */ #define ATA_E_ILI 0x01 /* illegal length */ #define ATA_E_NM 0x02 /* no media */ #define ATA_E_ABORT 0x04 /* command aborted */ #define ATA_E_MCR 0x08 /* media change request */ #define ATA_E_IDNF 0x10 /* ID not found */ #define ATA_E_MC 0x20 /* media changed */ #define ATA_E_UNC 0x40 /* uncorrectable data */ #define ATA_E_ICRC 0x80 /* UDMA crc error */ #define ATA_E_ATAPI_SENSE_MASK 0xf0 /* ATAPI sense key mask */ #define ATA_IREASON 9 /* (R) interrupt reason */ #define ATA_I_CMD 0x01 /* cmd (1) | data (0) */ #define ATA_I_IN 0x02 /* read (1) | write (0) */ #define ATA_I_RELEASE 0x04 /* released bus (1) */ #define ATA_I_TAGMASK 0xf8 /* tag mask */ #define ATA_STATUS 10 /* (R) status */ #define ATA_ALTSTAT 11 /* (R) alternate status */ #define ATA_S_ERROR 0x01 /* error */ #define ATA_S_INDEX 0x02 /* index */ #define ATA_S_CORR 0x04 /* data corrected */ #define ATA_S_DRQ 0x08 /* data request */ #define ATA_S_DSC 0x10 /* drive seek completed */ #define ATA_S_SERVICE 0x10 /* drive needs service */ #define ATA_S_DWF 0x20 /* drive write fault */ #define ATA_S_DMA 0x20 /* DMA ready */ #define ATA_S_READY 0x40 /* drive ready */ #define ATA_S_BUSY 0x80 /* busy */ #define ATA_CONTROL 12 /* (W) control */ #define ATA_A_IDS 0x02 /* disable interrupts */ #define ATA_A_RESET 0x04 /* RESET controller */ #define ATA_A_4BIT 0x08 /* 4 head bits */ #define ATA_A_HOB 0x80 /* High Order Byte enable */ /* SATA register defines */ #define ATA_SSTATUS 13 #define ATA_SS_DET_MASK 0x0000000f #define ATA_SS_DET_NO_DEVICE 0x00000000 #define ATA_SS_DET_DEV_PRESENT 0x00000001 #define ATA_SS_DET_PHY_ONLINE 0x00000003 #define ATA_SS_DET_PHY_OFFLINE 0x00000004 #define ATA_SS_SPD_MASK 0x000000f0 #define ATA_SS_SPD_NO_SPEED 0x00000000 #define ATA_SS_SPD_GEN1 0x00000010 #define ATA_SS_SPD_GEN2 0x00000020 #define ATA_SS_SPD_GEN3 0x00000030 #define ATA_SS_IPM_MASK 0x00000f00 #define ATA_SS_IPM_NO_DEVICE 0x00000000 #define ATA_SS_IPM_ACTIVE 0x00000100 #define ATA_SS_IPM_PARTIAL 0x00000200 #define ATA_SS_IPM_SLUMBER 0x00000600 #define ATA_SS_IPM_DEVSLEEP 0x00000800 #define ATA_SERROR 14 #define ATA_SE_DATA_CORRECTED 0x00000001 #define ATA_SE_COMM_CORRECTED 0x00000002 #define ATA_SE_DATA_ERR 0x00000100 #define ATA_SE_COMM_ERR 0x00000200 #define ATA_SE_PROT_ERR 0x00000400 #define ATA_SE_HOST_ERR 0x00000800 #define ATA_SE_PHY_CHANGED 0x00010000 #define ATA_SE_PHY_IERROR 0x00020000 #define ATA_SE_COMM_WAKE 0x00040000 #define ATA_SE_DECODE_ERR 0x00080000 #define ATA_SE_PARITY_ERR 0x00100000 #define ATA_SE_CRC_ERR 0x00200000 #define ATA_SE_HANDSHAKE_ERR 0x00400000 #define ATA_SE_LINKSEQ_ERR 0x00800000 #define ATA_SE_TRANSPORT_ERR 0x01000000 #define ATA_SE_UNKNOWN_FIS 0x02000000 #define ATA_SE_EXCHANGED 0x04000000 #define ATA_SCONTROL 15 #define ATA_SC_DET_MASK 0x0000000f #define ATA_SC_DET_IDLE 0x00000000 #define ATA_SC_DET_RESET 0x00000001 #define ATA_SC_DET_DISABLE 0x00000004 #define ATA_SC_SPD_MASK 0x000000f0 #define ATA_SC_SPD_NO_SPEED 0x00000000 #define ATA_SC_SPD_SPEED_GEN1 0x00000010 #define ATA_SC_SPD_SPEED_GEN2 0x00000020 #define ATA_SC_SPD_SPEED_GEN3 0x00000030 #define ATA_SC_IPM_MASK 0x00000f00 #define ATA_SC_IPM_NONE 0x00000000 #define ATA_SC_IPM_DIS_PARTIAL 0x00000100 #define ATA_SC_IPM_DIS_SLUMBER 0x00000200 #define ATA_SC_IPM_DIS_DEVSLEEP 0x00000400 #define ATA_SACTIVE 16 #define AHCI_MAX_PORTS 32 #define AHCI_MAX_SLOTS 32 #define AHCI_MAX_IRQS 16 /* SATA AHCI v1.0 register defines */ #define AHCI_CAP 0x00 #define AHCI_CAP_NPMASK 0x0000001f #define AHCI_CAP_SXS 0x00000020 #define AHCI_CAP_EMS 0x00000040 #define AHCI_CAP_CCCS 0x00000080 #define AHCI_CAP_NCS 0x00001F00 #define AHCI_CAP_NCS_SHIFT 8 #define AHCI_CAP_PSC 0x00002000 #define AHCI_CAP_SSC 0x00004000 #define AHCI_CAP_PMD 0x00008000 #define AHCI_CAP_FBSS 0x00010000 #define AHCI_CAP_SPM 0x00020000 #define AHCI_CAP_SAM 0x00080000 #define AHCI_CAP_ISS 0x00F00000 #define AHCI_CAP_ISS_SHIFT 20 #define AHCI_CAP_SCLO 0x01000000 #define AHCI_CAP_SAL 0x02000000 #define AHCI_CAP_SALP 0x04000000 #define AHCI_CAP_SSS 0x08000000 #define AHCI_CAP_SMPS 0x10000000 #define AHCI_CAP_SSNTF 0x20000000 #define AHCI_CAP_SNCQ 0x40000000 #define AHCI_CAP_64BIT 0x80000000 #define AHCI_GHC 0x04 #define AHCI_GHC_AE 0x80000000 #define AHCI_GHC_MRSM 0x00000004 #define AHCI_GHC_IE 0x00000002 #define AHCI_GHC_HR 0x00000001 #define AHCI_IS 0x08 #define AHCI_PI 0x0c #define AHCI_VS 0x10 #define AHCI_CCCC 0x14 #define AHCI_CCCC_TV_MASK 0xffff0000 #define AHCI_CCCC_TV_SHIFT 16 #define AHCI_CCCC_CC_MASK 0x0000ff00 #define AHCI_CCCC_CC_SHIFT 8 #define AHCI_CCCC_INT_MASK 0x000000f8 #define AHCI_CCCC_INT_SHIFT 3 #define AHCI_CCCC_EN 0x00000001 #define AHCI_CCCP 0x18 #define AHCI_EM_LOC 0x1C #define AHCI_EM_CTL 0x20 #define AHCI_EM_MR 0x00000001 #define AHCI_EM_TM 0x00000100 #define AHCI_EM_RST 0x00000200 #define AHCI_EM_LED 0x00010000 #define AHCI_EM_SAFTE 0x00020000 #define AHCI_EM_SES2 0x00040000 #define AHCI_EM_SGPIO 0x00080000 #define AHCI_EM_SMB 0x01000000 #define AHCI_EM_XMT 0x02000000 #define AHCI_EM_ALHD 0x04000000 #define AHCI_EM_PM 0x08000000 #define AHCI_CAP2 0x24 #define AHCI_CAP2_BOH 0x00000001 #define AHCI_CAP2_NVMP 0x00000002 #define AHCI_CAP2_APST 0x00000004 #define AHCI_CAP2_SDS 0x00000008 #define AHCI_CAP2_SADM 0x00000010 #define AHCI_CAP2_DESO 0x00000020 #define AHCI_OFFSET 0x100 #define AHCI_STEP 0x80 #define AHCI_P_CLB 0x00 #define AHCI_P_CLBU 0x04 #define AHCI_P_FB 0x08 #define AHCI_P_FBU 0x0c #define AHCI_P_IS 0x10 #define AHCI_P_IE 0x14 #define AHCI_P_IX_DHR 0x00000001 #define AHCI_P_IX_PS 0x00000002 #define AHCI_P_IX_DS 0x00000004 #define AHCI_P_IX_SDB 0x00000008 #define AHCI_P_IX_UF 0x00000010 #define AHCI_P_IX_DP 0x00000020 #define AHCI_P_IX_PC 0x00000040 #define AHCI_P_IX_MP 0x00000080 #define AHCI_P_IX_PRC 0x00400000 #define AHCI_P_IX_IPM 0x00800000 #define AHCI_P_IX_OF 0x01000000 #define AHCI_P_IX_INF 0x04000000 #define AHCI_P_IX_IF 0x08000000 #define AHCI_P_IX_HBD 0x10000000 #define AHCI_P_IX_HBF 0x20000000 #define AHCI_P_IX_TFE 0x40000000 #define AHCI_P_IX_CPD 0x80000000 #define AHCI_P_CMD 0x18 #define AHCI_P_CMD_ST 0x00000001 #define AHCI_P_CMD_SUD 0x00000002 #define AHCI_P_CMD_POD 0x00000004 #define AHCI_P_CMD_CLO 0x00000008 #define AHCI_P_CMD_FRE 0x00000010 #define AHCI_P_CMD_CCS_MASK 0x00001f00 #define AHCI_P_CMD_CCS_SHIFT 8 #define AHCI_P_CMD_ISS 0x00002000 #define AHCI_P_CMD_FR 0x00004000 #define AHCI_P_CMD_CR 0x00008000 #define AHCI_P_CMD_CPS 0x00010000 #define AHCI_P_CMD_PMA 0x00020000 #define AHCI_P_CMD_HPCP 0x00040000 #define AHCI_P_CMD_MPSP 0x00080000 #define AHCI_P_CMD_CPD 0x00100000 #define AHCI_P_CMD_ESP 0x00200000 #define AHCI_P_CMD_FBSCP 0x00400000 #define AHCI_P_CMD_APSTE 0x00800000 #define AHCI_P_CMD_ATAPI 0x01000000 #define AHCI_P_CMD_DLAE 0x02000000 #define AHCI_P_CMD_ALPE 0x04000000 #define AHCI_P_CMD_ASP 0x08000000 #define AHCI_P_CMD_ICC_MASK 0xf0000000 #define AHCI_P_CMD_NOOP 0x00000000 #define AHCI_P_CMD_ACTIVE 0x10000000 #define AHCI_P_CMD_PARTIAL 0x20000000 #define AHCI_P_CMD_SLUMBER 0x60000000 #define AHCI_P_CMD_DEVSLEEP 0x80000000 #define AHCI_P_TFD 0x20 #define AHCI_P_SIG 0x24 #define AHCI_P_SSTS 0x28 #define AHCI_P_SCTL 0x2c #define AHCI_P_SERR 0x30 #define AHCI_P_SACT 0x34 #define AHCI_P_CI 0x38 #define AHCI_P_SNTF 0x3C #define AHCI_P_FBS 0x40 #define AHCI_P_FBS_EN 0x00000001 #define AHCI_P_FBS_DEC 0x00000002 #define AHCI_P_FBS_SDE 0x00000004 #define AHCI_P_FBS_DEV 0x00000f00 #define AHCI_P_FBS_DEV_SHIFT 8 #define AHCI_P_FBS_ADO 0x0000f000 #define AHCI_P_FBS_ADO_SHIFT 12 #define AHCI_P_FBS_DWE 0x000f0000 #define AHCI_P_FBS_DWE_SHIFT 16 #define AHCI_P_DEVSLP 0x44 #define AHCI_P_DEVSLP_ADSE 0x00000001 #define AHCI_P_DEVSLP_DSP 0x00000002 #define AHCI_P_DEVSLP_DETO 0x000003fc #define AHCI_P_DEVSLP_DETO_SHIFT 2 #define AHCI_P_DEVSLP_MDAT 0x00007c00 #define AHCI_P_DEVSLP_MDAT_SHIFT 10 #define AHCI_P_DEVSLP_DITO 0x01ff8000 #define AHCI_P_DEVSLP_DITO_SHIFT 15 #define AHCI_P_DEVSLP_DM 0x0e000000 #define AHCI_P_DEVSLP_DM_SHIFT 25 /* Just to be sure, if building as module. */ #if MAXPHYS < 512 * 1024 #undef MAXPHYS #define MAXPHYS 512 * 1024 #endif /* Pessimistic prognosis on number of required S/G entries */ #define AHCI_SG_ENTRIES (roundup(btoc(MAXPHYS) + 1, 8)) /* Command list. 32 commands. First, 1Kbyte aligned. */ #define AHCI_CL_OFFSET 0 #define AHCI_CL_SIZE 32 /* Command tables. Up to 32 commands, Each, 128byte aligned. */ #define AHCI_CT_OFFSET (AHCI_CL_OFFSET + AHCI_CL_SIZE * AHCI_MAX_SLOTS) #define AHCI_CT_SIZE (128 + AHCI_SG_ENTRIES * 16) /* Total main work area. */ #define AHCI_WORK_SIZE (AHCI_CT_OFFSET + AHCI_CT_SIZE * ch->numslots) struct ahci_dma_prd { u_int64_t dba; u_int32_t reserved; u_int32_t dbc; /* 0 based */ #define AHCI_PRD_MASK 0x003fffff /* max 4MB */ #define AHCI_PRD_MAX (AHCI_PRD_MASK + 1) #define AHCI_PRD_IPC (1U << 31) } __packed; struct ahci_cmd_tab { u_int8_t cfis[64]; u_int8_t acmd[32]; u_int8_t reserved[32]; struct ahci_dma_prd prd_tab[AHCI_SG_ENTRIES]; } __packed; struct ahci_cmd_list { u_int16_t cmd_flags; #define AHCI_CMD_ATAPI 0x0020 #define AHCI_CMD_WRITE 0x0040 #define AHCI_CMD_PREFETCH 0x0080 #define AHCI_CMD_RESET 0x0100 #define AHCI_CMD_BIST 0x0200 #define AHCI_CMD_CLR_BUSY 0x0400 u_int16_t prd_length; /* PRD entries */ u_int32_t bytecount; u_int64_t cmd_table_phys; /* 128byte aligned */ } __packed; /* misc defines */ #define ATA_IRQ_RID 0 #define ATA_INTR_FLAGS (INTR_MPSAFE|INTR_TYPE_BIO|INTR_ENTROPY) struct ata_dmaslot { bus_dmamap_t data_map; /* data DMA map */ int nsegs; /* Number of segs loaded */ }; /* structure holding DMA related information */ struct ata_dma { bus_dma_tag_t work_tag; /* workspace DMA tag */ bus_dmamap_t work_map; /* workspace DMA map */ uint8_t *work; /* workspace */ bus_addr_t work_bus; /* bus address of work */ bus_dma_tag_t rfis_tag; /* RFIS list DMA tag */ bus_dmamap_t rfis_map; /* RFIS list DMA map */ uint8_t *rfis; /* FIS receive area */ bus_addr_t rfis_bus; /* bus address of rfis */ bus_dma_tag_t data_tag; /* data DMA tag */ }; enum ahci_slot_states { AHCI_SLOT_EMPTY, AHCI_SLOT_LOADING, AHCI_SLOT_RUNNING, AHCI_SLOT_EXECUTING }; struct ahci_slot { struct ahci_channel *ch; /* Channel */ u_int8_t slot; /* Number of this slot */ enum ahci_slot_states state; /* Slot state */ union ccb *ccb; /* CCB occupying slot */ struct ata_dmaslot dma; /* DMA data of this slot */ struct callout timeout; /* Execution timeout */ }; struct ahci_device { int revision; int mode; u_int bytecount; u_int atapi; u_int tags; u_int caps; }; struct ahci_led { device_t dev; /* Device handle */ struct cdev *led; uint8_t num; /* Number of this led */ uint8_t state; /* State of this led */ }; #define AHCI_NUM_LEDS 3 /* structure describing an ATA channel */ struct ahci_channel { device_t dev; /* Device handle */ int unit; /* Physical channel */ struct resource *r_mem; /* Memory of this channel */ struct resource *r_irq; /* Interrupt of this channel */ void *ih; /* Interrupt handle */ struct ata_dma dma; /* DMA data */ struct cam_sim *sim; struct cam_path *path; uint32_t caps; /* Controller capabilities */ uint32_t caps2; /* Controller capabilities */ uint32_t chcaps; /* Channel capabilities */ uint32_t chscaps; /* Channel sleep capabilities */ uint16_t vendorid; /* Vendor ID from the bus */ uint16_t deviceid; /* Device ID from the bus */ uint16_t subvendorid; /* Subvendor ID from the bus */ uint16_t subdeviceid; /* Subdevice ID from the bus */ int quirks; int numslots; /* Number of present slots */ int pm_level; /* power management level */ int devices; /* What is present */ int pm_present; /* PM presence reported */ int fbs_enabled; /* FIS-based switching enabled */ union ccb *hold[AHCI_MAX_SLOTS]; struct ahci_slot slot[AHCI_MAX_SLOTS]; uint32_t oslots; /* Occupied slots */ uint32_t rslots; /* Running slots */ uint32_t aslots; /* Slots with atomic commands */ uint32_t eslots; /* Slots in error */ uint32_t toslots; /* Slots in timeout */ int lastslot; /* Last used slot */ int taggedtarget; /* Last tagged target */ int numrslots; /* Number of running slots */ int numrslotspd[16];/* Number of running slots per dev */ int numtslots; /* Number of tagged slots */ int numtslotspd[16];/* Number of tagged slots per dev */ int numhslots; /* Number of held slots */ int recoverycmd; /* Our READ LOG active */ int fatalerr; /* Fatal error happend */ int resetting; /* Hard-reset in progress. */ int resetpolldiv; /* Hard-reset poll divider. */ int listening; /* SUD bit is cleared. */ int wrongccs; /* CCS field in CMD was wrong */ union ccb *frozen; /* Frozen command */ struct callout pm_timer; /* Power management events */ struct callout reset_timer; /* Hard-reset timeout */ struct ahci_device user[16]; /* User-specified settings */ struct ahci_device curr[16]; /* Current settings */ struct mtx_padalign mtx; /* state lock */ STAILQ_HEAD(, ccb_hdr) doneq; /* queue of completed CCBs */ int batch; /* doneq is in use */ }; struct ahci_enclosure { device_t dev; /* Device handle */ struct resource *r_memc; /* Control register */ struct resource *r_memt; /* Transmit buffer */ struct resource *r_memr; /* Recieve buffer */ struct cam_sim *sim; struct cam_path *path; struct mtx mtx; /* state lock */ struct ahci_led leds[AHCI_MAX_PORTS * 3]; uint32_t capsem; /* Controller capabilities */ uint8_t status[AHCI_MAX_PORTS][4]; /* ArrayDev statuses */ int quirks; int channels; int ichannels; }; /* structure describing a AHCI controller */ struct ahci_controller { device_t dev; bus_dma_tag_t dma_tag; int r_rid; uint16_t vendorid; /* Vendor ID from the bus */ uint16_t deviceid; /* Device ID from the bus */ uint16_t subvendorid; /* Subvendor ID from the bus */ uint16_t subdeviceid; /* Subdevice ID from the bus */ struct resource *r_mem; struct rman sc_iomem; struct ahci_controller_irq { struct ahci_controller *ctlr; struct resource *r_irq; void *handle; int r_irq_rid; int mode; #define AHCI_IRQ_MODE_ALL 0 #define AHCI_IRQ_MODE_AFTER 1 #define AHCI_IRQ_MODE_ONE 2 } irqs[AHCI_MAX_IRQS]; uint32_t caps; /* Controller capabilities */ uint32_t caps2; /* Controller capabilities */ uint32_t capsem; /* Controller capabilities */ uint32_t emloc; /* EM buffer location */ int quirks; int numirqs; int channels; int ichannels; int ccc; /* CCC timeout */ int cccv; /* CCC vector */ int direct; /* Direct command completion */ int msi; /* MSI interupts */ struct { void (*function)(void *); void *argument; } interrupt[AHCI_MAX_PORTS]; }; enum ahci_err_type { AHCI_ERR_NONE, /* No error */ AHCI_ERR_INVALID, /* Error detected by us before submitting. */ AHCI_ERR_INNOCENT, /* Innocent victim. */ AHCI_ERR_TFE, /* Task File Error. */ AHCI_ERR_SATA, /* SATA error. */ AHCI_ERR_TIMEOUT, /* Command execution timeout. */ AHCI_ERR_NCQ, /* NCQ command error. CCB should be put on hold * until READ LOG executed to reveal error. */ }; /* macros to hide busspace uglyness */ #define ATA_INB(res, offset) \ bus_read_1((res), (offset)) #define ATA_INW(res, offset) \ bus_read_2((res), (offset)) #define ATA_INL(res, offset) \ bus_read_4((res), (offset)) #define ATA_INSW(res, offset, addr, count) \ bus_read_multi_2((res), (offset), (addr), (count)) #define ATA_INSW_STRM(res, offset, addr, count) \ bus_read_multi_stream_2((res), (offset), (addr), (count)) #define ATA_INSL(res, offset, addr, count) \ bus_read_multi_4((res), (offset), (addr), (count)) #define ATA_INSL_STRM(res, offset, addr, count) \ bus_read_multi_stream_4((res), (offset), (addr), (count)) #define ATA_OUTB(res, offset, value) \ bus_write_1((res), (offset), (value)) #define ATA_OUTW(res, offset, value) \ bus_write_2((res), (offset), (value)) #define ATA_OUTL(res, offset, value) \ bus_write_4((res), (offset), (value)) #define ATA_OUTSW(res, offset, addr, count) \ bus_write_multi_2((res), (offset), (addr), (count)) #define ATA_OUTSW_STRM(res, offset, addr, count) \ bus_write_multi_stream_2((res), (offset), (addr), (count)) #define ATA_OUTSL(res, offset, addr, count) \ bus_write_multi_4((res), (offset), (addr), (count)) #define ATA_OUTSL_STRM(res, offset, addr, count) \ bus_write_multi_stream_4((res), (offset), (addr), (count)) #define AHCI_Q_NOFORCE 0x00000001 #define AHCI_Q_NOPMP 0x00000002 #define AHCI_Q_NONCQ 0x00000004 #define AHCI_Q_1CH 0x00000008 #define AHCI_Q_2CH 0x00000010 #define AHCI_Q_4CH 0x00000020 #define AHCI_Q_EDGEIS 0x00000040 #define AHCI_Q_SATA2 0x00000080 #define AHCI_Q_NOBSYRES 0x00000100 #define AHCI_Q_NOAA 0x00000200 #define AHCI_Q_NOCOUNT 0x00000400 #define AHCI_Q_ALTSIG 0x00000800 #define AHCI_Q_NOMSI 0x00001000 #define AHCI_Q_ATI_PMP_BUG 0x00002000 #define AHCI_Q_MAXIO_64K 0x00004000 #define AHCI_Q_SATA1_UNIT0 0x00008000 /* need better method for this */ #define AHCI_Q_ABAR0 0x00010000 #define AHCI_Q_1MSI 0x00020000 #define AHCI_Q_FORCE_PI 0x00040000 #define AHCI_Q_RESTORE_CAP 0x00080000 #define AHCI_Q_BIT_STRING \ - "\021" \ + "\020" \ "\001NOFORCE" \ "\002NOPMP" \ "\003NONCQ" \ "\0041CH" \ "\0052CH" \ "\0064CH" \ "\007EDGEIS" \ "\010SATA2" \ "\011NOBSYRES" \ "\012NOAA" \ "\013NOCOUNT" \ "\014ALTSIG" \ "\015NOMSI" \ "\016ATI_PMP_BUG" \ "\017MAXIO_64K" \ "\020SATA1_UNIT0" \ "\021ABAR0" \ "\0221MSI" \ "\022FORCE_PI" \ "\023RESTORE_CAP" int ahci_attach(device_t dev); int ahci_detach(device_t dev); int ahci_setup_interrupt(device_t dev); int ahci_print_child(device_t dev, device_t child); struct resource *ahci_alloc_resource(device_t dev, device_t child, int type, int *rid, u_long start, u_long end, u_long count, u_int flags); int ahci_release_resource(device_t dev, device_t child, int type, int rid, struct resource *r); int ahci_setup_intr(device_t dev, device_t child, struct resource *irq, int flags, driver_filter_t *filter, driver_intr_t *function, void *argument, void **cookiep); int ahci_teardown_intr(device_t dev, device_t child, struct resource *irq, void *cookie); int ahci_child_location_str(device_t dev, device_t child, char *buf, size_t buflen); bus_dma_tag_t ahci_get_dma_tag(device_t dev, device_t child); int ahci_ctlr_reset(device_t dev); int ahci_ctlr_setup(device_t dev); Index: projects/em_mq/sys/dev/hwpmc/hwpmc_mod.c =================================================================== --- projects/em_mq/sys/dev/hwpmc/hwpmc_mod.c (revision 283943) +++ projects/em_mq/sys/dev/hwpmc/hwpmc_mod.c (revision 283944) @@ -1,5125 +1,5125 @@ /*- * Copyright (c) 2003-2008 Joseph Koshy * Copyright (c) 2007 The FreeBSD Foundation * All rights reserved. * * Portions of this software were developed by A. Joseph Koshy under * sponsorship from the FreeBSD Foundation and Google, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* needs to be after */ #include #include #include #include #include #include #include #include "hwpmc_soft.h" /* * Types */ enum pmc_flags { PMC_FLAG_NONE = 0x00, /* do nothing */ PMC_FLAG_REMOVE = 0x01, /* atomically remove entry from hash */ PMC_FLAG_ALLOCATE = 0x02, /* add entry to hash if not found */ }; /* * The offset in sysent where the syscall is allocated. */ static int pmc_syscall_num = NO_SYSCALL; struct pmc_cpu **pmc_pcpu; /* per-cpu state */ pmc_value_t *pmc_pcpu_saved; /* saved PMC values: CSW handling */ #define PMC_PCPU_SAVED(C,R) pmc_pcpu_saved[(R) + md->pmd_npmc*(C)] struct mtx_pool *pmc_mtxpool; static int *pmc_pmcdisp; /* PMC row dispositions */ #define PMC_ROW_DISP_IS_FREE(R) (pmc_pmcdisp[(R)] == 0) #define PMC_ROW_DISP_IS_THREAD(R) (pmc_pmcdisp[(R)] > 0) #define PMC_ROW_DISP_IS_STANDALONE(R) (pmc_pmcdisp[(R)] < 0) #define PMC_MARK_ROW_FREE(R) do { \ pmc_pmcdisp[(R)] = 0; \ } while (0) #define PMC_MARK_ROW_STANDALONE(R) do { \ KASSERT(pmc_pmcdisp[(R)] <= 0, ("[pmc,%d] row disposition error", \ __LINE__)); \ atomic_add_int(&pmc_pmcdisp[(R)], -1); \ KASSERT(pmc_pmcdisp[(R)] >= (-pmc_cpu_max_active()), \ ("[pmc,%d] row disposition error", __LINE__)); \ } while (0) #define PMC_UNMARK_ROW_STANDALONE(R) do { \ atomic_add_int(&pmc_pmcdisp[(R)], 1); \ KASSERT(pmc_pmcdisp[(R)] <= 0, ("[pmc,%d] row disposition error", \ __LINE__)); \ } while (0) #define PMC_MARK_ROW_THREAD(R) do { \ KASSERT(pmc_pmcdisp[(R)] >= 0, ("[pmc,%d] row disposition error", \ __LINE__)); \ atomic_add_int(&pmc_pmcdisp[(R)], 1); \ } while (0) #define PMC_UNMARK_ROW_THREAD(R) do { \ atomic_add_int(&pmc_pmcdisp[(R)], -1); \ KASSERT(pmc_pmcdisp[(R)] >= 0, ("[pmc,%d] row disposition error", \ __LINE__)); \ } while (0) /* various event handlers */ static eventhandler_tag pmc_exit_tag, pmc_fork_tag, pmc_kld_load_tag, pmc_kld_unload_tag; /* Module statistics */ struct pmc_op_getdriverstats pmc_stats; /* Machine/processor dependent operations */ static struct pmc_mdep *md; /* * Hash tables mapping owner processes and target threads to PMCs. */ struct mtx pmc_processhash_mtx; /* spin mutex */ static u_long pmc_processhashmask; static LIST_HEAD(pmc_processhash, pmc_process) *pmc_processhash; /* * Hash table of PMC owner descriptors. This table is protected by * the shared PMC "sx" lock. */ static u_long pmc_ownerhashmask; static LIST_HEAD(pmc_ownerhash, pmc_owner) *pmc_ownerhash; /* * List of PMC owners with system-wide sampling PMCs. */ static LIST_HEAD(, pmc_owner) pmc_ss_owners; /* * A map of row indices to classdep structures. */ static struct pmc_classdep **pmc_rowindex_to_classdep; /* * Prototypes */ #ifdef HWPMC_DEBUG static int pmc_debugflags_sysctl_handler(SYSCTL_HANDLER_ARGS); static int pmc_debugflags_parse(char *newstr, char *fence); #endif static int load(struct module *module, int cmd, void *arg); static int pmc_attach_process(struct proc *p, struct pmc *pm); static struct pmc *pmc_allocate_pmc_descriptor(void); static struct pmc_owner *pmc_allocate_owner_descriptor(struct proc *p); static int pmc_attach_one_process(struct proc *p, struct pmc *pm); static int pmc_can_allocate_rowindex(struct proc *p, unsigned int ri, int cpu); static int pmc_can_attach(struct pmc *pm, struct proc *p); static void pmc_capture_user_callchain(int cpu, int soft, struct trapframe *tf); static void pmc_cleanup(void); static int pmc_detach_process(struct proc *p, struct pmc *pm); static int pmc_detach_one_process(struct proc *p, struct pmc *pm, int flags); static void pmc_destroy_owner_descriptor(struct pmc_owner *po); static void pmc_destroy_pmc_descriptor(struct pmc *pm); static struct pmc_owner *pmc_find_owner_descriptor(struct proc *p); static int pmc_find_pmc(pmc_id_t pmcid, struct pmc **pm); static struct pmc *pmc_find_pmc_descriptor_in_process(struct pmc_owner *po, pmc_id_t pmc); static struct pmc_process *pmc_find_process_descriptor(struct proc *p, uint32_t mode); static void pmc_force_context_switch(void); static void pmc_link_target_process(struct pmc *pm, struct pmc_process *pp); static void pmc_log_all_process_mappings(struct pmc_owner *po); static void pmc_log_kernel_mappings(struct pmc *pm); static void pmc_log_process_mappings(struct pmc_owner *po, struct proc *p); static void pmc_maybe_remove_owner(struct pmc_owner *po); static void pmc_process_csw_in(struct thread *td); static void pmc_process_csw_out(struct thread *td); static void pmc_process_exit(void *arg, struct proc *p); static void pmc_process_fork(void *arg, struct proc *p1, struct proc *p2, int n); static void pmc_process_samples(int cpu, int soft); static void pmc_release_pmc_descriptor(struct pmc *pmc); static void pmc_remove_owner(struct pmc_owner *po); static void pmc_remove_process_descriptor(struct pmc_process *pp); static void pmc_restore_cpu_binding(struct pmc_binding *pb); static void pmc_save_cpu_binding(struct pmc_binding *pb); static void pmc_select_cpu(int cpu); static int pmc_start(struct pmc *pm); static int pmc_stop(struct pmc *pm); static int pmc_syscall_handler(struct thread *td, void *syscall_args); static void pmc_unlink_target_process(struct pmc *pmc, struct pmc_process *pp); static int generic_switch_in(struct pmc_cpu *pc, struct pmc_process *pp); static int generic_switch_out(struct pmc_cpu *pc, struct pmc_process *pp); static struct pmc_mdep *pmc_generic_cpu_initialize(void); static void pmc_generic_cpu_finalize(struct pmc_mdep *md); /* * Kernel tunables and sysctl(8) interface. */ SYSCTL_DECL(_kern_hwpmc); static int pmc_callchaindepth = PMC_CALLCHAIN_DEPTH; SYSCTL_INT(_kern_hwpmc, OID_AUTO, callchaindepth, CTLFLAG_RDTUN, &pmc_callchaindepth, 0, "depth of call chain records"); #ifdef HWPMC_DEBUG struct pmc_debugflags pmc_debugflags = PMC_DEBUG_DEFAULT_FLAGS; char pmc_debugstr[PMC_DEBUG_STRSIZE]; TUNABLE_STR(PMC_SYSCTL_NAME_PREFIX "debugflags", pmc_debugstr, sizeof(pmc_debugstr)); SYSCTL_PROC(_kern_hwpmc, OID_AUTO, debugflags, CTLTYPE_STRING | CTLFLAG_RWTUN | CTLFLAG_NOFETCH, 0, 0, pmc_debugflags_sysctl_handler, "A", "debug flags"); #endif /* * kern.hwpmc.hashrows -- determines the number of rows in the * of the hash table used to look up threads */ static int pmc_hashsize = PMC_HASH_SIZE; SYSCTL_INT(_kern_hwpmc, OID_AUTO, hashsize, CTLFLAG_RDTUN, &pmc_hashsize, 0, "rows in hash tables"); /* * kern.hwpmc.nsamples --- number of PC samples/callchain stacks per CPU */ static int pmc_nsamples = PMC_NSAMPLES; SYSCTL_INT(_kern_hwpmc, OID_AUTO, nsamples, CTLFLAG_RDTUN, &pmc_nsamples, 0, "number of PC samples per CPU"); /* * kern.hwpmc.mtxpoolsize -- number of mutexes in the mutex pool. */ static int pmc_mtxpool_size = PMC_MTXPOOL_SIZE; SYSCTL_INT(_kern_hwpmc, OID_AUTO, mtxpoolsize, CTLFLAG_RDTUN, &pmc_mtxpool_size, 0, "size of spin mutex pool"); /* * security.bsd.unprivileged_syspmcs -- allow non-root processes to * allocate system-wide PMCs. * * Allowing unprivileged processes to allocate system PMCs is convenient * if system-wide measurements need to be taken concurrently with other * per-process measurements. This feature is turned off by default. */ static int pmc_unprivileged_syspmcs = 0; SYSCTL_INT(_security_bsd, OID_AUTO, unprivileged_syspmcs, CTLFLAG_RWTUN, &pmc_unprivileged_syspmcs, 0, "allow unprivileged process to allocate system PMCs"); /* * Hash function. Discard the lower 2 bits of the pointer since * these are always zero for our uses. The hash multiplier is * round((2^LONG_BIT) * ((sqrt(5)-1)/2)). */ #if LONG_BIT == 64 #define _PMC_HM 11400714819323198486u #elif LONG_BIT == 32 #define _PMC_HM 2654435769u #else #error Must know the size of 'long' to compile #endif #define PMC_HASH_PTR(P,M) ((((unsigned long) (P) >> 2) * _PMC_HM) & (M)) /* * Syscall structures */ /* The `sysent' for the new syscall */ static struct sysent pmc_sysent = { 2, /* sy_narg */ pmc_syscall_handler /* sy_call */ }; static struct syscall_module_data pmc_syscall_mod = { load, NULL, &pmc_syscall_num, &pmc_sysent, #if (__FreeBSD_version >= 1100000) { 0, NULL }, SY_THR_STATIC_KLD, #else { 0, NULL } #endif }; static moduledata_t pmc_mod = { PMC_MODULE_NAME, syscall_module_handler, &pmc_syscall_mod }; DECLARE_MODULE(pmc, pmc_mod, SI_SUB_SMP, SI_ORDER_ANY); MODULE_VERSION(pmc, PMC_VERSION); #ifdef HWPMC_DEBUG enum pmc_dbgparse_state { PMCDS_WS, /* in whitespace */ PMCDS_MAJOR, /* seen a major keyword */ PMCDS_MINOR }; static int pmc_debugflags_parse(char *newstr, char *fence) { char c, *p, *q; struct pmc_debugflags *tmpflags; int error, found, *newbits, tmp; size_t kwlen; tmpflags = malloc(sizeof(*tmpflags), M_PMC, M_WAITOK|M_ZERO); p = newstr; error = 0; for (; p < fence && (c = *p); p++) { /* skip white space */ if (c == ' ' || c == '\t') continue; /* look for a keyword followed by "=" */ for (q = p; p < fence && (c = *p) && c != '='; p++) ; if (c != '=') { error = EINVAL; goto done; } kwlen = p - q; newbits = NULL; /* lookup flag group name */ #define DBG_SET_FLAG_MAJ(S,F) \ if (kwlen == sizeof(S)-1 && strncmp(q, S, kwlen) == 0) \ newbits = &tmpflags->pdb_ ## F; DBG_SET_FLAG_MAJ("cpu", CPU); DBG_SET_FLAG_MAJ("csw", CSW); DBG_SET_FLAG_MAJ("logging", LOG); DBG_SET_FLAG_MAJ("module", MOD); DBG_SET_FLAG_MAJ("md", MDP); DBG_SET_FLAG_MAJ("owner", OWN); DBG_SET_FLAG_MAJ("pmc", PMC); DBG_SET_FLAG_MAJ("process", PRC); DBG_SET_FLAG_MAJ("sampling", SAM); if (newbits == NULL) { error = EINVAL; goto done; } p++; /* skip the '=' */ /* Now parse the individual flags */ tmp = 0; newflag: for (q = p; p < fence && (c = *p); p++) if (c == ' ' || c == '\t' || c == ',') break; /* p == fence or c == ws or c == "," or c == 0 */ if ((kwlen = p - q) == 0) { *newbits = tmp; continue; } found = 0; #define DBG_SET_FLAG_MIN(S,F) \ if (kwlen == sizeof(S)-1 && strncmp(q, S, kwlen) == 0) \ tmp |= found = (1 << PMC_DEBUG_MIN_ ## F) /* a '*' denotes all possible flags in the group */ if (kwlen == 1 && *q == '*') tmp = found = ~0; /* look for individual flag names */ DBG_SET_FLAG_MIN("allocaterow", ALR); DBG_SET_FLAG_MIN("allocate", ALL); DBG_SET_FLAG_MIN("attach", ATT); DBG_SET_FLAG_MIN("bind", BND); DBG_SET_FLAG_MIN("config", CFG); DBG_SET_FLAG_MIN("exec", EXC); DBG_SET_FLAG_MIN("exit", EXT); DBG_SET_FLAG_MIN("find", FND); DBG_SET_FLAG_MIN("flush", FLS); DBG_SET_FLAG_MIN("fork", FRK); DBG_SET_FLAG_MIN("getbuf", GTB); DBG_SET_FLAG_MIN("hook", PMH); DBG_SET_FLAG_MIN("init", INI); DBG_SET_FLAG_MIN("intr", INT); DBG_SET_FLAG_MIN("linktarget", TLK); DBG_SET_FLAG_MIN("mayberemove", OMR); DBG_SET_FLAG_MIN("ops", OPS); DBG_SET_FLAG_MIN("read", REA); DBG_SET_FLAG_MIN("register", REG); DBG_SET_FLAG_MIN("release", REL); DBG_SET_FLAG_MIN("remove", ORM); DBG_SET_FLAG_MIN("sample", SAM); DBG_SET_FLAG_MIN("scheduleio", SIO); DBG_SET_FLAG_MIN("select", SEL); DBG_SET_FLAG_MIN("signal", SIG); DBG_SET_FLAG_MIN("swi", SWI); DBG_SET_FLAG_MIN("swo", SWO); DBG_SET_FLAG_MIN("start", STA); DBG_SET_FLAG_MIN("stop", STO); DBG_SET_FLAG_MIN("syscall", PMS); DBG_SET_FLAG_MIN("unlinktarget", TUL); DBG_SET_FLAG_MIN("write", WRI); if (found == 0) { /* unrecognized flag name */ error = EINVAL; goto done; } if (c == 0 || c == ' ' || c == '\t') { /* end of flag group */ *newbits = tmp; continue; } p++; goto newflag; } /* save the new flag set */ bcopy(tmpflags, &pmc_debugflags, sizeof(pmc_debugflags)); done: free(tmpflags, M_PMC); return error; } static int pmc_debugflags_sysctl_handler(SYSCTL_HANDLER_ARGS) { char *fence, *newstr; int error; unsigned int n; (void) arg1; (void) arg2; /* unused parameters */ n = sizeof(pmc_debugstr); newstr = malloc(n, M_PMC, M_WAITOK|M_ZERO); (void) strlcpy(newstr, pmc_debugstr, n); error = sysctl_handle_string(oidp, newstr, n, req); /* if there is a new string, parse and copy it */ if (error == 0 && req->newptr != NULL) { fence = newstr + (n < req->newlen ? n : req->newlen + 1); if ((error = pmc_debugflags_parse(newstr, fence)) == 0) (void) strlcpy(pmc_debugstr, newstr, sizeof(pmc_debugstr)); } free(newstr, M_PMC); return error; } #endif /* * Map a row index to a classdep structure and return the adjusted row * index for the PMC class index. */ static struct pmc_classdep * pmc_ri_to_classdep(struct pmc_mdep *md, int ri, int *adjri) { struct pmc_classdep *pcd; (void) md; KASSERT(ri >= 0 && ri < md->pmd_npmc, ("[pmc,%d] illegal row-index %d", __LINE__, ri)); pcd = pmc_rowindex_to_classdep[ri]; KASSERT(pcd != NULL, ("[pmc,%d] ri %d null pcd", __LINE__, ri)); *adjri = ri - pcd->pcd_ri; KASSERT(*adjri >= 0 && *adjri < pcd->pcd_num, ("[pmc,%d] adjusted row-index %d", __LINE__, *adjri)); return (pcd); } /* * Concurrency Control * * The driver manages the following data structures: * * - target process descriptors, one per target process * - owner process descriptors (and attached lists), one per owner process * - lookup hash tables for owner and target processes * - PMC descriptors (and attached lists) * - per-cpu hardware state * - the 'hook' variable through which the kernel calls into * this module * - the machine hardware state (managed by the MD layer) * * These data structures are accessed from: * * - thread context-switch code * - interrupt handlers (possibly on multiple cpus) * - kernel threads on multiple cpus running on behalf of user * processes doing system calls * - this driver's private kernel threads * * = Locks and Locking strategy = * * The driver uses four locking strategies for its operation: * * - The global SX lock "pmc_sx" is used to protect internal * data structures. * * Calls into the module by syscall() start with this lock being * held in exclusive mode. Depending on the requested operation, * the lock may be downgraded to 'shared' mode to allow more * concurrent readers into the module. Calls into the module from * other parts of the kernel acquire the lock in shared mode. * * This SX lock is held in exclusive mode for any operations that * modify the linkages between the driver's internal data structures. * * The 'pmc_hook' function pointer is also protected by this lock. * It is only examined with the sx lock held in exclusive mode. The * kernel module is allowed to be unloaded only with the sx lock held * in exclusive mode. In normal syscall handling, after acquiring the * pmc_sx lock we first check that 'pmc_hook' is non-null before * proceeding. This prevents races between the thread unloading the module * and other threads seeking to use the module. * * - Lookups of target process structures and owner process structures * cannot use the global "pmc_sx" SX lock because these lookups need * to happen during context switches and in other critical sections * where sleeping is not allowed. We protect these lookup tables * with their own private spin-mutexes, "pmc_processhash_mtx" and * "pmc_ownerhash_mtx". * * - Interrupt handlers work in a lock free manner. At interrupt * time, handlers look at the PMC pointer (phw->phw_pmc) configured * when the PMC was started. If this pointer is NULL, the interrupt * is ignored after updating driver statistics. We ensure that this * pointer is set (using an atomic operation if necessary) before the * PMC hardware is started. Conversely, this pointer is unset atomically * only after the PMC hardware is stopped. * * We ensure that everything needed for the operation of an * interrupt handler is available without it needing to acquire any * locks. We also ensure that a PMC's software state is destroyed only * after the PMC is taken off hardware (on all CPUs). * * - Context-switch handling with process-private PMCs needs more * care. * * A given process may be the target of multiple PMCs. For example, * PMCATTACH and PMCDETACH may be requested by a process on one CPU * while the target process is running on another. A PMC could also * be getting released because its owner is exiting. We tackle * these situations in the following manner: * * - each target process structure 'pmc_process' has an array * of 'struct pmc *' pointers, one for each hardware PMC. * * - At context switch IN time, each "target" PMC in RUNNING state * gets started on hardware and a pointer to each PMC is copied into * the per-cpu phw array. The 'runcount' for the PMC is * incremented. * * - At context switch OUT time, all process-virtual PMCs are stopped * on hardware. The saved value is added to the PMCs value field * only if the PMC is in a non-deleted state (the PMCs state could * have changed during the current time slice). * * Note that since in-between a switch IN on a processor and a switch * OUT, the PMC could have been released on another CPU. Therefore * context switch OUT always looks at the hardware state to turn * OFF PMCs and will update a PMC's saved value only if reachable * from the target process record. * * - OP PMCRELEASE could be called on a PMC at any time (the PMC could * be attached to many processes at the time of the call and could * be active on multiple CPUs). * * We prevent further scheduling of the PMC by marking it as in * state 'DELETED'. If the runcount of the PMC is non-zero then * this PMC is currently running on a CPU somewhere. The thread * doing the PMCRELEASE operation waits by repeatedly doing a * pause() till the runcount comes to zero. * * The contents of a PMC descriptor (struct pmc) are protected using * a spin-mutex. In order to save space, we use a mutex pool. * * In terms of lock types used by witness(4), we use: * - Type "pmc-sx", used by the global SX lock. * - Type "pmc-sleep", for sleep mutexes used by logger threads. * - Type "pmc-per-proc", for protecting PMC owner descriptors. * - Type "pmc-leaf", used for all other spin mutexes. */ /* * save the cpu binding of the current kthread */ static void pmc_save_cpu_binding(struct pmc_binding *pb) { PMCDBG0(CPU,BND,2, "save-cpu"); thread_lock(curthread); pb->pb_bound = sched_is_bound(curthread); pb->pb_cpu = curthread->td_oncpu; thread_unlock(curthread); PMCDBG1(CPU,BND,2, "save-cpu cpu=%d", pb->pb_cpu); } /* * restore the cpu binding of the current thread */ static void pmc_restore_cpu_binding(struct pmc_binding *pb) { PMCDBG2(CPU,BND,2, "restore-cpu curcpu=%d restore=%d", curthread->td_oncpu, pb->pb_cpu); thread_lock(curthread); if (pb->pb_bound) sched_bind(curthread, pb->pb_cpu); else sched_unbind(curthread); thread_unlock(curthread); PMCDBG0(CPU,BND,2, "restore-cpu done"); } /* * move execution over the specified cpu and bind it there. */ static void pmc_select_cpu(int cpu) { KASSERT(cpu >= 0 && cpu < pmc_cpu_max(), ("[pmc,%d] bad cpu number %d", __LINE__, cpu)); /* Never move to an inactive CPU. */ KASSERT(pmc_cpu_is_active(cpu), ("[pmc,%d] selecting inactive " "CPU %d", __LINE__, cpu)); PMCDBG1(CPU,SEL,2, "select-cpu cpu=%d", cpu); thread_lock(curthread); sched_bind(curthread, cpu); thread_unlock(curthread); KASSERT(curthread->td_oncpu == cpu, ("[pmc,%d] CPU not bound [cpu=%d, curr=%d]", __LINE__, cpu, curthread->td_oncpu)); PMCDBG1(CPU,SEL,2, "select-cpu cpu=%d ok", cpu); } /* * Force a context switch. * * We do this by pause'ing for 1 tick -- invoking mi_switch() is not * guaranteed to force a context switch. */ static void pmc_force_context_switch(void) { pause("pmcctx", 1); } /* * Get the file name for an executable. This is a simple wrapper * around vn_fullpath(9). */ static void pmc_getfilename(struct vnode *v, char **fullpath, char **freepath) { *fullpath = "unknown"; *freepath = NULL; vn_fullpath(curthread, v, fullpath, freepath); } /* * remove an process owning PMCs */ void pmc_remove_owner(struct pmc_owner *po) { struct pmc *pm, *tmp; sx_assert(&pmc_sx, SX_XLOCKED); PMCDBG1(OWN,ORM,1, "remove-owner po=%p", po); /* Remove descriptor from the owner hash table */ LIST_REMOVE(po, po_next); /* release all owned PMC descriptors */ LIST_FOREACH_SAFE(pm, &po->po_pmcs, pm_next, tmp) { PMCDBG1(OWN,ORM,2, "pmc=%p", pm); KASSERT(pm->pm_owner == po, ("[pmc,%d] owner %p != po %p", __LINE__, pm->pm_owner, po)); pmc_release_pmc_descriptor(pm); /* will unlink from the list */ pmc_destroy_pmc_descriptor(pm); } KASSERT(po->po_sscount == 0, ("[pmc,%d] SS count not zero", __LINE__)); KASSERT(LIST_EMPTY(&po->po_pmcs), ("[pmc,%d] PMC list not empty", __LINE__)); /* de-configure the log file if present */ if (po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_deconfigure_log(po); } /* * remove an owner process record if all conditions are met. */ static void pmc_maybe_remove_owner(struct pmc_owner *po) { PMCDBG1(OWN,OMR,1, "maybe-remove-owner po=%p", po); /* * Remove owner record if * - this process does not own any PMCs * - this process has not allocated a system-wide sampling buffer */ if (LIST_EMPTY(&po->po_pmcs) && ((po->po_flags & PMC_PO_OWNS_LOGFILE) == 0)) { pmc_remove_owner(po); pmc_destroy_owner_descriptor(po); } } /* * Add an association between a target process and a PMC. */ static void pmc_link_target_process(struct pmc *pm, struct pmc_process *pp) { int ri; struct pmc_target *pt; sx_assert(&pmc_sx, SX_XLOCKED); KASSERT(pm != NULL && pp != NULL, ("[pmc,%d] Null pm %p or pp %p", __LINE__, pm, pp)); KASSERT(PMC_IS_VIRTUAL_MODE(PMC_TO_MODE(pm)), ("[pmc,%d] Attaching a non-process-virtual pmc=%p to pid=%d", __LINE__, pm, pp->pp_proc->p_pid)); KASSERT(pp->pp_refcnt >= 0 && pp->pp_refcnt <= ((int) md->pmd_npmc - 1), ("[pmc,%d] Illegal reference count %d for process record %p", __LINE__, pp->pp_refcnt, (void *) pp)); ri = PMC_TO_ROWINDEX(pm); PMCDBG3(PRC,TLK,1, "link-target pmc=%p ri=%d pmc-process=%p", pm, ri, pp); #ifdef HWPMC_DEBUG LIST_FOREACH(pt, &pm->pm_targets, pt_next) if (pt->pt_process == pp) KASSERT(0, ("[pmc,%d] pp %p already in pmc %p targets", __LINE__, pp, pm)); #endif pt = malloc(sizeof(struct pmc_target), M_PMC, M_WAITOK|M_ZERO); pt->pt_process = pp; LIST_INSERT_HEAD(&pm->pm_targets, pt, pt_next); atomic_store_rel_ptr((uintptr_t *)&pp->pp_pmcs[ri].pp_pmc, (uintptr_t)pm); if (pm->pm_owner->po_owner == pp->pp_proc) pm->pm_flags |= PMC_F_ATTACHED_TO_OWNER; /* * Initialize the per-process values at this row index. */ pp->pp_pmcs[ri].pp_pmcval = PMC_TO_MODE(pm) == PMC_MODE_TS ? pm->pm_sc.pm_reloadcount : 0; pp->pp_refcnt++; } /* * Removes the association between a target process and a PMC. */ static void pmc_unlink_target_process(struct pmc *pm, struct pmc_process *pp) { int ri; struct proc *p; struct pmc_target *ptgt; sx_assert(&pmc_sx, SX_XLOCKED); KASSERT(pm != NULL && pp != NULL, ("[pmc,%d] Null pm %p or pp %p", __LINE__, pm, pp)); KASSERT(pp->pp_refcnt >= 1 && pp->pp_refcnt <= (int) md->pmd_npmc, ("[pmc,%d] Illegal ref count %d on process record %p", __LINE__, pp->pp_refcnt, (void *) pp)); ri = PMC_TO_ROWINDEX(pm); PMCDBG3(PRC,TUL,1, "unlink-target pmc=%p ri=%d pmc-process=%p", pm, ri, pp); KASSERT(pp->pp_pmcs[ri].pp_pmc == pm, ("[pmc,%d] PMC ri %d mismatch pmc %p pp->[ri] %p", __LINE__, ri, pm, pp->pp_pmcs[ri].pp_pmc)); pp->pp_pmcs[ri].pp_pmc = NULL; pp->pp_pmcs[ri].pp_pmcval = (pmc_value_t) 0; /* Remove owner-specific flags */ if (pm->pm_owner->po_owner == pp->pp_proc) { pp->pp_flags &= ~PMC_PP_ENABLE_MSR_ACCESS; pm->pm_flags &= ~PMC_F_ATTACHED_TO_OWNER; } pp->pp_refcnt--; /* Remove the target process from the PMC structure */ LIST_FOREACH(ptgt, &pm->pm_targets, pt_next) if (ptgt->pt_process == pp) break; KASSERT(ptgt != NULL, ("[pmc,%d] process %p (pp: %p) not found " "in pmc %p", __LINE__, pp->pp_proc, pp, pm)); LIST_REMOVE(ptgt, pt_next); free(ptgt, M_PMC); /* if the PMC now lacks targets, send the owner a SIGIO */ if (LIST_EMPTY(&pm->pm_targets)) { p = pm->pm_owner->po_owner; PROC_LOCK(p); kern_psignal(p, SIGIO); PROC_UNLOCK(p); PMCDBG2(PRC,SIG,2, "signalling proc=%p signal=%d", p, SIGIO); } } /* * Check if PMC 'pm' may be attached to target process 't'. */ static int pmc_can_attach(struct pmc *pm, struct proc *t) { struct proc *o; /* pmc owner */ struct ucred *oc, *tc; /* owner, target credentials */ int decline_attach, i; /* * A PMC's owner can always attach that PMC to itself. */ if ((o = pm->pm_owner->po_owner) == t) return 0; PROC_LOCK(o); oc = o->p_ucred; crhold(oc); PROC_UNLOCK(o); PROC_LOCK(t); tc = t->p_ucred; crhold(tc); PROC_UNLOCK(t); /* * The effective uid of the PMC owner should match at least one * of the {effective,real,saved} uids of the target process. */ decline_attach = oc->cr_uid != tc->cr_uid && oc->cr_uid != tc->cr_svuid && oc->cr_uid != tc->cr_ruid; /* * Every one of the target's group ids, must be in the owner's * group list. */ for (i = 0; !decline_attach && i < tc->cr_ngroups; i++) decline_attach = !groupmember(tc->cr_groups[i], oc); /* check the read and saved gids too */ if (decline_attach == 0) decline_attach = !groupmember(tc->cr_rgid, oc) || !groupmember(tc->cr_svgid, oc); crfree(tc); crfree(oc); return !decline_attach; } /* * Attach a process to a PMC. */ static int pmc_attach_one_process(struct proc *p, struct pmc *pm) { int ri; char *fullpath, *freepath; struct pmc_process *pp; sx_assert(&pmc_sx, SX_XLOCKED); PMCDBG5(PRC,ATT,2, "attach-one pm=%p ri=%d proc=%p (%d, %s)", pm, PMC_TO_ROWINDEX(pm), p, p->p_pid, p->p_comm); /* * Locate the process descriptor corresponding to process 'p', * allocating space as needed. * * Verify that rowindex 'pm_rowindex' is free in the process * descriptor. * * If not, allocate space for a descriptor and link the * process descriptor and PMC. */ ri = PMC_TO_ROWINDEX(pm); if ((pp = pmc_find_process_descriptor(p, PMC_FLAG_ALLOCATE)) == NULL) return ENOMEM; if (pp->pp_pmcs[ri].pp_pmc == pm) /* already present at slot [ri] */ return EEXIST; if (pp->pp_pmcs[ri].pp_pmc != NULL) return EBUSY; pmc_link_target_process(pm, pp); if (PMC_IS_SAMPLING_MODE(PMC_TO_MODE(pm)) && (pm->pm_flags & PMC_F_ATTACHED_TO_OWNER) == 0) pm->pm_flags |= PMC_F_NEEDS_LOGFILE; pm->pm_flags |= PMC_F_ATTACH_DONE; /* mark as attached */ /* issue an attach event to a configured log file */ if (pm->pm_owner->po_flags & PMC_PO_OWNS_LOGFILE) { pmc_getfilename(p->p_textvp, &fullpath, &freepath); if (p->p_flag & P_KTHREAD) { fullpath = kernelname; freepath = NULL; } else pmclog_process_pmcattach(pm, p->p_pid, fullpath); if (freepath) free(freepath, M_TEMP); if (PMC_IS_SAMPLING_MODE(PMC_TO_MODE(pm))) pmc_log_process_mappings(pm->pm_owner, p); } /* mark process as using HWPMCs */ PROC_LOCK(p); p->p_flag |= P_HWPMC; PROC_UNLOCK(p); return 0; } /* * Attach a process and optionally its children */ static int pmc_attach_process(struct proc *p, struct pmc *pm) { int error; struct proc *top; sx_assert(&pmc_sx, SX_XLOCKED); PMCDBG5(PRC,ATT,1, "attach pm=%p ri=%d proc=%p (%d, %s)", pm, PMC_TO_ROWINDEX(pm), p, p->p_pid, p->p_comm); /* * If this PMC successfully allowed a GETMSR operation * in the past, disallow further ATTACHes. */ if ((pm->pm_flags & PMC_PP_ENABLE_MSR_ACCESS) != 0) return EPERM; if ((pm->pm_flags & PMC_F_DESCENDANTS) == 0) return pmc_attach_one_process(p, pm); /* * Traverse all child processes, attaching them to * this PMC. */ sx_slock(&proctree_lock); top = p; for (;;) { if ((error = pmc_attach_one_process(p, pm)) != 0) break; if (!LIST_EMPTY(&p->p_children)) p = LIST_FIRST(&p->p_children); else for (;;) { if (p == top) goto done; if (LIST_NEXT(p, p_sibling)) { p = LIST_NEXT(p, p_sibling); break; } p = p->p_pptr; } } if (error) (void) pmc_detach_process(top, pm); done: sx_sunlock(&proctree_lock); return error; } /* * Detach a process from a PMC. If there are no other PMCs tracking * this process, remove the process structure from its hash table. If * 'flags' contains PMC_FLAG_REMOVE, then free the process structure. */ static int pmc_detach_one_process(struct proc *p, struct pmc *pm, int flags) { int ri; struct pmc_process *pp; sx_assert(&pmc_sx, SX_XLOCKED); KASSERT(pm != NULL, ("[pmc,%d] null pm pointer", __LINE__)); ri = PMC_TO_ROWINDEX(pm); PMCDBG6(PRC,ATT,2, "detach-one pm=%p ri=%d proc=%p (%d, %s) flags=0x%x", pm, ri, p, p->p_pid, p->p_comm, flags); if ((pp = pmc_find_process_descriptor(p, 0)) == NULL) return ESRCH; if (pp->pp_pmcs[ri].pp_pmc != pm) return EINVAL; pmc_unlink_target_process(pm, pp); /* Issue a detach entry if a log file is configured */ if (pm->pm_owner->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_pmcdetach(pm, p->p_pid); /* * If there are no PMCs targetting this process, we remove its * descriptor from the target hash table and unset the P_HWPMC * flag in the struct proc. */ KASSERT(pp->pp_refcnt >= 0 && pp->pp_refcnt <= (int) md->pmd_npmc, ("[pmc,%d] Illegal refcnt %d for process struct %p", __LINE__, pp->pp_refcnt, pp)); if (pp->pp_refcnt != 0) /* still a target of some PMC */ return 0; pmc_remove_process_descriptor(pp); if (flags & PMC_FLAG_REMOVE) free(pp, M_PMC); PROC_LOCK(p); p->p_flag &= ~P_HWPMC; PROC_UNLOCK(p); return 0; } /* * Detach a process and optionally its descendants from a PMC. */ static int pmc_detach_process(struct proc *p, struct pmc *pm) { struct proc *top; sx_assert(&pmc_sx, SX_XLOCKED); PMCDBG5(PRC,ATT,1, "detach pm=%p ri=%d proc=%p (%d, %s)", pm, PMC_TO_ROWINDEX(pm), p, p->p_pid, p->p_comm); if ((pm->pm_flags & PMC_F_DESCENDANTS) == 0) return pmc_detach_one_process(p, pm, PMC_FLAG_REMOVE); /* * Traverse all children, detaching them from this PMC. We * ignore errors since we could be detaching a PMC from a * partially attached proc tree. */ sx_slock(&proctree_lock); top = p; for (;;) { (void) pmc_detach_one_process(p, pm, PMC_FLAG_REMOVE); if (!LIST_EMPTY(&p->p_children)) p = LIST_FIRST(&p->p_children); else for (;;) { if (p == top) goto done; if (LIST_NEXT(p, p_sibling)) { p = LIST_NEXT(p, p_sibling); break; } p = p->p_pptr; } } done: sx_sunlock(&proctree_lock); if (LIST_EMPTY(&pm->pm_targets)) pm->pm_flags &= ~PMC_F_ATTACH_DONE; return 0; } /* * Thread context switch IN */ static void pmc_process_csw_in(struct thread *td) { int cpu; unsigned int adjri, ri; struct pmc *pm; struct proc *p; struct pmc_cpu *pc; struct pmc_hw *phw; pmc_value_t newvalue; struct pmc_process *pp; struct pmc_classdep *pcd; p = td->td_proc; if ((pp = pmc_find_process_descriptor(p, PMC_FLAG_NONE)) == NULL) return; KASSERT(pp->pp_proc == td->td_proc, ("[pmc,%d] not my thread state", __LINE__)); critical_enter(); /* no preemption from this point */ cpu = PCPU_GET(cpuid); /* td->td_oncpu is invalid */ PMCDBG5(CSW,SWI,1, "cpu=%d proc=%p (%d, %s) pp=%p", cpu, p, p->p_pid, p->p_comm, pp); KASSERT(cpu >= 0 && cpu < pmc_cpu_max(), ("[pmc,%d] wierd CPU id %d", __LINE__, cpu)); pc = pmc_pcpu[cpu]; for (ri = 0; ri < md->pmd_npmc; ri++) { if ((pm = pp->pp_pmcs[ri].pp_pmc) == NULL) continue; KASSERT(PMC_IS_VIRTUAL_MODE(PMC_TO_MODE(pm)), ("[pmc,%d] Target PMC in non-virtual mode (%d)", __LINE__, PMC_TO_MODE(pm))); KASSERT(PMC_TO_ROWINDEX(pm) == ri, ("[pmc,%d] Row index mismatch pmc %d != ri %d", __LINE__, PMC_TO_ROWINDEX(pm), ri)); /* * Only PMCs that are marked as 'RUNNING' need * be placed on hardware. */ if (pm->pm_state != PMC_STATE_RUNNING) continue; /* increment PMC runcount */ atomic_add_rel_int(&pm->pm_runcount, 1); /* configure the HWPMC we are going to use. */ pcd = pmc_ri_to_classdep(md, ri, &adjri); pcd->pcd_config_pmc(cpu, adjri, pm); phw = pc->pc_hwpmcs[ri]; KASSERT(phw != NULL, ("[pmc,%d] null hw pointer", __LINE__)); KASSERT(phw->phw_pmc == pm, ("[pmc,%d] hw->pmc %p != pmc %p", __LINE__, phw->phw_pmc, pm)); /* * Write out saved value and start the PMC. * * Sampling PMCs use a per-process value, while * counting mode PMCs use a per-pmc value that is * inherited across descendants. */ if (PMC_TO_MODE(pm) == PMC_MODE_TS) { mtx_pool_lock_spin(pmc_mtxpool, pm); newvalue = PMC_PCPU_SAVED(cpu,ri) = pp->pp_pmcs[ri].pp_pmcval; mtx_pool_unlock_spin(pmc_mtxpool, pm); } else { KASSERT(PMC_TO_MODE(pm) == PMC_MODE_TC, ("[pmc,%d] illegal mode=%d", __LINE__, PMC_TO_MODE(pm))); mtx_pool_lock_spin(pmc_mtxpool, pm); newvalue = PMC_PCPU_SAVED(cpu, ri) = pm->pm_gv.pm_savedvalue; mtx_pool_unlock_spin(pmc_mtxpool, pm); } PMCDBG3(CSW,SWI,1,"cpu=%d ri=%d new=%jd", cpu, ri, newvalue); pcd->pcd_write_pmc(cpu, adjri, newvalue); pcd->pcd_start_pmc(cpu, adjri); } /* * perform any other architecture/cpu dependent thread * switch-in actions. */ (void) (*md->pmd_switch_in)(pc, pp); critical_exit(); } /* * Thread context switch OUT. */ static void pmc_process_csw_out(struct thread *td) { int cpu; int64_t tmp; struct pmc *pm; struct proc *p; enum pmc_mode mode; struct pmc_cpu *pc; pmc_value_t newvalue; unsigned int adjri, ri; struct pmc_process *pp; struct pmc_classdep *pcd; /* * Locate our process descriptor; this may be NULL if * this process is exiting and we have already removed * the process from the target process table. * * Note that due to kernel preemption, multiple * context switches may happen while the process is * exiting. * * Note also that if the target process cannot be * found we still need to deconfigure any PMCs that * are currently running on hardware. */ p = td->td_proc; pp = pmc_find_process_descriptor(p, PMC_FLAG_NONE); /* * save PMCs */ critical_enter(); cpu = PCPU_GET(cpuid); /* td->td_oncpu is invalid */ PMCDBG5(CSW,SWO,1, "cpu=%d proc=%p (%d, %s) pp=%p", cpu, p, p->p_pid, p->p_comm, pp); KASSERT(cpu >= 0 && cpu < pmc_cpu_max(), ("[pmc,%d wierd CPU id %d", __LINE__, cpu)); pc = pmc_pcpu[cpu]; /* * When a PMC gets unlinked from a target PMC, it will * be removed from the target's pp_pmc[] array. * * However, on a MP system, the target could have been * executing on another CPU at the time of the unlink. * So, at context switch OUT time, we need to look at * the hardware to determine if a PMC is scheduled on * it. */ for (ri = 0; ri < md->pmd_npmc; ri++) { pcd = pmc_ri_to_classdep(md, ri, &adjri); pm = NULL; (void) (*pcd->pcd_get_config)(cpu, adjri, &pm); if (pm == NULL) /* nothing at this row index */ continue; mode = PMC_TO_MODE(pm); if (!PMC_IS_VIRTUAL_MODE(mode)) continue; /* not a process virtual PMC */ KASSERT(PMC_TO_ROWINDEX(pm) == ri, ("[pmc,%d] ri mismatch pmc(%d) ri(%d)", __LINE__, PMC_TO_ROWINDEX(pm), ri)); /* Stop hardware if not already stopped */ if (pm->pm_stalled == 0) pcd->pcd_stop_pmc(cpu, adjri); /* reduce this PMC's runcount */ atomic_subtract_rel_int(&pm->pm_runcount, 1); /* * If this PMC is associated with this process, * save the reading. */ if (pp != NULL && pp->pp_pmcs[ri].pp_pmc != NULL) { KASSERT(pm == pp->pp_pmcs[ri].pp_pmc, ("[pmc,%d] pm %p != pp_pmcs[%d] %p", __LINE__, pm, ri, pp->pp_pmcs[ri].pp_pmc)); KASSERT(pp->pp_refcnt > 0, ("[pmc,%d] pp refcnt = %d", __LINE__, pp->pp_refcnt)); pcd->pcd_read_pmc(cpu, adjri, &newvalue); tmp = newvalue - PMC_PCPU_SAVED(cpu,ri); PMCDBG3(CSW,SWO,1,"cpu=%d ri=%d tmp=%jd", cpu, ri, tmp); if (mode == PMC_MODE_TS) { /* * For sampling process-virtual PMCs, * we expect the count to be * decreasing as the 'value' * programmed into the PMC is the * number of events to be seen till * the next sampling interrupt. */ if (tmp < 0) tmp += pm->pm_sc.pm_reloadcount; mtx_pool_lock_spin(pmc_mtxpool, pm); pp->pp_pmcs[ri].pp_pmcval -= tmp; if ((int64_t) pp->pp_pmcs[ri].pp_pmcval <= 0) pp->pp_pmcs[ri].pp_pmcval += pm->pm_sc.pm_reloadcount; mtx_pool_unlock_spin(pmc_mtxpool, pm); } else { /* * For counting process-virtual PMCs, * we expect the count to be * increasing monotonically, modulo a 64 * bit wraparound. */ KASSERT((int64_t) tmp >= 0, ("[pmc,%d] negative increment cpu=%d " "ri=%d newvalue=%jx saved=%jx " "incr=%jx", __LINE__, cpu, ri, newvalue, PMC_PCPU_SAVED(cpu,ri), tmp)); mtx_pool_lock_spin(pmc_mtxpool, pm); pm->pm_gv.pm_savedvalue += tmp; pp->pp_pmcs[ri].pp_pmcval += tmp; mtx_pool_unlock_spin(pmc_mtxpool, pm); if (pm->pm_flags & PMC_F_LOG_PROCCSW) pmclog_process_proccsw(pm, pp, tmp); } } /* mark hardware as free */ pcd->pcd_config_pmc(cpu, adjri, NULL); } /* * perform any other architecture/cpu dependent thread * switch out functions. */ (void) (*md->pmd_switch_out)(pc, pp); critical_exit(); } /* * A mapping change for a process. */ static void pmc_process_mmap(struct thread *td, struct pmckern_map_in *pkm) { int ri; pid_t pid; char *fullpath, *freepath; const struct pmc *pm; struct pmc_owner *po; const struct pmc_process *pp; freepath = fullpath = NULL; pmc_getfilename((struct vnode *) pkm->pm_file, &fullpath, &freepath); pid = td->td_proc->p_pid; /* Inform owners of all system-wide sampling PMCs. */ LIST_FOREACH(po, &pmc_ss_owners, po_ssnext) if (po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_map_in(po, pid, pkm->pm_address, fullpath); if ((pp = pmc_find_process_descriptor(td->td_proc, 0)) == NULL) goto done; /* * Inform sampling PMC owners tracking this process. */ for (ri = 0; ri < md->pmd_npmc; ri++) if ((pm = pp->pp_pmcs[ri].pp_pmc) != NULL && PMC_IS_SAMPLING_MODE(PMC_TO_MODE(pm))) pmclog_process_map_in(pm->pm_owner, pid, pkm->pm_address, fullpath); done: if (freepath) free(freepath, M_TEMP); } /* * Log an munmap request. */ static void pmc_process_munmap(struct thread *td, struct pmckern_map_out *pkm) { int ri; pid_t pid; struct pmc_owner *po; const struct pmc *pm; const struct pmc_process *pp; pid = td->td_proc->p_pid; LIST_FOREACH(po, &pmc_ss_owners, po_ssnext) if (po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_map_out(po, pid, pkm->pm_address, pkm->pm_address + pkm->pm_size); if ((pp = pmc_find_process_descriptor(td->td_proc, 0)) == NULL) return; for (ri = 0; ri < md->pmd_npmc; ri++) if ((pm = pp->pp_pmcs[ri].pp_pmc) != NULL && PMC_IS_SAMPLING_MODE(PMC_TO_MODE(pm))) pmclog_process_map_out(pm->pm_owner, pid, pkm->pm_address, pkm->pm_address + pkm->pm_size); } /* * Log mapping information about the kernel. */ static void pmc_log_kernel_mappings(struct pmc *pm) { struct pmc_owner *po; struct pmckern_map_in *km, *kmbase; sx_assert(&pmc_sx, SX_LOCKED); KASSERT(PMC_IS_SAMPLING_MODE(PMC_TO_MODE(pm)), ("[pmc,%d] non-sampling PMC (%p) desires mapping information", __LINE__, (void *) pm)); po = pm->pm_owner; if (po->po_flags & PMC_PO_INITIAL_MAPPINGS_DONE) return; /* * Log the current set of kernel modules. */ kmbase = linker_hwpmc_list_objects(); for (km = kmbase; km->pm_file != NULL; km++) { PMCDBG2(LOG,REG,1,"%s %p", (char *) km->pm_file, (void *) km->pm_address); pmclog_process_map_in(po, (pid_t) -1, km->pm_address, km->pm_file); } free(kmbase, M_LINKER); po->po_flags |= PMC_PO_INITIAL_MAPPINGS_DONE; } /* * Log the mappings for a single process. */ static void pmc_log_process_mappings(struct pmc_owner *po, struct proc *p) { vm_map_t map; struct vnode *vp; struct vmspace *vm; vm_map_entry_t entry; vm_offset_t last_end; u_int last_timestamp; struct vnode *last_vp; vm_offset_t start_addr; vm_object_t obj, lobj, tobj; char *fullpath, *freepath; last_vp = NULL; last_end = (vm_offset_t) 0; fullpath = freepath = NULL; if ((vm = vmspace_acquire_ref(p)) == NULL) return; map = &vm->vm_map; vm_map_lock_read(map); for (entry = map->header.next; entry != &map->header; entry = entry->next) { if (entry == NULL) { PMCDBG2(LOG,OPS,2, "hwpmc: vm_map entry unexpectedly " "NULL! pid=%d vm_map=%p\n", p->p_pid, map); break; } /* * We only care about executable map entries. */ if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) || !(entry->protection & VM_PROT_EXECUTE) || (entry->object.vm_object == NULL)) { continue; } obj = entry->object.vm_object; VM_OBJECT_RLOCK(obj); /* * Walk the backing_object list to find the base * (non-shadowed) vm_object. */ for (lobj = tobj = obj; tobj != NULL; tobj = tobj->backing_object) { if (tobj != obj) VM_OBJECT_RLOCK(tobj); if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); lobj = tobj; } /* * At this point lobj is the base vm_object and it is locked. */ if (lobj == NULL) { PMCDBG3(LOG,OPS,2, "hwpmc: lobj unexpectedly NULL! pid=%d " "vm_map=%p vm_obj=%p\n", p->p_pid, map, obj); VM_OBJECT_RUNLOCK(obj); continue; } - if (lobj->type != OBJT_VNODE || lobj->handle == NULL) { + vp = vm_object_vnode(lobj); + if (vp == NULL) { if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); VM_OBJECT_RUNLOCK(obj); continue; } /* * Skip contiguous regions that point to the same * vnode, so we don't emit redundant MAP-IN * directives. */ - if (entry->start == last_end && lobj->handle == last_vp) { + if (entry->start == last_end && vp == last_vp) { last_end = entry->end; if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); VM_OBJECT_RUNLOCK(obj); continue; } /* * We don't want to keep the proc's vm_map or this * vm_object locked while we walk the pathname, since * vn_fullpath() can sleep. However, if we drop the * lock, it's possible for concurrent activity to * modify the vm_map list. To protect against this, * we save the vm_map timestamp before we release the * lock, and check it after we reacquire the lock * below. */ start_addr = entry->start; last_end = entry->end; last_timestamp = map->timestamp; vm_map_unlock_read(map); - vp = lobj->handle; vref(vp); if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); VM_OBJECT_RUNLOCK(obj); freepath = NULL; pmc_getfilename(vp, &fullpath, &freepath); last_vp = vp; vrele(vp); vp = NULL; pmclog_process_map_in(po, p->p_pid, start_addr, fullpath); if (freepath) free(freepath, M_TEMP); vm_map_lock_read(map); /* * If our saved timestamp doesn't match, this means * that the vm_map was modified out from under us and * we can't trust our current "entry" pointer. Do a * new lookup for this entry. If there is no entry * for this address range, vm_map_lookup_entry() will * return the previous one, so we always want to go to * entry->next on the next loop iteration. * * There is an edge condition here that can occur if * there is no entry at or before this address. In * this situation, vm_map_lookup_entry returns * &map->header, which would cause our loop to abort * without processing the rest of the map. However, * in practice this will never happen for process * vm_map. This is because the executable's text * segment is the first mapping in the proc's address * space, and this mapping is never removed until the * process exits, so there will always be a non-header * entry at or before the requested address for * vm_map_lookup_entry to return. */ if (map->timestamp != last_timestamp) vm_map_lookup_entry(map, last_end - 1, &entry); } vm_map_unlock_read(map); vmspace_free(vm); return; } /* * Log mappings for all processes in the system. */ static void pmc_log_all_process_mappings(struct pmc_owner *po) { struct proc *p, *top; sx_assert(&pmc_sx, SX_XLOCKED); if ((p = pfind(1)) == NULL) panic("[pmc,%d] Cannot find init", __LINE__); PROC_UNLOCK(p); sx_slock(&proctree_lock); top = p; for (;;) { pmc_log_process_mappings(po, p); if (!LIST_EMPTY(&p->p_children)) p = LIST_FIRST(&p->p_children); else for (;;) { if (p == top) goto done; if (LIST_NEXT(p, p_sibling)) { p = LIST_NEXT(p, p_sibling); break; } p = p->p_pptr; } } done: sx_sunlock(&proctree_lock); } /* * The 'hook' invoked from the kernel proper */ #ifdef HWPMC_DEBUG const char *pmc_hooknames[] = { /* these strings correspond to PMC_FN_* in */ "", "EXEC", "CSW-IN", "CSW-OUT", "SAMPLE", "UNUSED1", "UNUSED2", "MMAP", "MUNMAP", "CALLCHAIN-NMI", "CALLCHAIN-SOFT", "SOFTSAMPLING" }; #endif static int pmc_hook_handler(struct thread *td, int function, void *arg) { PMCDBG4(MOD,PMH,1, "hook td=%p func=%d \"%s\" arg=%p", td, function, pmc_hooknames[function], arg); switch (function) { /* * Process exec() */ case PMC_FN_PROCESS_EXEC: { char *fullpath, *freepath; unsigned int ri; int is_using_hwpmcs; struct pmc *pm; struct proc *p; struct pmc_owner *po; struct pmc_process *pp; struct pmckern_procexec *pk; sx_assert(&pmc_sx, SX_XLOCKED); p = td->td_proc; pmc_getfilename(p->p_textvp, &fullpath, &freepath); pk = (struct pmckern_procexec *) arg; /* Inform owners of SS mode PMCs of the exec event. */ LIST_FOREACH(po, &pmc_ss_owners, po_ssnext) if (po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_procexec(po, PMC_ID_INVALID, p->p_pid, pk->pm_entryaddr, fullpath); PROC_LOCK(p); is_using_hwpmcs = p->p_flag & P_HWPMC; PROC_UNLOCK(p); if (!is_using_hwpmcs) { if (freepath) free(freepath, M_TEMP); break; } /* * PMCs are not inherited across an exec(): remove any * PMCs that this process is the owner of. */ if ((po = pmc_find_owner_descriptor(p)) != NULL) { pmc_remove_owner(po); pmc_destroy_owner_descriptor(po); } /* * If the process being exec'ed is not the target of any * PMC, we are done. */ if ((pp = pmc_find_process_descriptor(p, 0)) == NULL) { if (freepath) free(freepath, M_TEMP); break; } /* * Log the exec event to all monitoring owners. Skip * owners who have already recieved the event because * they had system sampling PMCs active. */ for (ri = 0; ri < md->pmd_npmc; ri++) if ((pm = pp->pp_pmcs[ri].pp_pmc) != NULL) { po = pm->pm_owner; if (po->po_sscount == 0 && po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_procexec(po, pm->pm_id, p->p_pid, pk->pm_entryaddr, fullpath); } if (freepath) free(freepath, M_TEMP); PMCDBG4(PRC,EXC,1, "exec proc=%p (%d, %s) cred-changed=%d", p, p->p_pid, p->p_comm, pk->pm_credentialschanged); if (pk->pm_credentialschanged == 0) /* no change */ break; /* * If the newly exec()'ed process has a different credential * than before, allow it to be the target of a PMC only if * the PMC's owner has sufficient priviledge. */ for (ri = 0; ri < md->pmd_npmc; ri++) if ((pm = pp->pp_pmcs[ri].pp_pmc) != NULL) if (pmc_can_attach(pm, td->td_proc) != 0) pmc_detach_one_process(td->td_proc, pm, PMC_FLAG_NONE); KASSERT(pp->pp_refcnt >= 0 && pp->pp_refcnt <= (int) md->pmd_npmc, ("[pmc,%d] Illegal ref count %d on pp %p", __LINE__, pp->pp_refcnt, pp)); /* * If this process is no longer the target of any * PMCs, we can remove the process entry and free * up space. */ if (pp->pp_refcnt == 0) { pmc_remove_process_descriptor(pp); free(pp, M_PMC); break; } } break; case PMC_FN_CSW_IN: pmc_process_csw_in(td); break; case PMC_FN_CSW_OUT: pmc_process_csw_out(td); break; /* * Process accumulated PC samples. * * This function is expected to be called by hardclock() for * each CPU that has accumulated PC samples. * * This function is to be executed on the CPU whose samples * are being processed. */ case PMC_FN_DO_SAMPLES: /* * Clear the cpu specific bit in the CPU mask before * do the rest of the processing. If the NMI handler * gets invoked after the "atomic_clear_int()" call * below but before "pmc_process_samples()" gets * around to processing the interrupt, then we will * come back here at the next hardclock() tick (and * may find nothing to do if "pmc_process_samples()" * had already processed the interrupt). We don't * lose the interrupt sample. */ CPU_CLR_ATOMIC(PCPU_GET(cpuid), &pmc_cpumask); pmc_process_samples(PCPU_GET(cpuid), PMC_HR); pmc_process_samples(PCPU_GET(cpuid), PMC_SR); break; case PMC_FN_MMAP: sx_assert(&pmc_sx, SX_LOCKED); pmc_process_mmap(td, (struct pmckern_map_in *) arg); break; case PMC_FN_MUNMAP: sx_assert(&pmc_sx, SX_LOCKED); pmc_process_munmap(td, (struct pmckern_map_out *) arg); break; case PMC_FN_USER_CALLCHAIN: /* * Record a call chain. */ KASSERT(td == curthread, ("[pmc,%d] td != curthread", __LINE__)); pmc_capture_user_callchain(PCPU_GET(cpuid), PMC_HR, (struct trapframe *) arg); td->td_pflags &= ~TDP_CALLCHAIN; break; case PMC_FN_USER_CALLCHAIN_SOFT: /* * Record a call chain. */ KASSERT(td == curthread, ("[pmc,%d] td != curthread", __LINE__)); pmc_capture_user_callchain(PCPU_GET(cpuid), PMC_SR, (struct trapframe *) arg); td->td_pflags &= ~TDP_CALLCHAIN; break; case PMC_FN_SOFT_SAMPLING: /* * Call soft PMC sampling intr. */ pmc_soft_intr((struct pmckern_soft *) arg); break; default: #ifdef HWPMC_DEBUG KASSERT(0, ("[pmc,%d] unknown hook %d\n", __LINE__, function)); #endif break; } return 0; } /* * allocate a 'struct pmc_owner' descriptor in the owner hash table. */ static struct pmc_owner * pmc_allocate_owner_descriptor(struct proc *p) { uint32_t hindex; struct pmc_owner *po; struct pmc_ownerhash *poh; hindex = PMC_HASH_PTR(p, pmc_ownerhashmask); poh = &pmc_ownerhash[hindex]; /* allocate space for N pointers and one descriptor struct */ po = malloc(sizeof(struct pmc_owner), M_PMC, M_WAITOK|M_ZERO); po->po_owner = p; LIST_INSERT_HEAD(poh, po, po_next); /* insert into hash table */ TAILQ_INIT(&po->po_logbuffers); mtx_init(&po->po_mtx, "pmc-owner-mtx", "pmc-per-proc", MTX_SPIN); PMCDBG4(OWN,ALL,1, "allocate-owner proc=%p (%d, %s) pmc-owner=%p", p, p->p_pid, p->p_comm, po); return po; } static void pmc_destroy_owner_descriptor(struct pmc_owner *po) { PMCDBG4(OWN,REL,1, "destroy-owner po=%p proc=%p (%d, %s)", po, po->po_owner, po->po_owner->p_pid, po->po_owner->p_comm); mtx_destroy(&po->po_mtx); free(po, M_PMC); } /* * find the descriptor corresponding to process 'p', adding or removing it * as specified by 'mode'. */ static struct pmc_process * pmc_find_process_descriptor(struct proc *p, uint32_t mode) { uint32_t hindex; struct pmc_process *pp, *ppnew; struct pmc_processhash *pph; hindex = PMC_HASH_PTR(p, pmc_processhashmask); pph = &pmc_processhash[hindex]; ppnew = NULL; /* * Pre-allocate memory in the FIND_ALLOCATE case since we * cannot call malloc(9) once we hold a spin lock. */ if (mode & PMC_FLAG_ALLOCATE) ppnew = malloc(sizeof(struct pmc_process) + md->pmd_npmc * sizeof(struct pmc_targetstate), M_PMC, M_WAITOK|M_ZERO); mtx_lock_spin(&pmc_processhash_mtx); LIST_FOREACH(pp, pph, pp_next) if (pp->pp_proc == p) break; if ((mode & PMC_FLAG_REMOVE) && pp != NULL) LIST_REMOVE(pp, pp_next); if ((mode & PMC_FLAG_ALLOCATE) && pp == NULL && ppnew != NULL) { ppnew->pp_proc = p; LIST_INSERT_HEAD(pph, ppnew, pp_next); pp = ppnew; ppnew = NULL; } mtx_unlock_spin(&pmc_processhash_mtx); if (pp != NULL && ppnew != NULL) free(ppnew, M_PMC); return pp; } /* * remove a process descriptor from the process hash table. */ static void pmc_remove_process_descriptor(struct pmc_process *pp) { KASSERT(pp->pp_refcnt == 0, ("[pmc,%d] Removing process descriptor %p with count %d", __LINE__, pp, pp->pp_refcnt)); mtx_lock_spin(&pmc_processhash_mtx); LIST_REMOVE(pp, pp_next); mtx_unlock_spin(&pmc_processhash_mtx); } /* * find an owner descriptor corresponding to proc 'p' */ static struct pmc_owner * pmc_find_owner_descriptor(struct proc *p) { uint32_t hindex; struct pmc_owner *po; struct pmc_ownerhash *poh; hindex = PMC_HASH_PTR(p, pmc_ownerhashmask); poh = &pmc_ownerhash[hindex]; po = NULL; LIST_FOREACH(po, poh, po_next) if (po->po_owner == p) break; PMCDBG5(OWN,FND,1, "find-owner proc=%p (%d, %s) hindex=0x%x -> " "pmc-owner=%p", p, p->p_pid, p->p_comm, hindex, po); return po; } /* * pmc_allocate_pmc_descriptor * * Allocate a pmc descriptor and initialize its * fields. */ static struct pmc * pmc_allocate_pmc_descriptor(void) { struct pmc *pmc; pmc = malloc(sizeof(struct pmc), M_PMC, M_WAITOK|M_ZERO); PMCDBG1(PMC,ALL,1, "allocate-pmc -> pmc=%p", pmc); return pmc; } /* * Destroy a pmc descriptor. */ static void pmc_destroy_pmc_descriptor(struct pmc *pm) { KASSERT(pm->pm_state == PMC_STATE_DELETED || pm->pm_state == PMC_STATE_FREE, ("[pmc,%d] destroying non-deleted PMC", __LINE__)); KASSERT(LIST_EMPTY(&pm->pm_targets), ("[pmc,%d] destroying pmc with targets", __LINE__)); KASSERT(pm->pm_owner == NULL, ("[pmc,%d] destroying pmc attached to an owner", __LINE__)); KASSERT(pm->pm_runcount == 0, ("[pmc,%d] pmc has non-zero run count %d", __LINE__, pm->pm_runcount)); free(pm, M_PMC); } static void pmc_wait_for_pmc_idle(struct pmc *pm) { #ifdef HWPMC_DEBUG volatile int maxloop; maxloop = 100 * pmc_cpu_max(); #endif /* * Loop (with a forced context switch) till the PMC's runcount * comes down to zero. */ while (atomic_load_acq_32(&pm->pm_runcount) > 0) { #ifdef HWPMC_DEBUG maxloop--; KASSERT(maxloop > 0, ("[pmc,%d] (ri%d, rc%d) waiting too long for " "pmc to be free", __LINE__, PMC_TO_ROWINDEX(pm), pm->pm_runcount)); #endif pmc_force_context_switch(); } } /* * This function does the following things: * * - detaches the PMC from hardware * - unlinks all target threads that were attached to it * - removes the PMC from its owner's list * - destroys the PMC private mutex * * Once this function completes, the given pmc pointer can be freed by * calling pmc_destroy_pmc_descriptor(). */ static void pmc_release_pmc_descriptor(struct pmc *pm) { enum pmc_mode mode; struct pmc_hw *phw; u_int adjri, ri, cpu; struct pmc_owner *po; struct pmc_binding pb; struct pmc_process *pp; struct pmc_classdep *pcd; struct pmc_target *ptgt, *tmp; sx_assert(&pmc_sx, SX_XLOCKED); KASSERT(pm, ("[pmc,%d] null pmc", __LINE__)); ri = PMC_TO_ROWINDEX(pm); pcd = pmc_ri_to_classdep(md, ri, &adjri); mode = PMC_TO_MODE(pm); PMCDBG3(PMC,REL,1, "release-pmc pmc=%p ri=%d mode=%d", pm, ri, mode); /* * First, we take the PMC off hardware. */ cpu = 0; if (PMC_IS_SYSTEM_MODE(mode)) { /* * A system mode PMC runs on a specific CPU. Switch * to this CPU and turn hardware off. */ pmc_save_cpu_binding(&pb); cpu = PMC_TO_CPU(pm); pmc_select_cpu(cpu); /* switch off non-stalled CPUs */ if (pm->pm_state == PMC_STATE_RUNNING && pm->pm_stalled == 0) { phw = pmc_pcpu[cpu]->pc_hwpmcs[ri]; KASSERT(phw->phw_pmc == pm, ("[pmc, %d] pmc ptr ri(%d) hw(%p) pm(%p)", __LINE__, ri, phw->phw_pmc, pm)); PMCDBG2(PMC,REL,2, "stopping cpu=%d ri=%d", cpu, ri); critical_enter(); pcd->pcd_stop_pmc(cpu, adjri); critical_exit(); } PMCDBG2(PMC,REL,2, "decfg cpu=%d ri=%d", cpu, ri); critical_enter(); pcd->pcd_config_pmc(cpu, adjri, NULL); critical_exit(); /* adjust the global and process count of SS mode PMCs */ if (mode == PMC_MODE_SS && pm->pm_state == PMC_STATE_RUNNING) { po = pm->pm_owner; po->po_sscount--; if (po->po_sscount == 0) { atomic_subtract_rel_int(&pmc_ss_count, 1); LIST_REMOVE(po, po_ssnext); } } pm->pm_state = PMC_STATE_DELETED; pmc_restore_cpu_binding(&pb); /* * We could have references to this PMC structure in * the per-cpu sample queues. Wait for the queue to * drain. */ pmc_wait_for_pmc_idle(pm); } else if (PMC_IS_VIRTUAL_MODE(mode)) { /* * A virtual PMC could be running on multiple CPUs at * a given instant. * * By marking its state as DELETED, we ensure that * this PMC is never further scheduled on hardware. * * Then we wait till all CPUs are done with this PMC. */ pm->pm_state = PMC_STATE_DELETED; /* Wait for the PMCs runcount to come to zero. */ pmc_wait_for_pmc_idle(pm); /* * At this point the PMC is off all CPUs and cannot be * freshly scheduled onto a CPU. It is now safe to * unlink all targets from this PMC. If a * process-record's refcount falls to zero, we remove * it from the hash table. The module-wide SX lock * protects us from races. */ LIST_FOREACH_SAFE(ptgt, &pm->pm_targets, pt_next, tmp) { pp = ptgt->pt_process; pmc_unlink_target_process(pm, pp); /* frees 'ptgt' */ PMCDBG1(PMC,REL,3, "pp->refcnt=%d", pp->pp_refcnt); /* * If the target process record shows that no * PMCs are attached to it, reclaim its space. */ if (pp->pp_refcnt == 0) { pmc_remove_process_descriptor(pp); free(pp, M_PMC); } } cpu = curthread->td_oncpu; /* setup cpu for pmd_release() */ } /* * Release any MD resources */ (void) pcd->pcd_release_pmc(cpu, adjri, pm); /* * Update row disposition */ if (PMC_IS_SYSTEM_MODE(PMC_TO_MODE(pm))) PMC_UNMARK_ROW_STANDALONE(ri); else PMC_UNMARK_ROW_THREAD(ri); /* unlink from the owner's list */ if (pm->pm_owner) { LIST_REMOVE(pm, pm_next); pm->pm_owner = NULL; } } /* * Register an owner and a pmc. */ static int pmc_register_owner(struct proc *p, struct pmc *pmc) { struct pmc_owner *po; sx_assert(&pmc_sx, SX_XLOCKED); if ((po = pmc_find_owner_descriptor(p)) == NULL) if ((po = pmc_allocate_owner_descriptor(p)) == NULL) return ENOMEM; KASSERT(pmc->pm_owner == NULL, ("[pmc,%d] attempting to own an initialized PMC", __LINE__)); pmc->pm_owner = po; LIST_INSERT_HEAD(&po->po_pmcs, pmc, pm_next); PROC_LOCK(p); p->p_flag |= P_HWPMC; PROC_UNLOCK(p); if (po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_pmcallocate(pmc); PMCDBG2(PMC,REG,1, "register-owner pmc-owner=%p pmc=%p", po, pmc); return 0; } /* * Return the current row disposition: * == 0 => FREE * > 0 => PROCESS MODE * < 0 => SYSTEM MODE */ int pmc_getrowdisp(int ri) { return pmc_pmcdisp[ri]; } /* * Check if a PMC at row index 'ri' can be allocated to the current * process. * * Allocation can fail if: * - the current process is already being profiled by a PMC at index 'ri', * attached to it via OP_PMCATTACH. * - the current process has already allocated a PMC at index 'ri' * via OP_ALLOCATE. */ static int pmc_can_allocate_rowindex(struct proc *p, unsigned int ri, int cpu) { enum pmc_mode mode; struct pmc *pm; struct pmc_owner *po; struct pmc_process *pp; PMCDBG5(PMC,ALR,1, "can-allocate-rowindex proc=%p (%d, %s) ri=%d " "cpu=%d", p, p->p_pid, p->p_comm, ri, cpu); /* * We shouldn't have already allocated a process-mode PMC at * row index 'ri'. * * We shouldn't have allocated a system-wide PMC on the same * CPU and same RI. */ if ((po = pmc_find_owner_descriptor(p)) != NULL) LIST_FOREACH(pm, &po->po_pmcs, pm_next) { if (PMC_TO_ROWINDEX(pm) == ri) { mode = PMC_TO_MODE(pm); if (PMC_IS_VIRTUAL_MODE(mode)) return EEXIST; if (PMC_IS_SYSTEM_MODE(mode) && (int) PMC_TO_CPU(pm) == cpu) return EEXIST; } } /* * We also shouldn't be the target of any PMC at this index * since otherwise a PMC_ATTACH to ourselves will fail. */ if ((pp = pmc_find_process_descriptor(p, 0)) != NULL) if (pp->pp_pmcs[ri].pp_pmc) return EEXIST; PMCDBG4(PMC,ALR,2, "can-allocate-rowindex proc=%p (%d, %s) ri=%d ok", p, p->p_pid, p->p_comm, ri); return 0; } /* * Check if a given PMC at row index 'ri' can be currently used in * mode 'mode'. */ static int pmc_can_allocate_row(int ri, enum pmc_mode mode) { enum pmc_disp disp; sx_assert(&pmc_sx, SX_XLOCKED); PMCDBG2(PMC,ALR,1, "can-allocate-row ri=%d mode=%d", ri, mode); if (PMC_IS_SYSTEM_MODE(mode)) disp = PMC_DISP_STANDALONE; else disp = PMC_DISP_THREAD; /* * check disposition for PMC row 'ri': * * Expected disposition Row-disposition Result * * STANDALONE STANDALONE or FREE proceed * STANDALONE THREAD fail * THREAD THREAD or FREE proceed * THREAD STANDALONE fail */ if (!PMC_ROW_DISP_IS_FREE(ri) && !(disp == PMC_DISP_THREAD && PMC_ROW_DISP_IS_THREAD(ri)) && !(disp == PMC_DISP_STANDALONE && PMC_ROW_DISP_IS_STANDALONE(ri))) return EBUSY; /* * All OK */ PMCDBG2(PMC,ALR,2, "can-allocate-row ri=%d mode=%d ok", ri, mode); return 0; } /* * Find a PMC descriptor with user handle 'pmcid' for thread 'td'. */ static struct pmc * pmc_find_pmc_descriptor_in_process(struct pmc_owner *po, pmc_id_t pmcid) { struct pmc *pm; KASSERT(PMC_ID_TO_ROWINDEX(pmcid) < md->pmd_npmc, ("[pmc,%d] Illegal pmc index %d (max %d)", __LINE__, PMC_ID_TO_ROWINDEX(pmcid), md->pmd_npmc)); LIST_FOREACH(pm, &po->po_pmcs, pm_next) if (pm->pm_id == pmcid) return pm; return NULL; } static int pmc_find_pmc(pmc_id_t pmcid, struct pmc **pmc) { struct pmc *pm; struct pmc_owner *po; PMCDBG1(PMC,FND,1, "find-pmc id=%d", pmcid); if ((po = pmc_find_owner_descriptor(curthread->td_proc)) == NULL) return ESRCH; if ((pm = pmc_find_pmc_descriptor_in_process(po, pmcid)) == NULL) return EINVAL; PMCDBG2(PMC,FND,2, "find-pmc id=%d -> pmc=%p", pmcid, pm); *pmc = pm; return 0; } /* * Start a PMC. */ static int pmc_start(struct pmc *pm) { enum pmc_mode mode; struct pmc_owner *po; struct pmc_binding pb; struct pmc_classdep *pcd; int adjri, error, cpu, ri; KASSERT(pm != NULL, ("[pmc,%d] null pm", __LINE__)); mode = PMC_TO_MODE(pm); ri = PMC_TO_ROWINDEX(pm); pcd = pmc_ri_to_classdep(md, ri, &adjri); error = 0; PMCDBG3(PMC,OPS,1, "start pmc=%p mode=%d ri=%d", pm, mode, ri); po = pm->pm_owner; /* * Disallow PMCSTART if a logfile is required but has not been * configured yet. */ if ((pm->pm_flags & PMC_F_NEEDS_LOGFILE) && (po->po_flags & PMC_PO_OWNS_LOGFILE) == 0) return (EDOOFUS); /* programming error */ /* * If this is a sampling mode PMC, log mapping information for * the kernel modules that are currently loaded. */ if (PMC_IS_SAMPLING_MODE(PMC_TO_MODE(pm))) pmc_log_kernel_mappings(pm); if (PMC_IS_VIRTUAL_MODE(mode)) { /* * If a PMCATTACH has never been done on this PMC, * attach it to its owner process. */ if (LIST_EMPTY(&pm->pm_targets)) error = (pm->pm_flags & PMC_F_ATTACH_DONE) ? ESRCH : pmc_attach_process(po->po_owner, pm); /* * If the PMC is attached to its owner, then force a context * switch to ensure that the MD state gets set correctly. */ if (error == 0) { pm->pm_state = PMC_STATE_RUNNING; if (pm->pm_flags & PMC_F_ATTACHED_TO_OWNER) pmc_force_context_switch(); } return (error); } /* * A system-wide PMC. * * Add the owner to the global list if this is a system-wide * sampling PMC. */ if (mode == PMC_MODE_SS) { if (po->po_sscount == 0) { LIST_INSERT_HEAD(&pmc_ss_owners, po, po_ssnext); atomic_add_rel_int(&pmc_ss_count, 1); PMCDBG1(PMC,OPS,1, "po=%p in global list", po); } po->po_sscount++; /* * Log mapping information for all existing processes in the * system. Subsequent mappings are logged as they happen; * see pmc_process_mmap(). */ if (po->po_logprocmaps == 0) { pmc_log_all_process_mappings(po); po->po_logprocmaps = 1; } } /* * Move to the CPU associated with this * PMC, and start the hardware. */ pmc_save_cpu_binding(&pb); cpu = PMC_TO_CPU(pm); if (!pmc_cpu_is_active(cpu)) return (ENXIO); pmc_select_cpu(cpu); /* * global PMCs are configured at allocation time * so write out the initial value and start the PMC. */ pm->pm_state = PMC_STATE_RUNNING; critical_enter(); if ((error = pcd->pcd_write_pmc(cpu, adjri, PMC_IS_SAMPLING_MODE(mode) ? pm->pm_sc.pm_reloadcount : pm->pm_sc.pm_initial)) == 0) error = pcd->pcd_start_pmc(cpu, adjri); critical_exit(); pmc_restore_cpu_binding(&pb); return (error); } /* * Stop a PMC. */ static int pmc_stop(struct pmc *pm) { struct pmc_owner *po; struct pmc_binding pb; struct pmc_classdep *pcd; int adjri, cpu, error, ri; KASSERT(pm != NULL, ("[pmc,%d] null pmc", __LINE__)); PMCDBG3(PMC,OPS,1, "stop pmc=%p mode=%d ri=%d", pm, PMC_TO_MODE(pm), PMC_TO_ROWINDEX(pm)); pm->pm_state = PMC_STATE_STOPPED; /* * If the PMC is a virtual mode one, changing the state to * non-RUNNING is enough to ensure that the PMC never gets * scheduled. * * If this PMC is current running on a CPU, then it will * handled correctly at the time its target process is context * switched out. */ if (PMC_IS_VIRTUAL_MODE(PMC_TO_MODE(pm))) return 0; /* * A system-mode PMC. Move to the CPU associated with * this PMC, and stop the hardware. We update the * 'initial count' so that a subsequent PMCSTART will * resume counting from the current hardware count. */ pmc_save_cpu_binding(&pb); cpu = PMC_TO_CPU(pm); KASSERT(cpu >= 0 && cpu < pmc_cpu_max(), ("[pmc,%d] illegal cpu=%d", __LINE__, cpu)); if (!pmc_cpu_is_active(cpu)) return ENXIO; pmc_select_cpu(cpu); ri = PMC_TO_ROWINDEX(pm); pcd = pmc_ri_to_classdep(md, ri, &adjri); critical_enter(); if ((error = pcd->pcd_stop_pmc(cpu, adjri)) == 0) error = pcd->pcd_read_pmc(cpu, adjri, &pm->pm_sc.pm_initial); critical_exit(); pmc_restore_cpu_binding(&pb); po = pm->pm_owner; /* remove this owner from the global list of SS PMC owners */ if (PMC_TO_MODE(pm) == PMC_MODE_SS) { po->po_sscount--; if (po->po_sscount == 0) { atomic_subtract_rel_int(&pmc_ss_count, 1); LIST_REMOVE(po, po_ssnext); PMCDBG1(PMC,OPS,2,"po=%p removed from global list", po); } } return (error); } #ifdef HWPMC_DEBUG static const char *pmc_op_to_name[] = { #undef __PMC_OP #define __PMC_OP(N, D) #N , __PMC_OPS() NULL }; #endif /* * The syscall interface */ #define PMC_GET_SX_XLOCK(...) do { \ sx_xlock(&pmc_sx); \ if (pmc_hook == NULL) { \ sx_xunlock(&pmc_sx); \ return __VA_ARGS__; \ } \ } while (0) #define PMC_DOWNGRADE_SX() do { \ sx_downgrade(&pmc_sx); \ is_sx_downgraded = 1; \ } while (0) static int pmc_syscall_handler(struct thread *td, void *syscall_args) { int error, is_sx_downgraded, is_sx_locked, op; struct pmc_syscall_args *c; void *arg; PMC_GET_SX_XLOCK(ENOSYS); DROP_GIANT(); is_sx_downgraded = 0; is_sx_locked = 1; c = (struct pmc_syscall_args *) syscall_args; op = c->pmop_code; arg = c->pmop_data; PMCDBG3(MOD,PMS,1, "syscall op=%d \"%s\" arg=%p", op, pmc_op_to_name[op], arg); error = 0; atomic_add_int(&pmc_stats.pm_syscalls, 1); switch(op) { /* * Configure a log file. * * XXX This OP will be reworked. */ case PMC_OP_CONFIGURELOG: { struct proc *p; struct pmc *pm; struct pmc_owner *po; struct pmc_op_configurelog cl; sx_assert(&pmc_sx, SX_XLOCKED); if ((error = copyin(arg, &cl, sizeof(cl))) != 0) break; /* mark this process as owning a log file */ p = td->td_proc; if ((po = pmc_find_owner_descriptor(p)) == NULL) if ((po = pmc_allocate_owner_descriptor(p)) == NULL) { error = ENOMEM; break; } /* * If a valid fd was passed in, try to configure that, * otherwise if 'fd' was less than zero and there was * a log file configured, flush its buffers and * de-configure it. */ if (cl.pm_logfd >= 0) { sx_xunlock(&pmc_sx); is_sx_locked = 0; error = pmclog_configure_log(md, po, cl.pm_logfd); } else if (po->po_flags & PMC_PO_OWNS_LOGFILE) { pmclog_process_closelog(po); error = pmclog_close(po); if (error == 0) { LIST_FOREACH(pm, &po->po_pmcs, pm_next) if (pm->pm_flags & PMC_F_NEEDS_LOGFILE && pm->pm_state == PMC_STATE_RUNNING) pmc_stop(pm); error = pmclog_deconfigure_log(po); } } else error = EINVAL; if (error) break; } break; /* * Flush a log file. */ case PMC_OP_FLUSHLOG: { struct pmc_owner *po; sx_assert(&pmc_sx, SX_XLOCKED); if ((po = pmc_find_owner_descriptor(td->td_proc)) == NULL) { error = EINVAL; break; } error = pmclog_flush(po); } break; /* * Close a log file. */ case PMC_OP_CLOSELOG: { struct pmc_owner *po; sx_assert(&pmc_sx, SX_XLOCKED); if ((po = pmc_find_owner_descriptor(td->td_proc)) == NULL) { error = EINVAL; break; } error = pmclog_close(po); } break; /* * Retrieve hardware configuration. */ case PMC_OP_GETCPUINFO: /* CPU information */ { struct pmc_op_getcpuinfo gci; struct pmc_classinfo *pci; struct pmc_classdep *pcd; int cl; gci.pm_cputype = md->pmd_cputype; gci.pm_ncpu = pmc_cpu_max(); gci.pm_npmc = md->pmd_npmc; gci.pm_nclass = md->pmd_nclass; pci = gci.pm_classes; pcd = md->pmd_classdep; for (cl = 0; cl < md->pmd_nclass; cl++, pci++, pcd++) { pci->pm_caps = pcd->pcd_caps; pci->pm_class = pcd->pcd_class; pci->pm_width = pcd->pcd_width; pci->pm_num = pcd->pcd_num; } error = copyout(&gci, arg, sizeof(gci)); } break; /* * Retrieve soft events list. */ case PMC_OP_GETDYNEVENTINFO: { enum pmc_class cl; enum pmc_event ev; struct pmc_op_getdyneventinfo *gei; struct pmc_dyn_event_descr dev; struct pmc_soft *ps; uint32_t nevent; sx_assert(&pmc_sx, SX_LOCKED); gei = (struct pmc_op_getdyneventinfo *) arg; if ((error = copyin(&gei->pm_class, &cl, sizeof(cl))) != 0) break; /* Only SOFT class is dynamic. */ if (cl != PMC_CLASS_SOFT) { error = EINVAL; break; } nevent = 0; for (ev = PMC_EV_SOFT_FIRST; (int)ev <= PMC_EV_SOFT_LAST; ev++) { ps = pmc_soft_ev_acquire(ev); if (ps == NULL) continue; bcopy(&ps->ps_ev, &dev, sizeof(dev)); pmc_soft_ev_release(ps); error = copyout(&dev, &gei->pm_events[nevent], sizeof(struct pmc_dyn_event_descr)); if (error != 0) break; nevent++; } if (error != 0) break; error = copyout(&nevent, &gei->pm_nevent, sizeof(nevent)); } break; /* * Get module statistics */ case PMC_OP_GETDRIVERSTATS: { struct pmc_op_getdriverstats gms; bcopy(&pmc_stats, &gms, sizeof(gms)); error = copyout(&gms, arg, sizeof(gms)); } break; /* * Retrieve module version number */ case PMC_OP_GETMODULEVERSION: { uint32_t cv, modv; /* retrieve the client's idea of the ABI version */ if ((error = copyin(arg, &cv, sizeof(uint32_t))) != 0) break; /* don't service clients newer than our driver */ modv = PMC_VERSION; if ((cv & 0xFFFF0000) > (modv & 0xFFFF0000)) { error = EPROGMISMATCH; break; } error = copyout(&modv, arg, sizeof(int)); } break; /* * Retrieve the state of all the PMCs on a given * CPU. */ case PMC_OP_GETPMCINFO: { int ari; struct pmc *pm; size_t pmcinfo_size; uint32_t cpu, n, npmc; struct pmc_owner *po; struct pmc_binding pb; struct pmc_classdep *pcd; struct pmc_info *p, *pmcinfo; struct pmc_op_getpmcinfo *gpi; PMC_DOWNGRADE_SX(); gpi = (struct pmc_op_getpmcinfo *) arg; if ((error = copyin(&gpi->pm_cpu, &cpu, sizeof(cpu))) != 0) break; if (cpu >= pmc_cpu_max()) { error = EINVAL; break; } if (!pmc_cpu_is_active(cpu)) { error = ENXIO; break; } /* switch to CPU 'cpu' */ pmc_save_cpu_binding(&pb); pmc_select_cpu(cpu); npmc = md->pmd_npmc; pmcinfo_size = npmc * sizeof(struct pmc_info); pmcinfo = malloc(pmcinfo_size, M_PMC, M_WAITOK); p = pmcinfo; for (n = 0; n < md->pmd_npmc; n++, p++) { pcd = pmc_ri_to_classdep(md, n, &ari); KASSERT(pcd != NULL, ("[pmc,%d] null pcd ri=%d", __LINE__, n)); if ((error = pcd->pcd_describe(cpu, ari, p, &pm)) != 0) break; if (PMC_ROW_DISP_IS_STANDALONE(n)) p->pm_rowdisp = PMC_DISP_STANDALONE; else if (PMC_ROW_DISP_IS_THREAD(n)) p->pm_rowdisp = PMC_DISP_THREAD; else p->pm_rowdisp = PMC_DISP_FREE; p->pm_ownerpid = -1; if (pm == NULL) /* no PMC associated */ continue; po = pm->pm_owner; KASSERT(po->po_owner != NULL, ("[pmc,%d] pmc_owner had a null proc pointer", __LINE__)); p->pm_ownerpid = po->po_owner->p_pid; p->pm_mode = PMC_TO_MODE(pm); p->pm_event = pm->pm_event; p->pm_flags = pm->pm_flags; if (PMC_IS_SAMPLING_MODE(PMC_TO_MODE(pm))) p->pm_reloadcount = pm->pm_sc.pm_reloadcount; } pmc_restore_cpu_binding(&pb); /* now copy out the PMC info collected */ if (error == 0) error = copyout(pmcinfo, &gpi->pm_pmcs, pmcinfo_size); free(pmcinfo, M_PMC); } break; /* * Set the administrative state of a PMC. I.e. whether * the PMC is to be used or not. */ case PMC_OP_PMCADMIN: { int cpu, ri; enum pmc_state request; struct pmc_cpu *pc; struct pmc_hw *phw; struct pmc_op_pmcadmin pma; struct pmc_binding pb; sx_assert(&pmc_sx, SX_XLOCKED); KASSERT(td == curthread, ("[pmc,%d] td != curthread", __LINE__)); error = priv_check(td, PRIV_PMC_MANAGE); if (error) break; if ((error = copyin(arg, &pma, sizeof(pma))) != 0) break; cpu = pma.pm_cpu; if (cpu < 0 || cpu >= (int) pmc_cpu_max()) { error = EINVAL; break; } if (!pmc_cpu_is_active(cpu)) { error = ENXIO; break; } request = pma.pm_state; if (request != PMC_STATE_DISABLED && request != PMC_STATE_FREE) { error = EINVAL; break; } ri = pma.pm_pmc; /* pmc id == row index */ if (ri < 0 || ri >= (int) md->pmd_npmc) { error = EINVAL; break; } /* * We can't disable a PMC with a row-index allocated * for process virtual PMCs. */ if (PMC_ROW_DISP_IS_THREAD(ri) && request == PMC_STATE_DISABLED) { error = EBUSY; break; } /* * otherwise, this PMC on this CPU is either free or * in system-wide mode. */ pmc_save_cpu_binding(&pb); pmc_select_cpu(cpu); pc = pmc_pcpu[cpu]; phw = pc->pc_hwpmcs[ri]; /* * XXX do we need some kind of 'forced' disable? */ if (phw->phw_pmc == NULL) { if (request == PMC_STATE_DISABLED && (phw->phw_state & PMC_PHW_FLAG_IS_ENABLED)) { phw->phw_state &= ~PMC_PHW_FLAG_IS_ENABLED; PMC_MARK_ROW_STANDALONE(ri); } else if (request == PMC_STATE_FREE && (phw->phw_state & PMC_PHW_FLAG_IS_ENABLED) == 0) { phw->phw_state |= PMC_PHW_FLAG_IS_ENABLED; PMC_UNMARK_ROW_STANDALONE(ri); } /* other cases are a no-op */ } else error = EBUSY; pmc_restore_cpu_binding(&pb); } break; /* * Allocate a PMC. */ case PMC_OP_PMCALLOCATE: { int adjri, n; u_int cpu; uint32_t caps; struct pmc *pmc; enum pmc_mode mode; struct pmc_hw *phw; struct pmc_binding pb; struct pmc_classdep *pcd; struct pmc_op_pmcallocate pa; if ((error = copyin(arg, &pa, sizeof(pa))) != 0) break; caps = pa.pm_caps; mode = pa.pm_mode; cpu = pa.pm_cpu; if ((mode != PMC_MODE_SS && mode != PMC_MODE_SC && mode != PMC_MODE_TS && mode != PMC_MODE_TC) || (cpu != (u_int) PMC_CPU_ANY && cpu >= pmc_cpu_max())) { error = EINVAL; break; } /* * Virtual PMCs should only ask for a default CPU. * System mode PMCs need to specify a non-default CPU. */ if ((PMC_IS_VIRTUAL_MODE(mode) && cpu != (u_int) PMC_CPU_ANY) || (PMC_IS_SYSTEM_MODE(mode) && cpu == (u_int) PMC_CPU_ANY)) { error = EINVAL; break; } /* * Check that an inactive CPU is not being asked for. */ if (PMC_IS_SYSTEM_MODE(mode) && !pmc_cpu_is_active(cpu)) { error = ENXIO; break; } /* * Refuse an allocation for a system-wide PMC if this * process has been jailed, or if this process lacks * super-user credentials and the sysctl tunable * 'security.bsd.unprivileged_syspmcs' is zero. */ if (PMC_IS_SYSTEM_MODE(mode)) { if (jailed(curthread->td_ucred)) { error = EPERM; break; } if (!pmc_unprivileged_syspmcs) { error = priv_check(curthread, PRIV_PMC_SYSTEM); if (error) break; } } /* * Look for valid values for 'pm_flags' */ if ((pa.pm_flags & ~(PMC_F_DESCENDANTS | PMC_F_LOG_PROCCSW | PMC_F_LOG_PROCEXIT | PMC_F_CALLCHAIN)) != 0) { error = EINVAL; break; } /* process logging options are not allowed for system PMCs */ if (PMC_IS_SYSTEM_MODE(mode) && (pa.pm_flags & (PMC_F_LOG_PROCCSW | PMC_F_LOG_PROCEXIT))) { error = EINVAL; break; } /* * All sampling mode PMCs need to be able to interrupt the * CPU. */ if (PMC_IS_SAMPLING_MODE(mode)) caps |= PMC_CAP_INTERRUPT; /* A valid class specifier should have been passed in. */ for (n = 0; n < md->pmd_nclass; n++) if (md->pmd_classdep[n].pcd_class == pa.pm_class) break; if (n == md->pmd_nclass) { error = EINVAL; break; } /* The requested PMC capabilities should be feasible. */ if ((md->pmd_classdep[n].pcd_caps & caps) != caps) { error = EOPNOTSUPP; break; } PMCDBG4(PMC,ALL,2, "event=%d caps=0x%x mode=%d cpu=%d", pa.pm_ev, caps, mode, cpu); pmc = pmc_allocate_pmc_descriptor(); pmc->pm_id = PMC_ID_MAKE_ID(cpu,pa.pm_mode,pa.pm_class, PMC_ID_INVALID); pmc->pm_event = pa.pm_ev; pmc->pm_state = PMC_STATE_FREE; pmc->pm_caps = caps; pmc->pm_flags = pa.pm_flags; /* switch thread to CPU 'cpu' */ pmc_save_cpu_binding(&pb); #define PMC_IS_SHAREABLE_PMC(cpu, n) \ (pmc_pcpu[(cpu)]->pc_hwpmcs[(n)]->phw_state & \ PMC_PHW_FLAG_IS_SHAREABLE) #define PMC_IS_UNALLOCATED(cpu, n) \ (pmc_pcpu[(cpu)]->pc_hwpmcs[(n)]->phw_pmc == NULL) if (PMC_IS_SYSTEM_MODE(mode)) { pmc_select_cpu(cpu); for (n = 0; n < (int) md->pmd_npmc; n++) { pcd = pmc_ri_to_classdep(md, n, &adjri); if (pmc_can_allocate_row(n, mode) == 0 && pmc_can_allocate_rowindex( curthread->td_proc, n, cpu) == 0 && (PMC_IS_UNALLOCATED(cpu, n) || PMC_IS_SHAREABLE_PMC(cpu, n)) && pcd->pcd_allocate_pmc(cpu, adjri, pmc, &pa) == 0) break; } } else { /* Process virtual mode */ for (n = 0; n < (int) md->pmd_npmc; n++) { pcd = pmc_ri_to_classdep(md, n, &adjri); if (pmc_can_allocate_row(n, mode) == 0 && pmc_can_allocate_rowindex( curthread->td_proc, n, PMC_CPU_ANY) == 0 && pcd->pcd_allocate_pmc(curthread->td_oncpu, adjri, pmc, &pa) == 0) break; } } #undef PMC_IS_UNALLOCATED #undef PMC_IS_SHAREABLE_PMC pmc_restore_cpu_binding(&pb); if (n == (int) md->pmd_npmc) { pmc_destroy_pmc_descriptor(pmc); pmc = NULL; error = EINVAL; break; } /* Fill in the correct value in the ID field */ pmc->pm_id = PMC_ID_MAKE_ID(cpu,mode,pa.pm_class,n); PMCDBG5(PMC,ALL,2, "ev=%d class=%d mode=%d n=%d -> pmcid=%x", pmc->pm_event, pa.pm_class, mode, n, pmc->pm_id); /* Process mode PMCs with logging enabled need log files */ if (pmc->pm_flags & (PMC_F_LOG_PROCEXIT | PMC_F_LOG_PROCCSW)) pmc->pm_flags |= PMC_F_NEEDS_LOGFILE; /* All system mode sampling PMCs require a log file */ if (PMC_IS_SAMPLING_MODE(mode) && PMC_IS_SYSTEM_MODE(mode)) pmc->pm_flags |= PMC_F_NEEDS_LOGFILE; /* * Configure global pmc's immediately */ if (PMC_IS_SYSTEM_MODE(PMC_TO_MODE(pmc))) { pmc_save_cpu_binding(&pb); pmc_select_cpu(cpu); phw = pmc_pcpu[cpu]->pc_hwpmcs[n]; pcd = pmc_ri_to_classdep(md, n, &adjri); if ((phw->phw_state & PMC_PHW_FLAG_IS_ENABLED) == 0 || (error = pcd->pcd_config_pmc(cpu, adjri, pmc)) != 0) { (void) pcd->pcd_release_pmc(cpu, adjri, pmc); pmc_destroy_pmc_descriptor(pmc); pmc = NULL; pmc_restore_cpu_binding(&pb); error = EPERM; break; } pmc_restore_cpu_binding(&pb); } pmc->pm_state = PMC_STATE_ALLOCATED; /* * mark row disposition */ if (PMC_IS_SYSTEM_MODE(mode)) PMC_MARK_ROW_STANDALONE(n); else PMC_MARK_ROW_THREAD(n); /* * Register this PMC with the current thread as its owner. */ if ((error = pmc_register_owner(curthread->td_proc, pmc)) != 0) { pmc_release_pmc_descriptor(pmc); pmc_destroy_pmc_descriptor(pmc); pmc = NULL; break; } /* * Return the allocated index. */ pa.pm_pmcid = pmc->pm_id; error = copyout(&pa, arg, sizeof(pa)); } break; /* * Attach a PMC to a process. */ case PMC_OP_PMCATTACH: { struct pmc *pm; struct proc *p; struct pmc_op_pmcattach a; sx_assert(&pmc_sx, SX_XLOCKED); if ((error = copyin(arg, &a, sizeof(a))) != 0) break; if (a.pm_pid < 0) { error = EINVAL; break; } else if (a.pm_pid == 0) a.pm_pid = td->td_proc->p_pid; if ((error = pmc_find_pmc(a.pm_pmc, &pm)) != 0) break; if (PMC_IS_SYSTEM_MODE(PMC_TO_MODE(pm))) { error = EINVAL; break; } /* PMCs may be (re)attached only when allocated or stopped */ if (pm->pm_state == PMC_STATE_RUNNING) { error = EBUSY; break; } else if (pm->pm_state != PMC_STATE_ALLOCATED && pm->pm_state != PMC_STATE_STOPPED) { error = EINVAL; break; } /* lookup pid */ if ((p = pfind(a.pm_pid)) == NULL) { error = ESRCH; break; } /* * Ignore processes that are working on exiting. */ if (p->p_flag & P_WEXIT) { error = ESRCH; PROC_UNLOCK(p); /* pfind() returns a locked process */ break; } /* * we are allowed to attach a PMC to a process if * we can debug it. */ error = p_candebug(curthread, p); PROC_UNLOCK(p); if (error == 0) error = pmc_attach_process(p, pm); } break; /* * Detach an attached PMC from a process. */ case PMC_OP_PMCDETACH: { struct pmc *pm; struct proc *p; struct pmc_op_pmcattach a; if ((error = copyin(arg, &a, sizeof(a))) != 0) break; if (a.pm_pid < 0) { error = EINVAL; break; } else if (a.pm_pid == 0) a.pm_pid = td->td_proc->p_pid; if ((error = pmc_find_pmc(a.pm_pmc, &pm)) != 0) break; if ((p = pfind(a.pm_pid)) == NULL) { error = ESRCH; break; } /* * Treat processes that are in the process of exiting * as if they were not present. */ if (p->p_flag & P_WEXIT) error = ESRCH; PROC_UNLOCK(p); /* pfind() returns a locked process */ if (error == 0) error = pmc_detach_process(p, pm); } break; /* * Retrieve the MSR number associated with the counter * 'pmc_id'. This allows processes to directly use RDPMC * instructions to read their PMCs, without the overhead of a * system call. */ case PMC_OP_PMCGETMSR: { int adjri, ri; struct pmc *pm; struct pmc_target *pt; struct pmc_op_getmsr gm; struct pmc_classdep *pcd; PMC_DOWNGRADE_SX(); if ((error = copyin(arg, &gm, sizeof(gm))) != 0) break; if ((error = pmc_find_pmc(gm.pm_pmcid, &pm)) != 0) break; /* * The allocated PMC has to be a process virtual PMC, * i.e., of type MODE_T[CS]. Global PMCs can only be * read using the PMCREAD operation since they may be * allocated on a different CPU than the one we could * be running on at the time of the RDPMC instruction. * * The GETMSR operation is not allowed for PMCs that * are inherited across processes. */ if (!PMC_IS_VIRTUAL_MODE(PMC_TO_MODE(pm)) || (pm->pm_flags & PMC_F_DESCENDANTS)) { error = EINVAL; break; } /* * It only makes sense to use a RDPMC (or its * equivalent instruction on non-x86 architectures) on * a process that has allocated and attached a PMC to * itself. Conversely the PMC is only allowed to have * one process attached to it -- its owner. */ if ((pt = LIST_FIRST(&pm->pm_targets)) == NULL || LIST_NEXT(pt, pt_next) != NULL || pt->pt_process->pp_proc != pm->pm_owner->po_owner) { error = EINVAL; break; } ri = PMC_TO_ROWINDEX(pm); pcd = pmc_ri_to_classdep(md, ri, &adjri); /* PMC class has no 'GETMSR' support */ if (pcd->pcd_get_msr == NULL) { error = ENOSYS; break; } if ((error = (*pcd->pcd_get_msr)(adjri, &gm.pm_msr)) < 0) break; if ((error = copyout(&gm, arg, sizeof(gm))) < 0) break; /* * Mark our process as using MSRs. Update machine * state using a forced context switch. */ pt->pt_process->pp_flags |= PMC_PP_ENABLE_MSR_ACCESS; pmc_force_context_switch(); } break; /* * Release an allocated PMC */ case PMC_OP_PMCRELEASE: { pmc_id_t pmcid; struct pmc *pm; struct pmc_owner *po; struct pmc_op_simple sp; /* * Find PMC pointer for the named PMC. * * Use pmc_release_pmc_descriptor() to switch off the * PMC, remove all its target threads, and remove the * PMC from its owner's list. * * Remove the owner record if this is the last PMC * owned. * * Free up space. */ if ((error = copyin(arg, &sp, sizeof(sp))) != 0) break; pmcid = sp.pm_pmcid; if ((error = pmc_find_pmc(pmcid, &pm)) != 0) break; po = pm->pm_owner; pmc_release_pmc_descriptor(pm); pmc_maybe_remove_owner(po); pmc_destroy_pmc_descriptor(pm); } break; /* * Read and/or write a PMC. */ case PMC_OP_PMCRW: { int adjri; struct pmc *pm; uint32_t cpu, ri; pmc_value_t oldvalue; struct pmc_binding pb; struct pmc_op_pmcrw prw; struct pmc_classdep *pcd; struct pmc_op_pmcrw *pprw; PMC_DOWNGRADE_SX(); if ((error = copyin(arg, &prw, sizeof(prw))) != 0) break; ri = 0; PMCDBG2(PMC,OPS,1, "rw id=%d flags=0x%x", prw.pm_pmcid, prw.pm_flags); /* must have at least one flag set */ if ((prw.pm_flags & (PMC_F_OLDVALUE|PMC_F_NEWVALUE)) == 0) { error = EINVAL; break; } /* locate pmc descriptor */ if ((error = pmc_find_pmc(prw.pm_pmcid, &pm)) != 0) break; /* Can't read a PMC that hasn't been started. */ if (pm->pm_state != PMC_STATE_ALLOCATED && pm->pm_state != PMC_STATE_STOPPED && pm->pm_state != PMC_STATE_RUNNING) { error = EINVAL; break; } /* writing a new value is allowed only for 'STOPPED' pmcs */ if (pm->pm_state == PMC_STATE_RUNNING && (prw.pm_flags & PMC_F_NEWVALUE)) { error = EBUSY; break; } if (PMC_IS_VIRTUAL_MODE(PMC_TO_MODE(pm))) { /* * If this PMC is attached to its owner (i.e., * the process requesting this operation) and * is running, then attempt to get an * upto-date reading from hardware for a READ. * Writes are only allowed when the PMC is * stopped, so only update the saved value * field. * * If the PMC is not running, or is not * attached to its owner, read/write to the * savedvalue field. */ ri = PMC_TO_ROWINDEX(pm); pcd = pmc_ri_to_classdep(md, ri, &adjri); mtx_pool_lock_spin(pmc_mtxpool, pm); cpu = curthread->td_oncpu; if (prw.pm_flags & PMC_F_OLDVALUE) { if ((pm->pm_flags & PMC_F_ATTACHED_TO_OWNER) && (pm->pm_state == PMC_STATE_RUNNING)) error = (*pcd->pcd_read_pmc)(cpu, adjri, &oldvalue); else oldvalue = pm->pm_gv.pm_savedvalue; } if (prw.pm_flags & PMC_F_NEWVALUE) pm->pm_gv.pm_savedvalue = prw.pm_value; mtx_pool_unlock_spin(pmc_mtxpool, pm); } else { /* System mode PMCs */ cpu = PMC_TO_CPU(pm); ri = PMC_TO_ROWINDEX(pm); pcd = pmc_ri_to_classdep(md, ri, &adjri); if (!pmc_cpu_is_active(cpu)) { error = ENXIO; break; } /* move this thread to CPU 'cpu' */ pmc_save_cpu_binding(&pb); pmc_select_cpu(cpu); critical_enter(); /* save old value */ if (prw.pm_flags & PMC_F_OLDVALUE) if ((error = (*pcd->pcd_read_pmc)(cpu, adjri, &oldvalue))) goto error; /* write out new value */ if (prw.pm_flags & PMC_F_NEWVALUE) error = (*pcd->pcd_write_pmc)(cpu, adjri, prw.pm_value); error: critical_exit(); pmc_restore_cpu_binding(&pb); if (error) break; } pprw = (struct pmc_op_pmcrw *) arg; #ifdef HWPMC_DEBUG if (prw.pm_flags & PMC_F_NEWVALUE) PMCDBG3(PMC,OPS,2, "rw id=%d new %jx -> old %jx", ri, prw.pm_value, oldvalue); else if (prw.pm_flags & PMC_F_OLDVALUE) PMCDBG2(PMC,OPS,2, "rw id=%d -> old %jx", ri, oldvalue); #endif /* return old value if requested */ if (prw.pm_flags & PMC_F_OLDVALUE) if ((error = copyout(&oldvalue, &pprw->pm_value, sizeof(prw.pm_value)))) break; } break; /* * Set the sampling rate for a sampling mode PMC and the * initial count for a counting mode PMC. */ case PMC_OP_PMCSETCOUNT: { struct pmc *pm; struct pmc_op_pmcsetcount sc; PMC_DOWNGRADE_SX(); if ((error = copyin(arg, &sc, sizeof(sc))) != 0) break; if ((error = pmc_find_pmc(sc.pm_pmcid, &pm)) != 0) break; if (pm->pm_state == PMC_STATE_RUNNING) { error = EBUSY; break; } if (PMC_IS_SAMPLING_MODE(PMC_TO_MODE(pm))) pm->pm_sc.pm_reloadcount = sc.pm_count; else pm->pm_sc.pm_initial = sc.pm_count; } break; /* * Start a PMC. */ case PMC_OP_PMCSTART: { pmc_id_t pmcid; struct pmc *pm; struct pmc_op_simple sp; sx_assert(&pmc_sx, SX_XLOCKED); if ((error = copyin(arg, &sp, sizeof(sp))) != 0) break; pmcid = sp.pm_pmcid; if ((error = pmc_find_pmc(pmcid, &pm)) != 0) break; KASSERT(pmcid == pm->pm_id, ("[pmc,%d] pmcid %x != id %x", __LINE__, pm->pm_id, pmcid)); if (pm->pm_state == PMC_STATE_RUNNING) /* already running */ break; else if (pm->pm_state != PMC_STATE_STOPPED && pm->pm_state != PMC_STATE_ALLOCATED) { error = EINVAL; break; } error = pmc_start(pm); } break; /* * Stop a PMC. */ case PMC_OP_PMCSTOP: { pmc_id_t pmcid; struct pmc *pm; struct pmc_op_simple sp; PMC_DOWNGRADE_SX(); if ((error = copyin(arg, &sp, sizeof(sp))) != 0) break; pmcid = sp.pm_pmcid; /* * Mark the PMC as inactive and invoke the MD stop * routines if needed. */ if ((error = pmc_find_pmc(pmcid, &pm)) != 0) break; KASSERT(pmcid == pm->pm_id, ("[pmc,%d] pmc id %x != pmcid %x", __LINE__, pm->pm_id, pmcid)); if (pm->pm_state == PMC_STATE_STOPPED) /* already stopped */ break; else if (pm->pm_state != PMC_STATE_RUNNING) { error = EINVAL; break; } error = pmc_stop(pm); } break; /* * Write a user supplied value to the log file. */ case PMC_OP_WRITELOG: { struct pmc_op_writelog wl; struct pmc_owner *po; PMC_DOWNGRADE_SX(); if ((error = copyin(arg, &wl, sizeof(wl))) != 0) break; if ((po = pmc_find_owner_descriptor(td->td_proc)) == NULL) { error = EINVAL; break; } if ((po->po_flags & PMC_PO_OWNS_LOGFILE) == 0) { error = EINVAL; break; } error = pmclog_process_userlog(po, &wl); } break; default: error = EINVAL; break; } if (is_sx_locked != 0) { if (is_sx_downgraded) sx_sunlock(&pmc_sx); else sx_xunlock(&pmc_sx); } if (error) atomic_add_int(&pmc_stats.pm_syscall_errors, 1); PICKUP_GIANT(); return error; } /* * Helper functions */ /* * Mark the thread as needing callchain capture and post an AST. The * actual callchain capture will be done in a context where it is safe * to take page faults. */ static void pmc_post_callchain_callback(void) { struct thread *td; td = curthread; /* * If there is multiple PMCs for the same interrupt ignore new post */ if (td->td_pflags & TDP_CALLCHAIN) return; /* * Mark this thread as needing callchain capture. * `td->td_pflags' will be safe to touch because this thread * was in user space when it was interrupted. */ td->td_pflags |= TDP_CALLCHAIN; /* * Don't let this thread migrate between CPUs until callchain * capture completes. */ sched_pin(); return; } /* * Interrupt processing. * * Find a free slot in the per-cpu array of samples and capture the * current callchain there. If a sample was successfully added, a bit * is set in mask 'pmc_cpumask' denoting that the DO_SAMPLES hook * needs to be invoked from the clock handler. * * This function is meant to be called from an NMI handler. It cannot * use any of the locking primitives supplied by the OS. */ int pmc_process_interrupt(int cpu, int ring, struct pmc *pm, struct trapframe *tf, int inuserspace) { int error, callchaindepth; struct thread *td; struct pmc_sample *ps; struct pmc_samplebuffer *psb; error = 0; /* * Allocate space for a sample buffer. */ psb = pmc_pcpu[cpu]->pc_sb[ring]; ps = psb->ps_write; if (ps->ps_nsamples) { /* in use, reader hasn't caught up */ pm->pm_stalled = 1; atomic_add_int(&pmc_stats.pm_intr_bufferfull, 1); PMCDBG6(SAM,INT,1,"(spc) cpu=%d pm=%p tf=%p um=%d wr=%d rd=%d", cpu, pm, (void *) tf, inuserspace, (int) (psb->ps_write - psb->ps_samples), (int) (psb->ps_read - psb->ps_samples)); error = ENOMEM; goto done; } /* Fill in entry. */ PMCDBG6(SAM,INT,1,"cpu=%d pm=%p tf=%p um=%d wr=%d rd=%d", cpu, pm, (void *) tf, inuserspace, (int) (psb->ps_write - psb->ps_samples), (int) (psb->ps_read - psb->ps_samples)); KASSERT(pm->pm_runcount >= 0, ("[pmc,%d] pm=%p runcount %d", __LINE__, (void *) pm, pm->pm_runcount)); atomic_add_rel_int(&pm->pm_runcount, 1); /* hold onto PMC */ ps->ps_pmc = pm; if ((td = curthread) && td->td_proc) ps->ps_pid = td->td_proc->p_pid; else ps->ps_pid = -1; ps->ps_cpu = cpu; ps->ps_td = td; ps->ps_flags = inuserspace ? PMC_CC_F_USERSPACE : 0; callchaindepth = (pm->pm_flags & PMC_F_CALLCHAIN) ? pmc_callchaindepth : 1; if (callchaindepth == 1) ps->ps_pc[0] = PMC_TRAPFRAME_TO_PC(tf); else { /* * Kernel stack traversals can be done immediately, * while we defer to an AST for user space traversals. */ if (!inuserspace) { callchaindepth = pmc_save_kernel_callchain(ps->ps_pc, callchaindepth, tf); } else { pmc_post_callchain_callback(); callchaindepth = PMC_SAMPLE_INUSE; } } ps->ps_nsamples = callchaindepth; /* mark entry as in use */ /* increment write pointer, modulo ring buffer size */ ps++; if (ps == psb->ps_fence) psb->ps_write = psb->ps_samples; else psb->ps_write = ps; done: /* mark CPU as needing processing */ CPU_SET_ATOMIC(cpu, &pmc_cpumask); return (error); } /* * Capture a user call chain. This function will be called from ast() * before control returns to userland and before the process gets * rescheduled. */ static void pmc_capture_user_callchain(int cpu, int ring, struct trapframe *tf) { int i; struct pmc *pm; struct thread *td; struct pmc_sample *ps; struct pmc_samplebuffer *psb; #ifdef INVARIANTS int ncallchains; #endif psb = pmc_pcpu[cpu]->pc_sb[ring]; td = curthread; KASSERT(td->td_pflags & TDP_CALLCHAIN, ("[pmc,%d] Retrieving callchain for thread that doesn't want it", __LINE__)); #ifdef INVARIANTS ncallchains = 0; #endif /* * Iterate through all deferred callchain requests. */ ps = psb->ps_samples; for (i = 0; i < pmc_nsamples; i++, ps++) { if (ps->ps_nsamples != PMC_SAMPLE_INUSE) continue; if (ps->ps_td != td) continue; KASSERT(ps->ps_cpu == cpu, ("[pmc,%d] cpu mismatch ps_cpu=%d pcpu=%d", __LINE__, ps->ps_cpu, PCPU_GET(cpuid))); pm = ps->ps_pmc; KASSERT(pm->pm_flags & PMC_F_CALLCHAIN, ("[pmc,%d] Retrieving callchain for PMC that doesn't " "want it", __LINE__)); KASSERT(pm->pm_runcount > 0, ("[pmc,%d] runcount %d", __LINE__, pm->pm_runcount)); /* * Retrieve the callchain and mark the sample buffer * as 'processable' by the timer tick sweep code. */ ps->ps_nsamples = pmc_save_user_callchain(ps->ps_pc, pmc_callchaindepth, tf); #ifdef INVARIANTS ncallchains++; #endif } KASSERT(ncallchains > 0, ("[pmc,%d] cpu %d didn't find a sample to collect", __LINE__, cpu)); KASSERT(td->td_pinned == 1, ("[pmc,%d] invalid td_pinned value", __LINE__)); sched_unpin(); /* Can migrate safely now. */ return; } /* * Process saved PC samples. */ static void pmc_process_samples(int cpu, int ring) { struct pmc *pm; int adjri, n; struct thread *td; struct pmc_owner *po; struct pmc_sample *ps; struct pmc_classdep *pcd; struct pmc_samplebuffer *psb; KASSERT(PCPU_GET(cpuid) == cpu, ("[pmc,%d] not on the correct CPU pcpu=%d cpu=%d", __LINE__, PCPU_GET(cpuid), cpu)); psb = pmc_pcpu[cpu]->pc_sb[ring]; for (n = 0; n < pmc_nsamples; n++) { /* bound on #iterations */ ps = psb->ps_read; if (ps->ps_nsamples == PMC_SAMPLE_FREE) break; pm = ps->ps_pmc; KASSERT(pm->pm_runcount > 0, ("[pmc,%d] pm=%p runcount %d", __LINE__, (void *) pm, pm->pm_runcount)); po = pm->pm_owner; KASSERT(PMC_IS_SAMPLING_MODE(PMC_TO_MODE(pm)), ("[pmc,%d] pmc=%p non-sampling mode=%d", __LINE__, pm, PMC_TO_MODE(pm))); /* Ignore PMCs that have been switched off */ if (pm->pm_state != PMC_STATE_RUNNING) goto entrydone; /* If there is a pending AST wait for completion */ if (ps->ps_nsamples == PMC_SAMPLE_INUSE) { /* Need a rescan at a later time. */ CPU_SET_ATOMIC(cpu, &pmc_cpumask); break; } PMCDBG6(SAM,OPS,1,"cpu=%d pm=%p n=%d fl=%x wr=%d rd=%d", cpu, pm, ps->ps_nsamples, ps->ps_flags, (int) (psb->ps_write - psb->ps_samples), (int) (psb->ps_read - psb->ps_samples)); /* * If this is a process-mode PMC that is attached to * its owner, and if the PC is in user mode, update * profiling statistics like timer-based profiling * would have done. */ if (pm->pm_flags & PMC_F_ATTACHED_TO_OWNER) { if (ps->ps_flags & PMC_CC_F_USERSPACE) { td = FIRST_THREAD_IN_PROC(po->po_owner); addupc_intr(td, ps->ps_pc[0], 1); } goto entrydone; } /* * Otherwise, this is either a sampling mode PMC that * is attached to a different process than its owner, * or a system-wide sampling PMC. Dispatch a log * entry to the PMC's owner process. */ pmclog_process_callchain(pm, ps); entrydone: ps->ps_nsamples = 0; /* mark entry as free */ atomic_subtract_rel_int(&pm->pm_runcount, 1); /* increment read pointer, modulo sample size */ if (++ps == psb->ps_fence) psb->ps_read = psb->ps_samples; else psb->ps_read = ps; } atomic_add_int(&pmc_stats.pm_log_sweeps, 1); /* Do not re-enable stalled PMCs if we failed to process any samples */ if (n == 0) return; /* * Restart any stalled sampling PMCs on this CPU. * * If the NMI handler sets the pm_stalled field of a PMC after * the check below, we'll end up processing the stalled PMC at * the next hardclock tick. */ for (n = 0; n < md->pmd_npmc; n++) { pcd = pmc_ri_to_classdep(md, n, &adjri); KASSERT(pcd != NULL, ("[pmc,%d] null pcd ri=%d", __LINE__, n)); (void) (*pcd->pcd_get_config)(cpu,adjri,&pm); if (pm == NULL || /* !cfg'ed */ pm->pm_state != PMC_STATE_RUNNING || /* !active */ !PMC_IS_SAMPLING_MODE(PMC_TO_MODE(pm)) || /* !sampling */ pm->pm_stalled == 0) /* !stalled */ continue; pm->pm_stalled = 0; (*pcd->pcd_start_pmc)(cpu, adjri); } } /* * Event handlers. */ /* * Handle a process exit. * * Remove this process from all hash tables. If this process * owned any PMCs, turn off those PMCs and deallocate them, * removing any associations with target processes. * * This function will be called by the last 'thread' of a * process. * * XXX This eventhandler gets called early in the exit process. * Consider using a 'hook' invocation from thread_exit() or equivalent * spot. Another negative is that kse_exit doesn't seem to call * exit1() [??]. * */ static void pmc_process_exit(void *arg __unused, struct proc *p) { struct pmc *pm; int adjri, cpu; unsigned int ri; int is_using_hwpmcs; struct pmc_owner *po; struct pmc_process *pp; struct pmc_classdep *pcd; pmc_value_t newvalue, tmp; PROC_LOCK(p); is_using_hwpmcs = p->p_flag & P_HWPMC; PROC_UNLOCK(p); /* * Log a sysexit event to all SS PMC owners. */ LIST_FOREACH(po, &pmc_ss_owners, po_ssnext) if (po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_sysexit(po, p->p_pid); if (!is_using_hwpmcs) return; PMC_GET_SX_XLOCK(); PMCDBG3(PRC,EXT,1,"process-exit proc=%p (%d, %s)", p, p->p_pid, p->p_comm); /* * Since this code is invoked by the last thread in an exiting * process, we would have context switched IN at some prior * point. However, with PREEMPTION, kernel mode context * switches may happen any time, so we want to disable a * context switch OUT till we get any PMCs targetting this * process off the hardware. * * We also need to atomically remove this process' * entry from our target process hash table, using * PMC_FLAG_REMOVE. */ PMCDBG3(PRC,EXT,1, "process-exit proc=%p (%d, %s)", p, p->p_pid, p->p_comm); critical_enter(); /* no preemption */ cpu = curthread->td_oncpu; if ((pp = pmc_find_process_descriptor(p, PMC_FLAG_REMOVE)) != NULL) { PMCDBG2(PRC,EXT,2, "process-exit proc=%p pmc-process=%p", p, pp); /* * The exiting process could the target of * some PMCs which will be running on * currently executing CPU. * * We need to turn these PMCs off like we * would do at context switch OUT time. */ for (ri = 0; ri < md->pmd_npmc; ri++) { /* * Pick up the pmc pointer from hardware * state similar to the CSW_OUT code. */ pm = NULL; pcd = pmc_ri_to_classdep(md, ri, &adjri); (void) (*pcd->pcd_get_config)(cpu, adjri, &pm); PMCDBG2(PRC,EXT,2, "ri=%d pm=%p", ri, pm); if (pm == NULL || !PMC_IS_VIRTUAL_MODE(PMC_TO_MODE(pm))) continue; PMCDBG4(PRC,EXT,2, "ppmcs[%d]=%p pm=%p " "state=%d", ri, pp->pp_pmcs[ri].pp_pmc, pm, pm->pm_state); KASSERT(PMC_TO_ROWINDEX(pm) == ri, ("[pmc,%d] ri mismatch pmc(%d) ri(%d)", __LINE__, PMC_TO_ROWINDEX(pm), ri)); KASSERT(pm == pp->pp_pmcs[ri].pp_pmc, ("[pmc,%d] pm %p != pp_pmcs[%d] %p", __LINE__, pm, ri, pp->pp_pmcs[ri].pp_pmc)); (void) pcd->pcd_stop_pmc(cpu, adjri); KASSERT(pm->pm_runcount > 0, ("[pmc,%d] bad runcount ri %d rc %d", __LINE__, ri, pm->pm_runcount)); /* Stop hardware only if it is actually running */ if (pm->pm_state == PMC_STATE_RUNNING && pm->pm_stalled == 0) { pcd->pcd_read_pmc(cpu, adjri, &newvalue); tmp = newvalue - PMC_PCPU_SAVED(cpu,ri); mtx_pool_lock_spin(pmc_mtxpool, pm); pm->pm_gv.pm_savedvalue += tmp; pp->pp_pmcs[ri].pp_pmcval += tmp; mtx_pool_unlock_spin(pmc_mtxpool, pm); } atomic_subtract_rel_int(&pm->pm_runcount,1); KASSERT((int) pm->pm_runcount >= 0, ("[pmc,%d] runcount is %d", __LINE__, ri)); (void) pcd->pcd_config_pmc(cpu, adjri, NULL); } /* * Inform the MD layer of this pseudo "context switch * out" */ (void) md->pmd_switch_out(pmc_pcpu[cpu], pp); critical_exit(); /* ok to be pre-empted now */ /* * Unlink this process from the PMCs that are * targetting it. This will send a signal to * all PMC owner's whose PMCs are orphaned. * * Log PMC value at exit time if requested. */ for (ri = 0; ri < md->pmd_npmc; ri++) if ((pm = pp->pp_pmcs[ri].pp_pmc) != NULL) { if (pm->pm_flags & PMC_F_NEEDS_LOGFILE && PMC_IS_COUNTING_MODE(PMC_TO_MODE(pm))) pmclog_process_procexit(pm, pp); pmc_unlink_target_process(pm, pp); } free(pp, M_PMC); } else critical_exit(); /* pp == NULL */ /* * If the process owned PMCs, free them up and free up * memory. */ if ((po = pmc_find_owner_descriptor(p)) != NULL) { pmc_remove_owner(po); pmc_destroy_owner_descriptor(po); } sx_xunlock(&pmc_sx); } /* * Handle a process fork. * * If the parent process 'p1' is under HWPMC monitoring, then copy * over any attached PMCs that have 'do_descendants' semantics. */ static void pmc_process_fork(void *arg __unused, struct proc *p1, struct proc *newproc, int flags) { int is_using_hwpmcs; unsigned int ri; uint32_t do_descendants; struct pmc *pm; struct pmc_owner *po; struct pmc_process *ppnew, *ppold; (void) flags; /* unused parameter */ PROC_LOCK(p1); is_using_hwpmcs = p1->p_flag & P_HWPMC; PROC_UNLOCK(p1); /* * If there are system-wide sampling PMCs active, we need to * log all fork events to their owner's logs. */ LIST_FOREACH(po, &pmc_ss_owners, po_ssnext) if (po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_procfork(po, p1->p_pid, newproc->p_pid); if (!is_using_hwpmcs) return; PMC_GET_SX_XLOCK(); PMCDBG4(PMC,FRK,1, "process-fork proc=%p (%d, %s) -> %p", p1, p1->p_pid, p1->p_comm, newproc); /* * If the parent process (curthread->td_proc) is a * target of any PMCs, look for PMCs that are to be * inherited, and link these into the new process * descriptor. */ if ((ppold = pmc_find_process_descriptor(curthread->td_proc, PMC_FLAG_NONE)) == NULL) goto done; /* nothing to do */ do_descendants = 0; for (ri = 0; ri < md->pmd_npmc; ri++) if ((pm = ppold->pp_pmcs[ri].pp_pmc) != NULL) do_descendants |= pm->pm_flags & PMC_F_DESCENDANTS; if (do_descendants == 0) /* nothing to do */ goto done; /* allocate a descriptor for the new process */ if ((ppnew = pmc_find_process_descriptor(newproc, PMC_FLAG_ALLOCATE)) == NULL) goto done; /* * Run through all PMCs that were targeting the old process * and which specified F_DESCENDANTS and attach them to the * new process. * * Log the fork event to all owners of PMCs attached to this * process, if not already logged. */ for (ri = 0; ri < md->pmd_npmc; ri++) if ((pm = ppold->pp_pmcs[ri].pp_pmc) != NULL && (pm->pm_flags & PMC_F_DESCENDANTS)) { pmc_link_target_process(pm, ppnew); po = pm->pm_owner; if (po->po_sscount == 0 && po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_procfork(po, p1->p_pid, newproc->p_pid); } /* * Now mark the new process as being tracked by this driver. */ PROC_LOCK(newproc); newproc->p_flag |= P_HWPMC; PROC_UNLOCK(newproc); done: sx_xunlock(&pmc_sx); } static void pmc_kld_load(void *arg __unused, linker_file_t lf) { struct pmc_owner *po; sx_slock(&pmc_sx); /* * Notify owners of system sampling PMCs about KLD operations. */ LIST_FOREACH(po, &pmc_ss_owners, po_ssnext) if (po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_map_in(po, (pid_t) -1, (uintfptr_t) lf->address, lf->filename); /* * TODO: Notify owners of (all) process-sampling PMCs too. */ sx_sunlock(&pmc_sx); } static void pmc_kld_unload(void *arg __unused, const char *filename __unused, caddr_t address, size_t size) { struct pmc_owner *po; sx_slock(&pmc_sx); LIST_FOREACH(po, &pmc_ss_owners, po_ssnext) if (po->po_flags & PMC_PO_OWNS_LOGFILE) pmclog_process_map_out(po, (pid_t) -1, (uintfptr_t) address, (uintfptr_t) address + size); /* * TODO: Notify owners of process-sampling PMCs. */ sx_sunlock(&pmc_sx); } /* * initialization */ static const char * pmc_name_of_pmcclass(enum pmc_class class) { switch (class) { #undef __PMC_CLASS #define __PMC_CLASS(S,V,D) \ case PMC_CLASS_##S: \ return #S; __PMC_CLASSES(); default: return (""); } } /* * Base class initializer: allocate structure and set default classes. */ struct pmc_mdep * pmc_mdep_alloc(int nclasses) { struct pmc_mdep *md; int n; /* SOFT + md classes */ n = 1 + nclasses; md = malloc(sizeof(struct pmc_mdep) + n * sizeof(struct pmc_classdep), M_PMC, M_WAITOK|M_ZERO); md->pmd_nclass = n; /* Add base class. */ pmc_soft_initialize(md); return md; } void pmc_mdep_free(struct pmc_mdep *md) { pmc_soft_finalize(md); free(md, M_PMC); } static int generic_switch_in(struct pmc_cpu *pc, struct pmc_process *pp) { (void) pc; (void) pp; return (0); } static int generic_switch_out(struct pmc_cpu *pc, struct pmc_process *pp) { (void) pc; (void) pp; return (0); } static struct pmc_mdep * pmc_generic_cpu_initialize(void) { struct pmc_mdep *md; md = pmc_mdep_alloc(0); md->pmd_cputype = PMC_CPU_GENERIC; md->pmd_pcpu_init = NULL; md->pmd_pcpu_fini = NULL; md->pmd_switch_in = generic_switch_in; md->pmd_switch_out = generic_switch_out; return (md); } static void pmc_generic_cpu_finalize(struct pmc_mdep *md) { (void) md; } static int pmc_initialize(void) { int c, cpu, error, n, ri; unsigned int maxcpu; struct pmc_binding pb; struct pmc_sample *ps; struct pmc_classdep *pcd; struct pmc_samplebuffer *sb; md = NULL; error = 0; #ifdef HWPMC_DEBUG /* parse debug flags first */ if (TUNABLE_STR_FETCH(PMC_SYSCTL_NAME_PREFIX "debugflags", pmc_debugstr, sizeof(pmc_debugstr))) pmc_debugflags_parse(pmc_debugstr, pmc_debugstr+strlen(pmc_debugstr)); #endif PMCDBG1(MOD,INI,0, "PMC Initialize (version %x)", PMC_VERSION); /* check kernel version */ if (pmc_kernel_version != PMC_VERSION) { if (pmc_kernel_version == 0) printf("hwpmc: this kernel has not been compiled with " "'options HWPMC_HOOKS'.\n"); else printf("hwpmc: kernel version (0x%x) does not match " "module version (0x%x).\n", pmc_kernel_version, PMC_VERSION); return EPROGMISMATCH; } /* * check sysctl parameters */ if (pmc_hashsize <= 0) { (void) printf("hwpmc: tunable \"hashsize\"=%d must be " "greater than zero.\n", pmc_hashsize); pmc_hashsize = PMC_HASH_SIZE; } if (pmc_nsamples <= 0 || pmc_nsamples > 65535) { (void) printf("hwpmc: tunable \"nsamples\"=%d out of " "range.\n", pmc_nsamples); pmc_nsamples = PMC_NSAMPLES; } if (pmc_callchaindepth <= 0 || pmc_callchaindepth > PMC_CALLCHAIN_DEPTH_MAX) { (void) printf("hwpmc: tunable \"callchaindepth\"=%d out of " "range - using %d.\n", pmc_callchaindepth, PMC_CALLCHAIN_DEPTH_MAX); pmc_callchaindepth = PMC_CALLCHAIN_DEPTH_MAX; } md = pmc_md_initialize(); if (md == NULL) { /* Default to generic CPU. */ md = pmc_generic_cpu_initialize(); if (md == NULL) return (ENOSYS); } KASSERT(md->pmd_nclass >= 1 && md->pmd_npmc >= 1, ("[pmc,%d] no classes or pmcs", __LINE__)); /* Compute the map from row-indices to classdep pointers. */ pmc_rowindex_to_classdep = malloc(sizeof(struct pmc_classdep *) * md->pmd_npmc, M_PMC, M_WAITOK|M_ZERO); for (n = 0; n < md->pmd_npmc; n++) pmc_rowindex_to_classdep[n] = NULL; for (ri = c = 0; c < md->pmd_nclass; c++) { pcd = &md->pmd_classdep[c]; for (n = 0; n < pcd->pcd_num; n++, ri++) pmc_rowindex_to_classdep[ri] = pcd; } KASSERT(ri == md->pmd_npmc, ("[pmc,%d] npmc miscomputed: ri=%d, md->npmc=%d", __LINE__, ri, md->pmd_npmc)); maxcpu = pmc_cpu_max(); /* allocate space for the per-cpu array */ pmc_pcpu = malloc(maxcpu * sizeof(struct pmc_cpu *), M_PMC, M_WAITOK|M_ZERO); /* per-cpu 'saved values' for managing process-mode PMCs */ pmc_pcpu_saved = malloc(sizeof(pmc_value_t) * maxcpu * md->pmd_npmc, M_PMC, M_WAITOK); /* Perform CPU-dependent initialization. */ pmc_save_cpu_binding(&pb); error = 0; for (cpu = 0; error == 0 && cpu < maxcpu; cpu++) { if (!pmc_cpu_is_active(cpu)) continue; pmc_select_cpu(cpu); pmc_pcpu[cpu] = malloc(sizeof(struct pmc_cpu) + md->pmd_npmc * sizeof(struct pmc_hw *), M_PMC, M_WAITOK|M_ZERO); if (md->pmd_pcpu_init) error = md->pmd_pcpu_init(md, cpu); for (n = 0; error == 0 && n < md->pmd_nclass; n++) error = md->pmd_classdep[n].pcd_pcpu_init(md, cpu); } pmc_restore_cpu_binding(&pb); if (error) return (error); /* allocate space for the sample array */ for (cpu = 0; cpu < maxcpu; cpu++) { if (!pmc_cpu_is_active(cpu)) continue; sb = malloc(sizeof(struct pmc_samplebuffer) + pmc_nsamples * sizeof(struct pmc_sample), M_PMC, M_WAITOK|M_ZERO); sb->ps_read = sb->ps_write = sb->ps_samples; sb->ps_fence = sb->ps_samples + pmc_nsamples; KASSERT(pmc_pcpu[cpu] != NULL, ("[pmc,%d] cpu=%d Null per-cpu data", __LINE__, cpu)); sb->ps_callchains = malloc(pmc_callchaindepth * pmc_nsamples * sizeof(uintptr_t), M_PMC, M_WAITOK|M_ZERO); for (n = 0, ps = sb->ps_samples; n < pmc_nsamples; n++, ps++) ps->ps_pc = sb->ps_callchains + (n * pmc_callchaindepth); pmc_pcpu[cpu]->pc_sb[PMC_HR] = sb; sb = malloc(sizeof(struct pmc_samplebuffer) + pmc_nsamples * sizeof(struct pmc_sample), M_PMC, M_WAITOK|M_ZERO); sb->ps_read = sb->ps_write = sb->ps_samples; sb->ps_fence = sb->ps_samples + pmc_nsamples; KASSERT(pmc_pcpu[cpu] != NULL, ("[pmc,%d] cpu=%d Null per-cpu data", __LINE__, cpu)); sb->ps_callchains = malloc(pmc_callchaindepth * pmc_nsamples * sizeof(uintptr_t), M_PMC, M_WAITOK|M_ZERO); for (n = 0, ps = sb->ps_samples; n < pmc_nsamples; n++, ps++) ps->ps_pc = sb->ps_callchains + (n * pmc_callchaindepth); pmc_pcpu[cpu]->pc_sb[PMC_SR] = sb; } /* allocate space for the row disposition array */ pmc_pmcdisp = malloc(sizeof(enum pmc_mode) * md->pmd_npmc, M_PMC, M_WAITOK|M_ZERO); /* mark all PMCs as available */ for (n = 0; n < (int) md->pmd_npmc; n++) PMC_MARK_ROW_FREE(n); /* allocate thread hash tables */ pmc_ownerhash = hashinit(pmc_hashsize, M_PMC, &pmc_ownerhashmask); pmc_processhash = hashinit(pmc_hashsize, M_PMC, &pmc_processhashmask); mtx_init(&pmc_processhash_mtx, "pmc-process-hash", "pmc-leaf", MTX_SPIN); LIST_INIT(&pmc_ss_owners); pmc_ss_count = 0; /* allocate a pool of spin mutexes */ pmc_mtxpool = mtx_pool_create("pmc-leaf", pmc_mtxpool_size, MTX_SPIN); PMCDBG4(MOD,INI,1, "pmc_ownerhash=%p, mask=0x%lx " "targethash=%p mask=0x%lx", pmc_ownerhash, pmc_ownerhashmask, pmc_processhash, pmc_processhashmask); /* register process {exit,fork,exec} handlers */ pmc_exit_tag = EVENTHANDLER_REGISTER(process_exit, pmc_process_exit, NULL, EVENTHANDLER_PRI_ANY); pmc_fork_tag = EVENTHANDLER_REGISTER(process_fork, pmc_process_fork, NULL, EVENTHANDLER_PRI_ANY); /* register kld event handlers */ pmc_kld_load_tag = EVENTHANDLER_REGISTER(kld_load, pmc_kld_load, NULL, EVENTHANDLER_PRI_ANY); pmc_kld_unload_tag = EVENTHANDLER_REGISTER(kld_unload, pmc_kld_unload, NULL, EVENTHANDLER_PRI_ANY); /* initialize logging */ pmclog_initialize(); /* set hook functions */ pmc_intr = md->pmd_intr; pmc_hook = pmc_hook_handler; if (error == 0) { printf(PMC_MODULE_NAME ":"); for (n = 0; n < (int) md->pmd_nclass; n++) { pcd = &md->pmd_classdep[n]; printf(" %s/%d/%d/0x%b", pmc_name_of_pmcclass(pcd->pcd_class), pcd->pcd_num, pcd->pcd_width, pcd->pcd_caps, "\20" "\1INT\2USR\3SYS\4EDG\5THR" "\6REA\7WRI\10INV\11QUA\12PRC" "\13TAG\14CSC"); } printf("\n"); } return (error); } /* prepare to be unloaded */ static void pmc_cleanup(void) { int c, cpu; unsigned int maxcpu; struct pmc_ownerhash *ph; struct pmc_owner *po, *tmp; struct pmc_binding pb; #ifdef HWPMC_DEBUG struct pmc_processhash *prh; #endif PMCDBG0(MOD,INI,0, "cleanup"); /* switch off sampling */ CPU_ZERO(&pmc_cpumask); pmc_intr = NULL; sx_xlock(&pmc_sx); if (pmc_hook == NULL) { /* being unloaded already */ sx_xunlock(&pmc_sx); return; } pmc_hook = NULL; /* prevent new threads from entering module */ /* deregister event handlers */ EVENTHANDLER_DEREGISTER(process_fork, pmc_fork_tag); EVENTHANDLER_DEREGISTER(process_exit, pmc_exit_tag); EVENTHANDLER_DEREGISTER(kld_load, pmc_kld_load_tag); EVENTHANDLER_DEREGISTER(kld_unload, pmc_kld_unload_tag); /* send SIGBUS to all owner threads, free up allocations */ if (pmc_ownerhash) for (ph = pmc_ownerhash; ph <= &pmc_ownerhash[pmc_ownerhashmask]; ph++) { LIST_FOREACH_SAFE(po, ph, po_next, tmp) { pmc_remove_owner(po); /* send SIGBUS to owner processes */ PMCDBG3(MOD,INI,2, "cleanup signal proc=%p " "(%d, %s)", po->po_owner, po->po_owner->p_pid, po->po_owner->p_comm); PROC_LOCK(po->po_owner); kern_psignal(po->po_owner, SIGBUS); PROC_UNLOCK(po->po_owner); pmc_destroy_owner_descriptor(po); } } /* reclaim allocated data structures */ if (pmc_mtxpool) mtx_pool_destroy(&pmc_mtxpool); mtx_destroy(&pmc_processhash_mtx); if (pmc_processhash) { #ifdef HWPMC_DEBUG struct pmc_process *pp; PMCDBG0(MOD,INI,3, "destroy process hash"); for (prh = pmc_processhash; prh <= &pmc_processhash[pmc_processhashmask]; prh++) LIST_FOREACH(pp, prh, pp_next) PMCDBG1(MOD,INI,3, "pid=%d", pp->pp_proc->p_pid); #endif hashdestroy(pmc_processhash, M_PMC, pmc_processhashmask); pmc_processhash = NULL; } if (pmc_ownerhash) { PMCDBG0(MOD,INI,3, "destroy owner hash"); hashdestroy(pmc_ownerhash, M_PMC, pmc_ownerhashmask); pmc_ownerhash = NULL; } KASSERT(LIST_EMPTY(&pmc_ss_owners), ("[pmc,%d] Global SS owner list not empty", __LINE__)); KASSERT(pmc_ss_count == 0, ("[pmc,%d] Global SS count not empty", __LINE__)); /* do processor and pmc-class dependent cleanup */ maxcpu = pmc_cpu_max(); PMCDBG0(MOD,INI,3, "md cleanup"); if (md) { pmc_save_cpu_binding(&pb); for (cpu = 0; cpu < maxcpu; cpu++) { PMCDBG2(MOD,INI,1,"pmc-cleanup cpu=%d pcs=%p", cpu, pmc_pcpu[cpu]); if (!pmc_cpu_is_active(cpu) || pmc_pcpu[cpu] == NULL) continue; pmc_select_cpu(cpu); for (c = 0; c < md->pmd_nclass; c++) md->pmd_classdep[c].pcd_pcpu_fini(md, cpu); if (md->pmd_pcpu_fini) md->pmd_pcpu_fini(md, cpu); } if (md->pmd_cputype == PMC_CPU_GENERIC) pmc_generic_cpu_finalize(md); else pmc_md_finalize(md); pmc_mdep_free(md); md = NULL; pmc_restore_cpu_binding(&pb); } /* Free per-cpu descriptors. */ for (cpu = 0; cpu < maxcpu; cpu++) { if (!pmc_cpu_is_active(cpu)) continue; KASSERT(pmc_pcpu[cpu]->pc_sb[PMC_HR] != NULL, ("[pmc,%d] Null hw cpu sample buffer cpu=%d", __LINE__, cpu)); KASSERT(pmc_pcpu[cpu]->pc_sb[PMC_SR] != NULL, ("[pmc,%d] Null sw cpu sample buffer cpu=%d", __LINE__, cpu)); free(pmc_pcpu[cpu]->pc_sb[PMC_HR]->ps_callchains, M_PMC); free(pmc_pcpu[cpu]->pc_sb[PMC_HR], M_PMC); free(pmc_pcpu[cpu]->pc_sb[PMC_SR]->ps_callchains, M_PMC); free(pmc_pcpu[cpu]->pc_sb[PMC_SR], M_PMC); free(pmc_pcpu[cpu], M_PMC); } free(pmc_pcpu, M_PMC); pmc_pcpu = NULL; free(pmc_pcpu_saved, M_PMC); pmc_pcpu_saved = NULL; if (pmc_pmcdisp) { free(pmc_pmcdisp, M_PMC); pmc_pmcdisp = NULL; } if (pmc_rowindex_to_classdep) { free(pmc_rowindex_to_classdep, M_PMC); pmc_rowindex_to_classdep = NULL; } pmclog_shutdown(); sx_xunlock(&pmc_sx); /* we are done */ } /* * The function called at load/unload. */ static int load (struct module *module __unused, int cmd, void *arg __unused) { int error; error = 0; switch (cmd) { case MOD_LOAD : /* initialize the subsystem */ error = pmc_initialize(); if (error != 0) break; PMCDBG2(MOD,INI,1, "syscall=%d maxcpu=%d", pmc_syscall_num, pmc_cpu_max()); break; case MOD_UNLOAD : case MOD_SHUTDOWN: pmc_cleanup(); PMCDBG0(MOD,INI,1, "unloaded"); break; default : error = EINVAL; /* XXX should panic(9) */ break; } return error; } /* memory pool */ MALLOC_DEFINE(M_PMC, "pmc", "Memory space for the PMC module"); Index: projects/em_mq/sys/dev/iwi/if_iwi.c =================================================================== --- projects/em_mq/sys/dev/iwi/if_iwi.c (revision 283943) +++ projects/em_mq/sys/dev/iwi/if_iwi.c (revision 283944) @@ -1,3636 +1,3636 @@ /*- * Copyright (c) 2004, 2005 * Damien Bergamini . All rights reserved. * Copyright (c) 2005-2006 Sam Leffler, Errno Consulting * Copyright (c) 2007 Andrew Thompson * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice unmodified, this list of conditions, and the following * disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); /*- * Intel(R) PRO/Wireless 2200BG/2225BG/2915ABG driver * http://www.intel.com/network/connectivity/products/wireless/prowireless_mobile.htm */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define IWI_DEBUG #ifdef IWI_DEBUG #define DPRINTF(x) do { if (iwi_debug > 0) printf x; } while (0) #define DPRINTFN(n, x) do { if (iwi_debug >= (n)) printf x; } while (0) int iwi_debug = 0; SYSCTL_INT(_debug, OID_AUTO, iwi, CTLFLAG_RW, &iwi_debug, 0, "iwi debug level"); static const char *iwi_fw_states[] = { "IDLE", /* IWI_FW_IDLE */ "LOADING", /* IWI_FW_LOADING */ "ASSOCIATING", /* IWI_FW_ASSOCIATING */ "DISASSOCIATING", /* IWI_FW_DISASSOCIATING */ "SCANNING", /* IWI_FW_SCANNING */ }; #else #define DPRINTF(x) #define DPRINTFN(n, x) #endif MODULE_DEPEND(iwi, pci, 1, 1, 1); MODULE_DEPEND(iwi, wlan, 1, 1, 1); MODULE_DEPEND(iwi, firmware, 1, 1, 1); enum { IWI_LED_TX, IWI_LED_RX, IWI_LED_POLL, }; struct iwi_ident { uint16_t vendor; uint16_t device; const char *name; }; static const struct iwi_ident iwi_ident_table[] = { { 0x8086, 0x4220, "Intel(R) PRO/Wireless 2200BG" }, { 0x8086, 0x4221, "Intel(R) PRO/Wireless 2225BG" }, { 0x8086, 0x4223, "Intel(R) PRO/Wireless 2915ABG" }, { 0x8086, 0x4224, "Intel(R) PRO/Wireless 2915ABG" }, { 0, 0, NULL } }; static struct ieee80211vap *iwi_vap_create(struct ieee80211com *, const char [IFNAMSIZ], int, enum ieee80211_opmode, int, const uint8_t [IEEE80211_ADDR_LEN], const uint8_t [IEEE80211_ADDR_LEN]); static void iwi_vap_delete(struct ieee80211vap *); static void iwi_dma_map_addr(void *, bus_dma_segment_t *, int, int); static int iwi_alloc_cmd_ring(struct iwi_softc *, struct iwi_cmd_ring *, int); static void iwi_reset_cmd_ring(struct iwi_softc *, struct iwi_cmd_ring *); static void iwi_free_cmd_ring(struct iwi_softc *, struct iwi_cmd_ring *); static int iwi_alloc_tx_ring(struct iwi_softc *, struct iwi_tx_ring *, int, bus_addr_t, bus_addr_t); static void iwi_reset_tx_ring(struct iwi_softc *, struct iwi_tx_ring *); static void iwi_free_tx_ring(struct iwi_softc *, struct iwi_tx_ring *); static int iwi_alloc_rx_ring(struct iwi_softc *, struct iwi_rx_ring *, int); static void iwi_reset_rx_ring(struct iwi_softc *, struct iwi_rx_ring *); static void iwi_free_rx_ring(struct iwi_softc *, struct iwi_rx_ring *); static struct ieee80211_node *iwi_node_alloc(struct ieee80211vap *, const uint8_t [IEEE80211_ADDR_LEN]); static void iwi_node_free(struct ieee80211_node *); static void iwi_media_status(struct ifnet *, struct ifmediareq *); static int iwi_newstate(struct ieee80211vap *, enum ieee80211_state, int); static void iwi_wme_init(struct iwi_softc *); -static int iwi_wme_setparams(struct iwi_softc *, struct ieee80211com *); +static int iwi_wme_setparams(struct iwi_softc *); static void iwi_update_wme(void *, int); static int iwi_wme_update(struct ieee80211com *); static uint16_t iwi_read_prom_word(struct iwi_softc *, uint8_t); static void iwi_frame_intr(struct iwi_softc *, struct iwi_rx_data *, int, struct iwi_frame *); static void iwi_notification_intr(struct iwi_softc *, struct iwi_notif *); static void iwi_rx_intr(struct iwi_softc *); static void iwi_tx_intr(struct iwi_softc *, struct iwi_tx_ring *); static void iwi_intr(void *); static int iwi_cmd(struct iwi_softc *, uint8_t, void *, uint8_t); static void iwi_write_ibssnode(struct iwi_softc *, const u_int8_t [], int); static int iwi_tx_start(struct ifnet *, struct mbuf *, struct ieee80211_node *, int); static int iwi_raw_xmit(struct ieee80211_node *, struct mbuf *, const struct ieee80211_bpf_params *); static void iwi_start_locked(struct ifnet *); static void iwi_start(struct ifnet *); static void iwi_watchdog(void *); static int iwi_ioctl(struct ifnet *, u_long, caddr_t); static void iwi_stop_master(struct iwi_softc *); static int iwi_reset(struct iwi_softc *); static int iwi_load_ucode(struct iwi_softc *, const struct iwi_fw *); static int iwi_load_firmware(struct iwi_softc *, const struct iwi_fw *); static void iwi_release_fw_dma(struct iwi_softc *sc); static int iwi_config(struct iwi_softc *); static int iwi_get_firmware(struct iwi_softc *, enum ieee80211_opmode); static void iwi_put_firmware(struct iwi_softc *); static void iwi_monitor_scan(void *, int); static int iwi_scanchan(struct iwi_softc *, unsigned long, int); static void iwi_scan_start(struct ieee80211com *); static void iwi_scan_end(struct ieee80211com *); static void iwi_set_channel(struct ieee80211com *); static void iwi_scan_curchan(struct ieee80211_scan_state *, unsigned long maxdwell); static void iwi_scan_mindwell(struct ieee80211_scan_state *); static int iwi_auth_and_assoc(struct iwi_softc *, struct ieee80211vap *); static void iwi_disassoc(void *, int); static int iwi_disassociate(struct iwi_softc *, int quiet); static void iwi_init_locked(struct iwi_softc *); static void iwi_init(void *); static int iwi_init_fw_dma(struct iwi_softc *, int); static void iwi_stop_locked(void *); static void iwi_stop(struct iwi_softc *); static void iwi_restart(void *, int); static int iwi_getrfkill(struct iwi_softc *); static void iwi_radio_on(void *, int); static void iwi_radio_off(void *, int); static void iwi_sysctlattach(struct iwi_softc *); static void iwi_led_event(struct iwi_softc *, int); static void iwi_ledattach(struct iwi_softc *); static int iwi_probe(device_t); static int iwi_attach(device_t); static int iwi_detach(device_t); static int iwi_shutdown(device_t); static int iwi_suspend(device_t); static int iwi_resume(device_t); static device_method_t iwi_methods[] = { /* Device interface */ DEVMETHOD(device_probe, iwi_probe), DEVMETHOD(device_attach, iwi_attach), DEVMETHOD(device_detach, iwi_detach), DEVMETHOD(device_shutdown, iwi_shutdown), DEVMETHOD(device_suspend, iwi_suspend), DEVMETHOD(device_resume, iwi_resume), DEVMETHOD_END }; static driver_t iwi_driver = { "iwi", iwi_methods, sizeof (struct iwi_softc) }; static devclass_t iwi_devclass; DRIVER_MODULE(iwi, pci, iwi_driver, iwi_devclass, NULL, NULL); MODULE_VERSION(iwi, 1); static __inline uint8_t MEM_READ_1(struct iwi_softc *sc, uint32_t addr) { CSR_WRITE_4(sc, IWI_CSR_INDIRECT_ADDR, addr); return CSR_READ_1(sc, IWI_CSR_INDIRECT_DATA); } static __inline uint32_t MEM_READ_4(struct iwi_softc *sc, uint32_t addr) { CSR_WRITE_4(sc, IWI_CSR_INDIRECT_ADDR, addr); return CSR_READ_4(sc, IWI_CSR_INDIRECT_DATA); } static int iwi_probe(device_t dev) { const struct iwi_ident *ident; for (ident = iwi_ident_table; ident->name != NULL; ident++) { if (pci_get_vendor(dev) == ident->vendor && pci_get_device(dev) == ident->device) { device_set_desc(dev, ident->name); return (BUS_PROBE_DEFAULT); } } return ENXIO; } static int iwi_attach(device_t dev) { struct iwi_softc *sc = device_get_softc(dev); struct ifnet *ifp; struct ieee80211com *ic; uint16_t val; int i, error; uint8_t bands; uint8_t macaddr[IEEE80211_ADDR_LEN]; sc->sc_dev = dev; ifp = sc->sc_ifp = if_alloc(IFT_IEEE80211); if (ifp == NULL) { device_printf(dev, "can not if_alloc()\n"); return ENXIO; } ic = ifp->if_l2com; IWI_LOCK_INIT(sc); sc->sc_unr = new_unrhdr(1, IWI_MAX_IBSSNODE-1, &sc->sc_mtx); TASK_INIT(&sc->sc_radiontask, 0, iwi_radio_on, sc); TASK_INIT(&sc->sc_radiofftask, 0, iwi_radio_off, sc); TASK_INIT(&sc->sc_restarttask, 0, iwi_restart, sc); TASK_INIT(&sc->sc_disassoctask, 0, iwi_disassoc, sc); TASK_INIT(&sc->sc_wmetask, 0, iwi_update_wme, sc); TASK_INIT(&sc->sc_monitortask, 0, iwi_monitor_scan, sc); callout_init_mtx(&sc->sc_wdtimer, &sc->sc_mtx, 0); callout_init_mtx(&sc->sc_rftimer, &sc->sc_mtx, 0); pci_write_config(dev, 0x41, 0, 1); /* enable bus-mastering */ pci_enable_busmaster(dev); i = PCIR_BAR(0); sc->mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &i, RF_ACTIVE); if (sc->mem == NULL) { device_printf(dev, "could not allocate memory resource\n"); goto fail; } sc->sc_st = rman_get_bustag(sc->mem); sc->sc_sh = rman_get_bushandle(sc->mem); i = 0; sc->irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &i, RF_ACTIVE | RF_SHAREABLE); if (sc->irq == NULL) { device_printf(dev, "could not allocate interrupt resource\n"); goto fail; } if (iwi_reset(sc) != 0) { device_printf(dev, "could not reset adapter\n"); goto fail; } /* * Allocate rings. */ if (iwi_alloc_cmd_ring(sc, &sc->cmdq, IWI_CMD_RING_COUNT) != 0) { device_printf(dev, "could not allocate Cmd ring\n"); goto fail; } for (i = 0; i < 4; i++) { error = iwi_alloc_tx_ring(sc, &sc->txq[i], IWI_TX_RING_COUNT, IWI_CSR_TX1_RIDX + i * 4, IWI_CSR_TX1_WIDX + i * 4); if (error != 0) { device_printf(dev, "could not allocate Tx ring %d\n", i+i); goto fail; } } if (iwi_alloc_rx_ring(sc, &sc->rxq, IWI_RX_RING_COUNT) != 0) { device_printf(dev, "could not allocate Rx ring\n"); goto fail; } iwi_wme_init(sc); ifp->if_softc = sc; if_initname(ifp, device_get_name(dev), device_get_unit(dev)); ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; ifp->if_init = iwi_init; ifp->if_ioctl = iwi_ioctl; ifp->if_start = iwi_start; IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); ifp->if_snd.ifq_drv_maxlen = ifqmaxlen; IFQ_SET_READY(&ifp->if_snd); ic->ic_ifp = ifp; ic->ic_softc = sc; ic->ic_name = device_get_nameunit(dev); ic->ic_opmode = IEEE80211_M_STA; ic->ic_phytype = IEEE80211_T_OFDM; /* not only, but not used */ /* set device capabilities */ ic->ic_caps = IEEE80211_C_STA /* station mode supported */ | IEEE80211_C_IBSS /* IBSS mode supported */ | IEEE80211_C_MONITOR /* monitor mode supported */ | IEEE80211_C_PMGT /* power save supported */ | IEEE80211_C_SHPREAMBLE /* short preamble supported */ | IEEE80211_C_WPA /* 802.11i */ | IEEE80211_C_WME /* 802.11e */ #if 0 | IEEE80211_C_BGSCAN /* capable of bg scanning */ #endif ; /* read MAC address from EEPROM */ val = iwi_read_prom_word(sc, IWI_EEPROM_MAC + 0); macaddr[0] = val & 0xff; macaddr[1] = val >> 8; val = iwi_read_prom_word(sc, IWI_EEPROM_MAC + 1); macaddr[2] = val & 0xff; macaddr[3] = val >> 8; val = iwi_read_prom_word(sc, IWI_EEPROM_MAC + 2); macaddr[4] = val & 0xff; macaddr[5] = val >> 8; bands = 0; setbit(&bands, IEEE80211_MODE_11B); setbit(&bands, IEEE80211_MODE_11G); if (pci_get_device(dev) >= 0x4223) setbit(&bands, IEEE80211_MODE_11A); ieee80211_init_channels(ic, NULL, &bands); ieee80211_ifattach(ic, macaddr); /* override default methods */ ic->ic_node_alloc = iwi_node_alloc; sc->sc_node_free = ic->ic_node_free; ic->ic_node_free = iwi_node_free; ic->ic_raw_xmit = iwi_raw_xmit; ic->ic_scan_start = iwi_scan_start; ic->ic_scan_end = iwi_scan_end; ic->ic_set_channel = iwi_set_channel; ic->ic_scan_curchan = iwi_scan_curchan; ic->ic_scan_mindwell = iwi_scan_mindwell; ic->ic_wme.wme_update = iwi_wme_update; ic->ic_vap_create = iwi_vap_create; ic->ic_vap_delete = iwi_vap_delete; ieee80211_radiotap_attach(ic, &sc->sc_txtap.wt_ihdr, sizeof(sc->sc_txtap), IWI_TX_RADIOTAP_PRESENT, &sc->sc_rxtap.wr_ihdr, sizeof(sc->sc_rxtap), IWI_RX_RADIOTAP_PRESENT); iwi_sysctlattach(sc); iwi_ledattach(sc); /* * Hook our interrupt after all initialization is complete. */ error = bus_setup_intr(dev, sc->irq, INTR_TYPE_NET | INTR_MPSAFE, NULL, iwi_intr, sc, &sc->sc_ih); if (error != 0) { device_printf(dev, "could not set up interrupt\n"); goto fail; } if (bootverbose) ieee80211_announce(ic); return 0; fail: /* XXX fix */ iwi_detach(dev); return ENXIO; } static int iwi_detach(device_t dev) { struct iwi_softc *sc = device_get_softc(dev); struct ifnet *ifp = sc->sc_ifp; struct ieee80211com *ic = ifp->if_l2com; bus_teardown_intr(dev, sc->irq, sc->sc_ih); /* NB: do early to drain any pending tasks */ ieee80211_draintask(ic, &sc->sc_radiontask); ieee80211_draintask(ic, &sc->sc_radiofftask); ieee80211_draintask(ic, &sc->sc_restarttask); ieee80211_draintask(ic, &sc->sc_disassoctask); ieee80211_draintask(ic, &sc->sc_monitortask); iwi_stop(sc); ieee80211_ifdetach(ic); iwi_put_firmware(sc); iwi_release_fw_dma(sc); iwi_free_cmd_ring(sc, &sc->cmdq); iwi_free_tx_ring(sc, &sc->txq[0]); iwi_free_tx_ring(sc, &sc->txq[1]); iwi_free_tx_ring(sc, &sc->txq[2]); iwi_free_tx_ring(sc, &sc->txq[3]); iwi_free_rx_ring(sc, &sc->rxq); bus_release_resource(dev, SYS_RES_IRQ, rman_get_rid(sc->irq), sc->irq); bus_release_resource(dev, SYS_RES_MEMORY, rman_get_rid(sc->mem), sc->mem); delete_unrhdr(sc->sc_unr); IWI_LOCK_DESTROY(sc); if_free(ifp); return 0; } static struct ieee80211vap * iwi_vap_create(struct ieee80211com *ic, const char name[IFNAMSIZ], int unit, enum ieee80211_opmode opmode, int flags, const uint8_t bssid[IEEE80211_ADDR_LEN], const uint8_t mac[IEEE80211_ADDR_LEN]) { struct ifnet *ifp = ic->ic_ifp; struct iwi_softc *sc = ifp->if_softc; struct iwi_vap *ivp; struct ieee80211vap *vap; int i; if (!TAILQ_EMPTY(&ic->ic_vaps)) /* only one at a time */ return NULL; /* * Get firmware image (and possibly dma memory) on mode change. */ if (iwi_get_firmware(sc, opmode)) return NULL; /* allocate DMA memory for mapping firmware image */ i = sc->fw_fw.size; if (sc->fw_boot.size > i) i = sc->fw_boot.size; /* XXX do we dma the ucode as well ? */ if (sc->fw_uc.size > i) i = sc->fw_uc.size; if (iwi_init_fw_dma(sc, i)) return NULL; ivp = (struct iwi_vap *) malloc(sizeof(struct iwi_vap), M_80211_VAP, M_NOWAIT | M_ZERO); if (ivp == NULL) return NULL; vap = &ivp->iwi_vap; ieee80211_vap_setup(ic, vap, name, unit, opmode, flags, bssid, mac); /* override the default, the setting comes from the linux driver */ vap->iv_bmissthreshold = 24; /* override with driver methods */ ivp->iwi_newstate = vap->iv_newstate; vap->iv_newstate = iwi_newstate; /* complete setup */ ieee80211_vap_attach(vap, ieee80211_media_change, iwi_media_status); ic->ic_opmode = opmode; return vap; } static void iwi_vap_delete(struct ieee80211vap *vap) { struct iwi_vap *ivp = IWI_VAP(vap); ieee80211_vap_detach(vap); free(ivp, M_80211_VAP); } static void iwi_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error) { if (error != 0) return; KASSERT(nseg == 1, ("too many DMA segments, %d should be 1", nseg)); *(bus_addr_t *)arg = segs[0].ds_addr; } static int iwi_alloc_cmd_ring(struct iwi_softc *sc, struct iwi_cmd_ring *ring, int count) { int error; ring->count = count; ring->queued = 0; ring->cur = ring->next = 0; error = bus_dma_tag_create(bus_get_dma_tag(sc->sc_dev), 4, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, count * IWI_CMD_DESC_SIZE, 1, count * IWI_CMD_DESC_SIZE, 0, NULL, NULL, &ring->desc_dmat); if (error != 0) { device_printf(sc->sc_dev, "could not create desc DMA tag\n"); goto fail; } error = bus_dmamem_alloc(ring->desc_dmat, (void **)&ring->desc, BUS_DMA_NOWAIT | BUS_DMA_ZERO, &ring->desc_map); if (error != 0) { device_printf(sc->sc_dev, "could not allocate DMA memory\n"); goto fail; } error = bus_dmamap_load(ring->desc_dmat, ring->desc_map, ring->desc, count * IWI_CMD_DESC_SIZE, iwi_dma_map_addr, &ring->physaddr, 0); if (error != 0) { device_printf(sc->sc_dev, "could not load desc DMA map\n"); goto fail; } return 0; fail: iwi_free_cmd_ring(sc, ring); return error; } static void iwi_reset_cmd_ring(struct iwi_softc *sc, struct iwi_cmd_ring *ring) { ring->queued = 0; ring->cur = ring->next = 0; } static void iwi_free_cmd_ring(struct iwi_softc *sc, struct iwi_cmd_ring *ring) { if (ring->desc != NULL) { bus_dmamap_sync(ring->desc_dmat, ring->desc_map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ring->desc_dmat, ring->desc_map); bus_dmamem_free(ring->desc_dmat, ring->desc, ring->desc_map); } if (ring->desc_dmat != NULL) bus_dma_tag_destroy(ring->desc_dmat); } static int iwi_alloc_tx_ring(struct iwi_softc *sc, struct iwi_tx_ring *ring, int count, bus_addr_t csr_ridx, bus_addr_t csr_widx) { int i, error; ring->count = count; ring->queued = 0; ring->cur = ring->next = 0; ring->csr_ridx = csr_ridx; ring->csr_widx = csr_widx; error = bus_dma_tag_create(bus_get_dma_tag(sc->sc_dev), 4, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, count * IWI_TX_DESC_SIZE, 1, count * IWI_TX_DESC_SIZE, 0, NULL, NULL, &ring->desc_dmat); if (error != 0) { device_printf(sc->sc_dev, "could not create desc DMA tag\n"); goto fail; } error = bus_dmamem_alloc(ring->desc_dmat, (void **)&ring->desc, BUS_DMA_NOWAIT | BUS_DMA_ZERO, &ring->desc_map); if (error != 0) { device_printf(sc->sc_dev, "could not allocate DMA memory\n"); goto fail; } error = bus_dmamap_load(ring->desc_dmat, ring->desc_map, ring->desc, count * IWI_TX_DESC_SIZE, iwi_dma_map_addr, &ring->physaddr, 0); if (error != 0) { device_printf(sc->sc_dev, "could not load desc DMA map\n"); goto fail; } ring->data = malloc(count * sizeof (struct iwi_tx_data), M_DEVBUF, M_NOWAIT | M_ZERO); if (ring->data == NULL) { device_printf(sc->sc_dev, "could not allocate soft data\n"); error = ENOMEM; goto fail; } error = bus_dma_tag_create(bus_get_dma_tag(sc->sc_dev), 1, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, MCLBYTES, IWI_MAX_NSEG, MCLBYTES, 0, NULL, NULL, &ring->data_dmat); if (error != 0) { device_printf(sc->sc_dev, "could not create data DMA tag\n"); goto fail; } for (i = 0; i < count; i++) { error = bus_dmamap_create(ring->data_dmat, 0, &ring->data[i].map); if (error != 0) { device_printf(sc->sc_dev, "could not create DMA map\n"); goto fail; } } return 0; fail: iwi_free_tx_ring(sc, ring); return error; } static void iwi_reset_tx_ring(struct iwi_softc *sc, struct iwi_tx_ring *ring) { struct iwi_tx_data *data; int i; for (i = 0; i < ring->count; i++) { data = &ring->data[i]; if (data->m != NULL) { bus_dmamap_sync(ring->data_dmat, data->map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ring->data_dmat, data->map); m_freem(data->m); data->m = NULL; } if (data->ni != NULL) { ieee80211_free_node(data->ni); data->ni = NULL; } } ring->queued = 0; ring->cur = ring->next = 0; } static void iwi_free_tx_ring(struct iwi_softc *sc, struct iwi_tx_ring *ring) { struct iwi_tx_data *data; int i; if (ring->desc != NULL) { bus_dmamap_sync(ring->desc_dmat, ring->desc_map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ring->desc_dmat, ring->desc_map); bus_dmamem_free(ring->desc_dmat, ring->desc, ring->desc_map); } if (ring->desc_dmat != NULL) bus_dma_tag_destroy(ring->desc_dmat); if (ring->data != NULL) { for (i = 0; i < ring->count; i++) { data = &ring->data[i]; if (data->m != NULL) { bus_dmamap_sync(ring->data_dmat, data->map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ring->data_dmat, data->map); m_freem(data->m); } if (data->ni != NULL) ieee80211_free_node(data->ni); if (data->map != NULL) bus_dmamap_destroy(ring->data_dmat, data->map); } free(ring->data, M_DEVBUF); } if (ring->data_dmat != NULL) bus_dma_tag_destroy(ring->data_dmat); } static int iwi_alloc_rx_ring(struct iwi_softc *sc, struct iwi_rx_ring *ring, int count) { struct iwi_rx_data *data; int i, error; ring->count = count; ring->cur = 0; ring->data = malloc(count * sizeof (struct iwi_rx_data), M_DEVBUF, M_NOWAIT | M_ZERO); if (ring->data == NULL) { device_printf(sc->sc_dev, "could not allocate soft data\n"); error = ENOMEM; goto fail; } error = bus_dma_tag_create(bus_get_dma_tag(sc->sc_dev), 1, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, MCLBYTES, 1, MCLBYTES, 0, NULL, NULL, &ring->data_dmat); if (error != 0) { device_printf(sc->sc_dev, "could not create data DMA tag\n"); goto fail; } for (i = 0; i < count; i++) { data = &ring->data[i]; error = bus_dmamap_create(ring->data_dmat, 0, &data->map); if (error != 0) { device_printf(sc->sc_dev, "could not create DMA map\n"); goto fail; } data->m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (data->m == NULL) { device_printf(sc->sc_dev, "could not allocate rx mbuf\n"); error = ENOMEM; goto fail; } error = bus_dmamap_load(ring->data_dmat, data->map, mtod(data->m, void *), MCLBYTES, iwi_dma_map_addr, &data->physaddr, 0); if (error != 0) { device_printf(sc->sc_dev, "could not load rx buf DMA map"); goto fail; } data->reg = IWI_CSR_RX_BASE + i * 4; } return 0; fail: iwi_free_rx_ring(sc, ring); return error; } static void iwi_reset_rx_ring(struct iwi_softc *sc, struct iwi_rx_ring *ring) { ring->cur = 0; } static void iwi_free_rx_ring(struct iwi_softc *sc, struct iwi_rx_ring *ring) { struct iwi_rx_data *data; int i; if (ring->data != NULL) { for (i = 0; i < ring->count; i++) { data = &ring->data[i]; if (data->m != NULL) { bus_dmamap_sync(ring->data_dmat, data->map, BUS_DMASYNC_POSTREAD); bus_dmamap_unload(ring->data_dmat, data->map); m_freem(data->m); } if (data->map != NULL) bus_dmamap_destroy(ring->data_dmat, data->map); } free(ring->data, M_DEVBUF); } if (ring->data_dmat != NULL) bus_dma_tag_destroy(ring->data_dmat); } static int iwi_shutdown(device_t dev) { struct iwi_softc *sc = device_get_softc(dev); iwi_stop(sc); iwi_put_firmware(sc); /* ??? XXX */ return 0; } static int iwi_suspend(device_t dev) { struct iwi_softc *sc = device_get_softc(dev); struct ieee80211com *ic = sc->sc_ifp->if_l2com; ieee80211_suspend_all(ic); return 0; } static int iwi_resume(device_t dev) { struct iwi_softc *sc = device_get_softc(dev); struct ieee80211com *ic = sc->sc_ifp->if_l2com; pci_write_config(dev, 0x41, 0, 1); ieee80211_resume_all(ic); return 0; } static struct ieee80211_node * iwi_node_alloc(struct ieee80211vap *vap, const uint8_t mac[IEEE80211_ADDR_LEN]) { struct iwi_node *in; in = malloc(sizeof (struct iwi_node), M_80211_NODE, M_NOWAIT | M_ZERO); if (in == NULL) return NULL; /* XXX assign sta table entry for adhoc */ in->in_station = -1; return &in->in_node; } static void iwi_node_free(struct ieee80211_node *ni) { struct ieee80211com *ic = ni->ni_ic; struct iwi_softc *sc = ic->ic_ifp->if_softc; struct iwi_node *in = (struct iwi_node *)ni; if (in->in_station != -1) { DPRINTF(("%s mac %6D station %u\n", __func__, ni->ni_macaddr, ":", in->in_station)); free_unr(sc->sc_unr, in->in_station); } sc->sc_node_free(ni); } /* * Convert h/w rate code to IEEE rate code. */ static int iwi_cvtrate(int iwirate) { switch (iwirate) { case IWI_RATE_DS1: return 2; case IWI_RATE_DS2: return 4; case IWI_RATE_DS5: return 11; case IWI_RATE_DS11: return 22; case IWI_RATE_OFDM6: return 12; case IWI_RATE_OFDM9: return 18; case IWI_RATE_OFDM12: return 24; case IWI_RATE_OFDM18: return 36; case IWI_RATE_OFDM24: return 48; case IWI_RATE_OFDM36: return 72; case IWI_RATE_OFDM48: return 96; case IWI_RATE_OFDM54: return 108; } return 0; } /* * The firmware automatically adapts the transmit speed. We report its current * value here. */ static void iwi_media_status(struct ifnet *ifp, struct ifmediareq *imr) { struct ieee80211vap *vap = ifp->if_softc; struct ieee80211com *ic = vap->iv_ic; struct iwi_softc *sc = ic->ic_ifp->if_softc; struct ieee80211_node *ni; /* read current transmission rate from adapter */ ni = ieee80211_ref_node(vap->iv_bss); ni->ni_txrate = iwi_cvtrate(CSR_READ_4(sc, IWI_CSR_CURRENT_TX_RATE)); ieee80211_free_node(ni); ieee80211_media_status(ifp, imr); } static int iwi_newstate(struct ieee80211vap *vap, enum ieee80211_state nstate, int arg) { struct iwi_vap *ivp = IWI_VAP(vap); struct ieee80211com *ic = vap->iv_ic; struct ifnet *ifp = ic->ic_ifp; struct iwi_softc *sc = ifp->if_softc; IWI_LOCK_DECL; DPRINTF(("%s: %s -> %s flags 0x%x\n", __func__, ieee80211_state_name[vap->iv_state], ieee80211_state_name[nstate], sc->flags)); IEEE80211_UNLOCK(ic); IWI_LOCK(sc); switch (nstate) { case IEEE80211_S_INIT: /* * NB: don't try to do this if iwi_stop_master has * shutdown the firmware and disabled interrupts. */ if (vap->iv_state == IEEE80211_S_RUN && (sc->flags & IWI_FLAG_FW_INITED)) iwi_disassociate(sc, 0); break; case IEEE80211_S_AUTH: iwi_auth_and_assoc(sc, vap); break; case IEEE80211_S_RUN: if (vap->iv_opmode == IEEE80211_M_IBSS && vap->iv_state == IEEE80211_S_SCAN) { /* * XXX when joining an ibss network we are called * with a SCAN -> RUN transition on scan complete. * Use that to call iwi_auth_and_assoc. On completing * the join we are then called again with an * AUTH -> RUN transition and we want to do nothing. * This is all totally bogus and needs to be redone. */ iwi_auth_and_assoc(sc, vap); } else if (vap->iv_opmode == IEEE80211_M_MONITOR) ieee80211_runtask(ic, &sc->sc_monitortask); break; case IEEE80211_S_ASSOC: /* * If we are transitioning from AUTH then just wait * for the ASSOC status to come back from the firmware. * Otherwise we need to issue the association request. */ if (vap->iv_state == IEEE80211_S_AUTH) break; iwi_auth_and_assoc(sc, vap); break; default: break; } IWI_UNLOCK(sc); IEEE80211_LOCK(ic); return ivp->iwi_newstate(vap, nstate, arg); } /* * WME parameters coming from IEEE 802.11e specification. These values are * already declared in ieee80211_proto.c, but they are static so they can't * be reused here. */ static const struct wmeParams iwi_wme_cck_params[WME_NUM_AC] = { { 0, 3, 5, 7, 0 }, /* WME_AC_BE */ { 0, 3, 5, 10, 0 }, /* WME_AC_BK */ { 0, 2, 4, 5, 188 }, /* WME_AC_VI */ { 0, 2, 3, 4, 102 } /* WME_AC_VO */ }; static const struct wmeParams iwi_wme_ofdm_params[WME_NUM_AC] = { { 0, 3, 4, 6, 0 }, /* WME_AC_BE */ { 0, 3, 4, 10, 0 }, /* WME_AC_BK */ { 0, 2, 3, 4, 94 }, /* WME_AC_VI */ { 0, 2, 2, 3, 47 } /* WME_AC_VO */ }; #define IWI_EXP2(v) htole16((1 << (v)) - 1) #define IWI_USEC(v) htole16(IEEE80211_TXOP_TO_US(v)) static void iwi_wme_init(struct iwi_softc *sc) { const struct wmeParams *wmep; int ac; memset(sc->wme, 0, sizeof sc->wme); for (ac = 0; ac < WME_NUM_AC; ac++) { /* set WME values for CCK modulation */ wmep = &iwi_wme_cck_params[ac]; sc->wme[1].aifsn[ac] = wmep->wmep_aifsn; sc->wme[1].cwmin[ac] = IWI_EXP2(wmep->wmep_logcwmin); sc->wme[1].cwmax[ac] = IWI_EXP2(wmep->wmep_logcwmax); sc->wme[1].burst[ac] = IWI_USEC(wmep->wmep_txopLimit); sc->wme[1].acm[ac] = wmep->wmep_acm; /* set WME values for OFDM modulation */ wmep = &iwi_wme_ofdm_params[ac]; sc->wme[2].aifsn[ac] = wmep->wmep_aifsn; sc->wme[2].cwmin[ac] = IWI_EXP2(wmep->wmep_logcwmin); sc->wme[2].cwmax[ac] = IWI_EXP2(wmep->wmep_logcwmax); sc->wme[2].burst[ac] = IWI_USEC(wmep->wmep_txopLimit); sc->wme[2].acm[ac] = wmep->wmep_acm; } } static int -iwi_wme_setparams(struct iwi_softc *sc, struct ieee80211com *ic) +iwi_wme_setparams(struct iwi_softc *sc) { + struct ieee80211com *ic = sc->sc_ifp->if_l2com; const struct wmeParams *wmep; int ac; for (ac = 0; ac < WME_NUM_AC; ac++) { /* set WME values for current operating mode */ wmep = &ic->ic_wme.wme_chanParams.cap_wmeParams[ac]; sc->wme[0].aifsn[ac] = wmep->wmep_aifsn; sc->wme[0].cwmin[ac] = IWI_EXP2(wmep->wmep_logcwmin); sc->wme[0].cwmax[ac] = IWI_EXP2(wmep->wmep_logcwmax); sc->wme[0].burst[ac] = IWI_USEC(wmep->wmep_txopLimit); sc->wme[0].acm[ac] = wmep->wmep_acm; } DPRINTF(("Setting WME parameters\n")); return iwi_cmd(sc, IWI_CMD_SET_WME_PARAMS, sc->wme, sizeof sc->wme); } #undef IWI_USEC #undef IWI_EXP2 static void iwi_update_wme(void *arg, int npending) { - struct ieee80211com *ic = arg; - struct iwi_softc *sc = ic->ic_ifp->if_softc; + struct iwi_softc *sc = arg; IWI_LOCK_DECL; IWI_LOCK(sc); - (void) iwi_wme_setparams(sc, ic); + (void) iwi_wme_setparams(sc); IWI_UNLOCK(sc); } static int iwi_wme_update(struct ieee80211com *ic) { struct iwi_softc *sc = ic->ic_ifp->if_softc; struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps); /* * We may be called to update the WME parameters in * the adapter at various places. If we're already * associated then initiate the request immediately; * otherwise we assume the params will get sent down * to the adapter as part of the work iwi_auth_and_assoc * does. */ if (vap->iv_state == IEEE80211_S_RUN) ieee80211_runtask(ic, &sc->sc_wmetask); return (0); } static int iwi_wme_setie(struct iwi_softc *sc) { struct ieee80211_wme_info wme; memset(&wme, 0, sizeof wme); wme.wme_id = IEEE80211_ELEMID_VENDOR; wme.wme_len = sizeof (struct ieee80211_wme_info) - 2; wme.wme_oui[0] = 0x00; wme.wme_oui[1] = 0x50; wme.wme_oui[2] = 0xf2; wme.wme_type = WME_OUI_TYPE; wme.wme_subtype = WME_INFO_OUI_SUBTYPE; wme.wme_version = WME_VERSION; wme.wme_info = 0; DPRINTF(("Setting WME IE (len=%u)\n", wme.wme_len)); return iwi_cmd(sc, IWI_CMD_SET_WMEIE, &wme, sizeof wme); } /* * Read 16 bits at address 'addr' from the serial EEPROM. */ static uint16_t iwi_read_prom_word(struct iwi_softc *sc, uint8_t addr) { uint32_t tmp; uint16_t val; int n; /* clock C once before the first command */ IWI_EEPROM_CTL(sc, 0); IWI_EEPROM_CTL(sc, IWI_EEPROM_S); IWI_EEPROM_CTL(sc, IWI_EEPROM_S | IWI_EEPROM_C); IWI_EEPROM_CTL(sc, IWI_EEPROM_S); /* write start bit (1) */ IWI_EEPROM_CTL(sc, IWI_EEPROM_S | IWI_EEPROM_D); IWI_EEPROM_CTL(sc, IWI_EEPROM_S | IWI_EEPROM_D | IWI_EEPROM_C); /* write READ opcode (10) */ IWI_EEPROM_CTL(sc, IWI_EEPROM_S | IWI_EEPROM_D); IWI_EEPROM_CTL(sc, IWI_EEPROM_S | IWI_EEPROM_D | IWI_EEPROM_C); IWI_EEPROM_CTL(sc, IWI_EEPROM_S); IWI_EEPROM_CTL(sc, IWI_EEPROM_S | IWI_EEPROM_C); /* write address A7-A0 */ for (n = 7; n >= 0; n--) { IWI_EEPROM_CTL(sc, IWI_EEPROM_S | (((addr >> n) & 1) << IWI_EEPROM_SHIFT_D)); IWI_EEPROM_CTL(sc, IWI_EEPROM_S | (((addr >> n) & 1) << IWI_EEPROM_SHIFT_D) | IWI_EEPROM_C); } IWI_EEPROM_CTL(sc, IWI_EEPROM_S); /* read data Q15-Q0 */ val = 0; for (n = 15; n >= 0; n--) { IWI_EEPROM_CTL(sc, IWI_EEPROM_S | IWI_EEPROM_C); IWI_EEPROM_CTL(sc, IWI_EEPROM_S); tmp = MEM_READ_4(sc, IWI_MEM_EEPROM_CTL); val |= ((tmp & IWI_EEPROM_Q) >> IWI_EEPROM_SHIFT_Q) << n; } IWI_EEPROM_CTL(sc, 0); /* clear Chip Select and clock C */ IWI_EEPROM_CTL(sc, IWI_EEPROM_S); IWI_EEPROM_CTL(sc, 0); IWI_EEPROM_CTL(sc, IWI_EEPROM_C); return val; } static void iwi_setcurchan(struct iwi_softc *sc, int chan) { struct ifnet *ifp = sc->sc_ifp; struct ieee80211com *ic = ifp->if_l2com; sc->curchan = chan; ieee80211_radiotap_chan_change(ic); } static void iwi_frame_intr(struct iwi_softc *sc, struct iwi_rx_data *data, int i, struct iwi_frame *frame) { struct ifnet *ifp = sc->sc_ifp; struct ieee80211com *ic = ifp->if_l2com; struct mbuf *mnew, *m; struct ieee80211_node *ni; int type, error, framelen; int8_t rssi, nf; IWI_LOCK_DECL; framelen = le16toh(frame->len); if (framelen < IEEE80211_MIN_LEN || framelen > MCLBYTES) { /* * XXX >MCLBYTES is bogus as it means the h/w dma'd * out of bounds; need to figure out how to limit * frame size in the firmware */ /* XXX stat */ DPRINTFN(1, ("drop rx frame len=%u chan=%u rssi=%u rssi_dbm=%u\n", le16toh(frame->len), frame->chan, frame->rssi, frame->rssi_dbm)); return; } DPRINTFN(5, ("received frame len=%u chan=%u rssi=%u rssi_dbm=%u\n", le16toh(frame->len), frame->chan, frame->rssi, frame->rssi_dbm)); if (frame->chan != sc->curchan) iwi_setcurchan(sc, frame->chan); /* * Try to allocate a new mbuf for this ring element and load it before * processing the current mbuf. If the ring element cannot be loaded, * drop the received packet and reuse the old mbuf. In the unlikely * case that the old mbuf can't be reloaded either, explicitly panic. */ mnew = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (mnew == NULL) { if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); return; } bus_dmamap_unload(sc->rxq.data_dmat, data->map); error = bus_dmamap_load(sc->rxq.data_dmat, data->map, mtod(mnew, void *), MCLBYTES, iwi_dma_map_addr, &data->physaddr, 0); if (error != 0) { m_freem(mnew); /* try to reload the old mbuf */ error = bus_dmamap_load(sc->rxq.data_dmat, data->map, mtod(data->m, void *), MCLBYTES, iwi_dma_map_addr, &data->physaddr, 0); if (error != 0) { /* very unlikely that it will fail... */ panic("%s: could not load old rx mbuf", device_get_name(sc->sc_dev)); } if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); return; } /* * New mbuf successfully loaded, update Rx ring and continue * processing. */ m = data->m; data->m = mnew; CSR_WRITE_4(sc, data->reg, data->physaddr); /* finalize mbuf */ m->m_pkthdr.rcvif = ifp; m->m_pkthdr.len = m->m_len = sizeof (struct iwi_hdr) + sizeof (struct iwi_frame) + framelen; m_adj(m, sizeof (struct iwi_hdr) + sizeof (struct iwi_frame)); rssi = frame->rssi_dbm; nf = -95; if (ieee80211_radiotap_active(ic)) { struct iwi_rx_radiotap_header *tap = &sc->sc_rxtap; tap->wr_flags = 0; tap->wr_antsignal = rssi; tap->wr_antnoise = nf; tap->wr_rate = iwi_cvtrate(frame->rate); tap->wr_antenna = frame->antenna; } IWI_UNLOCK(sc); ni = ieee80211_find_rxnode(ic, mtod(m, struct ieee80211_frame_min *)); if (ni != NULL) { type = ieee80211_input(ni, m, rssi, nf); ieee80211_free_node(ni); } else type = ieee80211_input_all(ic, m, rssi, nf); IWI_LOCK(sc); if (sc->sc_softled) { /* * Blink for any data frame. Otherwise do a * heartbeat-style blink when idle. The latter * is mainly for station mode where we depend on * periodic beacon frames to trigger the poll event. */ if (type == IEEE80211_FC0_TYPE_DATA) { sc->sc_rxrate = frame->rate; iwi_led_event(sc, IWI_LED_RX); } else if (ticks - sc->sc_ledevent >= sc->sc_ledidle) iwi_led_event(sc, IWI_LED_POLL); } } /* * Check for an association response frame to see if QoS * has been negotiated. We parse just enough to figure * out if we're supposed to use QoS. The proper solution * is to pass the frame up so ieee80211_input can do the * work but that's made hard by how things currently are * done in the driver. */ static void iwi_checkforqos(struct ieee80211vap *vap, const struct ieee80211_frame *wh, int len) { #define SUBTYPE(wh) ((wh)->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK) const uint8_t *frm, *efrm, *wme; struct ieee80211_node *ni; uint16_t capinfo, status, associd; /* NB: +8 for capinfo, status, associd, and first ie */ if (!(sizeof(*wh)+8 < len && len < IEEE80211_MAX_LEN) || SUBTYPE(wh) != IEEE80211_FC0_SUBTYPE_ASSOC_RESP) return; /* * asresp frame format * [2] capability information * [2] status * [2] association ID * [tlv] supported rates * [tlv] extended supported rates * [tlv] WME */ frm = (const uint8_t *)&wh[1]; efrm = ((const uint8_t *) wh) + len; capinfo = le16toh(*(const uint16_t *)frm); frm += 2; status = le16toh(*(const uint16_t *)frm); frm += 2; associd = le16toh(*(const uint16_t *)frm); frm += 2; wme = NULL; while (efrm - frm > 1) { IEEE80211_VERIFY_LENGTH(efrm - frm, frm[1] + 2, return); switch (*frm) { case IEEE80211_ELEMID_VENDOR: if (iswmeoui(frm)) wme = frm; break; } frm += frm[1] + 2; } ni = ieee80211_ref_node(vap->iv_bss); ni->ni_capinfo = capinfo; ni->ni_associd = associd & 0x3fff; if (wme != NULL) ni->ni_flags |= IEEE80211_NODE_QOS; else ni->ni_flags &= ~IEEE80211_NODE_QOS; ieee80211_free_node(ni); #undef SUBTYPE } static void iwi_notif_link_quality(struct iwi_softc *sc, struct iwi_notif *notif) { struct iwi_notif_link_quality *lq; int len; len = le16toh(notif->len); DPRINTFN(5, ("Notification (%u) - len=%d, sizeof=%zu\n", notif->type, len, sizeof(struct iwi_notif_link_quality) )); /* enforce length */ if (len != sizeof(struct iwi_notif_link_quality)) { DPRINTFN(5, ("Notification: (%u) too short (%d)\n", notif->type, len)); return; } lq = (struct iwi_notif_link_quality *)(notif + 1); memcpy(&sc->sc_linkqual, lq, sizeof(sc->sc_linkqual)); sc->sc_linkqual_valid = 1; } /* * Task queue callbacks for iwi_notification_intr used to avoid LOR's. */ static void iwi_notification_intr(struct iwi_softc *sc, struct iwi_notif *notif) { struct ifnet *ifp = sc->sc_ifp; struct ieee80211com *ic = ifp->if_l2com; struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps); struct iwi_notif_scan_channel *chan; struct iwi_notif_scan_complete *scan; struct iwi_notif_authentication *auth; struct iwi_notif_association *assoc; struct iwi_notif_beacon_state *beacon; switch (notif->type) { case IWI_NOTIF_TYPE_SCAN_CHANNEL: chan = (struct iwi_notif_scan_channel *)(notif + 1); DPRINTFN(3, ("Scan of channel %u complete (%u)\n", ieee80211_ieee2mhz(chan->nchan, 0), chan->nchan)); /* Reset the timer, the scan is still going */ sc->sc_state_timer = 3; break; case IWI_NOTIF_TYPE_SCAN_COMPLETE: scan = (struct iwi_notif_scan_complete *)(notif + 1); DPRINTFN(2, ("Scan completed (%u, %u)\n", scan->nchan, scan->status)); IWI_STATE_END(sc, IWI_FW_SCANNING); /* * Monitor mode works by doing a passive scan to set * the channel and enable rx. Because we don't want * to abort a scan lest the firmware crash we scan * for a short period of time and automatically restart * the scan when notified the sweep has completed. */ if (vap->iv_opmode == IEEE80211_M_MONITOR) { ieee80211_runtask(ic, &sc->sc_monitortask); break; } if (scan->status == IWI_SCAN_COMPLETED) { /* NB: don't need to defer, net80211 does it for us */ ieee80211_scan_next(vap); } break; case IWI_NOTIF_TYPE_AUTHENTICATION: auth = (struct iwi_notif_authentication *)(notif + 1); switch (auth->state) { case IWI_AUTH_SUCCESS: DPRINTFN(2, ("Authentication succeeeded\n")); ieee80211_new_state(vap, IEEE80211_S_ASSOC, -1); break; case IWI_AUTH_FAIL: /* * These are delivered as an unsolicited deauth * (e.g. due to inactivity) or in response to an * associate request. */ sc->flags &= ~IWI_FLAG_ASSOCIATED; if (vap->iv_state != IEEE80211_S_RUN) { DPRINTFN(2, ("Authentication failed\n")); vap->iv_stats.is_rx_auth_fail++; IWI_STATE_END(sc, IWI_FW_ASSOCIATING); } else { DPRINTFN(2, ("Deauthenticated\n")); vap->iv_stats.is_rx_deauth++; } ieee80211_new_state(vap, IEEE80211_S_SCAN, -1); break; case IWI_AUTH_SENT_1: case IWI_AUTH_RECV_2: case IWI_AUTH_SEQ1_PASS: break; case IWI_AUTH_SEQ1_FAIL: DPRINTFN(2, ("Initial authentication handshake failed; " "you probably need shared key\n")); vap->iv_stats.is_rx_auth_fail++; IWI_STATE_END(sc, IWI_FW_ASSOCIATING); /* XXX retry shared key when in auto */ break; default: device_printf(sc->sc_dev, "unknown authentication state %u\n", auth->state); break; } break; case IWI_NOTIF_TYPE_ASSOCIATION: assoc = (struct iwi_notif_association *)(notif + 1); switch (assoc->state) { case IWI_AUTH_SUCCESS: /* re-association, do nothing */ break; case IWI_ASSOC_SUCCESS: DPRINTFN(2, ("Association succeeded\n")); sc->flags |= IWI_FLAG_ASSOCIATED; IWI_STATE_END(sc, IWI_FW_ASSOCIATING); iwi_checkforqos(vap, (const struct ieee80211_frame *)(assoc+1), le16toh(notif->len) - sizeof(*assoc) - 1); ieee80211_new_state(vap, IEEE80211_S_RUN, -1); break; case IWI_ASSOC_INIT: sc->flags &= ~IWI_FLAG_ASSOCIATED; switch (sc->fw_state) { case IWI_FW_ASSOCIATING: DPRINTFN(2, ("Association failed\n")); IWI_STATE_END(sc, IWI_FW_ASSOCIATING); ieee80211_new_state(vap, IEEE80211_S_SCAN, -1); break; case IWI_FW_DISASSOCIATING: DPRINTFN(2, ("Dissassociated\n")); IWI_STATE_END(sc, IWI_FW_DISASSOCIATING); vap->iv_stats.is_rx_disassoc++; ieee80211_new_state(vap, IEEE80211_S_SCAN, -1); break; } break; default: device_printf(sc->sc_dev, "unknown association state %u\n", assoc->state); break; } break; case IWI_NOTIF_TYPE_BEACON: /* XXX check struct length */ beacon = (struct iwi_notif_beacon_state *)(notif + 1); DPRINTFN(5, ("Beacon state (%u, %u)\n", beacon->state, le32toh(beacon->number))); if (beacon->state == IWI_BEACON_MISS) { /* * The firmware notifies us of every beacon miss * so we need to track the count against the * configured threshold before notifying the * 802.11 layer. * XXX try to roam, drop assoc only on much higher count */ if (le32toh(beacon->number) >= vap->iv_bmissthreshold) { DPRINTF(("Beacon miss: %u >= %u\n", le32toh(beacon->number), vap->iv_bmissthreshold)); vap->iv_stats.is_beacon_miss++; /* * It's pointless to notify the 802.11 layer * as it'll try to send a probe request (which * we'll discard) and then timeout and drop us * into scan state. Instead tell the firmware * to disassociate and then on completion we'll * kick the state machine to scan. */ ieee80211_runtask(ic, &sc->sc_disassoctask); } } break; case IWI_NOTIF_TYPE_CALIBRATION: case IWI_NOTIF_TYPE_NOISE: /* XXX handle? */ DPRINTFN(5, ("Notification (%u)\n", notif->type)); break; case IWI_NOTIF_TYPE_LINK_QUALITY: iwi_notif_link_quality(sc, notif); break; default: DPRINTF(("unknown notification type %u flags 0x%x len %u\n", notif->type, notif->flags, le16toh(notif->len))); break; } } static void iwi_rx_intr(struct iwi_softc *sc) { struct iwi_rx_data *data; struct iwi_hdr *hdr; uint32_t hw; hw = CSR_READ_4(sc, IWI_CSR_RX_RIDX); for (; sc->rxq.cur != hw;) { data = &sc->rxq.data[sc->rxq.cur]; bus_dmamap_sync(sc->rxq.data_dmat, data->map, BUS_DMASYNC_POSTREAD); hdr = mtod(data->m, struct iwi_hdr *); switch (hdr->type) { case IWI_HDR_TYPE_FRAME: iwi_frame_intr(sc, data, sc->rxq.cur, (struct iwi_frame *)(hdr + 1)); break; case IWI_HDR_TYPE_NOTIF: iwi_notification_intr(sc, (struct iwi_notif *)(hdr + 1)); break; default: device_printf(sc->sc_dev, "unknown hdr type %u\n", hdr->type); } DPRINTFN(15, ("rx done idx=%u\n", sc->rxq.cur)); sc->rxq.cur = (sc->rxq.cur + 1) % IWI_RX_RING_COUNT; } /* tell the firmware what we have processed */ hw = (hw == 0) ? IWI_RX_RING_COUNT - 1 : hw - 1; CSR_WRITE_4(sc, IWI_CSR_RX_WIDX, hw); } static void iwi_tx_intr(struct iwi_softc *sc, struct iwi_tx_ring *txq) { struct ifnet *ifp = sc->sc_ifp; struct iwi_tx_data *data; uint32_t hw; hw = CSR_READ_4(sc, txq->csr_ridx); for (; txq->next != hw;) { data = &txq->data[txq->next]; bus_dmamap_sync(txq->data_dmat, data->map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(txq->data_dmat, data->map); if (data->m->m_flags & M_TXCB) ieee80211_process_callback(data->ni, data->m, 0/*XXX*/); m_freem(data->m); data->m = NULL; ieee80211_free_node(data->ni); data->ni = NULL; DPRINTFN(15, ("tx done idx=%u\n", txq->next)); if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); txq->queued--; txq->next = (txq->next + 1) % IWI_TX_RING_COUNT; } sc->sc_tx_timer = 0; ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; if (sc->sc_softled) iwi_led_event(sc, IWI_LED_TX); iwi_start_locked(ifp); } static void iwi_fatal_error_intr(struct iwi_softc *sc) { struct ifnet *ifp = sc->sc_ifp; struct ieee80211com *ic = ifp->if_l2com; struct ieee80211vap *vap = TAILQ_FIRST(&ic->ic_vaps); device_printf(sc->sc_dev, "firmware error\n"); if (vap != NULL) ieee80211_cancel_scan(vap); ieee80211_runtask(ic, &sc->sc_restarttask); sc->flags &= ~IWI_FLAG_BUSY; sc->sc_busy_timer = 0; wakeup(sc); } static void iwi_radio_off_intr(struct iwi_softc *sc) { struct ifnet *ifp = sc->sc_ifp; struct ieee80211com *ic = ifp->if_l2com; ieee80211_runtask(ic, &sc->sc_radiofftask); } static void iwi_intr(void *arg) { struct iwi_softc *sc = arg; uint32_t r; IWI_LOCK_DECL; IWI_LOCK(sc); if ((r = CSR_READ_4(sc, IWI_CSR_INTR)) == 0 || r == 0xffffffff) { IWI_UNLOCK(sc); return; } /* acknowledge interrupts */ CSR_WRITE_4(sc, IWI_CSR_INTR, r); if (r & IWI_INTR_FATAL_ERROR) { iwi_fatal_error_intr(sc); goto done; } if (r & IWI_INTR_FW_INITED) { if (!(r & (IWI_INTR_FATAL_ERROR | IWI_INTR_PARITY_ERROR))) wakeup(sc); } if (r & IWI_INTR_RADIO_OFF) iwi_radio_off_intr(sc); if (r & IWI_INTR_CMD_DONE) { sc->flags &= ~IWI_FLAG_BUSY; sc->sc_busy_timer = 0; wakeup(sc); } if (r & IWI_INTR_TX1_DONE) iwi_tx_intr(sc, &sc->txq[0]); if (r & IWI_INTR_TX2_DONE) iwi_tx_intr(sc, &sc->txq[1]); if (r & IWI_INTR_TX3_DONE) iwi_tx_intr(sc, &sc->txq[2]); if (r & IWI_INTR_TX4_DONE) iwi_tx_intr(sc, &sc->txq[3]); if (r & IWI_INTR_RX_DONE) iwi_rx_intr(sc); if (r & IWI_INTR_PARITY_ERROR) { /* XXX rate-limit */ device_printf(sc->sc_dev, "parity error\n"); } done: IWI_UNLOCK(sc); } static int iwi_cmd(struct iwi_softc *sc, uint8_t type, void *data, uint8_t len) { struct iwi_cmd_desc *desc; IWI_LOCK_ASSERT(sc); if (sc->flags & IWI_FLAG_BUSY) { device_printf(sc->sc_dev, "%s: cmd %d not sent, busy\n", __func__, type); return EAGAIN; } sc->flags |= IWI_FLAG_BUSY; sc->sc_busy_timer = 2; desc = &sc->cmdq.desc[sc->cmdq.cur]; desc->hdr.type = IWI_HDR_TYPE_COMMAND; desc->hdr.flags = IWI_HDR_FLAG_IRQ; desc->type = type; desc->len = len; memcpy(desc->data, data, len); bus_dmamap_sync(sc->cmdq.desc_dmat, sc->cmdq.desc_map, BUS_DMASYNC_PREWRITE); DPRINTFN(2, ("sending command idx=%u type=%u len=%u\n", sc->cmdq.cur, type, len)); sc->cmdq.cur = (sc->cmdq.cur + 1) % IWI_CMD_RING_COUNT; CSR_WRITE_4(sc, IWI_CSR_CMD_WIDX, sc->cmdq.cur); return msleep(sc, &sc->sc_mtx, 0, "iwicmd", hz); } static void iwi_write_ibssnode(struct iwi_softc *sc, const u_int8_t addr[IEEE80211_ADDR_LEN], int entry) { struct iwi_ibssnode node; /* write node information into NIC memory */ memset(&node, 0, sizeof node); IEEE80211_ADDR_COPY(node.bssid, addr); DPRINTF(("%s mac %6D station %u\n", __func__, node.bssid, ":", entry)); CSR_WRITE_REGION_1(sc, IWI_CSR_NODE_BASE + entry * sizeof node, (uint8_t *)&node, sizeof node); } static int iwi_tx_start(struct ifnet *ifp, struct mbuf *m0, struct ieee80211_node *ni, int ac) { struct iwi_softc *sc = ifp->if_softc; struct ieee80211vap *vap = ni->ni_vap; struct ieee80211com *ic = ni->ni_ic; struct iwi_node *in = (struct iwi_node *)ni; const struct ieee80211_frame *wh; struct ieee80211_key *k; const struct chanAccParams *cap; struct iwi_tx_ring *txq = &sc->txq[ac]; struct iwi_tx_data *data; struct iwi_tx_desc *desc; struct mbuf *mnew; bus_dma_segment_t segs[IWI_MAX_NSEG]; int error, nsegs, hdrlen, i; int ismcast, flags, xflags, staid; IWI_LOCK_ASSERT(sc); wh = mtod(m0, const struct ieee80211_frame *); /* NB: only data frames use this path */ hdrlen = ieee80211_hdrsize(wh); ismcast = IEEE80211_IS_MULTICAST(wh->i_addr1); flags = xflags = 0; if (!ismcast) flags |= IWI_DATA_FLAG_NEED_ACK; if (vap->iv_flags & IEEE80211_F_SHPREAMBLE) flags |= IWI_DATA_FLAG_SHPREAMBLE; if (IEEE80211_QOS_HAS_SEQ(wh)) { xflags |= IWI_DATA_XFLAG_QOS; cap = &ic->ic_wme.wme_chanParams; if (!cap->cap_wmeParams[ac].wmep_noackPolicy) flags &= ~IWI_DATA_FLAG_NEED_ACK; } /* * This is only used in IBSS mode where the firmware expect an index * in a h/w table instead of a destination address. */ if (vap->iv_opmode == IEEE80211_M_IBSS) { if (!ismcast) { if (in->in_station == -1) { in->in_station = alloc_unr(sc->sc_unr); if (in->in_station == -1) { /* h/w table is full */ m_freem(m0); ieee80211_free_node(ni); if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); return 0; } iwi_write_ibssnode(sc, ni->ni_macaddr, in->in_station); } staid = in->in_station; } else { /* * Multicast addresses have no associated node * so there will be no station entry. We reserve * entry 0 for one mcast address and use that. * If there are many being used this will be * expensive and we'll need to do a better job * but for now this handles the broadcast case. */ if (!IEEE80211_ADDR_EQ(wh->i_addr1, sc->sc_mcast)) { IEEE80211_ADDR_COPY(sc->sc_mcast, wh->i_addr1); iwi_write_ibssnode(sc, sc->sc_mcast, 0); } staid = 0; } } else staid = 0; if (wh->i_fc[1] & IEEE80211_FC1_PROTECTED) { k = ieee80211_crypto_encap(ni, m0); if (k == NULL) { m_freem(m0); return ENOBUFS; } /* packet header may have moved, reset our local pointer */ wh = mtod(m0, struct ieee80211_frame *); } if (ieee80211_radiotap_active_vap(vap)) { struct iwi_tx_radiotap_header *tap = &sc->sc_txtap; tap->wt_flags = 0; ieee80211_radiotap_tx(vap, m0); } data = &txq->data[txq->cur]; desc = &txq->desc[txq->cur]; /* save and trim IEEE802.11 header */ m_copydata(m0, 0, hdrlen, (caddr_t)&desc->wh); m_adj(m0, hdrlen); error = bus_dmamap_load_mbuf_sg(txq->data_dmat, data->map, m0, segs, &nsegs, 0); if (error != 0 && error != EFBIG) { device_printf(sc->sc_dev, "could not map mbuf (error %d)\n", error); m_freem(m0); return error; } if (error != 0) { mnew = m_defrag(m0, M_NOWAIT); if (mnew == NULL) { device_printf(sc->sc_dev, "could not defragment mbuf\n"); m_freem(m0); return ENOBUFS; } m0 = mnew; error = bus_dmamap_load_mbuf_sg(txq->data_dmat, data->map, m0, segs, &nsegs, 0); if (error != 0) { device_printf(sc->sc_dev, "could not map mbuf (error %d)\n", error); m_freem(m0); return error; } } data->m = m0; data->ni = ni; desc->hdr.type = IWI_HDR_TYPE_DATA; desc->hdr.flags = IWI_HDR_FLAG_IRQ; desc->station = staid; desc->cmd = IWI_DATA_CMD_TX; desc->len = htole16(m0->m_pkthdr.len); desc->flags = flags; desc->xflags = xflags; #if 0 if (vap->iv_flags & IEEE80211_F_PRIVACY) desc->wep_txkey = vap->iv_def_txkey; else #endif desc->flags |= IWI_DATA_FLAG_NO_WEP; desc->nseg = htole32(nsegs); for (i = 0; i < nsegs; i++) { desc->seg_addr[i] = htole32(segs[i].ds_addr); desc->seg_len[i] = htole16(segs[i].ds_len); } bus_dmamap_sync(txq->data_dmat, data->map, BUS_DMASYNC_PREWRITE); bus_dmamap_sync(txq->desc_dmat, txq->desc_map, BUS_DMASYNC_PREWRITE); DPRINTFN(5, ("sending data frame txq=%u idx=%u len=%u nseg=%u\n", ac, txq->cur, le16toh(desc->len), nsegs)); txq->queued++; txq->cur = (txq->cur + 1) % IWI_TX_RING_COUNT; CSR_WRITE_4(sc, txq->csr_widx, txq->cur); return 0; } static int iwi_raw_xmit(struct ieee80211_node *ni, struct mbuf *m, const struct ieee80211_bpf_params *params) { /* no support; just discard */ m_freem(m); ieee80211_free_node(ni); return 0; } static void iwi_start_locked(struct ifnet *ifp) { struct iwi_softc *sc = ifp->if_softc; struct mbuf *m; struct ieee80211_node *ni; int ac; IWI_LOCK_ASSERT(sc); if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) return; for (;;) { IFQ_DRV_DEQUEUE(&ifp->if_snd, m); if (m == NULL) break; ac = M_WME_GETAC(m); if (sc->txq[ac].queued > IWI_TX_RING_COUNT - 8) { /* there is no place left in this ring; tail drop */ /* XXX tail drop */ IFQ_DRV_PREPEND(&ifp->if_snd, m); ifp->if_drv_flags |= IFF_DRV_OACTIVE; break; } ni = (struct ieee80211_node *) m->m_pkthdr.rcvif; if (iwi_tx_start(ifp, m, ni, ac) != 0) { ieee80211_free_node(ni); if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); break; } sc->sc_tx_timer = 5; } } static void iwi_start(struct ifnet *ifp) { struct iwi_softc *sc = ifp->if_softc; IWI_LOCK_DECL; IWI_LOCK(sc); iwi_start_locked(ifp); IWI_UNLOCK(sc); } static void iwi_watchdog(void *arg) { struct iwi_softc *sc = arg; struct ifnet *ifp = sc->sc_ifp; struct ieee80211com *ic = ifp->if_l2com; IWI_LOCK_ASSERT(sc); if (sc->sc_tx_timer > 0) { if (--sc->sc_tx_timer == 0) { if_printf(ifp, "device timeout\n"); if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); ieee80211_runtask(ic, &sc->sc_restarttask); } } if (sc->sc_state_timer > 0) { if (--sc->sc_state_timer == 0) { if_printf(ifp, "firmware stuck in state %d, resetting\n", sc->fw_state); if (sc->fw_state == IWI_FW_SCANNING) { struct ieee80211com *ic = ifp->if_l2com; ieee80211_cancel_scan(TAILQ_FIRST(&ic->ic_vaps)); } ieee80211_runtask(ic, &sc->sc_restarttask); sc->sc_state_timer = 3; } } if (sc->sc_busy_timer > 0) { if (--sc->sc_busy_timer == 0) { if_printf(ifp, "firmware command timeout, resetting\n"); ieee80211_runtask(ic, &sc->sc_restarttask); } } callout_reset(&sc->sc_wdtimer, hz, iwi_watchdog, sc); } static int iwi_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) { struct iwi_softc *sc = ifp->if_softc; struct ieee80211com *ic = ifp->if_l2com; struct ifreq *ifr = (struct ifreq *) data; int error = 0, startall = 0; IWI_LOCK_DECL; switch (cmd) { case SIOCSIFFLAGS: IWI_LOCK(sc); if (ifp->if_flags & IFF_UP) { if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) { iwi_init_locked(sc); startall = 1; } } else { if (ifp->if_drv_flags & IFF_DRV_RUNNING) iwi_stop_locked(sc); } IWI_UNLOCK(sc); if (startall) ieee80211_start_all(ic); break; case SIOCGIFMEDIA: error = ifmedia_ioctl(ifp, ifr, &ic->ic_media, cmd); break; case SIOCGIFADDR: error = ether_ioctl(ifp, cmd, data); break; case SIOCGIWISTATS: IWI_LOCK(sc); /* XXX validate permissions/memory/etc? */ error = copyout(&sc->sc_linkqual, ifr->ifr_data, sizeof(struct iwi_notif_link_quality)); IWI_UNLOCK(sc); break; case SIOCZIWISTATS: IWI_LOCK(sc); memset(&sc->sc_linkqual, 0, sizeof(struct iwi_notif_link_quality)); IWI_UNLOCK(sc); error = 0; break; default: error = EINVAL; break; } return error; } static void iwi_stop_master(struct iwi_softc *sc) { uint32_t tmp; int ntries; /* disable interrupts */ CSR_WRITE_4(sc, IWI_CSR_INTR_MASK, 0); CSR_WRITE_4(sc, IWI_CSR_RST, IWI_RST_STOP_MASTER); for (ntries = 0; ntries < 5; ntries++) { if (CSR_READ_4(sc, IWI_CSR_RST) & IWI_RST_MASTER_DISABLED) break; DELAY(10); } if (ntries == 5) device_printf(sc->sc_dev, "timeout waiting for master\n"); tmp = CSR_READ_4(sc, IWI_CSR_RST); CSR_WRITE_4(sc, IWI_CSR_RST, tmp | IWI_RST_PRINCETON_RESET); sc->flags &= ~IWI_FLAG_FW_INITED; } static int iwi_reset(struct iwi_softc *sc) { uint32_t tmp; int i, ntries; iwi_stop_master(sc); tmp = CSR_READ_4(sc, IWI_CSR_CTL); CSR_WRITE_4(sc, IWI_CSR_CTL, tmp | IWI_CTL_INIT); CSR_WRITE_4(sc, IWI_CSR_READ_INT, IWI_READ_INT_INIT_HOST); /* wait for clock stabilization */ for (ntries = 0; ntries < 1000; ntries++) { if (CSR_READ_4(sc, IWI_CSR_CTL) & IWI_CTL_CLOCK_READY) break; DELAY(200); } if (ntries == 1000) { device_printf(sc->sc_dev, "timeout waiting for clock stabilization\n"); return EIO; } tmp = CSR_READ_4(sc, IWI_CSR_RST); CSR_WRITE_4(sc, IWI_CSR_RST, tmp | IWI_RST_SOFT_RESET); DELAY(10); tmp = CSR_READ_4(sc, IWI_CSR_CTL); CSR_WRITE_4(sc, IWI_CSR_CTL, tmp | IWI_CTL_INIT); /* clear NIC memory */ CSR_WRITE_4(sc, IWI_CSR_AUTOINC_ADDR, 0); for (i = 0; i < 0xc000; i++) CSR_WRITE_4(sc, IWI_CSR_AUTOINC_DATA, 0); return 0; } static const struct iwi_firmware_ohdr * iwi_setup_ofw(struct iwi_softc *sc, struct iwi_fw *fw) { const struct firmware *fp = fw->fp; const struct iwi_firmware_ohdr *hdr; if (fp->datasize < sizeof (struct iwi_firmware_ohdr)) { device_printf(sc->sc_dev, "image '%s' too small\n", fp->name); return NULL; } hdr = (const struct iwi_firmware_ohdr *)fp->data; if ((IWI_FW_GET_MAJOR(le32toh(hdr->version)) != IWI_FW_REQ_MAJOR) || (IWI_FW_GET_MINOR(le32toh(hdr->version)) != IWI_FW_REQ_MINOR)) { device_printf(sc->sc_dev, "version for '%s' %d.%d != %d.%d\n", fp->name, IWI_FW_GET_MAJOR(le32toh(hdr->version)), IWI_FW_GET_MINOR(le32toh(hdr->version)), IWI_FW_REQ_MAJOR, IWI_FW_REQ_MINOR); return NULL; } fw->data = ((const char *) fp->data) + sizeof(struct iwi_firmware_ohdr); fw->size = fp->datasize - sizeof(struct iwi_firmware_ohdr); fw->name = fp->name; return hdr; } static const struct iwi_firmware_ohdr * iwi_setup_oucode(struct iwi_softc *sc, struct iwi_fw *fw) { const struct iwi_firmware_ohdr *hdr; hdr = iwi_setup_ofw(sc, fw); if (hdr != NULL && le32toh(hdr->mode) != IWI_FW_MODE_UCODE) { device_printf(sc->sc_dev, "%s is not a ucode image\n", fw->name); hdr = NULL; } return hdr; } static void iwi_getfw(struct iwi_fw *fw, const char *fwname, struct iwi_fw *uc, const char *ucname) { if (fw->fp == NULL) fw->fp = firmware_get(fwname); /* NB: pre-3.0 ucode is packaged separately */ if (uc->fp == NULL && fw->fp != NULL && fw->fp->version < 300) uc->fp = firmware_get(ucname); } /* * Get the required firmware images if not already loaded. * Note that we hold firmware images so long as the device * is marked up in case we need to reload them on device init. * This is necessary because we re-init the device sometimes * from a context where we cannot read from the filesystem * (e.g. from the taskqueue thread when rfkill is re-enabled). * XXX return 0 on success, 1 on error. * * NB: the order of get'ing and put'ing images here is * intentional to support handling firmware images bundled * by operating mode and/or all together in one file with * the boot firmware as "master". */ static int iwi_get_firmware(struct iwi_softc *sc, enum ieee80211_opmode opmode) { const struct iwi_firmware_hdr *hdr; const struct firmware *fp; /* invalidate cached firmware on mode change */ if (sc->fw_mode != opmode) iwi_put_firmware(sc); switch (opmode) { case IEEE80211_M_STA: iwi_getfw(&sc->fw_fw, "iwi_bss", &sc->fw_uc, "iwi_ucode_bss"); break; case IEEE80211_M_IBSS: iwi_getfw(&sc->fw_fw, "iwi_ibss", &sc->fw_uc, "iwi_ucode_ibss"); break; case IEEE80211_M_MONITOR: iwi_getfw(&sc->fw_fw, "iwi_monitor", &sc->fw_uc, "iwi_ucode_monitor"); break; default: device_printf(sc->sc_dev, "unknown opmode %d\n", opmode); return EINVAL; } fp = sc->fw_fw.fp; if (fp == NULL) { device_printf(sc->sc_dev, "could not load firmware\n"); goto bad; } if (fp->version < 300) { /* * Firmware prior to 3.0 was packaged as separate * boot, firmware, and ucode images. Verify the * ucode image was read in, retrieve the boot image * if needed, and check version stamps for consistency. * The version stamps in the data are also checked * above; this is a bit paranoid but is a cheap * safeguard against mis-packaging. */ if (sc->fw_uc.fp == NULL) { device_printf(sc->sc_dev, "could not load ucode\n"); goto bad; } if (sc->fw_boot.fp == NULL) { sc->fw_boot.fp = firmware_get("iwi_boot"); if (sc->fw_boot.fp == NULL) { device_printf(sc->sc_dev, "could not load boot firmware\n"); goto bad; } } if (sc->fw_boot.fp->version != sc->fw_fw.fp->version || sc->fw_boot.fp->version != sc->fw_uc.fp->version) { device_printf(sc->sc_dev, "firmware version mismatch: " "'%s' is %d, '%s' is %d, '%s' is %d\n", sc->fw_boot.fp->name, sc->fw_boot.fp->version, sc->fw_uc.fp->name, sc->fw_uc.fp->version, sc->fw_fw.fp->name, sc->fw_fw.fp->version ); goto bad; } /* * Check and setup each image. */ if (iwi_setup_oucode(sc, &sc->fw_uc) == NULL || iwi_setup_ofw(sc, &sc->fw_boot) == NULL || iwi_setup_ofw(sc, &sc->fw_fw) == NULL) goto bad; } else { /* * Check and setup combined image. */ if (fp->datasize < sizeof(struct iwi_firmware_hdr)) { device_printf(sc->sc_dev, "image '%s' too small\n", fp->name); goto bad; } hdr = (const struct iwi_firmware_hdr *)fp->data; if (fp->datasize < sizeof(*hdr) + le32toh(hdr->bsize) + le32toh(hdr->usize) + le32toh(hdr->fsize)) { device_printf(sc->sc_dev, "image '%s' too small (2)\n", fp->name); goto bad; } sc->fw_boot.data = ((const char *) fp->data) + sizeof(*hdr); sc->fw_boot.size = le32toh(hdr->bsize); sc->fw_boot.name = fp->name; sc->fw_uc.data = sc->fw_boot.data + sc->fw_boot.size; sc->fw_uc.size = le32toh(hdr->usize); sc->fw_uc.name = fp->name; sc->fw_fw.data = sc->fw_uc.data + sc->fw_uc.size; sc->fw_fw.size = le32toh(hdr->fsize); sc->fw_fw.name = fp->name; } #if 0 device_printf(sc->sc_dev, "boot %d ucode %d fw %d bytes\n", sc->fw_boot.size, sc->fw_uc.size, sc->fw_fw.size); #endif sc->fw_mode = opmode; return 0; bad: iwi_put_firmware(sc); return 1; } static void iwi_put_fw(struct iwi_fw *fw) { if (fw->fp != NULL) { firmware_put(fw->fp, FIRMWARE_UNLOAD); fw->fp = NULL; } fw->data = NULL; fw->size = 0; fw->name = NULL; } /* * Release any cached firmware images. */ static void iwi_put_firmware(struct iwi_softc *sc) { iwi_put_fw(&sc->fw_uc); iwi_put_fw(&sc->fw_fw); iwi_put_fw(&sc->fw_boot); } static int iwi_load_ucode(struct iwi_softc *sc, const struct iwi_fw *fw) { uint32_t tmp; const uint16_t *w; const char *uc = fw->data; size_t size = fw->size; int i, ntries, error; IWI_LOCK_ASSERT(sc); error = 0; CSR_WRITE_4(sc, IWI_CSR_RST, CSR_READ_4(sc, IWI_CSR_RST) | IWI_RST_STOP_MASTER); for (ntries = 0; ntries < 5; ntries++) { if (CSR_READ_4(sc, IWI_CSR_RST) & IWI_RST_MASTER_DISABLED) break; DELAY(10); } if (ntries == 5) { device_printf(sc->sc_dev, "timeout waiting for master\n"); error = EIO; goto fail; } MEM_WRITE_4(sc, 0x3000e0, 0x80000000); DELAY(5000); tmp = CSR_READ_4(sc, IWI_CSR_RST); tmp &= ~IWI_RST_PRINCETON_RESET; CSR_WRITE_4(sc, IWI_CSR_RST, tmp); DELAY(5000); MEM_WRITE_4(sc, 0x3000e0, 0); DELAY(1000); MEM_WRITE_4(sc, IWI_MEM_EEPROM_EVENT, 1); DELAY(1000); MEM_WRITE_4(sc, IWI_MEM_EEPROM_EVENT, 0); DELAY(1000); MEM_WRITE_1(sc, 0x200000, 0x00); MEM_WRITE_1(sc, 0x200000, 0x40); DELAY(1000); /* write microcode into adapter memory */ for (w = (const uint16_t *)uc; size > 0; w++, size -= 2) MEM_WRITE_2(sc, 0x200010, htole16(*w)); MEM_WRITE_1(sc, 0x200000, 0x00); MEM_WRITE_1(sc, 0x200000, 0x80); /* wait until we get an answer */ for (ntries = 0; ntries < 100; ntries++) { if (MEM_READ_1(sc, 0x200000) & 1) break; DELAY(100); } if (ntries == 100) { device_printf(sc->sc_dev, "timeout waiting for ucode to initialize\n"); error = EIO; goto fail; } /* read the answer or the firmware will not initialize properly */ for (i = 0; i < 7; i++) MEM_READ_4(sc, 0x200004); MEM_WRITE_1(sc, 0x200000, 0x00); fail: return error; } /* macro to handle unaligned little endian data in firmware image */ #define GETLE32(p) ((p)[0] | (p)[1] << 8 | (p)[2] << 16 | (p)[3] << 24) static int iwi_load_firmware(struct iwi_softc *sc, const struct iwi_fw *fw) { u_char *p, *end; uint32_t sentinel, ctl, src, dst, sum, len, mlen, tmp; int ntries, error; IWI_LOCK_ASSERT(sc); /* copy firmware image to DMA memory */ memcpy(sc->fw_virtaddr, fw->data, fw->size); /* make sure the adapter will get up-to-date values */ bus_dmamap_sync(sc->fw_dmat, sc->fw_map, BUS_DMASYNC_PREWRITE); /* tell the adapter where the command blocks are stored */ MEM_WRITE_4(sc, 0x3000a0, 0x27000); /* * Store command blocks into adapter's internal memory using register * indirections. The adapter will read the firmware image through DMA * using information stored in command blocks. */ src = sc->fw_physaddr; p = sc->fw_virtaddr; end = p + fw->size; CSR_WRITE_4(sc, IWI_CSR_AUTOINC_ADDR, 0x27000); while (p < end) { dst = GETLE32(p); p += 4; src += 4; len = GETLE32(p); p += 4; src += 4; p += len; while (len > 0) { mlen = min(len, IWI_CB_MAXDATALEN); ctl = IWI_CB_DEFAULT_CTL | mlen; sum = ctl ^ src ^ dst; /* write a command block */ CSR_WRITE_4(sc, IWI_CSR_AUTOINC_DATA, ctl); CSR_WRITE_4(sc, IWI_CSR_AUTOINC_DATA, src); CSR_WRITE_4(sc, IWI_CSR_AUTOINC_DATA, dst); CSR_WRITE_4(sc, IWI_CSR_AUTOINC_DATA, sum); src += mlen; dst += mlen; len -= mlen; } } /* write a fictive final command block (sentinel) */ sentinel = CSR_READ_4(sc, IWI_CSR_AUTOINC_ADDR); CSR_WRITE_4(sc, IWI_CSR_AUTOINC_DATA, 0); tmp = CSR_READ_4(sc, IWI_CSR_RST); tmp &= ~(IWI_RST_MASTER_DISABLED | IWI_RST_STOP_MASTER); CSR_WRITE_4(sc, IWI_CSR_RST, tmp); /* tell the adapter to start processing command blocks */ MEM_WRITE_4(sc, 0x3000a4, 0x540100); /* wait until the adapter reaches the sentinel */ for (ntries = 0; ntries < 400; ntries++) { if (MEM_READ_4(sc, 0x3000d0) >= sentinel) break; DELAY(100); } /* sync dma, just in case */ bus_dmamap_sync(sc->fw_dmat, sc->fw_map, BUS_DMASYNC_POSTWRITE); if (ntries == 400) { device_printf(sc->sc_dev, "timeout processing command blocks for %s firmware\n", fw->name); return EIO; } /* we're done with command blocks processing */ MEM_WRITE_4(sc, 0x3000a4, 0x540c00); /* allow interrupts so we know when the firmware is ready */ CSR_WRITE_4(sc, IWI_CSR_INTR_MASK, IWI_INTR_MASK); /* tell the adapter to initialize the firmware */ CSR_WRITE_4(sc, IWI_CSR_RST, 0); tmp = CSR_READ_4(sc, IWI_CSR_CTL); CSR_WRITE_4(sc, IWI_CSR_CTL, tmp | IWI_CTL_ALLOW_STANDBY); /* wait at most one second for firmware initialization to complete */ if ((error = msleep(sc, &sc->sc_mtx, 0, "iwiinit", hz)) != 0) { device_printf(sc->sc_dev, "timeout waiting for %s firmware " "initialization to complete\n", fw->name); } return error; } static int iwi_setpowermode(struct iwi_softc *sc, struct ieee80211vap *vap) { uint32_t data; if (vap->iv_flags & IEEE80211_F_PMGTON) { /* XXX set more fine-grained operation */ data = htole32(IWI_POWER_MODE_MAX); } else data = htole32(IWI_POWER_MODE_CAM); DPRINTF(("Setting power mode to %u\n", le32toh(data))); return iwi_cmd(sc, IWI_CMD_SET_POWER_MODE, &data, sizeof data); } static int iwi_setwepkeys(struct iwi_softc *sc, struct ieee80211vap *vap) { struct iwi_wep_key wepkey; struct ieee80211_key *wk; int error, i; for (i = 0; i < IEEE80211_WEP_NKID; i++) { wk = &vap->iv_nw_keys[i]; wepkey.cmd = IWI_WEP_KEY_CMD_SETKEY; wepkey.idx = i; wepkey.len = wk->wk_keylen; memset(wepkey.key, 0, sizeof wepkey.key); memcpy(wepkey.key, wk->wk_key, wk->wk_keylen); DPRINTF(("Setting wep key index %u len %u\n", wepkey.idx, wepkey.len)); error = iwi_cmd(sc, IWI_CMD_SET_WEP_KEY, &wepkey, sizeof wepkey); if (error != 0) return error; } return 0; } static int iwi_config(struct iwi_softc *sc) { struct ifnet *ifp = sc->sc_ifp; struct ieee80211com *ic = ifp->if_l2com; struct iwi_configuration config; struct iwi_rateset rs; struct iwi_txpower power; uint32_t data; int error, i; IWI_LOCK_ASSERT(sc); DPRINTF(("Setting MAC address to %6D\n", IF_LLADDR(ifp), ":")); error = iwi_cmd(sc, IWI_CMD_SET_MAC_ADDRESS, IF_LLADDR(ifp), IEEE80211_ADDR_LEN); if (error != 0) return error; memset(&config, 0, sizeof config); config.bluetooth_coexistence = sc->bluetooth; config.silence_threshold = 0x1e; config.antenna = sc->antenna; config.multicast_enabled = 1; config.answer_pbreq = (ic->ic_opmode == IEEE80211_M_IBSS) ? 1 : 0; config.disable_unicast_decryption = 1; config.disable_multicast_decryption = 1; if (ic->ic_opmode == IEEE80211_M_MONITOR) { config.allow_invalid_frames = 1; config.allow_beacon_and_probe_resp = 1; config.allow_mgt = 1; } DPRINTF(("Configuring adapter\n")); error = iwi_cmd(sc, IWI_CMD_SET_CONFIG, &config, sizeof config); if (error != 0) return error; if (ic->ic_opmode == IEEE80211_M_IBSS) { power.mode = IWI_MODE_11B; power.nchan = 11; for (i = 0; i < 11; i++) { power.chan[i].chan = i + 1; power.chan[i].power = IWI_TXPOWER_MAX; } DPRINTF(("Setting .11b channels tx power\n")); error = iwi_cmd(sc, IWI_CMD_SET_TX_POWER, &power, sizeof power); if (error != 0) return error; power.mode = IWI_MODE_11G; DPRINTF(("Setting .11g channels tx power\n")); error = iwi_cmd(sc, IWI_CMD_SET_TX_POWER, &power, sizeof power); if (error != 0) return error; } memset(&rs, 0, sizeof rs); rs.mode = IWI_MODE_11G; rs.type = IWI_RATESET_TYPE_SUPPORTED; rs.nrates = ic->ic_sup_rates[IEEE80211_MODE_11G].rs_nrates; memcpy(rs.rates, ic->ic_sup_rates[IEEE80211_MODE_11G].rs_rates, rs.nrates); DPRINTF(("Setting .11bg supported rates (%u)\n", rs.nrates)); error = iwi_cmd(sc, IWI_CMD_SET_RATES, &rs, sizeof rs); if (error != 0) return error; memset(&rs, 0, sizeof rs); rs.mode = IWI_MODE_11A; rs.type = IWI_RATESET_TYPE_SUPPORTED; rs.nrates = ic->ic_sup_rates[IEEE80211_MODE_11A].rs_nrates; memcpy(rs.rates, ic->ic_sup_rates[IEEE80211_MODE_11A].rs_rates, rs.nrates); DPRINTF(("Setting .11a supported rates (%u)\n", rs.nrates)); error = iwi_cmd(sc, IWI_CMD_SET_RATES, &rs, sizeof rs); if (error != 0) return error; data = htole32(arc4random()); DPRINTF(("Setting initialization vector to %u\n", le32toh(data))); error = iwi_cmd(sc, IWI_CMD_SET_IV, &data, sizeof data); if (error != 0) return error; /* enable adapter */ DPRINTF(("Enabling adapter\n")); return iwi_cmd(sc, IWI_CMD_ENABLE, NULL, 0); } static __inline void set_scan_type(struct iwi_scan_ext *scan, int ix, int scan_type) { uint8_t *st = &scan->scan_type[ix / 2]; if (ix % 2) *st = (*st & 0xf0) | ((scan_type & 0xf) << 0); else *st = (*st & 0x0f) | ((scan_type & 0xf) << 4); } static int scan_type(const struct ieee80211_scan_state *ss, const struct ieee80211_channel *chan) { /* We can only set one essid for a directed scan */ if (ss->ss_nssid != 0) return IWI_SCAN_TYPE_BDIRECTED; if ((ss->ss_flags & IEEE80211_SCAN_ACTIVE) && (chan->ic_flags & IEEE80211_CHAN_PASSIVE) == 0) return IWI_SCAN_TYPE_BROADCAST; return IWI_SCAN_TYPE_PASSIVE; } static __inline int scan_band(const struct ieee80211_channel *c) { return IEEE80211_IS_CHAN_5GHZ(c) ? IWI_CHAN_5GHZ : IWI_CHAN_2GHZ; } static void iwi_monitor_scan(void *arg, int npending) { struct iwi_softc *sc = arg; IWI_LOCK_DECL; IWI_LOCK(sc); (void) iwi_scanchan(sc, 2000, 0); IWI_UNLOCK(sc); } /* * Start a scan on the current channel or all channels. */ static int iwi_scanchan(struct iwi_softc *sc, unsigned long maxdwell, int allchan) { struct ieee80211com *ic; struct ieee80211_channel *chan; struct ieee80211_scan_state *ss; struct iwi_scan_ext scan; int error = 0; IWI_LOCK_ASSERT(sc); if (sc->fw_state == IWI_FW_SCANNING) { /* * This should not happen as we only trigger scan_next after * completion */ DPRINTF(("%s: called too early - still scanning\n", __func__)); return (EBUSY); } IWI_STATE_BEGIN(sc, IWI_FW_SCANNING); ic = sc->sc_ifp->if_l2com; ss = ic->ic_scan; memset(&scan, 0, sizeof scan); scan.full_scan_index = htole32(++sc->sc_scangen); scan.dwell_time[IWI_SCAN_TYPE_PASSIVE] = htole16(maxdwell); if (ic->ic_flags_ext & IEEE80211_FEXT_BGSCAN) { /* * Use very short dwell times for when we send probe request * frames. Without this bg scans hang. Ideally this should * be handled with early-termination as done by net80211 but * that's not feasible (aborting a scan is problematic). */ scan.dwell_time[IWI_SCAN_TYPE_BROADCAST] = htole16(30); scan.dwell_time[IWI_SCAN_TYPE_BDIRECTED] = htole16(30); } else { scan.dwell_time[IWI_SCAN_TYPE_BROADCAST] = htole16(maxdwell); scan.dwell_time[IWI_SCAN_TYPE_BDIRECTED] = htole16(maxdwell); } /* We can only set one essid for a directed scan */ if (ss->ss_nssid != 0) { error = iwi_cmd(sc, IWI_CMD_SET_ESSID, ss->ss_ssid[0].ssid, ss->ss_ssid[0].len); if (error) return (error); } if (allchan) { int i, next, band, b, bstart; /* * Convert scan list to run-length encoded channel list * the firmware requires (preserving the order setup by * net80211). The first entry in each run specifies the * band and the count of items in the run. */ next = 0; /* next open slot */ bstart = 0; /* NB: not needed, silence compiler */ band = -1; /* NB: impossible value */ KASSERT(ss->ss_last > 0, ("no channels")); for (i = 0; i < ss->ss_last; i++) { chan = ss->ss_chans[i]; b = scan_band(chan); if (b != band) { if (band != -1) scan.channels[bstart] = (next - bstart) | band; /* NB: this allocates a slot for the run-len */ band = b, bstart = next++; } if (next >= IWI_SCAN_CHANNELS) { DPRINTF(("truncating scan list\n")); break; } scan.channels[next] = ieee80211_chan2ieee(ic, chan); set_scan_type(&scan, next, scan_type(ss, chan)); next++; } scan.channels[bstart] = (next - bstart) | band; } else { /* Scan the current channel only */ chan = ic->ic_curchan; scan.channels[0] = 1 | scan_band(chan); scan.channels[1] = ieee80211_chan2ieee(ic, chan); set_scan_type(&scan, 1, scan_type(ss, chan)); } #ifdef IWI_DEBUG if (iwi_debug > 0) { static const char *scantype[8] = { "PSTOP", "PASV", "DIR", "BCAST", "BDIR", "5", "6", "7" }; int i; printf("Scan request: index %u dwell %d/%d/%d\n" , le32toh(scan.full_scan_index) , le16toh(scan.dwell_time[IWI_SCAN_TYPE_PASSIVE]) , le16toh(scan.dwell_time[IWI_SCAN_TYPE_BROADCAST]) , le16toh(scan.dwell_time[IWI_SCAN_TYPE_BDIRECTED]) ); i = 0; do { int run = scan.channels[i]; if (run == 0) break; printf("Scan %d %s channels:", run & 0x3f, run & IWI_CHAN_2GHZ ? "2.4GHz" : "5GHz"); for (run &= 0x3f, i++; run > 0; run--, i++) { uint8_t type = scan.scan_type[i/2]; printf(" %u/%s", scan.channels[i], scantype[(i & 1 ? type : type>>4) & 7]); } printf("\n"); } while (i < IWI_SCAN_CHANNELS); } #endif return (iwi_cmd(sc, IWI_CMD_SCAN_EXT, &scan, sizeof scan)); } static int iwi_set_sensitivity(struct iwi_softc *sc, int8_t rssi_dbm) { struct iwi_sensitivity sens; DPRINTF(("Setting sensitivity to %d\n", rssi_dbm)); memset(&sens, 0, sizeof sens); sens.rssi = htole16(rssi_dbm); return iwi_cmd(sc, IWI_CMD_SET_SENSITIVITY, &sens, sizeof sens); } static int iwi_auth_and_assoc(struct iwi_softc *sc, struct ieee80211vap *vap) { struct ieee80211com *ic = vap->iv_ic; struct ifnet *ifp = vap->iv_ifp; struct ieee80211_node *ni; struct iwi_configuration config; struct iwi_associate *assoc = &sc->assoc; struct iwi_rateset rs; uint16_t capinfo; uint32_t data; int error, mode; IWI_LOCK_ASSERT(sc); ni = ieee80211_ref_node(vap->iv_bss); if (sc->flags & IWI_FLAG_ASSOCIATED) { DPRINTF(("Already associated\n")); return (-1); } IWI_STATE_BEGIN(sc, IWI_FW_ASSOCIATING); error = 0; mode = 0; if (IEEE80211_IS_CHAN_A(ic->ic_curchan)) mode = IWI_MODE_11A; else if (IEEE80211_IS_CHAN_G(ic->ic_curchan)) mode = IWI_MODE_11G; if (IEEE80211_IS_CHAN_B(ic->ic_curchan)) mode = IWI_MODE_11B; if (IEEE80211_IS_CHAN_2GHZ(ic->ic_curchan)) { memset(&config, 0, sizeof config); config.bluetooth_coexistence = sc->bluetooth; config.antenna = sc->antenna; config.multicast_enabled = 1; if (mode == IWI_MODE_11G) config.use_protection = 1; config.answer_pbreq = (vap->iv_opmode == IEEE80211_M_IBSS) ? 1 : 0; config.disable_unicast_decryption = 1; config.disable_multicast_decryption = 1; DPRINTF(("Configuring adapter\n")); error = iwi_cmd(sc, IWI_CMD_SET_CONFIG, &config, sizeof config); if (error != 0) goto done; } #ifdef IWI_DEBUG if (iwi_debug > 0) { printf("Setting ESSID to "); ieee80211_print_essid(ni->ni_essid, ni->ni_esslen); printf("\n"); } #endif error = iwi_cmd(sc, IWI_CMD_SET_ESSID, ni->ni_essid, ni->ni_esslen); if (error != 0) goto done; error = iwi_setpowermode(sc, vap); if (error != 0) goto done; data = htole32(vap->iv_rtsthreshold); DPRINTF(("Setting RTS threshold to %u\n", le32toh(data))); error = iwi_cmd(sc, IWI_CMD_SET_RTS_THRESHOLD, &data, sizeof data); if (error != 0) goto done; data = htole32(vap->iv_fragthreshold); DPRINTF(("Setting fragmentation threshold to %u\n", le32toh(data))); error = iwi_cmd(sc, IWI_CMD_SET_FRAG_THRESHOLD, &data, sizeof data); if (error != 0) goto done; /* the rate set has already been "negotiated" */ memset(&rs, 0, sizeof rs); rs.mode = mode; rs.type = IWI_RATESET_TYPE_NEGOTIATED; rs.nrates = ni->ni_rates.rs_nrates; if (rs.nrates > IWI_RATESET_SIZE) { DPRINTF(("Truncating negotiated rate set from %u\n", rs.nrates)); rs.nrates = IWI_RATESET_SIZE; } memcpy(rs.rates, ni->ni_rates.rs_rates, rs.nrates); DPRINTF(("Setting negotiated rates (%u)\n", rs.nrates)); error = iwi_cmd(sc, IWI_CMD_SET_RATES, &rs, sizeof rs); if (error != 0) goto done; memset(assoc, 0, sizeof *assoc); if ((vap->iv_flags & IEEE80211_F_WME) && ni->ni_ies.wme_ie != NULL) { /* NB: don't treat WME setup as failure */ - if (iwi_wme_setparams(sc, ic) == 0 && iwi_wme_setie(sc) == 0) + if (iwi_wme_setparams(sc) == 0 && iwi_wme_setie(sc) == 0) assoc->policy |= htole16(IWI_POLICY_WME); /* XXX complain on failure? */ } if (vap->iv_appie_wpa != NULL) { struct ieee80211_appie *ie = vap->iv_appie_wpa; DPRINTF(("Setting optional IE (len=%u)\n", ie->ie_len)); error = iwi_cmd(sc, IWI_CMD_SET_OPTIE, ie->ie_data, ie->ie_len); if (error != 0) goto done; } error = iwi_set_sensitivity(sc, ic->ic_node_getrssi(ni)); if (error != 0) goto done; assoc->mode = mode; assoc->chan = ic->ic_curchan->ic_ieee; /* * NB: do not arrange for shared key auth w/o privacy * (i.e. a wep key); it causes a firmware error. */ if ((vap->iv_flags & IEEE80211_F_PRIVACY) && ni->ni_authmode == IEEE80211_AUTH_SHARED) { assoc->auth = IWI_AUTH_SHARED; /* * It's possible to have privacy marked but no default * key setup. This typically is due to a user app bug * but if we blindly grab the key the firmware will * barf so avoid it for now. */ if (vap->iv_def_txkey != IEEE80211_KEYIX_NONE) assoc->auth |= vap->iv_def_txkey << 4; error = iwi_setwepkeys(sc, vap); if (error != 0) goto done; } if (vap->iv_flags & IEEE80211_F_WPA) assoc->policy |= htole16(IWI_POLICY_WPA); if (vap->iv_opmode == IEEE80211_M_IBSS && ni->ni_tstamp.tsf == 0) assoc->type = IWI_HC_IBSS_START; else assoc->type = IWI_HC_ASSOC; memcpy(assoc->tstamp, ni->ni_tstamp.data, 8); if (vap->iv_opmode == IEEE80211_M_IBSS) capinfo = IEEE80211_CAPINFO_IBSS; else capinfo = IEEE80211_CAPINFO_ESS; if (vap->iv_flags & IEEE80211_F_PRIVACY) capinfo |= IEEE80211_CAPINFO_PRIVACY; if ((ic->ic_flags & IEEE80211_F_SHPREAMBLE) && IEEE80211_IS_CHAN_2GHZ(ic->ic_curchan)) capinfo |= IEEE80211_CAPINFO_SHORT_PREAMBLE; if (ni->ni_capinfo & IEEE80211_CAPINFO_SHORT_SLOTTIME) capinfo |= IEEE80211_CAPINFO_SHORT_SLOTTIME; assoc->capinfo = htole16(capinfo); assoc->lintval = htole16(ic->ic_lintval); assoc->intval = htole16(ni->ni_intval); IEEE80211_ADDR_COPY(assoc->bssid, ni->ni_bssid); if (vap->iv_opmode == IEEE80211_M_IBSS) IEEE80211_ADDR_COPY(assoc->dst, ifp->if_broadcastaddr); else IEEE80211_ADDR_COPY(assoc->dst, ni->ni_bssid); DPRINTF(("%s bssid %6D dst %6D channel %u policy 0x%x " "auth %u capinfo 0x%x lintval %u bintval %u\n", assoc->type == IWI_HC_IBSS_START ? "Start" : "Join", assoc->bssid, ":", assoc->dst, ":", assoc->chan, le16toh(assoc->policy), assoc->auth, le16toh(assoc->capinfo), le16toh(assoc->lintval), le16toh(assoc->intval))); error = iwi_cmd(sc, IWI_CMD_ASSOCIATE, assoc, sizeof *assoc); done: ieee80211_free_node(ni); if (error) IWI_STATE_END(sc, IWI_FW_ASSOCIATING); return (error); } static void iwi_disassoc(void *arg, int pending) { struct iwi_softc *sc = arg; IWI_LOCK_DECL; IWI_LOCK(sc); iwi_disassociate(sc, 0); IWI_UNLOCK(sc); } static int iwi_disassociate(struct iwi_softc *sc, int quiet) { struct iwi_associate *assoc = &sc->assoc; if ((sc->flags & IWI_FLAG_ASSOCIATED) == 0) { DPRINTF(("Not associated\n")); return (-1); } IWI_STATE_BEGIN(sc, IWI_FW_DISASSOCIATING); if (quiet) assoc->type = IWI_HC_DISASSOC_QUIET; else assoc->type = IWI_HC_DISASSOC; DPRINTF(("Trying to disassociate from %6D channel %u\n", assoc->bssid, ":", assoc->chan)); return iwi_cmd(sc, IWI_CMD_ASSOCIATE, assoc, sizeof *assoc); } /* * release dma resources for the firmware */ static void iwi_release_fw_dma(struct iwi_softc *sc) { if (sc->fw_flags & IWI_FW_HAVE_PHY) bus_dmamap_unload(sc->fw_dmat, sc->fw_map); if (sc->fw_flags & IWI_FW_HAVE_MAP) bus_dmamem_free(sc->fw_dmat, sc->fw_virtaddr, sc->fw_map); if (sc->fw_flags & IWI_FW_HAVE_DMAT) bus_dma_tag_destroy(sc->fw_dmat); sc->fw_flags = 0; sc->fw_dma_size = 0; sc->fw_dmat = NULL; sc->fw_map = NULL; sc->fw_physaddr = 0; sc->fw_virtaddr = NULL; } /* * allocate the dma descriptor for the firmware. * Return 0 on success, 1 on error. * Must be called unlocked, protected by IWI_FLAG_FW_LOADING. */ static int iwi_init_fw_dma(struct iwi_softc *sc, int size) { if (sc->fw_dma_size >= size) return 0; if (bus_dma_tag_create(bus_get_dma_tag(sc->sc_dev), 4, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, size, 1, size, 0, NULL, NULL, &sc->fw_dmat) != 0) { device_printf(sc->sc_dev, "could not create firmware DMA tag\n"); goto error; } sc->fw_flags |= IWI_FW_HAVE_DMAT; if (bus_dmamem_alloc(sc->fw_dmat, &sc->fw_virtaddr, 0, &sc->fw_map) != 0) { device_printf(sc->sc_dev, "could not allocate firmware DMA memory\n"); goto error; } sc->fw_flags |= IWI_FW_HAVE_MAP; if (bus_dmamap_load(sc->fw_dmat, sc->fw_map, sc->fw_virtaddr, size, iwi_dma_map_addr, &sc->fw_physaddr, 0) != 0) { device_printf(sc->sc_dev, "could not load firmware DMA map\n"); goto error; } sc->fw_flags |= IWI_FW_HAVE_PHY; sc->fw_dma_size = size; return 0; error: iwi_release_fw_dma(sc); return 1; } static void iwi_init_locked(struct iwi_softc *sc) { struct ifnet *ifp = sc->sc_ifp; struct iwi_rx_data *data; int i; IWI_LOCK_ASSERT(sc); if (sc->fw_state == IWI_FW_LOADING) { device_printf(sc->sc_dev, "%s: already loading\n", __func__); return; /* XXX: condvar? */ } iwi_stop_locked(sc); IWI_STATE_BEGIN(sc, IWI_FW_LOADING); if (iwi_reset(sc) != 0) { device_printf(sc->sc_dev, "could not reset adapter\n"); goto fail; } if (iwi_load_firmware(sc, &sc->fw_boot) != 0) { device_printf(sc->sc_dev, "could not load boot firmware %s\n", sc->fw_boot.name); goto fail; } if (iwi_load_ucode(sc, &sc->fw_uc) != 0) { device_printf(sc->sc_dev, "could not load microcode %s\n", sc->fw_uc.name); goto fail; } iwi_stop_master(sc); CSR_WRITE_4(sc, IWI_CSR_CMD_BASE, sc->cmdq.physaddr); CSR_WRITE_4(sc, IWI_CSR_CMD_SIZE, sc->cmdq.count); CSR_WRITE_4(sc, IWI_CSR_CMD_WIDX, sc->cmdq.cur); CSR_WRITE_4(sc, IWI_CSR_TX1_BASE, sc->txq[0].physaddr); CSR_WRITE_4(sc, IWI_CSR_TX1_SIZE, sc->txq[0].count); CSR_WRITE_4(sc, IWI_CSR_TX1_WIDX, sc->txq[0].cur); CSR_WRITE_4(sc, IWI_CSR_TX2_BASE, sc->txq[1].physaddr); CSR_WRITE_4(sc, IWI_CSR_TX2_SIZE, sc->txq[1].count); CSR_WRITE_4(sc, IWI_CSR_TX2_WIDX, sc->txq[1].cur); CSR_WRITE_4(sc, IWI_CSR_TX3_BASE, sc->txq[2].physaddr); CSR_WRITE_4(sc, IWI_CSR_TX3_SIZE, sc->txq[2].count); CSR_WRITE_4(sc, IWI_CSR_TX3_WIDX, sc->txq[2].cur); CSR_WRITE_4(sc, IWI_CSR_TX4_BASE, sc->txq[3].physaddr); CSR_WRITE_4(sc, IWI_CSR_TX4_SIZE, sc->txq[3].count); CSR_WRITE_4(sc, IWI_CSR_TX4_WIDX, sc->txq[3].cur); for (i = 0; i < sc->rxq.count; i++) { data = &sc->rxq.data[i]; CSR_WRITE_4(sc, data->reg, data->physaddr); } CSR_WRITE_4(sc, IWI_CSR_RX_WIDX, sc->rxq.count - 1); if (iwi_load_firmware(sc, &sc->fw_fw) != 0) { device_printf(sc->sc_dev, "could not load main firmware %s\n", sc->fw_fw.name); goto fail; } sc->flags |= IWI_FLAG_FW_INITED; IWI_STATE_END(sc, IWI_FW_LOADING); if (iwi_config(sc) != 0) { device_printf(sc->sc_dev, "unable to enable adapter\n"); goto fail2; } callout_reset(&sc->sc_wdtimer, hz, iwi_watchdog, sc); ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; ifp->if_drv_flags |= IFF_DRV_RUNNING; return; fail: IWI_STATE_END(sc, IWI_FW_LOADING); fail2: iwi_stop_locked(sc); } static void iwi_init(void *priv) { struct iwi_softc *sc = priv; struct ifnet *ifp = sc->sc_ifp; struct ieee80211com *ic = ifp->if_l2com; IWI_LOCK_DECL; IWI_LOCK(sc); iwi_init_locked(sc); IWI_UNLOCK(sc); if (ifp->if_drv_flags & IFF_DRV_RUNNING) ieee80211_start_all(ic); } static void iwi_stop_locked(void *priv) { struct iwi_softc *sc = priv; struct ifnet *ifp = sc->sc_ifp; IWI_LOCK_ASSERT(sc); ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE); if (sc->sc_softled) { callout_stop(&sc->sc_ledtimer); sc->sc_blinking = 0; } callout_stop(&sc->sc_wdtimer); callout_stop(&sc->sc_rftimer); iwi_stop_master(sc); CSR_WRITE_4(sc, IWI_CSR_RST, IWI_RST_SOFT_RESET); /* reset rings */ iwi_reset_cmd_ring(sc, &sc->cmdq); iwi_reset_tx_ring(sc, &sc->txq[0]); iwi_reset_tx_ring(sc, &sc->txq[1]); iwi_reset_tx_ring(sc, &sc->txq[2]); iwi_reset_tx_ring(sc, &sc->txq[3]); iwi_reset_rx_ring(sc, &sc->rxq); sc->sc_tx_timer = 0; sc->sc_state_timer = 0; sc->sc_busy_timer = 0; sc->flags &= ~(IWI_FLAG_BUSY | IWI_FLAG_ASSOCIATED); sc->fw_state = IWI_FW_IDLE; wakeup(sc); } static void iwi_stop(struct iwi_softc *sc) { IWI_LOCK_DECL; IWI_LOCK(sc); iwi_stop_locked(sc); IWI_UNLOCK(sc); } static void iwi_restart(void *arg, int npending) { struct iwi_softc *sc = arg; iwi_init(sc); } /* * Return whether or not the radio is enabled in hardware * (i.e. the rfkill switch is "off"). */ static int iwi_getrfkill(struct iwi_softc *sc) { return (CSR_READ_4(sc, IWI_CSR_IO) & IWI_IO_RADIO_ENABLED) == 0; } static void iwi_radio_on(void *arg, int pending) { struct iwi_softc *sc = arg; struct ieee80211com *ic = sc->sc_ifp->if_l2com; device_printf(sc->sc_dev, "radio turned on\n"); iwi_init(sc); ieee80211_notify_radio(ic, 1); } static void iwi_rfkill_poll(void *arg) { struct iwi_softc *sc = arg; IWI_LOCK_ASSERT(sc); /* * Check for a change in rfkill state. We get an * interrupt when a radio is disabled but not when * it is enabled so we must poll for the latter. */ if (!iwi_getrfkill(sc)) { struct ifnet *ifp = sc->sc_ifp; struct ieee80211com *ic = ifp->if_l2com; ieee80211_runtask(ic, &sc->sc_radiontask); return; } callout_reset(&sc->sc_rftimer, 2*hz, iwi_rfkill_poll, sc); } static void iwi_radio_off(void *arg, int pending) { struct iwi_softc *sc = arg; struct ieee80211com *ic = sc->sc_ifp->if_l2com; IWI_LOCK_DECL; device_printf(sc->sc_dev, "radio turned off\n"); ieee80211_notify_radio(ic, 0); IWI_LOCK(sc); iwi_stop_locked(sc); iwi_rfkill_poll(sc); IWI_UNLOCK(sc); } static int iwi_sysctl_stats(SYSCTL_HANDLER_ARGS) { struct iwi_softc *sc = arg1; uint32_t size, buf[128]; memset(buf, 0, sizeof buf); if (!(sc->flags & IWI_FLAG_FW_INITED)) return SYSCTL_OUT(req, buf, sizeof buf); size = min(CSR_READ_4(sc, IWI_CSR_TABLE0_SIZE), 128 - 1); CSR_READ_REGION_4(sc, IWI_CSR_TABLE0_BASE, &buf[1], size); return SYSCTL_OUT(req, buf, size); } static int iwi_sysctl_radio(SYSCTL_HANDLER_ARGS) { struct iwi_softc *sc = arg1; int val = !iwi_getrfkill(sc); return SYSCTL_OUT(req, &val, sizeof val); } /* * Add sysctl knobs. */ static void iwi_sysctlattach(struct iwi_softc *sc) { struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(sc->sc_dev); struct sysctl_oid *tree = device_get_sysctl_tree(sc->sc_dev); SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(tree), OID_AUTO, "radio", CTLTYPE_INT | CTLFLAG_RD, sc, 0, iwi_sysctl_radio, "I", "radio transmitter switch state (0=off, 1=on)"); SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(tree), OID_AUTO, "stats", CTLTYPE_OPAQUE | CTLFLAG_RD, sc, 0, iwi_sysctl_stats, "S", "statistics"); sc->bluetooth = 0; SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO, "bluetooth", CTLFLAG_RW, &sc->bluetooth, 0, "bluetooth coexistence"); sc->antenna = IWI_ANTENNA_AUTO; SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO, "antenna", CTLFLAG_RW, &sc->antenna, 0, "antenna (0=auto)"); } /* * LED support. * * Different cards have different capabilities. Some have three * led's while others have only one. The linux ipw driver defines * led's for link state (associated or not), band (11a, 11g, 11b), * and for link activity. We use one led and vary the blink rate * according to the tx/rx traffic a la the ath driver. */ static __inline uint32_t iwi_toggle_event(uint32_t r) { return r &~ (IWI_RST_STANDBY | IWI_RST_GATE_ODMA | IWI_RST_GATE_IDMA | IWI_RST_GATE_ADMA); } static uint32_t iwi_read_event(struct iwi_softc *sc) { return MEM_READ_4(sc, IWI_MEM_EEPROM_EVENT); } static void iwi_write_event(struct iwi_softc *sc, uint32_t v) { MEM_WRITE_4(sc, IWI_MEM_EEPROM_EVENT, v); } static void iwi_led_done(void *arg) { struct iwi_softc *sc = arg; sc->sc_blinking = 0; } /* * Turn the activity LED off: flip the pin and then set a timer so no * update will happen for the specified duration. */ static void iwi_led_off(void *arg) { struct iwi_softc *sc = arg; uint32_t v; v = iwi_read_event(sc); v &= ~sc->sc_ledpin; iwi_write_event(sc, iwi_toggle_event(v)); callout_reset(&sc->sc_ledtimer, sc->sc_ledoff, iwi_led_done, sc); } /* * Blink the LED according to the specified on/off times. */ static void iwi_led_blink(struct iwi_softc *sc, int on, int off) { uint32_t v; v = iwi_read_event(sc); v |= sc->sc_ledpin; iwi_write_event(sc, iwi_toggle_event(v)); sc->sc_blinking = 1; sc->sc_ledoff = off; callout_reset(&sc->sc_ledtimer, on, iwi_led_off, sc); } static void iwi_led_event(struct iwi_softc *sc, int event) { #define N(a) (sizeof(a)/sizeof(a[0])) /* NB: on/off times from the Atheros NDIS driver, w/ permission */ static const struct { u_int rate; /* tx/rx iwi rate */ u_int16_t timeOn; /* LED on time (ms) */ u_int16_t timeOff; /* LED off time (ms) */ } blinkrates[] = { { IWI_RATE_OFDM54, 40, 10 }, { IWI_RATE_OFDM48, 44, 11 }, { IWI_RATE_OFDM36, 50, 13 }, { IWI_RATE_OFDM24, 57, 14 }, { IWI_RATE_OFDM18, 67, 16 }, { IWI_RATE_OFDM12, 80, 20 }, { IWI_RATE_DS11, 100, 25 }, { IWI_RATE_OFDM9, 133, 34 }, { IWI_RATE_OFDM6, 160, 40 }, { IWI_RATE_DS5, 200, 50 }, { 6, 240, 58 }, /* XXX 3Mb/s if it existed */ { IWI_RATE_DS2, 267, 66 }, { IWI_RATE_DS1, 400, 100 }, { 0, 500, 130 }, /* unknown rate/polling */ }; uint32_t txrate; int j = 0; /* XXX silence compiler */ sc->sc_ledevent = ticks; /* time of last event */ if (sc->sc_blinking) /* don't interrupt active blink */ return; switch (event) { case IWI_LED_POLL: j = N(blinkrates)-1; break; case IWI_LED_TX: /* read current transmission rate from adapter */ txrate = CSR_READ_4(sc, IWI_CSR_CURRENT_TX_RATE); if (blinkrates[sc->sc_txrix].rate != txrate) { for (j = 0; j < N(blinkrates)-1; j++) if (blinkrates[j].rate == txrate) break; sc->sc_txrix = j; } else j = sc->sc_txrix; break; case IWI_LED_RX: if (blinkrates[sc->sc_rxrix].rate != sc->sc_rxrate) { for (j = 0; j < N(blinkrates)-1; j++) if (blinkrates[j].rate == sc->sc_rxrate) break; sc->sc_rxrix = j; } else j = sc->sc_rxrix; break; } /* XXX beware of overflow */ iwi_led_blink(sc, (blinkrates[j].timeOn * hz) / 1000, (blinkrates[j].timeOff * hz) / 1000); #undef N } static int iwi_sysctl_softled(SYSCTL_HANDLER_ARGS) { struct iwi_softc *sc = arg1; int softled = sc->sc_softled; int error; error = sysctl_handle_int(oidp, &softled, 0, req); if (error || !req->newptr) return error; softled = (softled != 0); if (softled != sc->sc_softled) { if (softled) { uint32_t v = iwi_read_event(sc); v &= ~sc->sc_ledpin; iwi_write_event(sc, iwi_toggle_event(v)); } sc->sc_softled = softled; } return 0; } static void iwi_ledattach(struct iwi_softc *sc) { struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(sc->sc_dev); struct sysctl_oid *tree = device_get_sysctl_tree(sc->sc_dev); sc->sc_blinking = 0; sc->sc_ledstate = 1; sc->sc_ledidle = (2700*hz)/1000; /* 2.7sec */ callout_init_mtx(&sc->sc_ledtimer, &sc->sc_mtx, 0); SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(tree), OID_AUTO, "softled", CTLTYPE_INT | CTLFLAG_RW, sc, 0, iwi_sysctl_softled, "I", "enable/disable software LED support"); SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO, "ledpin", CTLFLAG_RW, &sc->sc_ledpin, 0, "pin setting to turn activity LED on"); SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO, "ledidle", CTLFLAG_RW, &sc->sc_ledidle, 0, "idle time for inactivity LED (ticks)"); /* XXX for debugging */ SYSCTL_ADD_UINT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO, "nictype", CTLFLAG_RD, &sc->sc_nictype, 0, "NIC type from EEPROM"); sc->sc_ledpin = IWI_RST_LED_ACTIVITY; sc->sc_softled = 1; sc->sc_nictype = (iwi_read_prom_word(sc, IWI_EEPROM_NIC) >> 8) & 0xff; if (sc->sc_nictype == 1) { /* * NB: led's are reversed. */ sc->sc_ledpin = IWI_RST_LED_ASSOCIATED; } } static void iwi_scan_start(struct ieee80211com *ic) { /* ignore */ } static void iwi_set_channel(struct ieee80211com *ic) { struct ifnet *ifp = ic->ic_ifp; struct iwi_softc *sc = ifp->if_softc; if (sc->fw_state == IWI_FW_IDLE) iwi_setcurchan(sc, ic->ic_curchan->ic_ieee); } static void iwi_scan_curchan(struct ieee80211_scan_state *ss, unsigned long maxdwell) { struct ieee80211vap *vap = ss->ss_vap; struct ifnet *ifp = vap->iv_ic->ic_ifp; struct iwi_softc *sc = ifp->if_softc; IWI_LOCK_DECL; IWI_LOCK(sc); if (iwi_scanchan(sc, maxdwell, 0)) ieee80211_cancel_scan(vap); IWI_UNLOCK(sc); } static void iwi_scan_mindwell(struct ieee80211_scan_state *ss) { /* NB: don't try to abort scan; wait for firmware to finish */ } static void iwi_scan_end(struct ieee80211com *ic) { struct ifnet *ifp = ic->ic_ifp; struct iwi_softc *sc = ifp->if_softc; IWI_LOCK_DECL; IWI_LOCK(sc); sc->flags &= ~IWI_FLAG_CHANNEL_SCAN; /* NB: make sure we're still scanning */ if (sc->fw_state == IWI_FW_SCANNING) iwi_cmd(sc, IWI_CMD_ABORT_SCAN, NULL, 0); IWI_UNLOCK(sc); } Index: projects/em_mq/sys/fs/procfs/procfs_map.c =================================================================== --- projects/em_mq/sys/fs/procfs/procfs_map.c (revision 283943) +++ projects/em_mq/sys/fs/procfs/procfs_map.c (revision 283944) @@ -1,244 +1,250 @@ /*- * Copyright (c) 1993 Jan-Simon Pendry * Copyright (c) 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Jan-Simon Pendry. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)procfs_status.c 8.3 (Berkeley) 2/17/94 * * $FreeBSD$ */ #include "opt_compat.h" #include #include #include #include #include #include #include #include #include #include #ifdef COMPAT_FREEBSD32 #include #endif #include #include #include #include #include #include #include #include #include #include #define MEBUFFERSIZE 256 /* * The map entries can *almost* be read with programs like cat. However, * large maps need special programs to read. It is not easy to implement * a program that can sense the required size of the buffer, and then * subsequently do a read with the appropriate size. This operation cannot * be atomic. The best that we can do is to allow the program to do a read * with an arbitrarily large buffer, and return as much as we can. We can * return an error code if the buffer is too small (EFBIG), then the program * can try a bigger buffer. */ int procfs_doprocmap(PFS_FILL_ARGS) { struct vmspace *vm; vm_map_t map; vm_map_entry_t entry, tmp_entry; struct vnode *vp; char *fullpath, *freepath; struct ucred *cred; int error; unsigned int last_timestamp; #ifdef COMPAT_FREEBSD32 int wrap32 = 0; #endif PROC_LOCK(p); error = p_candebug(td, p); PROC_UNLOCK(p); if (error) return (error); if (uio->uio_rw != UIO_READ) return (EOPNOTSUPP); #ifdef COMPAT_FREEBSD32 if (SV_CURPROC_FLAG(SV_ILP32)) { if (!(SV_PROC_FLAG(p, SV_ILP32))) return (EOPNOTSUPP); wrap32 = 1; } #endif vm = vmspace_acquire_ref(p); if (vm == NULL) return (ESRCH); map = &vm->vm_map; vm_map_lock_read(map); for (entry = map->header.next; entry != &map->header; entry = entry->next) { vm_object_t obj, tobj, lobj; int ref_count, shadow_count, flags; vm_offset_t e_start, e_end, addr; int resident, privateresident; char *type; vm_eflags_t e_eflags; vm_prot_t e_prot; if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) continue; e_eflags = entry->eflags; e_prot = entry->protection; e_start = entry->start; e_end = entry->end; privateresident = 0; obj = entry->object.vm_object; if (obj != NULL) { VM_OBJECT_RLOCK(obj); if (obj->shadow_count == 1) privateresident = obj->resident_page_count; } cred = (entry->cred) ? entry->cred : (obj ? obj->cred : NULL); resident = 0; addr = entry->start; while (addr < entry->end) { if (pmap_extract(map->pmap, addr)) resident++; addr += PAGE_SIZE; } for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { if (tobj != obj) VM_OBJECT_RLOCK(tobj); if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); lobj = tobj; } last_timestamp = map->timestamp; vm_map_unlock_read(map); freepath = NULL; fullpath = "-"; if (lobj) { + vp = NULL; switch (lobj->type) { default: case OBJT_DEFAULT: type = "default"; - vp = NULL; break; case OBJT_VNODE: type = "vnode"; vp = lobj->handle; vref(vp); break; case OBJT_SWAP: - type = "swap"; - vp = NULL; + if ((lobj->flags & OBJ_TMPFS_NODE) != 0) { + type = "vnode"; + if ((lobj->flags & OBJ_TMPFS) != 0) { + vp = lobj->un_pager.swp.swp_tmpfs; + vref(vp); + } + } else { + type = "swap"; + } break; case OBJT_SG: case OBJT_DEVICE: type = "device"; - vp = NULL; break; } if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); flags = obj->flags; ref_count = obj->ref_count; shadow_count = obj->shadow_count; VM_OBJECT_RUNLOCK(obj); if (vp != NULL) { vn_fullpath(td, vp, &fullpath, &freepath); vrele(vp); } } else { type = "none"; flags = 0; ref_count = 0; shadow_count = 0; } /* * format: * start, end, resident, private resident, cow, access, type, * charged, charged uid. */ error = sbuf_printf(sb, "0x%lx 0x%lx %d %d %p %s%s%s %d %d 0x%x %s %s %s %s %s %d\n", (u_long)e_start, (u_long)e_end, resident, privateresident, #ifdef COMPAT_FREEBSD32 wrap32 ? NULL : obj, /* Hide 64 bit value */ #else obj, #endif (e_prot & VM_PROT_READ)?"r":"-", (e_prot & VM_PROT_WRITE)?"w":"-", (e_prot & VM_PROT_EXECUTE)?"x":"-", ref_count, shadow_count, flags, (e_eflags & MAP_ENTRY_COW)?"COW":"NCOW", (e_eflags & MAP_ENTRY_NEEDS_COPY)?"NC":"NNC", type, fullpath, cred ? "CH":"NCH", cred ? cred->cr_ruid : -1); if (freepath != NULL) free(freepath, M_TEMP); vm_map_lock_read(map); if (error == -1) { error = 0; break; } if (last_timestamp != map->timestamp) { /* * Look again for the entry because the map was * modified while it was unlocked. Specifically, * the entry may have been clipped, merged, or deleted. */ vm_map_lookup_entry(map, e_end - 1, &tmp_entry); entry = tmp_entry; } } vm_map_unlock_read(map); vmspace_free(vm); return (error); } Index: projects/em_mq/sys/kern/kern_proc.c =================================================================== --- projects/em_mq/sys/kern/kern_proc.c (revision 283943) +++ projects/em_mq/sys/kern/kern_proc.c (revision 283944) @@ -1,3041 +1,3057 @@ /*- * Copyright (c) 1982, 1986, 1989, 1991, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include "opt_ddb.h" #include "opt_ktrace.h" #include "opt_kstack_pages.h" #include "opt_stack.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #endif #include #include #include #include #include #include #include #include #ifdef COMPAT_FREEBSD32 #include #include #endif SDT_PROVIDER_DEFINE(proc); SDT_PROBE_DEFINE4(proc, kernel, ctor, entry, "struct proc *", "int", "void *", "int"); SDT_PROBE_DEFINE4(proc, kernel, ctor, return, "struct proc *", "int", "void *", "int"); SDT_PROBE_DEFINE4(proc, kernel, dtor, entry, "struct proc *", "int", "void *", "struct thread *"); SDT_PROBE_DEFINE3(proc, kernel, dtor, return, "struct proc *", "int", "void *"); SDT_PROBE_DEFINE3(proc, kernel, init, entry, "struct proc *", "int", "int"); SDT_PROBE_DEFINE3(proc, kernel, init, return, "struct proc *", "int", "int"); MALLOC_DEFINE(M_PGRP, "pgrp", "process group header"); MALLOC_DEFINE(M_SESSION, "session", "session header"); static MALLOC_DEFINE(M_PROC, "proc", "Proc structures"); MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); static void doenterpgrp(struct proc *, struct pgrp *); static void orphanpg(struct pgrp *pg); static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp); static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp); static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread); static void pgadjustjobc(struct pgrp *pgrp, int entering); static void pgdelete(struct pgrp *); static int proc_ctor(void *mem, int size, void *arg, int flags); static void proc_dtor(void *mem, int size, void *arg); static int proc_init(void *mem, int size, int flags); static void proc_fini(void *mem, int size); static void pargs_free(struct pargs *pa); static struct proc *zpfind_locked(pid_t pid); /* * Other process lists */ struct pidhashhead *pidhashtbl; u_long pidhash; struct pgrphashhead *pgrphashtbl; u_long pgrphash; struct proclist allproc; struct proclist zombproc; struct sx allproc_lock; struct sx proctree_lock; struct mtx ppeers_lock; uma_zone_t proc_zone; int kstack_pages = KSTACK_PAGES; SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, "Kernel stack size in pages"); static int vmmap_skip_res_cnt = 0; SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW, &vmmap_skip_res_cnt, 0, "Skip calculation of the pages resident count in kern.proc.vmmap"); CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE); #ifdef COMPAT_FREEBSD32 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE); #endif /* * Initialize global process hashing structures. */ void procinit() { sx_init(&allproc_lock, "allproc"); sx_init(&proctree_lock, "proctree"); mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF); LIST_INIT(&allproc); LIST_INIT(&zombproc); pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash); pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); proc_zone = uma_zcreate("PROC", sched_sizeof_proc(), proc_ctor, proc_dtor, proc_init, proc_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uihashinit(); } /* * Prepare a proc for use. */ static int proc_ctor(void *mem, int size, void *arg, int flags) { struct proc *p; p = (struct proc *)mem; SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0); EVENTHANDLER_INVOKE(process_ctor, p); SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0); return (0); } /* * Reclaim a proc after use. */ static void proc_dtor(void *mem, int size, void *arg) { struct proc *p; struct thread *td; /* INVARIANTS checks go here */ p = (struct proc *)mem; td = FIRST_THREAD_IN_PROC(p); SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0); if (td != NULL) { #ifdef INVARIANTS KASSERT((p->p_numthreads == 1), ("bad number of threads in exiting process")); KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr")); #endif /* Free all OSD associated to this thread. */ osd_thread_exit(td); } EVENTHANDLER_INVOKE(process_dtor, p); if (p->p_ksi != NULL) KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue")); SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0); } /* * Initialize type-stable parts of a proc (when newly created). */ static int proc_init(void *mem, int size, int flags) { struct proc *p; p = (struct proc *)mem; SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0); p->p_sched = (struct p_sched *)&p[1]; mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW); mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW); mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW); mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW); mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW); cv_init(&p->p_pwait, "ppwait"); cv_init(&p->p_dbgwait, "dbgwait"); TAILQ_INIT(&p->p_threads); /* all threads in proc */ EVENTHANDLER_INVOKE(process_init, p); p->p_stats = pstats_alloc(); SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0); return (0); } /* * UMA should ensure that this function is never called. * Freeing a proc structure would violate type stability. */ static void proc_fini(void *mem, int size) { #ifdef notnow struct proc *p; p = (struct proc *)mem; EVENTHANDLER_INVOKE(process_fini, p); pstats_free(p->p_stats); thread_free(FIRST_THREAD_IN_PROC(p)); mtx_destroy(&p->p_mtx); if (p->p_ksi != NULL) ksiginfo_free(p->p_ksi); #else panic("proc reclaimed"); #endif } /* * Is p an inferior of the current process? */ int inferior(struct proc *p) { sx_assert(&proctree_lock, SX_LOCKED); PROC_LOCK_ASSERT(p, MA_OWNED); for (; p != curproc; p = proc_realparent(p)) { if (p->p_pid == 0) return (0); } return (1); } struct proc * pfind_locked(pid_t pid) { struct proc *p; sx_assert(&allproc_lock, SX_LOCKED); LIST_FOREACH(p, PIDHASH(pid), p_hash) { if (p->p_pid == pid) { PROC_LOCK(p); if (p->p_state == PRS_NEW) { PROC_UNLOCK(p); p = NULL; } break; } } return (p); } /* * Locate a process by number; return only "live" processes -- i.e., neither * zombies nor newly born but incompletely initialized processes. By not * returning processes in the PRS_NEW state, we allow callers to avoid * testing for that condition to avoid dereferencing p_ucred, et al. */ struct proc * pfind(pid_t pid) { struct proc *p; sx_slock(&allproc_lock); p = pfind_locked(pid); sx_sunlock(&allproc_lock); return (p); } static struct proc * pfind_tid_locked(pid_t tid) { struct proc *p; struct thread *td; sx_assert(&allproc_lock, SX_LOCKED); FOREACH_PROC_IN_SYSTEM(p) { PROC_LOCK(p); if (p->p_state == PRS_NEW) { PROC_UNLOCK(p); continue; } FOREACH_THREAD_IN_PROC(p, td) { if (td->td_tid == tid) goto found; } PROC_UNLOCK(p); } found: return (p); } /* * Locate a process group by number. * The caller must hold proctree_lock. */ struct pgrp * pgfind(pgid) register pid_t pgid; { register struct pgrp *pgrp; sx_assert(&proctree_lock, SX_LOCKED); LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) { if (pgrp->pg_id == pgid) { PGRP_LOCK(pgrp); return (pgrp); } } return (NULL); } /* * Locate process and do additional manipulations, depending on flags. */ int pget(pid_t pid, int flags, struct proc **pp) { struct proc *p; int error; sx_slock(&allproc_lock); if (pid <= PID_MAX) { p = pfind_locked(pid); if (p == NULL && (flags & PGET_NOTWEXIT) == 0) p = zpfind_locked(pid); } else if ((flags & PGET_NOTID) == 0) { p = pfind_tid_locked(pid); } else { p = NULL; } sx_sunlock(&allproc_lock); if (p == NULL) return (ESRCH); if ((flags & PGET_CANSEE) != 0) { error = p_cansee(curthread, p); if (error != 0) goto errout; } if ((flags & PGET_CANDEBUG) != 0) { error = p_candebug(curthread, p); if (error != 0) goto errout; } if ((flags & PGET_ISCURRENT) != 0 && curproc != p) { error = EPERM; goto errout; } if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) { error = ESRCH; goto errout; } if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) { /* * XXXRW: Not clear ESRCH is the right error during proc * execve(). */ error = ESRCH; goto errout; } if ((flags & PGET_HOLD) != 0) { _PHOLD(p); PROC_UNLOCK(p); } *pp = p; return (0); errout: PROC_UNLOCK(p); return (error); } /* * Create a new process group. * pgid must be equal to the pid of p. * Begin a new session if required. */ int enterpgrp(p, pgid, pgrp, sess) register struct proc *p; pid_t pgid; struct pgrp *pgrp; struct session *sess; { sx_assert(&proctree_lock, SX_XLOCKED); KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL")); KASSERT(p->p_pid == pgid, ("enterpgrp: new pgrp and pid != pgid")); KASSERT(pgfind(pgid) == NULL, ("enterpgrp: pgrp with pgid exists")); KASSERT(!SESS_LEADER(p), ("enterpgrp: session leader attempted setpgrp")); mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK); if (sess != NULL) { /* * new session */ mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF); PROC_LOCK(p); p->p_flag &= ~P_CONTROLT; PROC_UNLOCK(p); PGRP_LOCK(pgrp); sess->s_leader = p; sess->s_sid = p->p_pid; refcount_init(&sess->s_count, 1); sess->s_ttyvp = NULL; sess->s_ttydp = NULL; sess->s_ttyp = NULL; bcopy(p->p_session->s_login, sess->s_login, sizeof(sess->s_login)); pgrp->pg_session = sess; KASSERT(p == curproc, ("enterpgrp: mksession and p != curproc")); } else { pgrp->pg_session = p->p_session; sess_hold(pgrp->pg_session); PGRP_LOCK(pgrp); } pgrp->pg_id = pgid; LIST_INIT(&pgrp->pg_members); /* * As we have an exclusive lock of proctree_lock, * this should not deadlock. */ LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash); pgrp->pg_jobc = 0; SLIST_INIT(&pgrp->pg_sigiolst); PGRP_UNLOCK(pgrp); doenterpgrp(p, pgrp); return (0); } /* * Move p to an existing process group */ int enterthispgrp(p, pgrp) register struct proc *p; struct pgrp *pgrp; { sx_assert(&proctree_lock, SX_XLOCKED); PROC_LOCK_ASSERT(p, MA_NOTOWNED); PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); KASSERT(pgrp->pg_session == p->p_session, ("%s: pgrp's session %p, p->p_session %p.\n", __func__, pgrp->pg_session, p->p_session)); KASSERT(pgrp != p->p_pgrp, ("%s: p belongs to pgrp.", __func__)); doenterpgrp(p, pgrp); return (0); } /* * Move p to a process group */ static void doenterpgrp(p, pgrp) struct proc *p; struct pgrp *pgrp; { struct pgrp *savepgrp; sx_assert(&proctree_lock, SX_XLOCKED); PROC_LOCK_ASSERT(p, MA_NOTOWNED); PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED); SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED); savepgrp = p->p_pgrp; /* * Adjust eligibility of affected pgrps to participate in job control. * Increment eligibility counts before decrementing, otherwise we * could reach 0 spuriously during the first call. */ fixjobc(p, pgrp, 1); fixjobc(p, p->p_pgrp, 0); PGRP_LOCK(pgrp); PGRP_LOCK(savepgrp); PROC_LOCK(p); LIST_REMOVE(p, p_pglist); p->p_pgrp = pgrp; PROC_UNLOCK(p); LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); PGRP_UNLOCK(savepgrp); PGRP_UNLOCK(pgrp); if (LIST_EMPTY(&savepgrp->pg_members)) pgdelete(savepgrp); } /* * remove process from process group */ int leavepgrp(p) register struct proc *p; { struct pgrp *savepgrp; sx_assert(&proctree_lock, SX_XLOCKED); savepgrp = p->p_pgrp; PGRP_LOCK(savepgrp); PROC_LOCK(p); LIST_REMOVE(p, p_pglist); p->p_pgrp = NULL; PROC_UNLOCK(p); PGRP_UNLOCK(savepgrp); if (LIST_EMPTY(&savepgrp->pg_members)) pgdelete(savepgrp); return (0); } /* * delete a process group */ static void pgdelete(pgrp) register struct pgrp *pgrp; { struct session *savesess; struct tty *tp; sx_assert(&proctree_lock, SX_XLOCKED); PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); /* * Reset any sigio structures pointing to us as a result of * F_SETOWN with our pgid. */ funsetownlst(&pgrp->pg_sigiolst); PGRP_LOCK(pgrp); tp = pgrp->pg_session->s_ttyp; LIST_REMOVE(pgrp, pg_hash); savesess = pgrp->pg_session; PGRP_UNLOCK(pgrp); /* Remove the reference to the pgrp before deallocating it. */ if (tp != NULL) { tty_lock(tp); tty_rel_pgrp(tp, pgrp); } mtx_destroy(&pgrp->pg_mtx); free(pgrp, M_PGRP); sess_release(savesess); } static void pgadjustjobc(pgrp, entering) struct pgrp *pgrp; int entering; { PGRP_LOCK(pgrp); if (entering) pgrp->pg_jobc++; else { --pgrp->pg_jobc; if (pgrp->pg_jobc == 0) orphanpg(pgrp); } PGRP_UNLOCK(pgrp); } /* * Adjust pgrp jobc counters when specified process changes process group. * We count the number of processes in each process group that "qualify" * the group for terminal job control (those with a parent in a different * process group of the same session). If that count reaches zero, the * process group becomes orphaned. Check both the specified process' * process group and that of its children. * entering == 0 => p is leaving specified group. * entering == 1 => p is entering specified group. */ void fixjobc(p, pgrp, entering) register struct proc *p; register struct pgrp *pgrp; int entering; { register struct pgrp *hispgrp; register struct session *mysession; sx_assert(&proctree_lock, SX_LOCKED); PROC_LOCK_ASSERT(p, MA_NOTOWNED); PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED); SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED); /* * Check p's parent to see whether p qualifies its own process * group; if so, adjust count for p's process group. */ mysession = pgrp->pg_session; if ((hispgrp = p->p_pptr->p_pgrp) != pgrp && hispgrp->pg_session == mysession) pgadjustjobc(pgrp, entering); /* * Check this process' children to see whether they qualify * their process groups; if so, adjust counts for children's * process groups. */ LIST_FOREACH(p, &p->p_children, p_sibling) { hispgrp = p->p_pgrp; if (hispgrp == pgrp || hispgrp->pg_session != mysession) continue; PROC_LOCK(p); if (p->p_state == PRS_ZOMBIE) { PROC_UNLOCK(p); continue; } PROC_UNLOCK(p); pgadjustjobc(hispgrp, entering); } } /* * A process group has become orphaned; * if there are any stopped processes in the group, * hang-up all process in that group. */ static void orphanpg(pg) struct pgrp *pg; { register struct proc *p; PGRP_LOCK_ASSERT(pg, MA_OWNED); LIST_FOREACH(p, &pg->pg_members, p_pglist) { PROC_LOCK(p); if (P_SHOULDSTOP(p)) { PROC_UNLOCK(p); LIST_FOREACH(p, &pg->pg_members, p_pglist) { PROC_LOCK(p); kern_psignal(p, SIGHUP); kern_psignal(p, SIGCONT); PROC_UNLOCK(p); } return; } PROC_UNLOCK(p); } } void sess_hold(struct session *s) { refcount_acquire(&s->s_count); } void sess_release(struct session *s) { if (refcount_release(&s->s_count)) { if (s->s_ttyp != NULL) { tty_lock(s->s_ttyp); tty_rel_sess(s->s_ttyp, s); } mtx_destroy(&s->s_mtx); free(s, M_SESSION); } } #ifdef DDB DB_SHOW_COMMAND(pgrpdump, pgrpdump) { register struct pgrp *pgrp; register struct proc *p; register int i; for (i = 0; i <= pgrphash; i++) { if (!LIST_EMPTY(&pgrphashtbl[i])) { printf("\tindx %d\n", i); LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) { printf( "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n", (void *)pgrp, (long)pgrp->pg_id, (void *)pgrp->pg_session, pgrp->pg_session->s_count, (void *)LIST_FIRST(&pgrp->pg_members)); LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { printf("\t\tpid %ld addr %p pgrp %p\n", (long)p->p_pid, (void *)p, (void *)p->p_pgrp); } } } } } #endif /* DDB */ /* * Calculate the kinfo_proc members which contain process-wide * informations. * Must be called with the target process locked. */ static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp) { struct thread *td; PROC_LOCK_ASSERT(p, MA_OWNED); kp->ki_estcpu = 0; kp->ki_pctcpu = 0; FOREACH_THREAD_IN_PROC(p, td) { thread_lock(td); kp->ki_pctcpu += sched_pctcpu(td); kp->ki_estcpu += td->td_estcpu; thread_unlock(td); } } /* * Clear kinfo_proc and fill in any information that is common * to all threads in the process. * Must be called with the target process locked. */ static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp) { struct thread *td0; struct tty *tp; struct session *sp; struct ucred *cred; struct sigacts *ps; /* For proc_realparent. */ sx_assert(&proctree_lock, SX_LOCKED); PROC_LOCK_ASSERT(p, MA_OWNED); bzero(kp, sizeof(*kp)); kp->ki_structsize = sizeof(*kp); kp->ki_paddr = p; kp->ki_addr =/* p->p_addr; */0; /* XXX */ kp->ki_args = p->p_args; kp->ki_textvp = p->p_textvp; #ifdef KTRACE kp->ki_tracep = p->p_tracevp; kp->ki_traceflag = p->p_traceflag; #endif kp->ki_fd = p->p_fd; kp->ki_vmspace = p->p_vmspace; kp->ki_flag = p->p_flag; kp->ki_flag2 = p->p_flag2; cred = p->p_ucred; if (cred) { kp->ki_uid = cred->cr_uid; kp->ki_ruid = cred->cr_ruid; kp->ki_svuid = cred->cr_svuid; kp->ki_cr_flags = 0; if (cred->cr_flags & CRED_FLAG_CAPMODE) kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE; /* XXX bde doesn't like KI_NGROUPS */ if (cred->cr_ngroups > KI_NGROUPS) { kp->ki_ngroups = KI_NGROUPS; kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW; } else kp->ki_ngroups = cred->cr_ngroups; bcopy(cred->cr_groups, kp->ki_groups, kp->ki_ngroups * sizeof(gid_t)); kp->ki_rgid = cred->cr_rgid; kp->ki_svgid = cred->cr_svgid; /* If jailed(cred), emulate the old P_JAILED flag. */ if (jailed(cred)) { kp->ki_flag |= P_JAILED; /* If inside the jail, use 0 as a jail ID. */ if (cred->cr_prison != curthread->td_ucred->cr_prison) kp->ki_jid = cred->cr_prison->pr_id; } strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name, sizeof(kp->ki_loginclass)); } ps = p->p_sigacts; if (ps) { mtx_lock(&ps->ps_mtx); kp->ki_sigignore = ps->ps_sigignore; kp->ki_sigcatch = ps->ps_sigcatch; mtx_unlock(&ps->ps_mtx); } if (p->p_state != PRS_NEW && p->p_state != PRS_ZOMBIE && p->p_vmspace != NULL) { struct vmspace *vm = p->p_vmspace; kp->ki_size = vm->vm_map.size; kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/ FOREACH_THREAD_IN_PROC(p, td0) { if (!TD_IS_SWAPPED(td0)) kp->ki_rssize += td0->td_kstack_pages; } kp->ki_swrss = vm->vm_swrss; kp->ki_tsize = vm->vm_tsize; kp->ki_dsize = vm->vm_dsize; kp->ki_ssize = vm->vm_ssize; } else if (p->p_state == PRS_ZOMBIE) kp->ki_stat = SZOMB; if (kp->ki_flag & P_INMEM) kp->ki_sflag = PS_INMEM; else kp->ki_sflag = 0; /* Calculate legacy swtime as seconds since 'swtick'. */ kp->ki_swtime = (ticks - p->p_swtick) / hz; kp->ki_pid = p->p_pid; kp->ki_nice = p->p_nice; kp->ki_fibnum = p->p_fibnum; kp->ki_start = p->p_stats->p_start; timevaladd(&kp->ki_start, &boottime); PROC_STATLOCK(p); rufetch(p, &kp->ki_rusage); kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime); calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime); PROC_STATUNLOCK(p); calccru(p, &kp->ki_childutime, &kp->ki_childstime); /* Some callers want child times in a single value. */ kp->ki_childtime = kp->ki_childstime; timevaladd(&kp->ki_childtime, &kp->ki_childutime); FOREACH_THREAD_IN_PROC(p, td0) kp->ki_cow += td0->td_cow; tp = NULL; if (p->p_pgrp) { kp->ki_pgid = p->p_pgrp->pg_id; kp->ki_jobc = p->p_pgrp->pg_jobc; sp = p->p_pgrp->pg_session; if (sp != NULL) { kp->ki_sid = sp->s_sid; SESS_LOCK(sp); strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login)); if (sp->s_ttyvp) kp->ki_kiflag |= KI_CTTY; if (SESS_LEADER(p)) kp->ki_kiflag |= KI_SLEADER; /* XXX proctree_lock */ tp = sp->s_ttyp; SESS_UNLOCK(sp); } } if ((p->p_flag & P_CONTROLT) && tp != NULL) { kp->ki_tdev = tty_udev(tp); kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID; if (tp->t_session) kp->ki_tsid = tp->t_session->s_sid; } else kp->ki_tdev = NODEV; if (p->p_comm[0] != '\0') strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm)); if (p->p_sysent && p->p_sysent->sv_name != NULL && p->p_sysent->sv_name[0] != '\0') strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul)); kp->ki_siglist = p->p_siglist; kp->ki_xstat = p->p_xstat; kp->ki_acflag = p->p_acflag; kp->ki_lock = p->p_lock; if (p->p_pptr) { kp->ki_ppid = proc_realparent(p)->p_pid; if (p->p_flag & P_TRACED) kp->ki_tracer = p->p_pptr->p_pid; } } /* * Fill in information that is thread specific. Must be called with * target process locked. If 'preferthread' is set, overwrite certain * process-related fields that are maintained for both threads and * processes. */ static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread) { struct proc *p; p = td->td_proc; kp->ki_tdaddr = td; PROC_LOCK_ASSERT(p, MA_OWNED); if (preferthread) PROC_STATLOCK(p); thread_lock(td); if (td->td_wmesg != NULL) strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg)); else bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg)); strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)); if (TD_ON_LOCK(td)) { kp->ki_kiflag |= KI_LOCKBLOCK; strlcpy(kp->ki_lockname, td->td_lockname, sizeof(kp->ki_lockname)); } else { kp->ki_kiflag &= ~KI_LOCKBLOCK; bzero(kp->ki_lockname, sizeof(kp->ki_lockname)); } if (p->p_state == PRS_NORMAL) { /* approximate. */ if (TD_ON_RUNQ(td) || TD_CAN_RUN(td) || TD_IS_RUNNING(td)) { kp->ki_stat = SRUN; } else if (P_SHOULDSTOP(p)) { kp->ki_stat = SSTOP; } else if (TD_IS_SLEEPING(td)) { kp->ki_stat = SSLEEP; } else if (TD_ON_LOCK(td)) { kp->ki_stat = SLOCK; } else { kp->ki_stat = SWAIT; } } else if (p->p_state == PRS_ZOMBIE) { kp->ki_stat = SZOMB; } else { kp->ki_stat = SIDL; } /* Things in the thread */ kp->ki_wchan = td->td_wchan; kp->ki_pri.pri_level = td->td_priority; kp->ki_pri.pri_native = td->td_base_pri; /* * Note: legacy fields; clamp at the old NOCPU value and/or * the maximum u_char CPU value. */ if (td->td_lastcpu == NOCPU) kp->ki_lastcpu_old = NOCPU_OLD; else if (td->td_lastcpu > MAXCPU_OLD) kp->ki_lastcpu_old = MAXCPU_OLD; else kp->ki_lastcpu_old = td->td_lastcpu; if (td->td_oncpu == NOCPU) kp->ki_oncpu_old = NOCPU_OLD; else if (td->td_oncpu > MAXCPU_OLD) kp->ki_oncpu_old = MAXCPU_OLD; else kp->ki_oncpu_old = td->td_oncpu; kp->ki_lastcpu = td->td_lastcpu; kp->ki_oncpu = td->td_oncpu; kp->ki_tdflags = td->td_flags; kp->ki_tid = td->td_tid; kp->ki_numthreads = p->p_numthreads; kp->ki_pcb = td->td_pcb; kp->ki_kstack = (void *)td->td_kstack; kp->ki_slptime = (ticks - td->td_slptick) / hz; kp->ki_pri.pri_class = td->td_pri_class; kp->ki_pri.pri_user = td->td_user_pri; if (preferthread) { rufetchtd(td, &kp->ki_rusage); kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime); kp->ki_pctcpu = sched_pctcpu(td); kp->ki_estcpu = td->td_estcpu; kp->ki_cow = td->td_cow; } /* We can't get this anymore but ps etc never used it anyway. */ kp->ki_rqindex = 0; if (preferthread) kp->ki_siglist = td->td_siglist; kp->ki_sigmask = td->td_sigmask; thread_unlock(td); if (preferthread) PROC_STATUNLOCK(p); } /* * Fill in a kinfo_proc structure for the specified process. * Must be called with the target process locked. */ void fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp) { MPASS(FIRST_THREAD_IN_PROC(p) != NULL); fill_kinfo_proc_only(p, kp); fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0); fill_kinfo_aggregate(p, kp); } struct pstats * pstats_alloc(void) { return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK)); } /* * Copy parts of p_stats; zero the rest of p_stats (statistics). */ void pstats_fork(struct pstats *src, struct pstats *dst) { bzero(&dst->pstat_startzero, __rangeof(struct pstats, pstat_startzero, pstat_endzero)); bcopy(&src->pstat_startcopy, &dst->pstat_startcopy, __rangeof(struct pstats, pstat_startcopy, pstat_endcopy)); } void pstats_free(struct pstats *ps) { free(ps, M_SUBPROC); } static struct proc * zpfind_locked(pid_t pid) { struct proc *p; sx_assert(&allproc_lock, SX_LOCKED); LIST_FOREACH(p, &zombproc, p_list) { if (p->p_pid == pid) { PROC_LOCK(p); break; } } return (p); } /* * Locate a zombie process by number */ struct proc * zpfind(pid_t pid) { struct proc *p; sx_slock(&allproc_lock); p = zpfind_locked(pid); sx_sunlock(&allproc_lock); return (p); } #ifdef COMPAT_FREEBSD32 /* * This function is typically used to copy out the kernel address, so * it can be replaced by assignment of zero. */ static inline uint32_t ptr32_trim(void *ptr) { uintptr_t uptr; uptr = (uintptr_t)ptr; return ((uptr > UINT_MAX) ? 0 : uptr); } #define PTRTRIM_CP(src,dst,fld) \ do { (dst).fld = ptr32_trim((src).fld); } while (0) static void freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32) { int i; bzero(ki32, sizeof(struct kinfo_proc32)); ki32->ki_structsize = sizeof(struct kinfo_proc32); CP(*ki, *ki32, ki_layout); PTRTRIM_CP(*ki, *ki32, ki_args); PTRTRIM_CP(*ki, *ki32, ki_paddr); PTRTRIM_CP(*ki, *ki32, ki_addr); PTRTRIM_CP(*ki, *ki32, ki_tracep); PTRTRIM_CP(*ki, *ki32, ki_textvp); PTRTRIM_CP(*ki, *ki32, ki_fd); PTRTRIM_CP(*ki, *ki32, ki_vmspace); PTRTRIM_CP(*ki, *ki32, ki_wchan); CP(*ki, *ki32, ki_pid); CP(*ki, *ki32, ki_ppid); CP(*ki, *ki32, ki_pgid); CP(*ki, *ki32, ki_tpgid); CP(*ki, *ki32, ki_sid); CP(*ki, *ki32, ki_tsid); CP(*ki, *ki32, ki_jobc); CP(*ki, *ki32, ki_tdev); CP(*ki, *ki32, ki_siglist); CP(*ki, *ki32, ki_sigmask); CP(*ki, *ki32, ki_sigignore); CP(*ki, *ki32, ki_sigcatch); CP(*ki, *ki32, ki_uid); CP(*ki, *ki32, ki_ruid); CP(*ki, *ki32, ki_svuid); CP(*ki, *ki32, ki_rgid); CP(*ki, *ki32, ki_svgid); CP(*ki, *ki32, ki_ngroups); for (i = 0; i < KI_NGROUPS; i++) CP(*ki, *ki32, ki_groups[i]); CP(*ki, *ki32, ki_size); CP(*ki, *ki32, ki_rssize); CP(*ki, *ki32, ki_swrss); CP(*ki, *ki32, ki_tsize); CP(*ki, *ki32, ki_dsize); CP(*ki, *ki32, ki_ssize); CP(*ki, *ki32, ki_xstat); CP(*ki, *ki32, ki_acflag); CP(*ki, *ki32, ki_pctcpu); CP(*ki, *ki32, ki_estcpu); CP(*ki, *ki32, ki_slptime); CP(*ki, *ki32, ki_swtime); CP(*ki, *ki32, ki_cow); CP(*ki, *ki32, ki_runtime); TV_CP(*ki, *ki32, ki_start); TV_CP(*ki, *ki32, ki_childtime); CP(*ki, *ki32, ki_flag); CP(*ki, *ki32, ki_kiflag); CP(*ki, *ki32, ki_traceflag); CP(*ki, *ki32, ki_stat); CP(*ki, *ki32, ki_nice); CP(*ki, *ki32, ki_lock); CP(*ki, *ki32, ki_rqindex); CP(*ki, *ki32, ki_oncpu); CP(*ki, *ki32, ki_lastcpu); /* XXX TODO: wrap cpu value as appropriate */ CP(*ki, *ki32, ki_oncpu_old); CP(*ki, *ki32, ki_lastcpu_old); bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1); bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1); bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1); bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1); bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1); bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1); bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1); CP(*ki, *ki32, ki_tracer); CP(*ki, *ki32, ki_flag2); CP(*ki, *ki32, ki_fibnum); CP(*ki, *ki32, ki_cr_flags); CP(*ki, *ki32, ki_jid); CP(*ki, *ki32, ki_numthreads); CP(*ki, *ki32, ki_tid); CP(*ki, *ki32, ki_pri); freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage); freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch); PTRTRIM_CP(*ki, *ki32, ki_pcb); PTRTRIM_CP(*ki, *ki32, ki_kstack); PTRTRIM_CP(*ki, *ki32, ki_udata); CP(*ki, *ki32, ki_sflag); CP(*ki, *ki32, ki_tdflags); } #endif int kern_proc_out(struct proc *p, struct sbuf *sb, int flags) { struct thread *td; struct kinfo_proc ki; #ifdef COMPAT_FREEBSD32 struct kinfo_proc32 ki32; #endif int error; PROC_LOCK_ASSERT(p, MA_OWNED); MPASS(FIRST_THREAD_IN_PROC(p) != NULL); error = 0; fill_kinfo_proc(p, &ki); if ((flags & KERN_PROC_NOTHREADS) != 0) { #ifdef COMPAT_FREEBSD32 if ((flags & KERN_PROC_MASK32) != 0) { freebsd32_kinfo_proc_out(&ki, &ki32); if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) error = ENOMEM; } else #endif if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) error = ENOMEM; } else { FOREACH_THREAD_IN_PROC(p, td) { fill_kinfo_thread(td, &ki, 1); #ifdef COMPAT_FREEBSD32 if ((flags & KERN_PROC_MASK32) != 0) { freebsd32_kinfo_proc_out(&ki, &ki32); if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0) error = ENOMEM; } else #endif if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0) error = ENOMEM; if (error != 0) break; } } PROC_UNLOCK(p); return (error); } static int sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags, int doingzomb) { struct sbuf sb; struct kinfo_proc ki; struct proc *np; int error, error2; pid_t pid; pid = p->p_pid; sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req); sbuf_clear_flags(&sb, SBUF_INCLUDENUL); error = kern_proc_out(p, &sb, flags); error2 = sbuf_finish(&sb); sbuf_delete(&sb); if (error != 0) return (error); else if (error2 != 0) return (error2); if (doingzomb) np = zpfind(pid); else { if (pid == 0) return (0); np = pfind(pid); } if (np == NULL) return (ESRCH); if (np != p) { PROC_UNLOCK(np); return (ESRCH); } PROC_UNLOCK(np); return (0); } static int sysctl_kern_proc(SYSCTL_HANDLER_ARGS) { int *name = (int *)arg1; u_int namelen = arg2; struct proc *p; int flags, doingzomb, oid_number; int error = 0; oid_number = oidp->oid_number; if (oid_number != KERN_PROC_ALL && (oid_number & KERN_PROC_INC_THREAD) == 0) flags = KERN_PROC_NOTHREADS; else { flags = 0; oid_number &= ~KERN_PROC_INC_THREAD; } #ifdef COMPAT_FREEBSD32 if (req->flags & SCTL_MASK32) flags |= KERN_PROC_MASK32; #endif if (oid_number == KERN_PROC_PID) { if (namelen != 1) return (EINVAL); error = sysctl_wire_old_buffer(req, 0); if (error) return (error); sx_slock(&proctree_lock); error = pget((pid_t)name[0], PGET_CANSEE, &p); if (error == 0) error = sysctl_out_proc(p, req, flags, 0); sx_sunlock(&proctree_lock); return (error); } switch (oid_number) { case KERN_PROC_ALL: if (namelen != 0) return (EINVAL); break; case KERN_PROC_PROC: if (namelen != 0 && namelen != 1) return (EINVAL); break; default: if (namelen != 1) return (EINVAL); break; } if (!req->oldptr) { /* overestimate by 5 procs */ error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5); if (error) return (error); } error = sysctl_wire_old_buffer(req, 0); if (error != 0) return (error); sx_slock(&proctree_lock); sx_slock(&allproc_lock); for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) { if (!doingzomb) p = LIST_FIRST(&allproc); else p = LIST_FIRST(&zombproc); for (; p != 0; p = LIST_NEXT(p, p_list)) { /* * Skip embryonic processes. */ PROC_LOCK(p); if (p->p_state == PRS_NEW) { PROC_UNLOCK(p); continue; } KASSERT(p->p_ucred != NULL, ("process credential is NULL for non-NEW proc")); /* * Show a user only appropriate processes. */ if (p_cansee(curthread, p)) { PROC_UNLOCK(p); continue; } /* * TODO - make more efficient (see notes below). * do by session. */ switch (oid_number) { case KERN_PROC_GID: if (p->p_ucred->cr_gid != (gid_t)name[0]) { PROC_UNLOCK(p); continue; } break; case KERN_PROC_PGRP: /* could do this by traversing pgrp */ if (p->p_pgrp == NULL || p->p_pgrp->pg_id != (pid_t)name[0]) { PROC_UNLOCK(p); continue; } break; case KERN_PROC_RGID: if (p->p_ucred->cr_rgid != (gid_t)name[0]) { PROC_UNLOCK(p); continue; } break; case KERN_PROC_SESSION: if (p->p_session == NULL || p->p_session->s_sid != (pid_t)name[0]) { PROC_UNLOCK(p); continue; } break; case KERN_PROC_TTY: if ((p->p_flag & P_CONTROLT) == 0 || p->p_session == NULL) { PROC_UNLOCK(p); continue; } /* XXX proctree_lock */ SESS_LOCK(p->p_session); if (p->p_session->s_ttyp == NULL || tty_udev(p->p_session->s_ttyp) != (dev_t)name[0]) { SESS_UNLOCK(p->p_session); PROC_UNLOCK(p); continue; } SESS_UNLOCK(p->p_session); break; case KERN_PROC_UID: if (p->p_ucred->cr_uid != (uid_t)name[0]) { PROC_UNLOCK(p); continue; } break; case KERN_PROC_RUID: if (p->p_ucred->cr_ruid != (uid_t)name[0]) { PROC_UNLOCK(p); continue; } break; case KERN_PROC_PROC: break; default: break; } error = sysctl_out_proc(p, req, flags, doingzomb); if (error) { sx_sunlock(&allproc_lock); sx_sunlock(&proctree_lock); return (error); } } } sx_sunlock(&allproc_lock); sx_sunlock(&proctree_lock); return (0); } struct pargs * pargs_alloc(int len) { struct pargs *pa; pa = malloc(sizeof(struct pargs) + len, M_PARGS, M_WAITOK); refcount_init(&pa->ar_ref, 1); pa->ar_length = len; return (pa); } static void pargs_free(struct pargs *pa) { free(pa, M_PARGS); } void pargs_hold(struct pargs *pa) { if (pa == NULL) return; refcount_acquire(&pa->ar_ref); } void pargs_drop(struct pargs *pa) { if (pa == NULL) return; if (refcount_release(&pa->ar_ref)) pargs_free(pa); } static int proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf, size_t len) { struct iovec iov; struct uio uio; iov.iov_base = (caddr_t)buf; iov.iov_len = len; uio.uio_iov = &iov; uio.uio_iovcnt = 1; uio.uio_offset = offset; uio.uio_resid = (ssize_t)len; uio.uio_segflg = UIO_SYSSPACE; uio.uio_rw = UIO_READ; uio.uio_td = td; return (proc_rwmem(p, &uio)); } static int proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf, size_t len) { size_t i; int error; error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len); /* * Reading the chunk may validly return EFAULT if the string is shorter * than the chunk and is aligned at the end of the page, assuming the * next page is not mapped. So if EFAULT is returned do a fallback to * one byte read loop. */ if (error == EFAULT) { for (i = 0; i < len; i++, buf++, sptr++) { error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1); if (error != 0) return (error); if (*buf == '\0') break; } error = 0; } return (error); } #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */ enum proc_vector_type { PROC_ARG, PROC_ENV, PROC_AUX, }; #ifdef COMPAT_FREEBSD32 static int get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp, size_t *vsizep, enum proc_vector_type type) { struct freebsd32_ps_strings pss; Elf32_Auxinfo aux; vm_offset_t vptr, ptr; uint32_t *proc_vector32; char **proc_vector; size_t vsize, size; int i, error; error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings), &pss, sizeof(pss)); if (error != 0) return (error); switch (type) { case PROC_ARG: vptr = (vm_offset_t)PTRIN(pss.ps_argvstr); vsize = pss.ps_nargvstr; if (vsize > ARG_MAX) return (ENOEXEC); size = vsize * sizeof(int32_t); break; case PROC_ENV: vptr = (vm_offset_t)PTRIN(pss.ps_envstr); vsize = pss.ps_nenvstr; if (vsize > ARG_MAX) return (ENOEXEC); size = vsize * sizeof(int32_t); break; case PROC_AUX: vptr = (vm_offset_t)PTRIN(pss.ps_envstr) + (pss.ps_nenvstr + 1) * sizeof(int32_t); if (vptr % 4 != 0) return (ENOEXEC); for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { error = proc_read_mem(td, p, ptr, &aux, sizeof(aux)); if (error != 0) return (error); if (aux.a_type == AT_NULL) break; ptr += sizeof(aux); } if (aux.a_type != AT_NULL) return (ENOEXEC); vsize = i + 1; size = vsize * sizeof(aux); break; default: KASSERT(0, ("Wrong proc vector type: %d", type)); return (EINVAL); } proc_vector32 = malloc(size, M_TEMP, M_WAITOK); error = proc_read_mem(td, p, vptr, proc_vector32, size); if (error != 0) goto done; if (type == PROC_AUX) { *proc_vectorp = (char **)proc_vector32; *vsizep = vsize; return (0); } proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK); for (i = 0; i < (int)vsize; i++) proc_vector[i] = PTRIN(proc_vector32[i]); *proc_vectorp = proc_vector; *vsizep = vsize; done: free(proc_vector32, M_TEMP); return (error); } #endif static int get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp, size_t *vsizep, enum proc_vector_type type) { struct ps_strings pss; Elf_Auxinfo aux; vm_offset_t vptr, ptr; char **proc_vector; size_t vsize, size; int error, i; #ifdef COMPAT_FREEBSD32 if (SV_PROC_FLAG(p, SV_ILP32) != 0) return (get_proc_vector32(td, p, proc_vectorp, vsizep, type)); #endif error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings), &pss, sizeof(pss)); if (error != 0) return (error); switch (type) { case PROC_ARG: vptr = (vm_offset_t)pss.ps_argvstr; vsize = pss.ps_nargvstr; if (vsize > ARG_MAX) return (ENOEXEC); size = vsize * sizeof(char *); break; case PROC_ENV: vptr = (vm_offset_t)pss.ps_envstr; vsize = pss.ps_nenvstr; if (vsize > ARG_MAX) return (ENOEXEC); size = vsize * sizeof(char *); break; case PROC_AUX: /* * The aux array is just above env array on the stack. Check * that the address is naturally aligned. */ vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1) * sizeof(char *); #if __ELF_WORD_SIZE == 64 if (vptr % sizeof(uint64_t) != 0) #else if (vptr % sizeof(uint32_t) != 0) #endif return (ENOEXEC); /* * We count the array size reading the aux vectors from the * stack until AT_NULL vector is returned. So (to keep the code * simple) we read the process stack twice: the first time here * to find the size and the second time when copying the vectors * to the allocated proc_vector. */ for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) { error = proc_read_mem(td, p, ptr, &aux, sizeof(aux)); if (error != 0) return (error); if (aux.a_type == AT_NULL) break; ptr += sizeof(aux); } /* * If the PROC_AUXV_MAX entries are iterated over, and we have * not reached AT_NULL, it is most likely we are reading wrong * data: either the process doesn't have auxv array or data has * been modified. Return the error in this case. */ if (aux.a_type != AT_NULL) return (ENOEXEC); vsize = i + 1; size = vsize * sizeof(aux); break; default: KASSERT(0, ("Wrong proc vector type: %d", type)); return (EINVAL); /* In case we are built without INVARIANTS. */ } proc_vector = malloc(size, M_TEMP, M_WAITOK); if (proc_vector == NULL) return (ENOMEM); error = proc_read_mem(td, p, vptr, proc_vector, size); if (error != 0) { free(proc_vector, M_TEMP); return (error); } *proc_vectorp = proc_vector; *vsizep = vsize; return (0); } #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */ static int get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb, enum proc_vector_type type) { size_t done, len, nchr, vsize; int error, i; char **proc_vector, *sptr; char pss_string[GET_PS_STRINGS_CHUNK_SZ]; PROC_ASSERT_HELD(p); /* * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes. */ nchr = 2 * (PATH_MAX + ARG_MAX); error = get_proc_vector(td, p, &proc_vector, &vsize, type); if (error != 0) return (error); for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) { /* * The program may have scribbled into its argv array, e.g. to * remove some arguments. If that has happened, break out * before trying to read from NULL. */ if (proc_vector[i] == NULL) break; for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) { error = proc_read_string(td, p, sptr, pss_string, sizeof(pss_string)); if (error != 0) goto done; len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ); if (done + len >= nchr) len = nchr - done - 1; sbuf_bcat(sb, pss_string, len); if (len != GET_PS_STRINGS_CHUNK_SZ) break; done += GET_PS_STRINGS_CHUNK_SZ; } sbuf_bcat(sb, "", 1); done += len + 1; } done: free(proc_vector, M_TEMP); return (error); } int proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb) { return (get_ps_strings(curthread, p, sb, PROC_ARG)); } int proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb) { return (get_ps_strings(curthread, p, sb, PROC_ENV)); } int proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb) { size_t vsize, size; char **auxv; int error; error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX); if (error == 0) { #ifdef COMPAT_FREEBSD32 if (SV_PROC_FLAG(p, SV_ILP32) != 0) size = vsize * sizeof(Elf32_Auxinfo); else #endif size = vsize * sizeof(Elf_Auxinfo); if (sbuf_bcat(sb, auxv, size) != 0) error = ENOMEM; free(auxv, M_TEMP); } return (error); } /* * This sysctl allows a process to retrieve the argument list or process * title for another process without groping around in the address space * of the other process. It also allow a process to set its own "process * title to a string of its own choice. */ static int sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS) { int *name = (int *)arg1; u_int namelen = arg2; struct pargs *newpa, *pa; struct proc *p; struct sbuf sb; int flags, error = 0, error2; if (namelen != 1) return (EINVAL); flags = PGET_CANSEE; if (req->newptr != NULL) flags |= PGET_ISCURRENT; error = pget((pid_t)name[0], flags, &p); if (error) return (error); pa = p->p_args; if (pa != NULL) { pargs_hold(pa); PROC_UNLOCK(p); error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length); pargs_drop(pa); } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) { _PHOLD(p); PROC_UNLOCK(p); sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); sbuf_clear_flags(&sb, SBUF_INCLUDENUL); error = proc_getargv(curthread, p, &sb); error2 = sbuf_finish(&sb); PRELE(p); sbuf_delete(&sb); if (error == 0 && error2 != 0) error = error2; } else { PROC_UNLOCK(p); } if (error != 0 || req->newptr == NULL) return (error); if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit) return (ENOMEM); newpa = pargs_alloc(req->newlen); error = SYSCTL_IN(req, newpa->ar_args, req->newlen); if (error != 0) { pargs_free(newpa); return (error); } PROC_LOCK(p); pa = p->p_args; p->p_args = newpa; PROC_UNLOCK(p); pargs_drop(pa); return (0); } /* * This sysctl allows a process to retrieve environment of another process. */ static int sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS) { int *name = (int *)arg1; u_int namelen = arg2; struct proc *p; struct sbuf sb; int error, error2; if (namelen != 1) return (EINVAL); error = pget((pid_t)name[0], PGET_WANTREAD, &p); if (error != 0) return (error); if ((p->p_flag & P_SYSTEM) != 0) { PRELE(p); return (0); } sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); sbuf_clear_flags(&sb, SBUF_INCLUDENUL); error = proc_getenvv(curthread, p, &sb); error2 = sbuf_finish(&sb); PRELE(p); sbuf_delete(&sb); return (error != 0 ? error : error2); } /* * This sysctl allows a process to retrieve ELF auxiliary vector of * another process. */ static int sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS) { int *name = (int *)arg1; u_int namelen = arg2; struct proc *p; struct sbuf sb; int error, error2; if (namelen != 1) return (EINVAL); error = pget((pid_t)name[0], PGET_WANTREAD, &p); if (error != 0) return (error); if ((p->p_flag & P_SYSTEM) != 0) { PRELE(p); return (0); } sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req); sbuf_clear_flags(&sb, SBUF_INCLUDENUL); error = proc_getauxv(curthread, p, &sb); error2 = sbuf_finish(&sb); PRELE(p); sbuf_delete(&sb); return (error != 0 ? error : error2); } /* * This sysctl allows a process to retrieve the path of the executable for * itself or another process. */ static int sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS) { pid_t *pidp = (pid_t *)arg1; unsigned int arglen = arg2; struct proc *p; struct vnode *vp; char *retbuf, *freebuf; int error; if (arglen != 1) return (EINVAL); if (*pidp == -1) { /* -1 means this process */ p = req->td->td_proc; } else { error = pget(*pidp, PGET_CANSEE, &p); if (error != 0) return (error); } vp = p->p_textvp; if (vp == NULL) { if (*pidp != -1) PROC_UNLOCK(p); return (0); } vref(vp); if (*pidp != -1) PROC_UNLOCK(p); error = vn_fullpath(req->td, vp, &retbuf, &freebuf); vrele(vp); if (error) return (error); error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1); free(freebuf, M_TEMP); return (error); } static int sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS) { struct proc *p; char *sv_name; int *name; int namelen; int error; namelen = arg2; if (namelen != 1) return (EINVAL); name = (int *)arg1; error = pget((pid_t)name[0], PGET_CANSEE, &p); if (error != 0) return (error); sv_name = p->p_sysent->sv_name; PROC_UNLOCK(p); return (sysctl_handle_string(oidp, sv_name, 0, req)); } #ifdef KINFO_OVMENTRY_SIZE CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE); #endif #ifdef COMPAT_FREEBSD7 static int sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS) { vm_map_entry_t entry, tmp_entry; unsigned int last_timestamp; char *fullpath, *freepath; struct kinfo_ovmentry *kve; struct vattr va; struct ucred *cred; int error, *name; struct vnode *vp; struct proc *p; vm_map_t map; struct vmspace *vm; name = (int *)arg1; error = pget((pid_t)name[0], PGET_WANTREAD, &p); if (error != 0) return (error); vm = vmspace_acquire_ref(p); if (vm == NULL) { PRELE(p); return (ESRCH); } kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); map = &vm->vm_map; vm_map_lock_read(map); for (entry = map->header.next; entry != &map->header; entry = entry->next) { vm_object_t obj, tobj, lobj; vm_offset_t addr; if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) continue; bzero(kve, sizeof(*kve)); kve->kve_structsize = sizeof(*kve); kve->kve_private_resident = 0; obj = entry->object.vm_object; if (obj != NULL) { VM_OBJECT_RLOCK(obj); if (obj->shadow_count == 1) kve->kve_private_resident = obj->resident_page_count; } kve->kve_resident = 0; addr = entry->start; while (addr < entry->end) { if (pmap_extract(map->pmap, addr)) kve->kve_resident++; addr += PAGE_SIZE; } for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) { if (tobj != obj) VM_OBJECT_RLOCK(tobj); if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); lobj = tobj; } kve->kve_start = (void*)entry->start; kve->kve_end = (void*)entry->end; kve->kve_offset = (off_t)entry->offset; if (entry->protection & VM_PROT_READ) kve->kve_protection |= KVME_PROT_READ; if (entry->protection & VM_PROT_WRITE) kve->kve_protection |= KVME_PROT_WRITE; if (entry->protection & VM_PROT_EXECUTE) kve->kve_protection |= KVME_PROT_EXEC; if (entry->eflags & MAP_ENTRY_COW) kve->kve_flags |= KVME_FLAG_COW; if (entry->eflags & MAP_ENTRY_NEEDS_COPY) kve->kve_flags |= KVME_FLAG_NEEDS_COPY; if (entry->eflags & MAP_ENTRY_NOCOREDUMP) kve->kve_flags |= KVME_FLAG_NOCOREDUMP; last_timestamp = map->timestamp; vm_map_unlock_read(map); kve->kve_fileid = 0; kve->kve_fsid = 0; freepath = NULL; fullpath = ""; if (lobj) { vp = NULL; switch (lobj->type) { case OBJT_DEFAULT: kve->kve_type = KVME_TYPE_DEFAULT; break; case OBJT_VNODE: kve->kve_type = KVME_TYPE_VNODE; vp = lobj->handle; vref(vp); break; case OBJT_SWAP: - kve->kve_type = KVME_TYPE_SWAP; + if ((lobj->flags & OBJ_TMPFS_NODE) != 0) { + kve->kve_type = KVME_TYPE_VNODE; + if ((lobj->flags & OBJ_TMPFS) != 0) { + vp = lobj->un_pager.swp.swp_tmpfs; + vref(vp); + } + } else { + kve->kve_type = KVME_TYPE_SWAP; + } break; case OBJT_DEVICE: kve->kve_type = KVME_TYPE_DEVICE; break; case OBJT_PHYS: kve->kve_type = KVME_TYPE_PHYS; break; case OBJT_DEAD: kve->kve_type = KVME_TYPE_DEAD; break; case OBJT_SG: kve->kve_type = KVME_TYPE_SG; break; default: kve->kve_type = KVME_TYPE_UNKNOWN; break; } if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); kve->kve_ref_count = obj->ref_count; kve->kve_shadow_count = obj->shadow_count; VM_OBJECT_RUNLOCK(obj); if (vp != NULL) { vn_fullpath(curthread, vp, &fullpath, &freepath); cred = curthread->td_ucred; vn_lock(vp, LK_SHARED | LK_RETRY); if (VOP_GETATTR(vp, &va, cred) == 0) { kve->kve_fileid = va.va_fileid; kve->kve_fsid = va.va_fsid; } vput(vp); } } else { kve->kve_type = KVME_TYPE_NONE; kve->kve_ref_count = 0; kve->kve_shadow_count = 0; } strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); if (freepath != NULL) free(freepath, M_TEMP); error = SYSCTL_OUT(req, kve, sizeof(*kve)); vm_map_lock_read(map); if (error) break; if (last_timestamp != map->timestamp) { vm_map_lookup_entry(map, addr - 1, &tmp_entry); entry = tmp_entry; } } vm_map_unlock_read(map); vmspace_free(vm); PRELE(p); free(kve, M_TEMP); return (error); } #endif /* COMPAT_FREEBSD7 */ #ifdef KINFO_VMENTRY_SIZE CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE); #endif static void kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry, struct kinfo_vmentry *kve) { vm_object_t obj, tobj; vm_page_t m, m_adv; vm_offset_t addr; vm_paddr_t locked_pa; vm_pindex_t pi, pi_adv, pindex; locked_pa = 0; obj = entry->object.vm_object; addr = entry->start; m_adv = NULL; pi = OFF_TO_IDX(entry->offset); for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) { if (m_adv != NULL) { m = m_adv; } else { pi_adv = OFF_TO_IDX(entry->end - addr); pindex = pi; for (tobj = obj;; tobj = tobj->backing_object) { m = vm_page_find_least(tobj, pindex); if (m != NULL) { if (m->pindex == pindex) break; if (pi_adv > m->pindex - pindex) { pi_adv = m->pindex - pindex; m_adv = m; } } if (tobj->backing_object == NULL) goto next; pindex += OFF_TO_IDX(tobj-> backing_object_offset); } } m_adv = NULL; if (m->psind != 0 && addr + pagesizes[1] <= entry->end && (addr & (pagesizes[1] - 1)) == 0 && (pmap_mincore(map->pmap, addr, &locked_pa) & MINCORE_SUPER) != 0) { kve->kve_flags |= KVME_FLAG_SUPER; pi_adv = OFF_TO_IDX(pagesizes[1]); } else { /* * We do not test the found page on validity. * Either the page is busy and being paged in, * or it was invalidated. The first case * should be counted as resident, the second * is not so clear; we do account both. */ pi_adv = 1; } kve->kve_resident += pi_adv; next:; } PA_UNLOCK_COND(locked_pa); } /* * Must be called with the process locked and will return unlocked. */ int kern_proc_vmmap_out(struct proc *p, struct sbuf *sb) { vm_map_entry_t entry, tmp_entry; struct vattr va; vm_map_t map; vm_object_t obj, tobj, lobj; char *fullpath, *freepath; struct kinfo_vmentry *kve; struct ucred *cred; struct vnode *vp; struct vmspace *vm; vm_offset_t addr; unsigned int last_timestamp; int error; PROC_LOCK_ASSERT(p, MA_OWNED); _PHOLD(p); PROC_UNLOCK(p); vm = vmspace_acquire_ref(p); if (vm == NULL) { PRELE(p); return (ESRCH); } kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK); error = 0; map = &vm->vm_map; vm_map_lock_read(map); for (entry = map->header.next; entry != &map->header; entry = entry->next) { if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) continue; addr = entry->end; bzero(kve, sizeof(*kve)); obj = entry->object.vm_object; if (obj != NULL) { for (tobj = obj; tobj != NULL; tobj = tobj->backing_object) { VM_OBJECT_RLOCK(tobj); lobj = tobj; } if (obj->backing_object == NULL) kve->kve_private_resident = obj->resident_page_count; if (!vmmap_skip_res_cnt) kern_proc_vmmap_resident(map, entry, kve); for (tobj = obj; tobj != NULL; tobj = tobj->backing_object) { if (tobj != obj && tobj != lobj) VM_OBJECT_RUNLOCK(tobj); } } else { lobj = NULL; } kve->kve_start = entry->start; kve->kve_end = entry->end; kve->kve_offset = entry->offset; if (entry->protection & VM_PROT_READ) kve->kve_protection |= KVME_PROT_READ; if (entry->protection & VM_PROT_WRITE) kve->kve_protection |= KVME_PROT_WRITE; if (entry->protection & VM_PROT_EXECUTE) kve->kve_protection |= KVME_PROT_EXEC; if (entry->eflags & MAP_ENTRY_COW) kve->kve_flags |= KVME_FLAG_COW; if (entry->eflags & MAP_ENTRY_NEEDS_COPY) kve->kve_flags |= KVME_FLAG_NEEDS_COPY; if (entry->eflags & MAP_ENTRY_NOCOREDUMP) kve->kve_flags |= KVME_FLAG_NOCOREDUMP; if (entry->eflags & MAP_ENTRY_GROWS_UP) kve->kve_flags |= KVME_FLAG_GROWS_UP; if (entry->eflags & MAP_ENTRY_GROWS_DOWN) kve->kve_flags |= KVME_FLAG_GROWS_DOWN; last_timestamp = map->timestamp; vm_map_unlock_read(map); freepath = NULL; fullpath = ""; if (lobj != NULL) { vp = NULL; switch (lobj->type) { case OBJT_DEFAULT: kve->kve_type = KVME_TYPE_DEFAULT; break; case OBJT_VNODE: kve->kve_type = KVME_TYPE_VNODE; vp = lobj->handle; vref(vp); break; case OBJT_SWAP: - kve->kve_type = KVME_TYPE_SWAP; + if ((lobj->flags & OBJ_TMPFS_NODE) != 0) { + kve->kve_type = KVME_TYPE_VNODE; + if ((lobj->flags & OBJ_TMPFS) != 0) { + vp = lobj->un_pager.swp.swp_tmpfs; + vref(vp); + } + } else { + kve->kve_type = KVME_TYPE_SWAP; + } break; case OBJT_DEVICE: kve->kve_type = KVME_TYPE_DEVICE; break; case OBJT_PHYS: kve->kve_type = KVME_TYPE_PHYS; break; case OBJT_DEAD: kve->kve_type = KVME_TYPE_DEAD; break; case OBJT_SG: kve->kve_type = KVME_TYPE_SG; break; case OBJT_MGTDEVICE: kve->kve_type = KVME_TYPE_MGTDEVICE; break; default: kve->kve_type = KVME_TYPE_UNKNOWN; break; } if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); kve->kve_ref_count = obj->ref_count; kve->kve_shadow_count = obj->shadow_count; VM_OBJECT_RUNLOCK(obj); if (vp != NULL) { vn_fullpath(curthread, vp, &fullpath, &freepath); kve->kve_vn_type = vntype_to_kinfo(vp->v_type); cred = curthread->td_ucred; vn_lock(vp, LK_SHARED | LK_RETRY); if (VOP_GETATTR(vp, &va, cred) == 0) { kve->kve_vn_fileid = va.va_fileid; kve->kve_vn_fsid = va.va_fsid; kve->kve_vn_mode = MAKEIMODE(va.va_type, va.va_mode); kve->kve_vn_size = va.va_size; kve->kve_vn_rdev = va.va_rdev; kve->kve_status = KF_ATTR_VALID; } vput(vp); } } else { kve->kve_type = KVME_TYPE_NONE; kve->kve_ref_count = 0; kve->kve_shadow_count = 0; } strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path)); if (freepath != NULL) free(freepath, M_TEMP); /* Pack record size down */ kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) + strlen(kve->kve_path) + 1; kve->kve_structsize = roundup(kve->kve_structsize, sizeof(uint64_t)); if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0) error = ENOMEM; vm_map_lock_read(map); if (error != 0) break; if (last_timestamp != map->timestamp) { vm_map_lookup_entry(map, addr - 1, &tmp_entry); entry = tmp_entry; } } vm_map_unlock_read(map); vmspace_free(vm); PRELE(p); free(kve, M_TEMP); return (error); } static int sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS) { struct proc *p; struct sbuf sb; int error, error2, *name; name = (int *)arg1; sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req); sbuf_clear_flags(&sb, SBUF_INCLUDENUL); error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p); if (error != 0) { sbuf_delete(&sb); return (error); } error = kern_proc_vmmap_out(p, &sb); error2 = sbuf_finish(&sb); sbuf_delete(&sb); return (error != 0 ? error : error2); } #if defined(STACK) || defined(DDB) static int sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS) { struct kinfo_kstack *kkstp; int error, i, *name, numthreads; lwpid_t *lwpidarray; struct thread *td; struct stack *st; struct sbuf sb; struct proc *p; name = (int *)arg1; error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p); if (error != 0) return (error); kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK); st = stack_create(); lwpidarray = NULL; numthreads = 0; PROC_LOCK(p); repeat: if (numthreads < p->p_numthreads) { if (lwpidarray != NULL) { free(lwpidarray, M_TEMP); lwpidarray = NULL; } numthreads = p->p_numthreads; PROC_UNLOCK(p); lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP, M_WAITOK | M_ZERO); PROC_LOCK(p); goto repeat; } i = 0; /* * XXXRW: During the below loop, execve(2) and countless other sorts * of changes could have taken place. Should we check to see if the * vmspace has been replaced, or the like, in order to prevent * giving a snapshot that spans, say, execve(2), with some threads * before and some after? Among other things, the credentials could * have changed, in which case the right to extract debug info might * no longer be assured. */ FOREACH_THREAD_IN_PROC(p, td) { KASSERT(i < numthreads, ("sysctl_kern_proc_kstack: numthreads")); lwpidarray[i] = td->td_tid; i++; } numthreads = i; for (i = 0; i < numthreads; i++) { td = thread_find(p, lwpidarray[i]); if (td == NULL) { continue; } bzero(kkstp, sizeof(*kkstp)); (void)sbuf_new(&sb, kkstp->kkst_trace, sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN); thread_lock(td); kkstp->kkst_tid = td->td_tid; if (TD_IS_SWAPPED(td)) kkstp->kkst_state = KKST_STATE_SWAPPED; else if (TD_IS_RUNNING(td)) kkstp->kkst_state = KKST_STATE_RUNNING; else { kkstp->kkst_state = KKST_STATE_STACKOK; stack_save_td(st, td); } thread_unlock(td); PROC_UNLOCK(p); stack_sbuf_print(&sb, st); sbuf_finish(&sb); sbuf_delete(&sb); error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp)); PROC_LOCK(p); if (error) break; } _PRELE(p); PROC_UNLOCK(p); if (lwpidarray != NULL) free(lwpidarray, M_TEMP); stack_destroy(st); free(kkstp, M_TEMP); return (error); } #endif /* * This sysctl allows a process to retrieve the full list of groups from * itself or another process. */ static int sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS) { pid_t *pidp = (pid_t *)arg1; unsigned int arglen = arg2; struct proc *p; struct ucred *cred; int error; if (arglen != 1) return (EINVAL); if (*pidp == -1) { /* -1 means this process */ p = req->td->td_proc; PROC_LOCK(p); } else { error = pget(*pidp, PGET_CANSEE, &p); if (error != 0) return (error); } cred = crhold(p->p_ucred); PROC_UNLOCK(p); error = SYSCTL_OUT(req, cred->cr_groups, cred->cr_ngroups * sizeof(gid_t)); crfree(cred); return (error); } /* * This sysctl allows a process to retrieve or/and set the resource limit for * another process. */ static int sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS) { int *name = (int *)arg1; u_int namelen = arg2; struct rlimit rlim; struct proc *p; u_int which; int flags, error; if (namelen != 2) return (EINVAL); which = (u_int)name[1]; if (which >= RLIM_NLIMITS) return (EINVAL); if (req->newptr != NULL && req->newlen != sizeof(rlim)) return (EINVAL); flags = PGET_HOLD | PGET_NOTWEXIT; if (req->newptr != NULL) flags |= PGET_CANDEBUG; else flags |= PGET_CANSEE; error = pget((pid_t)name[0], flags, &p); if (error != 0) return (error); /* * Retrieve limit. */ if (req->oldptr != NULL) { PROC_LOCK(p); lim_rlimit(p, which, &rlim); PROC_UNLOCK(p); } error = SYSCTL_OUT(req, &rlim, sizeof(rlim)); if (error != 0) goto errout; /* * Set limit. */ if (req->newptr != NULL) { error = SYSCTL_IN(req, &rlim, sizeof(rlim)); if (error == 0) error = kern_proc_setrlimit(curthread, p, which, &rlim); } errout: PRELE(p); return (error); } /* * This sysctl allows a process to retrieve ps_strings structure location of * another process. */ static int sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS) { int *name = (int *)arg1; u_int namelen = arg2; struct proc *p; vm_offset_t ps_strings; int error; #ifdef COMPAT_FREEBSD32 uint32_t ps_strings32; #endif if (namelen != 1) return (EINVAL); error = pget((pid_t)name[0], PGET_CANDEBUG, &p); if (error != 0) return (error); #ifdef COMPAT_FREEBSD32 if ((req->flags & SCTL_MASK32) != 0) { /* * We return 0 if the 32 bit emulation request is for a 64 bit * process. */ ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ? PTROUT(p->p_sysent->sv_psstrings) : 0; PROC_UNLOCK(p); error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32)); return (error); } #endif ps_strings = p->p_sysent->sv_psstrings; PROC_UNLOCK(p); error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings)); return (error); } /* * This sysctl allows a process to retrieve umask of another process. */ static int sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS) { int *name = (int *)arg1; u_int namelen = arg2; struct proc *p; int error; u_short fd_cmask; if (namelen != 1) return (EINVAL); error = pget((pid_t)name[0], PGET_WANTREAD, &p); if (error != 0) return (error); FILEDESC_SLOCK(p->p_fd); fd_cmask = p->p_fd->fd_cmask; FILEDESC_SUNLOCK(p->p_fd); PRELE(p); error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask)); return (error); } /* * This sysctl allows a process to set and retrieve binary osreldate of * another process. */ static int sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS) { int *name = (int *)arg1; u_int namelen = arg2; struct proc *p; int flags, error, osrel; if (namelen != 1) return (EINVAL); if (req->newptr != NULL && req->newlen != sizeof(osrel)) return (EINVAL); flags = PGET_HOLD | PGET_NOTWEXIT; if (req->newptr != NULL) flags |= PGET_CANDEBUG; else flags |= PGET_CANSEE; error = pget((pid_t)name[0], flags, &p); if (error != 0) return (error); error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel)); if (error != 0) goto errout; if (req->newptr != NULL) { error = SYSCTL_IN(req, &osrel, sizeof(osrel)); if (error != 0) goto errout; if (osrel < 0) { error = EINVAL; goto errout; } p->p_osrel = osrel; } errout: PRELE(p); return (error); } static int sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS) { int *name = (int *)arg1; u_int namelen = arg2; struct proc *p; struct kinfo_sigtramp kst; const struct sysentvec *sv; int error; #ifdef COMPAT_FREEBSD32 struct kinfo_sigtramp32 kst32; #endif if (namelen != 1) return (EINVAL); error = pget((pid_t)name[0], PGET_CANDEBUG, &p); if (error != 0) return (error); sv = p->p_sysent; #ifdef COMPAT_FREEBSD32 if ((req->flags & SCTL_MASK32) != 0) { bzero(&kst32, sizeof(kst32)); if (SV_PROC_FLAG(p, SV_ILP32)) { if (sv->sv_sigcode_base != 0) { kst32.ksigtramp_start = sv->sv_sigcode_base; kst32.ksigtramp_end = sv->sv_sigcode_base + *sv->sv_szsigcode; } else { kst32.ksigtramp_start = sv->sv_psstrings - *sv->sv_szsigcode; kst32.ksigtramp_end = sv->sv_psstrings; } } PROC_UNLOCK(p); error = SYSCTL_OUT(req, &kst32, sizeof(kst32)); return (error); } #endif bzero(&kst, sizeof(kst)); if (sv->sv_sigcode_base != 0) { kst.ksigtramp_start = (char *)sv->sv_sigcode_base; kst.ksigtramp_end = (char *)sv->sv_sigcode_base + *sv->sv_szsigcode; } else { kst.ksigtramp_start = (char *)sv->sv_psstrings - *sv->sv_szsigcode; kst.ksigtramp_end = (char *)sv->sv_psstrings; } PROC_UNLOCK(p); error = SYSCTL_OUT(req, &kst, sizeof(kst)); return (error); } SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table"); SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT| CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc", "Return entire process table"); static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Return process table, no threads"); static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_args, "Process argument list"); static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_env, "Process environment"); static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector"); static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path"); static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name, "Process syscall vector name (ABI type)"); static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD), sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table"); static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Return process table, no threads"); #ifdef COMPAT_FREEBSD7 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries"); #endif static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries"); #if defined(STACK) || defined(DDB) static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks"); #endif static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups"); static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit, "Process resource limits"); static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings, "Process ps_strings location"); static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask"); static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel, "Process binary osreldate"); static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp, "Process signal trampoline location"); int allproc_gen; void stop_all_proc(void) { struct proc *cp, *p; int r, gen; bool restart, seen_stopped, seen_exiting, stopped_some; cp = curproc; /* * stop_all_proc() assumes that all process which have * usermode must be stopped, except current process, for * obvious reasons. Since other threads in the process * establishing global stop could unstop something, disable * calls from multithreaded processes as precaution. The * service must not be user-callable anyway. */ KASSERT((cp->p_flag & P_HADTHREADS) == 0 || (cp->p_flag & P_KTHREAD) != 0, ("mt stop_all_proc")); allproc_loop: sx_xlock(&allproc_lock); gen = allproc_gen; seen_exiting = seen_stopped = stopped_some = restart = false; LIST_REMOVE(cp, p_list); LIST_INSERT_HEAD(&allproc, cp, p_list); for (;;) { p = LIST_NEXT(cp, p_list); if (p == NULL) break; LIST_REMOVE(cp, p_list); LIST_INSERT_AFTER(p, cp, p_list); PROC_LOCK(p); if ((p->p_flag & (P_KTHREAD | P_SYSTEM | P_TOTAL_STOP)) != 0) { PROC_UNLOCK(p); continue; } if ((p->p_flag & P_WEXIT) != 0) { seen_exiting = true; PROC_UNLOCK(p); continue; } if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) { /* * Stopped processes are tolerated when there * are no other processes which might continue * them. P_STOPPED_SINGLE but not * P_TOTAL_STOP process still has at least one * thread running. */ seen_stopped = true; PROC_UNLOCK(p); continue; } _PHOLD(p); sx_xunlock(&allproc_lock); r = thread_single(p, SINGLE_ALLPROC); if (r != 0) restart = true; else stopped_some = true; _PRELE(p); PROC_UNLOCK(p); sx_xlock(&allproc_lock); } /* Catch forked children we did not see in iteration. */ if (gen != allproc_gen) restart = true; sx_xunlock(&allproc_lock); if (restart || stopped_some || seen_exiting || seen_stopped) { kern_yield(PRI_USER); goto allproc_loop; } } void resume_all_proc(void) { struct proc *cp, *p; cp = curproc; sx_xlock(&allproc_lock); LIST_REMOVE(cp, p_list); LIST_INSERT_HEAD(&allproc, cp, p_list); for (;;) { p = LIST_NEXT(cp, p_list); if (p == NULL) break; LIST_REMOVE(cp, p_list); LIST_INSERT_AFTER(p, cp, p_list); PROC_LOCK(p); if ((p->p_flag & P_TOTAL_STOP) != 0) { sx_xunlock(&allproc_lock); _PHOLD(p); thread_single_end(p, SINGLE_ALLPROC); _PRELE(p); PROC_UNLOCK(p); sx_xlock(&allproc_lock); } else { PROC_UNLOCK(p); } } sx_xunlock(&allproc_lock); } #define TOTAL_STOP_DEBUG 1 #ifdef TOTAL_STOP_DEBUG volatile static int ap_resume; #include static int sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS) { int error, val; val = 0; ap_resume = 0; error = sysctl_handle_int(oidp, &val, 0, req); if (error != 0 || req->newptr == NULL) return (error); if (val != 0) { stop_all_proc(); syncer_suspend(); while (ap_resume == 0) ; syncer_resume(); resume_all_proc(); } return (0); } SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0, sysctl_debug_stop_all_proc, "I", ""); #endif Index: projects/em_mq/sys/kern/sys_process.c =================================================================== --- projects/em_mq/sys/kern/sys_process.c (revision 283943) +++ projects/em_mq/sys/kern/sys_process.c (revision 283944) @@ -1,1309 +1,1309 @@ /*- * Copyright (c) 1994, Sean Eric Fagan * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by Sean Eric Fagan. * 4. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef COMPAT_FREEBSD32 #include #include struct ptrace_io_desc32 { int piod_op; uint32_t piod_offs; uint32_t piod_addr; uint32_t piod_len; }; struct ptrace_vm_entry32 { int pve_entry; int pve_timestamp; uint32_t pve_start; uint32_t pve_end; uint32_t pve_offset; u_int pve_prot; u_int pve_pathlen; int32_t pve_fileid; u_int pve_fsid; uint32_t pve_path; }; struct ptrace_lwpinfo32 { lwpid_t pl_lwpid; /* LWP described. */ int pl_event; /* Event that stopped the LWP. */ int pl_flags; /* LWP flags. */ sigset_t pl_sigmask; /* LWP signal mask */ sigset_t pl_siglist; /* LWP pending signal */ struct siginfo32 pl_siginfo; /* siginfo for signal */ char pl_tdname[MAXCOMLEN + 1]; /* LWP name. */ int pl_child_pid; /* New child pid */ }; #endif /* * Functions implemented using PROC_ACTION(): * * proc_read_regs(proc, regs) * Get the current user-visible register set from the process * and copy it into the regs structure (). * The process is stopped at the time read_regs is called. * * proc_write_regs(proc, regs) * Update the current register set from the passed in regs * structure. Take care to avoid clobbering special CPU * registers or privileged bits in the PSL. * Depending on the architecture this may have fix-up work to do, * especially if the IAR or PCW are modified. * The process is stopped at the time write_regs is called. * * proc_read_fpregs, proc_write_fpregs * deal with the floating point register set, otherwise as above. * * proc_read_dbregs, proc_write_dbregs * deal with the processor debug register set, otherwise as above. * * proc_sstep(proc) * Arrange for the process to trap after executing a single instruction. */ #define PROC_ACTION(action) do { \ int error; \ \ PROC_LOCK_ASSERT(td->td_proc, MA_OWNED); \ if ((td->td_proc->p_flag & P_INMEM) == 0) \ error = EIO; \ else \ error = (action); \ return (error); \ } while(0) int proc_read_regs(struct thread *td, struct reg *regs) { PROC_ACTION(fill_regs(td, regs)); } int proc_write_regs(struct thread *td, struct reg *regs) { PROC_ACTION(set_regs(td, regs)); } int proc_read_dbregs(struct thread *td, struct dbreg *dbregs) { PROC_ACTION(fill_dbregs(td, dbregs)); } int proc_write_dbregs(struct thread *td, struct dbreg *dbregs) { PROC_ACTION(set_dbregs(td, dbregs)); } /* * Ptrace doesn't support fpregs at all, and there are no security holes * or translations for fpregs, so we can just copy them. */ int proc_read_fpregs(struct thread *td, struct fpreg *fpregs) { PROC_ACTION(fill_fpregs(td, fpregs)); } int proc_write_fpregs(struct thread *td, struct fpreg *fpregs) { PROC_ACTION(set_fpregs(td, fpregs)); } #ifdef COMPAT_FREEBSD32 /* For 32 bit binaries, we need to expose the 32 bit regs layouts. */ int proc_read_regs32(struct thread *td, struct reg32 *regs32) { PROC_ACTION(fill_regs32(td, regs32)); } int proc_write_regs32(struct thread *td, struct reg32 *regs32) { PROC_ACTION(set_regs32(td, regs32)); } int proc_read_dbregs32(struct thread *td, struct dbreg32 *dbregs32) { PROC_ACTION(fill_dbregs32(td, dbregs32)); } int proc_write_dbregs32(struct thread *td, struct dbreg32 *dbregs32) { PROC_ACTION(set_dbregs32(td, dbregs32)); } int proc_read_fpregs32(struct thread *td, struct fpreg32 *fpregs32) { PROC_ACTION(fill_fpregs32(td, fpregs32)); } int proc_write_fpregs32(struct thread *td, struct fpreg32 *fpregs32) { PROC_ACTION(set_fpregs32(td, fpregs32)); } #endif int proc_sstep(struct thread *td) { PROC_ACTION(ptrace_single_step(td)); } int proc_rwmem(struct proc *p, struct uio *uio) { vm_map_t map; vm_offset_t pageno; /* page number */ vm_prot_t reqprot; int error, fault_flags, page_offset, writing; /* * Assert that someone has locked this vmspace. (Should be * curthread but we can't assert that.) This keeps the process * from exiting out from under us until this operation completes. */ KASSERT(p->p_lock >= 1, ("%s: process %p (pid %d) not held", __func__, p, p->p_pid)); /* * The map we want... */ map = &p->p_vmspace->vm_map; /* * If we are writing, then we request vm_fault() to create a private * copy of each page. Since these copies will not be writeable by the * process, we must explicity request that they be dirtied. */ writing = uio->uio_rw == UIO_WRITE; reqprot = writing ? VM_PROT_COPY | VM_PROT_READ : VM_PROT_READ; fault_flags = writing ? VM_FAULT_DIRTY : VM_FAULT_NORMAL; /* * Only map in one page at a time. We don't have to, but it * makes things easier. This way is trivial - right? */ do { vm_offset_t uva; u_int len; vm_page_t m; uva = (vm_offset_t)uio->uio_offset; /* * Get the page number of this segment. */ pageno = trunc_page(uva); page_offset = uva - pageno; /* * How many bytes to copy */ len = min(PAGE_SIZE - page_offset, uio->uio_resid); /* * Fault and hold the page on behalf of the process. */ error = vm_fault_hold(map, pageno, reqprot, fault_flags, &m); if (error != KERN_SUCCESS) { if (error == KERN_RESOURCE_SHORTAGE) error = ENOMEM; else error = EFAULT; break; } /* * Now do the i/o move. */ error = uiomove_fromphys(&m, page_offset, len, uio); /* Make the I-cache coherent for breakpoints. */ if (writing && error == 0) { vm_map_lock_read(map); if (vm_map_check_protection(map, pageno, pageno + PAGE_SIZE, VM_PROT_EXECUTE)) vm_sync_icache(map, uva, len); vm_map_unlock_read(map); } /* * Release the page. */ vm_page_lock(m); vm_page_unhold(m); vm_page_unlock(m); } while (error == 0 && uio->uio_resid > 0); return (error); } static int ptrace_vm_entry(struct thread *td, struct proc *p, struct ptrace_vm_entry *pve) { struct vattr vattr; vm_map_t map; vm_map_entry_t entry; vm_object_t obj, tobj, lobj; struct vmspace *vm; struct vnode *vp; char *freepath, *fullpath; u_int pathlen; int error, index; error = 0; obj = NULL; vm = vmspace_acquire_ref(p); map = &vm->vm_map; vm_map_lock_read(map); do { entry = map->header.next; index = 0; while (index < pve->pve_entry && entry != &map->header) { entry = entry->next; index++; } if (index != pve->pve_entry) { error = EINVAL; break; } while (entry != &map->header && (entry->eflags & MAP_ENTRY_IS_SUB_MAP) != 0) { entry = entry->next; index++; } if (entry == &map->header) { error = ENOENT; break; } /* We got an entry. */ pve->pve_entry = index + 1; pve->pve_timestamp = map->timestamp; pve->pve_start = entry->start; pve->pve_end = entry->end - 1; pve->pve_offset = entry->offset; pve->pve_prot = entry->protection; /* Backing object's path needed? */ if (pve->pve_pathlen == 0) break; pathlen = pve->pve_pathlen; pve->pve_pathlen = 0; obj = entry->object.vm_object; if (obj != NULL) VM_OBJECT_RLOCK(obj); } while (0); vm_map_unlock_read(map); vmspace_free(vm); pve->pve_fsid = VNOVAL; pve->pve_fileid = VNOVAL; if (error == 0 && obj != NULL) { lobj = obj; for (tobj = obj; tobj != NULL; tobj = tobj->backing_object) { if (tobj != obj) VM_OBJECT_RLOCK(tobj); if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); lobj = tobj; pve->pve_offset += tobj->backing_object_offset; } - vp = (lobj->type == OBJT_VNODE) ? lobj->handle : NULL; + vp = vm_object_vnode(lobj); if (vp != NULL) vref(vp); if (lobj != obj) VM_OBJECT_RUNLOCK(lobj); VM_OBJECT_RUNLOCK(obj); if (vp != NULL) { freepath = NULL; fullpath = NULL; vn_fullpath(td, vp, &fullpath, &freepath); vn_lock(vp, LK_SHARED | LK_RETRY); if (VOP_GETATTR(vp, &vattr, td->td_ucred) == 0) { pve->pve_fileid = vattr.va_fileid; pve->pve_fsid = vattr.va_fsid; } vput(vp); if (fullpath != NULL) { pve->pve_pathlen = strlen(fullpath) + 1; if (pve->pve_pathlen <= pathlen) { error = copyout(fullpath, pve->pve_path, pve->pve_pathlen); } else error = ENAMETOOLONG; } if (freepath != NULL) free(freepath, M_TEMP); } } if (error == 0) CTR3(KTR_PTRACE, "PT_VM_ENTRY: pid %d, entry %d, start %p", p->p_pid, pve->pve_entry, pve->pve_start); return (error); } #ifdef COMPAT_FREEBSD32 static int ptrace_vm_entry32(struct thread *td, struct proc *p, struct ptrace_vm_entry32 *pve32) { struct ptrace_vm_entry pve; int error; pve.pve_entry = pve32->pve_entry; pve.pve_pathlen = pve32->pve_pathlen; pve.pve_path = (void *)(uintptr_t)pve32->pve_path; error = ptrace_vm_entry(td, p, &pve); if (error == 0) { pve32->pve_entry = pve.pve_entry; pve32->pve_timestamp = pve.pve_timestamp; pve32->pve_start = pve.pve_start; pve32->pve_end = pve.pve_end; pve32->pve_offset = pve.pve_offset; pve32->pve_prot = pve.pve_prot; pve32->pve_fileid = pve.pve_fileid; pve32->pve_fsid = pve.pve_fsid; } pve32->pve_pathlen = pve.pve_pathlen; return (error); } static void ptrace_lwpinfo_to32(const struct ptrace_lwpinfo *pl, struct ptrace_lwpinfo32 *pl32) { pl32->pl_lwpid = pl->pl_lwpid; pl32->pl_event = pl->pl_event; pl32->pl_flags = pl->pl_flags; pl32->pl_sigmask = pl->pl_sigmask; pl32->pl_siglist = pl->pl_siglist; siginfo_to_siginfo32(&pl->pl_siginfo, &pl32->pl_siginfo); strcpy(pl32->pl_tdname, pl->pl_tdname); pl32->pl_child_pid = pl->pl_child_pid; } #endif /* COMPAT_FREEBSD32 */ /* * Process debugging system call. */ #ifndef _SYS_SYSPROTO_H_ struct ptrace_args { int req; pid_t pid; caddr_t addr; int data; }; #endif #ifdef COMPAT_FREEBSD32 /* * This CPP subterfuge is to try and reduce the number of ifdefs in * the body of the code. * COPYIN(uap->addr, &r.reg, sizeof r.reg); * becomes either: * copyin(uap->addr, &r.reg, sizeof r.reg); * or * copyin(uap->addr, &r.reg32, sizeof r.reg32); * .. except this is done at runtime. */ #define COPYIN(u, k, s) wrap32 ? \ copyin(u, k ## 32, s ## 32) : \ copyin(u, k, s) #define COPYOUT(k, u, s) wrap32 ? \ copyout(k ## 32, u, s ## 32) : \ copyout(k, u, s) #else #define COPYIN(u, k, s) copyin(u, k, s) #define COPYOUT(k, u, s) copyout(k, u, s) #endif int sys_ptrace(struct thread *td, struct ptrace_args *uap) { /* * XXX this obfuscation is to reduce stack usage, but the register * structs may be too large to put on the stack anyway. */ union { struct ptrace_io_desc piod; struct ptrace_lwpinfo pl; struct ptrace_vm_entry pve; struct dbreg dbreg; struct fpreg fpreg; struct reg reg; #ifdef COMPAT_FREEBSD32 struct dbreg32 dbreg32; struct fpreg32 fpreg32; struct reg32 reg32; struct ptrace_io_desc32 piod32; struct ptrace_lwpinfo32 pl32; struct ptrace_vm_entry32 pve32; #endif } r; void *addr; int error = 0; #ifdef COMPAT_FREEBSD32 int wrap32 = 0; if (SV_CURPROC_FLAG(SV_ILP32)) wrap32 = 1; #endif AUDIT_ARG_PID(uap->pid); AUDIT_ARG_CMD(uap->req); AUDIT_ARG_VALUE(uap->data); addr = &r; switch (uap->req) { case PT_GETREGS: case PT_GETFPREGS: case PT_GETDBREGS: case PT_LWPINFO: break; case PT_SETREGS: error = COPYIN(uap->addr, &r.reg, sizeof r.reg); break; case PT_SETFPREGS: error = COPYIN(uap->addr, &r.fpreg, sizeof r.fpreg); break; case PT_SETDBREGS: error = COPYIN(uap->addr, &r.dbreg, sizeof r.dbreg); break; case PT_IO: error = COPYIN(uap->addr, &r.piod, sizeof r.piod); break; case PT_VM_ENTRY: error = COPYIN(uap->addr, &r.pve, sizeof r.pve); break; default: addr = uap->addr; break; } if (error) return (error); error = kern_ptrace(td, uap->req, uap->pid, addr, uap->data); if (error) return (error); switch (uap->req) { case PT_VM_ENTRY: error = COPYOUT(&r.pve, uap->addr, sizeof r.pve); break; case PT_IO: error = COPYOUT(&r.piod, uap->addr, sizeof r.piod); break; case PT_GETREGS: error = COPYOUT(&r.reg, uap->addr, sizeof r.reg); break; case PT_GETFPREGS: error = COPYOUT(&r.fpreg, uap->addr, sizeof r.fpreg); break; case PT_GETDBREGS: error = COPYOUT(&r.dbreg, uap->addr, sizeof r.dbreg); break; case PT_LWPINFO: error = copyout(&r.pl, uap->addr, uap->data); break; } return (error); } #undef COPYIN #undef COPYOUT #ifdef COMPAT_FREEBSD32 /* * PROC_READ(regs, td2, addr); * becomes either: * proc_read_regs(td2, addr); * or * proc_read_regs32(td2, addr); * .. except this is done at runtime. There is an additional * complication in that PROC_WRITE disallows 32 bit consumers * from writing to 64 bit address space targets. */ #define PROC_READ(w, t, a) wrap32 ? \ proc_read_ ## w ## 32(t, a) : \ proc_read_ ## w (t, a) #define PROC_WRITE(w, t, a) wrap32 ? \ (safe ? proc_write_ ## w ## 32(t, a) : EINVAL ) : \ proc_write_ ## w (t, a) #else #define PROC_READ(w, t, a) proc_read_ ## w (t, a) #define PROC_WRITE(w, t, a) proc_write_ ## w (t, a) #endif int kern_ptrace(struct thread *td, int req, pid_t pid, void *addr, int data) { struct iovec iov; struct uio uio; struct proc *curp, *p, *pp; struct thread *td2 = NULL, *td3; struct ptrace_io_desc *piod = NULL; struct ptrace_lwpinfo *pl; int error, write, tmp, num; int proctree_locked = 0; lwpid_t tid = 0, *buf; #ifdef COMPAT_FREEBSD32 int wrap32 = 0, safe = 0; struct ptrace_io_desc32 *piod32 = NULL; struct ptrace_lwpinfo32 *pl32 = NULL; struct ptrace_lwpinfo plr; #endif curp = td->td_proc; /* Lock proctree before locking the process. */ switch (req) { case PT_TRACE_ME: case PT_ATTACH: case PT_STEP: case PT_CONTINUE: case PT_TO_SCE: case PT_TO_SCX: case PT_SYSCALL: case PT_FOLLOW_FORK: case PT_DETACH: sx_xlock(&proctree_lock); proctree_locked = 1; break; default: break; } write = 0; if (req == PT_TRACE_ME) { p = td->td_proc; PROC_LOCK(p); } else { if (pid <= PID_MAX) { if ((p = pfind(pid)) == NULL) { if (proctree_locked) sx_xunlock(&proctree_lock); return (ESRCH); } } else { td2 = tdfind(pid, -1); if (td2 == NULL) { if (proctree_locked) sx_xunlock(&proctree_lock); return (ESRCH); } p = td2->td_proc; tid = pid; pid = p->p_pid; } } AUDIT_ARG_PROCESS(p); if ((p->p_flag & P_WEXIT) != 0) { error = ESRCH; goto fail; } if ((error = p_cansee(td, p)) != 0) goto fail; if ((error = p_candebug(td, p)) != 0) goto fail; /* * System processes can't be debugged. */ if ((p->p_flag & P_SYSTEM) != 0) { error = EINVAL; goto fail; } if (tid == 0) { if ((p->p_flag & P_STOPPED_TRACE) != 0) { KASSERT(p->p_xthread != NULL, ("NULL p_xthread")); td2 = p->p_xthread; } else { td2 = FIRST_THREAD_IN_PROC(p); } tid = td2->td_tid; } #ifdef COMPAT_FREEBSD32 /* * Test if we're a 32 bit client and what the target is. * Set the wrap controls accordingly. */ if (SV_CURPROC_FLAG(SV_ILP32)) { if (SV_PROC_FLAG(td2->td_proc, SV_ILP32)) safe = 1; wrap32 = 1; } #endif /* * Permissions check */ switch (req) { case PT_TRACE_ME: /* Always legal. */ break; case PT_ATTACH: /* Self */ if (p->p_pid == td->td_proc->p_pid) { error = EINVAL; goto fail; } /* Already traced */ if (p->p_flag & P_TRACED) { error = EBUSY; goto fail; } /* Can't trace an ancestor if you're being traced. */ if (curp->p_flag & P_TRACED) { for (pp = curp->p_pptr; pp != NULL; pp = pp->p_pptr) { if (pp == p) { error = EINVAL; goto fail; } } } /* OK */ break; case PT_CLEARSTEP: /* Allow thread to clear single step for itself */ if (td->td_tid == tid) break; /* FALLTHROUGH */ default: /* not being traced... */ if ((p->p_flag & P_TRACED) == 0) { error = EPERM; goto fail; } /* not being traced by YOU */ if (p->p_pptr != td->td_proc) { error = EBUSY; goto fail; } /* not currently stopped */ if ((p->p_flag & (P_STOPPED_SIG | P_STOPPED_TRACE)) == 0 || p->p_suspcount != p->p_numthreads || (p->p_flag & P_WAITED) == 0) { error = EBUSY; goto fail; } if ((p->p_flag & P_STOPPED_TRACE) == 0) { static int count = 0; if (count++ == 0) printf("P_STOPPED_TRACE not set.\n"); } /* OK */ break; } /* Keep this process around until we finish this request. */ _PHOLD(p); #ifdef FIX_SSTEP /* * Single step fixup ala procfs */ FIX_SSTEP(td2); #endif /* * Actually do the requests */ td->td_retval[0] = 0; switch (req) { case PT_TRACE_ME: /* set my trace flag and "owner" so it can read/write me */ p->p_flag |= P_TRACED; if (p->p_flag & P_PPWAIT) p->p_flag |= P_PPTRACE; p->p_oppid = p->p_pptr->p_pid; CTR1(KTR_PTRACE, "PT_TRACE_ME: pid %d", p->p_pid); break; case PT_ATTACH: /* security check done above */ /* * It would be nice if the tracing relationship was separate * from the parent relationship but that would require * another set of links in the proc struct or for "wait" * to scan the entire proc table. To make life easier, * we just re-parent the process we're trying to trace. * The old parent is remembered so we can put things back * on a "detach". */ p->p_flag |= P_TRACED; p->p_oppid = p->p_pptr->p_pid; if (p->p_pptr != td->td_proc) { proc_reparent(p, td->td_proc); } data = SIGSTOP; CTR2(KTR_PTRACE, "PT_ATTACH: pid %d, oppid %d", p->p_pid, p->p_oppid); goto sendsig; /* in PT_CONTINUE below */ case PT_CLEARSTEP: CTR2(KTR_PTRACE, "PT_CLEARSTEP: tid %d (pid %d)", td2->td_tid, p->p_pid); error = ptrace_clear_single_step(td2); break; case PT_SETSTEP: CTR2(KTR_PTRACE, "PT_SETSTEP: tid %d (pid %d)", td2->td_tid, p->p_pid); error = ptrace_single_step(td2); break; case PT_SUSPEND: CTR2(KTR_PTRACE, "PT_SUSPEND: tid %d (pid %d)", td2->td_tid, p->p_pid); td2->td_dbgflags |= TDB_SUSPEND; thread_lock(td2); td2->td_flags |= TDF_NEEDSUSPCHK; thread_unlock(td2); break; case PT_RESUME: CTR2(KTR_PTRACE, "PT_RESUME: tid %d (pid %d)", td2->td_tid, p->p_pid); td2->td_dbgflags &= ~TDB_SUSPEND; break; case PT_FOLLOW_FORK: CTR3(KTR_PTRACE, "PT_FOLLOW_FORK: pid %d %s -> %s", p->p_pid, p->p_flag & P_FOLLOWFORK ? "enabled" : "disabled", data ? "enabled" : "disabled"); if (data) p->p_flag |= P_FOLLOWFORK; else p->p_flag &= ~P_FOLLOWFORK; break; case PT_STEP: case PT_CONTINUE: case PT_TO_SCE: case PT_TO_SCX: case PT_SYSCALL: case PT_DETACH: /* Zero means do not send any signal */ if (data < 0 || data > _SIG_MAXSIG) { error = EINVAL; break; } switch (req) { case PT_STEP: CTR2(KTR_PTRACE, "PT_STEP: tid %d (pid %d)", td2->td_tid, p->p_pid); error = ptrace_single_step(td2); if (error) goto out; break; case PT_CONTINUE: case PT_TO_SCE: case PT_TO_SCX: case PT_SYSCALL: if (addr != (void *)1) { error = ptrace_set_pc(td2, (u_long)(uintfptr_t)addr); if (error) goto out; } switch (req) { case PT_TO_SCE: p->p_stops |= S_PT_SCE; CTR2(KTR_PTRACE, "PT_TO_SCE: pid %d, stops = %#x", p->p_pid, p->p_stops); break; case PT_TO_SCX: p->p_stops |= S_PT_SCX; CTR2(KTR_PTRACE, "PT_TO_SCX: pid %d, stops = %#x", p->p_pid, p->p_stops); break; case PT_SYSCALL: p->p_stops |= S_PT_SCE | S_PT_SCX; CTR2(KTR_PTRACE, "PT_SYSCALL: pid %d, stops = %#x", p->p_pid, p->p_stops); break; case PT_CONTINUE: CTR1(KTR_PTRACE, "PT_CONTINUE: pid %d", p->p_pid); break; } break; case PT_DETACH: /* reset process parent */ if (p->p_oppid != p->p_pptr->p_pid) { PROC_LOCK(p->p_pptr); sigqueue_take(p->p_ksi); PROC_UNLOCK(p->p_pptr); pp = proc_realparent(p); proc_reparent(p, pp); if (pp == initproc) p->p_sigparent = SIGCHLD; CTR2(KTR_PTRACE, "PT_DETACH: pid %d reparented to pid %d", p->p_pid, pp->p_pid); } else CTR1(KTR_PTRACE, "PT_DETACH: pid %d", p->p_pid); p->p_oppid = 0; p->p_flag &= ~(P_TRACED | P_WAITED | P_FOLLOWFORK); p->p_stops = 0; /* should we send SIGCHLD? */ /* childproc_continued(p); */ break; } sendsig: if (proctree_locked) { sx_xunlock(&proctree_lock); proctree_locked = 0; } p->p_xstat = data; p->p_xthread = NULL; if ((p->p_flag & (P_STOPPED_SIG | P_STOPPED_TRACE)) != 0) { /* deliver or queue signal */ td2->td_dbgflags &= ~TDB_XSIG; td2->td_xsig = data; if (req == PT_DETACH) { FOREACH_THREAD_IN_PROC(p, td3) td3->td_dbgflags &= ~TDB_SUSPEND; } /* * unsuspend all threads, to not let a thread run, * you should use PT_SUSPEND to suspend it before * continuing process. */ PROC_SLOCK(p); p->p_flag &= ~(P_STOPPED_TRACE|P_STOPPED_SIG|P_WAITED); thread_unsuspend(p); PROC_SUNLOCK(p); if (req == PT_ATTACH) kern_psignal(p, data); } else { if (data) kern_psignal(p, data); } break; case PT_WRITE_I: case PT_WRITE_D: td2->td_dbgflags |= TDB_USERWR; write = 1; /* FALLTHROUGH */ case PT_READ_I: case PT_READ_D: PROC_UNLOCK(p); tmp = 0; /* write = 0 set above */ iov.iov_base = write ? (caddr_t)&data : (caddr_t)&tmp; iov.iov_len = sizeof(int); uio.uio_iov = &iov; uio.uio_iovcnt = 1; uio.uio_offset = (off_t)(uintptr_t)addr; uio.uio_resid = sizeof(int); uio.uio_segflg = UIO_SYSSPACE; /* i.e.: the uap */ uio.uio_rw = write ? UIO_WRITE : UIO_READ; uio.uio_td = td; error = proc_rwmem(p, &uio); if (uio.uio_resid != 0) { /* * XXX proc_rwmem() doesn't currently return ENOSPC, * so I think write() can bogusly return 0. * XXX what happens for short writes? We don't want * to write partial data. * XXX proc_rwmem() returns EPERM for other invalid * addresses. Convert this to EINVAL. Does this * clobber returns of EPERM for other reasons? */ if (error == 0 || error == ENOSPC || error == EPERM) error = EINVAL; /* EOF */ } if (!write) td->td_retval[0] = tmp; if (error == 0) { if (write) CTR3(KTR_PTRACE, "PT_WRITE: pid %d: %p <= %#x", p->p_pid, addr, data); else CTR3(KTR_PTRACE, "PT_READ: pid %d: %p >= %#x", p->p_pid, addr, tmp); } PROC_LOCK(p); break; case PT_IO: #ifdef COMPAT_FREEBSD32 if (wrap32) { piod32 = addr; iov.iov_base = (void *)(uintptr_t)piod32->piod_addr; iov.iov_len = piod32->piod_len; uio.uio_offset = (off_t)(uintptr_t)piod32->piod_offs; uio.uio_resid = piod32->piod_len; } else #endif { piod = addr; iov.iov_base = piod->piod_addr; iov.iov_len = piod->piod_len; uio.uio_offset = (off_t)(uintptr_t)piod->piod_offs; uio.uio_resid = piod->piod_len; } uio.uio_iov = &iov; uio.uio_iovcnt = 1; uio.uio_segflg = UIO_USERSPACE; uio.uio_td = td; #ifdef COMPAT_FREEBSD32 tmp = wrap32 ? piod32->piod_op : piod->piod_op; #else tmp = piod->piod_op; #endif switch (tmp) { case PIOD_READ_D: case PIOD_READ_I: CTR3(KTR_PTRACE, "PT_IO: pid %d: READ (%p, %#x)", p->p_pid, (uintptr_t)uio.uio_offset, uio.uio_resid); uio.uio_rw = UIO_READ; break; case PIOD_WRITE_D: case PIOD_WRITE_I: CTR3(KTR_PTRACE, "PT_IO: pid %d: WRITE (%p, %#x)", p->p_pid, (uintptr_t)uio.uio_offset, uio.uio_resid); td2->td_dbgflags |= TDB_USERWR; uio.uio_rw = UIO_WRITE; break; default: error = EINVAL; goto out; } PROC_UNLOCK(p); error = proc_rwmem(p, &uio); #ifdef COMPAT_FREEBSD32 if (wrap32) piod32->piod_len -= uio.uio_resid; else #endif piod->piod_len -= uio.uio_resid; PROC_LOCK(p); break; case PT_KILL: CTR1(KTR_PTRACE, "PT_KILL: pid %d", p->p_pid); data = SIGKILL; goto sendsig; /* in PT_CONTINUE above */ case PT_SETREGS: CTR2(KTR_PTRACE, "PT_SETREGS: tid %d (pid %d)", td2->td_tid, p->p_pid); td2->td_dbgflags |= TDB_USERWR; error = PROC_WRITE(regs, td2, addr); break; case PT_GETREGS: CTR2(KTR_PTRACE, "PT_GETREGS: tid %d (pid %d)", td2->td_tid, p->p_pid); error = PROC_READ(regs, td2, addr); break; case PT_SETFPREGS: CTR2(KTR_PTRACE, "PT_SETFPREGS: tid %d (pid %d)", td2->td_tid, p->p_pid); td2->td_dbgflags |= TDB_USERWR; error = PROC_WRITE(fpregs, td2, addr); break; case PT_GETFPREGS: CTR2(KTR_PTRACE, "PT_GETFPREGS: tid %d (pid %d)", td2->td_tid, p->p_pid); error = PROC_READ(fpregs, td2, addr); break; case PT_SETDBREGS: CTR2(KTR_PTRACE, "PT_SETDBREGS: tid %d (pid %d)", td2->td_tid, p->p_pid); td2->td_dbgflags |= TDB_USERWR; error = PROC_WRITE(dbregs, td2, addr); break; case PT_GETDBREGS: CTR2(KTR_PTRACE, "PT_GETDBREGS: tid %d (pid %d)", td2->td_tid, p->p_pid); error = PROC_READ(dbregs, td2, addr); break; case PT_LWPINFO: if (data <= 0 || #ifdef COMPAT_FREEBSD32 (!wrap32 && data > sizeof(*pl)) || (wrap32 && data > sizeof(*pl32))) { #else data > sizeof(*pl)) { #endif error = EINVAL; break; } #ifdef COMPAT_FREEBSD32 if (wrap32) { pl = &plr; pl32 = addr; } else #endif pl = addr; pl->pl_lwpid = td2->td_tid; pl->pl_event = PL_EVENT_NONE; pl->pl_flags = 0; if (td2->td_dbgflags & TDB_XSIG) { pl->pl_event = PL_EVENT_SIGNAL; if (td2->td_dbgksi.ksi_signo != 0 && #ifdef COMPAT_FREEBSD32 ((!wrap32 && data >= offsetof(struct ptrace_lwpinfo, pl_siginfo) + sizeof(pl->pl_siginfo)) || (wrap32 && data >= offsetof(struct ptrace_lwpinfo32, pl_siginfo) + sizeof(struct siginfo32))) #else data >= offsetof(struct ptrace_lwpinfo, pl_siginfo) + sizeof(pl->pl_siginfo) #endif ){ pl->pl_flags |= PL_FLAG_SI; pl->pl_siginfo = td2->td_dbgksi.ksi_info; } } if ((pl->pl_flags & PL_FLAG_SI) == 0) bzero(&pl->pl_siginfo, sizeof(pl->pl_siginfo)); if (td2->td_dbgflags & TDB_SCE) pl->pl_flags |= PL_FLAG_SCE; else if (td2->td_dbgflags & TDB_SCX) pl->pl_flags |= PL_FLAG_SCX; if (td2->td_dbgflags & TDB_EXEC) pl->pl_flags |= PL_FLAG_EXEC; if (td2->td_dbgflags & TDB_FORK) { pl->pl_flags |= PL_FLAG_FORKED; pl->pl_child_pid = td2->td_dbg_forked; } if (td2->td_dbgflags & TDB_CHILD) pl->pl_flags |= PL_FLAG_CHILD; pl->pl_sigmask = td2->td_sigmask; pl->pl_siglist = td2->td_siglist; strcpy(pl->pl_tdname, td2->td_name); #ifdef COMPAT_FREEBSD32 if (wrap32) ptrace_lwpinfo_to32(pl, pl32); #endif CTR5(KTR_PTRACE, "PT_LWPINFO: tid %d (pid %d) event %d flags %#x child pid %d", td2->td_tid, p->p_pid, pl->pl_event, pl->pl_flags, pl->pl_child_pid); break; case PT_GETNUMLWPS: CTR2(KTR_PTRACE, "PT_GETNUMLWPS: pid %d: %d threads", p->p_pid, p->p_numthreads); td->td_retval[0] = p->p_numthreads; break; case PT_GETLWPLIST: CTR3(KTR_PTRACE, "PT_GETLWPLIST: pid %d: data %d, actual %d", p->p_pid, data, p->p_numthreads); if (data <= 0) { error = EINVAL; break; } num = imin(p->p_numthreads, data); PROC_UNLOCK(p); buf = malloc(num * sizeof(lwpid_t), M_TEMP, M_WAITOK); tmp = 0; PROC_LOCK(p); FOREACH_THREAD_IN_PROC(p, td2) { if (tmp >= num) break; buf[tmp++] = td2->td_tid; } PROC_UNLOCK(p); error = copyout(buf, addr, tmp * sizeof(lwpid_t)); free(buf, M_TEMP); if (!error) td->td_retval[0] = tmp; PROC_LOCK(p); break; case PT_VM_TIMESTAMP: CTR2(KTR_PTRACE, "PT_VM_TIMESTAMP: pid %d: timestamp %d", p->p_pid, p->p_vmspace->vm_map.timestamp); td->td_retval[0] = p->p_vmspace->vm_map.timestamp; break; case PT_VM_ENTRY: PROC_UNLOCK(p); #ifdef COMPAT_FREEBSD32 if (wrap32) error = ptrace_vm_entry32(td, p, addr); else #endif error = ptrace_vm_entry(td, p, addr); PROC_LOCK(p); break; default: #ifdef __HAVE_PTRACE_MACHDEP if (req >= PT_FIRSTMACH) { PROC_UNLOCK(p); error = cpu_ptrace(td2, req, addr, data); PROC_LOCK(p); } else #endif /* Unknown request. */ error = EINVAL; break; } out: /* Drop our hold on this process now that the request has completed. */ _PRELE(p); fail: PROC_UNLOCK(p); if (proctree_locked) sx_xunlock(&proctree_lock); return (error); } #undef PROC_READ #undef PROC_WRITE /* * Stop a process because of a debugging event; * stay stopped until p->p_step is cleared * (cleared by PIOCCONT in procfs). */ void stopevent(struct proc *p, unsigned int event, unsigned int val) { PROC_LOCK_ASSERT(p, MA_OWNED); p->p_step = 1; CTR3(KTR_PTRACE, "stopevent: pid %d event %u val %u", p->p_pid, event, val); do { p->p_xstat = val; p->p_xthread = NULL; p->p_stype = event; /* Which event caused the stop? */ wakeup(&p->p_stype); /* Wake up any PIOCWAIT'ing procs */ msleep(&p->p_step, &p->p_mtx, PWAIT, "stopevent", 0); } while (p->p_step); } Index: projects/em_mq/sys/vm/vm_object.c =================================================================== --- projects/em_mq/sys/vm/vm_object.c (revision 283943) +++ projects/em_mq/sys/vm/vm_object.c (revision 283944) @@ -1,2659 +1,2671 @@ /*- * Copyright (c) 1991, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * The Mach Operating System project at Carnegie-Mellon University. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: @(#)vm_object.c 8.5 (Berkeley) 3/22/94 * * * Copyright (c) 1987, 1990 Carnegie-Mellon University. * All rights reserved. * * Authors: Avadis Tevanian, Jr., Michael Wayne Young * * Permission to use, copy, modify and distribute this software and * its documentation is hereby granted, provided that both the copyright * notice and this permission notice appear in all copies of the * software, derivative works or modified versions, and any portions * thereof, and that both notices appear in supporting documentation. * * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. * * Carnegie Mellon requests users of this software to return to * * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU * School of Computer Science * Carnegie Mellon University * Pittsburgh PA 15213-3890 * * any improvements or extensions that they make and grant Carnegie the * rights to redistribute these changes. */ /* * Virtual memory object module. */ #include __FBSDID("$FreeBSD$"); #include "opt_vm.h" #include #include #include #include #include #include #include #include #include /* for curproc, pageproc */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static int old_msync; SYSCTL_INT(_vm, OID_AUTO, old_msync, CTLFLAG_RW, &old_msync, 0, "Use old (insecure) msync behavior"); static int vm_object_page_collect_flush(vm_object_t object, vm_page_t p, int pagerflags, int flags, boolean_t *clearobjflags, boolean_t *eio); static boolean_t vm_object_page_remove_write(vm_page_t p, int flags, boolean_t *clearobjflags); static void vm_object_qcollapse(vm_object_t object); static void vm_object_vndeallocate(vm_object_t object); /* * Virtual memory objects maintain the actual data * associated with allocated virtual memory. A given * page of memory exists within exactly one object. * * An object is only deallocated when all "references" * are given up. Only one "reference" to a given * region of an object should be writeable. * * Associated with each object is a list of all resident * memory pages belonging to that object; this list is * maintained by the "vm_page" module, and locked by the object's * lock. * * Each object also records a "pager" routine which is * used to retrieve (and store) pages to the proper backing * storage. In addition, objects may be backed by other * objects from which they were virtual-copied. * * The only items within the object structure which are * modified after time of creation are: * reference count locked by object's lock * pager routine locked by object's lock * */ struct object_q vm_object_list; struct mtx vm_object_list_mtx; /* lock for object list and count */ struct vm_object kernel_object_store; struct vm_object kmem_object_store; static SYSCTL_NODE(_vm_stats, OID_AUTO, object, CTLFLAG_RD, 0, "VM object stats"); static long object_collapses; SYSCTL_LONG(_vm_stats_object, OID_AUTO, collapses, CTLFLAG_RD, &object_collapses, 0, "VM object collapses"); static long object_bypasses; SYSCTL_LONG(_vm_stats_object, OID_AUTO, bypasses, CTLFLAG_RD, &object_bypasses, 0, "VM object bypasses"); static uma_zone_t obj_zone; static int vm_object_zinit(void *mem, int size, int flags); #ifdef INVARIANTS static void vm_object_zdtor(void *mem, int size, void *arg); static void vm_object_zdtor(void *mem, int size, void *arg) { vm_object_t object; object = (vm_object_t)mem; KASSERT(object->ref_count == 0, ("object %p ref_count = %d", object, object->ref_count)); KASSERT(TAILQ_EMPTY(&object->memq), ("object %p has resident pages in its memq", object)); KASSERT(vm_radix_is_empty(&object->rtree), ("object %p has resident pages in its trie", object)); #if VM_NRESERVLEVEL > 0 KASSERT(LIST_EMPTY(&object->rvq), ("object %p has reservations", object)); #endif KASSERT(vm_object_cache_is_empty(object), ("object %p has cached pages", object)); KASSERT(object->paging_in_progress == 0, ("object %p paging_in_progress = %d", object, object->paging_in_progress)); KASSERT(object->resident_page_count == 0, ("object %p resident_page_count = %d", object, object->resident_page_count)); KASSERT(object->shadow_count == 0, ("object %p shadow_count = %d", object, object->shadow_count)); KASSERT(object->type == OBJT_DEAD, ("object %p has non-dead type %d", object, object->type)); } #endif static int vm_object_zinit(void *mem, int size, int flags) { vm_object_t object; object = (vm_object_t)mem; rw_init_flags(&object->lock, "vm object", RW_DUPOK | RW_NEW); /* These are true for any object that has been freed */ object->type = OBJT_DEAD; object->ref_count = 0; object->rtree.rt_root = 0; object->rtree.rt_flags = 0; object->paging_in_progress = 0; object->resident_page_count = 0; object->shadow_count = 0; object->cache.rt_root = 0; object->cache.rt_flags = 0; mtx_lock(&vm_object_list_mtx); TAILQ_INSERT_TAIL(&vm_object_list, object, object_list); mtx_unlock(&vm_object_list_mtx); return (0); } static void _vm_object_allocate(objtype_t type, vm_pindex_t size, vm_object_t object) { TAILQ_INIT(&object->memq); LIST_INIT(&object->shadow_head); object->type = type; switch (type) { case OBJT_DEAD: panic("_vm_object_allocate: can't create OBJT_DEAD"); case OBJT_DEFAULT: case OBJT_SWAP: object->flags = OBJ_ONEMAPPING; break; case OBJT_DEVICE: case OBJT_SG: object->flags = OBJ_FICTITIOUS | OBJ_UNMANAGED; break; case OBJT_MGTDEVICE: object->flags = OBJ_FICTITIOUS; break; case OBJT_PHYS: object->flags = OBJ_UNMANAGED; break; case OBJT_VNODE: object->flags = 0; break; default: panic("_vm_object_allocate: type %d is undefined", type); } object->size = size; object->generation = 1; object->ref_count = 1; object->memattr = VM_MEMATTR_DEFAULT; object->cred = NULL; object->charge = 0; object->handle = NULL; object->backing_object = NULL; object->backing_object_offset = (vm_ooffset_t) 0; #if VM_NRESERVLEVEL > 0 LIST_INIT(&object->rvq); #endif } /* * vm_object_init: * * Initialize the VM objects module. */ void vm_object_init(void) { TAILQ_INIT(&vm_object_list); mtx_init(&vm_object_list_mtx, "vm object_list", NULL, MTX_DEF); rw_init(&kernel_object->lock, "kernel vm object"); _vm_object_allocate(OBJT_PHYS, OFF_TO_IDX(VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS), kernel_object); #if VM_NRESERVLEVEL > 0 kernel_object->flags |= OBJ_COLORED; kernel_object->pg_color = (u_short)atop(VM_MIN_KERNEL_ADDRESS); #endif rw_init(&kmem_object->lock, "kmem vm object"); _vm_object_allocate(OBJT_PHYS, OFF_TO_IDX(VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS), kmem_object); #if VM_NRESERVLEVEL > 0 kmem_object->flags |= OBJ_COLORED; kmem_object->pg_color = (u_short)atop(VM_MIN_KERNEL_ADDRESS); #endif /* * The lock portion of struct vm_object must be type stable due * to vm_pageout_fallback_object_lock locking a vm object * without holding any references to it. */ obj_zone = uma_zcreate("VM OBJECT", sizeof (struct vm_object), NULL, #ifdef INVARIANTS vm_object_zdtor, #else NULL, #endif vm_object_zinit, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); vm_radix_init(); } void vm_object_clear_flag(vm_object_t object, u_short bits) { VM_OBJECT_ASSERT_WLOCKED(object); object->flags &= ~bits; } /* * Sets the default memory attribute for the specified object. Pages * that are allocated to this object are by default assigned this memory * attribute. * * Presently, this function must be called before any pages are allocated * to the object. In the future, this requirement may be relaxed for * "default" and "swap" objects. */ int vm_object_set_memattr(vm_object_t object, vm_memattr_t memattr) { VM_OBJECT_ASSERT_WLOCKED(object); switch (object->type) { case OBJT_DEFAULT: case OBJT_DEVICE: case OBJT_MGTDEVICE: case OBJT_PHYS: case OBJT_SG: case OBJT_SWAP: case OBJT_VNODE: if (!TAILQ_EMPTY(&object->memq)) return (KERN_FAILURE); break; case OBJT_DEAD: return (KERN_INVALID_ARGUMENT); default: panic("vm_object_set_memattr: object %p is of undefined type", object); } object->memattr = memattr; return (KERN_SUCCESS); } void vm_object_pip_add(vm_object_t object, short i) { VM_OBJECT_ASSERT_WLOCKED(object); object->paging_in_progress += i; } void vm_object_pip_subtract(vm_object_t object, short i) { VM_OBJECT_ASSERT_WLOCKED(object); object->paging_in_progress -= i; } void vm_object_pip_wakeup(vm_object_t object) { VM_OBJECT_ASSERT_WLOCKED(object); object->paging_in_progress--; if ((object->flags & OBJ_PIPWNT) && object->paging_in_progress == 0) { vm_object_clear_flag(object, OBJ_PIPWNT); wakeup(object); } } void vm_object_pip_wakeupn(vm_object_t object, short i) { VM_OBJECT_ASSERT_WLOCKED(object); if (i) object->paging_in_progress -= i; if ((object->flags & OBJ_PIPWNT) && object->paging_in_progress == 0) { vm_object_clear_flag(object, OBJ_PIPWNT); wakeup(object); } } void vm_object_pip_wait(vm_object_t object, char *waitid) { VM_OBJECT_ASSERT_WLOCKED(object); while (object->paging_in_progress) { object->flags |= OBJ_PIPWNT; VM_OBJECT_SLEEP(object, object, PVM, waitid, 0); } } /* * vm_object_allocate: * * Returns a new object with the given size. */ vm_object_t vm_object_allocate(objtype_t type, vm_pindex_t size) { vm_object_t object; object = (vm_object_t)uma_zalloc(obj_zone, M_WAITOK); _vm_object_allocate(type, size, object); return (object); } /* * vm_object_reference: * * Gets another reference to the given object. Note: OBJ_DEAD * objects can be referenced during final cleaning. */ void vm_object_reference(vm_object_t object) { if (object == NULL) return; VM_OBJECT_WLOCK(object); vm_object_reference_locked(object); VM_OBJECT_WUNLOCK(object); } /* * vm_object_reference_locked: * * Gets another reference to the given object. * * The object must be locked. */ void vm_object_reference_locked(vm_object_t object) { struct vnode *vp; VM_OBJECT_ASSERT_WLOCKED(object); object->ref_count++; if (object->type == OBJT_VNODE) { vp = object->handle; vref(vp); } } /* * Handle deallocating an object of type OBJT_VNODE. */ static void vm_object_vndeallocate(vm_object_t object) { struct vnode *vp = (struct vnode *) object->handle; VM_OBJECT_ASSERT_WLOCKED(object); KASSERT(object->type == OBJT_VNODE, ("vm_object_vndeallocate: not a vnode object")); KASSERT(vp != NULL, ("vm_object_vndeallocate: missing vp")); #ifdef INVARIANTS if (object->ref_count == 0) { vprint("vm_object_vndeallocate", vp); panic("vm_object_vndeallocate: bad object reference count"); } #endif /* * The test for text of vp vnode does not need a bypass to * reach right VV_TEXT there, since it is obtained from * object->handle. */ if (object->ref_count > 1 || (vp->v_vflag & VV_TEXT) == 0) { object->ref_count--; VM_OBJECT_WUNLOCK(object); /* vrele may need the vnode lock. */ vrele(vp); } else { vhold(vp); VM_OBJECT_WUNLOCK(object); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); vdrop(vp); VM_OBJECT_WLOCK(object); object->ref_count--; if (object->type == OBJT_DEAD) { VM_OBJECT_WUNLOCK(object); VOP_UNLOCK(vp, 0); } else { if (object->ref_count == 0) VOP_UNSET_TEXT(vp); VM_OBJECT_WUNLOCK(object); vput(vp); } } } /* * vm_object_deallocate: * * Release a reference to the specified object, * gained either through a vm_object_allocate * or a vm_object_reference call. When all references * are gone, storage associated with this object * may be relinquished. * * No object may be locked. */ void vm_object_deallocate(vm_object_t object) { vm_object_t temp; struct vnode *vp; while (object != NULL) { VM_OBJECT_WLOCK(object); if (object->type == OBJT_VNODE) { vm_object_vndeallocate(object); return; } KASSERT(object->ref_count != 0, ("vm_object_deallocate: object deallocated too many times: %d", object->type)); /* * If the reference count goes to 0 we start calling * vm_object_terminate() on the object chain. * A ref count of 1 may be a special case depending on the * shadow count being 0 or 1. */ object->ref_count--; if (object->ref_count > 1) { VM_OBJECT_WUNLOCK(object); return; } else if (object->ref_count == 1) { if (object->type == OBJT_SWAP && (object->flags & OBJ_TMPFS) != 0) { vp = object->un_pager.swp.swp_tmpfs; vhold(vp); VM_OBJECT_WUNLOCK(object); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); VM_OBJECT_WLOCK(object); if (object->type == OBJT_DEAD || object->ref_count != 1) { VM_OBJECT_WUNLOCK(object); VOP_UNLOCK(vp, 0); vdrop(vp); return; } if ((object->flags & OBJ_TMPFS) != 0) VOP_UNSET_TEXT(vp); VOP_UNLOCK(vp, 0); vdrop(vp); } if (object->shadow_count == 0 && object->handle == NULL && (object->type == OBJT_DEFAULT || (object->type == OBJT_SWAP && (object->flags & OBJ_TMPFS_NODE) == 0))) { vm_object_set_flag(object, OBJ_ONEMAPPING); } else if ((object->shadow_count == 1) && (object->handle == NULL) && (object->type == OBJT_DEFAULT || object->type == OBJT_SWAP)) { vm_object_t robject; robject = LIST_FIRST(&object->shadow_head); KASSERT(robject != NULL, ("vm_object_deallocate: ref_count: %d, shadow_count: %d", object->ref_count, object->shadow_count)); KASSERT((robject->flags & OBJ_TMPFS_NODE) == 0, ("shadowed tmpfs v_object %p", object)); if (!VM_OBJECT_TRYWLOCK(robject)) { /* * Avoid a potential deadlock. */ object->ref_count++; VM_OBJECT_WUNLOCK(object); /* * More likely than not the thread * holding robject's lock has lower * priority than the current thread. * Let the lower priority thread run. */ pause("vmo_de", 1); continue; } /* * Collapse object into its shadow unless its * shadow is dead. In that case, object will * be deallocated by the thread that is * deallocating its shadow. */ if ((robject->flags & OBJ_DEAD) == 0 && (robject->handle == NULL) && (robject->type == OBJT_DEFAULT || robject->type == OBJT_SWAP)) { robject->ref_count++; retry: if (robject->paging_in_progress) { VM_OBJECT_WUNLOCK(object); vm_object_pip_wait(robject, "objde1"); temp = robject->backing_object; if (object == temp) { VM_OBJECT_WLOCK(object); goto retry; } } else if (object->paging_in_progress) { VM_OBJECT_WUNLOCK(robject); object->flags |= OBJ_PIPWNT; VM_OBJECT_SLEEP(object, object, PDROP | PVM, "objde2", 0); VM_OBJECT_WLOCK(robject); temp = robject->backing_object; if (object == temp) { VM_OBJECT_WLOCK(object); goto retry; } } else VM_OBJECT_WUNLOCK(object); if (robject->ref_count == 1) { robject->ref_count--; object = robject; goto doterm; } object = robject; vm_object_collapse(object); VM_OBJECT_WUNLOCK(object); continue; } VM_OBJECT_WUNLOCK(robject); } VM_OBJECT_WUNLOCK(object); return; } doterm: temp = object->backing_object; if (temp != NULL) { KASSERT((object->flags & OBJ_TMPFS_NODE) == 0, ("shadowed tmpfs v_object 2 %p", object)); VM_OBJECT_WLOCK(temp); LIST_REMOVE(object, shadow_list); temp->shadow_count--; VM_OBJECT_WUNLOCK(temp); object->backing_object = NULL; } /* * Don't double-terminate, we could be in a termination * recursion due to the terminate having to sync data * to disk. */ if ((object->flags & OBJ_DEAD) == 0) vm_object_terminate(object); else VM_OBJECT_WUNLOCK(object); object = temp; } } /* * vm_object_destroy removes the object from the global object list * and frees the space for the object. */ void vm_object_destroy(vm_object_t object) { /* * Release the allocation charge. */ if (object->cred != NULL) { swap_release_by_cred(object->charge, object->cred); object->charge = 0; crfree(object->cred); object->cred = NULL; } /* * Free the space for the object. */ uma_zfree(obj_zone, object); } /* * vm_object_terminate actually destroys the specified object, freeing * up all previously used resources. * * The object must be locked. * This routine may block. */ void vm_object_terminate(vm_object_t object) { vm_page_t p, p_next; VM_OBJECT_ASSERT_WLOCKED(object); /* * Make sure no one uses us. */ vm_object_set_flag(object, OBJ_DEAD); /* * wait for the pageout daemon to be done with the object */ vm_object_pip_wait(object, "objtrm"); KASSERT(!object->paging_in_progress, ("vm_object_terminate: pageout in progress")); /* * Clean and free the pages, as appropriate. All references to the * object are gone, so we don't need to lock it. */ if (object->type == OBJT_VNODE) { struct vnode *vp = (struct vnode *)object->handle; /* * Clean pages and flush buffers. */ vm_object_page_clean(object, 0, 0, OBJPC_SYNC); VM_OBJECT_WUNLOCK(object); vinvalbuf(vp, V_SAVE, 0, 0); VM_OBJECT_WLOCK(object); } KASSERT(object->ref_count == 0, ("vm_object_terminate: object with references, ref_count=%d", object->ref_count)); /* * Free any remaining pageable pages. This also removes them from the * paging queues. However, don't free wired pages, just remove them * from the object. Rather than incrementally removing each page from * the object, the page and object are reset to any empty state. */ TAILQ_FOREACH_SAFE(p, &object->memq, listq, p_next) { vm_page_assert_unbusied(p); vm_page_lock(p); /* * Optimize the page's removal from the object by resetting * its "object" field. Specifically, if the page is not * wired, then the effect of this assignment is that * vm_page_free()'s call to vm_page_remove() will return * immediately without modifying the page or the object. */ p->object = NULL; if (p->wire_count == 0) { vm_page_free(p); PCPU_INC(cnt.v_pfree); } vm_page_unlock(p); } /* * If the object contained any pages, then reset it to an empty state. * None of the object's fields, including "resident_page_count", were * modified by the preceding loop. */ if (object->resident_page_count != 0) { vm_radix_reclaim_allnodes(&object->rtree); TAILQ_INIT(&object->memq); object->resident_page_count = 0; if (object->type == OBJT_VNODE) vdrop(object->handle); } #if VM_NRESERVLEVEL > 0 if (__predict_false(!LIST_EMPTY(&object->rvq))) vm_reserv_break_all(object); #endif if (__predict_false(!vm_object_cache_is_empty(object))) vm_page_cache_free(object, 0, 0); KASSERT(object->cred == NULL || object->type == OBJT_DEFAULT || object->type == OBJT_SWAP, ("%s: non-swap obj %p has cred", __func__, object)); /* * Let the pager know object is dead. */ vm_pager_deallocate(object); VM_OBJECT_WUNLOCK(object); vm_object_destroy(object); } /* * Make the page read-only so that we can clear the object flags. However, if * this is a nosync mmap then the object is likely to stay dirty so do not * mess with the page and do not clear the object flags. Returns TRUE if the * page should be flushed, and FALSE otherwise. */ static boolean_t vm_object_page_remove_write(vm_page_t p, int flags, boolean_t *clearobjflags) { /* * If we have been asked to skip nosync pages and this is a * nosync page, skip it. Note that the object flags were not * cleared in this case so we do not have to set them. */ if ((flags & OBJPC_NOSYNC) != 0 && (p->oflags & VPO_NOSYNC) != 0) { *clearobjflags = FALSE; return (FALSE); } else { pmap_remove_write(p); return (p->dirty != 0); } } /* * vm_object_page_clean * * Clean all dirty pages in the specified range of object. Leaves page * on whatever queue it is currently on. If NOSYNC is set then do not * write out pages with VPO_NOSYNC set (originally comes from MAP_NOSYNC), * leaving the object dirty. * * When stuffing pages asynchronously, allow clustering. XXX we need a * synchronous clustering mode implementation. * * Odd semantics: if start == end, we clean everything. * * The object must be locked. * * Returns FALSE if some page from the range was not written, as * reported by the pager, and TRUE otherwise. */ boolean_t vm_object_page_clean(vm_object_t object, vm_ooffset_t start, vm_ooffset_t end, int flags) { vm_page_t np, p; vm_pindex_t pi, tend, tstart; int curgeneration, n, pagerflags; boolean_t clearobjflags, eio, res; VM_OBJECT_ASSERT_WLOCKED(object); /* * The OBJ_MIGHTBEDIRTY flag is only set for OBJT_VNODE * objects. The check below prevents the function from * operating on non-vnode objects. */ if ((object->flags & OBJ_MIGHTBEDIRTY) == 0 || object->resident_page_count == 0) return (TRUE); pagerflags = (flags & (OBJPC_SYNC | OBJPC_INVAL)) != 0 ? VM_PAGER_PUT_SYNC : VM_PAGER_CLUSTER_OK; pagerflags |= (flags & OBJPC_INVAL) != 0 ? VM_PAGER_PUT_INVAL : 0; tstart = OFF_TO_IDX(start); tend = (end == 0) ? object->size : OFF_TO_IDX(end + PAGE_MASK); clearobjflags = tstart == 0 && tend >= object->size; res = TRUE; rescan: curgeneration = object->generation; for (p = vm_page_find_least(object, tstart); p != NULL; p = np) { pi = p->pindex; if (pi >= tend) break; np = TAILQ_NEXT(p, listq); if (p->valid == 0) continue; if (vm_page_sleep_if_busy(p, "vpcwai")) { if (object->generation != curgeneration) { if ((flags & OBJPC_SYNC) != 0) goto rescan; else clearobjflags = FALSE; } np = vm_page_find_least(object, pi); continue; } if (!vm_object_page_remove_write(p, flags, &clearobjflags)) continue; n = vm_object_page_collect_flush(object, p, pagerflags, flags, &clearobjflags, &eio); if (eio) { res = FALSE; clearobjflags = FALSE; } if (object->generation != curgeneration) { if ((flags & OBJPC_SYNC) != 0) goto rescan; else clearobjflags = FALSE; } /* * If the VOP_PUTPAGES() did a truncated write, so * that even the first page of the run is not fully * written, vm_pageout_flush() returns 0 as the run * length. Since the condition that caused truncated * write may be permanent, e.g. exhausted free space, * accepting n == 0 would cause an infinite loop. * * Forwarding the iterator leaves the unwritten page * behind, but there is not much we can do there if * filesystem refuses to write it. */ if (n == 0) { n = 1; clearobjflags = FALSE; } np = vm_page_find_least(object, pi + n); } #if 0 VOP_FSYNC(vp, (pagerflags & VM_PAGER_PUT_SYNC) ? MNT_WAIT : 0); #endif if (clearobjflags) vm_object_clear_flag(object, OBJ_MIGHTBEDIRTY); return (res); } static int vm_object_page_collect_flush(vm_object_t object, vm_page_t p, int pagerflags, int flags, boolean_t *clearobjflags, boolean_t *eio) { vm_page_t ma[vm_pageout_page_count], p_first, tp; int count, i, mreq, runlen; vm_page_lock_assert(p, MA_NOTOWNED); VM_OBJECT_ASSERT_WLOCKED(object); count = 1; mreq = 0; for (tp = p; count < vm_pageout_page_count; count++) { tp = vm_page_next(tp); if (tp == NULL || vm_page_busied(tp)) break; if (!vm_object_page_remove_write(tp, flags, clearobjflags)) break; } for (p_first = p; count < vm_pageout_page_count; count++) { tp = vm_page_prev(p_first); if (tp == NULL || vm_page_busied(tp)) break; if (!vm_object_page_remove_write(tp, flags, clearobjflags)) break; p_first = tp; mreq++; } for (tp = p_first, i = 0; i < count; tp = TAILQ_NEXT(tp, listq), i++) ma[i] = tp; vm_pageout_flush(ma, count, pagerflags, mreq, &runlen, eio); return (runlen); } /* * Note that there is absolutely no sense in writing out * anonymous objects, so we track down the vnode object * to write out. * We invalidate (remove) all pages from the address space * for semantic correctness. * * If the backing object is a device object with unmanaged pages, then any * mappings to the specified range of pages must be removed before this * function is called. * * Note: certain anonymous maps, such as MAP_NOSYNC maps, * may start out with a NULL object. */ boolean_t vm_object_sync(vm_object_t object, vm_ooffset_t offset, vm_size_t size, boolean_t syncio, boolean_t invalidate) { vm_object_t backing_object; struct vnode *vp; struct mount *mp; int error, flags, fsync_after; boolean_t res; if (object == NULL) return (TRUE); res = TRUE; error = 0; VM_OBJECT_WLOCK(object); while ((backing_object = object->backing_object) != NULL) { VM_OBJECT_WLOCK(backing_object); offset += object->backing_object_offset; VM_OBJECT_WUNLOCK(object); object = backing_object; if (object->size < OFF_TO_IDX(offset + size)) size = IDX_TO_OFF(object->size) - offset; } /* * Flush pages if writing is allowed, invalidate them * if invalidation requested. Pages undergoing I/O * will be ignored by vm_object_page_remove(). * * We cannot lock the vnode and then wait for paging * to complete without deadlocking against vm_fault. * Instead we simply call vm_object_page_remove() and * allow it to block internally on a page-by-page * basis when it encounters pages undergoing async * I/O. */ if (object->type == OBJT_VNODE && (object->flags & OBJ_MIGHTBEDIRTY) != 0) { vp = object->handle; VM_OBJECT_WUNLOCK(object); (void) vn_start_write(vp, &mp, V_WAIT); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); if (syncio && !invalidate && offset == 0 && OFF_TO_IDX(size) == object->size) { /* * If syncing the whole mapping of the file, * it is faster to schedule all the writes in * async mode, also allowing the clustering, * and then wait for i/o to complete. */ flags = 0; fsync_after = TRUE; } else { flags = (syncio || invalidate) ? OBJPC_SYNC : 0; flags |= invalidate ? (OBJPC_SYNC | OBJPC_INVAL) : 0; fsync_after = FALSE; } VM_OBJECT_WLOCK(object); res = vm_object_page_clean(object, offset, offset + size, flags); VM_OBJECT_WUNLOCK(object); if (fsync_after) error = VOP_FSYNC(vp, MNT_WAIT, curthread); VOP_UNLOCK(vp, 0); vn_finished_write(mp); if (error != 0) res = FALSE; VM_OBJECT_WLOCK(object); } if ((object->type == OBJT_VNODE || object->type == OBJT_DEVICE) && invalidate) { if (object->type == OBJT_DEVICE) /* * The option OBJPR_NOTMAPPED must be passed here * because vm_object_page_remove() cannot remove * unmanaged mappings. */ flags = OBJPR_NOTMAPPED; else if (old_msync) flags = OBJPR_NOTWIRED; else flags = OBJPR_CLEANONLY | OBJPR_NOTWIRED; vm_object_page_remove(object, OFF_TO_IDX(offset), OFF_TO_IDX(offset + size + PAGE_MASK), flags); } VM_OBJECT_WUNLOCK(object); return (res); } /* * vm_object_madvise: * * Implements the madvise function at the object/page level. * * MADV_WILLNEED (any object) * * Activate the specified pages if they are resident. * * MADV_DONTNEED (any object) * * Deactivate the specified pages if they are resident. * * MADV_FREE (OBJT_DEFAULT/OBJT_SWAP objects, * OBJ_ONEMAPPING only) * * Deactivate and clean the specified pages if they are * resident. This permits the process to reuse the pages * without faulting or the kernel to reclaim the pages * without I/O. */ void vm_object_madvise(vm_object_t object, vm_pindex_t pindex, vm_pindex_t end, int advise) { vm_pindex_t tpindex; vm_object_t backing_object, tobject; vm_page_t m; if (object == NULL) return; VM_OBJECT_WLOCK(object); /* * Locate and adjust resident pages */ for (; pindex < end; pindex += 1) { relookup: tobject = object; tpindex = pindex; shadowlookup: /* * MADV_FREE only operates on OBJT_DEFAULT or OBJT_SWAP pages * and those pages must be OBJ_ONEMAPPING. */ if (advise == MADV_FREE) { if ((tobject->type != OBJT_DEFAULT && tobject->type != OBJT_SWAP) || (tobject->flags & OBJ_ONEMAPPING) == 0) { goto unlock_tobject; } } else if ((tobject->flags & OBJ_UNMANAGED) != 0) goto unlock_tobject; m = vm_page_lookup(tobject, tpindex); if (m == NULL && advise == MADV_WILLNEED) { /* * If the page is cached, reactivate it. */ m = vm_page_alloc(tobject, tpindex, VM_ALLOC_IFCACHED | VM_ALLOC_NOBUSY); } if (m == NULL) { /* * There may be swap even if there is no backing page */ if (advise == MADV_FREE && tobject->type == OBJT_SWAP) swap_pager_freespace(tobject, tpindex, 1); /* * next object */ backing_object = tobject->backing_object; if (backing_object == NULL) goto unlock_tobject; VM_OBJECT_WLOCK(backing_object); tpindex += OFF_TO_IDX(tobject->backing_object_offset); if (tobject != object) VM_OBJECT_WUNLOCK(tobject); tobject = backing_object; goto shadowlookup; } else if (m->valid != VM_PAGE_BITS_ALL) goto unlock_tobject; /* * If the page is not in a normal state, skip it. */ vm_page_lock(m); if (m->hold_count != 0 || m->wire_count != 0) { vm_page_unlock(m); goto unlock_tobject; } KASSERT((m->flags & PG_FICTITIOUS) == 0, ("vm_object_madvise: page %p is fictitious", m)); KASSERT((m->oflags & VPO_UNMANAGED) == 0, ("vm_object_madvise: page %p is not managed", m)); if (vm_page_busied(m)) { if (advise == MADV_WILLNEED) { /* * Reference the page before unlocking and * sleeping so that the page daemon is less * likely to reclaim it. */ vm_page_aflag_set(m, PGA_REFERENCED); } if (object != tobject) VM_OBJECT_WUNLOCK(object); VM_OBJECT_WUNLOCK(tobject); vm_page_busy_sleep(m, "madvpo"); VM_OBJECT_WLOCK(object); goto relookup; } if (advise == MADV_WILLNEED) { vm_page_activate(m); } else { vm_page_advise(m, advise); } vm_page_unlock(m); if (advise == MADV_FREE && tobject->type == OBJT_SWAP) swap_pager_freespace(tobject, tpindex, 1); unlock_tobject: if (tobject != object) VM_OBJECT_WUNLOCK(tobject); } VM_OBJECT_WUNLOCK(object); } /* * vm_object_shadow: * * Create a new object which is backed by the * specified existing object range. The source * object reference is deallocated. * * The new object and offset into that object * are returned in the source parameters. */ void vm_object_shadow( vm_object_t *object, /* IN/OUT */ vm_ooffset_t *offset, /* IN/OUT */ vm_size_t length) { vm_object_t source; vm_object_t result; source = *object; /* * Don't create the new object if the old object isn't shared. */ if (source != NULL) { VM_OBJECT_WLOCK(source); if (source->ref_count == 1 && source->handle == NULL && (source->type == OBJT_DEFAULT || source->type == OBJT_SWAP)) { VM_OBJECT_WUNLOCK(source); return; } VM_OBJECT_WUNLOCK(source); } /* * Allocate a new object with the given length. */ result = vm_object_allocate(OBJT_DEFAULT, atop(length)); /* * The new object shadows the source object, adding a reference to it. * Our caller changes his reference to point to the new object, * removing a reference to the source object. Net result: no change * of reference count. * * Try to optimize the result object's page color when shadowing * in order to maintain page coloring consistency in the combined * shadowed object. */ result->backing_object = source; /* * Store the offset into the source object, and fix up the offset into * the new object. */ result->backing_object_offset = *offset; if (source != NULL) { VM_OBJECT_WLOCK(source); LIST_INSERT_HEAD(&source->shadow_head, result, shadow_list); source->shadow_count++; #if VM_NRESERVLEVEL > 0 result->flags |= source->flags & OBJ_COLORED; result->pg_color = (source->pg_color + OFF_TO_IDX(*offset)) & ((1 << (VM_NFREEORDER - 1)) - 1); #endif VM_OBJECT_WUNLOCK(source); } /* * Return the new things */ *offset = 0; *object = result; } /* * vm_object_split: * * Split the pages in a map entry into a new object. This affords * easier removal of unused pages, and keeps object inheritance from * being a negative impact on memory usage. */ void vm_object_split(vm_map_entry_t entry) { vm_page_t m, m_next; vm_object_t orig_object, new_object, source; vm_pindex_t idx, offidxstart; vm_size_t size; orig_object = entry->object.vm_object; if (orig_object->type != OBJT_DEFAULT && orig_object->type != OBJT_SWAP) return; if (orig_object->ref_count <= 1) return; VM_OBJECT_WUNLOCK(orig_object); offidxstart = OFF_TO_IDX(entry->offset); size = atop(entry->end - entry->start); /* * If swap_pager_copy() is later called, it will convert new_object * into a swap object. */ new_object = vm_object_allocate(OBJT_DEFAULT, size); /* * At this point, the new object is still private, so the order in * which the original and new objects are locked does not matter. */ VM_OBJECT_WLOCK(new_object); VM_OBJECT_WLOCK(orig_object); source = orig_object->backing_object; if (source != NULL) { VM_OBJECT_WLOCK(source); if ((source->flags & OBJ_DEAD) != 0) { VM_OBJECT_WUNLOCK(source); VM_OBJECT_WUNLOCK(orig_object); VM_OBJECT_WUNLOCK(new_object); vm_object_deallocate(new_object); VM_OBJECT_WLOCK(orig_object); return; } LIST_INSERT_HEAD(&source->shadow_head, new_object, shadow_list); source->shadow_count++; vm_object_reference_locked(source); /* for new_object */ vm_object_clear_flag(source, OBJ_ONEMAPPING); VM_OBJECT_WUNLOCK(source); new_object->backing_object_offset = orig_object->backing_object_offset + entry->offset; new_object->backing_object = source; } if (orig_object->cred != NULL) { new_object->cred = orig_object->cred; crhold(orig_object->cred); new_object->charge = ptoa(size); KASSERT(orig_object->charge >= ptoa(size), ("orig_object->charge < 0")); orig_object->charge -= ptoa(size); } retry: m = vm_page_find_least(orig_object, offidxstart); for (; m != NULL && (idx = m->pindex - offidxstart) < size; m = m_next) { m_next = TAILQ_NEXT(m, listq); /* * We must wait for pending I/O to complete before we can * rename the page. * * We do not have to VM_PROT_NONE the page as mappings should * not be changed by this operation. */ if (vm_page_busied(m)) { VM_OBJECT_WUNLOCK(new_object); vm_page_lock(m); VM_OBJECT_WUNLOCK(orig_object); vm_page_busy_sleep(m, "spltwt"); VM_OBJECT_WLOCK(orig_object); VM_OBJECT_WLOCK(new_object); goto retry; } /* vm_page_rename() will handle dirty and cache. */ if (vm_page_rename(m, new_object, idx)) { VM_OBJECT_WUNLOCK(new_object); VM_OBJECT_WUNLOCK(orig_object); VM_WAIT; VM_OBJECT_WLOCK(orig_object); VM_OBJECT_WLOCK(new_object); goto retry; } #if VM_NRESERVLEVEL > 0 /* * If some of the reservation's allocated pages remain with * the original object, then transferring the reservation to * the new object is neither particularly beneficial nor * particularly harmful as compared to leaving the reservation * with the original object. If, however, all of the * reservation's allocated pages are transferred to the new * object, then transferring the reservation is typically * beneficial. Determining which of these two cases applies * would be more costly than unconditionally renaming the * reservation. */ vm_reserv_rename(m, new_object, orig_object, offidxstart); #endif if (orig_object->type == OBJT_SWAP) vm_page_xbusy(m); } if (orig_object->type == OBJT_SWAP) { /* * swap_pager_copy() can sleep, in which case the orig_object's * and new_object's locks are released and reacquired. */ swap_pager_copy(orig_object, new_object, offidxstart, 0); TAILQ_FOREACH(m, &new_object->memq, listq) vm_page_xunbusy(m); /* * Transfer any cached pages from orig_object to new_object. * If swap_pager_copy() found swapped out pages within the * specified range of orig_object, then it changed * new_object's type to OBJT_SWAP when it transferred those * pages to new_object. Otherwise, new_object's type * should still be OBJT_DEFAULT and orig_object should not * contain any cached pages within the specified range. */ if (__predict_false(!vm_object_cache_is_empty(orig_object))) vm_page_cache_transfer(orig_object, offidxstart, new_object); } VM_OBJECT_WUNLOCK(orig_object); VM_OBJECT_WUNLOCK(new_object); entry->object.vm_object = new_object; entry->offset = 0LL; vm_object_deallocate(orig_object); VM_OBJECT_WLOCK(new_object); } #define OBSC_TEST_ALL_SHADOWED 0x0001 #define OBSC_COLLAPSE_NOWAIT 0x0002 #define OBSC_COLLAPSE_WAIT 0x0004 static int vm_object_backing_scan(vm_object_t object, int op) { int r = 1; vm_page_t p; vm_object_t backing_object; vm_pindex_t backing_offset_index; VM_OBJECT_ASSERT_WLOCKED(object); VM_OBJECT_ASSERT_WLOCKED(object->backing_object); backing_object = object->backing_object; backing_offset_index = OFF_TO_IDX(object->backing_object_offset); /* * Initial conditions */ if (op & OBSC_TEST_ALL_SHADOWED) { /* * We do not want to have to test for the existence of cache * or swap pages in the backing object. XXX but with the * new swapper this would be pretty easy to do. * * XXX what about anonymous MAP_SHARED memory that hasn't * been ZFOD faulted yet? If we do not test for this, the * shadow test may succeed! XXX */ if (backing_object->type != OBJT_DEFAULT) { return (0); } } if (op & OBSC_COLLAPSE_WAIT) { vm_object_set_flag(backing_object, OBJ_DEAD); } /* * Our scan */ p = TAILQ_FIRST(&backing_object->memq); while (p) { vm_page_t next = TAILQ_NEXT(p, listq); vm_pindex_t new_pindex = p->pindex - backing_offset_index; if (op & OBSC_TEST_ALL_SHADOWED) { vm_page_t pp; /* * Ignore pages outside the parent object's range * and outside the parent object's mapping of the * backing object. * * note that we do not busy the backing object's * page. */ if ( p->pindex < backing_offset_index || new_pindex >= object->size ) { p = next; continue; } /* * See if the parent has the page or if the parent's * object pager has the page. If the parent has the * page but the page is not valid, the parent's * object pager must have the page. * * If this fails, the parent does not completely shadow * the object and we might as well give up now. */ pp = vm_page_lookup(object, new_pindex); if ( (pp == NULL || pp->valid == 0) && !vm_pager_has_page(object, new_pindex, NULL, NULL) ) { r = 0; break; } } /* * Check for busy page */ if (op & (OBSC_COLLAPSE_WAIT | OBSC_COLLAPSE_NOWAIT)) { vm_page_t pp; if (op & OBSC_COLLAPSE_NOWAIT) { if (!p->valid || vm_page_busied(p)) { p = next; continue; } } else if (op & OBSC_COLLAPSE_WAIT) { if (vm_page_busied(p)) { VM_OBJECT_WUNLOCK(object); vm_page_lock(p); VM_OBJECT_WUNLOCK(backing_object); vm_page_busy_sleep(p, "vmocol"); VM_OBJECT_WLOCK(object); VM_OBJECT_WLOCK(backing_object); /* * If we slept, anything could have * happened. Since the object is * marked dead, the backing offset * should not have changed so we * just restart our scan. */ p = TAILQ_FIRST(&backing_object->memq); continue; } } KASSERT( p->object == backing_object, ("vm_object_backing_scan: object mismatch") ); if ( p->pindex < backing_offset_index || new_pindex >= object->size ) { if (backing_object->type == OBJT_SWAP) swap_pager_freespace(backing_object, p->pindex, 1); /* * Page is out of the parent object's range, we * can simply destroy it. */ vm_page_lock(p); KASSERT(!pmap_page_is_mapped(p), ("freeing mapped page %p", p)); if (p->wire_count == 0) vm_page_free(p); else vm_page_remove(p); vm_page_unlock(p); p = next; continue; } pp = vm_page_lookup(object, new_pindex); if ( (op & OBSC_COLLAPSE_NOWAIT) != 0 && (pp != NULL && pp->valid == 0) ) { if (backing_object->type == OBJT_SWAP) swap_pager_freespace(backing_object, p->pindex, 1); /* * The page in the parent is not (yet) valid. * We don't know anything about the state of * the original page. It might be mapped, * so we must avoid the next if here. * * This is due to a race in vm_fault() where * we must unbusy the original (backing_obj) * page before we can (re)lock the parent. * Hence we can get here. */ p = next; continue; } if ( pp != NULL || vm_pager_has_page(object, new_pindex, NULL, NULL) ) { if (backing_object->type == OBJT_SWAP) swap_pager_freespace(backing_object, p->pindex, 1); /* * page already exists in parent OR swap exists * for this location in the parent. Destroy * the original page from the backing object. * * Leave the parent's page alone */ vm_page_lock(p); KASSERT(!pmap_page_is_mapped(p), ("freeing mapped page %p", p)); if (p->wire_count == 0) vm_page_free(p); else vm_page_remove(p); vm_page_unlock(p); p = next; continue; } /* * Page does not exist in parent, rename the * page from the backing object to the main object. * * If the page was mapped to a process, it can remain * mapped through the rename. * vm_page_rename() will handle dirty and cache. */ if (vm_page_rename(p, object, new_pindex)) { if (op & OBSC_COLLAPSE_NOWAIT) { p = next; continue; } VM_OBJECT_WUNLOCK(backing_object); VM_OBJECT_WUNLOCK(object); VM_WAIT; VM_OBJECT_WLOCK(object); VM_OBJECT_WLOCK(backing_object); p = TAILQ_FIRST(&backing_object->memq); continue; } /* Use the old pindex to free the right page. */ if (backing_object->type == OBJT_SWAP) swap_pager_freespace(backing_object, new_pindex + backing_offset_index, 1); #if VM_NRESERVLEVEL > 0 /* * Rename the reservation. */ vm_reserv_rename(p, object, backing_object, backing_offset_index); #endif } p = next; } return (r); } /* * this version of collapse allows the operation to occur earlier and * when paging_in_progress is true for an object... This is not a complete * operation, but should plug 99.9% of the rest of the leaks. */ static void vm_object_qcollapse(vm_object_t object) { vm_object_t backing_object = object->backing_object; VM_OBJECT_ASSERT_WLOCKED(object); VM_OBJECT_ASSERT_WLOCKED(backing_object); if (backing_object->ref_count != 1) return; vm_object_backing_scan(object, OBSC_COLLAPSE_NOWAIT); } /* * vm_object_collapse: * * Collapse an object with the object backing it. * Pages in the backing object are moved into the * parent, and the backing object is deallocated. */ void vm_object_collapse(vm_object_t object) { VM_OBJECT_ASSERT_WLOCKED(object); while (TRUE) { vm_object_t backing_object; /* * Verify that the conditions are right for collapse: * * The object exists and the backing object exists. */ if ((backing_object = object->backing_object) == NULL) break; /* * we check the backing object first, because it is most likely * not collapsable. */ VM_OBJECT_WLOCK(backing_object); if (backing_object->handle != NULL || (backing_object->type != OBJT_DEFAULT && backing_object->type != OBJT_SWAP) || (backing_object->flags & OBJ_DEAD) || object->handle != NULL || (object->type != OBJT_DEFAULT && object->type != OBJT_SWAP) || (object->flags & OBJ_DEAD)) { VM_OBJECT_WUNLOCK(backing_object); break; } if ( object->paging_in_progress != 0 || backing_object->paging_in_progress != 0 ) { vm_object_qcollapse(object); VM_OBJECT_WUNLOCK(backing_object); break; } /* * We know that we can either collapse the backing object (if * the parent is the only reference to it) or (perhaps) have * the parent bypass the object if the parent happens to shadow * all the resident pages in the entire backing object. * * This is ignoring pager-backed pages such as swap pages. * vm_object_backing_scan fails the shadowing test in this * case. */ if (backing_object->ref_count == 1) { /* * If there is exactly one reference to the backing * object, we can collapse it into the parent. */ vm_object_backing_scan(object, OBSC_COLLAPSE_WAIT); #if VM_NRESERVLEVEL > 0 /* * Break any reservations from backing_object. */ if (__predict_false(!LIST_EMPTY(&backing_object->rvq))) vm_reserv_break_all(backing_object); #endif /* * Move the pager from backing_object to object. */ if (backing_object->type == OBJT_SWAP) { /* * swap_pager_copy() can sleep, in which case * the backing_object's and object's locks are * released and reacquired. * Since swap_pager_copy() is being asked to * destroy the source, it will change the * backing_object's type to OBJT_DEFAULT. */ swap_pager_copy( backing_object, object, OFF_TO_IDX(object->backing_object_offset), TRUE); /* * Free any cached pages from backing_object. */ if (__predict_false( !vm_object_cache_is_empty(backing_object))) vm_page_cache_free(backing_object, 0, 0); } /* * Object now shadows whatever backing_object did. * Note that the reference to * backing_object->backing_object moves from within * backing_object to within object. */ LIST_REMOVE(object, shadow_list); backing_object->shadow_count--; if (backing_object->backing_object) { VM_OBJECT_WLOCK(backing_object->backing_object); LIST_REMOVE(backing_object, shadow_list); LIST_INSERT_HEAD( &backing_object->backing_object->shadow_head, object, shadow_list); /* * The shadow_count has not changed. */ VM_OBJECT_WUNLOCK(backing_object->backing_object); } object->backing_object = backing_object->backing_object; object->backing_object_offset += backing_object->backing_object_offset; /* * Discard backing_object. * * Since the backing object has no pages, no pager left, * and no object references within it, all that is * necessary is to dispose of it. */ KASSERT(backing_object->ref_count == 1, ( "backing_object %p was somehow re-referenced during collapse!", backing_object)); backing_object->type = OBJT_DEAD; backing_object->ref_count = 0; VM_OBJECT_WUNLOCK(backing_object); vm_object_destroy(backing_object); object_collapses++; } else { vm_object_t new_backing_object; /* * If we do not entirely shadow the backing object, * there is nothing we can do so we give up. */ if (object->resident_page_count != object->size && vm_object_backing_scan(object, OBSC_TEST_ALL_SHADOWED) == 0) { VM_OBJECT_WUNLOCK(backing_object); break; } /* * Make the parent shadow the next object in the * chain. Deallocating backing_object will not remove * it, since its reference count is at least 2. */ LIST_REMOVE(object, shadow_list); backing_object->shadow_count--; new_backing_object = backing_object->backing_object; if ((object->backing_object = new_backing_object) != NULL) { VM_OBJECT_WLOCK(new_backing_object); LIST_INSERT_HEAD( &new_backing_object->shadow_head, object, shadow_list ); new_backing_object->shadow_count++; vm_object_reference_locked(new_backing_object); VM_OBJECT_WUNLOCK(new_backing_object); object->backing_object_offset += backing_object->backing_object_offset; } /* * Drop the reference count on backing_object. Since * its ref_count was at least 2, it will not vanish. */ backing_object->ref_count--; VM_OBJECT_WUNLOCK(backing_object); object_bypasses++; } /* * Try again with this object's new backing object. */ } } /* * vm_object_page_remove: * * For the given object, either frees or invalidates each of the * specified pages. In general, a page is freed. However, if a page is * wired for any reason other than the existence of a managed, wired * mapping, then it may be invalidated but not removed from the object. * Pages are specified by the given range ["start", "end") and the option * OBJPR_CLEANONLY. As a special case, if "end" is zero, then the range * extends from "start" to the end of the object. If the option * OBJPR_CLEANONLY is specified, then only the non-dirty pages within the * specified range are affected. If the option OBJPR_NOTMAPPED is * specified, then the pages within the specified range must have no * mappings. Otherwise, if this option is not specified, any mappings to * the specified pages are removed before the pages are freed or * invalidated. * * In general, this operation should only be performed on objects that * contain managed pages. There are, however, two exceptions. First, it * is performed on the kernel and kmem objects by vm_map_entry_delete(). * Second, it is used by msync(..., MS_INVALIDATE) to invalidate device- * backed pages. In both of these cases, the option OBJPR_CLEANONLY must * not be specified and the option OBJPR_NOTMAPPED must be specified. * * The object must be locked. */ void vm_object_page_remove(vm_object_t object, vm_pindex_t start, vm_pindex_t end, int options) { vm_page_t p, next; int wirings; VM_OBJECT_ASSERT_WLOCKED(object); KASSERT((object->flags & OBJ_UNMANAGED) == 0 || (options & (OBJPR_CLEANONLY | OBJPR_NOTMAPPED)) == OBJPR_NOTMAPPED, ("vm_object_page_remove: illegal options for object %p", object)); if (object->resident_page_count == 0) goto skipmemq; vm_object_pip_add(object, 1); again: p = vm_page_find_least(object, start); /* * Here, the variable "p" is either (1) the page with the least pindex * greater than or equal to the parameter "start" or (2) NULL. */ for (; p != NULL && (p->pindex < end || end == 0); p = next) { next = TAILQ_NEXT(p, listq); /* * If the page is wired for any reason besides the existence * of managed, wired mappings, then it cannot be freed. For * example, fictitious pages, which represent device memory, * are inherently wired and cannot be freed. They can, * however, be invalidated if the option OBJPR_CLEANONLY is * not specified. */ vm_page_lock(p); if (vm_page_xbusied(p)) { VM_OBJECT_WUNLOCK(object); vm_page_busy_sleep(p, "vmopax"); VM_OBJECT_WLOCK(object); goto again; } if ((wirings = p->wire_count) != 0 && (wirings = pmap_page_wired_mappings(p)) != p->wire_count) { if ((options & (OBJPR_NOTWIRED | OBJPR_NOTMAPPED)) == 0) { pmap_remove_all(p); /* Account for removal of wired mappings. */ if (wirings != 0) p->wire_count -= wirings; } if ((options & OBJPR_CLEANONLY) == 0) { p->valid = 0; vm_page_undirty(p); } goto next; } if (vm_page_busied(p)) { VM_OBJECT_WUNLOCK(object); vm_page_busy_sleep(p, "vmopar"); VM_OBJECT_WLOCK(object); goto again; } KASSERT((p->flags & PG_FICTITIOUS) == 0, ("vm_object_page_remove: page %p is fictitious", p)); if ((options & OBJPR_CLEANONLY) != 0 && p->valid != 0) { if ((options & OBJPR_NOTMAPPED) == 0) pmap_remove_write(p); if (p->dirty) goto next; } if ((options & OBJPR_NOTMAPPED) == 0) { if ((options & OBJPR_NOTWIRED) != 0 && wirings != 0) goto next; pmap_remove_all(p); /* Account for removal of wired mappings. */ if (wirings != 0) { KASSERT(p->wire_count == wirings, ("inconsistent wire count %d %d %p", p->wire_count, wirings, p)); p->wire_count = 0; atomic_subtract_int(&vm_cnt.v_wire_count, 1); } } vm_page_free(p); next: vm_page_unlock(p); } vm_object_pip_wakeup(object); skipmemq: if (__predict_false(!vm_object_cache_is_empty(object))) vm_page_cache_free(object, start, end); } /* * vm_object_page_cache: * * For the given object, attempt to move the specified clean * pages to the cache queue. If a page is wired for any reason, * then it will not be changed. Pages are specified by the given * range ["start", "end"). As a special case, if "end" is zero, * then the range extends from "start" to the end of the object. * Any mappings to the specified pages are removed before the * pages are moved to the cache queue. * * This operation should only be performed on objects that * contain non-fictitious, managed pages. * * The object must be locked. */ void vm_object_page_cache(vm_object_t object, vm_pindex_t start, vm_pindex_t end) { struct mtx *mtx, *new_mtx; vm_page_t p, next; VM_OBJECT_ASSERT_WLOCKED(object); KASSERT((object->flags & (OBJ_FICTITIOUS | OBJ_UNMANAGED)) == 0, ("vm_object_page_cache: illegal object %p", object)); if (object->resident_page_count == 0) return; p = vm_page_find_least(object, start); /* * Here, the variable "p" is either (1) the page with the least pindex * greater than or equal to the parameter "start" or (2) NULL. */ mtx = NULL; for (; p != NULL && (p->pindex < end || end == 0); p = next) { next = TAILQ_NEXT(p, listq); /* * Avoid releasing and reacquiring the same page lock. */ new_mtx = vm_page_lockptr(p); if (mtx != new_mtx) { if (mtx != NULL) mtx_unlock(mtx); mtx = new_mtx; mtx_lock(mtx); } vm_page_try_to_cache(p); } if (mtx != NULL) mtx_unlock(mtx); } /* * Populate the specified range of the object with valid pages. Returns * TRUE if the range is successfully populated and FALSE otherwise. * * Note: This function should be optimized to pass a larger array of * pages to vm_pager_get_pages() before it is applied to a non- * OBJT_DEVICE object. * * The object must be locked. */ boolean_t vm_object_populate(vm_object_t object, vm_pindex_t start, vm_pindex_t end) { vm_page_t m, ma[1]; vm_pindex_t pindex; int rv; VM_OBJECT_ASSERT_WLOCKED(object); for (pindex = start; pindex < end; pindex++) { m = vm_page_grab(object, pindex, VM_ALLOC_NORMAL); if (m->valid != VM_PAGE_BITS_ALL) { ma[0] = m; rv = vm_pager_get_pages(object, ma, 1, 0); m = vm_page_lookup(object, pindex); if (m == NULL) break; if (rv != VM_PAGER_OK) { vm_page_lock(m); vm_page_free(m); vm_page_unlock(m); break; } } /* * Keep "m" busy because a subsequent iteration may unlock * the object. */ } if (pindex > start) { m = vm_page_lookup(object, start); while (m != NULL && m->pindex < pindex) { vm_page_xunbusy(m); m = TAILQ_NEXT(m, listq); } } return (pindex == end); } /* * Routine: vm_object_coalesce * Function: Coalesces two objects backing up adjoining * regions of memory into a single object. * * returns TRUE if objects were combined. * * NOTE: Only works at the moment if the second object is NULL - * if it's not, which object do we lock first? * * Parameters: * prev_object First object to coalesce * prev_offset Offset into prev_object * prev_size Size of reference to prev_object * next_size Size of reference to the second object * reserved Indicator that extension region has * swap accounted for * * Conditions: * The object must *not* be locked. */ boolean_t vm_object_coalesce(vm_object_t prev_object, vm_ooffset_t prev_offset, vm_size_t prev_size, vm_size_t next_size, boolean_t reserved) { vm_pindex_t next_pindex; if (prev_object == NULL) return (TRUE); VM_OBJECT_WLOCK(prev_object); if ((prev_object->type != OBJT_DEFAULT && prev_object->type != OBJT_SWAP) || (prev_object->flags & OBJ_TMPFS_NODE) != 0) { VM_OBJECT_WUNLOCK(prev_object); return (FALSE); } /* * Try to collapse the object first */ vm_object_collapse(prev_object); /* * Can't coalesce if: . more than one reference . paged out . shadows * another object . has a copy elsewhere (any of which mean that the * pages not mapped to prev_entry may be in use anyway) */ if (prev_object->backing_object != NULL) { VM_OBJECT_WUNLOCK(prev_object); return (FALSE); } prev_size >>= PAGE_SHIFT; next_size >>= PAGE_SHIFT; next_pindex = OFF_TO_IDX(prev_offset) + prev_size; if ((prev_object->ref_count > 1) && (prev_object->size != next_pindex)) { VM_OBJECT_WUNLOCK(prev_object); return (FALSE); } /* * Account for the charge. */ if (prev_object->cred != NULL) { /* * If prev_object was charged, then this mapping, * althought not charged now, may become writable * later. Non-NULL cred in the object would prevent * swap reservation during enabling of the write * access, so reserve swap now. Failed reservation * cause allocation of the separate object for the map * entry, and swap reservation for this entry is * managed in appropriate time. */ if (!reserved && !swap_reserve_by_cred(ptoa(next_size), prev_object->cred)) { return (FALSE); } prev_object->charge += ptoa(next_size); } /* * Remove any pages that may still be in the object from a previous * deallocation. */ if (next_pindex < prev_object->size) { vm_object_page_remove(prev_object, next_pindex, next_pindex + next_size, 0); if (prev_object->type == OBJT_SWAP) swap_pager_freespace(prev_object, next_pindex, next_size); #if 0 if (prev_object->cred != NULL) { KASSERT(prev_object->charge >= ptoa(prev_object->size - next_pindex), ("object %p overcharged 1 %jx %jx", prev_object, (uintmax_t)next_pindex, (uintmax_t)next_size)); prev_object->charge -= ptoa(prev_object->size - next_pindex); } #endif } /* * Extend the object if necessary. */ if (next_pindex + next_size > prev_object->size) prev_object->size = next_pindex + next_size; VM_OBJECT_WUNLOCK(prev_object); return (TRUE); } void vm_object_set_writeable_dirty(vm_object_t object) { VM_OBJECT_ASSERT_WLOCKED(object); if (object->type != OBJT_VNODE) { if ((object->flags & OBJ_TMPFS_NODE) != 0) { KASSERT(object->type == OBJT_SWAP, ("non-swap tmpfs")); vm_object_set_flag(object, OBJ_TMPFS_DIRTY); } return; } object->generation++; if ((object->flags & OBJ_MIGHTBEDIRTY) != 0) return; vm_object_set_flag(object, OBJ_MIGHTBEDIRTY); } /* * vm_object_unwire: * * For each page offset within the specified range of the given object, * find the highest-level page in the shadow chain and unwire it. A page * must exist at every page offset, and the highest-level page must be * wired. */ void vm_object_unwire(vm_object_t object, vm_ooffset_t offset, vm_size_t length, uint8_t queue) { vm_object_t tobject; vm_page_t m, tm; vm_pindex_t end_pindex, pindex, tpindex; int depth, locked_depth; KASSERT((offset & PAGE_MASK) == 0, ("vm_object_unwire: offset is not page aligned")); KASSERT((length & PAGE_MASK) == 0, ("vm_object_unwire: length is not a multiple of PAGE_SIZE")); /* The wired count of a fictitious page never changes. */ if ((object->flags & OBJ_FICTITIOUS) != 0) return; pindex = OFF_TO_IDX(offset); end_pindex = pindex + atop(length); locked_depth = 1; VM_OBJECT_RLOCK(object); m = vm_page_find_least(object, pindex); while (pindex < end_pindex) { if (m == NULL || pindex < m->pindex) { /* * The first object in the shadow chain doesn't * contain a page at the current index. Therefore, * the page must exist in a backing object. */ tobject = object; tpindex = pindex; depth = 0; do { tpindex += OFF_TO_IDX(tobject->backing_object_offset); tobject = tobject->backing_object; KASSERT(tobject != NULL, ("vm_object_unwire: missing page")); if ((tobject->flags & OBJ_FICTITIOUS) != 0) goto next_page; depth++; if (depth == locked_depth) { locked_depth++; VM_OBJECT_RLOCK(tobject); } } while ((tm = vm_page_lookup(tobject, tpindex)) == NULL); } else { tm = m; m = TAILQ_NEXT(m, listq); } vm_page_lock(tm); vm_page_unwire(tm, queue); vm_page_unlock(tm); next_page: pindex++; } /* Release the accumulated object locks. */ for (depth = 0; depth < locked_depth; depth++) { tobject = object->backing_object; VM_OBJECT_RUNLOCK(object); object = tobject; } } +struct vnode * +vm_object_vnode(vm_object_t object) +{ + + VM_OBJECT_ASSERT_LOCKED(object); + if (object->type == OBJT_VNODE) + return (object->handle); + if (object->type == OBJT_SWAP && (object->flags & OBJ_TMPFS) != 0) + return (object->un_pager.swp.swp_tmpfs); + return (NULL); +} + static int sysctl_vm_object_list(SYSCTL_HANDLER_ARGS) { struct kinfo_vmobject kvo; char *fullpath, *freepath; struct vnode *vp; struct vattr va; vm_object_t obj; vm_page_t m; int count, error; if (req->oldptr == NULL) { /* * If an old buffer has not been provided, generate an * estimate of the space needed for a subsequent call. */ mtx_lock(&vm_object_list_mtx); count = 0; TAILQ_FOREACH(obj, &vm_object_list, object_list) { if (obj->type == OBJT_DEAD) continue; count++; } mtx_unlock(&vm_object_list_mtx); return (SYSCTL_OUT(req, NULL, sizeof(struct kinfo_vmobject) * count * 11 / 10)); } error = 0; /* * VM objects are type stable and are never removed from the * list once added. This allows us to safely read obj->object_list * after reacquiring the VM object lock. */ mtx_lock(&vm_object_list_mtx); TAILQ_FOREACH(obj, &vm_object_list, object_list) { if (obj->type == OBJT_DEAD) continue; VM_OBJECT_RLOCK(obj); if (obj->type == OBJT_DEAD) { VM_OBJECT_RUNLOCK(obj); continue; } mtx_unlock(&vm_object_list_mtx); kvo.kvo_size = ptoa(obj->size); kvo.kvo_resident = obj->resident_page_count; kvo.kvo_ref_count = obj->ref_count; kvo.kvo_shadow_count = obj->shadow_count; kvo.kvo_memattr = obj->memattr; kvo.kvo_active = 0; kvo.kvo_inactive = 0; TAILQ_FOREACH(m, &obj->memq, listq) { /* * A page may belong to the object but be * dequeued and set to PQ_NONE while the * object lock is not held. This makes the * reads of m->queue below racy, and we do not * count pages set to PQ_NONE. However, this * sysctl is only meant to give an * approximation of the system anyway. */ if (m->queue == PQ_ACTIVE) kvo.kvo_active++; else if (m->queue == PQ_INACTIVE) kvo.kvo_inactive++; } kvo.kvo_vn_fileid = 0; kvo.kvo_vn_fsid = 0; freepath = NULL; fullpath = ""; vp = NULL; switch (obj->type) { case OBJT_DEFAULT: kvo.kvo_type = KVME_TYPE_DEFAULT; break; case OBJT_VNODE: kvo.kvo_type = KVME_TYPE_VNODE; vp = obj->handle; vref(vp); break; case OBJT_SWAP: kvo.kvo_type = KVME_TYPE_SWAP; break; case OBJT_DEVICE: kvo.kvo_type = KVME_TYPE_DEVICE; break; case OBJT_PHYS: kvo.kvo_type = KVME_TYPE_PHYS; break; case OBJT_DEAD: kvo.kvo_type = KVME_TYPE_DEAD; break; case OBJT_SG: kvo.kvo_type = KVME_TYPE_SG; break; case OBJT_MGTDEVICE: kvo.kvo_type = KVME_TYPE_MGTDEVICE; break; default: kvo.kvo_type = KVME_TYPE_UNKNOWN; break; } VM_OBJECT_RUNLOCK(obj); if (vp != NULL) { vn_fullpath(curthread, vp, &fullpath, &freepath); vn_lock(vp, LK_SHARED | LK_RETRY); if (VOP_GETATTR(vp, &va, curthread->td_ucred) == 0) { kvo.kvo_vn_fileid = va.va_fileid; kvo.kvo_vn_fsid = va.va_fsid; } vput(vp); } strlcpy(kvo.kvo_path, fullpath, sizeof(kvo.kvo_path)); if (freepath != NULL) free(freepath, M_TEMP); /* Pack record size down */ kvo.kvo_structsize = offsetof(struct kinfo_vmobject, kvo_path) + strlen(kvo.kvo_path) + 1; kvo.kvo_structsize = roundup(kvo.kvo_structsize, sizeof(uint64_t)); error = SYSCTL_OUT(req, &kvo, kvo.kvo_structsize); mtx_lock(&vm_object_list_mtx); if (error) break; } mtx_unlock(&vm_object_list_mtx); return (error); } SYSCTL_PROC(_vm, OID_AUTO, objects, CTLTYPE_STRUCT | CTLFLAG_RW | CTLFLAG_SKIP | CTLFLAG_MPSAFE, NULL, 0, sysctl_vm_object_list, "S,kinfo_vmobject", "List of VM objects"); #include "opt_ddb.h" #ifdef DDB #include #include #include static int _vm_object_in_map(vm_map_t map, vm_object_t object, vm_map_entry_t entry) { vm_map_t tmpm; vm_map_entry_t tmpe; vm_object_t obj; int entcount; if (map == 0) return 0; if (entry == 0) { tmpe = map->header.next; entcount = map->nentries; while (entcount-- && (tmpe != &map->header)) { if (_vm_object_in_map(map, object, tmpe)) { return 1; } tmpe = tmpe->next; } } else if (entry->eflags & MAP_ENTRY_IS_SUB_MAP) { tmpm = entry->object.sub_map; tmpe = tmpm->header.next; entcount = tmpm->nentries; while (entcount-- && tmpe != &tmpm->header) { if (_vm_object_in_map(tmpm, object, tmpe)) { return 1; } tmpe = tmpe->next; } } else if ((obj = entry->object.vm_object) != NULL) { for (; obj; obj = obj->backing_object) if (obj == object) { return 1; } } return 0; } static int vm_object_in_map(vm_object_t object) { struct proc *p; /* sx_slock(&allproc_lock); */ FOREACH_PROC_IN_SYSTEM(p) { if (!p->p_vmspace /* || (p->p_flag & (P_SYSTEM|P_WEXIT)) */) continue; if (_vm_object_in_map(&p->p_vmspace->vm_map, object, 0)) { /* sx_sunlock(&allproc_lock); */ return 1; } } /* sx_sunlock(&allproc_lock); */ if (_vm_object_in_map(kernel_map, object, 0)) return 1; return 0; } DB_SHOW_COMMAND(vmochk, vm_object_check) { vm_object_t object; /* * make sure that internal objs are in a map somewhere * and none have zero ref counts. */ TAILQ_FOREACH(object, &vm_object_list, object_list) { if (object->handle == NULL && (object->type == OBJT_DEFAULT || object->type == OBJT_SWAP)) { if (object->ref_count == 0) { db_printf("vmochk: internal obj has zero ref count: %ld\n", (long)object->size); } if (!vm_object_in_map(object)) { db_printf( "vmochk: internal obj is not in a map: " "ref: %d, size: %lu: 0x%lx, backing_object: %p\n", object->ref_count, (u_long)object->size, (u_long)object->size, (void *)object->backing_object); } } } } /* * vm_object_print: [ debug ] */ DB_SHOW_COMMAND(object, vm_object_print_static) { /* XXX convert args. */ vm_object_t object = (vm_object_t)addr; boolean_t full = have_addr; vm_page_t p; /* XXX count is an (unused) arg. Avoid shadowing it. */ #define count was_count int count; if (object == NULL) return; db_iprintf( "Object %p: type=%d, size=0x%jx, res=%d, ref=%d, flags=0x%x ruid %d charge %jx\n", object, (int)object->type, (uintmax_t)object->size, object->resident_page_count, object->ref_count, object->flags, object->cred ? object->cred->cr_ruid : -1, (uintmax_t)object->charge); db_iprintf(" sref=%d, backing_object(%d)=(%p)+0x%jx\n", object->shadow_count, object->backing_object ? object->backing_object->ref_count : 0, object->backing_object, (uintmax_t)object->backing_object_offset); if (!full) return; db_indent += 2; count = 0; TAILQ_FOREACH(p, &object->memq, listq) { if (count == 0) db_iprintf("memory:="); else if (count == 6) { db_printf("\n"); db_iprintf(" ..."); count = 0; } else db_printf(","); count++; db_printf("(off=0x%jx,page=0x%jx)", (uintmax_t)p->pindex, (uintmax_t)VM_PAGE_TO_PHYS(p)); } if (count != 0) db_printf("\n"); db_indent -= 2; } /* XXX. */ #undef count /* XXX need this non-static entry for calling from vm_map_print. */ void vm_object_print( /* db_expr_t */ long addr, boolean_t have_addr, /* db_expr_t */ long count, char *modif) { vm_object_print_static(addr, have_addr, count, modif); } DB_SHOW_COMMAND(vmopag, vm_object_print_pages) { vm_object_t object; vm_pindex_t fidx; vm_paddr_t pa; vm_page_t m, prev_m; int rcount, nl, c; nl = 0; TAILQ_FOREACH(object, &vm_object_list, object_list) { db_printf("new object: %p\n", (void *)object); if (nl > 18) { c = cngetc(); if (c != ' ') return; nl = 0; } nl++; rcount = 0; fidx = 0; pa = -1; TAILQ_FOREACH(m, &object->memq, listq) { if (m->pindex > 128) break; if ((prev_m = TAILQ_PREV(m, pglist, listq)) != NULL && prev_m->pindex + 1 != m->pindex) { if (rcount) { db_printf(" index(%ld)run(%d)pa(0x%lx)\n", (long)fidx, rcount, (long)pa); if (nl > 18) { c = cngetc(); if (c != ' ') return; nl = 0; } nl++; rcount = 0; } } if (rcount && (VM_PAGE_TO_PHYS(m) == pa + rcount * PAGE_SIZE)) { ++rcount; continue; } if (rcount) { db_printf(" index(%ld)run(%d)pa(0x%lx)\n", (long)fidx, rcount, (long)pa); if (nl > 18) { c = cngetc(); if (c != ' ') return; nl = 0; } nl++; } fidx = m->pindex; pa = VM_PAGE_TO_PHYS(m); rcount = 1; } if (rcount) { db_printf(" index(%ld)run(%d)pa(0x%lx)\n", (long)fidx, rcount, (long)pa); if (nl > 18) { c = cngetc(); if (c != ' ') return; nl = 0; } nl++; } } } #endif /* DDB */ Index: projects/em_mq/sys/vm/vm_object.h =================================================================== --- projects/em_mq/sys/vm/vm_object.h (revision 283943) +++ projects/em_mq/sys/vm/vm_object.h (revision 283944) @@ -1,327 +1,328 @@ /*- * Copyright (c) 1991, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * The Mach Operating System project at Carnegie-Mellon University. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: @(#)vm_object.h 8.3 (Berkeley) 1/12/94 * * * Copyright (c) 1987, 1990 Carnegie-Mellon University. * All rights reserved. * * Authors: Avadis Tevanian, Jr., Michael Wayne Young * * Permission to use, copy, modify and distribute this software and * its documentation is hereby granted, provided that both the copyright * notice and this permission notice appear in all copies of the * software, derivative works or modified versions, and any portions * thereof, and that both notices appear in supporting documentation. * * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. * * Carnegie Mellon requests users of this software to return to * * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU * School of Computer Science * Carnegie Mellon University * Pittsburgh PA 15213-3890 * * any improvements or extensions that they make and grant Carnegie the * rights to redistribute these changes. * * $FreeBSD$ */ /* * Virtual memory object module definitions. */ #ifndef _VM_OBJECT_ #define _VM_OBJECT_ #include #include #include #include #include /* * Types defined: * * vm_object_t Virtual memory object. * * The root of cached pages pool is protected by both the per-object lock * and the free pages queue mutex. * On insert in the cache radix trie, the per-object lock is expected * to be already held and the free pages queue mutex will be * acquired during the operation too. * On remove and lookup from the cache radix trie, only the free * pages queue mutex is expected to be locked. * These rules allow for reliably checking for the presence of cached * pages with only the per-object lock held, thereby reducing contention * for the free pages queue mutex. * * List of locks * (c) const until freed * (o) per-object lock * (f) free pages queue mutex * */ struct vm_object { struct rwlock lock; TAILQ_ENTRY(vm_object) object_list; /* list of all objects */ LIST_HEAD(, vm_object) shadow_head; /* objects that this is a shadow for */ LIST_ENTRY(vm_object) shadow_list; /* chain of shadow objects */ TAILQ_HEAD(respgs, vm_page) memq; /* list of resident pages */ struct vm_radix rtree; /* root of the resident page radix trie*/ vm_pindex_t size; /* Object size */ int generation; /* generation ID */ int ref_count; /* How many refs?? */ int shadow_count; /* how many objects that this is a shadow for */ vm_memattr_t memattr; /* default memory attribute for pages */ objtype_t type; /* type of pager */ u_short flags; /* see below */ u_short pg_color; /* (c) color of first page in obj */ u_int paging_in_progress; /* Paging (in or out) so don't collapse or destroy */ int resident_page_count; /* number of resident pages */ struct vm_object *backing_object; /* object that I'm a shadow of */ vm_ooffset_t backing_object_offset;/* Offset in backing object */ TAILQ_ENTRY(vm_object) pager_object_list; /* list of all objects of this pager type */ LIST_HEAD(, vm_reserv) rvq; /* list of reservations */ struct vm_radix cache; /* (o + f) root of the cache page radix trie */ void *handle; union { /* * VNode pager * * vnp_size - current size of file */ struct { off_t vnp_size; vm_ooffset_t writemappings; } vnp; /* * Device pager * * devp_pglist - list of allocated pages */ struct { TAILQ_HEAD(, vm_page) devp_pglist; struct cdev_pager_ops *ops; struct cdev *dev; } devp; /* * SG pager * * sgp_pglist - list of allocated pages */ struct { TAILQ_HEAD(, vm_page) sgp_pglist; } sgp; /* * Swap pager * * swp_tmpfs - back-pointer to the tmpfs vnode, * if any, which uses the vm object * as backing store. The handle * cannot be reused for linking, * because the vnode can be * reclaimed and recreated, making * the handle changed and hash-chain * invalid. * * swp_bcount - number of swap 'swblock' metablocks, each * contains up to 16 swapblk assignments. * see vm/swap_pager.h */ struct { void *swp_tmpfs; int swp_bcount; } swp; } un_pager; struct ucred *cred; vm_ooffset_t charge; }; /* * Flags */ #define OBJ_FICTITIOUS 0x0001 /* (c) contains fictitious pages */ #define OBJ_UNMANAGED 0x0002 /* (c) contains unmanaged pages */ #define OBJ_ACTIVE 0x0004 /* active objects */ #define OBJ_DEAD 0x0008 /* dead objects (during rundown) */ #define OBJ_NOSPLIT 0x0010 /* dont split this object */ #define OBJ_PIPWNT 0x0040 /* paging in progress wanted */ #define OBJ_MIGHTBEDIRTY 0x0100 /* object might be dirty, only for vnode */ #define OBJ_TMPFS_NODE 0x0200 /* object belongs to tmpfs VREG node */ #define OBJ_TMPFS_DIRTY 0x0400 /* dirty tmpfs obj */ #define OBJ_COLORED 0x1000 /* pg_color is defined */ #define OBJ_ONEMAPPING 0x2000 /* One USE (a single, non-forked) mapping flag */ #define OBJ_DISCONNECTWNT 0x4000 /* disconnect from vnode wanted */ #define OBJ_TMPFS 0x8000 /* has tmpfs vnode allocated */ #define IDX_TO_OFF(idx) (((vm_ooffset_t)(idx)) << PAGE_SHIFT) #define OFF_TO_IDX(off) ((vm_pindex_t)(((vm_ooffset_t)(off)) >> PAGE_SHIFT)) #ifdef _KERNEL #define OBJPC_SYNC 0x1 /* sync I/O */ #define OBJPC_INVAL 0x2 /* invalidate */ #define OBJPC_NOSYNC 0x4 /* skip if VPO_NOSYNC */ /* * The following options are supported by vm_object_page_remove(). */ #define OBJPR_CLEANONLY 0x1 /* Don't remove dirty pages. */ #define OBJPR_NOTMAPPED 0x2 /* Don't unmap pages. */ #define OBJPR_NOTWIRED 0x4 /* Don't remove wired pages. */ TAILQ_HEAD(object_q, vm_object); extern struct object_q vm_object_list; /* list of allocated objects */ extern struct mtx vm_object_list_mtx; /* lock for object list and count */ extern struct vm_object kernel_object_store; extern struct vm_object kmem_object_store; #define kernel_object (&kernel_object_store) #define kmem_object (&kmem_object_store) #define VM_OBJECT_ASSERT_LOCKED(object) \ rw_assert(&(object)->lock, RA_LOCKED) #define VM_OBJECT_ASSERT_RLOCKED(object) \ rw_assert(&(object)->lock, RA_RLOCKED) #define VM_OBJECT_ASSERT_WLOCKED(object) \ rw_assert(&(object)->lock, RA_WLOCKED) #define VM_OBJECT_ASSERT_UNLOCKED(object) \ rw_assert(&(object)->lock, RA_UNLOCKED) #define VM_OBJECT_LOCK_DOWNGRADE(object) \ rw_downgrade(&(object)->lock) #define VM_OBJECT_RLOCK(object) \ rw_rlock(&(object)->lock) #define VM_OBJECT_RUNLOCK(object) \ rw_runlock(&(object)->lock) #define VM_OBJECT_SLEEP(object, wchan, pri, wmesg, timo) \ rw_sleep((wchan), &(object)->lock, (pri), (wmesg), (timo)) #define VM_OBJECT_TRYRLOCK(object) \ rw_try_rlock(&(object)->lock) #define VM_OBJECT_TRYWLOCK(object) \ rw_try_wlock(&(object)->lock) #define VM_OBJECT_TRYUPGRADE(object) \ rw_try_upgrade(&(object)->lock) #define VM_OBJECT_WLOCK(object) \ rw_wlock(&(object)->lock) #define VM_OBJECT_WUNLOCK(object) \ rw_wunlock(&(object)->lock) /* * The object must be locked or thread private. */ static __inline void vm_object_set_flag(vm_object_t object, u_short bits) { object->flags |= bits; } /* * Conditionally set the object's color, which (1) enables the allocation * of physical memory reservations for anonymous objects and larger-than- * superpage-sized named objects and (2) determines the first page offset * within the object at which a reservation may be allocated. In other * words, the color determines the alignment of the object with respect * to the largest superpage boundary. When mapping named objects, like * files or POSIX shared memory objects, the color should be set to zero * before a virtual address is selected for the mapping. In contrast, * for anonymous objects, the color may be set after the virtual address * is selected. * * The object must be locked. */ static __inline void vm_object_color(vm_object_t object, u_short color) { if ((object->flags & OBJ_COLORED) == 0) { object->pg_color = color; object->flags |= OBJ_COLORED; } } void vm_object_clear_flag(vm_object_t object, u_short bits); void vm_object_pip_add(vm_object_t object, short i); void vm_object_pip_subtract(vm_object_t object, short i); void vm_object_pip_wakeup(vm_object_t object); void vm_object_pip_wakeupn(vm_object_t object, short i); void vm_object_pip_wait(vm_object_t object, char *waitid); static __inline boolean_t vm_object_cache_is_empty(vm_object_t object) { return (vm_radix_is_empty(&object->cache)); } vm_object_t vm_object_allocate (objtype_t, vm_pindex_t); boolean_t vm_object_coalesce(vm_object_t, vm_ooffset_t, vm_size_t, vm_size_t, boolean_t); void vm_object_collapse (vm_object_t); void vm_object_deallocate (vm_object_t); void vm_object_destroy (vm_object_t); void vm_object_terminate (vm_object_t); void vm_object_set_writeable_dirty (vm_object_t); void vm_object_init (void); void vm_object_madvise(vm_object_t, vm_pindex_t, vm_pindex_t, int); void vm_object_page_cache(vm_object_t object, vm_pindex_t start, vm_pindex_t end); boolean_t vm_object_page_clean(vm_object_t object, vm_ooffset_t start, vm_ooffset_t end, int flags); void vm_object_page_remove(vm_object_t object, vm_pindex_t start, vm_pindex_t end, int options); boolean_t vm_object_populate(vm_object_t, vm_pindex_t, vm_pindex_t); void vm_object_print(long addr, boolean_t have_addr, long count, char *modif); void vm_object_reference (vm_object_t); void vm_object_reference_locked(vm_object_t); int vm_object_set_memattr(vm_object_t object, vm_memattr_t memattr); void vm_object_shadow (vm_object_t *, vm_ooffset_t *, vm_size_t); void vm_object_split(vm_map_entry_t); boolean_t vm_object_sync(vm_object_t, vm_ooffset_t, vm_size_t, boolean_t, boolean_t); void vm_object_unwire(vm_object_t object, vm_ooffset_t offset, vm_size_t length, uint8_t queue); +struct vnode *vm_object_vnode(vm_object_t object); #endif /* _KERNEL */ #endif /* _VM_OBJECT_ */ Index: projects/em_mq/sys =================================================================== --- projects/em_mq/sys (revision 283943) +++ projects/em_mq/sys (revision 283944) Property changes on: projects/em_mq/sys ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/sys:r283924-283943 Index: projects/em_mq =================================================================== --- projects/em_mq (revision 283943) +++ projects/em_mq (revision 283944) Property changes on: projects/em_mq ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head:r283924-283943