Index: head/sys/conf/kern.pre.mk =================================================================== --- head/sys/conf/kern.pre.mk (revision 360793) +++ head/sys/conf/kern.pre.mk (revision 360794) @@ -1,357 +1,357 @@ # $FreeBSD$ # Part of a unified Makefile for building kernels. This part contains all # of the definitions that need to be before %BEFORE_DEPEND. # Allow user to configure things that only effect src tree builds. # Note: This is duplicated from src.sys.mk to ensure that we include # /etc/src.conf when building the kernel. Kernels can be built without # the rest of /usr/src, but they still always process SRCCONF even though # the normal mechanisms to prevent that (compiling out of tree) won't # work. To ensure they do work, we have to duplicate thee few lines here. SRCCONF?= /etc/src.conf .if (exists(${SRCCONF}) || ${SRCCONF} != "/etc/src.conf") && !target(_srcconf_included_) .include "${SRCCONF}" _srcconf_included_: .endif .include .include .include "kern.opts.mk" # The kernel build always occurs in the object directory which is .CURDIR. .if ${.MAKE.MODE:Unormal:Mmeta} .MAKE.MODE+= curdirOk=yes .endif # The kernel build always expects .OBJDIR=.CURDIR. .OBJDIR: ${.CURDIR} .if defined(NO_OBJWALK) || ${MK_AUTO_OBJ} == "yes" NO_OBJWALK= t NO_MODULES_OBJ= t .endif .if !defined(NO_OBJWALK) _obj= obj .endif # Can be overridden by makeoptions or /etc/make.conf KERNEL_KO?= kernel KERNEL?= kernel KODIR?= /boot/${KERNEL} LDSCRIPT_NAME?= ldscript.$M LDSCRIPT?= $S/conf/${LDSCRIPT_NAME} M= ${MACHINE} AWK?= awk CP?= cp NM?= nm OBJCOPY?= objcopy SIZE?= size .if defined(DEBUG) _MINUS_O= -O CTFFLAGS+= -g .else .if ${MACHINE_CPUARCH} == "powerpc" _MINUS_O= -O # gcc miscompiles some code at -O2 .else _MINUS_O= -O2 .endif .endif .if ${MACHINE_CPUARCH} == "amd64" .if ${COMPILER_TYPE} == "clang" COPTFLAGS?=-O2 -pipe .else COPTFLAGS?=-O2 -frename-registers -pipe .endif .else COPTFLAGS?=${_MINUS_O} -pipe .endif .if !empty(COPTFLAGS:M-O[23s]) && empty(COPTFLAGS:M-fno-strict-aliasing) COPTFLAGS+= -fno-strict-aliasing .endif .if !defined(NO_CPU_COPTFLAGS) COPTFLAGS+= ${_CPUCFLAGS} .endif NOSTDINC= -nostdinc INCLUDES= ${NOSTDINC} ${INCLMAGIC} -I. -I$S -I$S/contrib/ck/include CFLAGS= ${COPTFLAGS} ${DEBUG} CFLAGS+= ${INCLUDES} -D_KERNEL -DHAVE_KERNEL_OPTION_HEADERS -include opt_global.h CFLAGS_PARAM_INLINE_UNIT_GROWTH?=100 CFLAGS_PARAM_LARGE_FUNCTION_GROWTH?=1000 .if ${MACHINE_CPUARCH} == "mips" CFLAGS_ARCH_PARAMS?=--param max-inline-insns-single=1000 -DMACHINE_ARCH='"${MACHINE_ARCH}"' .endif CFLAGS.gcc+= -fms-extensions -finline-limit=${INLINE_LIMIT} CFLAGS.gcc+= --param inline-unit-growth=${CFLAGS_PARAM_INLINE_UNIT_GROWTH} CFLAGS.gcc+= --param large-function-growth=${CFLAGS_PARAM_LARGE_FUNCTION_GROWTH} CFLAGS.gcc+= -fms-extensions .if defined(CFLAGS_ARCH_PARAMS) CFLAGS.gcc+=${CFLAGS_ARCH_PARAMS} .endif .if ${COMPILER_TYPE} == "gcc" && ${COMPILER_VERSION} < 50000 WERROR?= -Wno-error .else WERROR?= -Werror .endif # The following should be removed no earlier than LLVM11 being imported into the # tree, to ensure we don't regress the build. LLVM11 and GCC10 will switch the # default over to -fno-common, making this redundant. CFLAGS+= -fno-common # XXX LOCORE means "don't declare C stuff" not "for locore.s". ASM_CFLAGS= -x assembler-with-cpp -DLOCORE ${CFLAGS} ${ASM_CFLAGS.${.IMPSRC:T}} .if defined(PROFLEVEL) && ${PROFLEVEL} >= 1 CFLAGS+= -DGPROF CFLAGS.gcc+= -falign-functions=16 .if ${PROFLEVEL} >= 2 CFLAGS+= -DGPROF4 -DGUPROF PROF= -pg .if ${COMPILER_TYPE} == "gcc" PROF+= -mprofiler-epilogue .endif .else PROF= -pg .endif .endif DEFINED_PROF= ${PROF} KCSAN_ENABLED!= grep KCSAN opt_global.h || true ; echo .if !empty(KCSAN_ENABLED) SAN_CFLAGS+= -fsanitize=thread .endif KUBSAN_ENABLED!= grep KUBSAN opt_global.h || true ; echo .if !empty(KUBSAN_ENABLED) SAN_CFLAGS+= -fsanitize=undefined .endif COVERAGE_ENABLED!= grep COVERAGE opt_global.h || true ; echo .if !empty(COVERAGE_ENABLED) .if ${COMPILER_TYPE} == "clang" || \ (${COMPILER_TYPE} == "gcc" && ${COMPILER_VERSION} >= 80100) SAN_CFLAGS+= -fsanitize-coverage=trace-pc,trace-cmp .else SAN_CFLAGS+= -fsanitize-coverage=trace-pc .endif .endif CFLAGS+= ${SAN_CFLAGS} GCOV_ENABLED!= grep GCOV opt_global.h || true ; echo .if !empty(GCOV_ENABLED) .if ${COMPILER_TYPE} == "gcc" GCOV_CFLAGS+= -fprofile-arcs -ftest-coverage .endif .endif CFLAGS+= ${GCOV_CFLAGS} # Put configuration-specific C flags last (except for ${PROF}) so that they # can override the others. CFLAGS+= ${CONF_CFLAGS} .if defined(LINKER_FEATURES) && ${LINKER_FEATURES:Mbuild-id} LDFLAGS+= --build-id=sha1 .endif .if (${MACHINE_CPUARCH} == "aarch64" || ${MACHINE_CPUARCH} == "amd64" || \ - ${MACHINE_CPUARCH} == "i386") && \ + ${MACHINE_CPUARCH} == "i386" || ${MACHINE} == "powerpc") && \ defined(LINKER_FEATURES) && ${LINKER_FEATURES:Mifunc} == "" -.error amd64/arm64/i386 kernel requires linker ifunc support +.error amd64/arm64/i386/ppc* kernel requires linker ifunc support .endif .if ${MACHINE_CPUARCH} == "amd64" LDFLAGS+= -z max-page-size=2097152 .if ${LINKER_TYPE} != "lld" LDFLAGS+= -z common-page-size=4096 .else LDFLAGS+= -z notext -z ifunc-noplt .endif .endif .if ${MACHINE_CPUARCH} == "riscv" # Hack: Work around undefined weak symbols being out of range when linking with # LLD (address is a PC-relative calculation, and BFD works around this by # rewriting the instructions to generate an absolute address of 0); -fPIE # avoids this since it uses the GOT for all extern symbols, which is overly # inefficient for us. Drop once undefined weak symbols work with medany. .if ${LINKER_TYPE} == "lld" CFLAGS+= -fPIE .endif .endif NORMAL_C= ${CC} -c ${CFLAGS} ${WERROR} ${PROF} ${.IMPSRC} NORMAL_S= ${CC:N${CCACHE_BIN}} -c ${ASM_CFLAGS} ${WERROR} ${.IMPSRC} PROFILE_C= ${CC} -c ${CFLAGS} ${WERROR} ${.IMPSRC} NORMAL_C_NOWERROR= ${CC} -c ${CFLAGS} ${PROF} ${.IMPSRC} NORMAL_M= ${AWK} -f $S/tools/makeobjops.awk ${.IMPSRC} -c ; \ ${CC} -c ${CFLAGS} ${WERROR} ${PROF} ${.PREFIX}.c NORMAL_FW= uudecode -o ${.TARGET} ${.ALLSRC} NORMAL_FWO= ${LD} -b binary --no-warn-mismatch -d -warn-common -r \ -m ${LD_EMULATION} -o ${.TARGET} ${.ALLSRC:M*.fw} # for ZSTD in the kernel (include zstd/lib/freebsd before other CFLAGS) ZSTD_C= ${CC} -c -DZSTD_HEAPMODE=1 -I$S/contrib/zstd/lib/freebsd ${CFLAGS} -I$S/contrib/zstd/lib -I$S/contrib/zstd/lib/common ${WERROR} -Wno-inline -Wno-missing-prototypes ${PROF} -U__BMI__ ${.IMPSRC} # https://github.com/facebook/zstd/commit/812e8f2a [zstd 1.4.1] # "Note that [GCC] autovectorization still does not do a good job on the # optimized version, so it's turned off via attribute and flag. I found # that neither attribute nor command-line flag were entirely successful in # turning off vectorization, which is why there were both." .if ${COMPILER_TYPE} == "gcc" ZSTD_DECOMPRESS_BLOCK_FLAGS= -fno-tree-vectorize .endif # Common for dtrace / zfs CDDL_CFLAGS= -DFREEBSD_NAMECACHE -nostdinc -I$S/cddl/compat/opensolaris -I$S/cddl/contrib/opensolaris/uts/common -I$S -I$S/cddl/contrib/opensolaris/common ${CFLAGS} -Wno-unknown-pragmas -Wno-missing-prototypes -Wno-undef -Wno-strict-prototypes -Wno-cast-qual -Wno-parentheses -Wno-redundant-decls -Wno-missing-braces -Wno-uninitialized -Wno-unused -Wno-inline -Wno-switch -Wno-pointer-arith -Wno-unknown-pragmas CDDL_CFLAGS+= -include $S/cddl/compat/opensolaris/sys/debug_compat.h CDDL_C= ${CC} -c ${CDDL_CFLAGS} ${WERROR} ${PROF} ${.IMPSRC} # Special flags for managing the compat compiles for ZFS ZFS_CFLAGS= -DBUILDING_ZFS -I$S/cddl/contrib/opensolaris/uts/common/fs/zfs ZFS_CFLAGS+= -I$S/cddl/contrib/opensolaris/uts/common/fs/zfs/lua ZFS_CFLAGS+= -I$S/cddl/contrib/opensolaris/uts/common/zmod ZFS_CFLAGS+= -I$S/cddl/contrib/opensolaris/common/lz4 ZFS_CFLAGS+= -I$S/cddl/contrib/opensolaris/common/zfs ZFS_CFLAGS+= ${CDDL_CFLAGS} ZFS_ASM_CFLAGS= -x assembler-with-cpp -DLOCORE ${ZFS_CFLAGS} ZFS_C= ${CC} -c ${ZFS_CFLAGS} ${WERROR} ${PROF} ${.IMPSRC} ZFS_S= ${CC} -c ${ZFS_ASM_CFLAGS} ${WERROR} ${.IMPSRC} # Special flags for managing the compat compiles for DTrace DTRACE_CFLAGS= -DBUILDING_DTRACE ${CDDL_CFLAGS} -I$S/cddl/dev/dtrace -I$S/cddl/dev/dtrace/${MACHINE_CPUARCH} .if ${MACHINE_CPUARCH} == "amd64" || ${MACHINE_CPUARCH} == "i386" DTRACE_CFLAGS+= -I$S/cddl/contrib/opensolaris/uts/intel -I$S/cddl/dev/dtrace/x86 .endif DTRACE_CFLAGS+= -I$S/cddl/contrib/opensolaris/common/util -I$S -DDIS_MEM -DSMP DTRACE_ASM_CFLAGS= -x assembler-with-cpp -DLOCORE ${DTRACE_CFLAGS} DTRACE_C= ${CC} -c ${DTRACE_CFLAGS} ${WERROR} ${PROF} ${.IMPSRC} DTRACE_S= ${CC} -c ${DTRACE_ASM_CFLAGS} ${WERROR} ${.IMPSRC} # Special flags for managing the compat compiles for DTrace/FBT FBT_CFLAGS= -DBUILDING_DTRACE -nostdinc -I$S/cddl/dev/fbt/${MACHINE_CPUARCH} -I$S/cddl/dev/fbt -I$S/cddl/compat/opensolaris -I$S/cddl/contrib/opensolaris/uts/common -I$S ${CDDL_CFLAGS} .if ${MACHINE_CPUARCH} == "amd64" || ${MACHINE_CPUARCH} == "i386" FBT_CFLAGS+= -I$S/cddl/dev/fbt/x86 .endif FBT_C= ${CC} -c ${FBT_CFLAGS} ${WERROR} ${PROF} ${.IMPSRC} .if ${MK_CTF} != "no" NORMAL_CTFCONVERT= ${CTFCONVERT} ${CTFFLAGS} ${.TARGET} .elif ${MAKE_VERSION} >= 5201111300 NORMAL_CTFCONVERT= .else NORMAL_CTFCONVERT= @: .endif # Linux Kernel Programming Interface C-flags LINUXKPI_INCLUDES= -I$S/compat/linuxkpi/common/include LINUXKPI_C= ${NORMAL_C} ${LINUXKPI_INCLUDES} # Infiniband C flags. Correct include paths and omit errors that linux # does not honor. OFEDINCLUDES= -I$S/ofed/include -I$S/ofed/include/uapi ${LINUXKPI_INCLUDES} OFEDNOERR= -Wno-cast-qual -Wno-pointer-arith OFEDCFLAGS= ${CFLAGS:N-I*} -DCONFIG_INFINIBAND_USER_MEM \ ${OFEDINCLUDES} ${CFLAGS:M-I*} ${OFEDNOERR} OFED_C_NOIMP= ${CC} -c -o ${.TARGET} ${OFEDCFLAGS} ${WERROR} ${PROF} OFED_C= ${OFED_C_NOIMP} ${.IMPSRC} # mlxfw C flags. MLXFW_C= ${OFED_C_NOIMP} \ -I${SRCTOP}/sys/contrib/xz-embedded/freebsd \ -I${SRCTOP}/sys/contrib/xz-embedded/linux/lib/xz \ ${.IMPSRC} GEN_CFILES= $S/$M/$M/genassym.c ${MFILES:T:S/.m$/.c/} SYSTEM_CFILES= config.c env.c hints.c vnode_if.c SYSTEM_DEP= Makefile ${SYSTEM_OBJS} SYSTEM_OBJS= locore.o ${MDOBJS} ${OBJS} SYSTEM_OBJS+= ${SYSTEM_CFILES:.c=.o} SYSTEM_OBJS+= hack.pico KEYMAP=kbdcontrol -P ${SRCTOP}/share/vt/keymaps -P ${SRCTOP}/share/syscons/keymaps KEYMAP_FIX=sed -e 's/^static keymap_t.* = /static keymap_t key_map = /' -e 's/^static accentmap_t.* = /static accentmap_t accent_map = /' MD_ROOT_SIZE_CONFIGURED!= grep MD_ROOT_SIZE opt_md.h || true ; echo .if ${MFS_IMAGE:Uno} != "no" .if empty(MD_ROOT_SIZE_CONFIGURED) SYSTEM_OBJS+= embedfs_${MFS_IMAGE:T:R}.o .endif .endif SYSTEM_LD_BASECMD= \ ${LD} -m ${LD_EMULATION} -Bdynamic -T ${LDSCRIPT} ${_LDFLAGS} \ --no-warn-mismatch --warn-common --export-dynamic \ --dynamic-linker /red/herring -X SYSTEM_LD= @${SYSTEM_LD_BASECMD} -o ${.TARGET} ${SYSTEM_OBJS} vers.o SYSTEM_LD_TAIL= @${OBJCOPY} --strip-symbol gcc2_compiled. ${.TARGET} ; \ ${SIZE} ${.TARGET} ; chmod 755 ${.TARGET} SYSTEM_DEP+= ${LDSCRIPT} # Calculate path for .m files early, if needed. .if !defined(NO_MODULES) && !defined(__MPATH) && !make(install) && \ (empty(.MAKEFLAGS:M-V) || defined(NO_SKIP_MPATH)) __MPATH!=find ${S:tA}/ -name \*_if.m .endif # MKMODULESENV is set here so that port makefiles can augment # them. MKMODULESENV+= MAKEOBJDIRPREFIX=${.OBJDIR}/modules KMODDIR=${KODIR} MKMODULESENV+= MACHINE_CPUARCH=${MACHINE_CPUARCH} MKMODULESENV+= MACHINE=${MACHINE} MACHINE_ARCH=${MACHINE_ARCH} MKMODULESENV+= MODULES_EXTRA="${MODULES_EXTRA}" WITHOUT_MODULES="${WITHOUT_MODULES}" MKMODULESENV+= ARCH_FLAGS="${ARCH_FLAGS}" .if (${KERN_IDENT} == LINT) MKMODULESENV+= ALL_MODULES=LINT .endif .if defined(MODULES_OVERRIDE) MKMODULESENV+= MODULES_OVERRIDE="${MODULES_OVERRIDE}" .endif .if defined(DEBUG) MKMODULESENV+= DEBUG_FLAGS="${DEBUG}" .endif .if !defined(NO_MODULES) MKMODULESENV+= __MPATH="${__MPATH}" .endif # Architecture and output format arguments for objcopy to convert image to # object file .if ${MFS_IMAGE:Uno} != "no" .if empty(MD_ROOT_SIZE_CONFIGURED) .if !defined(EMBEDFS_FORMAT.${MACHINE_ARCH}) EMBEDFS_FORMAT.${MACHINE_ARCH}!= awk -F'"' '/OUTPUT_FORMAT/ {print $$2}' ${LDSCRIPT} .if empty(EMBEDFS_FORMAT.${MACHINE_ARCH}) .undef EMBEDFS_FORMAT.${MACHINE_ARCH} .endif .endif .if !defined(EMBEDFS_ARCH.${MACHINE_ARCH}) EMBEDFS_ARCH.${MACHINE_ARCH}!= sed -n '/OUTPUT_ARCH/s/.*(\(.*\)).*/\1/p' ${LDSCRIPT} .if empty(EMBEDFS_ARCH.${MACHINE_ARCH}) .undef EMBEDFS_ARCH.${MACHINE_ARCH} .endif .endif EMBEDFS_FORMAT.arm?= elf32-littlearm EMBEDFS_FORMAT.armv6?= elf32-littlearm EMBEDFS_FORMAT.armv7?= elf32-littlearm EMBEDFS_FORMAT.aarch64?= elf64-littleaarch64 EMBEDFS_FORMAT.mips?= elf32-tradbigmips EMBEDFS_FORMAT.mipsel?= elf32-tradlittlemips EMBEDFS_FORMAT.mips64?= elf64-tradbigmips EMBEDFS_FORMAT.mips64el?= elf64-tradlittlemips EMBEDFS_FORMAT.riscv64?= elf64-littleriscv .endif .endif # Detect kernel config options that force stack frames to be turned on. DDB_ENABLED!= grep DDB opt_ddb.h || true ; echo DTR_ENABLED!= grep KDTRACE_FRAME opt_kdtrace.h || true ; echo HWPMC_ENABLED!= grep HWPMC opt_hwpmc_hooks.h || true ; echo Index: head/sys/conf/ldscript.powerpc =================================================================== --- head/sys/conf/ldscript.powerpc (revision 360793) +++ head/sys/conf/ldscript.powerpc (revision 360794) @@ -1,150 +1,155 @@ /* $FreeBSD$ */ OUTPUT_FORMAT("elf32-powerpc-freebsd", "elf32-powerpc-freebsd", "elf32-powerpc-freebsd") OUTPUT_ARCH(powerpc) ENTRY(__start) SEARCH_DIR(/usr/lib); PROVIDE (__stack = 0); +PHDRS +{ + kernel PT_LOAD; + dynamic PT_DYNAMIC; +} SECTIONS { /* Read-only sections, merged into text segment: */ . = kernbase + SIZEOF_HEADERS; PROVIDE (begin = . - SIZEOF_HEADERS); .text : { *(.glink) *(.text) *(.stub) /* .gnu.warning sections are handled specially by elf32.em. */ *(.gnu.warning) *(.gnu.linkonce.t*) - } =0 + } :kernel =0 _etext = .; PROVIDE (etext = .); .interp : { *(.interp) } .hash : { *(.hash) } .dynsym : { *(.dynsym) } .dynstr : { *(.dynstr) } .gnu.version : { *(.gnu.version) } .gnu.version_d : { *(.gnu.version_d) } .gnu.version_r : { *(.gnu.version_r) } .rela.text : { *(.rela.text) *(.rela.gnu.linkonce.t*) } .rela.data : { *(.rela.data) *(.rela.gnu.linkonce.d*) } .rela.rodata : { *(.rela.rodata) *(.rela.gnu.linkonce.r*) } .rela.got : { *(.rela.got) } .rela.got1 : { *(.rela.got1) } .rela.got2 : { *(.rela.got2) } .rela.ctors : { *(.rela.ctors) } .rela.dtors : { *(.rela.dtors) } .rela.init : { *(.rela.init) } .rela.fini : { *(.rela.fini) } .rela.bss : { *(.rela.bss) } .rela.plt : { *(.rela.plt) } .rela.sdata : { *(.rela.sdata) } .rela.sbss : { *(.rela.sbss) } .rela.sdata2 : { *(.rela.sdata2) } .rela.sbss2 : { *(.rela.sbss2) } .init : { *(.init) } =0 .fini : { *(.fini) } =0 .rodata : { *(.rodata) *(.gnu.linkonce.r*) } .rodata1 : { *(.rodata1) } .note.gnu.build-id : { PROVIDE (__build_id_start = .); *(.note.gnu.build-id) PROVIDE (__build_id_end = .); } .sdata2 : { *(.sdata2) } .sbss2 : { *(.sbss2) } /* Adjust the address for the data segment to the next page up. */ . = ((. + 0x1000) & ~(0x1000 - 1)); .data : { *(.data) *(.gnu.linkonce.d*) } .data1 : { *(.data1) } .got1 : { *(.got1) } . = ALIGN(4096); .got : { *(.got) } .got.plt : { *(.got.plt) } - .dynamic : { *(.dynamic) } + .dynamic : { *(.dynamic) } :kernel :dynamic /* Put .ctors and .dtors next to the .got2 section, so that the pointers get relocated with -mrelocatable. Also put in the .fixup pointers. The current compiler no longer needs this, but keep it around for 2.7.2 */ PROVIDE (_GOT2_START_ = .); .got2 : { *(.got2) } PROVIDE (__CTOR_LIST__ = .); .ctors : { *(.ctors) } PROVIDE (__CTOR_END__ = .); PROVIDE (__DTOR_LIST__ = .); .dtors : { *(.dtors) } PROVIDE (__DTOR_END__ = .); PROVIDE (_FIXUP_START_ = .); .fixup : { *(.fixup) } PROVIDE (_FIXUP_END_ = .); PROVIDE (_GOT2_END_ = .); /* We want the small data sections together, so single-instruction offsets can access them all, and initialized data all before uninitialized, so we can shorten the on-disk segment size. */ - .sdata : { *(.sdata) } + .sdata : { *(.sdata) } :kernel _edata = .; PROVIDE (edata = .); .sbss : { PROVIDE (__sbss_start = .); *(.sbss) *(.scommon) *(.dynsbss) PROVIDE (__sbss_end = .); } .plt : { *(.plt) } .bss : { PROVIDE (__bss_start = .); *(.dynbss) *(.bss) *(COMMON) } _end = . ; PROVIDE (end = .); /* Stabs debugging sections. */ .stab 0 : { *(.stab) } .stabstr 0 : { *(.stabstr) } /* DWARF debug sections. Symbols in the DWARF debugging sections are relative to the beginning of the section so we begin them at 0. */ /* DWARF 1 */ .debug 0 : { *(.debug) } .line 0 : { *(.line) } /* GNU DWARF 1 extensions */ .debug_srcinfo 0 : { *(.debug_srcinfo) } .debug_sfnames 0 : { *(.debug_sfnames) } /* DWARF 1.1 and DWARF 2 */ .debug_aranges 0 : { *(.debug_aranges) } .debug_pubnames 0 : { *(.debug_pubnames) } /* DWARF 2 */ .debug_info 0 : { *(.debug_info) } .debug_abbrev 0 : { *(.debug_abbrev) } .debug_line 0 : { *(.debug_line) } .debug_frame 0 : { *(.debug_frame) } .debug_str 0 : { *(.debug_str) } .debug_loc 0 : { *(.debug_loc) } .debug_macinfo 0 : { *(.debug_macinfo) } /* SGI/MIPS DWARF 2 extensions */ .debug_weaknames 0 : { *(.debug_weaknames) } .debug_funcnames 0 : { *(.debug_funcnames) } .debug_typenames 0 : { *(.debug_typenames) } .debug_varnames 0 : { *(.debug_varnames) } /* These must appear regardless of . */ } Index: head/sys/conf/ldscript.powerpc64 =================================================================== --- head/sys/conf/ldscript.powerpc64 (revision 360793) +++ head/sys/conf/ldscript.powerpc64 (revision 360794) @@ -1,179 +1,179 @@ /* $FreeBSD$ */ OUTPUT_FORMAT("elf64-powerpc-freebsd", "elf64-powerpc-freebsd", "elf64-powerpc-freebsd") OUTPUT_ARCH(powerpc:common64) ENTRY(__start) SEARCH_DIR(/usr/lib); PROVIDE (__stack = 0); PHDRS { - text PT_LOAD ; - dynamic PT_DYNAMIC ; + kernel PT_LOAD; + dynamic PT_DYNAMIC; } SECTIONS { /* Low-address wrapper for bootloaders (kexec/kboot) that can't parse ELF */ . = kernbase - 0x100; - .kboot : { *(.text.kboot) } :text + .kboot : { *(.text.kboot) } :kernel /* Read-only sections, merged into text segment: */ . = kernbase; PROVIDE (begin = .); .text : { *(.glink) *(.text) *(.stub) /* .gnu.warning sections are handled specially by elf32.em. */ *(.gnu.warning) *(.gnu.linkonce.t*) } =0 _etext = .; PROVIDE (etext = .); /* Do not emit PT_INTERP section, which confuses some loaders (kexec-lite) */ /DISCARD/ : { *(.interp) } /* Also delete notes */ /DISCARD/ : { *(.note.*) } .hash : { *(.hash) } .dynsym : { *(.dynsym) } .dynstr : { *(.dynstr) } .gnu.version : { *(.gnu.version) } .gnu.version_d : { *(.gnu.version_d) } .gnu.version_r : { *(.gnu.version_r) } .note.gnu.build-id : { PROVIDE (__build_id_start = .); *(.note.gnu.build-id) PROVIDE (__build_id_end = .); } .rela.text : { *(.rela.text) *(.rela.gnu.linkonce.t*) } .rela.data : { *(.rela.data) *(.rela.gnu.linkonce.d*) } .rela.rodata : { *(.rela.rodata) *(.rela.gnu.linkonce.r*) } .rela.got : { *(.rela.got) } .rela.got1 : { *(.rela.got1) } .rela.got2 : { *(.rela.got2) } .rela.ctors : { *(.rela.ctors) } .rela.dtors : { *(.rela.dtors) } .rela.init : { *(.rela.init) } .rela.fini : { *(.rela.fini) } .rela.bss : { *(.rela.bss) } .rela.plt : { *(.rela.plt) } .rela.sdata : { *(.rela.sdata) } .rela.sbss : { *(.rela.sbss) } .rela.sdata2 : { *(.rela.sdata2) } .rela.sbss2 : { *(.rela.sbss2) } .init : { *(.init) } =0 .fini : { *(.fini) } =0 .rodata : { *(.rodata) *(.gnu.linkonce.r*) } .rodata1 : { *(.rodata1) } .sdata2 : { *(.sdata2) } .sbss2 : { *(.sbss2) } /* Adjust the address for the data segment to the next page up. */ . = ALIGN(4096); .data.read_frequently : { *(SORT_BY_ALIGNMENT(.data.read_frequently)) } .data.read_mostly : { *(.data.read_mostly) } . = ALIGN(128); .data.exclusive_cache_line : { *(.data.exclusive_cache_line) } . = ALIGN(128); .data : { *(.data) *(.gnu.linkonce.d*) } .data1 : { *(.data1) } .toc1 : ALIGN(8) { *(.toc1) } .opd : ALIGN(8) { KEEP (*(.opd)) } .branch_lt : ALIGN(8) { *(.branch_lt) } . = ALIGN(4096); .got : ALIGN(8) { __tocbase = .; *(.got) } .toc : ALIGN(8) { *(.toc) } - .dynamic : { *(.dynamic) } :text :dynamic + .dynamic : { *(.dynamic) } :kernel :dynamic /* Put .ctors and .dtors next to the .got2 section, so that the pointers get relocated with -mrelocatable. Also put in the .fixup pointers. The current compiler no longer needs this, but keep it around for 2.7.2 */ PROVIDE (_GOT2_START_ = .); .got2 : { *(.got2) } PROVIDE (__CTOR_LIST__ = .); .ctors : { *(.ctors) } PROVIDE (__CTOR_END__ = .); PROVIDE (__DTOR_LIST__ = .); .dtors : { *(.dtors) } PROVIDE (__DTOR_END__ = .); PROVIDE (_FIXUP_START_ = .); .fixup : { *(.fixup) } PROVIDE (_FIXUP_END_ = .); PROVIDE (_GOT2_END_ = .); /* We want the small data sections together, so single-instruction offsets can access them all, and initialized data all before uninitialized, so we can shorten the on-disk segment size. */ - .sdata : { *(.sdata) } + .sdata : { *(.sdata) } :kernel _edata = .; PROVIDE (edata = .); .sbss : { PROVIDE (__sbss_start = .); *(.sbss) *(.scommon) *(.dynsbss) PROVIDE (__sbss_end = .); } .plt : { *(.plt) } .bss : { PROVIDE (__bss_start = .); *(.dynbss) *(.bss) *(COMMON) } _end = . ; PROVIDE (end = .); /* Stabs debugging sections. */ .stab 0 : { *(.stab) } .stabstr 0 : { *(.stabstr) } /* DWARF debug sections. Symbols in the DWARF debugging sections are relative to the beginning of the section so we begin them at 0. */ /* DWARF 1 */ .debug 0 : { *(.debug) } .line 0 : { *(.line) } /* GNU DWARF 1 extensions */ .debug_srcinfo 0 : { *(.debug_srcinfo) } .debug_sfnames 0 : { *(.debug_sfnames) } /* DWARF 1.1 and DWARF 2 */ .debug_aranges 0 : { *(.debug_aranges) } .debug_pubnames 0 : { *(.debug_pubnames) } /* DWARF 2 */ .debug_info 0 : { *(.debug_info) } .debug_abbrev 0 : { *(.debug_abbrev) } .debug_line 0 : { *(.debug_line) } .debug_frame 0 : { *(.debug_frame) } .debug_str 0 : { *(.debug_str) } .debug_loc 0 : { *(.debug_loc) } .debug_macinfo 0 : { *(.debug_macinfo) } /* SGI/MIPS DWARF 2 extensions */ .debug_weaknames 0 : { *(.debug_weaknames) } .debug_funcnames 0 : { *(.debug_funcnames) } .debug_typenames 0 : { *(.debug_typenames) } .debug_varnames 0 : { *(.debug_varnames) } /* These must appear regardless of . */ } Index: head/sys/conf/ldscript.powerpcspe =================================================================== --- head/sys/conf/ldscript.powerpcspe (revision 360793) +++ head/sys/conf/ldscript.powerpcspe (revision 360794) @@ -1,151 +1,156 @@ /* $FreeBSD$ */ OUTPUT_FORMAT("elf32-powerpc-freebsd", "elf32-powerpc-freebsd", "elf32-powerpc-freebsd") OUTPUT_ARCH(powerpc) ENTRY(__start) SEARCH_DIR(/usr/lib); PROVIDE (__stack = 0); +PHDRS +{ + kernel PT_LOAD; + dynamic PT_DYNAMIC; +} SECTIONS { /* Read-only sections, merged into text segment: */ . = kernbase + SIZEOF_HEADERS; PROVIDE (begin = . - SIZEOF_HEADERS); .text : { *(.glink) *(.text) *(.stub) /* .gnu.warning sections are handled specially by elf32.em. */ *(.gnu.warning) *(.gnu.linkonce.t*) - } =0 + } :kernel =0 _etext = .; PROVIDE (etext = .); .interp : { *(.interp) } .hash : { *(.hash) } .dynsym : { *(.dynsym) } .dynstr : { *(.dynstr) } .gnu.version : { *(.gnu.version) } .gnu.version_d : { *(.gnu.version_d) } .gnu.version_r : { *(.gnu.version_r) } .rela.text : { *(.rela.text) *(.rela.gnu.linkonce.t*) } .rela.data : { *(.rela.data) *(.rela.gnu.linkonce.d*) } .rela.rodata : { *(.rela.rodata) *(.rela.gnu.linkonce.r*) } .rela.got : { *(.rela.got) } .rela.got1 : { *(.rela.got1) } .rela.got2 : { *(.rela.got2) } .rela.ctors : { *(.rela.ctors) } .rela.dtors : { *(.rela.dtors) } .rela.init : { *(.rela.init) } .rela.fini : { *(.rela.fini) } .rela.bss : { *(.rela.bss) } .rela.plt : { *(.rela.plt) } .rela.sdata : { *(.rela.sdata) } .rela.sbss : { *(.rela.sbss) } .rela.sdata2 : { *(.rela.sdata2) } .rela.sbss2 : { *(.rela.sbss2) } .init : { *(.init) } =0 .fini : { *(.fini) } =0 .rodata : { *(.rodata) *(.gnu.linkonce.r*) } .rodata1 : { *(.rodata1) } .note.gnu.build-id : { PROVIDE (__build_id_start = .); *(.note.gnu.build-id) PROVIDE (__build_id_end = .); } .sdata2 : { *(.sdata2) } .sbss2 : { *(.sbss2) } /* Adjust the address for the data segment to the next page up. */ . = ((. + 0x1000) & ~(0x1000 - 1)); .data : { *(.data) *(.gnu.linkonce.d*) CONSTRUCTORS } .data1 : { *(.data1) } .got1 : { *(.got1) } . = ALIGN(4096); .got : { *(.got) } .got.plt : { *(.got.plt) } - .dynamic : { *(.dynamic) } + .dynamic : { *(.dynamic) } :kernel :dynamic /* Put .ctors and .dtors next to the .got2 section, so that the pointers get relocated with -mrelocatable. Also put in the .fixup pointers. The current compiler no longer needs this, but keep it around for 2.7.2 */ PROVIDE (_GOT2_START_ = .); .got2 : { *(.got2) } PROVIDE (__CTOR_LIST__ = .); .ctors : { *(.ctors) } PROVIDE (__CTOR_END__ = .); PROVIDE (__DTOR_LIST__ = .); .dtors : { *(.dtors) } PROVIDE (__DTOR_END__ = .); PROVIDE (_FIXUP_START_ = .); .fixup : { *(.fixup) } PROVIDE (_FIXUP_END_ = .); PROVIDE (_GOT2_END_ = .); /* We want the small data sections together, so single-instruction offsets can access them all, and initialized data all before uninitialized, so we can shorten the on-disk segment size. */ - .sdata : { *(.sdata) } + .sdata : { *(.sdata) } :kernel _edata = .; PROVIDE (edata = .); .sbss : { PROVIDE (__sbss_start = .); *(.sbss) *(.scommon) *(.dynsbss) PROVIDE (__sbss_end = .); } .plt : { *(.plt) } .bss : { PROVIDE (__bss_start = .); *(.dynbss) *(.bss) *(COMMON) } _end = . ; PROVIDE (end = .); /* Stabs debugging sections. */ .stab 0 : { *(.stab) } .stabstr 0 : { *(.stabstr) } /* DWARF debug sections. Symbols in the DWARF debugging sections are relative to the beginning of the section so we begin them at 0. */ /* DWARF 1 */ .debug 0 : { *(.debug) } .line 0 : { *(.line) } /* GNU DWARF 1 extensions */ .debug_srcinfo 0 : { *(.debug_srcinfo) } .debug_sfnames 0 : { *(.debug_sfnames) } /* DWARF 1.1 and DWARF 2 */ .debug_aranges 0 : { *(.debug_aranges) } .debug_pubnames 0 : { *(.debug_pubnames) } /* DWARF 2 */ .debug_info 0 : { *(.debug_info) } .debug_abbrev 0 : { *(.debug_abbrev) } .debug_line 0 : { *(.debug_line) } .debug_frame 0 : { *(.debug_frame) } .debug_str 0 : { *(.debug_str) } .debug_loc 0 : { *(.debug_loc) } .debug_macinfo 0 : { *(.debug_macinfo) } /* SGI/MIPS DWARF 2 extensions */ .debug_weaknames 0 : { *(.debug_weaknames) } .debug_funcnames 0 : { *(.debug_funcnames) } .debug_typenames 0 : { *(.debug_typenames) } .debug_varnames 0 : { *(.debug_varnames) } /* These must appear regardless of . */ } Index: head/sys/kern/link_elf.c =================================================================== --- head/sys/kern/link_elf.c (revision 360793) +++ head/sys/kern/link_elf.c (revision 360794) @@ -1,1904 +1,1912 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 1998-2000 Doug Rabson * 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. */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include "opt_gdb.h" #include #include #ifdef GPROF #include #endif #include #include #include #ifdef SPARSE_MAPPING #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef SPARSE_MAPPING #include #include #include #endif #include #include #include #include "linker_if.h" #define MAXSEGS 4 typedef struct elf_file { struct linker_file lf; /* Common fields */ int preloaded; /* Was file pre-loaded */ caddr_t address; /* Relocation address */ #ifdef SPARSE_MAPPING vm_object_t object; /* VM object to hold file pages */ #endif Elf_Dyn *dynamic; /* Symbol table etc. */ Elf_Hashelt nbuckets; /* DT_HASH info */ Elf_Hashelt nchains; const Elf_Hashelt *buckets; const Elf_Hashelt *chains; caddr_t hash; caddr_t strtab; /* DT_STRTAB */ int strsz; /* DT_STRSZ */ const Elf_Sym *symtab; /* DT_SYMTAB */ Elf_Addr *got; /* DT_PLTGOT */ const Elf_Rel *pltrel; /* DT_JMPREL */ int pltrelsize; /* DT_PLTRELSZ */ const Elf_Rela *pltrela; /* DT_JMPREL */ int pltrelasize; /* DT_PLTRELSZ */ const Elf_Rel *rel; /* DT_REL */ int relsize; /* DT_RELSZ */ const Elf_Rela *rela; /* DT_RELA */ int relasize; /* DT_RELASZ */ caddr_t modptr; const Elf_Sym *ddbsymtab; /* The symbol table we are using */ long ddbsymcnt; /* Number of symbols */ caddr_t ddbstrtab; /* String table */ long ddbstrcnt; /* number of bytes in string table */ caddr_t symbase; /* malloc'ed symbold base */ caddr_t strbase; /* malloc'ed string base */ caddr_t ctftab; /* CTF table */ long ctfcnt; /* number of bytes in CTF table */ caddr_t ctfoff; /* CTF offset table */ caddr_t typoff; /* Type offset table */ long typlen; /* Number of type entries. */ Elf_Addr pcpu_start; /* Pre-relocation pcpu set start. */ Elf_Addr pcpu_stop; /* Pre-relocation pcpu set stop. */ Elf_Addr pcpu_base; /* Relocated pcpu set address. */ #ifdef VIMAGE Elf_Addr vnet_start; /* Pre-relocation vnet set start. */ Elf_Addr vnet_stop; /* Pre-relocation vnet set stop. */ Elf_Addr vnet_base; /* Relocated vnet set address. */ #endif #ifdef GDB struct link_map gdb; /* hooks for gdb */ #endif } *elf_file_t; struct elf_set { Elf_Addr es_start; Elf_Addr es_stop; Elf_Addr es_base; TAILQ_ENTRY(elf_set) es_link; }; TAILQ_HEAD(elf_set_head, elf_set); #include static int link_elf_link_common_finish(linker_file_t); static int link_elf_link_preload(linker_class_t cls, const char *, linker_file_t *); static int link_elf_link_preload_finish(linker_file_t); static int link_elf_load_file(linker_class_t, const char *, linker_file_t *); static int link_elf_lookup_symbol(linker_file_t, const char *, c_linker_sym_t *); static int link_elf_symbol_values(linker_file_t, c_linker_sym_t, linker_symval_t *); static int link_elf_search_symbol(linker_file_t, caddr_t, c_linker_sym_t *, long *); static void link_elf_unload_file(linker_file_t); static void link_elf_unload_preload(linker_file_t); static int link_elf_lookup_set(linker_file_t, const char *, void ***, void ***, int *); static int link_elf_each_function_name(linker_file_t, int (*)(const char *, void *), void *); static int link_elf_each_function_nameval(linker_file_t, linker_function_nameval_callback_t, void *); static void link_elf_reloc_local(linker_file_t); static long link_elf_symtab_get(linker_file_t, const Elf_Sym **); static long link_elf_strtab_get(linker_file_t, caddr_t *); static int elf_lookup(linker_file_t, Elf_Size, int, Elf_Addr *); static kobj_method_t link_elf_methods[] = { KOBJMETHOD(linker_lookup_symbol, link_elf_lookup_symbol), KOBJMETHOD(linker_symbol_values, link_elf_symbol_values), KOBJMETHOD(linker_search_symbol, link_elf_search_symbol), KOBJMETHOD(linker_unload, link_elf_unload_file), KOBJMETHOD(linker_load_file, link_elf_load_file), KOBJMETHOD(linker_link_preload, link_elf_link_preload), KOBJMETHOD(linker_link_preload_finish, link_elf_link_preload_finish), KOBJMETHOD(linker_lookup_set, link_elf_lookup_set), KOBJMETHOD(linker_each_function_name, link_elf_each_function_name), KOBJMETHOD(linker_each_function_nameval, link_elf_each_function_nameval), KOBJMETHOD(linker_ctf_get, link_elf_ctf_get), KOBJMETHOD(linker_symtab_get, link_elf_symtab_get), KOBJMETHOD(linker_strtab_get, link_elf_strtab_get), KOBJMETHOD_END }; static struct linker_class link_elf_class = { #if ELF_TARG_CLASS == ELFCLASS32 "elf32", #else "elf64", #endif link_elf_methods, sizeof(struct elf_file) }; typedef int (*elf_reloc_fn)(linker_file_t lf, Elf_Addr relocbase, const void *data, int type, elf_lookup_fn lookup); static int parse_dynamic(elf_file_t); static int relocate_file(elf_file_t); static int relocate_file1(elf_file_t ef, elf_lookup_fn lookup, elf_reloc_fn reloc, bool ifuncs); static int link_elf_preload_parse_symbols(elf_file_t); static struct elf_set_head set_pcpu_list; #ifdef VIMAGE static struct elf_set_head set_vnet_list; #endif static void elf_set_add(struct elf_set_head *list, Elf_Addr start, Elf_Addr stop, Elf_Addr base) { struct elf_set *set, *iter; set = malloc(sizeof(*set), M_LINKER, M_WAITOK); set->es_start = start; set->es_stop = stop; set->es_base = base; TAILQ_FOREACH(iter, list, es_link) { KASSERT((set->es_start < iter->es_start && set->es_stop < iter->es_stop) || (set->es_start > iter->es_start && set->es_stop > iter->es_stop), ("linker sets intersection: to insert: 0x%jx-0x%jx; inserted: 0x%jx-0x%jx", (uintmax_t)set->es_start, (uintmax_t)set->es_stop, (uintmax_t)iter->es_start, (uintmax_t)iter->es_stop)); if (iter->es_start > set->es_start) { TAILQ_INSERT_BEFORE(iter, set, es_link); break; } } if (iter == NULL) TAILQ_INSERT_TAIL(list, set, es_link); } static int elf_set_find(struct elf_set_head *list, Elf_Addr addr, Elf_Addr *start, Elf_Addr *base) { struct elf_set *set; TAILQ_FOREACH(set, list, es_link) { if (addr < set->es_start) return (0); if (addr < set->es_stop) { *start = set->es_start; *base = set->es_base; return (1); } } return (0); } static void elf_set_delete(struct elf_set_head *list, Elf_Addr start) { struct elf_set *set; TAILQ_FOREACH(set, list, es_link) { if (start < set->es_start) break; if (start == set->es_start) { TAILQ_REMOVE(list, set, es_link); free(set, M_LINKER); return; } } KASSERT(0, ("deleting unknown linker set (start = 0x%jx)", (uintmax_t)start)); } #ifdef GDB static void r_debug_state(struct r_debug *, struct link_map *); /* * A list of loaded modules for GDB to use for loading symbols. */ struct r_debug r_debug; #define GDB_STATE(s) do { \ r_debug.r_state = s; r_debug_state(NULL, NULL); \ } while (0) /* * Function for the debugger to set a breakpoint on to gain control. */ static void r_debug_state(struct r_debug *dummy_one __unused, struct link_map *dummy_two __unused) { } static void link_elf_add_gdb(struct link_map *l) { struct link_map *prev; l->l_next = NULL; if (r_debug.r_map == NULL) { /* Add first. */ l->l_prev = NULL; r_debug.r_map = l; } else { /* Append to list. */ for (prev = r_debug.r_map; prev->l_next != NULL; prev = prev->l_next) ; l->l_prev = prev; prev->l_next = l; } } static void link_elf_delete_gdb(struct link_map *l) { if (l->l_prev == NULL) { /* Remove first. */ if ((r_debug.r_map = l->l_next) != NULL) l->l_next->l_prev = NULL; } else { /* Remove any but first. */ if ((l->l_prev->l_next = l->l_next) != NULL) l->l_next->l_prev = l->l_prev; } } #endif /* GDB */ /* * The kernel symbol table starts here. */ extern struct _dynamic _DYNAMIC; static void link_elf_error(const char *filename, const char *s) { if (filename == NULL) printf("kldload: %s\n", s); else printf("kldload: %s: %s\n", filename, s); } static void link_elf_invoke_ctors(caddr_t addr, size_t size) { void (**ctor)(void); size_t i, cnt; if (addr == NULL || size == 0) return; cnt = size / sizeof(*ctor); ctor = (void *)addr; for (i = 0; i < cnt; i++) { if (ctor[i] != NULL) (*ctor[i])(); } } /* * Actions performed after linking/loading both the preloaded kernel and any * modules; whether preloaded or dynamicly loaded. */ static int link_elf_link_common_finish(linker_file_t lf) { #ifdef GDB elf_file_t ef = (elf_file_t)lf; char *newfilename; #endif int error; /* Notify MD code that a module is being loaded. */ error = elf_cpu_load_file(lf); if (error != 0) return (error); #ifdef GDB GDB_STATE(RT_ADD); ef->gdb.l_addr = lf->address; newfilename = malloc(strlen(lf->filename) + 1, M_LINKER, M_WAITOK); strcpy(newfilename, lf->filename); ef->gdb.l_name = newfilename; ef->gdb.l_ld = ef->dynamic; link_elf_add_gdb(&ef->gdb); GDB_STATE(RT_CONSISTENT); #endif /* Invoke .ctors */ link_elf_invoke_ctors(lf->ctors_addr, lf->ctors_size); return (0); } +#ifdef RELOCATABLE_KERNEL extern vm_offset_t __startkernel, __endkernel; +#endif static unsigned long kern_relbase = KERNBASE; SYSCTL_ULONG(_kern, OID_AUTO, base_address, CTLFLAG_RD, SYSCTL_NULL_ULONG_PTR, KERNBASE, "Kernel base address"); SYSCTL_ULONG(_kern, OID_AUTO, relbase_address, CTLFLAG_RD, &kern_relbase, 0, "Kernel relocated base address"); static void link_elf_init(void* arg) { Elf_Dyn *dp; Elf_Addr *ctors_addrp; Elf_Size *ctors_sizep; caddr_t modptr, baseptr, sizeptr; elf_file_t ef; const char *modname; linker_add_class(&link_elf_class); dp = (Elf_Dyn *)&_DYNAMIC; modname = NULL; modptr = preload_search_by_type("elf" __XSTRING(__ELF_WORD_SIZE) " kernel"); if (modptr == NULL) modptr = preload_search_by_type("elf kernel"); modname = (char *)preload_search_info(modptr, MODINFO_NAME); if (modname == NULL) modname = "kernel"; linker_kernel_file = linker_make_file(modname, &link_elf_class); if (linker_kernel_file == NULL) panic("%s: Can't create linker structures for kernel", __func__); ef = (elf_file_t) linker_kernel_file; ef->preloaded = 1; -#ifdef __powerpc__ +#ifdef RELOCATABLE_KERNEL ef->address = (caddr_t) (__startkernel - KERNBASE); #else ef->address = 0; #endif #ifdef SPARSE_MAPPING ef->object = NULL; #endif ef->dynamic = dp; if (dp != NULL) parse_dynamic(ef); -#ifdef __powerpc__ +#ifdef RELOCATABLE_KERNEL linker_kernel_file->address = (caddr_t)__startkernel; linker_kernel_file->size = (intptr_t)(__endkernel - __startkernel); kern_relbase = (unsigned long)__startkernel; #else linker_kernel_file->address += KERNBASE; linker_kernel_file->size = -(intptr_t)linker_kernel_file->address; #endif if (modptr != NULL) { ef->modptr = modptr; baseptr = preload_search_info(modptr, MODINFO_ADDR); if (baseptr != NULL) linker_kernel_file->address = *(caddr_t *)baseptr; sizeptr = preload_search_info(modptr, MODINFO_SIZE); if (sizeptr != NULL) linker_kernel_file->size = *(size_t *)sizeptr; ctors_addrp = (Elf_Addr *)preload_search_info(modptr, MODINFO_METADATA | MODINFOMD_CTORS_ADDR); ctors_sizep = (Elf_Size *)preload_search_info(modptr, MODINFO_METADATA | MODINFOMD_CTORS_SIZE); if (ctors_addrp != NULL && ctors_sizep != NULL) { linker_kernel_file->ctors_addr = ef->address + *ctors_addrp; linker_kernel_file->ctors_size = *ctors_sizep; } } (void)link_elf_preload_parse_symbols(ef); #ifdef GDB r_debug.r_map = NULL; r_debug.r_brk = r_debug_state; r_debug.r_state = RT_CONSISTENT; #endif (void)link_elf_link_common_finish(linker_kernel_file); linker_kernel_file->flags |= LINKER_FILE_LINKED; TAILQ_INIT(&set_pcpu_list); #ifdef VIMAGE TAILQ_INIT(&set_vnet_list); #endif } SYSINIT(link_elf, SI_SUB_KLD, SI_ORDER_THIRD, link_elf_init, NULL); static int link_elf_preload_parse_symbols(elf_file_t ef) { caddr_t pointer; caddr_t ssym, esym, base; caddr_t strtab; int strcnt; Elf_Sym *symtab; int symcnt; if (ef->modptr == NULL) return (0); pointer = preload_search_info(ef->modptr, MODINFO_METADATA | MODINFOMD_SSYM); if (pointer == NULL) return (0); ssym = *(caddr_t *)pointer; pointer = preload_search_info(ef->modptr, MODINFO_METADATA | MODINFOMD_ESYM); if (pointer == NULL) return (0); esym = *(caddr_t *)pointer; base = ssym; symcnt = *(long *)base; base += sizeof(long); symtab = (Elf_Sym *)base; base += roundup(symcnt, sizeof(long)); if (base > esym || base < ssym) { printf("Symbols are corrupt!\n"); return (EINVAL); } strcnt = *(long *)base; base += sizeof(long); strtab = base; base += roundup(strcnt, sizeof(long)); if (base > esym || base < ssym) { printf("Symbols are corrupt!\n"); return (EINVAL); } ef->ddbsymtab = symtab; ef->ddbsymcnt = symcnt / sizeof(Elf_Sym); ef->ddbstrtab = strtab; ef->ddbstrcnt = strcnt; return (0); } static int parse_dynamic(elf_file_t ef) { Elf_Dyn *dp; int plttype = DT_REL; for (dp = ef->dynamic; dp->d_tag != DT_NULL; dp++) { switch (dp->d_tag) { case DT_HASH: { /* From src/libexec/rtld-elf/rtld.c */ const Elf_Hashelt *hashtab = (const Elf_Hashelt *) (ef->address + dp->d_un.d_ptr); ef->nbuckets = hashtab[0]; ef->nchains = hashtab[1]; ef->buckets = hashtab + 2; ef->chains = ef->buckets + ef->nbuckets; break; } case DT_STRTAB: ef->strtab = (caddr_t) (ef->address + dp->d_un.d_ptr); break; case DT_STRSZ: ef->strsz = dp->d_un.d_val; break; case DT_SYMTAB: ef->symtab = (Elf_Sym*) (ef->address + dp->d_un.d_ptr); break; case DT_SYMENT: if (dp->d_un.d_val != sizeof(Elf_Sym)) return (ENOEXEC); break; case DT_PLTGOT: ef->got = (Elf_Addr *) (ef->address + dp->d_un.d_ptr); break; case DT_REL: ef->rel = (const Elf_Rel *) (ef->address + dp->d_un.d_ptr); break; case DT_RELSZ: ef->relsize = dp->d_un.d_val; break; case DT_RELENT: if (dp->d_un.d_val != sizeof(Elf_Rel)) return (ENOEXEC); break; case DT_JMPREL: ef->pltrel = (const Elf_Rel *) (ef->address + dp->d_un.d_ptr); break; case DT_PLTRELSZ: ef->pltrelsize = dp->d_un.d_val; break; case DT_RELA: ef->rela = (const Elf_Rela *) (ef->address + dp->d_un.d_ptr); break; case DT_RELASZ: ef->relasize = dp->d_un.d_val; break; case DT_RELAENT: if (dp->d_un.d_val != sizeof(Elf_Rela)) return (ENOEXEC); break; case DT_PLTREL: plttype = dp->d_un.d_val; if (plttype != DT_REL && plttype != DT_RELA) return (ENOEXEC); break; #ifdef GDB case DT_DEBUG: dp->d_un.d_ptr = (Elf_Addr)&r_debug; break; #endif } } if (plttype == DT_RELA) { ef->pltrela = (const Elf_Rela *)ef->pltrel; ef->pltrel = NULL; ef->pltrelasize = ef->pltrelsize; ef->pltrelsize = 0; } ef->ddbsymtab = ef->symtab; ef->ddbsymcnt = ef->nchains; ef->ddbstrtab = ef->strtab; ef->ddbstrcnt = ef->strsz; return elf_cpu_parse_dynamic(ef->address, ef->dynamic); } #define LS_PADDING 0x90909090 static int parse_dpcpu(elf_file_t ef) { int error, size; #if defined(__i386__) uint32_t pad; #endif ef->pcpu_start = 0; ef->pcpu_stop = 0; error = link_elf_lookup_set(&ef->lf, "pcpu", (void ***)&ef->pcpu_start, (void ***)&ef->pcpu_stop, NULL); /* Error just means there is no pcpu set to relocate. */ if (error != 0) return (0); size = (uintptr_t)ef->pcpu_stop - (uintptr_t)ef->pcpu_start; /* Empty set? */ if (size < 1) return (0); #if defined(__i386__) /* In case we do find __start/stop_set_ symbols double-check. */ if (size < 4) { uprintf("Kernel module '%s' must be recompiled with " "linker script\n", ef->lf.pathname); return (ENOEXEC); } /* Padding from linker-script correct? */ pad = *(uint32_t *)((uintptr_t)ef->pcpu_stop - sizeof(pad)); if (pad != LS_PADDING) { uprintf("Kernel module '%s' must be recompiled with " "linker script, invalid padding %#04x (%#04x)\n", ef->lf.pathname, pad, LS_PADDING); return (ENOEXEC); } /* If we only have valid padding, nothing to do. */ if (size == 4) return (0); #endif /* * Allocate space in the primary pcpu area. Copy in our * initialization from the data section and then initialize * all per-cpu storage from that. */ ef->pcpu_base = (Elf_Addr)(uintptr_t)dpcpu_alloc(size); if (ef->pcpu_base == 0) { printf("%s: pcpu module space is out of space; " "cannot allocate %d for %s\n", __func__, size, ef->lf.pathname); return (ENOSPC); } memcpy((void *)ef->pcpu_base, (void *)ef->pcpu_start, size); dpcpu_copy((void *)ef->pcpu_base, size); elf_set_add(&set_pcpu_list, ef->pcpu_start, ef->pcpu_stop, ef->pcpu_base); return (0); } #ifdef VIMAGE static int parse_vnet(elf_file_t ef) { int error, size; #if defined(__i386__) uint32_t pad; #endif ef->vnet_start = 0; ef->vnet_stop = 0; error = link_elf_lookup_set(&ef->lf, "vnet", (void ***)&ef->vnet_start, (void ***)&ef->vnet_stop, NULL); /* Error just means there is no vnet data set to relocate. */ if (error != 0) return (0); size = (uintptr_t)ef->vnet_stop - (uintptr_t)ef->vnet_start; /* Empty set? */ if (size < 1) return (0); #if defined(__i386__) /* In case we do find __start/stop_set_ symbols double-check. */ if (size < 4) { uprintf("Kernel module '%s' must be recompiled with " "linker script\n", ef->lf.pathname); return (ENOEXEC); } /* Padding from linker-script correct? */ pad = *(uint32_t *)((uintptr_t)ef->vnet_stop - sizeof(pad)); if (pad != LS_PADDING) { uprintf("Kernel module '%s' must be recompiled with " "linker script, invalid padding %#04x (%#04x)\n", ef->lf.pathname, pad, LS_PADDING); return (ENOEXEC); } /* If we only have valid padding, nothing to do. */ if (size == 4) return (0); #endif /* * Allocate space in the primary vnet area. Copy in our * initialization from the data section and then initialize * all per-vnet storage from that. */ ef->vnet_base = (Elf_Addr)(uintptr_t)vnet_data_alloc(size); if (ef->vnet_base == 0) { printf("%s: vnet module space is out of space; " "cannot allocate %d for %s\n", __func__, size, ef->lf.pathname); return (ENOSPC); } memcpy((void *)ef->vnet_base, (void *)ef->vnet_start, size); vnet_data_copy((void *)ef->vnet_base, size); elf_set_add(&set_vnet_list, ef->vnet_start, ef->vnet_stop, ef->vnet_base); return (0); } #endif #undef LS_PADDING /* * Apply the specified protection to the loadable segments of a preloaded linker * file. */ static int preload_protect(elf_file_t ef, vm_prot_t prot) { #ifdef __amd64__ Elf_Ehdr *hdr; Elf_Phdr *phdr, *phlimit; vm_prot_t nprot; int error; error = 0; hdr = (Elf_Ehdr *)ef->address; phdr = (Elf_Phdr *)(ef->address + hdr->e_phoff); phlimit = phdr + hdr->e_phnum; for (; phdr < phlimit; phdr++) { if (phdr->p_type != PT_LOAD) continue; nprot = prot | VM_PROT_READ; if ((phdr->p_flags & PF_W) != 0) nprot |= VM_PROT_WRITE; if ((phdr->p_flags & PF_X) != 0) nprot |= VM_PROT_EXECUTE; error = pmap_change_prot((vm_offset_t)ef->address + phdr->p_vaddr, round_page(phdr->p_memsz), nprot); if (error != 0) break; } return (error); #else return (0); #endif } #ifdef __arm__ /* * Locate the ARM exception/unwind table info for DDB and stack(9) use by * searching for the section header that describes it. There may be no unwind * info, for example in a module containing only data. */ static void link_elf_locate_exidx(linker_file_t lf, Elf_Shdr *shdr, int nhdr) { int i; for (i = 0; i < nhdr; i++) { if (shdr[i].sh_type == SHT_ARM_EXIDX) { lf->exidx_addr = shdr[i].sh_addr + lf->address; lf->exidx_size = shdr[i].sh_size; break; } } } /* * Locate the section headers metadata in a preloaded module, then use it to * locate the exception/unwind table in the module. The size of the metadata * block is stored in a uint32 word immediately before the data itself, and a * comment in preload_search_info() says it is safe to rely on that. */ static void link_elf_locate_exidx_preload(struct linker_file *lf, caddr_t modptr) { uint32_t *modinfo; Elf_Shdr *shdr; uint32_t nhdr; modinfo = (uint32_t *)preload_search_info(modptr, MODINFO_METADATA | MODINFOMD_SHDR); if (modinfo != NULL) { shdr = (Elf_Shdr *)modinfo; nhdr = modinfo[-1] / sizeof(Elf_Shdr); link_elf_locate_exidx(lf, shdr, nhdr); } } #endif /* __arm__ */ static int link_elf_link_preload(linker_class_t cls, const char *filename, linker_file_t *result) { Elf_Addr *ctors_addrp; Elf_Size *ctors_sizep; caddr_t modptr, baseptr, sizeptr, dynptr; char *type; elf_file_t ef; linker_file_t lf; int error; vm_offset_t dp; /* Look to see if we have the file preloaded */ modptr = preload_search_by_name(filename); if (modptr == NULL) return (ENOENT); type = (char *)preload_search_info(modptr, MODINFO_TYPE); baseptr = preload_search_info(modptr, MODINFO_ADDR); sizeptr = preload_search_info(modptr, MODINFO_SIZE); dynptr = preload_search_info(modptr, MODINFO_METADATA | MODINFOMD_DYNAMIC); if (type == NULL || (strcmp(type, "elf" __XSTRING(__ELF_WORD_SIZE) " module") != 0 && strcmp(type, "elf module") != 0)) return (EFTYPE); if (baseptr == NULL || sizeptr == NULL || dynptr == NULL) return (EINVAL); lf = linker_make_file(filename, &link_elf_class); if (lf == NULL) return (ENOMEM); ef = (elf_file_t) lf; ef->preloaded = 1; ef->modptr = modptr; ef->address = *(caddr_t *)baseptr; #ifdef SPARSE_MAPPING ef->object = NULL; #endif dp = (vm_offset_t)ef->address + *(vm_offset_t *)dynptr; ef->dynamic = (Elf_Dyn *)dp; lf->address = ef->address; lf->size = *(size_t *)sizeptr; ctors_addrp = (Elf_Addr *)preload_search_info(modptr, MODINFO_METADATA | MODINFOMD_CTORS_ADDR); ctors_sizep = (Elf_Size *)preload_search_info(modptr, MODINFO_METADATA | MODINFOMD_CTORS_SIZE); if (ctors_addrp != NULL && ctors_sizep != NULL) { lf->ctors_addr = ef->address + *ctors_addrp; lf->ctors_size = *ctors_sizep; } #ifdef __arm__ link_elf_locate_exidx_preload(lf, modptr); #endif error = parse_dynamic(ef); if (error == 0) error = parse_dpcpu(ef); #ifdef VIMAGE if (error == 0) error = parse_vnet(ef); #endif if (error == 0) error = preload_protect(ef, VM_PROT_ALL); if (error != 0) { linker_file_unload(lf, LINKER_UNLOAD_FORCE); return (error); } link_elf_reloc_local(lf); *result = lf; return (0); } static int link_elf_link_preload_finish(linker_file_t lf) { elf_file_t ef; int error; ef = (elf_file_t) lf; error = relocate_file(ef); if (error == 0) error = preload_protect(ef, VM_PROT_NONE); if (error != 0) return (error); (void)link_elf_preload_parse_symbols(ef); return (link_elf_link_common_finish(lf)); } static int link_elf_load_file(linker_class_t cls, const char* filename, linker_file_t* result) { struct nameidata nd; struct thread* td = curthread; /* XXX */ Elf_Ehdr *hdr; caddr_t firstpage, segbase; int nbytes, i; Elf_Phdr *phdr; Elf_Phdr *phlimit; Elf_Phdr *segs[MAXSEGS]; int nsegs; Elf_Phdr *phdyn; caddr_t mapbase; size_t mapsize; Elf_Addr base_vaddr; Elf_Addr base_vlimit; int error = 0; ssize_t resid; int flags; elf_file_t ef; linker_file_t lf; Elf_Shdr *shdr; int symtabindex; int symstrindex; int shstrindex; int symcnt; int strcnt; char *shstrs; shdr = NULL; lf = NULL; shstrs = NULL; NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, filename, td); flags = FREAD; error = vn_open(&nd, &flags, 0, NULL); if (error != 0) return (error); NDFREE(&nd, NDF_ONLY_PNBUF); if (nd.ni_vp->v_type != VREG) { error = ENOEXEC; firstpage = NULL; goto out; } #ifdef MAC error = mac_kld_check_load(curthread->td_ucred, nd.ni_vp); if (error != 0) { firstpage = NULL; goto out; } #endif /* * Read the elf header from the file. */ firstpage = malloc(PAGE_SIZE, M_LINKER, M_WAITOK); hdr = (Elf_Ehdr *)firstpage; error = vn_rdwr(UIO_READ, nd.ni_vp, firstpage, PAGE_SIZE, 0, UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred, NOCRED, &resid, td); nbytes = PAGE_SIZE - resid; if (error != 0) goto out; if (!IS_ELF(*hdr)) { error = ENOEXEC; goto out; } if (hdr->e_ident[EI_CLASS] != ELF_TARG_CLASS || hdr->e_ident[EI_DATA] != ELF_TARG_DATA) { link_elf_error(filename, "Unsupported file layout"); error = ENOEXEC; goto out; } if (hdr->e_ident[EI_VERSION] != EV_CURRENT || hdr->e_version != EV_CURRENT) { link_elf_error(filename, "Unsupported file version"); error = ENOEXEC; goto out; } if (hdr->e_type != ET_EXEC && hdr->e_type != ET_DYN) { error = ENOSYS; goto out; } if (hdr->e_machine != ELF_TARG_MACH) { link_elf_error(filename, "Unsupported machine"); error = ENOEXEC; goto out; } /* * We rely on the program header being in the first page. * This is not strictly required by the ABI specification, but * it seems to always true in practice. And, it simplifies * things considerably. */ if (!((hdr->e_phentsize == sizeof(Elf_Phdr)) && (hdr->e_phoff + hdr->e_phnum*sizeof(Elf_Phdr) <= PAGE_SIZE) && (hdr->e_phoff + hdr->e_phnum*sizeof(Elf_Phdr) <= nbytes))) link_elf_error(filename, "Unreadable program headers"); /* * Scan the program header entries, and save key information. * * We rely on there being exactly two load segments, text and data, * in that order. */ phdr = (Elf_Phdr *) (firstpage + hdr->e_phoff); phlimit = phdr + hdr->e_phnum; nsegs = 0; phdyn = NULL; while (phdr < phlimit) { switch (phdr->p_type) { case PT_LOAD: if (nsegs == MAXSEGS) { link_elf_error(filename, "Too many sections"); error = ENOEXEC; goto out; } /* * XXX: We just trust they come in right order ?? */ segs[nsegs] = phdr; ++nsegs; break; case PT_DYNAMIC: phdyn = phdr; break; case PT_INTERP: error = ENOSYS; goto out; } ++phdr; } if (phdyn == NULL) { link_elf_error(filename, "Object is not dynamically-linked"); error = ENOEXEC; goto out; } if (nsegs == 0) { link_elf_error(filename, "No sections"); error = ENOEXEC; goto out; } /* * Allocate the entire address space of the object, to stake * out our contiguous region, and to establish the base * address for relocation. */ base_vaddr = trunc_page(segs[0]->p_vaddr); base_vlimit = round_page(segs[nsegs - 1]->p_vaddr + segs[nsegs - 1]->p_memsz); mapsize = base_vlimit - base_vaddr; lf = linker_make_file(filename, &link_elf_class); if (lf == NULL) { error = ENOMEM; goto out; } ef = (elf_file_t) lf; #ifdef SPARSE_MAPPING ef->object = vm_object_allocate(OBJT_PHYS, atop(mapsize)); if (ef->object == NULL) { error = ENOMEM; goto out; } #ifdef __amd64__ mapbase = (caddr_t)KERNBASE; #else mapbase = (caddr_t)vm_map_min(kernel_map); #endif /* * Mapping protections are downgraded after relocation processing. */ error = vm_map_find(kernel_map, ef->object, 0, (vm_offset_t *)&mapbase, mapsize, 0, VMFS_OPTIMAL_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0); if (error != 0) { vm_object_deallocate(ef->object); ef->object = NULL; goto out; } #else mapbase = malloc(mapsize, M_LINKER, M_EXEC | M_WAITOK); #endif ef->address = mapbase; /* * Read the text and data sections and zero the bss. */ for (i = 0; i < nsegs; i++) { segbase = mapbase + segs[i]->p_vaddr - base_vaddr; #ifdef SPARSE_MAPPING /* * Consecutive segments may have different mapping permissions, * so be strict and verify that their mappings do not overlap. */ if (((vm_offset_t)segbase & PAGE_MASK) != 0) { error = EINVAL; goto out; } error = vm_map_wire(kernel_map, (vm_offset_t)segbase, (vm_offset_t)segbase + round_page(segs[i]->p_memsz), VM_MAP_WIRE_SYSTEM | VM_MAP_WIRE_NOHOLES); if (error != KERN_SUCCESS) { error = ENOMEM; goto out; } #endif error = vn_rdwr(UIO_READ, nd.ni_vp, segbase, segs[i]->p_filesz, segs[i]->p_offset, UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred, NOCRED, &resid, td); if (error != 0) goto out; bzero(segbase + segs[i]->p_filesz, segs[i]->p_memsz - segs[i]->p_filesz); } #ifdef GPROF /* Update profiling information with the new text segment. */ mtx_lock(&Giant); kmupetext((uintfptr_t)(mapbase + segs[0]->p_vaddr - base_vaddr + segs[0]->p_memsz)); mtx_unlock(&Giant); #endif ef->dynamic = (Elf_Dyn *) (mapbase + phdyn->p_vaddr - base_vaddr); lf->address = ef->address; lf->size = mapsize; error = parse_dynamic(ef); if (error != 0) goto out; error = parse_dpcpu(ef); if (error != 0) goto out; #ifdef VIMAGE error = parse_vnet(ef); if (error != 0) goto out; #endif link_elf_reloc_local(lf); VOP_UNLOCK(nd.ni_vp); error = linker_load_dependencies(lf); vn_lock(nd.ni_vp, LK_EXCLUSIVE | LK_RETRY); if (error != 0) goto out; error = relocate_file(ef); if (error != 0) goto out; #ifdef SPARSE_MAPPING /* * Downgrade permissions on text segment mappings now that relocation * processing is complete. Restrict permissions on read-only segments. */ for (i = 0; i < nsegs; i++) { vm_prot_t prot; if (segs[i]->p_type != PT_LOAD) continue; prot = VM_PROT_READ; if ((segs[i]->p_flags & PF_W) != 0) prot |= VM_PROT_WRITE; if ((segs[i]->p_flags & PF_X) != 0) prot |= VM_PROT_EXECUTE; segbase = mapbase + segs[i]->p_vaddr - base_vaddr; error = vm_map_protect(kernel_map, (vm_offset_t)segbase, (vm_offset_t)segbase + round_page(segs[i]->p_memsz), prot, FALSE); if (error != KERN_SUCCESS) { error = ENOMEM; goto out; } } #endif /* * Try and load the symbol table if it's present. (you can * strip it!) */ nbytes = hdr->e_shnum * hdr->e_shentsize; if (nbytes == 0 || hdr->e_shoff == 0) goto nosyms; shdr = malloc(nbytes, M_LINKER, M_WAITOK | M_ZERO); error = vn_rdwr(UIO_READ, nd.ni_vp, (caddr_t)shdr, nbytes, hdr->e_shoff, UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred, NOCRED, &resid, td); if (error != 0) goto out; /* Read section string table */ shstrindex = hdr->e_shstrndx; if (shstrindex != 0 && shdr[shstrindex].sh_type == SHT_STRTAB && shdr[shstrindex].sh_size != 0) { nbytes = shdr[shstrindex].sh_size; shstrs = malloc(nbytes, M_LINKER, M_WAITOK | M_ZERO); error = vn_rdwr(UIO_READ, nd.ni_vp, (caddr_t)shstrs, nbytes, shdr[shstrindex].sh_offset, UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred, NOCRED, &resid, td); if (error) goto out; } symtabindex = -1; symstrindex = -1; for (i = 0; i < hdr->e_shnum; i++) { if (shdr[i].sh_type == SHT_SYMTAB) { symtabindex = i; symstrindex = shdr[i].sh_link; } else if (shstrs != NULL && shdr[i].sh_name != 0 && strcmp(shstrs + shdr[i].sh_name, ".ctors") == 0) { /* Record relocated address and size of .ctors. */ lf->ctors_addr = mapbase + shdr[i].sh_addr - base_vaddr; lf->ctors_size = shdr[i].sh_size; } } if (symtabindex < 0 || symstrindex < 0) goto nosyms; symcnt = shdr[symtabindex].sh_size; ef->symbase = malloc(symcnt, M_LINKER, M_WAITOK); strcnt = shdr[symstrindex].sh_size; ef->strbase = malloc(strcnt, M_LINKER, M_WAITOK); error = vn_rdwr(UIO_READ, nd.ni_vp, ef->symbase, symcnt, shdr[symtabindex].sh_offset, UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred, NOCRED, &resid, td); if (error != 0) goto out; error = vn_rdwr(UIO_READ, nd.ni_vp, ef->strbase, strcnt, shdr[symstrindex].sh_offset, UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred, NOCRED, &resid, td); if (error != 0) goto out; ef->ddbsymcnt = symcnt / sizeof(Elf_Sym); ef->ddbsymtab = (const Elf_Sym *)ef->symbase; ef->ddbstrcnt = strcnt; ef->ddbstrtab = ef->strbase; nosyms: #ifdef __arm__ link_elf_locate_exidx(lf, shdr, hdr->e_shnum); #endif error = link_elf_link_common_finish(lf); if (error != 0) goto out; *result = lf; out: VOP_UNLOCK(nd.ni_vp); vn_close(nd.ni_vp, FREAD, td->td_ucred, td); if (error != 0 && lf != NULL) linker_file_unload(lf, LINKER_UNLOAD_FORCE); free(shdr, M_LINKER); free(firstpage, M_LINKER); free(shstrs, M_LINKER); return (error); } Elf_Addr elf_relocaddr(linker_file_t lf, Elf_Addr x) { elf_file_t ef; KASSERT(lf->ops->cls == (kobj_class_t)&link_elf_class, ("elf_relocaddr: unexpected linker file %p", lf)); ef = (elf_file_t)lf; if (x >= ef->pcpu_start && x < ef->pcpu_stop) return ((x - ef->pcpu_start) + ef->pcpu_base); #ifdef VIMAGE if (x >= ef->vnet_start && x < ef->vnet_stop) return ((x - ef->vnet_start) + ef->vnet_base); #endif return (x); } static void link_elf_unload_file(linker_file_t file) { elf_file_t ef = (elf_file_t) file; if (ef->pcpu_base != 0) { dpcpu_free((void *)ef->pcpu_base, ef->pcpu_stop - ef->pcpu_start); elf_set_delete(&set_pcpu_list, ef->pcpu_start); } #ifdef VIMAGE if (ef->vnet_base != 0) { vnet_data_free((void *)ef->vnet_base, ef->vnet_stop - ef->vnet_start); elf_set_delete(&set_vnet_list, ef->vnet_start); } #endif #ifdef GDB if (ef->gdb.l_ld != NULL) { GDB_STATE(RT_DELETE); free((void *)(uintptr_t)ef->gdb.l_name, M_LINKER); link_elf_delete_gdb(&ef->gdb); GDB_STATE(RT_CONSISTENT); } #endif /* Notify MD code that a module is being unloaded. */ elf_cpu_unload_file(file); if (ef->preloaded) { link_elf_unload_preload(file); return; } #ifdef SPARSE_MAPPING if (ef->object != NULL) { vm_map_remove(kernel_map, (vm_offset_t) ef->address, (vm_offset_t) ef->address + (ef->object->size << PAGE_SHIFT)); } #else free(ef->address, M_LINKER); #endif free(ef->symbase, M_LINKER); free(ef->strbase, M_LINKER); free(ef->ctftab, M_LINKER); free(ef->ctfoff, M_LINKER); free(ef->typoff, M_LINKER); } static void link_elf_unload_preload(linker_file_t file) { if (file->pathname != NULL) preload_delete_name(file->pathname); } static const char * symbol_name(elf_file_t ef, Elf_Size r_info) { const Elf_Sym *ref; if (ELF_R_SYM(r_info)) { ref = ef->symtab + ELF_R_SYM(r_info); return (ef->strtab + ref->st_name); } return (NULL); } static int symbol_type(elf_file_t ef, Elf_Size r_info) { const Elf_Sym *ref; if (ELF_R_SYM(r_info)) { ref = ef->symtab + ELF_R_SYM(r_info); return (ELF_ST_TYPE(ref->st_info)); } return (STT_NOTYPE); } static int relocate_file1(elf_file_t ef, elf_lookup_fn lookup, elf_reloc_fn reloc, bool ifuncs) { const Elf_Rel *rel; const Elf_Rela *rela; const char *symname; #define APPLY_RELOCS(iter, tbl, tblsize, type) do { \ for ((iter) = (tbl); (iter) != NULL && \ (iter) < (tbl) + (tblsize) / sizeof(*(iter)); (iter)++) { \ if ((symbol_type(ef, (iter)->r_info) == \ STT_GNU_IFUNC || \ elf_is_ifunc_reloc((iter)->r_info)) != ifuncs) \ continue; \ if (reloc(&ef->lf, (Elf_Addr)ef->address, \ (iter), (type), lookup)) { \ symname = symbol_name(ef, (iter)->r_info); \ printf("link_elf: symbol %s undefined\n", \ symname); \ return (ENOENT); \ } \ } \ } while (0) APPLY_RELOCS(rel, ef->rel, ef->relsize, ELF_RELOC_REL); APPLY_RELOCS(rela, ef->rela, ef->relasize, ELF_RELOC_RELA); APPLY_RELOCS(rel, ef->pltrel, ef->pltrelsize, ELF_RELOC_REL); APPLY_RELOCS(rela, ef->pltrela, ef->pltrelasize, ELF_RELOC_RELA); #undef APPLY_RELOCS return (0); } static int relocate_file(elf_file_t ef) { int error; error = relocate_file1(ef, elf_lookup, elf_reloc, false); if (error == 0) error = relocate_file1(ef, elf_lookup, elf_reloc, true); return (error); } /* * Hash function for symbol table lookup. Don't even think about changing * this. It is specified by the System V ABI. */ static unsigned long elf_hash(const char *name) { const unsigned char *p = (const unsigned char *) name; unsigned long h = 0; unsigned long g; while (*p != '\0') { h = (h << 4) + *p++; if ((g = h & 0xf0000000) != 0) h ^= g >> 24; h &= ~g; } return (h); } static int link_elf_lookup_symbol(linker_file_t lf, const char *name, c_linker_sym_t *sym) { elf_file_t ef = (elf_file_t) lf; unsigned long symnum; const Elf_Sym* symp; const char *strp; unsigned long hash; int i; /* If we don't have a hash, bail. */ if (ef->buckets == NULL || ef->nbuckets == 0) { printf("link_elf_lookup_symbol: missing symbol hash table\n"); return (ENOENT); } /* First, search hashed global symbols */ hash = elf_hash(name); symnum = ef->buckets[hash % ef->nbuckets]; while (symnum != STN_UNDEF) { if (symnum >= ef->nchains) { printf("%s: corrupt symbol table\n", __func__); return (ENOENT); } symp = ef->symtab + symnum; if (symp->st_name == 0) { printf("%s: corrupt symbol table\n", __func__); return (ENOENT); } strp = ef->strtab + symp->st_name; if (strcmp(name, strp) == 0) { if (symp->st_shndx != SHN_UNDEF || (symp->st_value != 0 && (ELF_ST_TYPE(symp->st_info) == STT_FUNC || ELF_ST_TYPE(symp->st_info) == STT_GNU_IFUNC))) { *sym = (c_linker_sym_t) symp; return (0); } return (ENOENT); } symnum = ef->chains[symnum]; } /* If we have not found it, look at the full table (if loaded) */ if (ef->symtab == ef->ddbsymtab) return (ENOENT); /* Exhaustive search */ for (i = 0, symp = ef->ddbsymtab; i < ef->ddbsymcnt; i++, symp++) { strp = ef->ddbstrtab + symp->st_name; if (strcmp(name, strp) == 0) { if (symp->st_shndx != SHN_UNDEF || (symp->st_value != 0 && (ELF_ST_TYPE(symp->st_info) == STT_FUNC || ELF_ST_TYPE(symp->st_info) == STT_GNU_IFUNC))) { *sym = (c_linker_sym_t) symp; return (0); } return (ENOENT); } } return (ENOENT); } static int link_elf_symbol_values(linker_file_t lf, c_linker_sym_t sym, linker_symval_t *symval) { elf_file_t ef; const Elf_Sym *es; caddr_t val; ef = (elf_file_t)lf; es = (const Elf_Sym *)sym; if (es >= ef->symtab && es < (ef->symtab + ef->nchains)) { symval->name = ef->strtab + es->st_name; val = (caddr_t)ef->address + es->st_value; if (ELF_ST_TYPE(es->st_info) == STT_GNU_IFUNC) val = ((caddr_t (*)(void))val)(); symval->value = val; symval->size = es->st_size; return (0); } if (ef->symtab == ef->ddbsymtab) return (ENOENT); if (es >= ef->ddbsymtab && es < (ef->ddbsymtab + ef->ddbsymcnt)) { symval->name = ef->ddbstrtab + es->st_name; val = (caddr_t)ef->address + es->st_value; if (ELF_ST_TYPE(es->st_info) == STT_GNU_IFUNC) val = ((caddr_t (*)(void))val)(); symval->value = val; symval->size = es->st_size; return (0); } return (ENOENT); } static int link_elf_search_symbol(linker_file_t lf, caddr_t value, c_linker_sym_t *sym, long *diffp) { elf_file_t ef = (elf_file_t) lf; u_long off = (uintptr_t) (void *) value; u_long diff = off; u_long st_value; const Elf_Sym* es; const Elf_Sym* best = NULL; int i; for (i = 0, es = ef->ddbsymtab; i < ef->ddbsymcnt; i++, es++) { if (es->st_name == 0) continue; st_value = es->st_value + (uintptr_t) (void *) ef->address; if (off >= st_value) { if (off - st_value < diff) { diff = off - st_value; best = es; if (diff == 0) break; } else if (off - st_value == diff) { best = es; } } } if (best == NULL) *diffp = off; else *diffp = diff; *sym = (c_linker_sym_t) best; return (0); } /* * Look up a linker set on an ELF system. */ static int link_elf_lookup_set(linker_file_t lf, const char *name, void ***startp, void ***stopp, int *countp) { c_linker_sym_t sym; linker_symval_t symval; char *setsym; void **start, **stop; int len, error = 0, count; len = strlen(name) + sizeof("__start_set_"); /* sizeof includes \0 */ setsym = malloc(len, M_LINKER, M_WAITOK); /* get address of first entry */ snprintf(setsym, len, "%s%s", "__start_set_", name); error = link_elf_lookup_symbol(lf, setsym, &sym); if (error != 0) goto out; link_elf_symbol_values(lf, sym, &symval); if (symval.value == 0) { error = ESRCH; goto out; } start = (void **)symval.value; /* get address of last entry */ snprintf(setsym, len, "%s%s", "__stop_set_", name); error = link_elf_lookup_symbol(lf, setsym, &sym); if (error != 0) goto out; link_elf_symbol_values(lf, sym, &symval); if (symval.value == 0) { error = ESRCH; goto out; } stop = (void **)symval.value; /* and the number of entries */ count = stop - start; /* and copy out */ if (startp != NULL) *startp = start; if (stopp != NULL) *stopp = stop; if (countp != NULL) *countp = count; out: free(setsym, M_LINKER); return (error); } static int link_elf_each_function_name(linker_file_t file, int (*callback)(const char *, void *), void *opaque) { elf_file_t ef = (elf_file_t)file; const Elf_Sym *symp; int i, error; /* Exhaustive search */ for (i = 0, symp = ef->ddbsymtab; i < ef->ddbsymcnt; i++, symp++) { if (symp->st_value != 0 && (ELF_ST_TYPE(symp->st_info) == STT_FUNC || ELF_ST_TYPE(symp->st_info) == STT_GNU_IFUNC)) { error = callback(ef->ddbstrtab + symp->st_name, opaque); if (error != 0) return (error); } } return (0); } static int link_elf_each_function_nameval(linker_file_t file, linker_function_nameval_callback_t callback, void *opaque) { linker_symval_t symval; elf_file_t ef = (elf_file_t)file; const Elf_Sym* symp; int i, error; /* Exhaustive search */ for (i = 0, symp = ef->ddbsymtab; i < ef->ddbsymcnt; i++, symp++) { if (symp->st_value != 0 && (ELF_ST_TYPE(symp->st_info) == STT_FUNC || ELF_ST_TYPE(symp->st_info) == STT_GNU_IFUNC)) { error = link_elf_symbol_values(file, (c_linker_sym_t) symp, &symval); if (error != 0) return (error); error = callback(file, i, &symval, opaque); if (error != 0) return (error); } } return (0); } const Elf_Sym * elf_get_sym(linker_file_t lf, Elf_Size symidx) { elf_file_t ef = (elf_file_t)lf; if (symidx >= ef->nchains) return (NULL); return (ef->symtab + symidx); } const char * elf_get_symname(linker_file_t lf, Elf_Size symidx) { elf_file_t ef = (elf_file_t)lf; const Elf_Sym *sym; if (symidx >= ef->nchains) return (NULL); sym = ef->symtab + symidx; return (ef->strtab + sym->st_name); } /* * Symbol lookup function that can be used when the symbol index is known (ie * in relocations). It uses the symbol index instead of doing a fully fledged * hash table based lookup when such is valid. For example for local symbols. * This is not only more efficient, it's also more correct. It's not always * the case that the symbol can be found through the hash table. */ static int elf_lookup(linker_file_t lf, Elf_Size symidx, int deps, Elf_Addr *res) { elf_file_t ef = (elf_file_t)lf; const Elf_Sym *sym; const char *symbol; Elf_Addr addr, start, base; /* Don't even try to lookup the symbol if the index is bogus. */ if (symidx >= ef->nchains) { *res = 0; return (EINVAL); } sym = ef->symtab + symidx; /* * Don't do a full lookup when the symbol is local. It may even * fail because it may not be found through the hash table. */ if (ELF_ST_BIND(sym->st_info) == STB_LOCAL) { /* Force lookup failure when we have an insanity. */ if (sym->st_shndx == SHN_UNDEF || sym->st_value == 0) { *res = 0; return (EINVAL); } *res = ((Elf_Addr)ef->address + sym->st_value); return (0); } /* * XXX we can avoid doing a hash table based lookup for global * symbols as well. This however is not always valid, so we'll * just do it the hard way for now. Performance tweaks can * always be added. */ symbol = ef->strtab + sym->st_name; /* Force a lookup failure if the symbol name is bogus. */ if (*symbol == 0) { *res = 0; return (EINVAL); } addr = ((Elf_Addr)linker_file_lookup_symbol(lf, symbol, deps)); if (addr == 0 && ELF_ST_BIND(sym->st_info) != STB_WEAK) { *res = 0; return (EINVAL); } if (elf_set_find(&set_pcpu_list, addr, &start, &base)) addr = addr - start + base; #ifdef VIMAGE else if (elf_set_find(&set_vnet_list, addr, &start, &base)) addr = addr - start + base; #endif *res = addr; return (0); } static void link_elf_reloc_local(linker_file_t lf) { const Elf_Rel *rellim; const Elf_Rel *rel; const Elf_Rela *relalim; const Elf_Rela *rela; elf_file_t ef = (elf_file_t)lf; /* Perform relocations without addend if there are any: */ if ((rel = ef->rel) != NULL) { rellim = (const Elf_Rel *)((const char *)ef->rel + ef->relsize); while (rel < rellim) { elf_reloc_local(lf, (Elf_Addr)ef->address, rel, ELF_RELOC_REL, elf_lookup); rel++; } } /* Perform relocations with addend if there are any: */ if ((rela = ef->rela) != NULL) { relalim = (const Elf_Rela *) ((const char *)ef->rela + ef->relasize); while (rela < relalim) { elf_reloc_local(lf, (Elf_Addr)ef->address, rela, ELF_RELOC_RELA, elf_lookup); rela++; } } } static long link_elf_symtab_get(linker_file_t lf, const Elf_Sym **symtab) { elf_file_t ef = (elf_file_t)lf; *symtab = ef->ddbsymtab; if (*symtab == NULL) return (0); return (ef->ddbsymcnt); } static long link_elf_strtab_get(linker_file_t lf, caddr_t *strtab) { elf_file_t ef = (elf_file_t)lf; *strtab = ef->ddbstrtab; if (*strtab == NULL) return (0); return (ef->ddbstrcnt); } -#if defined(__i386__) || defined(__amd64__) || defined(__aarch64__) +#if defined(__i386__) || defined(__amd64__) || defined(__aarch64__) || defined(__powerpc__) /* * Use this lookup routine when performing relocations early during boot. * The generic lookup routine depends on kobj, which is not initialized * at that point. */ static int elf_lookup_ifunc(linker_file_t lf, Elf_Size symidx, int deps __unused, Elf_Addr *res) { elf_file_t ef; const Elf_Sym *symp; caddr_t val; ef = (elf_file_t)lf; symp = ef->symtab + symidx; if (ELF_ST_TYPE(symp->st_info) == STT_GNU_IFUNC) { val = (caddr_t)ef->address + symp->st_value; *res = ((Elf_Addr (*)(void))val)(); return (0); } return (ENOENT); } void link_elf_ireloc(caddr_t kmdp) { struct elf_file eff; elf_file_t ef; ef = &eff; bzero_early(ef, sizeof(*ef)); ef->modptr = kmdp; ef->dynamic = (Elf_Dyn *)&_DYNAMIC; - parse_dynamic(ef); + +#ifdef RELOCATABLE_KERNEL + ef->address = (caddr_t) (__startkernel - KERNBASE); +#else ef->address = 0; +#endif + parse_dynamic(ef); + link_elf_preload_parse_symbols(ef); relocate_file1(ef, elf_lookup_ifunc, elf_reloc, true); } #endif Index: head/sys/powerpc/aim/aim_machdep.c =================================================================== --- head/sys/powerpc/aim/aim_machdep.c (revision 360793) +++ head/sys/powerpc/aim/aim_machdep.c (revision 360794) @@ -1,714 +1,724 @@ /*- * Copyright (C) 1995, 1996 Wolfgang Solfrank. * Copyright (C) 1995, 1996 TooLs GmbH. * 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 TooLs GmbH. * 4. The name of TooLs GmbH may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``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 TOOLS GMBH 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. */ /*- * Copyright (C) 2001 Benno Rice * 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 Benno Rice ``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 TOOLS GMBH 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: machdep.c,v 1.74.2.1 2000/11/01 16:13:48 tv Exp $ */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include "opt_kstack_pages.h" #include "opt_platform.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 #include #include #include #include #include #include #include #include #ifndef __powerpc64__ #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef __powerpc64__ #include "mmu_oea64.h" #endif #ifndef __powerpc64__ struct bat battable[16]; #endif #ifndef __powerpc64__ /* Bits for running on 64-bit systems in 32-bit mode. */ extern void *testppc64, *testppc64size; extern void *restorebridge, *restorebridgesize; extern void *rfid_patch, *rfi_patch1, *rfi_patch2; extern void *trapcode64; extern Elf_Addr _GLOBAL_OFFSET_TABLE_[]; #endif extern void *rstcode, *rstcodeend; extern void *trapcode, *trapcodeend; extern void *hypertrapcode, *hypertrapcodeend; extern void *generictrap, *generictrap64; extern void *alitrap, *aliend; extern void *dsitrap, *dsiend; extern void *decrint, *decrsize; extern void *extint, *extsize; extern void *dblow, *dbend; extern void *imisstrap, *imisssize; extern void *dlmisstrap, *dlmisssize; extern void *dsmisstrap, *dsmisssize; extern void *ap_pcpu; extern void __restartkernel(vm_offset_t, vm_offset_t, vm_offset_t, void *, uint32_t, register_t offset, register_t msr); +extern void __restartkernel_virtual(vm_offset_t, vm_offset_t, vm_offset_t, void *, uint32_t, register_t offset, register_t msr); void aim_early_init(vm_offset_t fdt, vm_offset_t toc, vm_offset_t ofentry, void *mdp, uint32_t mdp_cookie); void aim_cpu_init(vm_offset_t toc); void aim_early_init(vm_offset_t fdt, vm_offset_t toc, vm_offset_t ofentry, void *mdp, uint32_t mdp_cookie) { register_t scratch; /* * If running from an FDT, make sure we are in real mode to avoid * tromping on firmware page tables. Everything in the kernel assumes * 1:1 mappings out of firmware, so this won't break anything not * already broken. This doesn't work if there is live OF, since OF * may internally use non-1:1 mappings. */ if (ofentry == 0) mtmsr(mfmsr() & ~(PSL_IR | PSL_DR)); #ifdef __powerpc64__ /* - * If in real mode, relocate to high memory so that the kernel + * Relocate to high memory so that the kernel * can execute from the direct map. + * + * If we are in virtual mode already, use a special entry point + * that sets up a temporary DMAP to execute from until we can + * properly set up the MMU. */ - if (!(mfmsr() & PSL_DR) && - (vm_offset_t)&aim_early_init < DMAP_BASE_ADDRESS) - __restartkernel(fdt, 0, ofentry, mdp, mdp_cookie, - DMAP_BASE_ADDRESS, mfmsr()); + if ((vm_offset_t)&aim_early_init < DMAP_BASE_ADDRESS) { + if (mfmsr() & PSL_DR) { + __restartkernel_virtual(fdt, 0, ofentry, mdp, + mdp_cookie, DMAP_BASE_ADDRESS, mfmsr()); + } else { + __restartkernel(fdt, 0, ofentry, mdp, mdp_cookie, + DMAP_BASE_ADDRESS, mfmsr()); + } + } #endif /* Various very early CPU fix ups */ switch (mfpvr() >> 16) { /* * PowerPC 970 CPUs have a misfeature requested by Apple that * makes them pretend they have a 32-byte cacheline. Turn this * off before we measure the cacheline size. */ case IBM970: case IBM970FX: case IBM970MP: case IBM970GX: scratch = mfspr(SPR_HID5); scratch &= ~HID5_970_DCBZ_SIZE_HI; mtspr(SPR_HID5, scratch); break; #ifdef __powerpc64__ case IBMPOWER7: case IBMPOWER7PLUS: case IBMPOWER8: case IBMPOWER8E: case IBMPOWER8NVL: case IBMPOWER9: /* XXX: get from ibm,slb-size in device tree */ n_slbs = 32; break; #endif } } void aim_cpu_init(vm_offset_t toc) { size_t trap_offset, trapsize; vm_offset_t trap; register_t msr; uint8_t *cache_check; int cacheline_warn; #ifndef __powerpc64__ register_t scratch; int ppc64; #endif trap_offset = 0; cacheline_warn = 0; /* General setup for AIM CPUs */ psl_kernset = PSL_EE | PSL_ME | PSL_IR | PSL_DR | PSL_RI; #ifdef __powerpc64__ psl_kernset |= PSL_SF; if (mfmsr() & PSL_HV) psl_kernset |= PSL_HV; #endif psl_userset = psl_kernset | PSL_PR; #ifdef __powerpc64__ psl_userset32 = psl_userset & ~PSL_SF; #endif /* * Zeroed bits in this variable signify that the value of the bit * in its position is allowed to vary between userspace contexts. * * All other bits are required to be identical for every userspace * context. The actual *value* of the bit is determined by * psl_userset and/or psl_userset32, and is not allowed to change. * * Remember to update this set when implementing support for * *conditionally* enabling a processor facility. Failing to do * this will cause swapcontext() in userspace to break when a * process uses a conditionally-enabled facility. * * When *unconditionally* implementing support for a processor * facility, update psl_userset / psl_userset32 instead. * * See the access control check in set_mcontext(). */ psl_userstatic = ~(PSL_VSX | PSL_VEC | PSL_FP | PSL_FE0 | PSL_FE1); /* * Mask bits from the SRR1 that aren't really the MSR: * Bits 1-4, 10-15 (ppc32), 33-36, 42-47 (ppc64) */ psl_userstatic &= ~0x783f0000UL; /* * Initialize the interrupt tables and figure out our cache line * size and whether or not we need the 64-bit bridge code. */ /* * Disable translation in case the vector area hasn't been * mapped (G5). Note that no OFW calls can be made until * translation is re-enabled. */ msr = mfmsr(); mtmsr((msr & ~(PSL_IR | PSL_DR)) | PSL_RI); /* * Measure the cacheline size using dcbz * * Use EXC_PGM as a playground. We are about to overwrite it * anyway, we know it exists, and we know it is cache-aligned. */ cache_check = (void *)EXC_PGM; for (cacheline_size = 0; cacheline_size < 0x100; cacheline_size++) cache_check[cacheline_size] = 0xff; __asm __volatile("dcbz 0,%0":: "r" (cache_check) : "memory"); /* Find the first byte dcbz did not zero to get the cache line size */ for (cacheline_size = 0; cacheline_size < 0x100 && cache_check[cacheline_size] == 0; cacheline_size++); /* Work around psim bug */ if (cacheline_size == 0) { cacheline_warn = 1; cacheline_size = 32; } #ifndef __powerpc64__ /* * Figure out whether we need to use the 64 bit PMAP. This works by * executing an instruction that is only legal on 64-bit PPC (mtmsrd), * and setting ppc64 = 0 if that causes a trap. */ ppc64 = 1; bcopy(&testppc64, (void *)EXC_PGM, (size_t)&testppc64size); __syncicache((void *)EXC_PGM, (size_t)&testppc64size); __asm __volatile("\ mfmsr %0; \ mtsprg2 %1; \ \ mtmsrd %0; \ mfsprg2 %1;" : "=r"(scratch), "=r"(ppc64)); if (ppc64) cpu_features |= PPC_FEATURE_64; /* * Now copy restorebridge into all the handlers, if necessary, * and set up the trap tables. */ if (cpu_features & PPC_FEATURE_64) { /* Patch the two instances of rfi -> rfid */ bcopy(&rfid_patch,&rfi_patch1,4); #ifdef KDB /* rfi_patch2 is at the end of dbleave */ bcopy(&rfid_patch,&rfi_patch2,4); #endif } #else /* powerpc64 */ cpu_features |= PPC_FEATURE_64; #endif trapsize = (size_t)&trapcodeend - (size_t)&trapcode; /* * Copy generic handler into every possible trap. Special cases will get * different ones in a minute. */ for (trap = EXC_RST; trap < EXC_LAST; trap += 0x20) bcopy(&trapcode, (void *)trap, trapsize); #ifndef __powerpc64__ if (cpu_features & PPC_FEATURE_64) { /* * Copy a code snippet to restore 32-bit bridge mode * to the top of every non-generic trap handler */ trap_offset += (size_t)&restorebridgesize; bcopy(&restorebridge, (void *)EXC_RST, trap_offset); bcopy(&restorebridge, (void *)EXC_DSI, trap_offset); bcopy(&restorebridge, (void *)EXC_ALI, trap_offset); bcopy(&restorebridge, (void *)EXC_PGM, trap_offset); bcopy(&restorebridge, (void *)EXC_MCHK, trap_offset); bcopy(&restorebridge, (void *)EXC_TRC, trap_offset); bcopy(&restorebridge, (void *)EXC_BPT, trap_offset); } #else trapsize = (size_t)&hypertrapcodeend - (size_t)&hypertrapcode; bcopy(&hypertrapcode, (void *)(EXC_HEA + trap_offset), trapsize); bcopy(&hypertrapcode, (void *)(EXC_HMI + trap_offset), trapsize); bcopy(&hypertrapcode, (void *)(EXC_HVI + trap_offset), trapsize); bcopy(&hypertrapcode, (void *)(EXC_SOFT_PATCH + trap_offset), trapsize); #endif bcopy(&rstcode, (void *)(EXC_RST + trap_offset), (size_t)&rstcodeend - (size_t)&rstcode); #ifdef KDB bcopy(&dblow, (void *)(EXC_MCHK + trap_offset), (size_t)&dbend - (size_t)&dblow); bcopy(&dblow, (void *)(EXC_PGM + trap_offset), (size_t)&dbend - (size_t)&dblow); bcopy(&dblow, (void *)(EXC_TRC + trap_offset), (size_t)&dbend - (size_t)&dblow); bcopy(&dblow, (void *)(EXC_BPT + trap_offset), (size_t)&dbend - (size_t)&dblow); #endif bcopy(&alitrap, (void *)(EXC_ALI + trap_offset), (size_t)&aliend - (size_t)&alitrap); bcopy(&dsitrap, (void *)(EXC_DSI + trap_offset), (size_t)&dsiend - (size_t)&dsitrap); /* Set address of generictrap for self-reloc calculations */ *((void **)TRAP_GENTRAP) = &generictrap; #ifdef __powerpc64__ /* Set TOC base so that the interrupt code can get at it */ *((void **)TRAP_ENTRY) = &generictrap; *((register_t *)TRAP_TOCBASE) = toc; #else /* Set branch address for trap code */ if (cpu_features & PPC_FEATURE_64) *((void **)TRAP_ENTRY) = &generictrap64; else *((void **)TRAP_ENTRY) = &generictrap; *((void **)TRAP_TOCBASE) = _GLOBAL_OFFSET_TABLE_; /* G2-specific TLB miss helper handlers */ bcopy(&imisstrap, (void *)EXC_IMISS, (size_t)&imisssize); bcopy(&dlmisstrap, (void *)EXC_DLMISS, (size_t)&dlmisssize); bcopy(&dsmisstrap, (void *)EXC_DSMISS, (size_t)&dsmisssize); #endif __syncicache(EXC_RSVD, EXC_LAST - EXC_RSVD); /* * Restore MSR */ mtmsr(msr); /* Warn if cachline size was not determined */ if (cacheline_warn == 1) { printf("WARNING: cacheline size undetermined, setting to 32\n"); } /* * Initialise virtual memory. Use BUS_PROBE_GENERIC priority * in case the platform module had a better idea of what we * should do. */ if (cpu_features & PPC_FEATURE_64) pmap_mmu_install(MMU_TYPE_G5, BUS_PROBE_GENERIC); else pmap_mmu_install(MMU_TYPE_OEA, BUS_PROBE_GENERIC); } /* * Shutdown the CPU as much as possible. */ void cpu_halt(void) { OF_exit(); } int ptrace_single_step(struct thread *td) { struct trapframe *tf; tf = td->td_frame; tf->srr1 |= PSL_SE; return (0); } int ptrace_clear_single_step(struct thread *td) { struct trapframe *tf; tf = td->td_frame; tf->srr1 &= ~PSL_SE; return (0); } void kdb_cpu_clear_singlestep(void) { kdb_frame->srr1 &= ~PSL_SE; } void kdb_cpu_set_singlestep(void) { kdb_frame->srr1 |= PSL_SE; } /* * Initialise a struct pcpu. */ void cpu_pcpu_init(struct pcpu *pcpu, int cpuid, size_t sz) { #ifdef __powerpc64__ /* Copy the SLB contents from the current CPU */ memcpy(pcpu->pc_aim.slb, PCPU_GET(aim.slb), sizeof(pcpu->pc_aim.slb)); #endif } #ifndef __powerpc64__ uint64_t va_to_vsid(pmap_t pm, vm_offset_t va) { return ((pm->pm_sr[(uintptr_t)va >> ADDR_SR_SHFT]) & SR_VSID_MASK); } #endif /* * These functions need to provide addresses that both (a) work in real mode * (or whatever mode/circumstances the kernel is in in early boot (now)) and * (b) can still, in principle, work once the kernel is going. Because these * rely on existing mappings/real mode, unmap is a no-op. */ vm_offset_t pmap_early_io_map(vm_paddr_t pa, vm_size_t size) { KASSERT(!pmap_bootstrapped, ("Not available after PMAP started!")); /* * If we have the MMU up in early boot, assume it is 1:1. Otherwise, * try to get the address in a memory region compatible with the * direct map for efficiency later. */ if (mfmsr() & PSL_DR) return (pa); else return (DMAP_BASE_ADDRESS + pa); } void pmap_early_io_unmap(vm_offset_t va, vm_size_t size) { KASSERT(!pmap_bootstrapped, ("Not available after PMAP started!")); } /* From p3-53 of the MPC7450 RISC Microprocessor Family Reference Manual */ void flush_disable_caches(void) { register_t msr; register_t msscr0; register_t cache_reg; volatile uint32_t *memp; uint32_t temp; int i; int x; msr = mfmsr(); powerpc_sync(); mtmsr(msr & ~(PSL_EE | PSL_DR)); msscr0 = mfspr(SPR_MSSCR0); msscr0 &= ~MSSCR0_L2PFE; mtspr(SPR_MSSCR0, msscr0); powerpc_sync(); isync(); __asm__ __volatile__("dssall; sync"); powerpc_sync(); isync(); __asm__ __volatile__("dcbf 0,%0" :: "r"(0)); __asm__ __volatile__("dcbf 0,%0" :: "r"(0)); __asm__ __volatile__("dcbf 0,%0" :: "r"(0)); /* Lock the L1 Data cache. */ mtspr(SPR_LDSTCR, mfspr(SPR_LDSTCR) | 0xFF); powerpc_sync(); isync(); mtspr(SPR_LDSTCR, 0); /* * Perform this in two stages: Flush the cache starting in RAM, then do it * from ROM. */ memp = (volatile uint32_t *)0x00000000; for (i = 0; i < 128 * 1024; i++) { temp = *memp; __asm__ __volatile__("dcbf 0,%0" :: "r"(memp)); memp += 32/sizeof(*memp); } memp = (volatile uint32_t *)0xfff00000; x = 0xfe; for (; x != 0xff;) { mtspr(SPR_LDSTCR, x); for (i = 0; i < 128; i++) { temp = *memp; __asm__ __volatile__("dcbf 0,%0" :: "r"(memp)); memp += 32/sizeof(*memp); } x = ((x << 1) | 1) & 0xff; } mtspr(SPR_LDSTCR, 0); cache_reg = mfspr(SPR_L2CR); if (cache_reg & L2CR_L2E) { cache_reg &= ~(L2CR_L2IO_7450 | L2CR_L2DO_7450); mtspr(SPR_L2CR, cache_reg); powerpc_sync(); mtspr(SPR_L2CR, cache_reg | L2CR_L2HWF); while (mfspr(SPR_L2CR) & L2CR_L2HWF) ; /* Busy wait for cache to flush */ powerpc_sync(); cache_reg &= ~L2CR_L2E; mtspr(SPR_L2CR, cache_reg); powerpc_sync(); mtspr(SPR_L2CR, cache_reg | L2CR_L2I); powerpc_sync(); while (mfspr(SPR_L2CR) & L2CR_L2I) ; /* Busy wait for L2 cache invalidate */ powerpc_sync(); } cache_reg = mfspr(SPR_L3CR); if (cache_reg & L3CR_L3E) { cache_reg &= ~(L3CR_L3IO | L3CR_L3DO); mtspr(SPR_L3CR, cache_reg); powerpc_sync(); mtspr(SPR_L3CR, cache_reg | L3CR_L3HWF); while (mfspr(SPR_L3CR) & L3CR_L3HWF) ; /* Busy wait for cache to flush */ powerpc_sync(); cache_reg &= ~L3CR_L3E; mtspr(SPR_L3CR, cache_reg); powerpc_sync(); mtspr(SPR_L3CR, cache_reg | L3CR_L3I); powerpc_sync(); while (mfspr(SPR_L3CR) & L3CR_L3I) ; /* Busy wait for L3 cache invalidate */ powerpc_sync(); } mtspr(SPR_HID0, mfspr(SPR_HID0) & ~HID0_DCE); powerpc_sync(); isync(); mtmsr(msr); } void cpu_sleep() { static u_quad_t timebase = 0; static register_t sprgs[4]; static register_t srrs[2]; jmp_buf resetjb; struct thread *fputd; struct thread *vectd; register_t hid0; register_t msr; register_t saved_msr; ap_pcpu = pcpup; PCPU_SET(restore, &resetjb); saved_msr = mfmsr(); fputd = PCPU_GET(fputhread); vectd = PCPU_GET(vecthread); if (fputd != NULL) save_fpu(fputd); if (vectd != NULL) save_vec(vectd); if (setjmp(resetjb) == 0) { sprgs[0] = mfspr(SPR_SPRG0); sprgs[1] = mfspr(SPR_SPRG1); sprgs[2] = mfspr(SPR_SPRG2); sprgs[3] = mfspr(SPR_SPRG3); srrs[0] = mfspr(SPR_SRR0); srrs[1] = mfspr(SPR_SRR1); timebase = mftb(); powerpc_sync(); flush_disable_caches(); hid0 = mfspr(SPR_HID0); hid0 = (hid0 & ~(HID0_DOZE | HID0_NAP)) | HID0_SLEEP; powerpc_sync(); isync(); msr = mfmsr() | PSL_POW; mtspr(SPR_HID0, hid0); powerpc_sync(); while (1) mtmsr(msr); } platform_smp_timebase_sync(timebase, 0); PCPU_SET(curthread, curthread); PCPU_SET(curpcb, curthread->td_pcb); pmap_activate(curthread); powerpc_sync(); mtspr(SPR_SPRG0, sprgs[0]); mtspr(SPR_SPRG1, sprgs[1]); mtspr(SPR_SPRG2, sprgs[2]); mtspr(SPR_SPRG3, sprgs[3]); mtspr(SPR_SRR0, srrs[0]); mtspr(SPR_SRR1, srrs[1]); mtmsr(saved_msr); if (fputd == curthread) enable_fpu(curthread); if (vectd == curthread) enable_vec(curthread); powerpc_sync(); } Index: head/sys/powerpc/aim/locore64.S =================================================================== --- head/sys/powerpc/aim/locore64.S (revision 360793) +++ head/sys/powerpc/aim/locore64.S (revision 360794) @@ -1,219 +1,270 @@ /* $FreeBSD$ */ /*- * Copyright (C) 2010-2016 Nathan Whitehorn * 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 ``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 TOOLS GMBH 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. * * $FreeBSD$ */ #include "assym.inc" #include #include #include #include #include #include #ifdef _CALL_ELF .abiversion _CALL_ELF #endif /* Glue for linker script */ .globl kernbase .set kernbase, KERNBASE /* * Globals */ .data .align 3 GLOBAL(__startkernel) .llong begin GLOBAL(__endkernel) .llong end GLOBAL(can_wakeup) .llong 0x0 .align 4 #define TMPSTKSZ 16384 /* 16K temporary stack */ GLOBAL(tmpstk) .space TMPSTKSZ TOC_ENTRY(tmpstk) TOC_ENTRY(can_wakeup) #ifdef KDB #define TRAPSTKSZ 8192 /* 8k trap stack */ GLOBAL(trapstk) .space TRAPSTKSZ TOC_ENTRY(trapstk) #endif /* * Entry point for bootloaders that do not fully implement ELF and start * at the beginning of the image (kexec, notably). In its own section so * that it ends up before any linker-generated call stubs and actually at * the beginning of the image. kexec on some systems also enters at * (start of image) + 0x60, so put a spin loop there. */ .section ".text.kboot", "x", @progbits kbootentry: b __start . = kbootentry + 0x40 /* Magic address used in platform layer */ .global smp_spin_sem ap_kexec_spin_sem: .long -1 . = kbootentry + 0x60 /* Entry point for kexec APs */ ap_kexec_start: /* At 0x60 past start, copied to 0x60 by kexec */ /* r3 set to CPU ID by kexec */ /* Invalidate icache for low-memory copy and jump there */ li %r0,0x80 dcbst 0,%r0 sync icbi 0,%r0 isync ba 0x80 /* Absolute branch to next inst */ . = kbootentry + 0x80 /* Aligned to cache line */ 1: or 31,31,31 /* yield */ sync lwz %r1,0x40(0) /* Spin on ap_kexec_spin_sem */ cmpw %r1,%r3 /* Until it equals our CPU ID */ bne 1b /* Released */ or 2,2,2 /* unyield */ /* Make sure that it will be software reset. Clear SRR1 */ li %r1,0 mtsrr1 %r1 ba EXC_RST /* * Now start the real text section */ .text .globl btext btext: /* * Main kernel entry point. * * Calling convention: * r3: Flattened Device Tree pointer (or zero) * r4: ignored * r5: OF client interface pointer (or zero) * r6: Loader metadata pointer (or zero) * r7: Magic cookie (0xfb5d104d) to indicate that r6 has loader metadata */ .text ASENTRY_NOPROF(__start) /* Set 64-bit mode if not yet set before branching to C */ mfmsr %r20 li %r21,1 insrdi %r20,%r21,1,0 mtmsrd %r20 isync nop /* Make this block a multiple of 8 bytes */ /* Set up the TOC pointer */ b 0f .align 3 0: nop bl 1f .llong __tocbase + 0x8000 - . 1: mflr %r2 ld %r1,0(%r2) add %r2,%r1,%r2 /* Get load offset */ ld %r31,-0x8000(%r2) /* First TOC entry is TOC base */ subf %r31,%r31,%r2 /* Subtract from real TOC base to get base */ /* Set up the stack pointer */ bl 1f .llong tmpstk + TMPSTKSZ - 96 - . 1: mflr %r30 ld %r1,0(%r30) add %r1,%r1,%r30 nop /* Relocate kernel */ std %r3,48(%r1) std %r4,56(%r1) std %r5,64(%r1) std %r6,72(%r1) std %r7,80(%r1) bl 1f .llong _DYNAMIC-. 1: mflr %r3 ld %r4,0(%r3) add %r3,%r4,%r3 mr %r4,%r31 bl elf_reloc_self nop ld %r3,48(%r1) ld %r4,56(%r1) ld %r5,64(%r1) ld %r6,72(%r1) ld %r7,80(%r1) /* Begin CPU init */ mr %r4,%r2 /* Replace ignored r4 with tocbase for trap handlers */ bl powerpc_init nop /* Set stack pointer to new value and branch to mi_startup */ mr %r1, %r3 li %r3, 0 std %r3, 0(%r1) bl mi_startup nop /* Unreachable */ b . +ASENTRY_NOPROF(__restartkernel_virtual) + /* + * When coming in via this entry point, we need to alter the SLB to + * shadow the segment register emulation entries in DMAP space. + * We need to do this dance because we are running with virtual-mode + * OpenFirmware and have not yet taken over the MMU. + * + * Assumptions: + * 1) The kernel is currently identity-mapped. + * 2) We are currently executing at an address compatible with + * real mode. + * 3) The first 16 SLB entries are emulating SRs. + * 4) The rest of the SLB is not in use. + * 5) OpenFirmware is not manipulating the SLB at runtime. + * 6) We are running on 64-bit AIM. + * + * Tested on a G5. + */ + mfmsr %r14 + /* Switch to real mode because we are about to mess with the SLB. */ + andi. %r14, %r14, ~(PSL_DR|PSL_IR|PSL_ME|PSL_RI)@l + mtmsr %r14 + isync + /* Prepare variables for later use. */ + li %r14, 0 + li %r18, 0 + oris %r18, %r18, 0xc000 + sldi %r18, %r18, 32 /* r18: 0xc000000000000000 */ +1: + /* + * Loop over the first 16 SLB entries. + * Offset the SLBE into the DMAP, add 16 to the index, and write + * it back to the SLB. + */ + /* XXX add more safety checks */ + slbmfev %r15, %r14 + slbmfee %r16, %r14 + or %r16, %r16, %r14 /* index is 0-15 */ + ori %r16, %r16, 0x10 /* add 16 to index. */ + or %r16, %r16, %r18 /* SLBE DMAP offset */ + rldicr %r17, %r16, 0, 37 /* Invalidation SLBE */ + + isync + slbie %r17 + /* isync */ + slbmte %r15, %r16 + isync + addi %r14, %r14, 1 + cmpdi %r14, 16 + blt 1b + ASENTRY_NOPROF(__restartkernel) /* * r3-r7: arguments to go to __start * r8: offset from current kernel address to apply * r9: MSR to set when (atomically) jumping to __start + r8 */ mtsrr1 %r9 bl 1f 1: mflr %r25 add %r25,%r8,%r25 addi %r25,%r25,2f-1b mtsrr0 %r25 rfid 2: bl __start nop #include Index: head/sys/powerpc/include/db_machdep.h =================================================================== --- head/sys/powerpc/include/db_machdep.h (revision 360793) +++ head/sys/powerpc/include/db_machdep.h (revision 360794) @@ -1,92 +1,88 @@ /*- * Mach Operating System * Copyright (c) 1992 Carnegie Mellon University * All Rights Reserved. * * 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 Mellon * the rights to redistribute these changes. * * $OpenBSD: db_machdep.h,v 1.2 1997/03/21 00:48:48 niklas Exp $ * $NetBSD: db_machdep.h,v 1.4.22.1 2000/08/05 11:10:43 wiz Exp $ * $FreeBSD$ */ /* * Machine-dependent defines for new kernel debugger. */ #ifndef _POWERPC_DB_MACHDEP_H_ #define _POWERPC_DB_MACHDEP_H_ #include #include #define DB_ELF_SYMBOLS #define DB_ELFSIZE __ELF_WORD_SIZE typedef vm_offset_t db_addr_t; /* address - unsigned */ typedef intptr_t db_expr_t; /* expression - signed */ #define PC_REGS(regs) ((db_addr_t)kdb_thrctx->pcb_lr) #define BKPT_INST 0x7C810808 /* breakpoint instruction */ #define BKPT_SIZE (4) /* size of breakpoint inst */ #define BKPT_SET(inst) (BKPT_INST) #define db_clear_single_step kdb_cpu_clear_singlestep #define db_set_single_step kdb_cpu_set_singlestep #if 0 #define SR_SINGLESTEP 0x400 #define db_clear_single_step(regs) ((regs)->msr &= ~SR_SINGLESTEP) #define db_set_single_step(regs) ((regs)->msr |= SR_SINGLESTEP) #endif #define T_BREAKPOINT 0xffff #define IS_BREAKPOINT_TRAP(type, code) ((type) == T_BREAKPOINT) #define T_WATCHPOINT 0xeeee #ifdef T_WATCHPOINT #define IS_WATCHPOINT_TRAP(type, code) ((type) == T_WATCHPOINT) #else #define IS_WATCHPOINT_TRAP(type, code) 0 #endif #define M_RTS 0xfc0007fe #define I_RTS 0x4c000020 #define M_BC 0xfc000000 #define I_BC 0x40000000 #define M_B 0xfc000000 #define I_B 0x50000000 #define M_RFI 0xfc0007fe #define I_RFI 0x4c000064 #define inst_trap_return(ins) (((ins)&M_RFI) == I_RFI) #define inst_return(ins) (((ins)&M_RTS) == I_RTS) #define inst_call(ins) (((ins)&M_BC ) == I_BC || \ ((ins)&M_B ) == I_B ) #define inst_load(ins) 0 #define inst_store(ins) 0 -#ifdef __powerpc64__ -#define DB_STOFFS(offs) ((offs) & ~DMAP_BASE_ADDRESS) -#endif - #endif /* _POWERPC_DB_MACHDEP_H_ */ Index: head/sys/powerpc/include/param.h =================================================================== --- head/sys/powerpc/include/param.h (revision 360793) +++ head/sys/powerpc/include/param.h (revision 360794) @@ -1,138 +1,140 @@ /*- * SPDX-License-Identifier: BSD-4-Clause * * Copyright (c) 2001 David E. O'Brien * Copyright (c) 1990 The Regents of the University of California. * All rights reserved. * * This code is derived from software contributed to Berkeley by * William Jolitz. * * 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. * * from: @(#)param.h 5.8 (Berkeley) 6/28/91 * $FreeBSD$ */ #ifndef _POWERPC_INCLUDE_PARAM_H_ #define _POWERPC_INCLUDE_PARAM_H_ /* * Machine dependent constants for PowerPC */ #include /* Needed to display interrupts on OFW PCI */ #define __PCI_REROUTE_INTERRUPT #ifndef MACHINE #define MACHINE "powerpc" #endif #ifndef MACHINE_ARCH #ifdef __powerpc64__ #define MACHINE_ARCH "powerpc64" #else #ifdef __SPE__ #define MACHINE_ARCH "powerpcspe" #else #define MACHINE_ARCH "powerpc" #endif #endif #endif #define MID_MACHINE MID_POWERPC #ifdef __powerpc64__ #ifndef MACHINE_ARCH32 #define MACHINE_ARCH32 "powerpc" #endif #endif #ifdef SMP #ifndef MAXCPU #define MAXCPU 256 #endif #else #define MAXCPU 1 #endif #ifndef MAXMEMDOM #define MAXMEMDOM 8 #endif #define ALIGNBYTES _ALIGNBYTES #define ALIGN(p) _ALIGN(p) /* * ALIGNED_POINTER is a boolean macro that checks whether an address * is valid to fetch data elements of type t from on this architecture. * This does not reflect the optimal alignment, just the possibility * (within reasonable limits). */ #define ALIGNED_POINTER(p, t) ((((uintptr_t)(p)) & (sizeof (t) - 1)) == 0) /* * CACHE_LINE_SIZE is the compile-time maximum cache line size for an * architecture. It should be used with appropriate caution. */ #define CACHE_LINE_SHIFT 7 #define CACHE_LINE_SIZE (1 << CACHE_LINE_SHIFT) #define PAGE_SHIFT 12 #define PAGE_SIZE (1 << PAGE_SHIFT) /* Page size */ #define PAGE_MASK (PAGE_SIZE - 1) #define NPTEPG (PAGE_SIZE/(sizeof (pt_entry_t))) #define MAXPAGESIZES 1 /* maximum number of supported page sizes */ +#define RELOCATABLE_KERNEL 1 /* kernel may relocate during startup */ + #ifndef KSTACK_PAGES #ifdef __powerpc64__ #define KSTACK_PAGES 8 /* includes pcb */ #else #define KSTACK_PAGES 4 /* includes pcb */ #endif #endif #define KSTACK_GUARD_PAGES 1 /* pages of kstack guard; 0 disables */ #define USPACE (kstack_pages * PAGE_SIZE) /* total size of pcb */ /* * Mach derived conversion macros */ #define trunc_page(x) ((x) & ~(PAGE_MASK)) #define round_page(x) (((x) + PAGE_MASK) & ~PAGE_MASK) #define atop(x) ((x) >> PAGE_SHIFT) #define ptoa(x) ((x) << PAGE_SHIFT) #define powerpc_btop(x) ((x) >> PAGE_SHIFT) #define powerpc_ptob(x) ((x) << PAGE_SHIFT) #define pgtok(x) ((x) * (PAGE_SIZE / 1024UL)) #define btoc(x) ((vm_offset_t)(((x)+PAGE_MASK)>>PAGE_SHIFT)) #endif /* !_POWERPC_INCLUDE_PARAM_H_ */ Index: head/sys/powerpc/ofw/ofw_initrd.c =================================================================== --- head/sys/powerpc/ofw/ofw_initrd.c (revision 360793) +++ head/sys/powerpc/ofw/ofw_initrd.c (revision 360794) @@ -1,100 +1,112 @@ /*- * Copyright (C) 2018 Breno Leitao * * 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 ``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 TOOLS GMBH 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 "opt_md.h" extern u_char *mfs_root; extern int mfs_root_size; static void ofw_initrd_probe_and_attach(void *junk); SYSINIT(ofw_initrd_probe_and_attach, SI_SUB_KMEM, SI_ORDER_ANY, ofw_initrd_probe_and_attach, NULL); static void ofw_initrd_probe_and_attach(void *junk) { phandle_t chosen; vm_paddr_t start, end; pcell_t cell[2]; ssize_t size; + u_char *taste; + Elf_Ehdr ehdr; if (!hw_direct_map) return; chosen = OF_finddevice("/chosen"); if (chosen <= 0) return; if (!OF_hasprop(chosen, "linux,initrd-start") || !OF_hasprop(chosen, "linux,initrd-end")) return; size = OF_getencprop(chosen, "linux,initrd-start", cell, sizeof(cell)); if (size == 4) start = cell[0]; else if (size == 8) start = (uint64_t)cell[0] << 32 | cell[1]; else { printf("ofw_initrd: Wrong linux,initrd-start size\n"); return; } size = OF_getencprop(chosen, "linux,initrd-end", cell, sizeof(cell)); if (size == 4) end = cell[0]; else if (size == 8) end = (uint64_t)cell[0] << 32 | cell[1]; else{ printf("ofw_initrd: Wrong linux,initrd-end size\n"); return; } if (end - start > 0) { - mfs_root = (u_char *) PHYS_TO_DMAP(start); + taste = (u_char*) PHYS_TO_DMAP(start); + memcpy(&ehdr, taste, sizeof(ehdr)); + + if (IS_ELF(ehdr)) { + printf("ofw_initrd: initrd is kernel image!\n"); + return; + } + + mfs_root = taste; mfs_root_size = end - start; printf("ofw_initrd: initrd loaded at 0x%08lx-0x%08lx\n", start, end); } } Index: head/sys/powerpc/powerpc/elf32_machdep.c =================================================================== --- head/sys/powerpc/powerpc/elf32_machdep.c (revision 360793) +++ head/sys/powerpc/powerpc/elf32_machdep.c (revision 360794) @@ -1,459 +1,466 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright 1996-1998 John D. Polstra. * 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 ``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. * * $FreeBSD$ */ #include #include #include #define __ELF_WORD_SIZE 32 #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 __powerpc64__ #include #include extern const char *freebsd32_syscallnames[]; static void ppc32_fixlimit(struct rlimit *rl, int which); static SYSCTL_NODE(_compat, OID_AUTO, ppc32, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, "32-bit mode"); #define PPC32_MAXDSIZ (1024*1024*1024) static u_long ppc32_maxdsiz = PPC32_MAXDSIZ; SYSCTL_ULONG(_compat_ppc32, OID_AUTO, maxdsiz, CTLFLAG_RWTUN, &ppc32_maxdsiz, 0, ""); #define PPC32_MAXSSIZ (64*1024*1024) u_long ppc32_maxssiz = PPC32_MAXSSIZ; SYSCTL_ULONG(_compat_ppc32, OID_AUTO, maxssiz, CTLFLAG_RWTUN, &ppc32_maxssiz, 0, ""); #else static void ppc32_runtime_resolve(void); #endif struct sysentvec elf32_freebsd_sysvec = { .sv_size = SYS_MAXSYSCALL, #ifdef __powerpc64__ .sv_table = freebsd32_sysent, #else .sv_table = sysent, #endif .sv_errsize = 0, .sv_errtbl = NULL, .sv_transtrap = NULL, .sv_fixup = __elfN(freebsd_fixup), .sv_copyout_auxargs = __elfN(powerpc_copyout_auxargs), .sv_sendsig = sendsig, .sv_sigcode = sigcode32, .sv_szsigcode = &szsigcode32, .sv_name = "FreeBSD ELF32", .sv_coredump = __elfN(coredump), .sv_imgact_try = NULL, .sv_minsigstksz = MINSIGSTKSZ, .sv_minuser = VM_MIN_ADDRESS, .sv_stackprot = VM_PROT_ALL, #ifdef __powerpc64__ .sv_maxuser = VM_MAXUSER_ADDRESS32, .sv_usrstack = FREEBSD32_USRSTACK, .sv_psstrings = FREEBSD32_PS_STRINGS, .sv_copyout_strings = freebsd32_copyout_strings, .sv_setregs = ppc32_setregs, .sv_syscallnames = freebsd32_syscallnames, .sv_fixlimit = ppc32_fixlimit, #else .sv_maxuser = VM_MAXUSER_ADDRESS, .sv_usrstack = USRSTACK, .sv_psstrings = PS_STRINGS, .sv_copyout_strings = exec_copyout_strings, .sv_setregs = exec_setregs, .sv_syscallnames = syscallnames, .sv_fixlimit = NULL, #endif .sv_maxssiz = NULL, .sv_flags = SV_ABI_FREEBSD | SV_ILP32 | SV_SHP | SV_ASLR, .sv_set_syscall_retval = cpu_set_syscall_retval, .sv_fetch_syscall_args = cpu_fetch_syscall_args, .sv_shared_page_base = FREEBSD32_SHAREDPAGE, .sv_shared_page_len = PAGE_SIZE, .sv_schedtail = NULL, .sv_thread_detach = NULL, .sv_trap = NULL, .sv_hwcap = &cpu_features, .sv_hwcap2 = &cpu_features2, }; INIT_SYSENTVEC(elf32_sysvec, &elf32_freebsd_sysvec); static Elf32_Brandinfo freebsd_brand_info = { .brand = ELFOSABI_FREEBSD, .machine = EM_PPC, .compat_3_brand = "FreeBSD", .emul_path = NULL, .interp_path = "/libexec/ld-elf.so.1", .sysvec = &elf32_freebsd_sysvec, #ifdef __powerpc64__ .interp_newpath = "/libexec/ld-elf32.so.1", #else .interp_newpath = NULL, #endif .brand_note = &elf32_freebsd_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; SYSINIT(elf32, SI_SUB_EXEC, SI_ORDER_FIRST, (sysinit_cfunc_t) elf32_insert_brand_entry, &freebsd_brand_info); static Elf32_Brandinfo freebsd_brand_oinfo = { .brand = ELFOSABI_FREEBSD, .machine = EM_PPC, .compat_3_brand = "FreeBSD", .emul_path = NULL, .interp_path = "/usr/libexec/ld-elf.so.1", .sysvec = &elf32_freebsd_sysvec, .interp_newpath = NULL, .brand_note = &elf32_freebsd_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE }; SYSINIT(oelf32, SI_SUB_EXEC, SI_ORDER_ANY, (sysinit_cfunc_t) elf32_insert_brand_entry, &freebsd_brand_oinfo); void elf_reloc_self(Elf_Dyn *dynp, Elf_Addr relocbase); void elf32_dump_thread(struct thread *td, void *dst, size_t *off) { size_t len; struct pcb *pcb; uint64_t vshr[32]; uint64_t *vsr_dw1; int vsr_idx; len = 0; pcb = td->td_pcb; if (pcb->pcb_flags & PCB_VEC) { save_vec_nodrop(td); if (dst != NULL) { len += elf32_populate_note(NT_PPC_VMX, &pcb->pcb_vec, (char *)dst + len, sizeof(pcb->pcb_vec), NULL); } else len += elf32_populate_note(NT_PPC_VMX, NULL, NULL, sizeof(pcb->pcb_vec), NULL); } if (pcb->pcb_flags & PCB_VSX) { save_fpu_nodrop(td); if (dst != NULL) { /* * Doubleword 0 of VSR0-VSR31 overlap with FPR0-FPR31 and * VSR32-VSR63 overlap with VR0-VR31, so we only copy * the non-overlapping data, which is doubleword 1 of VSR0-VSR31. */ for (vsr_idx = 0; vsr_idx < nitems(vshr); vsr_idx++) { vsr_dw1 = (uint64_t *)&pcb->pcb_fpu.fpr[vsr_idx].vsr[2]; vshr[vsr_idx] = *vsr_dw1; } len += elf32_populate_note(NT_PPC_VSX, vshr, (char *)dst + len, sizeof(vshr), NULL); } else len += elf32_populate_note(NT_PPC_VSX, NULL, NULL, sizeof(vshr), NULL); } *off = len; } #ifndef __powerpc64__ bool -elf_is_ifunc_reloc(Elf_Size r_info __unused) +elf_is_ifunc_reloc(Elf_Size r_info) { - return (false); + return (ELF_R_TYPE(r_info) == R_PPC_IRELATIVE); } /* Process one elf relocation with addend. */ static int elf_reloc_internal(linker_file_t lf, Elf_Addr relocbase, const void *data, int type, int local, elf_lookup_fn lookup) { Elf_Addr *where; Elf_Half *hwhere; Elf_Addr addr; - Elf_Addr addend; + Elf_Addr addend, val; Elf_Word rtype, symidx; const Elf_Rela *rela; int error; switch (type) { case ELF_RELOC_REL: panic("PPC only supports RELA relocations"); break; case ELF_RELOC_RELA: rela = (const Elf_Rela *)data; where = (Elf_Addr *) ((uintptr_t)relocbase + rela->r_offset); hwhere = (Elf_Half *) ((uintptr_t)relocbase + rela->r_offset); addend = rela->r_addend; rtype = ELF_R_TYPE(rela->r_info); symidx = ELF_R_SYM(rela->r_info); break; default: panic("elf_reloc: unknown relocation mode %d\n", type); } switch (rtype) { case R_PPC_NONE: break; case R_PPC_ADDR32: /* word32 S + A */ error = lookup(lf, symidx, 1, &addr); if (error != 0) return -1; *where = elf_relocaddr(lf, addr + addend); break; case R_PPC_ADDR16_LO: /* #lo(S) */ error = lookup(lf, symidx, 1, &addr); if (error != 0) return -1; /* * addend values are sometimes relative to sections * (i.e. .rodata) in rela, where in reality they * are relative to relocbase. Detect this condition. */ if (addr > relocbase && addr <= (relocbase + addend)) addr = relocbase; addr = elf_relocaddr(lf, addr + addend); *hwhere = addr & 0xffff; break; case R_PPC_ADDR16_HA: /* #ha(S) */ error = lookup(lf, symidx, 1, &addr); if (error != 0) return -1; /* * addend values are sometimes relative to sections * (i.e. .rodata) in rela, where in reality they * are relative to relocbase. Detect this condition. */ if (addr > relocbase && addr <= (relocbase + addend)) addr = relocbase; addr = elf_relocaddr(lf, addr + addend); *hwhere = ((addr >> 16) + ((addr & 0x8000) ? 1 : 0)) & 0xffff; break; case R_PPC_RELATIVE: /* word32 B + A */ *where = elf_relocaddr(lf, relocbase + addend); break; case R_PPC_JMP_SLOT: /* PLT jump slot entry */ /* * We currently only support Secure-PLT jump slots. * Given that we reject BSS-PLT modules during load, we * don't need to check again. * The method we are using here is equivilent to * LD_BIND_NOW. */ error = lookup(lf, symidx, 1, &addr); if (error != 0) return -1; *where = elf_relocaddr(lf, addr + addend); + break; + + case R_PPC_IRELATIVE: + addr = relocbase + addend; + val = ((Elf32_Addr (*)(void))addr)(); + if (*where != val) + *where = val; break; default: printf("kldload: unexpected relocation type %d\n", (int) rtype); return -1; } return(0); } void elf_reloc_self(Elf_Dyn *dynp, Elf_Addr relocbase) { Elf_Rela *rela = NULL, *relalim; Elf_Addr relasz = 0; Elf_Addr *where; /* * Extract the rela/relasz values from the dynamic section */ for (; dynp->d_tag != DT_NULL; dynp++) { switch (dynp->d_tag) { case DT_RELA: rela = (Elf_Rela *)(relocbase+dynp->d_un.d_ptr); break; case DT_RELASZ: relasz = dynp->d_un.d_val; break; } } /* * Relocate these values */ relalim = (Elf_Rela *)((caddr_t)rela + relasz); for (; rela < relalim; rela++) { if (ELF_R_TYPE(rela->r_info) != R_PPC_RELATIVE) continue; where = (Elf_Addr *)(relocbase + rela->r_offset); *where = (Elf_Addr)(relocbase + rela->r_addend); } } int elf_reloc(linker_file_t lf, Elf_Addr relocbase, const void *data, int type, elf_lookup_fn lookup) { return (elf_reloc_internal(lf, relocbase, data, type, 0, lookup)); } int elf_reloc_local(linker_file_t lf, Elf_Addr relocbase, const void *data, int type, elf_lookup_fn lookup) { return (elf_reloc_internal(lf, relocbase, data, type, 1, lookup)); } int elf_cpu_load_file(linker_file_t lf) { /* Only sync the cache for non-kernel modules */ if (lf->id != 1) __syncicache(lf->address, lf->size); return (0); } int elf_cpu_unload_file(linker_file_t lf __unused) { return (0); } static void ppc32_runtime_resolve() { /* * Since we don't support lazy binding, panic immediately if anyone * manages to call the runtime resolver. */ panic("kldload: Runtime resolver was called unexpectedly!"); } int elf_cpu_parse_dynamic(caddr_t loadbase, Elf_Dyn *dynamic) { Elf_Dyn *dp; bool has_plt = false; bool secure_plt = false; Elf_Addr *got; for (dp = dynamic; dp->d_tag != DT_NULL; dp++) { switch (dp->d_tag) { case DT_PPC_GOT: secure_plt = true; got = (Elf_Addr *)(loadbase + dp->d_un.d_ptr); /* Install runtime resolver canary. */ got[1] = (Elf_Addr)ppc32_runtime_resolve; got[2] = (Elf_Addr)0; break; case DT_PLTGOT: has_plt = true; break; } } if (has_plt && !secure_plt) { printf("kldload: BSS-PLT modules are not supported.\n"); return (-1); } return (0); } #endif #ifdef __powerpc64__ static void ppc32_fixlimit(struct rlimit *rl, int which) { switch (which) { case RLIMIT_DATA: if (ppc32_maxdsiz != 0) { if (rl->rlim_cur > ppc32_maxdsiz) rl->rlim_cur = ppc32_maxdsiz; if (rl->rlim_max > ppc32_maxdsiz) rl->rlim_max = ppc32_maxdsiz; } break; case RLIMIT_STACK: if (ppc32_maxssiz != 0) { if (rl->rlim_cur > ppc32_maxssiz) rl->rlim_cur = ppc32_maxssiz; if (rl->rlim_max > ppc32_maxssiz) rl->rlim_max = ppc32_maxssiz; } break; } } #endif Index: head/sys/powerpc/powerpc/elf64_machdep.c =================================================================== --- head/sys/powerpc/powerpc/elf64_machdep.c (revision 360793) +++ head/sys/powerpc/powerpc/elf64_machdep.c (revision 360794) @@ -1,423 +1,430 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright 1996-1998 John D. Polstra. * 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 ``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. * * $FreeBSD$ */ #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 static void exec_setregs_funcdesc(struct thread *td, struct image_params *imgp, uintptr_t stack); struct sysentvec elf64_freebsd_sysvec_v1 = { .sv_size = SYS_MAXSYSCALL, .sv_table = sysent, .sv_errsize = 0, .sv_errtbl = NULL, .sv_transtrap = NULL, .sv_fixup = __elfN(freebsd_fixup), .sv_sendsig = sendsig, .sv_sigcode = sigcode64, .sv_szsigcode = &szsigcode64, .sv_name = "FreeBSD ELF64", .sv_coredump = __elfN(coredump), .sv_imgact_try = NULL, .sv_minsigstksz = MINSIGSTKSZ, .sv_minuser = VM_MIN_ADDRESS, .sv_maxuser = VM_MAXUSER_ADDRESS, .sv_usrstack = USRSTACK, .sv_psstrings = PS_STRINGS, .sv_stackprot = VM_PROT_ALL, .sv_copyout_auxargs = __elfN(powerpc_copyout_auxargs), .sv_copyout_strings = exec_copyout_strings, .sv_setregs = exec_setregs_funcdesc, .sv_fixlimit = NULL, .sv_maxssiz = NULL, .sv_flags = SV_ABI_FREEBSD | SV_LP64 | SV_SHP | SV_ASLR, .sv_set_syscall_retval = cpu_set_syscall_retval, .sv_fetch_syscall_args = cpu_fetch_syscall_args, .sv_syscallnames = syscallnames, .sv_shared_page_base = SHAREDPAGE, .sv_shared_page_len = PAGE_SIZE, .sv_schedtail = NULL, .sv_thread_detach = NULL, .sv_trap = NULL, .sv_hwcap = &cpu_features, .sv_hwcap2 = &cpu_features2, }; INIT_SYSENTVEC(elf64_sysvec_v1, &elf64_freebsd_sysvec_v1); struct sysentvec elf64_freebsd_sysvec_v2 = { .sv_size = SYS_MAXSYSCALL, .sv_table = sysent, .sv_errsize = 0, .sv_errtbl = NULL, .sv_transtrap = NULL, .sv_fixup = __elfN(freebsd_fixup), .sv_sendsig = sendsig, .sv_sigcode = sigcode64_elfv2, .sv_szsigcode = &szsigcode64_elfv2, .sv_name = "FreeBSD ELF64 V2", .sv_coredump = __elfN(coredump), .sv_imgact_try = NULL, .sv_minsigstksz = MINSIGSTKSZ, .sv_minuser = VM_MIN_ADDRESS, .sv_maxuser = VM_MAXUSER_ADDRESS, .sv_usrstack = USRSTACK, .sv_psstrings = PS_STRINGS, .sv_stackprot = VM_PROT_ALL, .sv_copyout_auxargs = __elfN(powerpc_copyout_auxargs), .sv_copyout_strings = exec_copyout_strings, .sv_setregs = exec_setregs, .sv_fixlimit = NULL, .sv_maxssiz = NULL, .sv_flags = SV_ABI_FREEBSD | SV_LP64 | SV_SHP, .sv_set_syscall_retval = cpu_set_syscall_retval, .sv_fetch_syscall_args = cpu_fetch_syscall_args, .sv_syscallnames = syscallnames, .sv_shared_page_base = SHAREDPAGE, .sv_shared_page_len = PAGE_SIZE, .sv_schedtail = NULL, .sv_thread_detach = NULL, .sv_trap = NULL, .sv_hwcap = &cpu_features, .sv_hwcap2 = &cpu_features2, }; INIT_SYSENTVEC(elf64_sysvec_v2, &elf64_freebsd_sysvec_v2); static boolean_t ppc64_elfv1_header_match(struct image_params *params); static boolean_t ppc64_elfv2_header_match(struct image_params *params); static Elf64_Brandinfo freebsd_brand_info_elfv1 = { .brand = ELFOSABI_FREEBSD, .machine = EM_PPC64, .compat_3_brand = "FreeBSD", .emul_path = NULL, .interp_path = "/libexec/ld-elf.so.1", .sysvec = &elf64_freebsd_sysvec_v1, .interp_newpath = NULL, .brand_note = &elf64_freebsd_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE, .header_supported = &ppc64_elfv1_header_match }; SYSINIT(elf64v1, SI_SUB_EXEC, SI_ORDER_ANY, (sysinit_cfunc_t) elf64_insert_brand_entry, &freebsd_brand_info_elfv1); static Elf64_Brandinfo freebsd_brand_info_elfv2 = { .brand = ELFOSABI_FREEBSD, .machine = EM_PPC64, .compat_3_brand = "FreeBSD", .emul_path = NULL, .interp_path = "/libexec/ld-elf.so.1", .sysvec = &elf64_freebsd_sysvec_v2, .interp_newpath = NULL, .brand_note = &elf64_freebsd_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE, .header_supported = &ppc64_elfv2_header_match }; SYSINIT(elf64v2, SI_SUB_EXEC, SI_ORDER_ANY, (sysinit_cfunc_t) elf64_insert_brand_entry, &freebsd_brand_info_elfv2); static Elf64_Brandinfo freebsd_brand_oinfo = { .brand = ELFOSABI_FREEBSD, .machine = EM_PPC64, .compat_3_brand = "FreeBSD", .emul_path = NULL, .interp_path = "/usr/libexec/ld-elf.so.1", .sysvec = &elf64_freebsd_sysvec_v1, .interp_newpath = NULL, .brand_note = &elf64_freebsd_brandnote, .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE, .header_supported = &ppc64_elfv1_header_match }; SYSINIT(oelf64, SI_SUB_EXEC, SI_ORDER_ANY, (sysinit_cfunc_t) elf64_insert_brand_entry, &freebsd_brand_oinfo); void elf_reloc_self(Elf_Dyn *dynp, Elf_Addr relocbase); static boolean_t ppc64_elfv1_header_match(struct image_params *params) { const Elf64_Ehdr *hdr = (const Elf64_Ehdr *)params->image_header; int abi = (hdr->e_flags & 3); return (abi == 0 || abi == 1); } static boolean_t ppc64_elfv2_header_match(struct image_params *params) { const Elf64_Ehdr *hdr = (const Elf64_Ehdr *)params->image_header; int abi = (hdr->e_flags & 3); return (abi == 2); } static void exec_setregs_funcdesc(struct thread *td, struct image_params *imgp, uintptr_t stack) { struct trapframe *tf; register_t entry_desc[3]; tf = trapframe(td); exec_setregs(td, imgp, stack); /* * For 64-bit ELFv1, we need to disentangle the function * descriptor * * 0. entry point * 1. TOC value (r2) * 2. Environment pointer (r11) */ (void)copyin((void *)imgp->entry_addr, entry_desc, sizeof(entry_desc)); tf->srr0 = entry_desc[0] + imgp->reloc_base; tf->fixreg[2] = entry_desc[1] + imgp->reloc_base; tf->fixreg[11] = entry_desc[2] + imgp->reloc_base; } void elf64_dump_thread(struct thread *td, void *dst, size_t *off) { size_t len; struct pcb *pcb; uint64_t vshr[32]; uint64_t *vsr_dw1; int vsr_idx; len = 0; pcb = td->td_pcb; if (pcb->pcb_flags & PCB_VEC) { save_vec_nodrop(td); if (dst != NULL) { len += elf64_populate_note(NT_PPC_VMX, &pcb->pcb_vec, (char *)dst + len, sizeof(pcb->pcb_vec), NULL); } else len += elf64_populate_note(NT_PPC_VMX, NULL, NULL, sizeof(pcb->pcb_vec), NULL); } if (pcb->pcb_flags & PCB_VSX) { save_fpu_nodrop(td); if (dst != NULL) { /* * Doubleword 0 of VSR0-VSR31 overlap with FPR0-FPR31 and * VSR32-VSR63 overlap with VR0-VR31, so we only copy * the non-overlapping data, which is doubleword 1 of VSR0-VSR31. */ for (vsr_idx = 0; vsr_idx < nitems(vshr); vsr_idx++) { vsr_dw1 = (uint64_t *)&pcb->pcb_fpu.fpr[vsr_idx].vsr[2]; vshr[vsr_idx] = *vsr_dw1; } len += elf64_populate_note(NT_PPC_VSX, vshr, (char *)dst + len, sizeof(vshr), NULL); } else len += elf64_populate_note(NT_PPC_VSX, NULL, NULL, sizeof(vshr), NULL); } *off = len; } bool -elf_is_ifunc_reloc(Elf_Size r_info __unused) +elf_is_ifunc_reloc(Elf_Size r_info) { - return (false); + return (ELF_R_TYPE(r_info) == R_PPC_IRELATIVE); } /* Process one elf relocation with addend. */ static int elf_reloc_internal(linker_file_t lf, Elf_Addr relocbase, const void *data, int type, int local, elf_lookup_fn lookup) { Elf_Addr *where; Elf_Addr addr; - Elf_Addr addend; + Elf_Addr addend, val; Elf_Word rtype, symidx; const Elf_Rela *rela; int error; switch (type) { case ELF_RELOC_REL: panic("PPC only supports RELA relocations"); break; case ELF_RELOC_RELA: rela = (const Elf_Rela *)data; where = (Elf_Addr *) (relocbase + rela->r_offset); addend = rela->r_addend; rtype = ELF_R_TYPE(rela->r_info); symidx = ELF_R_SYM(rela->r_info); break; default: panic("elf_reloc: unknown relocation mode %d\n", type); } switch (rtype) { case R_PPC_NONE: break; case R_PPC64_ADDR64: /* doubleword64 S + A */ error = lookup(lf, symidx, 1, &addr); if (error != 0) return -1; addr += addend; *where = addr; break; case R_PPC_RELATIVE: /* doubleword64 B + A */ *where = elf_relocaddr(lf, relocbase + addend); break; case R_PPC_JMP_SLOT: /* function descriptor copy */ lookup(lf, symidx, 1, &addr); #if !defined(_CALL_ELF) || _CALL_ELF == 1 memcpy(where, (Elf_Addr *)addr, 3*sizeof(Elf_Addr)); #else *where = addr; #endif __asm __volatile("dcbst 0,%0; sync" :: "r"(where) : "memory"); + break; + + case R_PPC_IRELATIVE: + addr = relocbase + addend; + val = ((Elf64_Addr (*)(void))addr)(); + if (*where != val) + *where = val; break; default: printf("kldload: unexpected relocation type %d\n", (int) rtype); return -1; } return(0); } void elf_reloc_self(Elf_Dyn *dynp, Elf_Addr relocbase) { Elf_Rela *rela = NULL, *relalim; Elf_Addr relasz = 0; Elf_Addr *where; /* * Extract the rela/relasz values from the dynamic section */ for (; dynp->d_tag != DT_NULL; dynp++) { switch (dynp->d_tag) { case DT_RELA: rela = (Elf_Rela *)(relocbase+dynp->d_un.d_ptr); break; case DT_RELASZ: relasz = dynp->d_un.d_val; break; } } /* * Relocate these values */ relalim = (Elf_Rela *)((caddr_t)rela + relasz); for (; rela < relalim; rela++) { if (ELF_R_TYPE(rela->r_info) != R_PPC_RELATIVE) continue; where = (Elf_Addr *)(relocbase + rela->r_offset); *where = (Elf_Addr)(relocbase + rela->r_addend); } } int elf_reloc(linker_file_t lf, Elf_Addr relocbase, const void *data, int type, elf_lookup_fn lookup) { return (elf_reloc_internal(lf, relocbase, data, type, 0, lookup)); } int elf_reloc_local(linker_file_t lf, Elf_Addr relocbase, const void *data, int type, elf_lookup_fn lookup) { return (elf_reloc_internal(lf, relocbase, data, type, 1, lookup)); } int elf_cpu_load_file(linker_file_t lf) { /* Only sync the cache for non-kernel modules */ if (lf->id != 1) __syncicache(lf->address, lf->size); return (0); } int elf_cpu_unload_file(linker_file_t lf __unused) { return (0); } int elf_cpu_parse_dynamic(caddr_t loadbase __unused, Elf_Dyn *dynamic __unused) { return (0); } Index: head/sys/powerpc/powerpc/machdep.c =================================================================== --- head/sys/powerpc/powerpc/machdep.c (revision 360793) +++ head/sys/powerpc/powerpc/machdep.c (revision 360794) @@ -1,655 +1,884 @@ /*- * Copyright (C) 1995, 1996 Wolfgang Solfrank. * Copyright (C) 1995, 1996 TooLs GmbH. * 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 TooLs GmbH. * 4. The name of TooLs GmbH may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``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 TOOLS GMBH 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. */ /*- * Copyright (C) 2001 Benno Rice * 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 Benno Rice ``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 TOOLS GMBH 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: machdep.c,v 1.74.2.1 2000/11/01 16:13:48 tv Exp $ */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include "opt_kstack_pages.h" #include "opt_platform.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 #include #include #include #include #include #include #include #include #include #ifndef __powerpc64__ #include #endif #include #include #include #include +#include #include #include #include #include #include #include #include #include #include #include #include #include #include #include int cold = 1; #ifdef __powerpc64__ int cacheline_size = 128; #else int cacheline_size = 32; #endif int hw_direct_map = 1; #ifdef BOOKE extern vm_paddr_t kernload; #endif extern void *ap_pcpu; struct pcpu __pcpu[MAXCPU]; static char init_kenv[2048]; static struct trapframe frame0; char machine[] = "powerpc"; SYSCTL_STRING(_hw, HW_MACHINE, machine, CTLFLAG_RD, machine, 0, ""); static void cpu_startup(void *); SYSINIT(cpu, SI_SUB_CPU, SI_ORDER_FIRST, cpu_startup, NULL); SYSCTL_INT(_machdep, CPU_CACHELINE, cacheline_size, CTLFLAG_RD, &cacheline_size, 0, ""); uintptr_t powerpc_init(vm_offset_t, vm_offset_t, vm_offset_t, void *, uint32_t); +static void fake_preload_metadata(void); + long Maxmem = 0; long realmem = 0; /* Default MSR values set in the AIM/Book-E early startup code */ register_t psl_kernset; register_t psl_userset; register_t psl_userstatic; #ifdef __powerpc64__ register_t psl_userset32; #endif struct kva_md_info kmi; static void cpu_startup(void *dummy) { /* * Initialise the decrementer-based clock. */ decr_init(); /* * Good {morning,afternoon,evening,night}. */ cpu_setup(PCPU_GET(cpuid)); #ifdef PERFMON perfmon_init(); #endif printf("real memory = %ju (%ju MB)\n", ptoa((uintmax_t)physmem), ptoa((uintmax_t)physmem) / 1048576); realmem = physmem; if (bootverbose) printf("available KVA = %zu (%zu MB)\n", virtual_end - virtual_avail, (virtual_end - virtual_avail) / 1048576); /* * Display any holes after the first chunk of extended memory. */ if (bootverbose) { int indx; printf("Physical memory chunk(s):\n"); for (indx = 0; phys_avail[indx + 1] != 0; indx += 2) { vm_paddr_t size1 = phys_avail[indx + 1] - phys_avail[indx]; #ifdef __powerpc64__ printf("0x%016jx - 0x%016jx, %ju bytes (%ju pages)\n", #else printf("0x%09jx - 0x%09jx, %ju bytes (%ju pages)\n", #endif (uintmax_t)phys_avail[indx], (uintmax_t)phys_avail[indx + 1] - 1, (uintmax_t)size1, (uintmax_t)size1 / PAGE_SIZE); } } vm_ksubmap_init(&kmi); printf("avail memory = %ju (%ju MB)\n", ptoa((uintmax_t)vm_free_count()), ptoa((uintmax_t)vm_free_count()) / 1048576); /* * Set up buffers, so they can be used to read disk labels. */ bufinit(); vm_pager_bufferinit(); } extern vm_offset_t __startkernel, __endkernel; extern unsigned char __bss_start[]; extern unsigned char __sbss_start[]; extern unsigned char __sbss_end[]; extern unsigned char _end[]; void aim_early_init(vm_offset_t fdt, vm_offset_t toc, vm_offset_t ofentry, void *mdp, uint32_t mdp_cookie); void aim_cpu_init(vm_offset_t toc); void booke_cpu_init(void); +#ifdef DDB +static void load_external_symtab(void); +static void displace_symbol_table(vm_offset_t, vm_offset_t, vm_offset_t); +#endif + uintptr_t powerpc_init(vm_offset_t fdt, vm_offset_t toc, vm_offset_t ofentry, void *mdp, uint32_t mdp_cookie) { struct pcpu *pc; struct cpuref bsp; vm_offset_t startkernel, endkernel; char *env; + void *kmdp = NULL; bool ofw_bootargs = false; + bool symbols_provided = false; #ifdef DDB vm_offset_t ksym_start; vm_offset_t ksym_end; + vm_offset_t ksym_sz; #endif /* First guess at start/end kernel positions */ startkernel = __startkernel; endkernel = __endkernel; /* * If the metadata pointer cookie is not set to the magic value, * the number in mdp should be treated as nonsense. */ if (mdp_cookie != 0xfb5d104d) mdp = NULL; #if !defined(BOOKE) /* * On BOOKE the BSS is already cleared and some variables * initialized. Do not wipe them out. */ bzero(__sbss_start, __sbss_end - __sbss_start); bzero(__bss_start, _end - __bss_start); #endif cpu_feature_setup(); #ifdef AIM aim_early_init(fdt, toc, ofentry, mdp, mdp_cookie); #endif /* + * At this point, we are executing in our correct memory space. + * Book-E started there, and AIM has done an rfi and restarted + * execution from _start. + * + * We may still be in real mode, however. If we are running out of + * the direct map on 64 bit, this is possible to do. + */ + + /* * Parse metadata if present and fetch parameters. Must be done * before console is inited so cninit gets the right value of * boothowto. */ if (mdp != NULL) { - void *kmdp = NULL; + /* + * Starting up from loader. + * + * Full metadata has been provided, but we need to figure + * out the correct address to relocate it to. + */ char *envp = NULL; uintptr_t md_offset = 0; - vm_paddr_t kernelendphys; + vm_paddr_t kernelstartphys, kernelendphys; #ifdef AIM if ((uintptr_t)&powerpc_init > DMAP_BASE_ADDRESS) md_offset = DMAP_BASE_ADDRESS; #else /* BOOKE */ md_offset = VM_MIN_KERNEL_ADDRESS - kernload; #endif preload_metadata = mdp; if (md_offset > 0) { + /* Translate phys offset into DMAP offset. */ preload_metadata += md_offset; preload_bootstrap_relocate(md_offset); } kmdp = preload_search_by_type("elf kernel"); if (kmdp != NULL) { boothowto = MD_FETCH(kmdp, MODINFOMD_HOWTO, int); envp = MD_FETCH(kmdp, MODINFOMD_ENVP, char *); if (envp != NULL) envp += md_offset; init_static_kenv(envp, 0); if (fdt == 0) { fdt = MD_FETCH(kmdp, MODINFOMD_DTBP, uintptr_t); if (fdt != 0) fdt += md_offset; } + kernelstartphys = MD_FETCH(kmdp, MODINFO_ADDR, + vm_offset_t); + /* kernelstartphys is already relocated. */ kernelendphys = MD_FETCH(kmdp, MODINFOMD_KERNEND, vm_offset_t); if (kernelendphys != 0) kernelendphys += md_offset; endkernel = ulmax(endkernel, kernelendphys); #ifdef DDB ksym_start = MD_FETCH(kmdp, MODINFOMD_SSYM, uintptr_t); ksym_end = MD_FETCH(kmdp, MODINFOMD_ESYM, uintptr_t); + ksym_sz = *(Elf_Size*)ksym_start; + + /* + * Loader already handled displacing to the load + * address, but we still need to displace it to the + * DMAP. + */ + displace_symbol_table( + (vm_offset_t)(ksym_start + sizeof(Elf_Size)), + ksym_sz, md_offset); + db_fetch_ksymtab(ksym_start, ksym_end); + symbols_provided = true; #endif } } else { + /* + * Self-loading kernel, we have to fake up metadata. + * + * Since we are creating the metadata from the final + * memory space, we don't need to call + * preload_boostrap_relocate(). + */ + fake_preload_metadata(); + kmdp = preload_search_by_type("elf kernel"); init_static_kenv(init_kenv, sizeof(init_kenv)); ofw_bootargs = true; } + /* Store boot environment state */ OF_initial_setup((void *)fdt, NULL, (int (*)(void *))ofentry); /* * Init params/tunables that can be overridden by the loader */ init_param1(); /* * Start initializing proc0 and thread0. */ proc_linkup0(&proc0, &thread0); thread0.td_frame = &frame0; #ifdef __powerpc64__ __asm __volatile("mr 13,%0" :: "r"(&thread0)); #else __asm __volatile("mr 2,%0" :: "r"(&thread0)); #endif /* * Init mutexes, which we use heavily in PMAP */ mutex_init(); /* * Install the OF client interface */ OF_bootstrap(); +#ifdef DDB + if (!symbols_provided && hw_direct_map) + load_external_symtab(); +#endif + if (ofw_bootargs) ofw_parse_bootargs(); /* * Initialize the console before printing anything. */ cninit(); #ifdef AIM aim_cpu_init(toc); #else /* BOOKE */ booke_cpu_init(); /* Make sure the kernel icache is valid before we go too much further */ __syncicache((caddr_t)startkernel, endkernel - startkernel); #endif /* * Choose a platform module so we can get the physical memory map. */ platform_probe_and_attach(); /* * Set up per-cpu data for the BSP now that the platform can tell * us which that is. */ if (platform_smp_get_bsp(&bsp) != 0) bsp.cr_cpuid = 0; pc = &__pcpu[bsp.cr_cpuid]; __asm __volatile("mtsprg 0, %0" :: "r"(pc)); pcpu_init(pc, bsp.cr_cpuid, sizeof(struct pcpu)); pc->pc_curthread = &thread0; thread0.td_oncpu = bsp.cr_cpuid; pc->pc_cpuid = bsp.cr_cpuid; pc->pc_hwref = bsp.cr_hwref; /* * Init KDB */ kdb_init(); /* * Bring up MMU */ pmap_bootstrap(startkernel, endkernel); mtmsr(psl_kernset & ~PSL_EE); + link_elf_ireloc(kmdp); /* * Initialize params/tunables that are derived from memsize */ init_param2(physmem); /* * Grab booted kernel's name */ env = kern_getenv("kernelname"); if (env != NULL) { strlcpy(kernelname, env, sizeof(kernelname)); freeenv(env); } /* * Finish setting up thread0. */ thread0.td_pcb = (struct pcb *) ((thread0.td_kstack + thread0.td_kstack_pages * PAGE_SIZE - sizeof(struct pcb)) & ~15UL); bzero((void *)thread0.td_pcb, sizeof(struct pcb)); pc->pc_curpcb = thread0.td_pcb; /* Initialise the message buffer. */ msgbufinit(msgbufp, msgbufsize); #ifdef KDB if (boothowto & RB_KDB) kdb_enter(KDB_WHY_BOOTFLAGS, "Boot flags requested debugger"); #endif return (((uintptr_t)thread0.td_pcb - (sizeof(struct callframe) - 3*sizeof(register_t))) & ~15UL); +} + +#ifdef DDB +/* + * XXX Figure out where to move this. + */ +static void +displace_symbol_table(vm_offset_t ksym_start, + vm_offset_t ksym_sz, vm_offset_t displacement) { + Elf_Sym *sym; + + /* + * Relocate the symbol table to our final load address. + */ + for (sym = (Elf_Sym *)ksym_start; + (vm_paddr_t)sym < (ksym_start + ksym_sz); + sym++) { + if (sym->st_name == 0 || + sym->st_shndx == SHN_UNDEF || + sym->st_value == 0) + continue; + if (ELF_ST_TYPE(sym->st_info) != STT_OBJECT && + ELF_ST_TYPE(sym->st_info) != STT_FUNC && + ELF_ST_TYPE(sym->st_info) != STT_NOTYPE) + continue; + /* Skip relocating any implausible symbols */ + if (sym->st_value > KERNBASE) + sym->st_value += displacement; + } +} + +/* + * On powernv, we might not have symbols loaded via loader. However, if the + * user passed the kernel in as the initrd as well, we can manually load it + * via reinterpreting the initrd copy of the kernel. + */ +static void +load_external_symtab(void) { + phandle_t chosen; + vm_paddr_t start, end; + pcell_t cell[2]; + ssize_t size; + u_char *kernelimg; + + int i; + + Elf_Ehdr *ehdr; + Elf_Phdr *phdr; + Elf_Shdr *shdr; + + vm_offset_t ksym_start, ksym_sz, kstr_start, kstr_sz; + + if (!hw_direct_map) + return; + + chosen = OF_finddevice("/chosen"); + if (chosen <= 0) + return; + + if (!OF_hasprop(chosen, "linux,initrd-start") || + !OF_hasprop(chosen, "linux,initrd-end")) + return; + + size = OF_getencprop(chosen, "linux,initrd-start", cell, sizeof(cell)); + if (size == 4) + start = cell[0]; + else if (size == 8) + start = (uint64_t)cell[0] << 32 | cell[1]; + else + return; + + size = OF_getencprop(chosen, "linux,initrd-end", cell, sizeof(cell)); + if (size == 4) + end = cell[0]; + else if (size == 8) + end = (uint64_t)cell[0] << 32 | cell[1]; + else + return; + + if (!(end - start > 0)) + return; + + kernelimg = (u_char *) PHYS_TO_DMAP(start); + + ehdr = (Elf_Ehdr *)kernelimg; + + if (!IS_ELF(*ehdr)) + return; + + phdr = (Elf_Phdr *)(kernelimg + ehdr->e_phoff); + shdr = (Elf_Shdr *)(kernelimg + ehdr->e_shoff); + + ksym_start = 0; + ksym_sz = 0; + kstr_start = 0; + kstr_sz = 0; + for (i = 0; i < ehdr->e_shnum; i++) { + if (shdr[i].sh_type == SHT_SYMTAB) { + ksym_start = (vm_offset_t)(kernelimg + + shdr[i].sh_offset); + ksym_sz = (vm_offset_t)(shdr[i].sh_size); + kstr_start = (vm_offset_t)(kernelimg + + shdr[shdr[i].sh_link].sh_offset); + kstr_sz = (vm_offset_t) + (shdr[shdr[i].sh_link].sh_size); + } + } + + if (ksym_start != 0 && kstr_start != 0 && ksym_sz != 0 && + kstr_sz != 0 && ksym_start < kstr_start) { + + displace_symbol_table(ksym_start, ksym_sz, + (__startkernel - KERNBASE)); + ksymtab = ksym_start; + ksymtab_size = ksym_sz; + kstrtab = kstr_start; + } + +}; +#endif + +/* + * When not being loaded from loader, we need to create our own metadata + * so we can interact with the kernel linker. + */ +static void +fake_preload_metadata(void) { + /* We depend on dword alignment here. */ + static uint32_t fake_preload[36] __aligned(8); + int i = 0; + + fake_preload[i++] = MODINFO_NAME; + fake_preload[i++] = strlen("kernel") + 1; + strcpy((char*)&fake_preload[i], "kernel"); + /* ['k' 'e' 'r' 'n'] ['e' 'l' '\0' ..] */ + i += 2; + + fake_preload[i++] = MODINFO_TYPE; + fake_preload[i++] = strlen("elf kernel") + 1; + strcpy((char*)&fake_preload[i], "elf kernel"); + /* ['e' 'l' 'f' ' '] ['k' 'e' 'r' 'n'] ['e' 'l' '\0' ..] */ + i += 3; + +#ifdef __powerpc64__ + /* Padding -- Fields start on u_long boundaries */ + fake_preload[i++] = 0; +#endif + + fake_preload[i++] = MODINFO_ADDR; + fake_preload[i++] = sizeof(vm_offset_t); + *(vm_offset_t *)&fake_preload[i] = + (vm_offset_t)(__startkernel); + i += (sizeof(vm_offset_t) / 4); + + fake_preload[i++] = MODINFO_SIZE; + fake_preload[i++] = sizeof(vm_offset_t); + *(vm_offset_t *)&fake_preload[i] = + (vm_offset_t)(__endkernel) - (vm_offset_t)(__startkernel); + i += (sizeof(vm_offset_t) / 4); + + /* + * MODINFOMD_SSYM and MODINFOMD_ESYM cannot be provided here, + * as the memory comes from outside the loaded ELF sections. + * + * If the symbols are being provided by other means (MFS), the + * tables will be loaded into the debugger directly. + */ + + /* Null field at end to mark end of data. */ + fake_preload[i++] = 0; + fake_preload[i] = 0; + preload_metadata = (void*)fake_preload; } /* * Flush the D-cache for non-DMA I/O so that the I-cache can * be made coherent later. */ void cpu_flush_dcache(void *ptr, size_t len) { register_t addr, off; /* * Align the address to a cacheline and adjust the length * accordingly. Then round the length to a multiple of the * cacheline for easy looping. */ addr = (uintptr_t)ptr; off = addr & (cacheline_size - 1); addr -= off; len = roundup2(len + off, cacheline_size); while (len > 0) { __asm __volatile ("dcbf 0,%0" :: "r"(addr)); __asm __volatile ("sync"); addr += cacheline_size; len -= cacheline_size; } } int ptrace_set_pc(struct thread *td, unsigned long addr) { struct trapframe *tf; tf = td->td_frame; tf->srr0 = (register_t)addr; return (0); } void spinlock_enter(void) { struct thread *td; register_t msr; td = curthread; if (td->td_md.md_spinlock_count == 0) { nop_prio_mhigh(); msr = intr_disable(); td->td_md.md_spinlock_count = 1; td->td_md.md_saved_msr = msr; critical_enter(); } else td->td_md.md_spinlock_count++; } void spinlock_exit(void) { struct thread *td; register_t msr; td = curthread; msr = td->td_md.md_saved_msr; td->td_md.md_spinlock_count--; if (td->td_md.md_spinlock_count == 0) { critical_exit(); intr_restore(msr); nop_prio_medium(); } } /* * Simple ddb(4) command/hack to view any SPR on the running CPU. * Uses a trivial asm function to perform the mfspr, and rewrites the mfspr * instruction each time. * XXX: Since it uses code modification, it won't work if the kernel code pages * are marked RO. */ extern register_t get_spr(int); #ifdef DDB DB_SHOW_COMMAND(spr, db_show_spr) { register_t spr; volatile uint32_t *p; int sprno, saved_sprno; if (!have_addr) return; saved_sprno = sprno = (intptr_t) addr; sprno = ((sprno & 0x3e0) >> 5) | ((sprno & 0x1f) << 5); p = (uint32_t *)(void *)&get_spr; #ifdef __powerpc64__ #if defined(_CALL_ELF) && _CALL_ELF == 2 /* Account for ELFv2 function prologue. */ p += 2; #else p = *(volatile uint32_t * volatile *)p; #endif #endif *p = (*p & ~0x001ff800) | (sprno << 11); __syncicache(__DEVOLATILE(uint32_t *, p), cacheline_size); spr = get_spr(sprno); db_printf("SPR %d(%x): %lx\n", saved_sprno, saved_sprno, (unsigned long)spr); } DB_SHOW_COMMAND(frame, db_show_frame) { struct trapframe *tf; long reg; int i; tf = have_addr ? (struct trapframe *)addr : curthread->td_frame; /* * Everything casts through long to simplify the printing. * 'long' is native register size anyway. */ db_printf("trap frame %p\n", tf); for (i = 0; i < nitems(tf->fixreg); i++) { reg = tf->fixreg[i]; db_printf(" r%d:\t%#lx (%ld)\n", i, reg, reg); } reg = tf->lr; db_printf(" lr:\t%#lx\n", reg); reg = tf->cr; db_printf(" cr:\t%#lx\n", reg); reg = tf->xer; db_printf(" xer:\t%#lx\n", reg); reg = tf->ctr; db_printf(" ctr:\t%#lx (%ld)\n", reg, reg); reg = tf->srr0; db_printf(" srr0:\t%#lx\n", reg); reg = tf->srr1; db_printf(" srr1:\t%#lx\n", reg); reg = tf->exc; db_printf(" exc:\t%#lx\n", reg); reg = tf->dar; db_printf(" dar:\t%#lx\n", reg); #ifdef AIM reg = tf->cpu.aim.dsisr; db_printf(" dsisr:\t%#lx\n", reg); #else reg = tf->cpu.booke.esr; db_printf(" esr:\t%#lx\n", reg); reg = tf->cpu.booke.dbcr0; db_printf(" dbcr0:\t%#lx\n", reg); #endif } #endif #undef bzero void bzero(void *buf, size_t len) { caddr_t p; p = buf; while (((vm_offset_t) p & (sizeof(u_long) - 1)) && len) { *p++ = 0; len--; } while (len >= sizeof(u_long) * 8) { *(u_long*) p = 0; *((u_long*) p + 1) = 0; *((u_long*) p + 2) = 0; *((u_long*) p + 3) = 0; len -= sizeof(u_long) * 8; *((u_long*) p + 4) = 0; *((u_long*) p + 5) = 0; *((u_long*) p + 6) = 0; *((u_long*) p + 7) = 0; p += sizeof(u_long) * 8; } while (len >= sizeof(u_long)) { *(u_long*) p = 0; len -= sizeof(u_long); p += sizeof(u_long); } while (len) { *p++ = 0; len--; } } /* __stack_chk_fail_local() is called in secure-plt (32-bit). */ #if !defined(__powerpc64__) extern void __stack_chk_fail(void); void __stack_chk_fail_local(void); void __stack_chk_fail_local(void) { __stack_chk_fail(); } #endif