Index: projects/release-pkg/bin/sh/jobs.c =================================================================== --- projects/release-pkg/bin/sh/jobs.c (revision 296327) +++ projects/release-pkg/bin/sh/jobs.c (revision 296328) @@ -1,1515 +1,1515 @@ /*- * Copyright (c) 1991, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Kenneth Almquist. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint #if 0 static char sccsid[] = "@(#)jobs.c 8.5 (Berkeley) 5/4/95"; #endif #endif /* not lint */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include "shell.h" #if JOBS #include #undef CEOF /* syntax.h redefines this */ #endif #include "redir.h" #include "exec.h" #include "show.h" #include "main.h" #include "parser.h" #include "nodes.h" #include "jobs.h" #include "options.h" #include "trap.h" #include "syntax.h" #include "input.h" #include "output.h" #include "memalloc.h" #include "error.h" #include "mystring.h" #include "var.h" #include "builtins.h" static struct job *jobtab; /* array of jobs */ static int njobs; /* size of array */ static pid_t backgndpid = -1; /* pid of last background process */ static struct job *bgjob = NULL; /* last background process */ #if JOBS static struct job *jobmru; /* most recently used job list */ static pid_t initialpgrp; /* pgrp of shell on invocation */ #endif static int ttyfd = -1; /* mode flags for dowait */ #define DOWAIT_BLOCK 0x1 /* wait until a child exits */ #define DOWAIT_SIG 0x2 /* if DOWAIT_BLOCK, abort on signal */ #define DOWAIT_SIG_TRAP 0x4 /* if DOWAIT_SIG, abort on trapped signal only */ #if JOBS static void restartjob(struct job *); #endif static void freejob(struct job *); static int waitcmdloop(struct job *); static struct job *getjob_nonotfound(const char *); static struct job *getjob(const char *); pid_t killjob(const char *, int); static pid_t dowait(int, struct job *); static void checkzombies(void); static void cmdtxt(union node *); static void cmdputs(const char *); #if JOBS static void setcurjob(struct job *); static void deljob(struct job *); static struct job *getcurjob(struct job *); #endif static void printjobcmd(struct job *); static void showjob(struct job *, int); /* * Turn job control on and off. */ static int jobctl; #if JOBS static void jobctl_notty(void) { if (ttyfd >= 0) { close(ttyfd); ttyfd = -1; } if (!iflag) { setsignal(SIGTSTP); setsignal(SIGTTOU); setsignal(SIGTTIN); jobctl = 1; return; } out2fmt_flush("sh: can't access tty; job control turned off\n"); mflag = 0; } void setjobctl(int on) { int i; if (on == jobctl || rootshell == 0) return; if (on) { if (ttyfd != -1) close(ttyfd); if ((ttyfd = open(_PATH_TTY, O_RDWR | O_CLOEXEC)) < 0) { i = 0; while (i <= 2 && !isatty(i)) i++; if (i > 2 || (ttyfd = fcntl(i, F_DUPFD_CLOEXEC, 10)) < 0) { jobctl_notty(); return; } } if (ttyfd < 10) { /* * Keep our TTY file descriptor out of the way of * the user's redirections. */ if ((i = fcntl(ttyfd, F_DUPFD_CLOEXEC, 10)) < 0) { jobctl_notty(); return; } close(ttyfd); ttyfd = i; } do { /* while we are in the background */ initialpgrp = tcgetpgrp(ttyfd); if (initialpgrp < 0) { jobctl_notty(); return; } if (initialpgrp != getpgrp()) { if (!iflag) { initialpgrp = -1; jobctl_notty(); return; } kill(0, SIGTTIN); continue; } } while (0); setsignal(SIGTSTP); setsignal(SIGTTOU); setsignal(SIGTTIN); setpgid(0, rootpid); tcsetpgrp(ttyfd, rootpid); } else { /* turning job control off */ setpgid(0, initialpgrp); if (ttyfd >= 0) { tcsetpgrp(ttyfd, initialpgrp); close(ttyfd); ttyfd = -1; } setsignal(SIGTSTP); setsignal(SIGTTOU); setsignal(SIGTTIN); } jobctl = on; } #endif #if JOBS int fgcmd(int argc __unused, char **argv __unused) { struct job *jp; pid_t pgrp; int status; nextopt(""); jp = getjob(*argptr); if (jp->jobctl == 0) error("job not created under job control"); printjobcmd(jp); flushout(&output); pgrp = jp->ps[0].pid; if (ttyfd >= 0) tcsetpgrp(ttyfd, pgrp); restartjob(jp); jp->foreground = 1; INTOFF; status = waitforjob(jp, (int *)NULL); INTON; return status; } int bgcmd(int argc __unused, char **argv __unused) { struct job *jp; nextopt(""); do { jp = getjob(*argptr); if (jp->jobctl == 0) error("job not created under job control"); if (jp->state == JOBDONE) continue; restartjob(jp); jp->foreground = 0; out1fmt("[%td] ", jp - jobtab + 1); printjobcmd(jp); } while (*argptr != NULL && *++argptr != NULL); return 0; } static void restartjob(struct job *jp) { struct procstat *ps; int i; if (jp->state == JOBDONE) return; setcurjob(jp); INTOFF; kill(-jp->ps[0].pid, SIGCONT); for (ps = jp->ps, i = jp->nprocs ; --i >= 0 ; ps++) { if (WIFSTOPPED(ps->status)) { ps->status = -1; jp->state = 0; } } INTON; } #endif int jobscmd(int argc __unused, char *argv[] __unused) { char *id; int ch, mode; mode = SHOWJOBS_DEFAULT; while ((ch = nextopt("lps")) != '\0') { switch (ch) { case 'l': mode = SHOWJOBS_VERBOSE; break; case 'p': mode = SHOWJOBS_PGIDS; break; case 's': mode = SHOWJOBS_PIDS; break; } } if (*argptr == NULL) showjobs(0, mode); else while ((id = *argptr++) != NULL) showjob(getjob(id), mode); return (0); } static void printjobcmd(struct job *jp) { struct procstat *ps; int i; for (ps = jp->ps, i = jp->nprocs ; --i >= 0 ; ps++) { out1str(ps->cmd); if (i > 0) out1str(" | "); } out1c('\n'); } static void showjob(struct job *jp, int mode) { char s[64]; - char statestr[64]; - const char *sigstr; + char statebuf[16]; + const char *statestr, *coredump; struct procstat *ps; struct job *j; int col, curr, i, jobno, prev, procno; char c; procno = (mode == SHOWJOBS_PGIDS) ? 1 : jp->nprocs; jobno = jp - jobtab + 1; curr = prev = 0; #if JOBS if ((j = getcurjob(NULL)) != NULL) { curr = j - jobtab + 1; if ((j = getcurjob(j)) != NULL) prev = j - jobtab + 1; } #endif + coredump = ""; ps = jp->ps + jp->nprocs - 1; if (jp->state == 0) { - strcpy(statestr, "Running"); + statestr = "Running"; #if JOBS } else if (jp->state == JOBSTOPPED) { while (!WIFSTOPPED(ps->status) && ps > jp->ps) ps--; if (WIFSTOPPED(ps->status)) i = WSTOPSIG(ps->status); else i = -1; - sigstr = strsignal(i); - if (sigstr != NULL) - strcpy(statestr, sigstr); - else - strcpy(statestr, "Suspended"); + statestr = strsignal(i); + if (statestr == NULL) + statestr = "Suspended"; #endif } else if (WIFEXITED(ps->status)) { if (WEXITSTATUS(ps->status) == 0) - strcpy(statestr, "Done"); - else - fmtstr(statestr, 64, "Done(%d)", + statestr = "Done"; + else { + fmtstr(statebuf, sizeof(statebuf), "Done(%d)", WEXITSTATUS(ps->status)); + statestr = statebuf; + } } else { i = WTERMSIG(ps->status); - sigstr = strsignal(i); - if (sigstr != NULL) - strcpy(statestr, sigstr); - else - strcpy(statestr, "Unknown signal"); + statestr = strsignal(i); + if (statestr == NULL) + statestr = "Unknown signal"; if (WCOREDUMP(ps->status)) - strcat(statestr, " (core dumped)"); + coredump = " (core dumped)"; } for (ps = jp->ps ; procno > 0 ; ps++, procno--) { /* for each process */ if (mode == SHOWJOBS_PIDS || mode == SHOWJOBS_PGIDS) { out1fmt("%d\n", (int)ps->pid); continue; } if (mode != SHOWJOBS_VERBOSE && ps != jp->ps) continue; if (jobno == curr && ps == jp->ps) c = '+'; else if (jobno == prev && ps == jp->ps) c = '-'; else c = ' '; if (ps == jp->ps) fmtstr(s, 64, "[%d] %c ", jobno, c); else fmtstr(s, 64, " %c ", c); out1str(s); col = strlen(s); if (mode == SHOWJOBS_VERBOSE) { fmtstr(s, 64, "%d ", (int)ps->pid); out1str(s); col += strlen(s); } if (ps == jp->ps) { out1str(statestr); - col += strlen(statestr); + out1str(coredump); + col += strlen(statestr) + strlen(coredump); } do { out1c(' '); col++; } while (col < 30); if (mode == SHOWJOBS_VERBOSE) { out1str(ps->cmd); out1c('\n'); } else printjobcmd(jp); } } /* * Print a list of jobs. If "change" is nonzero, only print jobs whose * statuses have changed since the last call to showjobs. * * If the shell is interrupted in the process of creating a job, the * result may be a job structure containing zero processes. Such structures * will be freed here. */ void showjobs(int change, int mode) { int jobno; struct job *jp; TRACE(("showjobs(%d) called\n", change)); checkzombies(); for (jobno = 1, jp = jobtab ; jobno <= njobs ; jobno++, jp++) { if (! jp->used) continue; if (jp->nprocs == 0) { freejob(jp); continue; } if (change && ! jp->changed) continue; showjob(jp, mode); if (mode == SHOWJOBS_DEFAULT || mode == SHOWJOBS_VERBOSE) { jp->changed = 0; /* Hack: discard jobs for which $! has not been * referenced in interactive mode when they terminate. */ if (jp->state == JOBDONE && !jp->remembered && (iflag || jp != bgjob)) { freejob(jp); } } } } /* * Mark a job structure as unused. */ static void freejob(struct job *jp) { struct procstat *ps; int i; INTOFF; if (bgjob == jp) bgjob = NULL; for (i = jp->nprocs, ps = jp->ps ; --i >= 0 ; ps++) { if (ps->cmd != nullstr) ckfree(ps->cmd); } if (jp->ps != &jp->ps0) ckfree(jp->ps); jp->used = 0; #if JOBS deljob(jp); #endif INTON; } int waitcmd(int argc __unused, char **argv __unused) { struct job *job; int retval; nextopt(""); if (*argptr == NULL) return (waitcmdloop(NULL)); do { job = getjob_nonotfound(*argptr); if (job == NULL) retval = 127; else retval = waitcmdloop(job); argptr++; } while (*argptr != NULL); return (retval); } static int waitcmdloop(struct job *job) { int status, retval, sig; struct job *jp; /* * Loop until a process is terminated or stopped, or a SIGINT is * received. */ do { if (job != NULL) { if (job->state == JOBDONE) { status = job->ps[job->nprocs - 1].status; if (WIFEXITED(status)) retval = WEXITSTATUS(status); else retval = WTERMSIG(status) + 128; if (! iflag || ! job->changed) freejob(job); else { job->remembered = 0; if (job == bgjob) bgjob = NULL; } return retval; } } else { for (jp = jobtab ; jp < jobtab + njobs; jp++) if (jp->used && jp->state == JOBDONE) { if (! iflag || ! jp->changed) freejob(jp); else { jp->remembered = 0; if (jp == bgjob) bgjob = NULL; } } for (jp = jobtab ; ; jp++) { if (jp >= jobtab + njobs) { /* no running procs */ return 0; } if (jp->used && jp->state == 0) break; } } } while (dowait(DOWAIT_BLOCK | DOWAIT_SIG, (struct job *)NULL) != -1); sig = pendingsig_waitcmd; pendingsig_waitcmd = 0; return sig + 128; } int jobidcmd(int argc __unused, char **argv __unused) { struct job *jp; int i; nextopt(""); jp = getjob(*argptr); for (i = 0 ; i < jp->nprocs ; ) { out1fmt("%d", (int)jp->ps[i].pid); out1c(++i < jp->nprocs? ' ' : '\n'); } return 0; } /* * Convert a job name to a job structure. */ static struct job * getjob_nonotfound(const char *name) { int jobno; struct job *found, *jp; size_t namelen; pid_t pid; int i; if (name == NULL) { #if JOBS name = "%+"; #else error("No current job"); #endif } if (name[0] == '%') { if (is_digit(name[1])) { jobno = number(name + 1); if (jobno > 0 && jobno <= njobs && jobtab[jobno - 1].used != 0) return &jobtab[jobno - 1]; #if JOBS } else if ((name[1] == '%' || name[1] == '+') && name[2] == '\0') { if ((jp = getcurjob(NULL)) == NULL) error("No current job"); return (jp); } else if (name[1] == '-' && name[2] == '\0') { if ((jp = getcurjob(NULL)) == NULL || (jp = getcurjob(jp)) == NULL) error("No previous job"); return (jp); #endif } else if (name[1] == '?') { found = NULL; for (jp = jobtab, i = njobs ; --i >= 0 ; jp++) { if (jp->used && jp->nprocs > 0 && strstr(jp->ps[0].cmd, name + 2) != NULL) { if (found) error("%s: ambiguous", name); found = jp; } } if (found != NULL) return (found); } else { namelen = strlen(name); found = NULL; for (jp = jobtab, i = njobs ; --i >= 0 ; jp++) { if (jp->used && jp->nprocs > 0 && strncmp(jp->ps[0].cmd, name + 1, namelen - 1) == 0) { if (found) error("%s: ambiguous", name); found = jp; } } if (found) return found; } } else if (is_number(name)) { pid = (pid_t)number(name); for (jp = jobtab, i = njobs ; --i >= 0 ; jp++) { if (jp->used && jp->nprocs > 0 && jp->ps[jp->nprocs - 1].pid == pid) return jp; } } return NULL; } static struct job * getjob(const char *name) { struct job *jp; jp = getjob_nonotfound(name); if (jp == NULL) error("No such job: %s", name); return (jp); } int killjob(const char *name, int sig) { struct job *jp; int i, ret; jp = getjob(name); if (jp->state == JOBDONE) return 0; if (jp->jobctl) return kill(-jp->ps[0].pid, sig); ret = -1; errno = ESRCH; for (i = 0; i < jp->nprocs; i++) if (jp->ps[i].status == -1 || WIFSTOPPED(jp->ps[i].status)) { if (kill(jp->ps[i].pid, sig) == 0) ret = 0; } else ret = 0; return ret; } /* * Return a new job structure, */ struct job * makejob(union node *node __unused, int nprocs) { int i; struct job *jp; for (i = njobs, jp = jobtab ; ; jp++) { if (--i < 0) { INTOFF; if (njobs == 0) { jobtab = ckmalloc(4 * sizeof jobtab[0]); #if JOBS jobmru = NULL; #endif } else { jp = ckmalloc((njobs + 4) * sizeof jobtab[0]); memcpy(jp, jobtab, njobs * sizeof jp[0]); #if JOBS /* Relocate `next' pointers and list head */ if (jobmru != NULL) jobmru = &jp[jobmru - jobtab]; for (i = 0; i < njobs; i++) if (jp[i].next != NULL) jp[i].next = &jp[jp[i].next - jobtab]; #endif if (bgjob != NULL) bgjob = &jp[bgjob - jobtab]; /* Relocate `ps' pointers */ for (i = 0; i < njobs; i++) if (jp[i].ps == &jobtab[i].ps0) jp[i].ps = &jp[i].ps0; ckfree(jobtab); jobtab = jp; } jp = jobtab + njobs; for (i = 4 ; --i >= 0 ; jobtab[njobs++].used = 0) ; INTON; break; } if (jp->used == 0) break; } INTOFF; jp->state = 0; jp->used = 1; jp->changed = 0; jp->nprocs = 0; jp->foreground = 0; jp->remembered = 0; #if JOBS jp->jobctl = jobctl; jp->next = NULL; #endif if (nprocs > 1) { jp->ps = ckmalloc(nprocs * sizeof (struct procstat)); } else { jp->ps = &jp->ps0; } INTON; TRACE(("makejob(%p, %d) returns %%%td\n", (void *)node, nprocs, jp - jobtab + 1)); return jp; } #if JOBS static void setcurjob(struct job *cj) { struct job *jp, *prev; for (prev = NULL, jp = jobmru; jp != NULL; prev = jp, jp = jp->next) { if (jp == cj) { if (prev != NULL) prev->next = jp->next; else jobmru = jp->next; jp->next = jobmru; jobmru = cj; return; } } cj->next = jobmru; jobmru = cj; } static void deljob(struct job *j) { struct job *jp, *prev; for (prev = NULL, jp = jobmru; jp != NULL; prev = jp, jp = jp->next) { if (jp == j) { if (prev != NULL) prev->next = jp->next; else jobmru = jp->next; return; } } } /* * Return the most recently used job that isn't `nj', and preferably one * that is stopped. */ static struct job * getcurjob(struct job *nj) { struct job *jp; /* Try to find a stopped one.. */ for (jp = jobmru; jp != NULL; jp = jp->next) if (jp->used && jp != nj && jp->state == JOBSTOPPED) return (jp); /* Otherwise the most recently used job that isn't `nj' */ for (jp = jobmru; jp != NULL; jp = jp->next) if (jp->used && jp != nj) return (jp); return (NULL); } #endif /* * Fork of a subshell. If we are doing job control, give the subshell its * own process group. Jp is a job structure that the job is to be added to. * N is the command that will be evaluated by the child. Both jp and n may * be NULL. The mode parameter can be one of the following: * FORK_FG - Fork off a foreground process. * FORK_BG - Fork off a background process. * FORK_NOJOB - Like FORK_FG, but don't give the process its own * process group even if job control is on. * * When job control is turned off, background processes have their standard * input redirected to /dev/null (except for the second and later processes * in a pipeline). */ pid_t forkshell(struct job *jp, union node *n, int mode) { pid_t pid; pid_t pgrp; TRACE(("forkshell(%%%td, %p, %d) called\n", jp - jobtab, (void *)n, mode)); INTOFF; if (mode == FORK_BG && (jp == NULL || jp->nprocs == 0)) checkzombies(); flushall(); pid = fork(); if (pid == -1) { TRACE(("Fork failed, errno=%d\n", errno)); INTON; error("Cannot fork: %s", strerror(errno)); } if (pid == 0) { struct job *p; int wasroot; int i; TRACE(("Child shell %d\n", (int)getpid())); wasroot = rootshell; rootshell = 0; handler = &main_handler; closescript(); INTON; forcelocal = 0; clear_traps(); #if JOBS jobctl = 0; /* do job control only in root shell */ if (wasroot && mode != FORK_NOJOB && mflag) { if (jp == NULL || jp->nprocs == 0) pgrp = getpid(); else pgrp = jp->ps[0].pid; if (setpgid(0, pgrp) == 0 && mode == FORK_FG && ttyfd >= 0) { /*** this causes superfluous TIOCSPGRPS ***/ if (tcsetpgrp(ttyfd, pgrp) < 0) error("tcsetpgrp failed, errno=%d", errno); } setsignal(SIGTSTP); setsignal(SIGTTOU); } else if (mode == FORK_BG) { ignoresig(SIGINT); ignoresig(SIGQUIT); if ((jp == NULL || jp->nprocs == 0) && ! fd0_redirected_p ()) { close(0); if (open(_PATH_DEVNULL, O_RDONLY) != 0) error("cannot open %s: %s", _PATH_DEVNULL, strerror(errno)); } } #else if (mode == FORK_BG) { ignoresig(SIGINT); ignoresig(SIGQUIT); if ((jp == NULL || jp->nprocs == 0) && ! fd0_redirected_p ()) { close(0); if (open(_PATH_DEVNULL, O_RDONLY) != 0) error("cannot open %s: %s", _PATH_DEVNULL, strerror(errno)); } } #endif INTOFF; for (i = njobs, p = jobtab ; --i >= 0 ; p++) if (p->used) freejob(p); INTON; if (wasroot && iflag) { setsignal(SIGINT); setsignal(SIGQUIT); setsignal(SIGTERM); } return pid; } if (rootshell && mode != FORK_NOJOB && mflag) { if (jp == NULL || jp->nprocs == 0) pgrp = pid; else pgrp = jp->ps[0].pid; setpgid(pid, pgrp); } if (mode == FORK_BG) { if (bgjob != NULL && bgjob->state == JOBDONE && !bgjob->remembered && !iflag) freejob(bgjob); backgndpid = pid; /* set $! */ bgjob = jp; } if (jp) { struct procstat *ps = &jp->ps[jp->nprocs++]; ps->pid = pid; ps->status = -1; ps->cmd = nullstr; if (iflag && rootshell && n) ps->cmd = commandtext(n); jp->foreground = mode == FORK_FG; #if JOBS setcurjob(jp); #endif } INTON; TRACE(("In parent shell: child = %d\n", (int)pid)); return pid; } pid_t vforkexecshell(struct job *jp, char **argv, char **envp, const char *path, int idx, int pip[2]) { pid_t pid; struct jmploc jmploc; struct jmploc *savehandler; TRACE(("vforkexecshell(%%%td, %s, %p) called\n", jp - jobtab, argv[0], (void *)pip)); INTOFF; flushall(); savehandler = handler; pid = vfork(); if (pid == -1) { TRACE(("Vfork failed, errno=%d\n", errno)); INTON; error("Cannot fork: %s", strerror(errno)); } if (pid == 0) { TRACE(("Child shell %d\n", (int)getpid())); if (setjmp(jmploc.loc)) _exit(exception == EXEXEC ? exerrno : 2); if (pip != NULL) { close(pip[0]); if (pip[1] != 1) { dup2(pip[1], 1); close(pip[1]); } } handler = &jmploc; shellexec(argv, envp, path, idx); } handler = savehandler; if (jp) { struct procstat *ps = &jp->ps[jp->nprocs++]; ps->pid = pid; ps->status = -1; ps->cmd = nullstr; jp->foreground = 1; #if JOBS setcurjob(jp); #endif } INTON; TRACE(("In parent shell: child = %d\n", (int)pid)); return pid; } /* * Wait for job to finish. * * Under job control we have the problem that while a child process is * running interrupts generated by the user are sent to the child but not * to the shell. This means that an infinite loop started by an inter- * active user may be hard to kill. With job control turned off, an * interactive user may place an interactive program inside a loop. If * the interactive program catches interrupts, the user doesn't want * these interrupts to also abort the loop. The approach we take here * is to have the shell ignore interrupt signals while waiting for a * foreground process to terminate, and then send itself an interrupt * signal if the child process was terminated by an interrupt signal. * Unfortunately, some programs want to do a bit of cleanup and then * exit on interrupt; unless these processes terminate themselves by * sending a signal to themselves (instead of calling exit) they will * confuse this approach. */ int waitforjob(struct job *jp, int *origstatus) { #if JOBS int propagate_int = jp->jobctl && jp->foreground; #endif int status; int st; INTOFF; TRACE(("waitforjob(%%%td) called\n", jp - jobtab + 1)); while (jp->state == 0) if (dowait(DOWAIT_BLOCK | (Tflag ? DOWAIT_SIG | DOWAIT_SIG_TRAP : 0), jp) == -1) dotrap(); #if JOBS if (jp->jobctl) { if (ttyfd >= 0 && tcsetpgrp(ttyfd, rootpid) < 0) error("tcsetpgrp failed, errno=%d\n", errno); } if (jp->state == JOBSTOPPED) setcurjob(jp); #endif status = jp->ps[jp->nprocs - 1].status; if (origstatus != NULL) *origstatus = status; /* convert to 8 bits */ if (WIFEXITED(status)) st = WEXITSTATUS(status); #if JOBS else if (WIFSTOPPED(status)) st = WSTOPSIG(status) + 128; #endif else st = WTERMSIG(status) + 128; if (! JOBS || jp->state == JOBDONE) freejob(jp); if (int_pending()) { if (!WIFSIGNALED(status) || WTERMSIG(status) != SIGINT) CLEAR_PENDING_INT; } #if JOBS else if (rootshell && propagate_int && WIFSIGNALED(status) && WTERMSIG(status) == SIGINT) kill(getpid(), SIGINT); #endif INTON; return st; } static void dummy_handler(int sig __unused) { } /* * Wait for a process to terminate. */ static pid_t dowait(int mode, struct job *job) { struct sigaction sa, osa; sigset_t mask, omask; pid_t pid; int status; struct procstat *sp; struct job *jp; struct job *thisjob; const char *sigstr; int done; int stopped; int sig; int coredump; int wflags; int restore_sigchld; TRACE(("dowait(%d, %p) called\n", mode, job)); restore_sigchld = 0; if ((mode & DOWAIT_SIG) != 0) { sigfillset(&mask); sigprocmask(SIG_BLOCK, &mask, &omask); INTOFF; if (!issigchldtrapped()) { restore_sigchld = 1; sa.sa_handler = dummy_handler; sa.sa_flags = 0; sigemptyset(&sa.sa_mask); sigaction(SIGCHLD, &sa, &osa); } } do { #if JOBS if (iflag) wflags = WUNTRACED | WCONTINUED; else #endif wflags = 0; if ((mode & (DOWAIT_BLOCK | DOWAIT_SIG)) != DOWAIT_BLOCK) wflags |= WNOHANG; pid = wait3(&status, wflags, (struct rusage *)NULL); TRACE(("wait returns %d, status=%d\n", (int)pid, status)); if (pid == 0 && (mode & DOWAIT_SIG) != 0) { pid = -1; if (((mode & DOWAIT_SIG_TRAP) != 0 ? pendingsig : pendingsig_waitcmd) != 0) { errno = EINTR; break; } sigsuspend(&omask); if (int_pending()) break; } } while (pid == -1 && errno == EINTR); if (pid == -1 && errno == ECHILD && job != NULL) job->state = JOBDONE; if ((mode & DOWAIT_SIG) != 0) { if (restore_sigchld) sigaction(SIGCHLD, &osa, NULL); sigprocmask(SIG_SETMASK, &omask, NULL); INTON; } if (pid <= 0) return pid; INTOFF; thisjob = NULL; for (jp = jobtab ; jp < jobtab + njobs ; jp++) { if (jp->used && jp->nprocs > 0) { done = 1; stopped = 1; for (sp = jp->ps ; sp < jp->ps + jp->nprocs ; sp++) { if (sp->pid == -1) continue; if (sp->pid == pid && (sp->status == -1 || WIFSTOPPED(sp->status))) { TRACE(("Changing status of proc %d from 0x%x to 0x%x\n", (int)pid, sp->status, status)); if (WIFCONTINUED(status)) { sp->status = -1; jp->state = 0; } else sp->status = status; thisjob = jp; } if (sp->status == -1) stopped = 0; else if (WIFSTOPPED(sp->status)) done = 0; } if (stopped) { /* stopped or done */ int state = done? JOBDONE : JOBSTOPPED; if (jp->state != state) { TRACE(("Job %td: changing state from %d to %d\n", jp - jobtab + 1, jp->state, state)); jp->state = state; if (jp != job) { if (done && !jp->remembered && !iflag && jp != bgjob) freejob(jp); #if JOBS else if (done) deljob(jp); #endif } } } } } INTON; if (!thisjob || thisjob->state == 0) ; else if ((!rootshell || !iflag || thisjob == job) && thisjob->foreground && thisjob->state != JOBSTOPPED) { sig = 0; coredump = 0; for (sp = thisjob->ps; sp < thisjob->ps + thisjob->nprocs; sp++) if (WIFSIGNALED(sp->status)) { sig = WTERMSIG(sp->status); coredump = WCOREDUMP(sp->status); } if (sig > 0 && sig != SIGINT && sig != SIGPIPE) { sigstr = strsignal(sig); if (sigstr != NULL) out2str(sigstr); else out2str("Unknown signal"); if (coredump) out2str(" (core dumped)"); out2c('\n'); flushout(out2); } } else { TRACE(("Not printing status, rootshell=%d, job=%p\n", rootshell, job)); thisjob->changed = 1; } return pid; } /* * return 1 if there are stopped jobs, otherwise 0 */ int job_warning = 0; int stoppedjobs(void) { int jobno; struct job *jp; if (job_warning) return (0); for (jobno = 1, jp = jobtab; jobno <= njobs; jobno++, jp++) { if (jp->used == 0) continue; if (jp->state == JOBSTOPPED) { out2fmt_flush("You have stopped jobs.\n"); job_warning = 2; return (1); } } return (0); } static void checkzombies(void) { while (njobs > 0 && dowait(0, NULL) > 0) ; } int backgndpidset(void) { return backgndpid != -1; } pid_t backgndpidval(void) { if (bgjob != NULL && !forcelocal) bgjob->remembered = 1; return backgndpid; } /* * Return a string identifying a command (to be printed by the * jobs command. */ static char *cmdnextc; static int cmdnleft; #define MAXCMDTEXT 200 char * commandtext(union node *n) { char *name; cmdnextc = name = ckmalloc(MAXCMDTEXT); cmdnleft = MAXCMDTEXT - 4; cmdtxt(n); *cmdnextc = '\0'; return name; } static void cmdtxtdogroup(union node *n) { cmdputs("; do "); cmdtxt(n); cmdputs("; done"); } static void cmdtxtredir(union node *n, const char *op, int deffd) { char s[2]; if (n->nfile.fd != deffd) { s[0] = n->nfile.fd + '0'; s[1] = '\0'; cmdputs(s); } cmdputs(op); if (n->type == NTOFD || n->type == NFROMFD) { if (n->ndup.dupfd >= 0) s[0] = n->ndup.dupfd + '0'; else s[0] = '-'; s[1] = '\0'; cmdputs(s); } else { cmdtxt(n->nfile.fname); } } static void cmdtxt(union node *n) { union node *np; struct nodelist *lp; if (n == NULL) return; switch (n->type) { case NSEMI: cmdtxt(n->nbinary.ch1); cmdputs("; "); cmdtxt(n->nbinary.ch2); break; case NAND: cmdtxt(n->nbinary.ch1); cmdputs(" && "); cmdtxt(n->nbinary.ch2); break; case NOR: cmdtxt(n->nbinary.ch1); cmdputs(" || "); cmdtxt(n->nbinary.ch2); break; case NPIPE: for (lp = n->npipe.cmdlist ; lp ; lp = lp->next) { cmdtxt(lp->n); if (lp->next) cmdputs(" | "); } break; case NSUBSHELL: cmdputs("("); cmdtxt(n->nredir.n); cmdputs(")"); break; case NREDIR: case NBACKGND: cmdtxt(n->nredir.n); break; case NIF: cmdputs("if "); cmdtxt(n->nif.test); cmdputs("; then "); cmdtxt(n->nif.ifpart); cmdputs("..."); break; case NWHILE: cmdputs("while "); cmdtxt(n->nbinary.ch1); cmdtxtdogroup(n->nbinary.ch2); break; case NUNTIL: cmdputs("until "); cmdtxt(n->nbinary.ch1); cmdtxtdogroup(n->nbinary.ch2); break; case NFOR: cmdputs("for "); cmdputs(n->nfor.var); cmdputs(" in ..."); break; case NCASE: cmdputs("case "); cmdputs(n->ncase.expr->narg.text); cmdputs(" in ..."); break; case NDEFUN: cmdputs(n->narg.text); cmdputs("() ..."); break; case NNOT: cmdputs("! "); cmdtxt(n->nnot.com); break; case NCMD: for (np = n->ncmd.args ; np ; np = np->narg.next) { cmdtxt(np); if (np->narg.next) cmdputs(" "); } for (np = n->ncmd.redirect ; np ; np = np->nfile.next) { cmdputs(" "); cmdtxt(np); } break; case NARG: cmdputs(n->narg.text); break; case NTO: cmdtxtredir(n, ">", 1); break; case NAPPEND: cmdtxtredir(n, ">>", 1); break; case NTOFD: cmdtxtredir(n, ">&", 1); break; case NCLOBBER: cmdtxtredir(n, ">|", 1); break; case NFROM: cmdtxtredir(n, "<", 0); break; case NFROMTO: cmdtxtredir(n, "<>", 0); break; case NFROMFD: cmdtxtredir(n, "<&", 0); break; case NHERE: case NXHERE: cmdputs("<<..."); break; default: cmdputs("???"); break; } } static void cmdputs(const char *s) { const char *p; char *q; char c; int subtype = 0; if (cmdnleft <= 0) return; p = s; q = cmdnextc; while ((c = *p++) != '\0') { if (c == CTLESC) *q++ = *p++; else if (c == CTLVAR) { *q++ = '$'; if (--cmdnleft > 0) *q++ = '{'; subtype = *p++; if ((subtype & VSTYPE) == VSLENGTH && --cmdnleft > 0) *q++ = '#'; } else if (c == '=' && subtype != 0) { *q = "}-+?=##%%\0X"[(subtype & VSTYPE) - VSNORMAL]; if (*q) q++; else cmdnleft++; if (((subtype & VSTYPE) == VSTRIMLEFTMAX || (subtype & VSTYPE) == VSTRIMRIGHTMAX) && --cmdnleft > 0) *q = q[-1], q++; subtype = 0; } else if (c == CTLENDVAR) { *q++ = '}'; } else if (c == CTLBACKQ || c == CTLBACKQ+CTLQUOTE) { cmdnleft -= 5; if (cmdnleft > 0) { *q++ = '$'; *q++ = '('; *q++ = '.'; *q++ = '.'; *q++ = '.'; *q++ = ')'; } } else if (c == CTLARI) { cmdnleft -= 2; if (cmdnleft > 0) { *q++ = '$'; *q++ = '('; *q++ = '('; } p++; } else if (c == CTLENDARI) { if (--cmdnleft > 0) { *q++ = ')'; *q++ = ')'; } } else if (c == CTLQUOTEMARK || c == CTLQUOTEEND) cmdnleft++; /* ignore */ else *q++ = c; if (--cmdnleft <= 0) { *q++ = '.'; *q++ = '.'; *q++ = '.'; break; } } cmdnextc = q; } Index: projects/release-pkg/bin/sh/parser.c =================================================================== --- projects/release-pkg/bin/sh/parser.c (revision 296327) +++ projects/release-pkg/bin/sh/parser.c (revision 296328) @@ -1,2115 +1,2115 @@ /*- * Copyright (c) 1991, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Kenneth Almquist. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint #if 0 static char sccsid[] = "@(#)parser.c 8.7 (Berkeley) 5/16/95"; #endif #endif /* not lint */ #include __FBSDID("$FreeBSD$"); #include #include #include #include "shell.h" #include "parser.h" #include "nodes.h" #include "expand.h" /* defines rmescapes() */ #include "syntax.h" #include "options.h" #include "input.h" #include "output.h" #include "var.h" #include "error.h" #include "memalloc.h" #include "mystring.h" #include "alias.h" #include "show.h" #include "eval.h" #include "exec.h" /* to check for special builtins */ #ifndef NO_HISTORY #include "myhistedit.h" #endif /* * Shell command parser. */ #define PROMPTLEN 128 /* values of checkkwd variable */ #define CHKALIAS 0x1 #define CHKKWD 0x2 #define CHKNL 0x4 /* values returned by readtoken */ #include "token.h" struct heredoc { struct heredoc *next; /* next here document in list */ union node *here; /* redirection node */ char *eofmark; /* string indicating end of input */ int striptabs; /* if set, strip leading tabs */ }; struct parser_temp { struct parser_temp *next; void *data; }; static struct heredoc *heredoclist; /* list of here documents to read */ static int doprompt; /* if set, prompt the user */ static int needprompt; /* true if interactive and at start of line */ static int lasttoken; /* last token read */ static int tokpushback; /* last token pushed back */ static char *wordtext; /* text of last word returned by readtoken */ static int checkkwd; static struct nodelist *backquotelist; static union node *redirnode; static struct heredoc *heredoc; static int quoteflag; /* set if (part of) last token was quoted */ static int startlinno; /* line # where last token started */ static int funclinno; /* line # where the current function started */ static struct parser_temp *parser_temp; #define NOEOFMARK ((const char *)&heredoclist) static union node *list(int); static union node *andor(void); static union node *pipeline(void); static union node *command(void); static union node *simplecmd(union node **, union node *); static union node *makename(void); static union node *makebinary(int type, union node *n1, union node *n2); static void parsefname(void); static void parseheredoc(void); static int peektoken(void); static int readtoken(void); static int xxreadtoken(void); static int readtoken1(int, const char *, const char *, int); static int noexpand(char *); static void consumetoken(int); static void synexpect(int) __dead2; static void synerror(const char *) __dead2; static void setprompt(int); static int pgetc_linecont(void); static void * parser_temp_alloc(size_t len) { struct parser_temp *t; INTOFF; t = ckmalloc(sizeof(*t)); t->data = NULL; t->next = parser_temp; parser_temp = t; t->data = ckmalloc(len); INTON; return t->data; } static void * parser_temp_realloc(void *ptr, size_t len) { struct parser_temp *t; INTOFF; t = parser_temp; if (ptr != t->data) error("bug: parser_temp_realloc misused"); t->data = ckrealloc(t->data, len); INTON; return t->data; } static void parser_temp_free_upto(void *ptr) { struct parser_temp *t; int done = 0; INTOFF; while (parser_temp != NULL && !done) { t = parser_temp; parser_temp = t->next; done = t->data == ptr; ckfree(t->data); ckfree(t); } INTON; if (!done) error("bug: parser_temp_free_upto misused"); } static void parser_temp_free_all(void) { struct parser_temp *t; INTOFF; while (parser_temp != NULL) { t = parser_temp; parser_temp = t->next; ckfree(t->data); ckfree(t); } INTON; } /* * Read and parse a command. Returns NEOF on end of file. (NULL is a * valid parse tree indicating a blank line.) */ union node * parsecmd(int interact) { int t; /* This assumes the parser is not re-entered, * which could happen if we add command substitution on PS1/PS2. */ parser_temp_free_all(); heredoclist = NULL; tokpushback = 0; checkkwd = 0; doprompt = interact; if (doprompt) setprompt(1); else setprompt(0); needprompt = 0; t = readtoken(); if (t == TEOF) return NEOF; if (t == TNL) return NULL; tokpushback++; return list(1); } /* * Read and parse words for wordexp. * Returns a list of NARG nodes; NULL if there are no words. */ union node * parsewordexp(void) { union node *n, *first = NULL, **pnext; int t; /* This assumes the parser is not re-entered, * which could happen if we add command substitution on PS1/PS2. */ parser_temp_free_all(); heredoclist = NULL; tokpushback = 0; checkkwd = 0; doprompt = 0; setprompt(0); needprompt = 0; pnext = &first; while ((t = readtoken()) != TEOF) { if (t != TWORD) synexpect(TWORD); n = makename(); *pnext = n; pnext = &n->narg.next; } return first; } static union node * list(int nlflag) { union node *ntop, *n1, *n2, *n3; int tok; checkkwd = CHKNL | CHKKWD | CHKALIAS; if (!nlflag && tokendlist[peektoken()]) return NULL; ntop = n1 = NULL; for (;;) { n2 = andor(); tok = readtoken(); if (tok == TBACKGND) { if (n2 != NULL && n2->type == NPIPE) { n2->npipe.backgnd = 1; } else if (n2 != NULL && n2->type == NREDIR) { n2->type = NBACKGND; } else { n3 = (union node *)stalloc(sizeof (struct nredir)); n3->type = NBACKGND; n3->nredir.n = n2; n3->nredir.redirect = NULL; n2 = n3; } } if (ntop == NULL) ntop = n2; else if (n1 == NULL) { n1 = makebinary(NSEMI, ntop, n2); ntop = n1; } else { n3 = makebinary(NSEMI, n1->nbinary.ch2, n2); n1->nbinary.ch2 = n3; n1 = n3; } switch (tok) { case TBACKGND: case TSEMI: tok = readtoken(); /* FALLTHROUGH */ case TNL: if (tok == TNL) { parseheredoc(); if (nlflag) return ntop; } else if (tok == TEOF && nlflag) { parseheredoc(); return ntop; } else { tokpushback++; } checkkwd = CHKNL | CHKKWD | CHKALIAS; if (!nlflag && tokendlist[peektoken()]) return ntop; break; case TEOF: if (heredoclist) parseheredoc(); else pungetc(); /* push back EOF on input */ return ntop; default: if (nlflag) synexpect(-1); tokpushback++; return ntop; } } } static union node * andor(void) { union node *n; int t; n = pipeline(); for (;;) { if ((t = readtoken()) == TAND) { t = NAND; } else if (t == TOR) { t = NOR; } else { tokpushback++; return n; } n = makebinary(t, n, pipeline()); } } static union node * pipeline(void) { union node *n1, *n2, *pipenode; struct nodelist *lp, *prev; int negate, t; negate = 0; checkkwd = CHKNL | CHKKWD | CHKALIAS; TRACE(("pipeline: entered\n")); while (readtoken() == TNOT) negate = !negate; tokpushback++; n1 = command(); if (readtoken() == TPIPE) { pipenode = (union node *)stalloc(sizeof (struct npipe)); pipenode->type = NPIPE; pipenode->npipe.backgnd = 0; lp = (struct nodelist *)stalloc(sizeof (struct nodelist)); pipenode->npipe.cmdlist = lp; lp->n = n1; do { prev = lp; lp = (struct nodelist *)stalloc(sizeof (struct nodelist)); checkkwd = CHKNL | CHKKWD | CHKALIAS; t = readtoken(); tokpushback++; if (t == TNOT) lp->n = pipeline(); else lp->n = command(); prev->next = lp; } while (readtoken() == TPIPE); lp->next = NULL; n1 = pipenode; } tokpushback++; if (negate) { n2 = (union node *)stalloc(sizeof (struct nnot)); n2->type = NNOT; n2->nnot.com = n1; return n2; } else return n1; } static union node * command(void) { union node *n1, *n2; union node *ap, **app; union node *cp, **cpp; union node *redir, **rpp; int t; int is_subshell; checkkwd = CHKNL | CHKKWD | CHKALIAS; is_subshell = 0; redir = NULL; n1 = NULL; rpp = &redir; /* Check for redirection which may precede command */ while (readtoken() == TREDIR) { *rpp = n2 = redirnode; rpp = &n2->nfile.next; parsefname(); } tokpushback++; switch (readtoken()) { case TIF: n1 = (union node *)stalloc(sizeof (struct nif)); n1->type = NIF; if ((n1->nif.test = list(0)) == NULL) synexpect(-1); consumetoken(TTHEN); n1->nif.ifpart = list(0); n2 = n1; while (readtoken() == TELIF) { n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif)); n2 = n2->nif.elsepart; n2->type = NIF; if ((n2->nif.test = list(0)) == NULL) synexpect(-1); consumetoken(TTHEN); n2->nif.ifpart = list(0); } if (lasttoken == TELSE) n2->nif.elsepart = list(0); else { n2->nif.elsepart = NULL; tokpushback++; } consumetoken(TFI); checkkwd = CHKKWD | CHKALIAS; break; case TWHILE: case TUNTIL: t = lasttoken; if ((n1 = list(0)) == NULL) synexpect(-1); consumetoken(TDO); n1 = makebinary((t == TWHILE)? NWHILE : NUNTIL, n1, list(0)); consumetoken(TDONE); checkkwd = CHKKWD | CHKALIAS; break; case TFOR: if (readtoken() != TWORD || quoteflag || ! goodname(wordtext)) synerror("Bad for loop variable"); n1 = (union node *)stalloc(sizeof (struct nfor)); n1->type = NFOR; n1->nfor.var = wordtext; while (readtoken() == TNL) ; if (lasttoken == TWORD && ! quoteflag && equal(wordtext, "in")) { app = ≈ while (readtoken() == TWORD) { n2 = makename(); *app = n2; app = &n2->narg.next; } *app = NULL; n1->nfor.args = ap; if (lasttoken != TNL && lasttoken != TSEMI) synexpect(-1); } else { static char argvars[5] = { CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0' }; n2 = (union node *)stalloc(sizeof (struct narg)); n2->type = NARG; n2->narg.text = argvars; n2->narg.backquote = NULL; n2->narg.next = NULL; n1->nfor.args = n2; /* * Newline or semicolon here is optional (but note * that the original Bourne shell only allowed NL). */ if (lasttoken != TNL && lasttoken != TSEMI) tokpushback++; } checkkwd = CHKNL | CHKKWD | CHKALIAS; if ((t = readtoken()) == TDO) t = TDONE; else if (t == TBEGIN) t = TEND; else synexpect(-1); n1->nfor.body = list(0); consumetoken(t); checkkwd = CHKKWD | CHKALIAS; break; case TCASE: n1 = (union node *)stalloc(sizeof (struct ncase)); n1->type = NCASE; consumetoken(TWORD); n1->ncase.expr = makename(); while (readtoken() == TNL); if (lasttoken != TWORD || ! equal(wordtext, "in")) synerror("expecting \"in\""); cpp = &n1->ncase.cases; checkkwd = CHKNL | CHKKWD, readtoken(); while (lasttoken != TESAC) { *cpp = cp = (union node *)stalloc(sizeof (struct nclist)); cp->type = NCLIST; app = &cp->nclist.pattern; if (lasttoken == TLP) readtoken(); for (;;) { *app = ap = makename(); checkkwd = CHKNL | CHKKWD; if (readtoken() != TPIPE) break; app = &ap->narg.next; readtoken(); } ap->narg.next = NULL; if (lasttoken != TRP) synexpect(TRP); cp->nclist.body = list(0); checkkwd = CHKNL | CHKKWD | CHKALIAS; if ((t = readtoken()) != TESAC) { if (t == TENDCASE) ; else if (t == TFALLTHRU) cp->type = NCLISTFALLTHRU; else synexpect(TENDCASE); checkkwd = CHKNL | CHKKWD, readtoken(); } cpp = &cp->nclist.next; } *cpp = NULL; checkkwd = CHKKWD | CHKALIAS; break; case TLP: n1 = (union node *)stalloc(sizeof (struct nredir)); n1->type = NSUBSHELL; n1->nredir.n = list(0); n1->nredir.redirect = NULL; consumetoken(TRP); checkkwd = CHKKWD | CHKALIAS; is_subshell = 1; break; case TBEGIN: n1 = list(0); consumetoken(TEND); checkkwd = CHKKWD | CHKALIAS; break; /* A simple command must have at least one redirection or word. */ case TBACKGND: case TSEMI: case TAND: case TOR: case TPIPE: case TENDCASE: case TFALLTHRU: case TEOF: case TNL: case TRP: if (!redir) synexpect(-1); case TWORD: tokpushback++; n1 = simplecmd(rpp, redir); return n1; default: synexpect(-1); } /* Now check for redirection which may follow command */ while (readtoken() == TREDIR) { *rpp = n2 = redirnode; rpp = &n2->nfile.next; parsefname(); } tokpushback++; *rpp = NULL; if (redir) { if (!is_subshell) { n2 = (union node *)stalloc(sizeof (struct nredir)); n2->type = NREDIR; n2->nredir.n = n1; n1 = n2; } n1->nredir.redirect = redir; } return n1; } static union node * simplecmd(union node **rpp, union node *redir) { union node *args, **app; union node **orig_rpp = rpp; union node *n = NULL; int special; int savecheckkwd; /* If we don't have any redirections already, then we must reset */ /* rpp to be the address of the local redir variable. */ if (redir == 0) rpp = &redir; args = NULL; app = &args; /* * We save the incoming value, because we need this for shell * functions. There can not be a redirect or an argument between * the function name and the open parenthesis. */ orig_rpp = rpp; savecheckkwd = CHKALIAS; for (;;) { checkkwd = savecheckkwd; if (readtoken() == TWORD) { n = makename(); *app = n; app = &n->narg.next; if (savecheckkwd != 0 && !isassignment(wordtext)) savecheckkwd = 0; } else if (lasttoken == TREDIR) { *rpp = n = redirnode; rpp = &n->nfile.next; parsefname(); /* read name of redirection file */ } else if (lasttoken == TLP && app == &args->narg.next && rpp == orig_rpp) { /* We have a function */ consumetoken(TRP); funclinno = plinno; /* * - Require plain text. * - Functions with '/' cannot be called. * - Reject name=(). * - Reject ksh extended glob patterns. */ if (!noexpand(n->narg.text) || quoteflag || strchr(n->narg.text, '/') || strchr("!%*+-=?@}~", n->narg.text[strlen(n->narg.text) - 1])) synerror("Bad function name"); rmescapes(n->narg.text); if (find_builtin(n->narg.text, &special) >= 0 && special) synerror("Cannot override a special builtin with a function"); n->type = NDEFUN; n->narg.next = command(); funclinno = 0; return n; } else { tokpushback++; break; } } *app = NULL; *rpp = NULL; n = (union node *)stalloc(sizeof (struct ncmd)); n->type = NCMD; n->ncmd.args = args; n->ncmd.redirect = redir; return n; } static union node * makename(void) { union node *n; n = (union node *)stalloc(sizeof (struct narg)); n->type = NARG; n->narg.next = NULL; n->narg.text = wordtext; n->narg.backquote = backquotelist; return n; } static union node * makebinary(int type, union node *n1, union node *n2) { union node *n; n = (union node *)stalloc(sizeof (struct nbinary)); n->type = type; n->nbinary.ch1 = n1; n->nbinary.ch2 = n2; return (n); } void forcealias(void) { checkkwd |= CHKALIAS; } void fixredir(union node *n, const char *text, int err) { TRACE(("Fix redir %s %d\n", text, err)); if (!err) n->ndup.vname = NULL; if (is_digit(text[0]) && text[1] == '\0') n->ndup.dupfd = digit_val(text[0]); else if (text[0] == '-' && text[1] == '\0') n->ndup.dupfd = -1; else { if (err) synerror("Bad fd number"); else n->ndup.vname = makename(); } } static void parsefname(void) { union node *n = redirnode; consumetoken(TWORD); if (n->type == NHERE) { struct heredoc *here = heredoc; struct heredoc *p; if (quoteflag == 0) n->type = NXHERE; TRACE(("Here document %d\n", n->type)); if (here->striptabs) { while (*wordtext == '\t') wordtext++; } if (! noexpand(wordtext)) synerror("Illegal eof marker for << redirection"); rmescapes(wordtext); here->eofmark = wordtext; here->next = NULL; if (heredoclist == NULL) heredoclist = here; else { for (p = heredoclist ; p->next ; p = p->next); p->next = here; } } else if (n->type == NTOFD || n->type == NFROMFD) { fixredir(n, wordtext, 0); } else { n->nfile.fname = makename(); } } /* * Input any here documents. */ static void parseheredoc(void) { struct heredoc *here; union node *n; while (heredoclist) { here = heredoclist; heredoclist = here->next; if (needprompt) { setprompt(2); needprompt = 0; } readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX, here->eofmark, here->striptabs); n = makename(); here->here->nhere.doc = n; } } static int peektoken(void) { int t; t = readtoken(); tokpushback++; return (t); } static int readtoken(void) { int t; struct alias *ap; #ifdef DEBUG int alreadyseen = tokpushback; #endif top: t = xxreadtoken(); /* * eat newlines */ if (checkkwd & CHKNL) { while (t == TNL) { parseheredoc(); t = xxreadtoken(); } } /* * check for keywords and aliases */ if (t == TWORD && !quoteflag) { const char * const *pp; if (checkkwd & CHKKWD) for (pp = parsekwd; *pp; pp++) { if (**pp == *wordtext && equal(*pp, wordtext)) { lasttoken = t = pp - parsekwd + KWDOFFSET; TRACE(("keyword %s recognized\n", tokname[t])); goto out; } } if (checkkwd & CHKALIAS && (ap = lookupalias(wordtext, 1)) != NULL) { pushstring(ap->val, strlen(ap->val), ap); goto top; } } out: if (t != TNOT) checkkwd = 0; #ifdef DEBUG if (!alreadyseen) TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : "")); else TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : "")); #endif return (t); } /* * Read the next input token. * If the token is a word, we set backquotelist to the list of cmds in * backquotes. We set quoteflag to true if any part of the word was * quoted. * If the token is TREDIR, then we set redirnode to a structure containing * the redirection. * In all cases, the variable startlinno is set to the number of the line * on which the token starts. * * [Change comment: here documents and internal procedures] * [Readtoken shouldn't have any arguments. Perhaps we should make the * word parsing code into a separate routine. In this case, readtoken * doesn't need to have any internal procedures, but parseword does. * We could also make parseoperator in essence the main routine, and * have parseword (readtoken1?) handle both words and redirection.] */ #define RETURN(token) return lasttoken = token static int xxreadtoken(void) { int c; if (tokpushback) { tokpushback = 0; return lasttoken; } if (needprompt) { setprompt(2); needprompt = 0; } startlinno = plinno; for (;;) { /* until token or start of word found */ c = pgetc_macro(); switch (c) { case ' ': case '\t': continue; case '#': while ((c = pgetc()) != '\n' && c != PEOF); pungetc(); continue; case '\\': if (pgetc() == '\n') { startlinno = ++plinno; if (doprompt) setprompt(2); else setprompt(0); continue; } pungetc(); /* FALLTHROUGH */ default: return readtoken1(c, BASESYNTAX, (char *)NULL, 0); case '\n': plinno++; needprompt = doprompt; RETURN(TNL); case PEOF: RETURN(TEOF); case '&': if (pgetc_linecont() == '&') RETURN(TAND); pungetc(); RETURN(TBACKGND); case '|': if (pgetc_linecont() == '|') RETURN(TOR); pungetc(); RETURN(TPIPE); case ';': c = pgetc_linecont(); if (c == ';') RETURN(TENDCASE); else if (c == '&') RETURN(TFALLTHRU); pungetc(); RETURN(TSEMI); case '(': RETURN(TLP); case ')': RETURN(TRP); } } #undef RETURN } #define MAXNEST_static 8 struct tokenstate { const char *syntax; /* *SYNTAX */ int parenlevel; /* levels of parentheses in arithmetic */ enum tokenstate_category { TSTATE_TOP, TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */ TSTATE_VAR_NEW, /* other ${var...}, own dquote state */ TSTATE_ARITH } category; }; /* * Check to see whether we are at the end of the here document. When this * is called, c is set to the first character of the next input line. If * we are at the end of the here document, this routine sets the c to PEOF. * The new value of c is returned. */ static int checkend(int c, const char *eofmark, int striptabs) { if (striptabs) { while (c == '\t') c = pgetc(); } if (c == *eofmark) { int c2; const char *q; for (q = eofmark + 1; c2 = pgetc(), *q != '\0' && c2 == *q; q++) ; if ((c2 == PEOF || c2 == '\n') && *q == '\0') { c = PEOF; if (c2 == '\n') { plinno++; needprompt = doprompt; } } else { pungetc(); pushstring(eofmark + 1, q - (eofmark + 1), NULL); } } else if (c == '\n' && *eofmark == '\0') { c = PEOF; plinno++; needprompt = doprompt; } return (c); } /* * Parse a redirection operator. The variable "out" points to a string * specifying the fd to be redirected. The variable "c" contains the * first character of the redirection operator. */ static void parseredir(char *out, int c) { char fd = *out; union node *np; np = (union node *)stalloc(sizeof (struct nfile)); if (c == '>') { np->nfile.fd = 1; c = pgetc_linecont(); if (c == '>') np->type = NAPPEND; else if (c == '&') np->type = NTOFD; else if (c == '|') np->type = NCLOBBER; else { np->type = NTO; pungetc(); } } else { /* c == '<' */ np->nfile.fd = 0; c = pgetc_linecont(); if (c == '<') { if (sizeof (struct nfile) != sizeof (struct nhere)) { np = (union node *)stalloc(sizeof (struct nhere)); np->nfile.fd = 0; } np->type = NHERE; heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc)); heredoc->here = np; if ((c = pgetc_linecont()) == '-') { heredoc->striptabs = 1; } else { heredoc->striptabs = 0; pungetc(); } } else if (c == '&') np->type = NFROMFD; else if (c == '>') np->type = NFROMTO; else { np->type = NFROM; pungetc(); } } if (fd != '\0') np->nfile.fd = digit_val(fd); redirnode = np; } /* * Called to parse command substitutions. */ static char * parsebackq(char *out, struct nodelist **pbqlist, int oldstyle, int dblquote, int quoted) { struct nodelist **nlpp; union node *n; char *volatile str; struct jmploc jmploc; struct jmploc *const savehandler = handler; size_t savelen; int saveprompt; const int bq_startlinno = plinno; char *volatile ostr = NULL; struct parsefile *const savetopfile = getcurrentfile(); struct heredoc *const saveheredoclist = heredoclist; struct heredoc *here; str = NULL; if (setjmp(jmploc.loc)) { popfilesupto(savetopfile); if (str) ckfree(str); if (ostr) ckfree(ostr); heredoclist = saveheredoclist; handler = savehandler; if (exception == EXERROR) { startlinno = bq_startlinno; synerror("Error in command substitution"); } longjmp(handler->loc, 1); } INTOFF; savelen = out - stackblock(); if (savelen > 0) { str = ckmalloc(savelen); memcpy(str, stackblock(), savelen); } handler = &jmploc; heredoclist = NULL; INTON; if (oldstyle) { /* We must read until the closing backquote, giving special treatment to some slashes, and then push the string and reread it as input, interpreting it normally. */ char *oout; int c; int olen; STARTSTACKSTR(oout); for (;;) { if (needprompt) { setprompt(2); needprompt = 0; } CHECKSTRSPACE(2, oout); c = pgetc_linecont(); if (c == '`') break; switch (c) { case '\\': c = pgetc(); if (c != '\\' && c != '`' && c != '$' && (!dblquote || c != '"')) USTPUTC('\\', oout); break; case '\n': plinno++; needprompt = doprompt; break; case PEOF: startlinno = plinno; synerror("EOF in backquote substitution"); break; default: break; } USTPUTC(c, oout); } USTPUTC('\0', oout); olen = oout - stackblock(); INTOFF; ostr = ckmalloc(olen); memcpy(ostr, stackblock(), olen); setinputstring(ostr, 1); INTON; } nlpp = pbqlist; while (*nlpp) nlpp = &(*nlpp)->next; *nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist)); (*nlpp)->next = NULL; if (oldstyle) { saveprompt = doprompt; doprompt = 0; } n = list(0); if (oldstyle) { if (peektoken() != TEOF) synexpect(-1); doprompt = saveprompt; } else consumetoken(TRP); (*nlpp)->n = n; if (oldstyle) { /* * Start reading from old file again, ignoring any pushed back * tokens left from the backquote parsing */ popfile(); tokpushback = 0; } STARTSTACKSTR(out); CHECKSTRSPACE(savelen + 1, out); INTOFF; if (str) { memcpy(out, str, savelen); STADJUST(savelen, out); ckfree(str); str = NULL; } if (ostr) { ckfree(ostr); ostr = NULL; } here = saveheredoclist; if (here != NULL) { while (here->next != NULL) here = here->next; here->next = heredoclist; heredoclist = saveheredoclist; } handler = savehandler; INTON; if (quoted) USTPUTC(CTLBACKQ | CTLQUOTE, out); else USTPUTC(CTLBACKQ, out); return out; } /* * Called to parse a backslash escape sequence inside $'...'. * The backslash has already been read. */ static char * readcstyleesc(char *out) { int c, vc, i, n; unsigned int v; c = pgetc(); switch (c) { case '\0': synerror("Unterminated quoted string"); case '\n': plinno++; if (doprompt) setprompt(2); else setprompt(0); return out; case '\\': case '\'': case '"': v = c; break; case 'a': v = '\a'; break; case 'b': v = '\b'; break; case 'e': v = '\033'; break; case 'f': v = '\f'; break; case 'n': v = '\n'; break; case 'r': v = '\r'; break; case 't': v = '\t'; break; case 'v': v = '\v'; break; case 'x': v = 0; for (;;) { c = pgetc(); if (c >= '0' && c <= '9') v = (v << 4) + c - '0'; else if (c >= 'A' && c <= 'F') v = (v << 4) + c - 'A' + 10; else if (c >= 'a' && c <= 'f') v = (v << 4) + c - 'a' + 10; else break; } pungetc(); break; case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': v = c - '0'; c = pgetc(); if (c >= '0' && c <= '7') { v <<= 3; v += c - '0'; c = pgetc(); if (c >= '0' && c <= '7') { v <<= 3; v += c - '0'; } else pungetc(); } else pungetc(); break; case 'c': c = pgetc(); if (c < 0x3f || c > 0x7a || c == 0x60) synerror("Bad escape sequence"); if (c == '\\' && pgetc() != '\\') synerror("Bad escape sequence"); if (c == '?') v = 127; else v = c & 0x1f; break; case 'u': case 'U': n = c == 'U' ? 8 : 4; v = 0; for (i = 0; i < n; i++) { c = pgetc(); if (c >= '0' && c <= '9') v = (v << 4) + c - '0'; else if (c >= 'A' && c <= 'F') v = (v << 4) + c - 'A' + 10; else if (c >= 'a' && c <= 'f') v = (v << 4) + c - 'a' + 10; else synerror("Bad escape sequence"); } if (v == 0 || (v >= 0xd800 && v <= 0xdfff)) synerror("Bad escape sequence"); /* We really need iconv here. */ if (initial_localeisutf8 && v > 127) { CHECKSTRSPACE(4, out); /* * We cannot use wctomb() as the locale may have * changed. */ if (v <= 0x7ff) { USTPUTC(0xc0 | v >> 6, out); USTPUTC(0x80 | (v & 0x3f), out); return out; } else if (v <= 0xffff) { USTPUTC(0xe0 | v >> 12, out); USTPUTC(0x80 | ((v >> 6) & 0x3f), out); USTPUTC(0x80 | (v & 0x3f), out); return out; } else if (v <= 0x10ffff) { USTPUTC(0xf0 | v >> 18, out); USTPUTC(0x80 | ((v >> 12) & 0x3f), out); USTPUTC(0x80 | ((v >> 6) & 0x3f), out); USTPUTC(0x80 | (v & 0x3f), out); return out; } } if (v > 127) v = '?'; break; default: synerror("Bad escape sequence"); } vc = (char)v; /* * We can't handle NUL bytes. * POSIX says we should skip till the closing quote. */ if (vc == '\0') { while ((c = pgetc()) != '\'') { if (c == '\\') c = pgetc(); if (c == PEOF) synerror("Unterminated quoted string"); if (c == '\n') { plinno++; if (doprompt) setprompt(2); else setprompt(0); } } pungetc(); return out; } if (SQSYNTAX[vc] == CCTL) USTPUTC(CTLESC, out); USTPUTC(vc, out); return out; } /* * If eofmark is NULL, read a word or a redirection symbol. If eofmark * is not NULL, read a here document. In the latter case, eofmark is the * word which marks the end of the document and striptabs is true if * leading tabs should be stripped from the document. The argument firstc * is the first character of the input token or document. * * Because C does not have internal subroutines, I have simulated them * using goto's to implement the subroutine linkage. The following macros * will run code that appears at the end of readtoken1. */ #define PARSESUB() {goto parsesub; parsesub_return:;} #define PARSEARITH() {goto parsearith; parsearith_return:;} static int readtoken1(int firstc, char const *initialsyntax, const char *eofmark, int striptabs) { int c = firstc; char *out; int len; struct nodelist *bqlist; int quotef; int newvarnest; int level; int synentry; struct tokenstate state_static[MAXNEST_static]; int maxnest = MAXNEST_static; struct tokenstate *state = state_static; int sqiscstyle = 0; startlinno = plinno; quotef = 0; bqlist = NULL; newvarnest = 0; level = 0; state[level].syntax = initialsyntax; state[level].parenlevel = 0; state[level].category = TSTATE_TOP; STARTSTACKSTR(out); loop: { /* for each line, until end of word */ if (eofmark && eofmark != NOEOFMARK) /* set c to PEOF if at end of here document */ c = checkend(c, eofmark, striptabs); for (;;) { /* until end of line or end of word */ CHECKSTRSPACE(4, out); /* permit 4 calls to USTPUTC */ synentry = state[level].syntax[c]; switch(synentry) { case CNL: /* '\n' */ if (state[level].syntax == BASESYNTAX) goto endword; /* exit outer loop */ USTPUTC(c, out); plinno++; if (doprompt) setprompt(2); else setprompt(0); c = pgetc(); goto loop; /* continue outer loop */ case CSBACK: if (sqiscstyle) { out = readcstyleesc(out); break; } /* FALLTHROUGH */ case CWORD: USTPUTC(c, out); break; case CCTL: if (eofmark == NULL || initialsyntax != SQSYNTAX) USTPUTC(CTLESC, out); USTPUTC(c, out); break; case CBACK: /* backslash */ c = pgetc(); if (c == PEOF) { USTPUTC('\\', out); pungetc(); } else if (c == '\n') { plinno++; if (doprompt) setprompt(2); else setprompt(0); } else { if (state[level].syntax == DQSYNTAX && c != '\\' && c != '`' && c != '$' && (c != '"' || (eofmark != NULL && newvarnest == 0)) && (c != '}' || state[level].category != TSTATE_VAR_OLD)) USTPUTC('\\', out); if ((eofmark == NULL || newvarnest > 0) && state[level].syntax == BASESYNTAX) USTPUTC(CTLQUOTEMARK, out); if (SQSYNTAX[c] == CCTL) USTPUTC(CTLESC, out); USTPUTC(c, out); if ((eofmark == NULL || newvarnest > 0) && state[level].syntax == BASESYNTAX && state[level].category == TSTATE_VAR_OLD) USTPUTC(CTLQUOTEEND, out); quotef++; } break; case CSQUOTE: USTPUTC(CTLQUOTEMARK, out); state[level].syntax = SQSYNTAX; sqiscstyle = 0; break; case CDQUOTE: USTPUTC(CTLQUOTEMARK, out); state[level].syntax = DQSYNTAX; break; case CENDQUOTE: if (eofmark != NULL && newvarnest == 0) USTPUTC(c, out); else { if (state[level].category == TSTATE_VAR_OLD) USTPUTC(CTLQUOTEEND, out); state[level].syntax = BASESYNTAX; quotef++; } break; case CVAR: /* '$' */ PARSESUB(); /* parse substitution */ break; case CENDVAR: /* '}' */ if (level > 0 && ((state[level].category == TSTATE_VAR_OLD && state[level].syntax == state[level - 1].syntax) || (state[level].category == TSTATE_VAR_NEW && state[level].syntax == BASESYNTAX))) { if (state[level].category == TSTATE_VAR_NEW) newvarnest--; level--; USTPUTC(CTLENDVAR, out); } else { USTPUTC(c, out); } break; case CLP: /* '(' in arithmetic */ state[level].parenlevel++; USTPUTC(c, out); break; case CRP: /* ')' in arithmetic */ if (state[level].parenlevel > 0) { USTPUTC(c, out); --state[level].parenlevel; } else { if (pgetc_linecont() == ')') { if (level > 0 && state[level].category == TSTATE_ARITH) { level--; USTPUTC(CTLENDARI, out); } else USTPUTC(')', out); } else { /* * unbalanced parens * (don't 2nd guess - no error) */ pungetc(); USTPUTC(')', out); } } break; case CBQUOTE: /* '`' */ out = parsebackq(out, &bqlist, 1, state[level].syntax == DQSYNTAX && (eofmark == NULL || newvarnest > 0), state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX); break; case CEOF: goto endword; /* exit outer loop */ case CIGN: break; default: if (level == 0) goto endword; /* exit outer loop */ USTPUTC(c, out); } c = pgetc_macro(); } } endword: if (state[level].syntax == ARISYNTAX) synerror("Missing '))'"); if (state[level].syntax != BASESYNTAX && eofmark == NULL) synerror("Unterminated quoted string"); if (state[level].category == TSTATE_VAR_OLD || state[level].category == TSTATE_VAR_NEW) { startlinno = plinno; synerror("Missing '}'"); } if (state != state_static) parser_temp_free_upto(state); USTPUTC('\0', out); len = out - stackblock(); out = stackblock(); if (eofmark == NULL) { if ((c == '>' || c == '<') && quotef == 0 && len <= 2 && (*out == '\0' || is_digit(*out))) { parseredir(out, c); return lasttoken = TREDIR; } else { pungetc(); } } quoteflag = quotef; backquotelist = bqlist; grabstackblock(len); wordtext = out; return lasttoken = TWORD; /* end of readtoken routine */ /* * Parse a substitution. At this point, we have read the dollar sign * and nothing else. */ parsesub: { char buf[10]; int subtype; int typeloc; int flags; char *p; static const char types[] = "}-+?="; int bracketed_name = 0; /* used to handle ${[0-9]*} variables */ int linno; int length; int c1; c = pgetc_linecont(); if (c == '(') { /* $(command) or $((arith)) */ if (pgetc_linecont() == '(') { PARSEARITH(); } else { pungetc(); out = parsebackq(out, &bqlist, 0, state[level].syntax == DQSYNTAX && (eofmark == NULL || newvarnest > 0), state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX); } } else if (c == '{' || is_name(c) || is_special(c)) { USTPUTC(CTLVAR, out); typeloc = out - stackblock(); USTPUTC(VSNORMAL, out); subtype = VSNORMAL; flags = 0; if (c == '{') { bracketed_name = 1; c = pgetc_linecont(); subtype = 0; } varname: if (!is_eof(c) && is_name(c)) { length = 0; do { STPUTC(c, out); c = pgetc_linecont(); length++; } while (!is_eof(c) && is_in_name(c)); if (length == 6 && strncmp(out - length, "LINENO", length) == 0) { /* Replace the variable name with the * current line number. */ linno = plinno; if (funclinno != 0) linno -= funclinno - 1; snprintf(buf, sizeof(buf), "%d", linno); STADJUST(-6, out); STPUTS(buf, out); flags |= VSLINENO; } } else if (is_digit(c)) { if (bracketed_name) { do { STPUTC(c, out); c = pgetc_linecont(); } while (is_digit(c)); } else { - STPUTC(c, out); + USTPUTC(c, out); c = pgetc_linecont(); } } else if (is_special(c)) { c1 = c; c = pgetc_linecont(); if (subtype == 0 && c1 == '#') { subtype = VSLENGTH; if (strchr(types, c) == NULL && c != ':' && c != '#' && c != '%') goto varname; c1 = c; c = pgetc_linecont(); if (c1 != '}' && c == '}') { pungetc(); c = c1; goto varname; } pungetc(); c = c1; c1 = '#'; subtype = 0; } USTPUTC(c1, out); } else { subtype = VSERROR; if (c == '}') pungetc(); else if (c == '\n' || c == PEOF) synerror("Unexpected end of line in substitution"); else if (BASESYNTAX[c] != CCTL) USTPUTC(c, out); } if (subtype == 0) { switch (c) { case ':': flags |= VSNUL; c = pgetc_linecont(); /*FALLTHROUGH*/ default: p = strchr(types, c); if (p == NULL) { if (c == '\n' || c == PEOF) synerror("Unexpected end of line in substitution"); if (flags == VSNUL) STPUTC(':', out); if (BASESYNTAX[c] != CCTL) STPUTC(c, out); subtype = VSERROR; } else subtype = p - types + VSNORMAL; break; case '%': case '#': { int cc = c; subtype = c == '#' ? VSTRIMLEFT : VSTRIMRIGHT; c = pgetc_linecont(); if (c == cc) subtype++; else pungetc(); break; } } } else if (subtype != VSERROR) { if (subtype == VSLENGTH && c != '}') subtype = VSERROR; pungetc(); } STPUTC('=', out); if (state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX) flags |= VSQUOTE; *(stackblock() + typeloc) = subtype | flags; if (subtype != VSNORMAL) { if (level + 1 >= maxnest) { maxnest *= 2; if (state == state_static) { state = parser_temp_alloc( maxnest * sizeof(*state)); memcpy(state, state_static, MAXNEST_static * sizeof(*state)); } else state = parser_temp_realloc(state, maxnest * sizeof(*state)); } level++; state[level].parenlevel = 0; if (subtype == VSMINUS || subtype == VSPLUS || subtype == VSQUESTION || subtype == VSASSIGN) { /* * For operators that were in the Bourne shell, * inherit the double-quote state. */ state[level].syntax = state[level - 1].syntax; state[level].category = TSTATE_VAR_OLD; } else { /* * The other operators take a pattern, * so go to BASESYNTAX. * Also, ' and " are now special, even * in here documents. */ state[level].syntax = BASESYNTAX; state[level].category = TSTATE_VAR_NEW; newvarnest++; } } } else if (c == '\'' && state[level].syntax == BASESYNTAX) { /* $'cstylequotes' */ USTPUTC(CTLQUOTEMARK, out); state[level].syntax = SQSYNTAX; sqiscstyle = 1; } else { USTPUTC('$', out); pungetc(); } goto parsesub_return; } /* * Parse an arithmetic expansion (indicate start of one and set state) */ parsearith: { if (level + 1 >= maxnest) { maxnest *= 2; if (state == state_static) { state = parser_temp_alloc( maxnest * sizeof(*state)); memcpy(state, state_static, MAXNEST_static * sizeof(*state)); } else state = parser_temp_realloc(state, maxnest * sizeof(*state)); } level++; state[level].syntax = ARISYNTAX; state[level].parenlevel = 0; state[level].category = TSTATE_ARITH; USTPUTC(CTLARI, out); if (state[level - 1].syntax == DQSYNTAX) USTPUTC('"',out); else USTPUTC(' ',out); goto parsearith_return; } } /* end of readtoken */ /* * Returns true if the text contains nothing to expand (no dollar signs * or backquotes). */ static int noexpand(char *text) { char *p; char c; p = text; while ((c = *p++) != '\0') { if ( c == CTLQUOTEMARK) continue; if (c == CTLESC) p++; else if (BASESYNTAX[(int)c] == CCTL) return 0; } return 1; } /* * Return true if the argument is a legal variable name (a letter or * underscore followed by zero or more letters, underscores, and digits). */ int goodname(const char *name) { const char *p; p = name; if (! is_name(*p)) return 0; while (*++p) { if (! is_in_name(*p)) return 0; } return 1; } int isassignment(const char *p) { if (!is_name(*p)) return 0; p++; for (;;) { if (*p == '=') return 1; else if (!is_in_name(*p)) return 0; p++; } } static void consumetoken(int token) { if (readtoken() != token) synexpect(token); } /* * Called when an unexpected token is read during the parse. The argument * is the token that is expected, or -1 if more than one type of token can * occur at this point. */ static void synexpect(int token) { char msg[64]; if (token >= 0) { fmtstr(msg, 64, "%s unexpected (expecting %s)", tokname[lasttoken], tokname[token]); } else { fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]); } synerror(msg); } static void synerror(const char *msg) { if (commandname) outfmt(out2, "%s: %d: ", commandname, startlinno); else if (arg0) outfmt(out2, "%s: ", arg0); outfmt(out2, "Syntax error: %s\n", msg); error((char *)NULL); } static void setprompt(int which) { whichprompt = which; if (which == 0) return; #ifndef NO_HISTORY if (!el) #endif { out2str(getprompt(NULL)); flushout(out2); } } static int pgetc_linecont(void) { int c; while ((c = pgetc_macro()) == '\\') { c = pgetc(); if (c == '\n') { plinno++; if (doprompt) setprompt(2); else setprompt(0); } else { pungetc(); /* Allow the backslash to be pushed back. */ pushstring("\\", 1, NULL); return (pgetc()); } } return (c); } /* * called by editline -- any expansions to the prompt * should be added here. */ char * getprompt(void *unused __unused) { static char ps[PROMPTLEN]; const char *fmt; const char *pwd; int i, trim; static char internal_error[] = "??"; /* * Select prompt format. */ switch (whichprompt) { case 0: fmt = ""; break; case 1: fmt = ps1val(); break; case 2: fmt = ps2val(); break; default: return internal_error; } /* * Format prompt string. */ for (i = 0; (i < 127) && (*fmt != '\0'); i++, fmt++) if (*fmt == '\\') switch (*++fmt) { /* * Hostname. * * \h specifies just the local hostname, * \H specifies fully-qualified hostname. */ case 'h': case 'H': ps[i] = '\0'; gethostname(&ps[i], PROMPTLEN - i); /* Skip to end of hostname. */ trim = (*fmt == 'h') ? '.' : '\0'; while ((ps[i+1] != '\0') && (ps[i+1] != trim)) i++; break; /* * Working directory. * * \W specifies just the final component, * \w specifies the entire path. */ case 'W': case 'w': pwd = lookupvar("PWD"); if (pwd == NULL) pwd = "?"; if (*fmt == 'W' && *pwd == '/' && pwd[1] != '\0') strlcpy(&ps[i], strrchr(pwd, '/') + 1, PROMPTLEN - i); else strlcpy(&ps[i], pwd, PROMPTLEN - i); /* Skip to end of path. */ while (ps[i + 1] != '\0') i++; break; /* * Superuser status. * * '$' for normal users, '#' for root. */ case '$': ps[i] = (geteuid() != 0) ? '$' : '#'; break; /* * A literal \. */ case '\\': ps[i] = '\\'; break; /* * Emit unrecognized formats verbatim. */ default: ps[i++] = '\\'; ps[i] = *fmt; break; } else ps[i] = *fmt; ps[i] = '\0'; return (ps); } const char * expandstr(const char *ps) { union node n; struct jmploc jmploc; struct jmploc *const savehandler = handler; const int saveprompt = doprompt; struct parsefile *const savetopfile = getcurrentfile(); struct parser_temp *const saveparser_temp = parser_temp; const char *result = NULL; if (!setjmp(jmploc.loc)) { handler = &jmploc; parser_temp = NULL; setinputstring(ps, 1); doprompt = 0; readtoken1(pgetc(), DQSYNTAX, NOEOFMARK, 0); if (backquotelist != NULL) error("Command substitution not allowed here"); n.narg.type = NARG; n.narg.next = NULL; n.narg.text = wordtext; n.narg.backquote = backquotelist; expandarg(&n, NULL, 0); result = stackblock(); INTOFF; } handler = savehandler; doprompt = saveprompt; popfilesupto(savetopfile); if (parser_temp != saveparser_temp) { parser_temp_free_all(); parser_temp = saveparser_temp; } if (result != NULL) { INTON; } else if (exception == EXINT) raise(SIGINT); return result; } Index: projects/release-pkg/libexec/rtld-elf/map_object.c =================================================================== --- projects/release-pkg/libexec/rtld-elf/map_object.c (revision 296327) +++ projects/release-pkg/libexec/rtld-elf/map_object.c (revision 296328) @@ -1,473 +1,472 @@ /*- * 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 "debug.h" #include "rtld.h" static Elf_Ehdr *get_elf_header(int, const char *, const struct stat *); -static int convert_prot(int); /* Elf flags -> mmap protection */ static int convert_flags(int); /* Elf flags -> mmap flags */ /* * Map a shared object into memory. The "fd" argument is a file descriptor, * which must be open on the object and positioned at its beginning. * The "path" argument is a pathname that is used only for error messages. * * The return value is a pointer to a newly-allocated Obj_Entry structure * for the shared object. Returns NULL on failure. */ Obj_Entry * map_object(int fd, const char *path, const struct stat *sb) { Obj_Entry *obj; Elf_Ehdr *hdr; int i; Elf_Phdr *phdr; Elf_Phdr *phlimit; Elf_Phdr **segs; int nsegs; Elf_Phdr *phdyn; Elf_Phdr *phinterp; Elf_Phdr *phtls; caddr_t mapbase; size_t mapsize; Elf_Addr base_vaddr; Elf_Addr base_vlimit; caddr_t base_addr; int base_flags; Elf_Off data_offset; Elf_Addr data_vaddr; Elf_Addr data_vlimit; caddr_t data_addr; int data_prot; int data_flags; Elf_Addr clear_vaddr; caddr_t clear_addr; caddr_t clear_page; Elf_Addr phdr_vaddr; size_t nclear, phsize; Elf_Addr bss_vaddr; Elf_Addr bss_vlimit; caddr_t bss_addr; Elf_Word stack_flags; Elf_Addr relro_page; size_t relro_size; Elf_Addr note_start; Elf_Addr note_end; char *note_map; size_t note_map_len; hdr = get_elf_header(fd, path, sb); if (hdr == NULL) return (NULL); /* * Scan the program header entries, and save key information. * * We expect that the loadable segments are ordered by load address. */ phdr = (Elf_Phdr *) ((char *)hdr + hdr->e_phoff); phsize = hdr->e_phnum * sizeof (phdr[0]); phlimit = phdr + hdr->e_phnum; nsegs = -1; phdyn = phinterp = phtls = NULL; phdr_vaddr = 0; relro_page = 0; relro_size = 0; note_start = 0; note_end = 0; note_map = NULL; segs = alloca(sizeof(segs[0]) * hdr->e_phnum); stack_flags = RTLD_DEFAULT_STACK_PF_EXEC | PF_R | PF_W; while (phdr < phlimit) { switch (phdr->p_type) { case PT_INTERP: phinterp = phdr; break; case PT_LOAD: segs[++nsegs] = phdr; if ((segs[nsegs]->p_align & (PAGE_SIZE - 1)) != 0) { _rtld_error("%s: PT_LOAD segment %d not page-aligned", path, nsegs); goto error; } break; case PT_PHDR: phdr_vaddr = phdr->p_vaddr; phsize = phdr->p_memsz; break; case PT_DYNAMIC: phdyn = phdr; break; case PT_TLS: phtls = phdr; break; case PT_GNU_STACK: stack_flags = phdr->p_flags; break; case PT_GNU_RELRO: relro_page = phdr->p_vaddr; relro_size = phdr->p_memsz; break; case PT_NOTE: if (phdr->p_offset > PAGE_SIZE || phdr->p_offset + phdr->p_filesz > PAGE_SIZE) { note_map_len = round_page(phdr->p_offset + phdr->p_filesz) - trunc_page(phdr->p_offset); note_map = mmap(NULL, note_map_len, PROT_READ, MAP_PRIVATE, fd, trunc_page(phdr->p_offset)); if (note_map == MAP_FAILED) { _rtld_error("%s: error mapping PT_NOTE (%d)", path, errno); goto error; } note_start = (Elf_Addr)(note_map + phdr->p_offset - trunc_page(phdr->p_offset)); } else { note_start = (Elf_Addr)(char *)hdr + phdr->p_offset; } note_end = note_start + phdr->p_filesz; break; } ++phdr; } if (phdyn == NULL) { _rtld_error("%s: object is not dynamically-linked", path); goto error; } if (nsegs < 0) { _rtld_error("%s: too few PT_LOAD segments", path); goto error; } /* * Map 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]->p_vaddr + segs[nsegs]->p_memsz); mapsize = base_vlimit - base_vaddr; base_addr = (caddr_t) base_vaddr; base_flags = MAP_PRIVATE | MAP_ANON | MAP_NOCORE; if (npagesizes > 1 && round_page(segs[0]->p_filesz) >= pagesizes[1]) base_flags |= MAP_ALIGNED_SUPER; mapbase = mmap(base_addr, mapsize, PROT_NONE, base_flags, -1, 0); if (mapbase == (caddr_t) -1) { _rtld_error("%s: mmap of entire address space failed: %s", path, rtld_strerror(errno)); goto error; } if (base_addr != NULL && mapbase != base_addr) { _rtld_error("%s: mmap returned wrong address: wanted %p, got %p", path, base_addr, mapbase); goto error1; } for (i = 0; i <= nsegs; i++) { /* Overlay the segment onto the proper region. */ data_offset = trunc_page(segs[i]->p_offset); data_vaddr = trunc_page(segs[i]->p_vaddr); data_vlimit = round_page(segs[i]->p_vaddr + segs[i]->p_filesz); data_addr = mapbase + (data_vaddr - base_vaddr); data_prot = convert_prot(segs[i]->p_flags); data_flags = convert_flags(segs[i]->p_flags) | MAP_FIXED; if (mmap(data_addr, data_vlimit - data_vaddr, data_prot, data_flags | MAP_PREFAULT_READ, fd, data_offset) == (caddr_t) -1) { _rtld_error("%s: mmap of data failed: %s", path, rtld_strerror(errno)); goto error1; } /* Do BSS setup */ if (segs[i]->p_filesz != segs[i]->p_memsz) { /* Clear any BSS in the last page of the segment. */ clear_vaddr = segs[i]->p_vaddr + segs[i]->p_filesz; clear_addr = mapbase + (clear_vaddr - base_vaddr); clear_page = mapbase + (trunc_page(clear_vaddr) - base_vaddr); if ((nclear = data_vlimit - clear_vaddr) > 0) { /* Make sure the end of the segment is writable */ if ((data_prot & PROT_WRITE) == 0 && -1 == mprotect(clear_page, PAGE_SIZE, data_prot|PROT_WRITE)) { _rtld_error("%s: mprotect failed: %s", path, rtld_strerror(errno)); goto error1; } memset(clear_addr, 0, nclear); /* Reset the data protection back */ if ((data_prot & PROT_WRITE) == 0) mprotect(clear_page, PAGE_SIZE, data_prot); } /* Overlay the BSS segment onto the proper region. */ bss_vaddr = data_vlimit; bss_vlimit = round_page(segs[i]->p_vaddr + segs[i]->p_memsz); bss_addr = mapbase + (bss_vaddr - base_vaddr); if (bss_vlimit > bss_vaddr) { /* There is something to do */ if (mmap(bss_addr, bss_vlimit - bss_vaddr, data_prot, data_flags | MAP_ANON, -1, 0) == (caddr_t)-1) { _rtld_error("%s: mmap of bss failed: %s", path, rtld_strerror(errno)); goto error1; } } } if (phdr_vaddr == 0 && data_offset <= hdr->e_phoff && (data_vlimit - data_vaddr + data_offset) >= (hdr->e_phoff + hdr->e_phnum * sizeof (Elf_Phdr))) { phdr_vaddr = data_vaddr + hdr->e_phoff - data_offset; } } obj = obj_new(); if (sb != NULL) { obj->dev = sb->st_dev; obj->ino = sb->st_ino; } obj->mapbase = mapbase; obj->mapsize = mapsize; obj->textsize = round_page(segs[0]->p_vaddr + segs[0]->p_memsz) - base_vaddr; obj->vaddrbase = base_vaddr; obj->relocbase = mapbase - base_vaddr; obj->dynamic = (const Elf_Dyn *) (obj->relocbase + phdyn->p_vaddr); if (hdr->e_entry != 0) obj->entry = (caddr_t) (obj->relocbase + hdr->e_entry); if (phdr_vaddr != 0) { obj->phdr = (const Elf_Phdr *) (obj->relocbase + phdr_vaddr); } else { obj->phdr = malloc(phsize); if (obj->phdr == NULL) { obj_free(obj); _rtld_error("%s: cannot allocate program header", path); goto error1; } memcpy((char *)obj->phdr, (char *)hdr + hdr->e_phoff, phsize); obj->phdr_alloc = true; } obj->phsize = phsize; if (phinterp != NULL) obj->interp = (const char *) (obj->relocbase + phinterp->p_vaddr); if (phtls != NULL) { tls_dtv_generation++; obj->tlsindex = ++tls_max_index; obj->tlssize = phtls->p_memsz; obj->tlsalign = phtls->p_align; obj->tlsinitsize = phtls->p_filesz; obj->tlsinit = mapbase + phtls->p_vaddr; } obj->stack_flags = stack_flags; obj->relro_page = obj->relocbase + trunc_page(relro_page); obj->relro_size = round_page(relro_size); if (note_start < note_end) digest_notes(obj, note_start, note_end); if (note_map != NULL) munmap(note_map, note_map_len); munmap(hdr, PAGE_SIZE); return (obj); error1: munmap(mapbase, mapsize); error: if (note_map != NULL && note_map != MAP_FAILED) munmap(note_map, note_map_len); munmap(hdr, PAGE_SIZE); return (NULL); } static Elf_Ehdr * get_elf_header(int fd, const char *path, const struct stat *sbp) { Elf_Ehdr *hdr; /* Make sure file has enough data for the ELF header */ if (sbp != NULL && sbp->st_size < sizeof(Elf_Ehdr)) { _rtld_error("%s: invalid file format", path); return (NULL); } hdr = mmap(NULL, PAGE_SIZE, PROT_READ, MAP_PRIVATE | MAP_PREFAULT_READ, fd, 0); if (hdr == (Elf_Ehdr *)MAP_FAILED) { _rtld_error("%s: read error: %s", path, rtld_strerror(errno)); return (NULL); } /* Make sure the file is valid */ if (!IS_ELF(*hdr)) { _rtld_error("%s: invalid file format", path); goto error; } if (hdr->e_ident[EI_CLASS] != ELF_TARG_CLASS || hdr->e_ident[EI_DATA] != ELF_TARG_DATA) { _rtld_error("%s: unsupported file layout", path); goto error; } if (hdr->e_ident[EI_VERSION] != EV_CURRENT || hdr->e_version != EV_CURRENT) { _rtld_error("%s: unsupported file version", path); goto error; } if (hdr->e_type != ET_EXEC && hdr->e_type != ET_DYN) { _rtld_error("%s: unsupported file type", path); goto error; } if (hdr->e_machine != ELF_TARG_MACH) { _rtld_error("%s: unsupported machine", path); goto error; } /* * 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)) { _rtld_error( "%s: invalid shared object: e_phentsize != sizeof(Elf_Phdr)", path); goto error; } if (hdr->e_phoff + hdr->e_phnum * sizeof(Elf_Phdr) > (size_t)PAGE_SIZE) { _rtld_error("%s: program header too large", path); goto error; } return (hdr); error: munmap(hdr, PAGE_SIZE); return (NULL); } void obj_free(Obj_Entry *obj) { Objlist_Entry *elm; if (obj->tls_done) free_tls_offset(obj); while (obj->needed != NULL) { Needed_Entry *needed = obj->needed; obj->needed = needed->next; free(needed); } while (!STAILQ_EMPTY(&obj->names)) { Name_Entry *entry = STAILQ_FIRST(&obj->names); STAILQ_REMOVE_HEAD(&obj->names, link); free(entry); } while (!STAILQ_EMPTY(&obj->dldags)) { elm = STAILQ_FIRST(&obj->dldags); STAILQ_REMOVE_HEAD(&obj->dldags, link); free(elm); } while (!STAILQ_EMPTY(&obj->dagmembers)) { elm = STAILQ_FIRST(&obj->dagmembers); STAILQ_REMOVE_HEAD(&obj->dagmembers, link); free(elm); } if (obj->vertab) free(obj->vertab); if (obj->origin_path) free(obj->origin_path); if (obj->z_origin) free(obj->rpath); if (obj->priv) free(obj->priv); if (obj->path) free(obj->path); if (obj->phdr_alloc) free((void *)obj->phdr); free(obj); } Obj_Entry * obj_new(void) { Obj_Entry *obj; obj = CNEW(Obj_Entry); STAILQ_INIT(&obj->dldags); STAILQ_INIT(&obj->dagmembers); STAILQ_INIT(&obj->names); return obj; } /* * Given a set of ELF protection flags, return the corresponding protection * flags for MMAP. */ -static int +int convert_prot(int elfflags) { int prot = 0; if (elfflags & PF_R) prot |= PROT_READ; if (elfflags & PF_W) prot |= PROT_WRITE; if (elfflags & PF_X) prot |= PROT_EXEC; return prot; } static int convert_flags(int elfflags) { int flags = MAP_PRIVATE; /* All mappings are private */ /* * Readonly mappings are marked "MAP_NOCORE", because they can be * reconstructed by a debugger. */ if (!(elfflags & PF_W)) flags |= MAP_NOCORE; return flags; } Index: projects/release-pkg/libexec/rtld-elf/rtld.c =================================================================== --- projects/release-pkg/libexec/rtld-elf/rtld.c (revision 296327) +++ projects/release-pkg/libexec/rtld-elf/rtld.c (revision 296328) @@ -1,5184 +1,5207 @@ /*- * Copyright 1996, 1997, 1998, 1999, 2000 John D. Polstra. * Copyright 2003 Alexander Kabaev . * Copyright 2009-2012 Konstantin Belousov . * Copyright 2012 John Marino . * 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. * * $FreeBSD$ */ /* * Dynamic linker for ELF. * * John Polstra . */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "debug.h" #include "rtld.h" #include "libmap.h" #include "paths.h" #include "rtld_tls.h" #include "rtld_printf.h" #include "notes.h" /* Types. */ typedef void (*func_ptr_type)(); typedef void * (*path_enum_proc) (const char *path, size_t len, void *arg); /* * Function declarations. */ static const char *basename(const char *); static void digest_dynamic1(Obj_Entry *, int, const Elf_Dyn **, const Elf_Dyn **, const Elf_Dyn **); static void digest_dynamic2(Obj_Entry *, const Elf_Dyn *, const Elf_Dyn *, const Elf_Dyn *); static void digest_dynamic(Obj_Entry *, int); static Obj_Entry *digest_phdr(const Elf_Phdr *, int, caddr_t, const char *); static Obj_Entry *dlcheck(void *); static Obj_Entry *dlopen_object(const char *name, int fd, Obj_Entry *refobj, int lo_flags, int mode, RtldLockState *lockstate); static Obj_Entry *do_load_object(int, const char *, char *, struct stat *, int); static int do_search_info(const Obj_Entry *obj, int, struct dl_serinfo *); static bool donelist_check(DoneList *, const Obj_Entry *); static void errmsg_restore(char *); static char *errmsg_save(void); static void *fill_search_info(const char *, size_t, void *); static char *find_library(const char *, const Obj_Entry *, int *); static const char *gethints(bool); static void init_dag(Obj_Entry *); static void init_pagesizes(Elf_Auxinfo **aux_info); static void init_rtld(caddr_t, Elf_Auxinfo **); static void initlist_add_neededs(Needed_Entry *, Objlist *); static void initlist_add_objects(Obj_Entry *, Obj_Entry *, Objlist *); static void linkmap_add(Obj_Entry *); static void linkmap_delete(Obj_Entry *); static void load_filtees(Obj_Entry *, int flags, RtldLockState *); static void unload_filtees(Obj_Entry *); static int load_needed_objects(Obj_Entry *, int); static int load_preload_objects(void); static Obj_Entry *load_object(const char *, int fd, const Obj_Entry *, int); static void map_stacks_exec(RtldLockState *); static Obj_Entry *obj_from_addr(const void *); static void objlist_call_fini(Objlist *, Obj_Entry *, RtldLockState *); static void objlist_call_init(Objlist *, RtldLockState *); static void objlist_clear(Objlist *); static Objlist_Entry *objlist_find(Objlist *, const Obj_Entry *); static void objlist_init(Objlist *); static void objlist_push_head(Objlist *, Obj_Entry *); static void objlist_push_tail(Objlist *, Obj_Entry *); static void objlist_put_after(Objlist *, Obj_Entry *, Obj_Entry *); static void objlist_remove(Objlist *, Obj_Entry *); static int parse_libdir(const char *); static void *path_enumerate(const char *, path_enum_proc, void *); static int relocate_object_dag(Obj_Entry *root, bool bind_now, Obj_Entry *rtldobj, int flags, RtldLockState *lockstate); static int relocate_object(Obj_Entry *obj, bool bind_now, Obj_Entry *rtldobj, int flags, RtldLockState *lockstate); static int relocate_objects(Obj_Entry *, bool, Obj_Entry *, int, RtldLockState *); static int resolve_objects_ifunc(Obj_Entry *first, bool bind_now, int flags, RtldLockState *lockstate); static int rtld_dirname(const char *, char *); static int rtld_dirname_abs(const char *, char *); static void *rtld_dlopen(const char *name, int fd, int mode); static void rtld_exit(void); static char *search_library_path(const char *, const char *); static char *search_library_pathfds(const char *, const char *, int *); static const void **get_program_var_addr(const char *, RtldLockState *); static void set_program_var(const char *, const void *); static int symlook_default(SymLook *, const Obj_Entry *refobj); static int symlook_global(SymLook *, DoneList *); static void symlook_init_from_req(SymLook *, const SymLook *); static int symlook_list(SymLook *, const Objlist *, DoneList *); static int symlook_needed(SymLook *, const Needed_Entry *, DoneList *); static int symlook_obj1_sysv(SymLook *, const Obj_Entry *); static int symlook_obj1_gnu(SymLook *, const Obj_Entry *); static void trace_loaded_objects(Obj_Entry *); static void unlink_object(Obj_Entry *); static void unload_object(Obj_Entry *); static void unref_dag(Obj_Entry *); static void ref_dag(Obj_Entry *); static char *origin_subst_one(Obj_Entry *, char *, const char *, const char *, bool); static char *origin_subst(Obj_Entry *, char *); static bool obj_resolve_origin(Obj_Entry *obj); static void preinit_main(void); static int rtld_verify_versions(const Objlist *); static int rtld_verify_object_versions(Obj_Entry *); static void object_add_name(Obj_Entry *, const char *); static int object_match_name(const Obj_Entry *, const char *); static void ld_utrace_log(int, void *, void *, size_t, int, const char *); static void rtld_fill_dl_phdr_info(const Obj_Entry *obj, struct dl_phdr_info *phdr_info); static uint32_t gnu_hash(const char *); static bool matched_symbol(SymLook *, const Obj_Entry *, Sym_Match_Result *, const unsigned long); void r_debug_state(struct r_debug *, struct link_map *) __noinline __exported; void _r_debug_postinit(struct link_map *) __noinline __exported; int __sys_openat(int, const char *, int, ...); /* * Data declarations. */ static char *error_message; /* Message for dlerror(), or NULL */ struct r_debug r_debug __exported; /* for GDB; */ static bool libmap_disable; /* Disable libmap */ static bool ld_loadfltr; /* Immediate filters processing */ static char *libmap_override; /* Maps to use in addition to libmap.conf */ static bool trust; /* False for setuid and setgid programs */ static bool dangerous_ld_env; /* True if environment variables have been used to affect the libraries loaded */ static char *ld_bind_now; /* Environment variable for immediate binding */ static char *ld_debug; /* Environment variable for debugging */ static char *ld_library_path; /* Environment variable for search path */ static char *ld_library_dirs; /* Environment variable for library descriptors */ static char *ld_preload; /* Environment variable for libraries to load first */ static char *ld_elf_hints_path; /* Environment variable for alternative hints path */ static char *ld_tracing; /* Called from ldd to print libs */ static char *ld_utrace; /* Use utrace() to log events. */ static struct obj_entry_q obj_list; /* Queue of all loaded objects */ static Obj_Entry *obj_main; /* The main program shared object */ static Obj_Entry obj_rtld; /* The dynamic linker shared object */ static unsigned int obj_count; /* Number of objects in obj_list */ static unsigned int obj_loads; /* Number of loads of objects (gen count) */ static Objlist list_global = /* Objects dlopened with RTLD_GLOBAL */ STAILQ_HEAD_INITIALIZER(list_global); static Objlist list_main = /* Objects loaded at program startup */ STAILQ_HEAD_INITIALIZER(list_main); static Objlist list_fini = /* Objects needing fini() calls */ STAILQ_HEAD_INITIALIZER(list_fini); Elf_Sym sym_zero; /* For resolving undefined weak refs. */ #define GDB_STATE(s,m) r_debug.r_state = s; r_debug_state(&r_debug,m); extern Elf_Dyn _DYNAMIC; #pragma weak _DYNAMIC #ifndef RTLD_IS_DYNAMIC #define RTLD_IS_DYNAMIC() (&_DYNAMIC != NULL) #endif int dlclose(void *) __exported; char *dlerror(void) __exported; void *dlopen(const char *, int) __exported; void *fdlopen(int, int) __exported; void *dlsym(void *, const char *) __exported; dlfunc_t dlfunc(void *, const char *) __exported; void *dlvsym(void *, const char *, const char *) __exported; int dladdr(const void *, Dl_info *) __exported; void dllockinit(void *, void *(*)(void *), void (*)(void *), void (*)(void *), void (*)(void *), void (*)(void *), void (*)(void *)) __exported; int dlinfo(void *, int , void *) __exported; int dl_iterate_phdr(__dl_iterate_hdr_callback, void *) __exported; int _rtld_addr_phdr(const void *, struct dl_phdr_info *) __exported; int _rtld_get_stack_prot(void) __exported; int _rtld_is_dlopened(void *) __exported; void _rtld_error(const char *, ...) __exported; int npagesizes, osreldate; size_t *pagesizes; long __stack_chk_guard[8] = {0, 0, 0, 0, 0, 0, 0, 0}; static int stack_prot = PROT_READ | PROT_WRITE | RTLD_DEFAULT_STACK_EXEC; static int max_stack_flags; /* * Global declarations normally provided by crt1. The dynamic linker is * not built with crt1, so we have to provide them ourselves. */ char *__progname; char **environ; /* * Used to pass argc, argv to init functions. */ int main_argc; char **main_argv; /* * Globals to control TLS allocation. */ size_t tls_last_offset; /* Static TLS offset of last module */ size_t tls_last_size; /* Static TLS size of last module */ size_t tls_static_space; /* Static TLS space allocated */ size_t tls_static_max_align; int tls_dtv_generation = 1; /* Used to detect when dtv size changes */ int tls_max_index = 1; /* Largest module index allocated */ bool ld_library_path_rpath = false; /* * Globals for path names, and such */ char *ld_elf_hints_default = _PATH_ELF_HINTS; char *ld_path_libmap_conf = _PATH_LIBMAP_CONF; char *ld_path_rtld = _PATH_RTLD; char *ld_standard_library_path = STANDARD_LIBRARY_PATH; char *ld_env_prefix = LD_; /* * Fill in a DoneList with an allocation large enough to hold all of * the currently-loaded objects. Keep this as a macro since it calls * alloca and we want that to occur within the scope of the caller. */ #define donelist_init(dlp) \ ((dlp)->objs = alloca(obj_count * sizeof (dlp)->objs[0]), \ assert((dlp)->objs != NULL), \ (dlp)->num_alloc = obj_count, \ (dlp)->num_used = 0) #define UTRACE_DLOPEN_START 1 #define UTRACE_DLOPEN_STOP 2 #define UTRACE_DLCLOSE_START 3 #define UTRACE_DLCLOSE_STOP 4 #define UTRACE_LOAD_OBJECT 5 #define UTRACE_UNLOAD_OBJECT 6 #define UTRACE_ADD_RUNDEP 7 #define UTRACE_PRELOAD_FINISHED 8 #define UTRACE_INIT_CALL 9 #define UTRACE_FINI_CALL 10 #define UTRACE_DLSYM_START 11 #define UTRACE_DLSYM_STOP 12 struct utrace_rtld { char sig[4]; /* 'RTLD' */ int event; void *handle; void *mapbase; /* Used for 'parent' and 'init/fini' */ size_t mapsize; int refcnt; /* Used for 'mode' */ char name[MAXPATHLEN]; }; #define LD_UTRACE(e, h, mb, ms, r, n) do { \ if (ld_utrace != NULL) \ ld_utrace_log(e, h, mb, ms, r, n); \ } while (0) static void ld_utrace_log(int event, void *handle, void *mapbase, size_t mapsize, int refcnt, const char *name) { struct utrace_rtld ut; ut.sig[0] = 'R'; ut.sig[1] = 'T'; ut.sig[2] = 'L'; ut.sig[3] = 'D'; ut.event = event; ut.handle = handle; ut.mapbase = mapbase; ut.mapsize = mapsize; ut.refcnt = refcnt; bzero(ut.name, sizeof(ut.name)); if (name) strlcpy(ut.name, name, sizeof(ut.name)); utrace(&ut, sizeof(ut)); } #ifdef RTLD_VARIANT_ENV_NAMES /* * construct the env variable based on the type of binary that's * running. */ static inline const char * _LD(const char *var) { static char buffer[128]; strlcpy(buffer, ld_env_prefix, sizeof(buffer)); strlcat(buffer, var, sizeof(buffer)); return (buffer); } #else #define _LD(x) LD_ x #endif /* * Main entry point for dynamic linking. The first argument is the * stack pointer. The stack is expected to be laid out as described * in the SVR4 ABI specification, Intel 386 Processor Supplement. * Specifically, the stack pointer points to a word containing * ARGC. Following that in the stack is a null-terminated sequence * of pointers to argument strings. Then comes a null-terminated * sequence of pointers to environment strings. Finally, there is a * sequence of "auxiliary vector" entries. * * The second argument points to a place to store the dynamic linker's * exit procedure pointer and the third to a place to store the main * program's object. * * The return value is the main program's entry point. */ func_ptr_type _rtld(Elf_Addr *sp, func_ptr_type *exit_proc, Obj_Entry **objp) { Elf_Auxinfo *aux_info[AT_COUNT]; int i; int argc; char **argv; char **env; Elf_Auxinfo *aux; Elf_Auxinfo *auxp; const char *argv0; Objlist_Entry *entry; Obj_Entry *obj; Obj_Entry *preload_tail; Obj_Entry *last_interposer; Objlist initlist; RtldLockState lockstate; char *library_path_rpath; int mib[2]; size_t len; /* * On entry, the dynamic linker itself has not been relocated yet. * Be very careful not to reference any global data until after * init_rtld has returned. It is OK to reference file-scope statics * and string constants, and to call static and global functions. */ /* Find the auxiliary vector on the stack. */ argc = *sp++; argv = (char **) sp; sp += argc + 1; /* Skip over arguments and NULL terminator */ env = (char **) sp; while (*sp++ != 0) /* Skip over environment, and NULL terminator */ ; aux = (Elf_Auxinfo *) sp; /* Digest the auxiliary vector. */ for (i = 0; i < AT_COUNT; i++) aux_info[i] = NULL; for (auxp = aux; auxp->a_type != AT_NULL; auxp++) { if (auxp->a_type < AT_COUNT) aux_info[auxp->a_type] = auxp; } /* Initialize and relocate ourselves. */ assert(aux_info[AT_BASE] != NULL); init_rtld((caddr_t) aux_info[AT_BASE]->a_un.a_ptr, aux_info); __progname = obj_rtld.path; argv0 = argv[0] != NULL ? argv[0] : "(null)"; environ = env; main_argc = argc; main_argv = argv; if (aux_info[AT_CANARY] != NULL && aux_info[AT_CANARY]->a_un.a_ptr != NULL) { i = aux_info[AT_CANARYLEN]->a_un.a_val; if (i > sizeof(__stack_chk_guard)) i = sizeof(__stack_chk_guard); memcpy(__stack_chk_guard, aux_info[AT_CANARY]->a_un.a_ptr, i); } else { mib[0] = CTL_KERN; mib[1] = KERN_ARND; len = sizeof(__stack_chk_guard); if (sysctl(mib, 2, __stack_chk_guard, &len, NULL, 0) == -1 || len != sizeof(__stack_chk_guard)) { /* If sysctl was unsuccessful, use the "terminator canary". */ ((unsigned char *)(void *)__stack_chk_guard)[0] = 0; ((unsigned char *)(void *)__stack_chk_guard)[1] = 0; ((unsigned char *)(void *)__stack_chk_guard)[2] = '\n'; ((unsigned char *)(void *)__stack_chk_guard)[3] = 255; } } trust = !issetugid(); md_abi_variant_hook(aux_info); ld_bind_now = getenv(_LD("BIND_NOW")); /* * If the process is tainted, then we un-set the dangerous environment * variables. The process will be marked as tainted until setuid(2) * is called. If any child process calls setuid(2) we do not want any * future processes to honor the potentially un-safe variables. */ if (!trust) { if (unsetenv(_LD("PRELOAD")) || unsetenv(_LD("LIBMAP")) || unsetenv(_LD("LIBRARY_PATH")) || unsetenv(_LD("LIBRARY_PATH_FDS")) || unsetenv(_LD("LIBMAP_DISABLE")) || unsetenv(_LD("DEBUG")) || unsetenv(_LD("ELF_HINTS_PATH")) || unsetenv(_LD("LOADFLTR")) || unsetenv(_LD("LIBRARY_PATH_RPATH"))) { _rtld_error("environment corrupt; aborting"); rtld_die(); } } ld_debug = getenv(_LD("DEBUG")); libmap_disable = getenv(_LD("LIBMAP_DISABLE")) != NULL; libmap_override = getenv(_LD("LIBMAP")); ld_library_path = getenv(_LD("LIBRARY_PATH")); ld_library_dirs = getenv(_LD("LIBRARY_PATH_FDS")); ld_preload = getenv(_LD("PRELOAD")); ld_elf_hints_path = getenv(_LD("ELF_HINTS_PATH")); ld_loadfltr = getenv(_LD("LOADFLTR")) != NULL; library_path_rpath = getenv(_LD("LIBRARY_PATH_RPATH")); if (library_path_rpath != NULL) { if (library_path_rpath[0] == 'y' || library_path_rpath[0] == 'Y' || library_path_rpath[0] == '1') ld_library_path_rpath = true; else ld_library_path_rpath = false; } dangerous_ld_env = libmap_disable || (libmap_override != NULL) || (ld_library_path != NULL) || (ld_preload != NULL) || (ld_elf_hints_path != NULL) || ld_loadfltr; ld_tracing = getenv(_LD("TRACE_LOADED_OBJECTS")); ld_utrace = getenv(_LD("UTRACE")); if ((ld_elf_hints_path == NULL) || strlen(ld_elf_hints_path) == 0) ld_elf_hints_path = ld_elf_hints_default; if (ld_debug != NULL && *ld_debug != '\0') debug = 1; dbg("%s is initialized, base address = %p", __progname, (caddr_t) aux_info[AT_BASE]->a_un.a_ptr); dbg("RTLD dynamic = %p", obj_rtld.dynamic); dbg("RTLD pltgot = %p", obj_rtld.pltgot); dbg("initializing thread locks"); lockdflt_init(); /* * Load the main program, or process its program header if it is * already loaded. */ if (aux_info[AT_EXECFD] != NULL) { /* Load the main program. */ int fd = aux_info[AT_EXECFD]->a_un.a_val; dbg("loading main program"); obj_main = map_object(fd, argv0, NULL); close(fd); if (obj_main == NULL) rtld_die(); max_stack_flags = obj->stack_flags; } else { /* Main program already loaded. */ const Elf_Phdr *phdr; int phnum; caddr_t entry; dbg("processing main program's program header"); assert(aux_info[AT_PHDR] != NULL); phdr = (const Elf_Phdr *) aux_info[AT_PHDR]->a_un.a_ptr; assert(aux_info[AT_PHNUM] != NULL); phnum = aux_info[AT_PHNUM]->a_un.a_val; assert(aux_info[AT_PHENT] != NULL); assert(aux_info[AT_PHENT]->a_un.a_val == sizeof(Elf_Phdr)); assert(aux_info[AT_ENTRY] != NULL); entry = (caddr_t) aux_info[AT_ENTRY]->a_un.a_ptr; if ((obj_main = digest_phdr(phdr, phnum, entry, argv0)) == NULL) rtld_die(); } if (aux_info[AT_EXECPATH] != 0) { char *kexecpath; char buf[MAXPATHLEN]; kexecpath = aux_info[AT_EXECPATH]->a_un.a_ptr; dbg("AT_EXECPATH %p %s", kexecpath, kexecpath); if (kexecpath[0] == '/') obj_main->path = kexecpath; else if (getcwd(buf, sizeof(buf)) == NULL || strlcat(buf, "/", sizeof(buf)) >= sizeof(buf) || strlcat(buf, kexecpath, sizeof(buf)) >= sizeof(buf)) obj_main->path = xstrdup(argv0); else obj_main->path = xstrdup(buf); } else { dbg("No AT_EXECPATH"); obj_main->path = xstrdup(argv0); } dbg("obj_main path %s", obj_main->path); obj_main->mainprog = true; if (aux_info[AT_STACKPROT] != NULL && aux_info[AT_STACKPROT]->a_un.a_val != 0) stack_prot = aux_info[AT_STACKPROT]->a_un.a_val; #ifndef COMPAT_32BIT /* * Get the actual dynamic linker pathname from the executable if * possible. (It should always be possible.) That ensures that * gdb will find the right dynamic linker even if a non-standard * one is being used. */ if (obj_main->interp != NULL && strcmp(obj_main->interp, obj_rtld.path) != 0) { free(obj_rtld.path); obj_rtld.path = xstrdup(obj_main->interp); __progname = obj_rtld.path; } #endif digest_dynamic(obj_main, 0); dbg("%s valid_hash_sysv %d valid_hash_gnu %d dynsymcount %d", obj_main->path, obj_main->valid_hash_sysv, obj_main->valid_hash_gnu, obj_main->dynsymcount); linkmap_add(obj_main); linkmap_add(&obj_rtld); /* Link the main program into the list of objects. */ TAILQ_INSERT_HEAD(&obj_list, obj_main, next); obj_count++; obj_loads++; /* Initialize a fake symbol for resolving undefined weak references. */ sym_zero.st_info = ELF_ST_INFO(STB_GLOBAL, STT_NOTYPE); sym_zero.st_shndx = SHN_UNDEF; sym_zero.st_value = -(uintptr_t)obj_main->relocbase; if (!libmap_disable) libmap_disable = (bool)lm_init(libmap_override); dbg("loading LD_PRELOAD libraries"); if (load_preload_objects() == -1) rtld_die(); preload_tail = globallist_curr(TAILQ_LAST(&obj_list, obj_entry_q)); dbg("loading needed objects"); if (load_needed_objects(obj_main, 0) == -1) rtld_die(); /* Make a list of all objects loaded at startup. */ last_interposer = obj_main; TAILQ_FOREACH(obj, &obj_list, next) { if (obj->marker) continue; if (obj->z_interpose && obj != obj_main) { objlist_put_after(&list_main, last_interposer, obj); last_interposer = obj; } else { objlist_push_tail(&list_main, obj); } obj->refcount++; } dbg("checking for required versions"); if (rtld_verify_versions(&list_main) == -1 && !ld_tracing) rtld_die(); if (ld_tracing) { /* We're done */ trace_loaded_objects(obj_main); exit(0); } if (getenv(_LD("DUMP_REL_PRE")) != NULL) { dump_relocations(obj_main); exit (0); } /* * Processing tls relocations requires having the tls offsets * initialized. Prepare offsets before starting initial * relocation processing. */ dbg("initializing initial thread local storage offsets"); STAILQ_FOREACH(entry, &list_main, link) { /* * Allocate all the initial objects out of the static TLS * block even if they didn't ask for it. */ allocate_tls_offset(entry->obj); } if (relocate_objects(obj_main, ld_bind_now != NULL && *ld_bind_now != '\0', &obj_rtld, SYMLOOK_EARLY, NULL) == -1) rtld_die(); dbg("doing copy relocations"); if (do_copy_relocations(obj_main) == -1) rtld_die(); if (getenv(_LD("DUMP_REL_POST")) != NULL) { dump_relocations(obj_main); exit (0); } /* * Setup TLS for main thread. This must be done after the * relocations are processed, since tls initialization section * might be the subject for relocations. */ dbg("initializing initial thread local storage"); allocate_initial_tls(globallist_curr(TAILQ_FIRST(&obj_list))); dbg("initializing key program variables"); set_program_var("__progname", argv[0] != NULL ? basename(argv[0]) : ""); set_program_var("environ", env); set_program_var("__elf_aux_vector", aux); /* Make a list of init functions to call. */ objlist_init(&initlist); initlist_add_objects(globallist_curr(TAILQ_FIRST(&obj_list)), preload_tail, &initlist); r_debug_state(NULL, &obj_main->linkmap); /* say hello to gdb! */ map_stacks_exec(NULL); dbg("resolving ifuncs"); if (resolve_objects_ifunc(obj_main, ld_bind_now != NULL && *ld_bind_now != '\0', SYMLOOK_EARLY, NULL) == -1) rtld_die(); if (!obj_main->crt_no_init) { /* * Make sure we don't call the main program's init and fini * functions for binaries linked with old crt1 which calls * _init itself. */ obj_main->init = obj_main->fini = (Elf_Addr)NULL; obj_main->preinit_array = obj_main->init_array = obj_main->fini_array = (Elf_Addr)NULL; } wlock_acquire(rtld_bind_lock, &lockstate); if (obj_main->crt_no_init) preinit_main(); objlist_call_init(&initlist, &lockstate); _r_debug_postinit(&obj_main->linkmap); objlist_clear(&initlist); dbg("loading filtees"); TAILQ_FOREACH(obj, &obj_list, next) { if (obj->marker) continue; if (ld_loadfltr || obj->z_loadfltr) load_filtees(obj, 0, &lockstate); } lock_release(rtld_bind_lock, &lockstate); dbg("transferring control to program entry point = %p", obj_main->entry); /* Return the exit procedure and the program entry point. */ *exit_proc = rtld_exit; *objp = obj_main; return (func_ptr_type) obj_main->entry; } void * rtld_resolve_ifunc(const Obj_Entry *obj, const Elf_Sym *def) { void *ptr; Elf_Addr target; ptr = (void *)make_function_pointer(def, obj); target = ((Elf_Addr (*)(void))ptr)(); return ((void *)target); } Elf_Addr _rtld_bind(Obj_Entry *obj, Elf_Size reloff) { const Elf_Rel *rel; const Elf_Sym *def; const Obj_Entry *defobj; Elf_Addr *where; Elf_Addr target; RtldLockState lockstate; rlock_acquire(rtld_bind_lock, &lockstate); if (sigsetjmp(lockstate.env, 0) != 0) lock_upgrade(rtld_bind_lock, &lockstate); if (obj->pltrel) rel = (const Elf_Rel *) ((caddr_t) obj->pltrel + reloff); else rel = (const Elf_Rel *) ((caddr_t) obj->pltrela + reloff); where = (Elf_Addr *) (obj->relocbase + rel->r_offset); def = find_symdef(ELF_R_SYM(rel->r_info), obj, &defobj, true, NULL, &lockstate); if (def == NULL) rtld_die(); if (ELF_ST_TYPE(def->st_info) == STT_GNU_IFUNC) target = (Elf_Addr)rtld_resolve_ifunc(defobj, def); else target = (Elf_Addr)(defobj->relocbase + def->st_value); dbg("\"%s\" in \"%s\" ==> %p in \"%s\"", defobj->strtab + def->st_name, basename(obj->path), (void *)target, basename(defobj->path)); /* * Write the new contents for the jmpslot. Note that depending on * architecture, the value which we need to return back to the * lazy binding trampoline may or may not be the target * address. The value returned from reloc_jmpslot() is the value * that the trampoline needs. */ target = reloc_jmpslot(where, target, defobj, obj, rel); lock_release(rtld_bind_lock, &lockstate); return target; } /* * Error reporting function. Use it like printf. If formats the message * into a buffer, and sets things up so that the next call to dlerror() * will return the message. */ void _rtld_error(const char *fmt, ...) { static char buf[512]; va_list ap; va_start(ap, fmt); rtld_vsnprintf(buf, sizeof buf, fmt, ap); error_message = buf; va_end(ap); } /* * Return a dynamically-allocated copy of the current error message, if any. */ static char * errmsg_save(void) { return error_message == NULL ? NULL : xstrdup(error_message); } /* * Restore the current error message from a copy which was previously saved * by errmsg_save(). The copy is freed. */ static void errmsg_restore(char *saved_msg) { if (saved_msg == NULL) error_message = NULL; else { _rtld_error("%s", saved_msg); free(saved_msg); } } static const char * basename(const char *name) { const char *p = strrchr(name, '/'); return p != NULL ? p + 1 : name; } static struct utsname uts; static char * origin_subst_one(Obj_Entry *obj, char *real, const char *kw, const char *subst, bool may_free) { char *p, *p1, *res, *resp; int subst_len, kw_len, subst_count, old_len, new_len; kw_len = strlen(kw); /* * First, count the number of the keyword occurences, to * preallocate the final string. */ for (p = real, subst_count = 0;; p = p1 + kw_len, subst_count++) { p1 = strstr(p, kw); if (p1 == NULL) break; } /* * If the keyword is not found, just return. * * Return non-substituted string if resolution failed. We * cannot do anything more reasonable, the failure mode of the * caller is unresolved library anyway. */ if (subst_count == 0 || (obj != NULL && !obj_resolve_origin(obj))) return (may_free ? real : xstrdup(real)); if (obj != NULL) subst = obj->origin_path; /* * There is indeed something to substitute. Calculate the * length of the resulting string, and allocate it. */ subst_len = strlen(subst); old_len = strlen(real); new_len = old_len + (subst_len - kw_len) * subst_count; res = xmalloc(new_len + 1); /* * Now, execute the substitution loop. */ for (p = real, resp = res, *resp = '\0';;) { p1 = strstr(p, kw); if (p1 != NULL) { /* Copy the prefix before keyword. */ memcpy(resp, p, p1 - p); resp += p1 - p; /* Keyword replacement. */ memcpy(resp, subst, subst_len); resp += subst_len; *resp = '\0'; p = p1 + kw_len; } else break; } /* Copy to the end of string and finish. */ strcat(resp, p); if (may_free) free(real); return (res); } static char * origin_subst(Obj_Entry *obj, char *real) { char *res1, *res2, *res3, *res4; if (obj == NULL || !trust) return (xstrdup(real)); if (uts.sysname[0] == '\0') { if (uname(&uts) != 0) { _rtld_error("utsname failed: %d", errno); return (NULL); } } res1 = origin_subst_one(obj, real, "$ORIGIN", NULL, false); res2 = origin_subst_one(NULL, res1, "$OSNAME", uts.sysname, true); res3 = origin_subst_one(NULL, res2, "$OSREL", uts.release, true); res4 = origin_subst_one(NULL, res3, "$PLATFORM", uts.machine, true); return (res4); } void rtld_die(void) { const char *msg = dlerror(); if (msg == NULL) msg = "Fatal error"; rtld_fdputstr(STDERR_FILENO, msg); rtld_fdputchar(STDERR_FILENO, '\n'); _exit(1); } /* * Process a shared object's DYNAMIC section, and save the important * information in its Obj_Entry structure. */ static void digest_dynamic1(Obj_Entry *obj, int early, const Elf_Dyn **dyn_rpath, const Elf_Dyn **dyn_soname, const Elf_Dyn **dyn_runpath) { const Elf_Dyn *dynp; Needed_Entry **needed_tail = &obj->needed; Needed_Entry **needed_filtees_tail = &obj->needed_filtees; Needed_Entry **needed_aux_filtees_tail = &obj->needed_aux_filtees; const Elf_Hashelt *hashtab; const Elf32_Word *hashval; Elf32_Word bkt, nmaskwords; int bloom_size32; int plttype = DT_REL; *dyn_rpath = NULL; *dyn_soname = NULL; *dyn_runpath = NULL; obj->bind_now = false; for (dynp = obj->dynamic; dynp->d_tag != DT_NULL; dynp++) { switch (dynp->d_tag) { case DT_REL: obj->rel = (const Elf_Rel *) (obj->relocbase + dynp->d_un.d_ptr); break; case DT_RELSZ: obj->relsize = dynp->d_un.d_val; break; case DT_RELENT: assert(dynp->d_un.d_val == sizeof(Elf_Rel)); break; case DT_JMPREL: obj->pltrel = (const Elf_Rel *) (obj->relocbase + dynp->d_un.d_ptr); break; case DT_PLTRELSZ: obj->pltrelsize = dynp->d_un.d_val; break; case DT_RELA: obj->rela = (const Elf_Rela *) (obj->relocbase + dynp->d_un.d_ptr); break; case DT_RELASZ: obj->relasize = dynp->d_un.d_val; break; case DT_RELAENT: assert(dynp->d_un.d_val == sizeof(Elf_Rela)); break; case DT_PLTREL: plttype = dynp->d_un.d_val; assert(dynp->d_un.d_val == DT_REL || plttype == DT_RELA); break; case DT_SYMTAB: obj->symtab = (const Elf_Sym *) (obj->relocbase + dynp->d_un.d_ptr); break; case DT_SYMENT: assert(dynp->d_un.d_val == sizeof(Elf_Sym)); break; case DT_STRTAB: obj->strtab = (const char *) (obj->relocbase + dynp->d_un.d_ptr); break; case DT_STRSZ: obj->strsize = dynp->d_un.d_val; break; case DT_VERNEED: obj->verneed = (const Elf_Verneed *) (obj->relocbase + dynp->d_un.d_val); break; case DT_VERNEEDNUM: obj->verneednum = dynp->d_un.d_val; break; case DT_VERDEF: obj->verdef = (const Elf_Verdef *) (obj->relocbase + dynp->d_un.d_val); break; case DT_VERDEFNUM: obj->verdefnum = dynp->d_un.d_val; break; case DT_VERSYM: obj->versyms = (const Elf_Versym *)(obj->relocbase + dynp->d_un.d_val); break; case DT_HASH: { hashtab = (const Elf_Hashelt *)(obj->relocbase + dynp->d_un.d_ptr); obj->nbuckets = hashtab[0]; obj->nchains = hashtab[1]; obj->buckets = hashtab + 2; obj->chains = obj->buckets + obj->nbuckets; obj->valid_hash_sysv = obj->nbuckets > 0 && obj->nchains > 0 && obj->buckets != NULL; } break; case DT_GNU_HASH: { hashtab = (const Elf_Hashelt *)(obj->relocbase + dynp->d_un.d_ptr); obj->nbuckets_gnu = hashtab[0]; obj->symndx_gnu = hashtab[1]; nmaskwords = hashtab[2]; bloom_size32 = (__ELF_WORD_SIZE / 32) * nmaskwords; obj->maskwords_bm_gnu = nmaskwords - 1; obj->shift2_gnu = hashtab[3]; obj->bloom_gnu = (Elf_Addr *) (hashtab + 4); obj->buckets_gnu = hashtab + 4 + bloom_size32; obj->chain_zero_gnu = obj->buckets_gnu + obj->nbuckets_gnu - obj->symndx_gnu; /* Number of bitmask words is required to be power of 2 */ obj->valid_hash_gnu = powerof2(nmaskwords) && obj->nbuckets_gnu > 0 && obj->buckets_gnu != NULL; } break; case DT_NEEDED: if (!obj->rtld) { Needed_Entry *nep = NEW(Needed_Entry); nep->name = dynp->d_un.d_val; nep->obj = NULL; nep->next = NULL; *needed_tail = nep; needed_tail = &nep->next; } break; case DT_FILTER: if (!obj->rtld) { Needed_Entry *nep = NEW(Needed_Entry); nep->name = dynp->d_un.d_val; nep->obj = NULL; nep->next = NULL; *needed_filtees_tail = nep; needed_filtees_tail = &nep->next; } break; case DT_AUXILIARY: if (!obj->rtld) { Needed_Entry *nep = NEW(Needed_Entry); nep->name = dynp->d_un.d_val; nep->obj = NULL; nep->next = NULL; *needed_aux_filtees_tail = nep; needed_aux_filtees_tail = &nep->next; } break; case DT_PLTGOT: obj->pltgot = (Elf_Addr *) (obj->relocbase + dynp->d_un.d_ptr); break; case DT_TEXTREL: obj->textrel = true; break; case DT_SYMBOLIC: obj->symbolic = true; break; case DT_RPATH: /* * We have to wait until later to process this, because we * might not have gotten the address of the string table yet. */ *dyn_rpath = dynp; break; case DT_SONAME: *dyn_soname = dynp; break; case DT_RUNPATH: *dyn_runpath = dynp; break; case DT_INIT: obj->init = (Elf_Addr) (obj->relocbase + dynp->d_un.d_ptr); break; case DT_PREINIT_ARRAY: obj->preinit_array = (Elf_Addr)(obj->relocbase + dynp->d_un.d_ptr); break; case DT_PREINIT_ARRAYSZ: obj->preinit_array_num = dynp->d_un.d_val / sizeof(Elf_Addr); break; case DT_INIT_ARRAY: obj->init_array = (Elf_Addr)(obj->relocbase + dynp->d_un.d_ptr); break; case DT_INIT_ARRAYSZ: obj->init_array_num = dynp->d_un.d_val / sizeof(Elf_Addr); break; case DT_FINI: obj->fini = (Elf_Addr) (obj->relocbase + dynp->d_un.d_ptr); break; case DT_FINI_ARRAY: obj->fini_array = (Elf_Addr)(obj->relocbase + dynp->d_un.d_ptr); break; case DT_FINI_ARRAYSZ: obj->fini_array_num = dynp->d_un.d_val / sizeof(Elf_Addr); break; /* * Don't process DT_DEBUG on MIPS as the dynamic section * is mapped read-only. DT_MIPS_RLD_MAP is used instead. */ #ifndef __mips__ case DT_DEBUG: if (!early) dbg("Filling in DT_DEBUG entry"); ((Elf_Dyn*)dynp)->d_un.d_ptr = (Elf_Addr) &r_debug; break; #endif case DT_FLAGS: if (dynp->d_un.d_val & DF_ORIGIN) obj->z_origin = true; if (dynp->d_un.d_val & DF_SYMBOLIC) obj->symbolic = true; if (dynp->d_un.d_val & DF_TEXTREL) obj->textrel = true; if (dynp->d_un.d_val & DF_BIND_NOW) obj->bind_now = true; /*if (dynp->d_un.d_val & DF_STATIC_TLS) ;*/ break; #ifdef __mips__ case DT_MIPS_LOCAL_GOTNO: obj->local_gotno = dynp->d_un.d_val; break; case DT_MIPS_SYMTABNO: obj->symtabno = dynp->d_un.d_val; break; case DT_MIPS_GOTSYM: obj->gotsym = dynp->d_un.d_val; break; case DT_MIPS_RLD_MAP: *((Elf_Addr *)(dynp->d_un.d_ptr)) = (Elf_Addr) &r_debug; break; #endif #ifdef __powerpc64__ case DT_PPC64_GLINK: obj->glink = (Elf_Addr) (obj->relocbase + dynp->d_un.d_ptr); break; #endif case DT_FLAGS_1: if (dynp->d_un.d_val & DF_1_NOOPEN) obj->z_noopen = true; if (dynp->d_un.d_val & DF_1_ORIGIN) obj->z_origin = true; if (dynp->d_un.d_val & DF_1_GLOBAL) obj->z_global = true; if (dynp->d_un.d_val & DF_1_BIND_NOW) obj->bind_now = true; if (dynp->d_un.d_val & DF_1_NODELETE) obj->z_nodelete = true; if (dynp->d_un.d_val & DF_1_LOADFLTR) obj->z_loadfltr = true; if (dynp->d_un.d_val & DF_1_INTERPOSE) obj->z_interpose = true; if (dynp->d_un.d_val & DF_1_NODEFLIB) obj->z_nodeflib = true; break; default: if (!early) { dbg("Ignoring d_tag %ld = %#lx", (long)dynp->d_tag, (long)dynp->d_tag); } break; } } obj->traced = false; if (plttype == DT_RELA) { obj->pltrela = (const Elf_Rela *) obj->pltrel; obj->pltrel = NULL; obj->pltrelasize = obj->pltrelsize; obj->pltrelsize = 0; } /* Determine size of dynsym table (equal to nchains of sysv hash) */ if (obj->valid_hash_sysv) obj->dynsymcount = obj->nchains; else if (obj->valid_hash_gnu) { obj->dynsymcount = 0; for (bkt = 0; bkt < obj->nbuckets_gnu; bkt++) { if (obj->buckets_gnu[bkt] == 0) continue; hashval = &obj->chain_zero_gnu[obj->buckets_gnu[bkt]]; do obj->dynsymcount++; while ((*hashval++ & 1u) == 0); } obj->dynsymcount += obj->symndx_gnu; } } static bool obj_resolve_origin(Obj_Entry *obj) { if (obj->origin_path != NULL) return (true); obj->origin_path = xmalloc(PATH_MAX); return (rtld_dirname_abs(obj->path, obj->origin_path) != -1); } static void digest_dynamic2(Obj_Entry *obj, const Elf_Dyn *dyn_rpath, const Elf_Dyn *dyn_soname, const Elf_Dyn *dyn_runpath) { if (obj->z_origin && !obj_resolve_origin(obj)) rtld_die(); if (dyn_runpath != NULL) { obj->runpath = (char *)obj->strtab + dyn_runpath->d_un.d_val; obj->runpath = origin_subst(obj, obj->runpath); } else if (dyn_rpath != NULL) { obj->rpath = (char *)obj->strtab + dyn_rpath->d_un.d_val; obj->rpath = origin_subst(obj, obj->rpath); } if (dyn_soname != NULL) object_add_name(obj, obj->strtab + dyn_soname->d_un.d_val); } static void digest_dynamic(Obj_Entry *obj, int early) { const Elf_Dyn *dyn_rpath; const Elf_Dyn *dyn_soname; const Elf_Dyn *dyn_runpath; digest_dynamic1(obj, early, &dyn_rpath, &dyn_soname, &dyn_runpath); digest_dynamic2(obj, dyn_rpath, dyn_soname, dyn_runpath); } /* * Process a shared object's program header. This is used only for the * main program, when the kernel has already loaded the main program * into memory before calling the dynamic linker. It creates and * returns an Obj_Entry structure. */ static Obj_Entry * digest_phdr(const Elf_Phdr *phdr, int phnum, caddr_t entry, const char *path) { Obj_Entry *obj; const Elf_Phdr *phlimit = phdr + phnum; const Elf_Phdr *ph; Elf_Addr note_start, note_end; int nsegs = 0; obj = obj_new(); for (ph = phdr; ph < phlimit; ph++) { if (ph->p_type != PT_PHDR) continue; obj->phdr = phdr; obj->phsize = ph->p_memsz; obj->relocbase = (caddr_t)phdr - ph->p_vaddr; break; } obj->stack_flags = PF_X | PF_R | PF_W; for (ph = phdr; ph < phlimit; ph++) { switch (ph->p_type) { case PT_INTERP: obj->interp = (const char *)(ph->p_vaddr + obj->relocbase); break; case PT_LOAD: if (nsegs == 0) { /* First load segment */ obj->vaddrbase = trunc_page(ph->p_vaddr); obj->mapbase = obj->vaddrbase + obj->relocbase; obj->textsize = round_page(ph->p_vaddr + ph->p_memsz) - obj->vaddrbase; } else { /* Last load segment */ obj->mapsize = round_page(ph->p_vaddr + ph->p_memsz) - obj->vaddrbase; } nsegs++; break; case PT_DYNAMIC: obj->dynamic = (const Elf_Dyn *)(ph->p_vaddr + obj->relocbase); break; case PT_TLS: obj->tlsindex = 1; obj->tlssize = ph->p_memsz; obj->tlsalign = ph->p_align; obj->tlsinitsize = ph->p_filesz; obj->tlsinit = (void*)(ph->p_vaddr + obj->relocbase); break; case PT_GNU_STACK: obj->stack_flags = ph->p_flags; break; case PT_GNU_RELRO: obj->relro_page = obj->relocbase + trunc_page(ph->p_vaddr); obj->relro_size = round_page(ph->p_memsz); break; case PT_NOTE: note_start = (Elf_Addr)obj->relocbase + ph->p_vaddr; note_end = note_start + ph->p_filesz; digest_notes(obj, note_start, note_end); break; } } if (nsegs < 1) { _rtld_error("%s: too few PT_LOAD segments", path); return NULL; } obj->entry = entry; return obj; } void digest_notes(Obj_Entry *obj, Elf_Addr note_start, Elf_Addr note_end) { const Elf_Note *note; const char *note_name; uintptr_t p; for (note = (const Elf_Note *)note_start; (Elf_Addr)note < note_end; note = (const Elf_Note *)((const char *)(note + 1) + roundup2(note->n_namesz, sizeof(Elf32_Addr)) + roundup2(note->n_descsz, sizeof(Elf32_Addr)))) { if (note->n_namesz != sizeof(NOTE_FREEBSD_VENDOR) || note->n_descsz != sizeof(int32_t)) continue; if (note->n_type != NT_FREEBSD_ABI_TAG && note->n_type != NT_FREEBSD_NOINIT_TAG) continue; note_name = (const char *)(note + 1); if (strncmp(NOTE_FREEBSD_VENDOR, note_name, sizeof(NOTE_FREEBSD_VENDOR)) != 0) continue; switch (note->n_type) { case NT_FREEBSD_ABI_TAG: /* FreeBSD osrel note */ p = (uintptr_t)(note + 1); p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); obj->osrel = *(const int32_t *)(p); dbg("note osrel %d", obj->osrel); break; case NT_FREEBSD_NOINIT_TAG: /* FreeBSD 'crt does not call init' note */ obj->crt_no_init = true; dbg("note crt_no_init"); break; } } } static Obj_Entry * dlcheck(void *handle) { Obj_Entry *obj; TAILQ_FOREACH(obj, &obj_list, next) { if (obj == (Obj_Entry *) handle) break; } if (obj == NULL || obj->refcount == 0 || obj->dl_refcount == 0) { _rtld_error("Invalid shared object handle %p", handle); return NULL; } return obj; } /* * If the given object is already in the donelist, return true. Otherwise * add the object to the list and return false. */ static bool donelist_check(DoneList *dlp, const Obj_Entry *obj) { unsigned int i; for (i = 0; i < dlp->num_used; i++) if (dlp->objs[i] == obj) return true; /* * Our donelist allocation should always be sufficient. But if * our threads locking isn't working properly, more shared objects * could have been loaded since we allocated the list. That should * never happen, but we'll handle it properly just in case it does. */ if (dlp->num_used < dlp->num_alloc) dlp->objs[dlp->num_used++] = obj; return false; } /* * Hash function for symbol table lookup. Don't even think about changing * this. It is specified by the System V ABI. */ 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; } /* * The GNU hash function is the Daniel J. Bernstein hash clipped to 32 bits * unsigned in case it's implemented with a wider type. */ static uint32_t gnu_hash(const char *s) { uint32_t h; unsigned char c; h = 5381; for (c = *s; c != '\0'; c = *++s) h = h * 33 + c; return (h & 0xffffffff); } /* * Find the library with the given name, and return its full pathname. * The returned string is dynamically allocated. Generates an error * message and returns NULL if the library cannot be found. * * If the second argument is non-NULL, then it refers to an already- * loaded shared object, whose library search path will be searched. * * If a library is successfully located via LD_LIBRARY_PATH_FDS, its * descriptor (which is close-on-exec) will be passed out via the third * argument. * * The search order is: * DT_RPATH in the referencing file _unless_ DT_RUNPATH is present (1) * DT_RPATH of the main object if DSO without defined DT_RUNPATH (1) * LD_LIBRARY_PATH * DT_RUNPATH in the referencing file * ldconfig hints (if -z nodefaultlib, filter out default library directories * from list) * /lib:/usr/lib _unless_ the referencing file is linked with -z nodefaultlib * * (1) Handled in digest_dynamic2 - rpath left NULL if runpath defined. */ static char * find_library(const char *xname, const Obj_Entry *refobj, int *fdp) { char *pathname; char *name; bool nodeflib, objgiven; objgiven = refobj != NULL; if (strchr(xname, '/') != NULL) { /* Hard coded pathname */ if (xname[0] != '/' && !trust) { _rtld_error("Absolute pathname required for shared object \"%s\"", xname); return NULL; } return (origin_subst(__DECONST(Obj_Entry *, refobj), __DECONST(char *, xname))); } if (libmap_disable || !objgiven || (name = lm_find(refobj->path, xname)) == NULL) name = (char *)xname; dbg(" Searching for \"%s\"", name); /* * If refobj->rpath != NULL, then refobj->runpath is NULL. Fall * back to pre-conforming behaviour if user requested so with * LD_LIBRARY_PATH_RPATH environment variable and ignore -z * nodeflib. */ if (objgiven && refobj->rpath != NULL && ld_library_path_rpath) { if ((pathname = search_library_path(name, ld_library_path)) != NULL || (refobj != NULL && (pathname = search_library_path(name, refobj->rpath)) != NULL) || (pathname = search_library_pathfds(name, ld_library_dirs, fdp)) != NULL || (pathname = search_library_path(name, gethints(false))) != NULL || (pathname = search_library_path(name, ld_standard_library_path)) != NULL) return (pathname); } else { nodeflib = objgiven ? refobj->z_nodeflib : false; if ((objgiven && (pathname = search_library_path(name, refobj->rpath)) != NULL) || (objgiven && refobj->runpath == NULL && refobj != obj_main && (pathname = search_library_path(name, obj_main->rpath)) != NULL) || (pathname = search_library_path(name, ld_library_path)) != NULL || (objgiven && (pathname = search_library_path(name, refobj->runpath)) != NULL) || (pathname = search_library_pathfds(name, ld_library_dirs, fdp)) != NULL || (pathname = search_library_path(name, gethints(nodeflib))) != NULL || (objgiven && !nodeflib && (pathname = search_library_path(name, ld_standard_library_path)) != NULL)) return (pathname); } if (objgiven && refobj->path != NULL) { _rtld_error("Shared object \"%s\" not found, required by \"%s\"", name, basename(refobj->path)); } else { _rtld_error("Shared object \"%s\" not found", name); } return NULL; } /* * Given a symbol number in a referencing object, find the corresponding * definition of the symbol. Returns a pointer to the symbol, or NULL if * no definition was found. Returns a pointer to the Obj_Entry of the * defining object via the reference parameter DEFOBJ_OUT. */ const Elf_Sym * find_symdef(unsigned long symnum, const Obj_Entry *refobj, const Obj_Entry **defobj_out, int flags, SymCache *cache, RtldLockState *lockstate) { const Elf_Sym *ref; const Elf_Sym *def; const Obj_Entry *defobj; SymLook req; const char *name; int res; /* * If we have already found this symbol, get the information from * the cache. */ if (symnum >= refobj->dynsymcount) return NULL; /* Bad object */ if (cache != NULL && cache[symnum].sym != NULL) { *defobj_out = cache[symnum].obj; return cache[symnum].sym; } ref = refobj->symtab + symnum; name = refobj->strtab + ref->st_name; def = NULL; defobj = NULL; /* * We don't have to do a full scale lookup if the symbol is local. * We know it will bind to the instance in this load module; to * which we already have a pointer (ie ref). By not doing a lookup, * we not only improve performance, but it also avoids unresolvable * symbols when local symbols are not in the hash table. This has * been seen with the ia64 toolchain. */ if (ELF_ST_BIND(ref->st_info) != STB_LOCAL) { if (ELF_ST_TYPE(ref->st_info) == STT_SECTION) { _rtld_error("%s: Bogus symbol table entry %lu", refobj->path, symnum); } symlook_init(&req, name); req.flags = flags; req.ventry = fetch_ventry(refobj, symnum); req.lockstate = lockstate; res = symlook_default(&req, refobj); if (res == 0) { def = req.sym_out; defobj = req.defobj_out; } } else { def = ref; defobj = refobj; } /* * If we found no definition and the reference is weak, treat the * symbol as having the value zero. */ if (def == NULL && ELF_ST_BIND(ref->st_info) == STB_WEAK) { def = &sym_zero; defobj = obj_main; } if (def != NULL) { *defobj_out = defobj; /* Record the information in the cache to avoid subsequent lookups. */ if (cache != NULL) { cache[symnum].sym = def; cache[symnum].obj = defobj; } } else { if (refobj != &obj_rtld) _rtld_error("%s: Undefined symbol \"%s\"", refobj->path, name); } return def; } /* * Return the search path from the ldconfig hints file, reading it if * necessary. If nostdlib is true, then the default search paths are * not added to result. * * Returns NULL if there are problems with the hints file, * or if the search path there is empty. */ static const char * gethints(bool nostdlib) { static char *hints, *filtered_path; struct elfhints_hdr hdr; struct fill_search_info_args sargs, hargs; struct dl_serinfo smeta, hmeta, *SLPinfo, *hintinfo; struct dl_serpath *SLPpath, *hintpath; char *p; unsigned int SLPndx, hintndx, fndx, fcount; int fd; size_t flen; bool skip; /* First call, read the hints file */ if (hints == NULL) { /* Keep from trying again in case the hints file is bad. */ hints = ""; if ((fd = open(ld_elf_hints_path, O_RDONLY | O_CLOEXEC)) == -1) return (NULL); if (read(fd, &hdr, sizeof hdr) != sizeof hdr || hdr.magic != ELFHINTS_MAGIC || hdr.version != 1) { close(fd); return (NULL); } p = xmalloc(hdr.dirlistlen + 1); if (lseek(fd, hdr.strtab + hdr.dirlist, SEEK_SET) == -1 || read(fd, p, hdr.dirlistlen + 1) != (ssize_t)hdr.dirlistlen + 1) { free(p); close(fd); return (NULL); } hints = p; close(fd); } /* * If caller agreed to receive list which includes the default * paths, we are done. Otherwise, if we still did not * calculated filtered result, do it now. */ if (!nostdlib) return (hints[0] != '\0' ? hints : NULL); if (filtered_path != NULL) goto filt_ret; /* * Obtain the list of all configured search paths, and the * list of the default paths. * * First estimate the size of the results. */ smeta.dls_size = __offsetof(struct dl_serinfo, dls_serpath); smeta.dls_cnt = 0; hmeta.dls_size = __offsetof(struct dl_serinfo, dls_serpath); hmeta.dls_cnt = 0; sargs.request = RTLD_DI_SERINFOSIZE; sargs.serinfo = &smeta; hargs.request = RTLD_DI_SERINFOSIZE; hargs.serinfo = &hmeta; path_enumerate(ld_standard_library_path, fill_search_info, &sargs); path_enumerate(p, fill_search_info, &hargs); SLPinfo = xmalloc(smeta.dls_size); hintinfo = xmalloc(hmeta.dls_size); /* * Next fetch both sets of paths. */ sargs.request = RTLD_DI_SERINFO; sargs.serinfo = SLPinfo; sargs.serpath = &SLPinfo->dls_serpath[0]; sargs.strspace = (char *)&SLPinfo->dls_serpath[smeta.dls_cnt]; hargs.request = RTLD_DI_SERINFO; hargs.serinfo = hintinfo; hargs.serpath = &hintinfo->dls_serpath[0]; hargs.strspace = (char *)&hintinfo->dls_serpath[hmeta.dls_cnt]; path_enumerate(ld_standard_library_path, fill_search_info, &sargs); path_enumerate(p, fill_search_info, &hargs); /* * Now calculate the difference between two sets, by excluding * standard paths from the full set. */ fndx = 0; fcount = 0; filtered_path = xmalloc(hdr.dirlistlen + 1); hintpath = &hintinfo->dls_serpath[0]; for (hintndx = 0; hintndx < hmeta.dls_cnt; hintndx++, hintpath++) { skip = false; SLPpath = &SLPinfo->dls_serpath[0]; /* * Check each standard path against current. */ for (SLPndx = 0; SLPndx < smeta.dls_cnt; SLPndx++, SLPpath++) { /* matched, skip the path */ if (!strcmp(hintpath->dls_name, SLPpath->dls_name)) { skip = true; break; } } if (skip) continue; /* * Not matched against any standard path, add the path * to result. Separate consequtive paths with ':'. */ if (fcount > 0) { filtered_path[fndx] = ':'; fndx++; } fcount++; flen = strlen(hintpath->dls_name); strncpy((filtered_path + fndx), hintpath->dls_name, flen); fndx += flen; } filtered_path[fndx] = '\0'; free(SLPinfo); free(hintinfo); filt_ret: return (filtered_path[0] != '\0' ? filtered_path : NULL); } static void init_dag(Obj_Entry *root) { const Needed_Entry *needed; const Objlist_Entry *elm; DoneList donelist; if (root->dag_inited) return; donelist_init(&donelist); /* Root object belongs to own DAG. */ objlist_push_tail(&root->dldags, root); objlist_push_tail(&root->dagmembers, root); donelist_check(&donelist, root); /* * Add dependencies of root object to DAG in breadth order * by exploiting the fact that each new object get added * to the tail of the dagmembers list. */ STAILQ_FOREACH(elm, &root->dagmembers, link) { for (needed = elm->obj->needed; needed != NULL; needed = needed->next) { if (needed->obj == NULL || donelist_check(&donelist, needed->obj)) continue; objlist_push_tail(&needed->obj->dldags, root); objlist_push_tail(&root->dagmembers, needed->obj); } } root->dag_inited = true; } Obj_Entry * globallist_curr(const Obj_Entry *obj) { for (;;) { if (obj == NULL) return (NULL); if (!obj->marker) return (__DECONST(Obj_Entry *, obj)); obj = TAILQ_PREV(obj, obj_entry_q, next); } } Obj_Entry * globallist_next(const Obj_Entry *obj) { for (;;) { obj = TAILQ_NEXT(obj, next); if (obj == NULL) return (NULL); if (!obj->marker) return (__DECONST(Obj_Entry *, obj)); } } static void process_z(Obj_Entry *root) { const Objlist_Entry *elm; Obj_Entry *obj; /* * Walk over object DAG and process every dependent object * that is marked as DF_1_NODELETE or DF_1_GLOBAL. They need * to grow their own DAG. * * For DF_1_GLOBAL, DAG is required for symbol lookups in * symlook_global() to work. * * For DF_1_NODELETE, the DAG should have its reference upped. */ STAILQ_FOREACH(elm, &root->dagmembers, link) { obj = elm->obj; if (obj == NULL) continue; if (obj->z_nodelete && !obj->ref_nodel) { dbg("obj %s -z nodelete", obj->path); init_dag(obj); ref_dag(obj); obj->ref_nodel = true; } if (obj->z_global && objlist_find(&list_global, obj) == NULL) { dbg("obj %s -z global", obj->path); objlist_push_tail(&list_global, obj); init_dag(obj); } } } /* * Initialize the dynamic linker. The argument is the address at which * the dynamic linker has been mapped into memory. The primary task of * this function is to relocate the dynamic linker. */ static void init_rtld(caddr_t mapbase, Elf_Auxinfo **aux_info) { Obj_Entry objtmp; /* Temporary rtld object */ const Elf_Dyn *dyn_rpath; const Elf_Dyn *dyn_soname; const Elf_Dyn *dyn_runpath; #ifdef RTLD_INIT_PAGESIZES_EARLY /* The page size is required by the dynamic memory allocator. */ init_pagesizes(aux_info); #endif /* * Conjure up an Obj_Entry structure for the dynamic linker. * * The "path" member can't be initialized yet because string constants * cannot yet be accessed. Below we will set it correctly. */ memset(&objtmp, 0, sizeof(objtmp)); objtmp.path = NULL; objtmp.rtld = true; objtmp.mapbase = mapbase; #ifdef PIC objtmp.relocbase = mapbase; #endif if (RTLD_IS_DYNAMIC()) { objtmp.dynamic = rtld_dynamic(&objtmp); digest_dynamic1(&objtmp, 1, &dyn_rpath, &dyn_soname, &dyn_runpath); assert(objtmp.needed == NULL); #if !defined(__mips__) /* MIPS has a bogus DT_TEXTREL. */ assert(!objtmp.textrel); #endif /* * Temporarily put the dynamic linker entry into the object list, so * that symbols can be found. */ relocate_objects(&objtmp, true, &objtmp, 0, NULL); } /* Initialize the object list. */ TAILQ_INIT(&obj_list); /* Now that non-local variables can be accesses, copy out obj_rtld. */ memcpy(&obj_rtld, &objtmp, sizeof(obj_rtld)); #ifndef RTLD_INIT_PAGESIZES_EARLY /* The page size is required by the dynamic memory allocator. */ init_pagesizes(aux_info); #endif if (aux_info[AT_OSRELDATE] != NULL) osreldate = aux_info[AT_OSRELDATE]->a_un.a_val; digest_dynamic2(&obj_rtld, dyn_rpath, dyn_soname, dyn_runpath); /* Replace the path with a dynamically allocated copy. */ obj_rtld.path = xstrdup(ld_path_rtld); r_debug.r_brk = r_debug_state; r_debug.r_state = RT_CONSISTENT; } /* * Retrieve the array of supported page sizes. The kernel provides the page * sizes in increasing order. */ static void init_pagesizes(Elf_Auxinfo **aux_info) { static size_t psa[MAXPAGESIZES]; int mib[2]; size_t len, size; if (aux_info[AT_PAGESIZES] != NULL && aux_info[AT_PAGESIZESLEN] != NULL) { size = aux_info[AT_PAGESIZESLEN]->a_un.a_val; pagesizes = aux_info[AT_PAGESIZES]->a_un.a_ptr; } else { len = 2; if (sysctlnametomib("hw.pagesizes", mib, &len) == 0) size = sizeof(psa); else { /* As a fallback, retrieve the base page size. */ size = sizeof(psa[0]); if (aux_info[AT_PAGESZ] != NULL) { psa[0] = aux_info[AT_PAGESZ]->a_un.a_val; goto psa_filled; } else { mib[0] = CTL_HW; mib[1] = HW_PAGESIZE; len = 2; } } if (sysctl(mib, len, psa, &size, NULL, 0) == -1) { _rtld_error("sysctl for hw.pagesize(s) failed"); rtld_die(); } psa_filled: pagesizes = psa; } npagesizes = size / sizeof(pagesizes[0]); /* Discard any invalid entries at the end of the array. */ while (npagesizes > 0 && pagesizes[npagesizes - 1] == 0) npagesizes--; } /* * Add the init functions from a needed object list (and its recursive * needed objects) to "list". This is not used directly; it is a helper * function for initlist_add_objects(). The write lock must be held * when this function is called. */ static void initlist_add_neededs(Needed_Entry *needed, Objlist *list) { /* Recursively process the successor needed objects. */ if (needed->next != NULL) initlist_add_neededs(needed->next, list); /* Process the current needed object. */ if (needed->obj != NULL) initlist_add_objects(needed->obj, needed->obj, list); } /* * Scan all of the DAGs rooted in the range of objects from "obj" to * "tail" and add their init functions to "list". This recurses over * the DAGs and ensure the proper init ordering such that each object's * needed libraries are initialized before the object itself. At the * same time, this function adds the objects to the global finalization * list "list_fini" in the opposite order. The write lock must be * held when this function is called. */ static void initlist_add_objects(Obj_Entry *obj, Obj_Entry *tail, Objlist *list) { Obj_Entry *nobj; if (obj->init_scanned || obj->init_done) return; obj->init_scanned = true; /* Recursively process the successor objects. */ nobj = globallist_next(obj); if (nobj != NULL && obj != tail) initlist_add_objects(nobj, tail, list); /* Recursively process the needed objects. */ if (obj->needed != NULL) initlist_add_neededs(obj->needed, list); if (obj->needed_filtees != NULL) initlist_add_neededs(obj->needed_filtees, list); if (obj->needed_aux_filtees != NULL) initlist_add_neededs(obj->needed_aux_filtees, list); /* Add the object to the init list. */ if (obj->preinit_array != (Elf_Addr)NULL || obj->init != (Elf_Addr)NULL || obj->init_array != (Elf_Addr)NULL) objlist_push_tail(list, obj); /* Add the object to the global fini list in the reverse order. */ if ((obj->fini != (Elf_Addr)NULL || obj->fini_array != (Elf_Addr)NULL) && !obj->on_fini_list) { objlist_push_head(&list_fini, obj); obj->on_fini_list = true; } } #ifndef FPTR_TARGET #define FPTR_TARGET(f) ((Elf_Addr) (f)) #endif static void free_needed_filtees(Needed_Entry *n) { Needed_Entry *needed, *needed1; for (needed = n; needed != NULL; needed = needed->next) { if (needed->obj != NULL) { dlclose(needed->obj); needed->obj = NULL; } } for (needed = n; needed != NULL; needed = needed1) { needed1 = needed->next; free(needed); } } static void unload_filtees(Obj_Entry *obj) { free_needed_filtees(obj->needed_filtees); obj->needed_filtees = NULL; free_needed_filtees(obj->needed_aux_filtees); obj->needed_aux_filtees = NULL; obj->filtees_loaded = false; } static void load_filtee1(Obj_Entry *obj, Needed_Entry *needed, int flags, RtldLockState *lockstate) { for (; needed != NULL; needed = needed->next) { needed->obj = dlopen_object(obj->strtab + needed->name, -1, obj, flags, ((ld_loadfltr || obj->z_loadfltr) ? RTLD_NOW : RTLD_LAZY) | RTLD_LOCAL, lockstate); } } static void load_filtees(Obj_Entry *obj, int flags, RtldLockState *lockstate) { lock_restart_for_upgrade(lockstate); if (!obj->filtees_loaded) { load_filtee1(obj, obj->needed_filtees, flags, lockstate); load_filtee1(obj, obj->needed_aux_filtees, flags, lockstate); obj->filtees_loaded = true; } } static int process_needed(Obj_Entry *obj, Needed_Entry *needed, int flags) { Obj_Entry *obj1; for (; needed != NULL; needed = needed->next) { obj1 = needed->obj = load_object(obj->strtab + needed->name, -1, obj, flags & ~RTLD_LO_NOLOAD); if (obj1 == NULL && !ld_tracing && (flags & RTLD_LO_FILTEES) == 0) return (-1); } return (0); } /* * Given a shared object, traverse its list of needed objects, and load * each of them. Returns 0 on success. Generates an error message and * returns -1 on failure. */ static int load_needed_objects(Obj_Entry *first, int flags) { Obj_Entry *obj; obj = first; TAILQ_FOREACH_FROM(obj, &obj_list, next) { if (obj->marker) continue; if (process_needed(obj, obj->needed, flags) == -1) return (-1); } return (0); } static int load_preload_objects(void) { char *p = ld_preload; Obj_Entry *obj; static const char delim[] = " \t:;"; if (p == NULL) return 0; p += strspn(p, delim); while (*p != '\0') { size_t len = strcspn(p, delim); char savech; savech = p[len]; p[len] = '\0'; obj = load_object(p, -1, NULL, 0); if (obj == NULL) return -1; /* XXX - cleanup */ obj->z_interpose = true; p[len] = savech; p += len; p += strspn(p, delim); } LD_UTRACE(UTRACE_PRELOAD_FINISHED, NULL, NULL, 0, 0, NULL); return 0; } static const char * printable_path(const char *path) { return (path == NULL ? "" : path); } /* * Load a shared object into memory, if it is not already loaded. The * object may be specified by name or by user-supplied file descriptor * fd_u. In the later case, the fd_u descriptor is not closed, but its * duplicate is. * * Returns a pointer to the Obj_Entry for the object. Returns NULL * on failure. */ static Obj_Entry * load_object(const char *name, int fd_u, const Obj_Entry *refobj, int flags) { Obj_Entry *obj; int fd; struct stat sb; char *path; fd = -1; if (name != NULL) { TAILQ_FOREACH(obj, &obj_list, next) { if (obj->marker) continue; if (object_match_name(obj, name)) return (obj); } path = find_library(name, refobj, &fd); if (path == NULL) return (NULL); } else path = NULL; if (fd >= 0) { /* * search_library_pathfds() opens a fresh file descriptor for the * library, so there is no need to dup(). */ } else if (fd_u == -1) { /* * If we didn't find a match by pathname, or the name is not * supplied, open the file and check again by device and inode. * This avoids false mismatches caused by multiple links or ".." * in pathnames. * * To avoid a race, we open the file and use fstat() rather than * using stat(). */ if ((fd = open(path, O_RDONLY | O_CLOEXEC | O_VERIFY)) == -1) { _rtld_error("Cannot open \"%s\"", path); free(path); return (NULL); } } else { fd = fcntl(fd_u, F_DUPFD_CLOEXEC, 0); if (fd == -1) { _rtld_error("Cannot dup fd"); free(path); return (NULL); } } if (fstat(fd, &sb) == -1) { _rtld_error("Cannot fstat \"%s\"", printable_path(path)); close(fd); free(path); return NULL; } TAILQ_FOREACH(obj, &obj_list, next) { if (obj->marker) continue; if (obj->ino == sb.st_ino && obj->dev == sb.st_dev) break; } if (obj != NULL && name != NULL) { object_add_name(obj, name); free(path); close(fd); return obj; } if (flags & RTLD_LO_NOLOAD) { free(path); close(fd); return (NULL); } /* First use of this object, so we must map it in */ obj = do_load_object(fd, name, path, &sb, flags); if (obj == NULL) free(path); close(fd); return obj; } static Obj_Entry * do_load_object(int fd, const char *name, char *path, struct stat *sbp, int flags) { Obj_Entry *obj; struct statfs fs; /* * but first, make sure that environment variables haven't been * used to circumvent the noexec flag on a filesystem. */ if (dangerous_ld_env) { if (fstatfs(fd, &fs) != 0) { _rtld_error("Cannot fstatfs \"%s\"", printable_path(path)); return NULL; } if (fs.f_flags & MNT_NOEXEC) { _rtld_error("Cannot execute objects on %s\n", fs.f_mntonname); return NULL; } } dbg("loading \"%s\"", printable_path(path)); obj = map_object(fd, printable_path(path), sbp); if (obj == NULL) return NULL; /* * If DT_SONAME is present in the object, digest_dynamic2 already * added it to the object names. */ if (name != NULL) object_add_name(obj, name); obj->path = path; digest_dynamic(obj, 0); dbg("%s valid_hash_sysv %d valid_hash_gnu %d dynsymcount %d", obj->path, obj->valid_hash_sysv, obj->valid_hash_gnu, obj->dynsymcount); if (obj->z_noopen && (flags & (RTLD_LO_DLOPEN | RTLD_LO_TRACE)) == RTLD_LO_DLOPEN) { dbg("refusing to load non-loadable \"%s\"", obj->path); _rtld_error("Cannot dlopen non-loadable %s", obj->path); munmap(obj->mapbase, obj->mapsize); obj_free(obj); return (NULL); } obj->dlopened = (flags & RTLD_LO_DLOPEN) != 0; TAILQ_INSERT_TAIL(&obj_list, obj, next); obj_count++; obj_loads++; linkmap_add(obj); /* for GDB & dlinfo() */ max_stack_flags |= obj->stack_flags; dbg(" %p .. %p: %s", obj->mapbase, obj->mapbase + obj->mapsize - 1, obj->path); if (obj->textrel) dbg(" WARNING: %s has impure text", obj->path); LD_UTRACE(UTRACE_LOAD_OBJECT, obj, obj->mapbase, obj->mapsize, 0, obj->path); return obj; } static Obj_Entry * obj_from_addr(const void *addr) { Obj_Entry *obj; TAILQ_FOREACH(obj, &obj_list, next) { if (obj->marker) continue; if (addr < (void *) obj->mapbase) continue; if (addr < (void *) (obj->mapbase + obj->mapsize)) return obj; } return NULL; } static void preinit_main(void) { Elf_Addr *preinit_addr; int index; preinit_addr = (Elf_Addr *)obj_main->preinit_array; if (preinit_addr == NULL) return; for (index = 0; index < obj_main->preinit_array_num; index++) { if (preinit_addr[index] != 0 && preinit_addr[index] != 1) { dbg("calling preinit function for %s at %p", obj_main->path, (void *)preinit_addr[index]); LD_UTRACE(UTRACE_INIT_CALL, obj_main, (void *)preinit_addr[index], 0, 0, obj_main->path); call_init_pointer(obj_main, preinit_addr[index]); } } } /* * Call the finalization functions for each of the objects in "list" * belonging to the DAG of "root" and referenced once. If NULL "root" * is specified, every finalization function will be called regardless * of the reference count and the list elements won't be freed. All of * the objects are expected to have non-NULL fini functions. */ static void objlist_call_fini(Objlist *list, Obj_Entry *root, RtldLockState *lockstate) { Objlist_Entry *elm; char *saved_msg; Elf_Addr *fini_addr; int index; assert(root == NULL || root->refcount == 1); /* * Preserve the current error message since a fini function might * call into the dynamic linker and overwrite it. */ saved_msg = errmsg_save(); do { STAILQ_FOREACH(elm, list, link) { if (root != NULL && (elm->obj->refcount != 1 || objlist_find(&root->dagmembers, elm->obj) == NULL)) continue; /* Remove object from fini list to prevent recursive invocation. */ STAILQ_REMOVE(list, elm, Struct_Objlist_Entry, link); /* * XXX: If a dlopen() call references an object while the * fini function is in progress, we might end up trying to * unload the referenced object in dlclose() or the object * won't be unloaded although its fini function has been * called. */ lock_release(rtld_bind_lock, lockstate); /* * It is legal to have both DT_FINI and DT_FINI_ARRAY defined. * When this happens, DT_FINI_ARRAY is processed first. */ fini_addr = (Elf_Addr *)elm->obj->fini_array; if (fini_addr != NULL && elm->obj->fini_array_num > 0) { for (index = elm->obj->fini_array_num - 1; index >= 0; index--) { if (fini_addr[index] != 0 && fini_addr[index] != 1) { dbg("calling fini function for %s at %p", elm->obj->path, (void *)fini_addr[index]); LD_UTRACE(UTRACE_FINI_CALL, elm->obj, (void *)fini_addr[index], 0, 0, elm->obj->path); call_initfini_pointer(elm->obj, fini_addr[index]); } } } if (elm->obj->fini != (Elf_Addr)NULL) { dbg("calling fini function for %s at %p", elm->obj->path, (void *)elm->obj->fini); LD_UTRACE(UTRACE_FINI_CALL, elm->obj, (void *)elm->obj->fini, 0, 0, elm->obj->path); call_initfini_pointer(elm->obj, elm->obj->fini); } wlock_acquire(rtld_bind_lock, lockstate); /* No need to free anything if process is going down. */ if (root != NULL) free(elm); /* * We must restart the list traversal after every fini call * because a dlclose() call from the fini function or from * another thread might have modified the reference counts. */ break; } } while (elm != NULL); errmsg_restore(saved_msg); } /* * Call the initialization functions for each of the objects in * "list". All of the objects are expected to have non-NULL init * functions. */ static void objlist_call_init(Objlist *list, RtldLockState *lockstate) { Objlist_Entry *elm; Obj_Entry *obj; char *saved_msg; Elf_Addr *init_addr; int index; /* * Clean init_scanned flag so that objects can be rechecked and * possibly initialized earlier if any of vectors called below * cause the change by using dlopen. */ TAILQ_FOREACH(obj, &obj_list, next) { if (obj->marker) continue; obj->init_scanned = false; } /* * Preserve the current error message since an init function might * call into the dynamic linker and overwrite it. */ saved_msg = errmsg_save(); STAILQ_FOREACH(elm, list, link) { if (elm->obj->init_done) /* Initialized early. */ continue; /* * Race: other thread might try to use this object before current * one completes the initilization. Not much can be done here * without better locking. */ elm->obj->init_done = true; lock_release(rtld_bind_lock, lockstate); /* * It is legal to have both DT_INIT and DT_INIT_ARRAY defined. * When this happens, DT_INIT is processed first. */ if (elm->obj->init != (Elf_Addr)NULL) { dbg("calling init function for %s at %p", elm->obj->path, (void *)elm->obj->init); LD_UTRACE(UTRACE_INIT_CALL, elm->obj, (void *)elm->obj->init, 0, 0, elm->obj->path); call_initfini_pointer(elm->obj, elm->obj->init); } init_addr = (Elf_Addr *)elm->obj->init_array; if (init_addr != NULL) { for (index = 0; index < elm->obj->init_array_num; index++) { if (init_addr[index] != 0 && init_addr[index] != 1) { dbg("calling init function for %s at %p", elm->obj->path, (void *)init_addr[index]); LD_UTRACE(UTRACE_INIT_CALL, elm->obj, (void *)init_addr[index], 0, 0, elm->obj->path); call_init_pointer(elm->obj, init_addr[index]); } } } wlock_acquire(rtld_bind_lock, lockstate); } errmsg_restore(saved_msg); } static void objlist_clear(Objlist *list) { Objlist_Entry *elm; while (!STAILQ_EMPTY(list)) { elm = STAILQ_FIRST(list); STAILQ_REMOVE_HEAD(list, link); free(elm); } } static Objlist_Entry * objlist_find(Objlist *list, const Obj_Entry *obj) { Objlist_Entry *elm; STAILQ_FOREACH(elm, list, link) if (elm->obj == obj) return elm; return NULL; } static void objlist_init(Objlist *list) { STAILQ_INIT(list); } static void objlist_push_head(Objlist *list, Obj_Entry *obj) { Objlist_Entry *elm; elm = NEW(Objlist_Entry); elm->obj = obj; STAILQ_INSERT_HEAD(list, elm, link); } static void objlist_push_tail(Objlist *list, Obj_Entry *obj) { Objlist_Entry *elm; elm = NEW(Objlist_Entry); elm->obj = obj; STAILQ_INSERT_TAIL(list, elm, link); } static void objlist_put_after(Objlist *list, Obj_Entry *listobj, Obj_Entry *obj) { Objlist_Entry *elm, *listelm; STAILQ_FOREACH(listelm, list, link) { if (listelm->obj == listobj) break; } elm = NEW(Objlist_Entry); elm->obj = obj; if (listelm != NULL) STAILQ_INSERT_AFTER(list, listelm, elm, link); else STAILQ_INSERT_TAIL(list, elm, link); } static void objlist_remove(Objlist *list, Obj_Entry *obj) { Objlist_Entry *elm; if ((elm = objlist_find(list, obj)) != NULL) { STAILQ_REMOVE(list, elm, Struct_Objlist_Entry, link); free(elm); } } /* * Relocate dag rooted in the specified object. * Returns 0 on success, or -1 on failure. */ static int relocate_object_dag(Obj_Entry *root, bool bind_now, Obj_Entry *rtldobj, int flags, RtldLockState *lockstate) { Objlist_Entry *elm; int error; error = 0; STAILQ_FOREACH(elm, &root->dagmembers, link) { error = relocate_object(elm->obj, bind_now, rtldobj, flags, lockstate); if (error == -1) break; } return (error); } /* + * Prepare for, or clean after, relocating an object marked with + * DT_TEXTREL or DF_TEXTREL. Before relocating, all read-only + * segments are remapped read-write. After relocations are done, the + * segment's permissions are returned back to the modes specified in + * the phdrs. If any relocation happened, or always for wired + * program, COW is triggered. + */ +static int +reloc_textrel_prot(Obj_Entry *obj, bool before) +{ + const Elf_Phdr *ph; + void *base; + size_t l, sz; + int prot; + + for (l = obj->phsize / sizeof(*ph), ph = obj->phdr; l > 0; + l--, ph++) { + if (ph->p_type != PT_LOAD || (ph->p_flags & PF_W) != 0) + continue; + base = obj->relocbase + trunc_page(ph->p_vaddr); + sz = round_page(ph->p_vaddr + ph->p_filesz) - + trunc_page(ph->p_vaddr); + prot = convert_prot(ph->p_flags) | (before ? PROT_WRITE : 0); + if (mprotect(base, sz, prot) == -1) { + _rtld_error("%s: Cannot write-%sable text segment: %s", + obj->path, before ? "en" : "dis", + rtld_strerror(errno)); + return (-1); + } + } + return (0); +} + +/* * Relocate single object. * Returns 0 on success, or -1 on failure. */ static int relocate_object(Obj_Entry *obj, bool bind_now, Obj_Entry *rtldobj, int flags, RtldLockState *lockstate) { if (obj->relocated) return (0); obj->relocated = true; if (obj != rtldobj) dbg("relocating \"%s\"", obj->path); if (obj->symtab == NULL || obj->strtab == NULL || !(obj->valid_hash_sysv || obj->valid_hash_gnu)) { _rtld_error("%s: Shared object has no run-time symbol table", obj->path); return (-1); } - if (obj->textrel) { - /* There are relocations to the write-protected text segment. */ - if (mprotect(obj->mapbase, obj->textsize, - PROT_READ|PROT_WRITE|PROT_EXEC) == -1) { - _rtld_error("%s: Cannot write-enable text segment: %s", - obj->path, rtld_strerror(errno)); - return (-1); - } - } + /* There are relocations to the write-protected text segment. */ + if (obj->textrel && reloc_textrel_prot(obj, true) != 0) + return (-1); /* Process the non-PLT non-IFUNC relocations. */ if (reloc_non_plt(obj, rtldobj, flags, lockstate)) return (-1); - if (obj->textrel) { /* Re-protected the text segment. */ - if (mprotect(obj->mapbase, obj->textsize, - PROT_READ|PROT_EXEC) == -1) { - _rtld_error("%s: Cannot write-protect text segment: %s", - obj->path, rtld_strerror(errno)); - return (-1); - } - } + /* Re-protected the text segment. */ + if (obj->textrel && reloc_textrel_prot(obj, false) != 0) + return (-1); /* Set the special PLT or GOT entries. */ init_pltgot(obj); /* Process the PLT relocations. */ if (reloc_plt(obj) == -1) return (-1); /* Relocate the jump slots if we are doing immediate binding. */ if (obj->bind_now || bind_now) if (reloc_jmpslots(obj, flags, lockstate) == -1) return (-1); /* * Process the non-PLT IFUNC relocations. The relocations are * processed in two phases, because IFUNC resolvers may * reference other symbols, which must be readily processed * before resolvers are called. */ if (obj->non_plt_gnu_ifunc && reloc_non_plt(obj, rtldobj, flags | SYMLOOK_IFUNC, lockstate)) return (-1); if (obj->relro_size > 0) { if (mprotect(obj->relro_page, obj->relro_size, PROT_READ) == -1) { _rtld_error("%s: Cannot enforce relro protection: %s", obj->path, rtld_strerror(errno)); return (-1); } } /* * Set up the magic number and version in the Obj_Entry. These * were checked in the crt1.o from the original ElfKit, so we * set them for backward compatibility. */ obj->magic = RTLD_MAGIC; obj->version = RTLD_VERSION; return (0); } /* * Relocate newly-loaded shared objects. The argument is a pointer to * the Obj_Entry for the first such object. All objects from the first * to the end of the list of objects are relocated. Returns 0 on success, * or -1 on failure. */ static int relocate_objects(Obj_Entry *first, bool bind_now, Obj_Entry *rtldobj, int flags, RtldLockState *lockstate) { Obj_Entry *obj; int error; error = 0; obj = first; TAILQ_FOREACH_FROM(obj, &obj_list, next) { if (obj->marker) continue; error = relocate_object(obj, bind_now, rtldobj, flags, lockstate); if (error == -1) break; } return (error); } /* * The handling of R_MACHINE_IRELATIVE relocations and jumpslots * referencing STT_GNU_IFUNC symbols is postponed till the other * relocations are done. The indirect functions specified as * ifunc are allowed to call other symbols, so we need to have * objects relocated before asking for resolution from indirects. * * The R_MACHINE_IRELATIVE slots are resolved in greedy fashion, * instead of the usual lazy handling of PLT slots. It is * consistent with how GNU does it. */ static int resolve_object_ifunc(Obj_Entry *obj, bool bind_now, int flags, RtldLockState *lockstate) { if (obj->irelative && reloc_iresolve(obj, lockstate) == -1) return (-1); if ((obj->bind_now || bind_now) && obj->gnu_ifunc && reloc_gnu_ifunc(obj, flags, lockstate) == -1) return (-1); return (0); } static int resolve_objects_ifunc(Obj_Entry *first, bool bind_now, int flags, RtldLockState *lockstate) { Obj_Entry *obj; obj = first; TAILQ_FOREACH_FROM(obj, &obj_list, next) { if (obj->marker) continue; if (resolve_object_ifunc(obj, bind_now, flags, lockstate) == -1) return (-1); } return (0); } static int initlist_objects_ifunc(Objlist *list, bool bind_now, int flags, RtldLockState *lockstate) { Objlist_Entry *elm; STAILQ_FOREACH(elm, list, link) { if (resolve_object_ifunc(elm->obj, bind_now, flags, lockstate) == -1) return (-1); } return (0); } /* * Cleanup procedure. It will be called (by the atexit mechanism) just * before the process exits. */ static void rtld_exit(void) { RtldLockState lockstate; wlock_acquire(rtld_bind_lock, &lockstate); dbg("rtld_exit()"); objlist_call_fini(&list_fini, NULL, &lockstate); /* No need to remove the items from the list, since we are exiting. */ if (!libmap_disable) lm_fini(); lock_release(rtld_bind_lock, &lockstate); } /* * Iterate over a search path, translate each element, and invoke the * callback on the result. */ static void * path_enumerate(const char *path, path_enum_proc callback, void *arg) { const char *trans; if (path == NULL) return (NULL); path += strspn(path, ":;"); while (*path != '\0') { size_t len; char *res; len = strcspn(path, ":;"); trans = lm_findn(NULL, path, len); if (trans) res = callback(trans, strlen(trans), arg); else res = callback(path, len, arg); if (res != NULL) return (res); path += len; path += strspn(path, ":;"); } return (NULL); } struct try_library_args { const char *name; size_t namelen; char *buffer; size_t buflen; }; static void * try_library_path(const char *dir, size_t dirlen, void *param) { struct try_library_args *arg; arg = param; if (*dir == '/' || trust) { char *pathname; if (dirlen + 1 + arg->namelen + 1 > arg->buflen) return (NULL); pathname = arg->buffer; strncpy(pathname, dir, dirlen); pathname[dirlen] = '/'; strcpy(pathname + dirlen + 1, arg->name); dbg(" Trying \"%s\"", pathname); if (access(pathname, F_OK) == 0) { /* We found it */ pathname = xmalloc(dirlen + 1 + arg->namelen + 1); strcpy(pathname, arg->buffer); return (pathname); } } return (NULL); } static char * search_library_path(const char *name, const char *path) { char *p; struct try_library_args arg; if (path == NULL) return NULL; arg.name = name; arg.namelen = strlen(name); arg.buffer = xmalloc(PATH_MAX); arg.buflen = PATH_MAX; p = path_enumerate(path, try_library_path, &arg); free(arg.buffer); return (p); } /* * Finds the library with the given name using the directory descriptors * listed in the LD_LIBRARY_PATH_FDS environment variable. * * Returns a freshly-opened close-on-exec file descriptor for the library, * or -1 if the library cannot be found. */ static char * search_library_pathfds(const char *name, const char *path, int *fdp) { char *envcopy, *fdstr, *found, *last_token; size_t len; int dirfd, fd; dbg("%s('%s', '%s', fdp)", __func__, name, path); /* Don't load from user-specified libdirs into setuid binaries. */ if (!trust) return (NULL); /* We can't do anything if LD_LIBRARY_PATH_FDS isn't set. */ if (path == NULL) return (NULL); /* LD_LIBRARY_PATH_FDS only works with relative paths. */ if (name[0] == '/') { dbg("Absolute path (%s) passed to %s", name, __func__); return (NULL); } /* * Use strtok_r() to walk the FD:FD:FD list. This requires a local * copy of the path, as strtok_r rewrites separator tokens * with '\0'. */ found = NULL; envcopy = xstrdup(path); for (fdstr = strtok_r(envcopy, ":", &last_token); fdstr != NULL; fdstr = strtok_r(NULL, ":", &last_token)) { dirfd = parse_libdir(fdstr); if (dirfd < 0) break; fd = __sys_openat(dirfd, name, O_RDONLY | O_CLOEXEC | O_VERIFY); if (fd >= 0) { *fdp = fd; len = strlen(fdstr) + strlen(name) + 3; found = xmalloc(len); if (rtld_snprintf(found, len, "#%d/%s", dirfd, name) < 0) { _rtld_error("error generating '%d/%s'", dirfd, name); rtld_die(); } dbg("open('%s') => %d", found, fd); break; } } free(envcopy); return (found); } int dlclose(void *handle) { Obj_Entry *root; RtldLockState lockstate; wlock_acquire(rtld_bind_lock, &lockstate); root = dlcheck(handle); if (root == NULL) { lock_release(rtld_bind_lock, &lockstate); return -1; } LD_UTRACE(UTRACE_DLCLOSE_START, handle, NULL, 0, root->dl_refcount, root->path); /* Unreference the object and its dependencies. */ root->dl_refcount--; if (root->refcount == 1) { /* * The object will be no longer referenced, so we must unload it. * First, call the fini functions. */ objlist_call_fini(&list_fini, root, &lockstate); unref_dag(root); /* Finish cleaning up the newly-unreferenced objects. */ GDB_STATE(RT_DELETE,&root->linkmap); unload_object(root); GDB_STATE(RT_CONSISTENT,NULL); } else unref_dag(root); LD_UTRACE(UTRACE_DLCLOSE_STOP, handle, NULL, 0, 0, NULL); lock_release(rtld_bind_lock, &lockstate); return 0; } char * dlerror(void) { char *msg = error_message; error_message = NULL; return msg; } /* * This function is deprecated and has no effect. */ void dllockinit(void *context, void *(*lock_create)(void *context), void (*rlock_acquire)(void *lock), void (*wlock_acquire)(void *lock), void (*lock_release)(void *lock), void (*lock_destroy)(void *lock), void (*context_destroy)(void *context)) { static void *cur_context; static void (*cur_context_destroy)(void *); /* Just destroy the context from the previous call, if necessary. */ if (cur_context_destroy != NULL) cur_context_destroy(cur_context); cur_context = context; cur_context_destroy = context_destroy; } void * dlopen(const char *name, int mode) { return (rtld_dlopen(name, -1, mode)); } void * fdlopen(int fd, int mode) { return (rtld_dlopen(NULL, fd, mode)); } static void * rtld_dlopen(const char *name, int fd, int mode) { RtldLockState lockstate; int lo_flags; LD_UTRACE(UTRACE_DLOPEN_START, NULL, NULL, 0, mode, name); ld_tracing = (mode & RTLD_TRACE) == 0 ? NULL : "1"; if (ld_tracing != NULL) { rlock_acquire(rtld_bind_lock, &lockstate); if (sigsetjmp(lockstate.env, 0) != 0) lock_upgrade(rtld_bind_lock, &lockstate); environ = (char **)*get_program_var_addr("environ", &lockstate); lock_release(rtld_bind_lock, &lockstate); } lo_flags = RTLD_LO_DLOPEN; if (mode & RTLD_NODELETE) lo_flags |= RTLD_LO_NODELETE; if (mode & RTLD_NOLOAD) lo_flags |= RTLD_LO_NOLOAD; if (ld_tracing != NULL) lo_flags |= RTLD_LO_TRACE; return (dlopen_object(name, fd, obj_main, lo_flags, mode & (RTLD_MODEMASK | RTLD_GLOBAL), NULL)); } static void dlopen_cleanup(Obj_Entry *obj) { obj->dl_refcount--; unref_dag(obj); if (obj->refcount == 0) unload_object(obj); } static Obj_Entry * dlopen_object(const char *name, int fd, Obj_Entry *refobj, int lo_flags, int mode, RtldLockState *lockstate) { Obj_Entry *old_obj_tail; Obj_Entry *obj; Objlist initlist; RtldLockState mlockstate; int result; objlist_init(&initlist); if (lockstate == NULL && !(lo_flags & RTLD_LO_EARLY)) { wlock_acquire(rtld_bind_lock, &mlockstate); lockstate = &mlockstate; } GDB_STATE(RT_ADD,NULL); old_obj_tail = globallist_curr(TAILQ_LAST(&obj_list, obj_entry_q)); obj = NULL; if (name == NULL && fd == -1) { obj = obj_main; obj->refcount++; } else { obj = load_object(name, fd, refobj, lo_flags); } if (obj) { obj->dl_refcount++; if (mode & RTLD_GLOBAL && objlist_find(&list_global, obj) == NULL) objlist_push_tail(&list_global, obj); if (globallist_next(old_obj_tail) != NULL) { /* We loaded something new. */ assert(globallist_next(old_obj_tail) == obj); result = load_needed_objects(obj, lo_flags & (RTLD_LO_DLOPEN | RTLD_LO_EARLY)); init_dag(obj); ref_dag(obj); if (result != -1) result = rtld_verify_versions(&obj->dagmembers); if (result != -1 && ld_tracing) goto trace; if (result == -1 || relocate_object_dag(obj, (mode & RTLD_MODEMASK) == RTLD_NOW, &obj_rtld, (lo_flags & RTLD_LO_EARLY) ? SYMLOOK_EARLY : 0, lockstate) == -1) { dlopen_cleanup(obj); obj = NULL; } else if (lo_flags & RTLD_LO_EARLY) { /* * Do not call the init functions for early loaded * filtees. The image is still not initialized enough * for them to work. * * Our object is found by the global object list and * will be ordered among all init calls done right * before transferring control to main. */ } else { /* Make list of init functions to call. */ initlist_add_objects(obj, obj, &initlist); } /* * Process all no_delete or global objects here, given * them own DAGs to prevent their dependencies from being * unloaded. This has to be done after we have loaded all * of the dependencies, so that we do not miss any. */ if (obj != NULL) process_z(obj); } else { /* * Bump the reference counts for objects on this DAG. If * this is the first dlopen() call for the object that was * already loaded as a dependency, initialize the dag * starting at it. */ init_dag(obj); ref_dag(obj); if ((lo_flags & RTLD_LO_TRACE) != 0) goto trace; } if (obj != NULL && ((lo_flags & RTLD_LO_NODELETE) != 0 || obj->z_nodelete) && !obj->ref_nodel) { dbg("obj %s nodelete", obj->path); ref_dag(obj); obj->z_nodelete = obj->ref_nodel = true; } } LD_UTRACE(UTRACE_DLOPEN_STOP, obj, NULL, 0, obj ? obj->dl_refcount : 0, name); GDB_STATE(RT_CONSISTENT,obj ? &obj->linkmap : NULL); if (!(lo_flags & RTLD_LO_EARLY)) { map_stacks_exec(lockstate); } if (initlist_objects_ifunc(&initlist, (mode & RTLD_MODEMASK) == RTLD_NOW, (lo_flags & RTLD_LO_EARLY) ? SYMLOOK_EARLY : 0, lockstate) == -1) { objlist_clear(&initlist); dlopen_cleanup(obj); if (lockstate == &mlockstate) lock_release(rtld_bind_lock, lockstate); return (NULL); } if (!(lo_flags & RTLD_LO_EARLY)) { /* Call the init functions. */ objlist_call_init(&initlist, lockstate); } objlist_clear(&initlist); if (lockstate == &mlockstate) lock_release(rtld_bind_lock, lockstate); return obj; trace: trace_loaded_objects(obj); if (lockstate == &mlockstate) lock_release(rtld_bind_lock, lockstate); exit(0); } static void * do_dlsym(void *handle, const char *name, void *retaddr, const Ver_Entry *ve, int flags) { DoneList donelist; const Obj_Entry *obj, *defobj; const Elf_Sym *def; SymLook req; RtldLockState lockstate; tls_index ti; void *sym; int res; def = NULL; defobj = NULL; symlook_init(&req, name); req.ventry = ve; req.flags = flags | SYMLOOK_IN_PLT; req.lockstate = &lockstate; LD_UTRACE(UTRACE_DLSYM_START, handle, NULL, 0, 0, name); rlock_acquire(rtld_bind_lock, &lockstate); if (sigsetjmp(lockstate.env, 0) != 0) lock_upgrade(rtld_bind_lock, &lockstate); if (handle == NULL || handle == RTLD_NEXT || handle == RTLD_DEFAULT || handle == RTLD_SELF) { if ((obj = obj_from_addr(retaddr)) == NULL) { _rtld_error("Cannot determine caller's shared object"); lock_release(rtld_bind_lock, &lockstate); LD_UTRACE(UTRACE_DLSYM_STOP, handle, NULL, 0, 0, name); return NULL; } if (handle == NULL) { /* Just the caller's shared object. */ res = symlook_obj(&req, obj); if (res == 0) { def = req.sym_out; defobj = req.defobj_out; } } else if (handle == RTLD_NEXT || /* Objects after caller's */ handle == RTLD_SELF) { /* ... caller included */ if (handle == RTLD_NEXT) obj = globallist_next(obj); TAILQ_FOREACH_FROM(obj, &obj_list, next) { if (obj->marker) continue; res = symlook_obj(&req, obj); if (res == 0) { if (def == NULL || ELF_ST_BIND(req.sym_out->st_info) != STB_WEAK) { def = req.sym_out; defobj = req.defobj_out; if (ELF_ST_BIND(def->st_info) != STB_WEAK) break; } } } /* * Search the dynamic linker itself, and possibly resolve the * symbol from there. This is how the application links to * dynamic linker services such as dlopen. */ if (def == NULL || ELF_ST_BIND(def->st_info) == STB_WEAK) { res = symlook_obj(&req, &obj_rtld); if (res == 0) { def = req.sym_out; defobj = req.defobj_out; } } } else { assert(handle == RTLD_DEFAULT); res = symlook_default(&req, obj); if (res == 0) { defobj = req.defobj_out; def = req.sym_out; } } } else { if ((obj = dlcheck(handle)) == NULL) { lock_release(rtld_bind_lock, &lockstate); LD_UTRACE(UTRACE_DLSYM_STOP, handle, NULL, 0, 0, name); return NULL; } donelist_init(&donelist); if (obj->mainprog) { /* Handle obtained by dlopen(NULL, ...) implies global scope. */ res = symlook_global(&req, &donelist); if (res == 0) { def = req.sym_out; defobj = req.defobj_out; } /* * Search the dynamic linker itself, and possibly resolve the * symbol from there. This is how the application links to * dynamic linker services such as dlopen. */ if (def == NULL || ELF_ST_BIND(def->st_info) == STB_WEAK) { res = symlook_obj(&req, &obj_rtld); if (res == 0) { def = req.sym_out; defobj = req.defobj_out; } } } else { /* Search the whole DAG rooted at the given object. */ res = symlook_list(&req, &obj->dagmembers, &donelist); if (res == 0) { def = req.sym_out; defobj = req.defobj_out; } } } if (def != NULL) { lock_release(rtld_bind_lock, &lockstate); /* * The value required by the caller is derived from the value * of the symbol. this is simply the relocated value of the * symbol. */ if (ELF_ST_TYPE(def->st_info) == STT_FUNC) sym = make_function_pointer(def, defobj); else if (ELF_ST_TYPE(def->st_info) == STT_GNU_IFUNC) sym = rtld_resolve_ifunc(defobj, def); else if (ELF_ST_TYPE(def->st_info) == STT_TLS) { ti.ti_module = defobj->tlsindex; ti.ti_offset = def->st_value; sym = __tls_get_addr(&ti); } else sym = defobj->relocbase + def->st_value; LD_UTRACE(UTRACE_DLSYM_STOP, handle, sym, 0, 0, name); return (sym); } _rtld_error("Undefined symbol \"%s\"", name); lock_release(rtld_bind_lock, &lockstate); LD_UTRACE(UTRACE_DLSYM_STOP, handle, NULL, 0, 0, name); return NULL; } void * dlsym(void *handle, const char *name) { return do_dlsym(handle, name, __builtin_return_address(0), NULL, SYMLOOK_DLSYM); } dlfunc_t dlfunc(void *handle, const char *name) { union { void *d; dlfunc_t f; } rv; rv.d = do_dlsym(handle, name, __builtin_return_address(0), NULL, SYMLOOK_DLSYM); return (rv.f); } void * dlvsym(void *handle, const char *name, const char *version) { Ver_Entry ventry; ventry.name = version; ventry.file = NULL; ventry.hash = elf_hash(version); ventry.flags= 0; return do_dlsym(handle, name, __builtin_return_address(0), &ventry, SYMLOOK_DLSYM); } int _rtld_addr_phdr(const void *addr, struct dl_phdr_info *phdr_info) { const Obj_Entry *obj; RtldLockState lockstate; rlock_acquire(rtld_bind_lock, &lockstate); obj = obj_from_addr(addr); if (obj == NULL) { _rtld_error("No shared object contains address"); lock_release(rtld_bind_lock, &lockstate); return (0); } rtld_fill_dl_phdr_info(obj, phdr_info); lock_release(rtld_bind_lock, &lockstate); return (1); } int dladdr(const void *addr, Dl_info *info) { const Obj_Entry *obj; const Elf_Sym *def; void *symbol_addr; unsigned long symoffset; RtldLockState lockstate; rlock_acquire(rtld_bind_lock, &lockstate); obj = obj_from_addr(addr); if (obj == NULL) { _rtld_error("No shared object contains address"); lock_release(rtld_bind_lock, &lockstate); return 0; } info->dli_fname = obj->path; info->dli_fbase = obj->mapbase; info->dli_saddr = (void *)0; info->dli_sname = NULL; /* * Walk the symbol list looking for the symbol whose address is * closest to the address sent in. */ for (symoffset = 0; symoffset < obj->dynsymcount; symoffset++) { def = obj->symtab + symoffset; /* * For skip the symbol if st_shndx is either SHN_UNDEF or * SHN_COMMON. */ if (def->st_shndx == SHN_UNDEF || def->st_shndx == SHN_COMMON) continue; /* * If the symbol is greater than the specified address, or if it * is further away from addr than the current nearest symbol, * then reject it. */ symbol_addr = obj->relocbase + def->st_value; if (symbol_addr > addr || symbol_addr < info->dli_saddr) continue; /* Update our idea of the nearest symbol. */ info->dli_sname = obj->strtab + def->st_name; info->dli_saddr = symbol_addr; /* Exact match? */ if (info->dli_saddr == addr) break; } lock_release(rtld_bind_lock, &lockstate); return 1; } int dlinfo(void *handle, int request, void *p) { const Obj_Entry *obj; RtldLockState lockstate; int error; rlock_acquire(rtld_bind_lock, &lockstate); if (handle == NULL || handle == RTLD_SELF) { void *retaddr; retaddr = __builtin_return_address(0); /* __GNUC__ only */ if ((obj = obj_from_addr(retaddr)) == NULL) _rtld_error("Cannot determine caller's shared object"); } else obj = dlcheck(handle); if (obj == NULL) { lock_release(rtld_bind_lock, &lockstate); return (-1); } error = 0; switch (request) { case RTLD_DI_LINKMAP: *((struct link_map const **)p) = &obj->linkmap; break; case RTLD_DI_ORIGIN: error = rtld_dirname(obj->path, p); break; case RTLD_DI_SERINFOSIZE: case RTLD_DI_SERINFO: error = do_search_info(obj, request, (struct dl_serinfo *)p); break; default: _rtld_error("Invalid request %d passed to dlinfo()", request); error = -1; } lock_release(rtld_bind_lock, &lockstate); return (error); } static void rtld_fill_dl_phdr_info(const Obj_Entry *obj, struct dl_phdr_info *phdr_info) { phdr_info->dlpi_addr = (Elf_Addr)obj->relocbase; phdr_info->dlpi_name = obj->path; phdr_info->dlpi_phdr = obj->phdr; phdr_info->dlpi_phnum = obj->phsize / sizeof(obj->phdr[0]); phdr_info->dlpi_tls_modid = obj->tlsindex; phdr_info->dlpi_tls_data = obj->tlsinit; phdr_info->dlpi_adds = obj_loads; phdr_info->dlpi_subs = obj_loads - obj_count; } int dl_iterate_phdr(__dl_iterate_hdr_callback callback, void *param) { struct dl_phdr_info phdr_info; Obj_Entry *obj, marker; RtldLockState bind_lockstate, phdr_lockstate; int error; bzero(&marker, sizeof(marker)); marker.marker = true; error = 0; wlock_acquire(rtld_phdr_lock, &phdr_lockstate); rlock_acquire(rtld_bind_lock, &bind_lockstate); for (obj = globallist_curr(TAILQ_FIRST(&obj_list)); obj != NULL;) { TAILQ_INSERT_AFTER(&obj_list, obj, &marker, next); rtld_fill_dl_phdr_info(obj, &phdr_info); lock_release(rtld_bind_lock, &bind_lockstate); error = callback(&phdr_info, sizeof phdr_info, param); rlock_acquire(rtld_bind_lock, &bind_lockstate); obj = globallist_next(&marker); TAILQ_REMOVE(&obj_list, &marker, next); if (error != 0) { lock_release(rtld_bind_lock, &bind_lockstate); lock_release(rtld_phdr_lock, &phdr_lockstate); return (error); } } if (error == 0) { rtld_fill_dl_phdr_info(&obj_rtld, &phdr_info); lock_release(rtld_bind_lock, &bind_lockstate); error = callback(&phdr_info, sizeof(phdr_info), param); } lock_release(rtld_phdr_lock, &phdr_lockstate); return (error); } static void * fill_search_info(const char *dir, size_t dirlen, void *param) { struct fill_search_info_args *arg; arg = param; if (arg->request == RTLD_DI_SERINFOSIZE) { arg->serinfo->dls_cnt ++; arg->serinfo->dls_size += sizeof(struct dl_serpath) + dirlen + 1; } else { struct dl_serpath *s_entry; s_entry = arg->serpath; s_entry->dls_name = arg->strspace; s_entry->dls_flags = arg->flags; strncpy(arg->strspace, dir, dirlen); arg->strspace[dirlen] = '\0'; arg->strspace += dirlen + 1; arg->serpath++; } return (NULL); } static int do_search_info(const Obj_Entry *obj, int request, struct dl_serinfo *info) { struct dl_serinfo _info; struct fill_search_info_args args; args.request = RTLD_DI_SERINFOSIZE; args.serinfo = &_info; _info.dls_size = __offsetof(struct dl_serinfo, dls_serpath); _info.dls_cnt = 0; path_enumerate(obj->rpath, fill_search_info, &args); path_enumerate(ld_library_path, fill_search_info, &args); path_enumerate(obj->runpath, fill_search_info, &args); path_enumerate(gethints(obj->z_nodeflib), fill_search_info, &args); if (!obj->z_nodeflib) path_enumerate(ld_standard_library_path, fill_search_info, &args); if (request == RTLD_DI_SERINFOSIZE) { info->dls_size = _info.dls_size; info->dls_cnt = _info.dls_cnt; return (0); } if (info->dls_cnt != _info.dls_cnt || info->dls_size != _info.dls_size) { _rtld_error("Uninitialized Dl_serinfo struct passed to dlinfo()"); return (-1); } args.request = RTLD_DI_SERINFO; args.serinfo = info; args.serpath = &info->dls_serpath[0]; args.strspace = (char *)&info->dls_serpath[_info.dls_cnt]; args.flags = LA_SER_RUNPATH; if (path_enumerate(obj->rpath, fill_search_info, &args) != NULL) return (-1); args.flags = LA_SER_LIBPATH; if (path_enumerate(ld_library_path, fill_search_info, &args) != NULL) return (-1); args.flags = LA_SER_RUNPATH; if (path_enumerate(obj->runpath, fill_search_info, &args) != NULL) return (-1); args.flags = LA_SER_CONFIG; if (path_enumerate(gethints(obj->z_nodeflib), fill_search_info, &args) != NULL) return (-1); args.flags = LA_SER_DEFAULT; if (!obj->z_nodeflib && path_enumerate(ld_standard_library_path, fill_search_info, &args) != NULL) return (-1); return (0); } static int rtld_dirname(const char *path, char *bname) { const char *endp; /* Empty or NULL string gets treated as "." */ if (path == NULL || *path == '\0') { bname[0] = '.'; bname[1] = '\0'; return (0); } /* Strip trailing slashes */ endp = path + strlen(path) - 1; while (endp > path && *endp == '/') endp--; /* Find the start of the dir */ while (endp > path && *endp != '/') endp--; /* Either the dir is "/" or there are no slashes */ if (endp == path) { bname[0] = *endp == '/' ? '/' : '.'; bname[1] = '\0'; return (0); } else { do { endp--; } while (endp > path && *endp == '/'); } if (endp - path + 2 > PATH_MAX) { _rtld_error("Filename is too long: %s", path); return(-1); } strncpy(bname, path, endp - path + 1); bname[endp - path + 1] = '\0'; return (0); } static int rtld_dirname_abs(const char *path, char *base) { char *last; if (realpath(path, base) == NULL) return (-1); dbg("%s -> %s", path, base); last = strrchr(base, '/'); if (last == NULL) return (-1); if (last != base) *last = '\0'; return (0); } static void linkmap_add(Obj_Entry *obj) { struct link_map *l = &obj->linkmap; struct link_map *prev; obj->linkmap.l_name = obj->path; obj->linkmap.l_addr = obj->mapbase; obj->linkmap.l_ld = obj->dynamic; #ifdef __mips__ /* GDB needs load offset on MIPS to use the symbols */ obj->linkmap.l_offs = obj->relocbase; #endif if (r_debug.r_map == NULL) { r_debug.r_map = l; return; } /* * Scan to the end of the list, but not past the entry for the * dynamic linker, which we want to keep at the very end. */ for (prev = r_debug.r_map; prev->l_next != NULL && prev->l_next != &obj_rtld.linkmap; prev = prev->l_next) ; /* Link in the new entry. */ l->l_prev = prev; l->l_next = prev->l_next; if (l->l_next != NULL) l->l_next->l_prev = l; prev->l_next = l; } static void linkmap_delete(Obj_Entry *obj) { struct link_map *l = &obj->linkmap; if (l->l_prev == NULL) { if ((r_debug.r_map = l->l_next) != NULL) l->l_next->l_prev = NULL; return; } if ((l->l_prev->l_next = l->l_next) != NULL) l->l_next->l_prev = l->l_prev; } /* * Function for the debugger to set a breakpoint on to gain control. * * The two parameters allow the debugger to easily find and determine * what the runtime loader is doing and to whom it is doing it. * * When the loadhook trap is hit (r_debug_state, set at program * initialization), the arguments can be found on the stack: * * +8 struct link_map *m * +4 struct r_debug *rd * +0 RetAddr */ void r_debug_state(struct r_debug* rd, struct link_map *m) { /* * The following is a hack to force the compiler to emit calls to * this function, even when optimizing. If the function is empty, * the compiler is not obliged to emit any code for calls to it, * even when marked __noinline. However, gdb depends on those * calls being made. */ __compiler_membar(); } /* * A function called after init routines have completed. This can be used to * break before a program's entry routine is called, and can be used when * main is not available in the symbol table. */ void _r_debug_postinit(struct link_map *m) { /* See r_debug_state(). */ __compiler_membar(); } /* * Get address of the pointer variable in the main program. * Prefer non-weak symbol over the weak one. */ static const void ** get_program_var_addr(const char *name, RtldLockState *lockstate) { SymLook req; DoneList donelist; symlook_init(&req, name); req.lockstate = lockstate; donelist_init(&donelist); if (symlook_global(&req, &donelist) != 0) return (NULL); if (ELF_ST_TYPE(req.sym_out->st_info) == STT_FUNC) return ((const void **)make_function_pointer(req.sym_out, req.defobj_out)); else if (ELF_ST_TYPE(req.sym_out->st_info) == STT_GNU_IFUNC) return ((const void **)rtld_resolve_ifunc(req.defobj_out, req.sym_out)); else return ((const void **)(req.defobj_out->relocbase + req.sym_out->st_value)); } /* * Set a pointer variable in the main program to the given value. This * is used to set key variables such as "environ" before any of the * init functions are called. */ static void set_program_var(const char *name, const void *value) { const void **addr; if ((addr = get_program_var_addr(name, NULL)) != NULL) { dbg("\"%s\": *%p <-- %p", name, addr, value); *addr = value; } } /* * Search the global objects, including dependencies and main object, * for the given symbol. */ static int symlook_global(SymLook *req, DoneList *donelist) { SymLook req1; const Objlist_Entry *elm; int res; symlook_init_from_req(&req1, req); /* Search all objects loaded at program start up. */ if (req->defobj_out == NULL || ELF_ST_BIND(req->sym_out->st_info) == STB_WEAK) { res = symlook_list(&req1, &list_main, donelist); if (res == 0 && (req->defobj_out == NULL || ELF_ST_BIND(req1.sym_out->st_info) != STB_WEAK)) { req->sym_out = req1.sym_out; req->defobj_out = req1.defobj_out; assert(req->defobj_out != NULL); } } /* Search all DAGs whose roots are RTLD_GLOBAL objects. */ STAILQ_FOREACH(elm, &list_global, link) { if (req->defobj_out != NULL && ELF_ST_BIND(req->sym_out->st_info) != STB_WEAK) break; res = symlook_list(&req1, &elm->obj->dagmembers, donelist); if (res == 0 && (req->defobj_out == NULL || ELF_ST_BIND(req1.sym_out->st_info) != STB_WEAK)) { req->sym_out = req1.sym_out; req->defobj_out = req1.defobj_out; assert(req->defobj_out != NULL); } } return (req->sym_out != NULL ? 0 : ESRCH); } /* * Given a symbol name in a referencing object, find the corresponding * definition of the symbol. Returns a pointer to the symbol, or NULL if * no definition was found. Returns a pointer to the Obj_Entry of the * defining object via the reference parameter DEFOBJ_OUT. */ static int symlook_default(SymLook *req, const Obj_Entry *refobj) { DoneList donelist; const Objlist_Entry *elm; SymLook req1; int res; donelist_init(&donelist); symlook_init_from_req(&req1, req); /* Look first in the referencing object if linked symbolically. */ if (refobj->symbolic && !donelist_check(&donelist, refobj)) { res = symlook_obj(&req1, refobj); if (res == 0) { req->sym_out = req1.sym_out; req->defobj_out = req1.defobj_out; assert(req->defobj_out != NULL); } } symlook_global(req, &donelist); /* Search all dlopened DAGs containing the referencing object. */ STAILQ_FOREACH(elm, &refobj->dldags, link) { if (req->sym_out != NULL && ELF_ST_BIND(req->sym_out->st_info) != STB_WEAK) break; res = symlook_list(&req1, &elm->obj->dagmembers, &donelist); if (res == 0 && (req->sym_out == NULL || ELF_ST_BIND(req1.sym_out->st_info) != STB_WEAK)) { req->sym_out = req1.sym_out; req->defobj_out = req1.defobj_out; assert(req->defobj_out != NULL); } } /* * Search the dynamic linker itself, and possibly resolve the * symbol from there. This is how the application links to * dynamic linker services such as dlopen. */ if (req->sym_out == NULL || ELF_ST_BIND(req->sym_out->st_info) == STB_WEAK) { res = symlook_obj(&req1, &obj_rtld); if (res == 0) { req->sym_out = req1.sym_out; req->defobj_out = req1.defobj_out; assert(req->defobj_out != NULL); } } return (req->sym_out != NULL ? 0 : ESRCH); } static int symlook_list(SymLook *req, const Objlist *objlist, DoneList *dlp) { const Elf_Sym *def; const Obj_Entry *defobj; const Objlist_Entry *elm; SymLook req1; int res; def = NULL; defobj = NULL; STAILQ_FOREACH(elm, objlist, link) { if (donelist_check(dlp, elm->obj)) continue; symlook_init_from_req(&req1, req); if ((res = symlook_obj(&req1, elm->obj)) == 0) { if (def == NULL || ELF_ST_BIND(req1.sym_out->st_info) != STB_WEAK) { def = req1.sym_out; defobj = req1.defobj_out; if (ELF_ST_BIND(def->st_info) != STB_WEAK) break; } } } if (def != NULL) { req->sym_out = def; req->defobj_out = defobj; return (0); } return (ESRCH); } /* * Search the chain of DAGS cointed to by the given Needed_Entry * for a symbol of the given name. Each DAG is scanned completely * before advancing to the next one. Returns a pointer to the symbol, * or NULL if no definition was found. */ static int symlook_needed(SymLook *req, const Needed_Entry *needed, DoneList *dlp) { const Elf_Sym *def; const Needed_Entry *n; const Obj_Entry *defobj; SymLook req1; int res; def = NULL; defobj = NULL; symlook_init_from_req(&req1, req); for (n = needed; n != NULL; n = n->next) { if (n->obj == NULL || (res = symlook_list(&req1, &n->obj->dagmembers, dlp)) != 0) continue; if (def == NULL || ELF_ST_BIND(req1.sym_out->st_info) != STB_WEAK) { def = req1.sym_out; defobj = req1.defobj_out; if (ELF_ST_BIND(def->st_info) != STB_WEAK) break; } } if (def != NULL) { req->sym_out = def; req->defobj_out = defobj; return (0); } return (ESRCH); } /* * Search the symbol table of a single shared object for a symbol of * the given name and version, if requested. Returns a pointer to the * symbol, or NULL if no definition was found. If the object is * filter, return filtered symbol from filtee. * * The symbol's hash value is passed in for efficiency reasons; that * eliminates many recomputations of the hash value. */ int symlook_obj(SymLook *req, const Obj_Entry *obj) { DoneList donelist; SymLook req1; int flags, res, mres; /* * If there is at least one valid hash at this point, we prefer to * use the faster GNU version if available. */ if (obj->valid_hash_gnu) mres = symlook_obj1_gnu(req, obj); else if (obj->valid_hash_sysv) mres = symlook_obj1_sysv(req, obj); else return (EINVAL); if (mres == 0) { if (obj->needed_filtees != NULL) { flags = (req->flags & SYMLOOK_EARLY) ? RTLD_LO_EARLY : 0; load_filtees(__DECONST(Obj_Entry *, obj), flags, req->lockstate); donelist_init(&donelist); symlook_init_from_req(&req1, req); res = symlook_needed(&req1, obj->needed_filtees, &donelist); if (res == 0) { req->sym_out = req1.sym_out; req->defobj_out = req1.defobj_out; } return (res); } if (obj->needed_aux_filtees != NULL) { flags = (req->flags & SYMLOOK_EARLY) ? RTLD_LO_EARLY : 0; load_filtees(__DECONST(Obj_Entry *, obj), flags, req->lockstate); donelist_init(&donelist); symlook_init_from_req(&req1, req); res = symlook_needed(&req1, obj->needed_aux_filtees, &donelist); if (res == 0) { req->sym_out = req1.sym_out; req->defobj_out = req1.defobj_out; return (res); } } } return (mres); } /* Symbol match routine common to both hash functions */ static bool matched_symbol(SymLook *req, const Obj_Entry *obj, Sym_Match_Result *result, const unsigned long symnum) { Elf_Versym verndx; const Elf_Sym *symp; const char *strp; symp = obj->symtab + symnum; strp = obj->strtab + symp->st_name; switch (ELF_ST_TYPE(symp->st_info)) { case STT_FUNC: case STT_NOTYPE: case STT_OBJECT: case STT_COMMON: case STT_GNU_IFUNC: if (symp->st_value == 0) return (false); /* fallthrough */ case STT_TLS: if (symp->st_shndx != SHN_UNDEF) break; #ifndef __mips__ else if (((req->flags & SYMLOOK_IN_PLT) == 0) && (ELF_ST_TYPE(symp->st_info) == STT_FUNC)) break; /* fallthrough */ #endif default: return (false); } if (req->name[0] != strp[0] || strcmp(req->name, strp) != 0) return (false); if (req->ventry == NULL) { if (obj->versyms != NULL) { verndx = VER_NDX(obj->versyms[symnum]); if (verndx > obj->vernum) { _rtld_error( "%s: symbol %s references wrong version %d", obj->path, obj->strtab + symnum, verndx); return (false); } /* * If we are not called from dlsym (i.e. this * is a normal relocation from unversioned * binary), accept the symbol immediately if * it happens to have first version after this * shared object became versioned. Otherwise, * if symbol is versioned and not hidden, * remember it. If it is the only symbol with * this name exported by the shared object, it * will be returned as a match by the calling * function. If symbol is global (verndx < 2) * accept it unconditionally. */ if ((req->flags & SYMLOOK_DLSYM) == 0 && verndx == VER_NDX_GIVEN) { result->sym_out = symp; return (true); } else if (verndx >= VER_NDX_GIVEN) { if ((obj->versyms[symnum] & VER_NDX_HIDDEN) == 0) { if (result->vsymp == NULL) result->vsymp = symp; result->vcount++; } return (false); } } result->sym_out = symp; return (true); } if (obj->versyms == NULL) { if (object_match_name(obj, req->ventry->name)) { _rtld_error("%s: object %s should provide version %s " "for symbol %s", obj_rtld.path, obj->path, req->ventry->name, obj->strtab + symnum); return (false); } } else { verndx = VER_NDX(obj->versyms[symnum]); if (verndx > obj->vernum) { _rtld_error("%s: symbol %s references wrong version %d", obj->path, obj->strtab + symnum, verndx); return (false); } if (obj->vertab[verndx].hash != req->ventry->hash || strcmp(obj->vertab[verndx].name, req->ventry->name)) { /* * Version does not match. Look if this is a * global symbol and if it is not hidden. If * global symbol (verndx < 2) is available, * use it. Do not return symbol if we are * called by dlvsym, because dlvsym looks for * a specific version and default one is not * what dlvsym wants. */ if ((req->flags & SYMLOOK_DLSYM) || (verndx >= VER_NDX_GIVEN) || (obj->versyms[symnum] & VER_NDX_HIDDEN)) return (false); } } result->sym_out = symp; return (true); } /* * Search for symbol using SysV hash function. * obj->buckets is known not to be NULL at this point; the test for this was * performed with the obj->valid_hash_sysv assignment. */ static int symlook_obj1_sysv(SymLook *req, const Obj_Entry *obj) { unsigned long symnum; Sym_Match_Result matchres; matchres.sym_out = NULL; matchres.vsymp = NULL; matchres.vcount = 0; for (symnum = obj->buckets[req->hash % obj->nbuckets]; symnum != STN_UNDEF; symnum = obj->chains[symnum]) { if (symnum >= obj->nchains) return (ESRCH); /* Bad object */ if (matched_symbol(req, obj, &matchres, symnum)) { req->sym_out = matchres.sym_out; req->defobj_out = obj; return (0); } } if (matchres.vcount == 1) { req->sym_out = matchres.vsymp; req->defobj_out = obj; return (0); } return (ESRCH); } /* Search for symbol using GNU hash function */ static int symlook_obj1_gnu(SymLook *req, const Obj_Entry *obj) { Elf_Addr bloom_word; const Elf32_Word *hashval; Elf32_Word bucket; Sym_Match_Result matchres; unsigned int h1, h2; unsigned long symnum; matchres.sym_out = NULL; matchres.vsymp = NULL; matchres.vcount = 0; /* Pick right bitmask word from Bloom filter array */ bloom_word = obj->bloom_gnu[(req->hash_gnu / __ELF_WORD_SIZE) & obj->maskwords_bm_gnu]; /* Calculate modulus word size of gnu hash and its derivative */ h1 = req->hash_gnu & (__ELF_WORD_SIZE - 1); h2 = ((req->hash_gnu >> obj->shift2_gnu) & (__ELF_WORD_SIZE - 1)); /* Filter out the "definitely not in set" queries */ if (((bloom_word >> h1) & (bloom_word >> h2) & 1) == 0) return (ESRCH); /* Locate hash chain and corresponding value element*/ bucket = obj->buckets_gnu[req->hash_gnu % obj->nbuckets_gnu]; if (bucket == 0) return (ESRCH); hashval = &obj->chain_zero_gnu[bucket]; do { if (((*hashval ^ req->hash_gnu) >> 1) == 0) { symnum = hashval - obj->chain_zero_gnu; if (matched_symbol(req, obj, &matchres, symnum)) { req->sym_out = matchres.sym_out; req->defobj_out = obj; return (0); } } } while ((*hashval++ & 1) == 0); if (matchres.vcount == 1) { req->sym_out = matchres.vsymp; req->defobj_out = obj; return (0); } return (ESRCH); } static void trace_loaded_objects(Obj_Entry *obj) { char *fmt1, *fmt2, *fmt, *main_local, *list_containers; int c; if ((main_local = getenv(_LD("TRACE_LOADED_OBJECTS_PROGNAME"))) == NULL) main_local = ""; if ((fmt1 = getenv(_LD("TRACE_LOADED_OBJECTS_FMT1"))) == NULL) fmt1 = "\t%o => %p (%x)\n"; if ((fmt2 = getenv(_LD("TRACE_LOADED_OBJECTS_FMT2"))) == NULL) fmt2 = "\t%o (%x)\n"; list_containers = getenv(_LD("TRACE_LOADED_OBJECTS_ALL")); TAILQ_FOREACH_FROM(obj, &obj_list, next) { Needed_Entry *needed; char *name, *path; bool is_lib; if (obj->marker) continue; if (list_containers && obj->needed != NULL) rtld_printf("%s:\n", obj->path); for (needed = obj->needed; needed; needed = needed->next) { if (needed->obj != NULL) { if (needed->obj->traced && !list_containers) continue; needed->obj->traced = true; path = needed->obj->path; } else path = "not found"; name = (char *)obj->strtab + needed->name; is_lib = strncmp(name, "lib", 3) == 0; /* XXX - bogus */ fmt = is_lib ? fmt1 : fmt2; while ((c = *fmt++) != '\0') { switch (c) { default: rtld_putchar(c); continue; case '\\': switch (c = *fmt) { case '\0': continue; case 'n': rtld_putchar('\n'); break; case 't': rtld_putchar('\t'); break; } break; case '%': switch (c = *fmt) { case '\0': continue; case '%': default: rtld_putchar(c); break; case 'A': rtld_putstr(main_local); break; case 'a': rtld_putstr(obj_main->path); break; case 'o': rtld_putstr(name); break; #if 0 case 'm': rtld_printf("%d", sodp->sod_major); break; case 'n': rtld_printf("%d", sodp->sod_minor); break; #endif case 'p': rtld_putstr(path); break; case 'x': rtld_printf("%p", needed->obj ? needed->obj->mapbase : 0); break; } break; } ++fmt; } } } } /* * Unload a dlopened object and its dependencies from memory and from * our data structures. It is assumed that the DAG rooted in the * object has already been unreferenced, and that the object has a * reference count of 0. */ static void unload_object(Obj_Entry *root) { Obj_Entry *obj, *obj1; assert(root->refcount == 0); /* * Pass over the DAG removing unreferenced objects from * appropriate lists. */ unlink_object(root); /* Unmap all objects that are no longer referenced. */ TAILQ_FOREACH_SAFE(obj, &obj_list, next, obj1) { if (obj->marker || obj->refcount != 0) continue; LD_UTRACE(UTRACE_UNLOAD_OBJECT, obj, obj->mapbase, obj->mapsize, 0, obj->path); dbg("unloading \"%s\"", obj->path); unload_filtees(root); munmap(obj->mapbase, obj->mapsize); linkmap_delete(obj); TAILQ_REMOVE(&obj_list, obj, next); obj_count--; obj_free(obj); } } static void unlink_object(Obj_Entry *root) { Objlist_Entry *elm; if (root->refcount == 0) { /* Remove the object from the RTLD_GLOBAL list. */ objlist_remove(&list_global, root); /* Remove the object from all objects' DAG lists. */ STAILQ_FOREACH(elm, &root->dagmembers, link) { objlist_remove(&elm->obj->dldags, root); if (elm->obj != root) unlink_object(elm->obj); } } } static void ref_dag(Obj_Entry *root) { Objlist_Entry *elm; assert(root->dag_inited); STAILQ_FOREACH(elm, &root->dagmembers, link) elm->obj->refcount++; } static void unref_dag(Obj_Entry *root) { Objlist_Entry *elm; assert(root->dag_inited); STAILQ_FOREACH(elm, &root->dagmembers, link) elm->obj->refcount--; } /* * Common code for MD __tls_get_addr(). */ static void *tls_get_addr_slow(Elf_Addr **, int, size_t) __noinline; static void * tls_get_addr_slow(Elf_Addr **dtvp, int index, size_t offset) { Elf_Addr *newdtv, *dtv; RtldLockState lockstate; int to_copy; dtv = *dtvp; /* Check dtv generation in case new modules have arrived */ if (dtv[0] != tls_dtv_generation) { wlock_acquire(rtld_bind_lock, &lockstate); newdtv = xcalloc(tls_max_index + 2, sizeof(Elf_Addr)); to_copy = dtv[1]; if (to_copy > tls_max_index) to_copy = tls_max_index; memcpy(&newdtv[2], &dtv[2], to_copy * sizeof(Elf_Addr)); newdtv[0] = tls_dtv_generation; newdtv[1] = tls_max_index; free(dtv); lock_release(rtld_bind_lock, &lockstate); dtv = *dtvp = newdtv; } /* Dynamically allocate module TLS if necessary */ if (dtv[index + 1] == 0) { /* Signal safe, wlock will block out signals. */ wlock_acquire(rtld_bind_lock, &lockstate); if (!dtv[index + 1]) dtv[index + 1] = (Elf_Addr)allocate_module_tls(index); lock_release(rtld_bind_lock, &lockstate); } return ((void *)(dtv[index + 1] + offset)); } void * tls_get_addr_common(Elf_Addr **dtvp, int index, size_t offset) { Elf_Addr *dtv; dtv = *dtvp; /* Check dtv generation in case new modules have arrived */ if (__predict_true(dtv[0] == tls_dtv_generation && dtv[index + 1] != 0)) return ((void *)(dtv[index + 1] + offset)); return (tls_get_addr_slow(dtvp, index, offset)); } #if defined(__aarch64__) || defined(__arm__) || defined(__mips__) || \ defined(__powerpc__) || defined(__riscv__) /* * Allocate Static TLS using the Variant I method. */ void * allocate_tls(Obj_Entry *objs, void *oldtcb, size_t tcbsize, size_t tcbalign) { Obj_Entry *obj; char *tcb; Elf_Addr **tls; Elf_Addr *dtv; Elf_Addr addr; int i; if (oldtcb != NULL && tcbsize == TLS_TCB_SIZE) return (oldtcb); assert(tcbsize >= TLS_TCB_SIZE); tcb = xcalloc(1, tls_static_space - TLS_TCB_SIZE + tcbsize); tls = (Elf_Addr **)(tcb + tcbsize - TLS_TCB_SIZE); if (oldtcb != NULL) { memcpy(tls, oldtcb, tls_static_space); free(oldtcb); /* Adjust the DTV. */ dtv = tls[0]; for (i = 0; i < dtv[1]; i++) { if (dtv[i+2] >= (Elf_Addr)oldtcb && dtv[i+2] < (Elf_Addr)oldtcb + tls_static_space) { dtv[i+2] = dtv[i+2] - (Elf_Addr)oldtcb + (Elf_Addr)tls; } } } else { dtv = xcalloc(tls_max_index + 2, sizeof(Elf_Addr)); tls[0] = dtv; dtv[0] = tls_dtv_generation; dtv[1] = tls_max_index; for (obj = globallist_curr(objs); obj != NULL; obj = globallist_next(obj)) { if (obj->tlsoffset > 0) { addr = (Elf_Addr)tls + obj->tlsoffset; if (obj->tlsinitsize > 0) memcpy((void*) addr, obj->tlsinit, obj->tlsinitsize); if (obj->tlssize > obj->tlsinitsize) memset((void*) (addr + obj->tlsinitsize), 0, obj->tlssize - obj->tlsinitsize); dtv[obj->tlsindex + 1] = addr; } } } return (tcb); } void free_tls(void *tcb, size_t tcbsize, size_t tcbalign) { Elf_Addr *dtv; Elf_Addr tlsstart, tlsend; int dtvsize, i; assert(tcbsize >= TLS_TCB_SIZE); tlsstart = (Elf_Addr)tcb + tcbsize - TLS_TCB_SIZE; tlsend = tlsstart + tls_static_space; dtv = *(Elf_Addr **)tlsstart; dtvsize = dtv[1]; for (i = 0; i < dtvsize; i++) { if (dtv[i+2] && (dtv[i+2] < tlsstart || dtv[i+2] >= tlsend)) { free((void*)dtv[i+2]); } } free(dtv); free(tcb); } #endif #if defined(__i386__) || defined(__amd64__) || defined(__sparc64__) /* * Allocate Static TLS using the Variant II method. */ void * allocate_tls(Obj_Entry *objs, void *oldtls, size_t tcbsize, size_t tcbalign) { Obj_Entry *obj; size_t size, ralign; char *tls; Elf_Addr *dtv, *olddtv; Elf_Addr segbase, oldsegbase, addr; int i; ralign = tcbalign; if (tls_static_max_align > ralign) ralign = tls_static_max_align; size = round(tls_static_space, ralign) + round(tcbsize, ralign); assert(tcbsize >= 2*sizeof(Elf_Addr)); tls = malloc_aligned(size, ralign); dtv = xcalloc(tls_max_index + 2, sizeof(Elf_Addr)); segbase = (Elf_Addr)(tls + round(tls_static_space, ralign)); ((Elf_Addr*)segbase)[0] = segbase; ((Elf_Addr*)segbase)[1] = (Elf_Addr) dtv; dtv[0] = tls_dtv_generation; dtv[1] = tls_max_index; if (oldtls) { /* * Copy the static TLS block over whole. */ oldsegbase = (Elf_Addr) oldtls; memcpy((void *)(segbase - tls_static_space), (const void *)(oldsegbase - tls_static_space), tls_static_space); /* * If any dynamic TLS blocks have been created tls_get_addr(), * move them over. */ olddtv = ((Elf_Addr**)oldsegbase)[1]; for (i = 0; i < olddtv[1]; i++) { if (olddtv[i+2] < oldsegbase - size || olddtv[i+2] > oldsegbase) { dtv[i+2] = olddtv[i+2]; olddtv[i+2] = 0; } } /* * We assume that this block was the one we created with * allocate_initial_tls(). */ free_tls(oldtls, 2*sizeof(Elf_Addr), sizeof(Elf_Addr)); } else { obj = objs; TAILQ_FOREACH_FROM(obj, &obj_list, next) { if (obj->marker || obj->tlsoffset == 0) continue; addr = segbase - obj->tlsoffset; memset((void*) (addr + obj->tlsinitsize), 0, obj->tlssize - obj->tlsinitsize); if (obj->tlsinit) memcpy((void*) addr, obj->tlsinit, obj->tlsinitsize); dtv[obj->tlsindex + 1] = addr; } } return (void*) segbase; } void free_tls(void *tls, size_t tcbsize, size_t tcbalign) { Elf_Addr* dtv; size_t size, ralign; int dtvsize, i; Elf_Addr tlsstart, tlsend; /* * Figure out the size of the initial TLS block so that we can * find stuff which ___tls_get_addr() allocated dynamically. */ ralign = tcbalign; if (tls_static_max_align > ralign) ralign = tls_static_max_align; size = round(tls_static_space, ralign); dtv = ((Elf_Addr**)tls)[1]; dtvsize = dtv[1]; tlsend = (Elf_Addr) tls; tlsstart = tlsend - size; for (i = 0; i < dtvsize; i++) { if (dtv[i + 2] != 0 && (dtv[i + 2] < tlsstart || dtv[i + 2] > tlsend)) { free_aligned((void *)dtv[i + 2]); } } free_aligned((void *)tlsstart); free((void*) dtv); } #endif /* * Allocate TLS block for module with given index. */ void * allocate_module_tls(int index) { Obj_Entry* obj; char* p; TAILQ_FOREACH(obj, &obj_list, next) { if (obj->marker) continue; if (obj->tlsindex == index) break; } if (!obj) { _rtld_error("Can't find module with TLS index %d", index); rtld_die(); } p = malloc_aligned(obj->tlssize, obj->tlsalign); memcpy(p, obj->tlsinit, obj->tlsinitsize); memset(p + obj->tlsinitsize, 0, obj->tlssize - obj->tlsinitsize); return p; } bool allocate_tls_offset(Obj_Entry *obj) { size_t off; if (obj->tls_done) return true; if (obj->tlssize == 0) { obj->tls_done = true; return true; } if (tls_last_offset == 0) off = calculate_first_tls_offset(obj->tlssize, obj->tlsalign); else off = calculate_tls_offset(tls_last_offset, tls_last_size, obj->tlssize, obj->tlsalign); /* * If we have already fixed the size of the static TLS block, we * must stay within that size. When allocating the static TLS, we * leave a small amount of space spare to be used for dynamically * loading modules which use static TLS. */ if (tls_static_space != 0) { if (calculate_tls_end(off, obj->tlssize) > tls_static_space) return false; } else if (obj->tlsalign > tls_static_max_align) { tls_static_max_align = obj->tlsalign; } tls_last_offset = obj->tlsoffset = off; tls_last_size = obj->tlssize; obj->tls_done = true; return true; } void free_tls_offset(Obj_Entry *obj) { /* * If we were the last thing to allocate out of the static TLS * block, we give our space back to the 'allocator'. This is a * simplistic workaround to allow libGL.so.1 to be loaded and * unloaded multiple times. */ if (calculate_tls_end(obj->tlsoffset, obj->tlssize) == calculate_tls_end(tls_last_offset, tls_last_size)) { tls_last_offset -= obj->tlssize; tls_last_size = 0; } } void * _rtld_allocate_tls(void *oldtls, size_t tcbsize, size_t tcbalign) { void *ret; RtldLockState lockstate; wlock_acquire(rtld_bind_lock, &lockstate); ret = allocate_tls(globallist_curr(TAILQ_FIRST(&obj_list)), oldtls, tcbsize, tcbalign); lock_release(rtld_bind_lock, &lockstate); return (ret); } void _rtld_free_tls(void *tcb, size_t tcbsize, size_t tcbalign) { RtldLockState lockstate; wlock_acquire(rtld_bind_lock, &lockstate); free_tls(tcb, tcbsize, tcbalign); lock_release(rtld_bind_lock, &lockstate); } static void object_add_name(Obj_Entry *obj, const char *name) { Name_Entry *entry; size_t len; len = strlen(name); entry = malloc(sizeof(Name_Entry) + len); if (entry != NULL) { strcpy(entry->name, name); STAILQ_INSERT_TAIL(&obj->names, entry, link); } } static int object_match_name(const Obj_Entry *obj, const char *name) { Name_Entry *entry; STAILQ_FOREACH(entry, &obj->names, link) { if (strcmp(name, entry->name) == 0) return (1); } return (0); } static Obj_Entry * locate_dependency(const Obj_Entry *obj, const char *name) { const Objlist_Entry *entry; const Needed_Entry *needed; STAILQ_FOREACH(entry, &list_main, link) { if (object_match_name(entry->obj, name)) return entry->obj; } for (needed = obj->needed; needed != NULL; needed = needed->next) { if (strcmp(obj->strtab + needed->name, name) == 0 || (needed->obj != NULL && object_match_name(needed->obj, name))) { /* * If there is DT_NEEDED for the name we are looking for, * we are all set. Note that object might not be found if * dependency was not loaded yet, so the function can * return NULL here. This is expected and handled * properly by the caller. */ return (needed->obj); } } _rtld_error("%s: Unexpected inconsistency: dependency %s not found", obj->path, name); rtld_die(); } static int check_object_provided_version(Obj_Entry *refobj, const Obj_Entry *depobj, const Elf_Vernaux *vna) { const Elf_Verdef *vd; const char *vername; vername = refobj->strtab + vna->vna_name; vd = depobj->verdef; if (vd == NULL) { _rtld_error("%s: version %s required by %s not defined", depobj->path, vername, refobj->path); return (-1); } for (;;) { if (vd->vd_version != VER_DEF_CURRENT) { _rtld_error("%s: Unsupported version %d of Elf_Verdef entry", depobj->path, vd->vd_version); return (-1); } if (vna->vna_hash == vd->vd_hash) { const Elf_Verdaux *aux = (const Elf_Verdaux *) ((char *)vd + vd->vd_aux); if (strcmp(vername, depobj->strtab + aux->vda_name) == 0) return (0); } if (vd->vd_next == 0) break; vd = (const Elf_Verdef *) ((char *)vd + vd->vd_next); } if (vna->vna_flags & VER_FLG_WEAK) return (0); _rtld_error("%s: version %s required by %s not found", depobj->path, vername, refobj->path); return (-1); } static int rtld_verify_object_versions(Obj_Entry *obj) { const Elf_Verneed *vn; const Elf_Verdef *vd; const Elf_Verdaux *vda; const Elf_Vernaux *vna; const Obj_Entry *depobj; int maxvernum, vernum; if (obj->ver_checked) return (0); obj->ver_checked = true; maxvernum = 0; /* * Walk over defined and required version records and figure out * max index used by any of them. Do very basic sanity checking * while there. */ vn = obj->verneed; while (vn != NULL) { if (vn->vn_version != VER_NEED_CURRENT) { _rtld_error("%s: Unsupported version %d of Elf_Verneed entry", obj->path, vn->vn_version); return (-1); } vna = (const Elf_Vernaux *) ((char *)vn + vn->vn_aux); for (;;) { vernum = VER_NEED_IDX(vna->vna_other); if (vernum > maxvernum) maxvernum = vernum; if (vna->vna_next == 0) break; vna = (const Elf_Vernaux *) ((char *)vna + vna->vna_next); } if (vn->vn_next == 0) break; vn = (const Elf_Verneed *) ((char *)vn + vn->vn_next); } vd = obj->verdef; while (vd != NULL) { if (vd->vd_version != VER_DEF_CURRENT) { _rtld_error("%s: Unsupported version %d of Elf_Verdef entry", obj->path, vd->vd_version); return (-1); } vernum = VER_DEF_IDX(vd->vd_ndx); if (vernum > maxvernum) maxvernum = vernum; if (vd->vd_next == 0) break; vd = (const Elf_Verdef *) ((char *)vd + vd->vd_next); } if (maxvernum == 0) return (0); /* * Store version information in array indexable by version index. * Verify that object version requirements are satisfied along the * way. */ obj->vernum = maxvernum + 1; obj->vertab = xcalloc(obj->vernum, sizeof(Ver_Entry)); vd = obj->verdef; while (vd != NULL) { if ((vd->vd_flags & VER_FLG_BASE) == 0) { vernum = VER_DEF_IDX(vd->vd_ndx); assert(vernum <= maxvernum); vda = (const Elf_Verdaux *)((char *)vd + vd->vd_aux); obj->vertab[vernum].hash = vd->vd_hash; obj->vertab[vernum].name = obj->strtab + vda->vda_name; obj->vertab[vernum].file = NULL; obj->vertab[vernum].flags = 0; } if (vd->vd_next == 0) break; vd = (const Elf_Verdef *) ((char *)vd + vd->vd_next); } vn = obj->verneed; while (vn != NULL) { depobj = locate_dependency(obj, obj->strtab + vn->vn_file); if (depobj == NULL) return (-1); vna = (const Elf_Vernaux *) ((char *)vn + vn->vn_aux); for (;;) { if (check_object_provided_version(obj, depobj, vna)) return (-1); vernum = VER_NEED_IDX(vna->vna_other); assert(vernum <= maxvernum); obj->vertab[vernum].hash = vna->vna_hash; obj->vertab[vernum].name = obj->strtab + vna->vna_name; obj->vertab[vernum].file = obj->strtab + vn->vn_file; obj->vertab[vernum].flags = (vna->vna_other & VER_NEED_HIDDEN) ? VER_INFO_HIDDEN : 0; if (vna->vna_next == 0) break; vna = (const Elf_Vernaux *) ((char *)vna + vna->vna_next); } if (vn->vn_next == 0) break; vn = (const Elf_Verneed *) ((char *)vn + vn->vn_next); } return 0; } static int rtld_verify_versions(const Objlist *objlist) { Objlist_Entry *entry; int rc; rc = 0; STAILQ_FOREACH(entry, objlist, link) { /* * Skip dummy objects or objects that have their version requirements * already checked. */ if (entry->obj->strtab == NULL || entry->obj->vertab != NULL) continue; if (rtld_verify_object_versions(entry->obj) == -1) { rc = -1; if (ld_tracing == NULL) break; } } if (rc == 0 || ld_tracing != NULL) rc = rtld_verify_object_versions(&obj_rtld); return rc; } const Ver_Entry * fetch_ventry(const Obj_Entry *obj, unsigned long symnum) { Elf_Versym vernum; if (obj->vertab) { vernum = VER_NDX(obj->versyms[symnum]); if (vernum >= obj->vernum) { _rtld_error("%s: symbol %s has wrong verneed value %d", obj->path, obj->strtab + symnum, vernum); } else if (obj->vertab[vernum].hash != 0) { return &obj->vertab[vernum]; } } return NULL; } int _rtld_get_stack_prot(void) { return (stack_prot); } int _rtld_is_dlopened(void *arg) { Obj_Entry *obj; RtldLockState lockstate; int res; rlock_acquire(rtld_bind_lock, &lockstate); obj = dlcheck(arg); if (obj == NULL) obj = obj_from_addr(arg); if (obj == NULL) { _rtld_error("No shared object contains address"); lock_release(rtld_bind_lock, &lockstate); return (-1); } res = obj->dlopened ? 1 : 0; lock_release(rtld_bind_lock, &lockstate); return (res); } static void map_stacks_exec(RtldLockState *lockstate) { void (*thr_map_stacks_exec)(void); if ((max_stack_flags & PF_X) == 0 || (stack_prot & PROT_EXEC) != 0) return; thr_map_stacks_exec = (void (*)(void))(uintptr_t) get_program_var_addr("__pthread_map_stacks_exec", lockstate); if (thr_map_stacks_exec != NULL) { stack_prot |= PROT_EXEC; thr_map_stacks_exec(); } } void symlook_init(SymLook *dst, const char *name) { bzero(dst, sizeof(*dst)); dst->name = name; dst->hash = elf_hash(name); dst->hash_gnu = gnu_hash(name); } static void symlook_init_from_req(SymLook *dst, const SymLook *src) { dst->name = src->name; dst->hash = src->hash; dst->hash_gnu = src->hash_gnu; dst->ventry = src->ventry; dst->flags = src->flags; dst->defobj_out = NULL; dst->sym_out = NULL; dst->lockstate = src->lockstate; } /* * Parse a file descriptor number without pulling in more of libc (e.g. atoi). */ static int parse_libdir(const char *str) { static const int RADIX = 10; /* XXXJA: possibly support hex? */ const char *orig; int fd; char c; orig = str; fd = 0; for (c = *str; c != '\0'; c = *++str) { if (c < '0' || c > '9') return (-1); fd *= RADIX; fd += c - '0'; } /* Make sure we actually parsed something. */ if (str == orig) { _rtld_error("failed to parse directory FD from '%s'", str); return (-1); } return (fd); } /* * Overrides for libc_pic-provided functions. */ int __getosreldate(void) { size_t len; int oid[2]; int error, osrel; if (osreldate != 0) return (osreldate); oid[0] = CTL_KERN; oid[1] = KERN_OSRELDATE; osrel = 0; len = sizeof(osrel); error = sysctl(oid, 2, &osrel, &len, NULL, 0); if (error == 0 && osrel > 0 && len == sizeof(osrel)) osreldate = osrel; return (osreldate); } void exit(int status) { _exit(status); } void (*__cleanup)(void); int __isthreaded = 0; int _thread_autoinit_dummy_decl = 1; /* * No unresolved symbols for rtld. */ void __pthread_cxa_finalize(struct dl_phdr_info *a) { } void __stack_chk_fail(void) { _rtld_error("stack overflow detected; terminated"); rtld_die(); } __weak_reference(__stack_chk_fail, __stack_chk_fail_local); void __chk_fail(void) { _rtld_error("buffer overflow detected; terminated"); rtld_die(); } const char * rtld_strerror(int errnum) { if (errnum < 0 || errnum >= sys_nerr) return ("Unknown error"); return (sys_errlist[errnum]); } Index: projects/release-pkg/libexec/rtld-elf/rtld.h =================================================================== --- projects/release-pkg/libexec/rtld-elf/rtld.h (revision 296327) +++ projects/release-pkg/libexec/rtld-elf/rtld.h (revision 296328) @@ -1,401 +1,402 @@ /*- * Copyright 1996, 1997, 1998, 1999, 2000 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. * * $FreeBSD$ */ #ifndef RTLD_H /* { */ #define RTLD_H 1 #include #include #include #include #include #include #include #include #include "rtld_lock.h" #include "rtld_machdep.h" #define NEW(type) ((type *) xmalloc(sizeof(type))) #define CNEW(type) ((type *) xcalloc(1, sizeof(type))) /* We might as well do booleans like C++. */ typedef unsigned char bool; #define false 0 #define true 1 extern size_t tls_last_offset; extern size_t tls_last_size; extern size_t tls_static_space; extern int tls_dtv_generation; extern int tls_max_index; extern int npagesizes; extern size_t *pagesizes; extern int main_argc; extern char **main_argv; extern char **environ; struct stat; struct Struct_Obj_Entry; /* Lists of shared objects */ typedef struct Struct_Objlist_Entry { STAILQ_ENTRY(Struct_Objlist_Entry) link; struct Struct_Obj_Entry *obj; } Objlist_Entry; typedef STAILQ_HEAD(Struct_Objlist, Struct_Objlist_Entry) Objlist; /* Types of init and fini functions */ typedef void (*InitFunc)(void); typedef void (*InitArrFunc)(int, char **, char **); /* Lists of shared object dependencies */ typedef struct Struct_Needed_Entry { struct Struct_Needed_Entry *next; struct Struct_Obj_Entry *obj; unsigned long name; /* Offset of name in string table */ } Needed_Entry; typedef struct Struct_Name_Entry { STAILQ_ENTRY(Struct_Name_Entry) link; char name[1]; } Name_Entry; /* Lock object */ typedef struct Struct_LockInfo { void *context; /* Client context for creating locks */ void *thelock; /* The one big lock */ /* Debugging aids. */ volatile int rcount; /* Number of readers holding lock */ volatile int wcount; /* Number of writers holding lock */ /* Methods */ void *(*lock_create)(void *context); void (*rlock_acquire)(void *lock); void (*wlock_acquire)(void *lock); void (*rlock_release)(void *lock); void (*wlock_release)(void *lock); void (*lock_destroy)(void *lock); void (*context_destroy)(void *context); } LockInfo; typedef struct Struct_Ver_Entry { Elf_Word hash; unsigned int flags; const char *name; const char *file; } Ver_Entry; typedef struct Struct_Sym_Match_Result { const Elf_Sym *sym_out; const Elf_Sym *vsymp; int vcount; } Sym_Match_Result; #define VER_INFO_HIDDEN 0x01 /* * Shared object descriptor. * * Items marked with "(%)" are dynamically allocated, and must be freed * when the structure is destroyed. * * CAUTION: It appears that the JDK port peeks into these structures. * It looks at "next" and "mapbase" at least. Don't add new members * near the front, until this can be straightened out. */ typedef struct Struct_Obj_Entry { /* * These two items have to be set right for compatibility with the * original ElfKit crt1.o. */ Elf_Size magic; /* Magic number (sanity check) */ Elf_Size version; /* Version number of struct format */ TAILQ_ENTRY(Struct_Obj_Entry) next; char *path; /* Pathname of underlying file (%) */ char *origin_path; /* Directory path of origin file */ int refcount; int dl_refcount; /* Number of times loaded by dlopen */ /* These items are computed by map_object() or by digest_phdr(). */ caddr_t mapbase; /* Base address of mapped region */ size_t mapsize; /* Size of mapped region in bytes */ size_t textsize; /* Size of text segment in bytes */ Elf_Addr vaddrbase; /* Base address in shared object file */ caddr_t relocbase; /* Relocation constant = mapbase - vaddrbase */ const Elf_Dyn *dynamic; /* Dynamic section */ caddr_t entry; /* Entry point */ const Elf_Phdr *phdr; /* Program header if it is mapped, else NULL */ size_t phsize; /* Size of program header in bytes */ const char *interp; /* Pathname of the interpreter, if any */ Elf_Word stack_flags; /* TLS information */ int tlsindex; /* Index in DTV for this module */ void *tlsinit; /* Base address of TLS init block */ size_t tlsinitsize; /* Size of TLS init block for this module */ size_t tlssize; /* Size of TLS block for this module */ size_t tlsoffset; /* Offset of static TLS block for this module */ size_t tlsalign; /* Alignment of static TLS block */ caddr_t relro_page; size_t relro_size; /* Items from the dynamic section. */ Elf_Addr *pltgot; /* PLT or GOT, depending on architecture */ const Elf_Rel *rel; /* Relocation entries */ unsigned long relsize; /* Size in bytes of relocation info */ const Elf_Rela *rela; /* Relocation entries with addend */ unsigned long relasize; /* Size in bytes of addend relocation info */ const Elf_Rel *pltrel; /* PLT relocation entries */ unsigned long pltrelsize; /* Size in bytes of PLT relocation info */ const Elf_Rela *pltrela; /* PLT relocation entries with addend */ unsigned long pltrelasize; /* Size in bytes of PLT addend reloc info */ const Elf_Sym *symtab; /* Symbol table */ const char *strtab; /* String table */ unsigned long strsize; /* Size in bytes of string table */ #ifdef __mips__ Elf_Word local_gotno; /* Number of local GOT entries */ Elf_Word symtabno; /* Number of dynamic symbols */ Elf_Word gotsym; /* First dynamic symbol in GOT */ #endif #ifdef __powerpc64__ Elf_Addr glink; /* GLINK PLT call stub section */ #endif const Elf_Verneed *verneed; /* Required versions. */ Elf_Word verneednum; /* Number of entries in verneed table */ const Elf_Verdef *verdef; /* Provided versions. */ Elf_Word verdefnum; /* Number of entries in verdef table */ const Elf_Versym *versyms; /* Symbol versions table */ const Elf_Hashelt *buckets; /* Hash table buckets array */ unsigned long nbuckets; /* Number of buckets */ const Elf_Hashelt *chains; /* Hash table chain array */ unsigned long nchains; /* Number of entries in chain array */ Elf32_Word nbuckets_gnu; /* Number of GNU hash buckets*/ Elf32_Word symndx_gnu; /* 1st accessible symbol on dynsym table */ Elf32_Word maskwords_bm_gnu; /* Bloom filter words - 1 (bitmask) */ Elf32_Word shift2_gnu; /* Bloom filter shift count */ Elf32_Word dynsymcount; /* Total entries in dynsym table */ Elf_Addr *bloom_gnu; /* Bloom filter used by GNU hash func */ const Elf_Hashelt *buckets_gnu; /* GNU hash table bucket array */ const Elf_Hashelt *chain_zero_gnu; /* GNU hash table value array (Zeroed) */ char *rpath; /* Search path specified in object */ char *runpath; /* Search path with different priority */ Needed_Entry *needed; /* Shared objects needed by this one (%) */ Needed_Entry *needed_filtees; Needed_Entry *needed_aux_filtees; STAILQ_HEAD(, Struct_Name_Entry) names; /* List of names for this object we know about. */ Ver_Entry *vertab; /* Versions required /defined by this object */ int vernum; /* Number of entries in vertab */ Elf_Addr init; /* Initialization function to call */ Elf_Addr fini; /* Termination function to call */ Elf_Addr preinit_array; /* Pre-initialization array of functions */ Elf_Addr init_array; /* Initialization array of functions */ Elf_Addr fini_array; /* Termination array of functions */ int preinit_array_num; /* Number of entries in preinit_array */ int init_array_num; /* Number of entries in init_array */ int fini_array_num; /* Number of entries in fini_array */ int32_t osrel; /* OSREL note value */ bool mainprog : 1; /* True if this is the main program */ bool rtld : 1; /* True if this is the dynamic linker */ bool relocated : 1; /* True if processed by relocate_objects() */ bool ver_checked : 1; /* True if processed by rtld_verify_object_versions */ bool textrel : 1; /* True if there are relocations to text seg */ bool symbolic : 1; /* True if generated with "-Bsymbolic" */ bool bind_now : 1; /* True if all relocations should be made first */ bool traced : 1; /* Already printed in ldd trace output */ bool jmpslots_done : 1; /* Already have relocated the jump slots */ bool init_done : 1; /* Already have added object to init list */ bool tls_done : 1; /* Already allocated offset for static TLS */ bool phdr_alloc : 1; /* Phdr is allocated and needs to be freed. */ bool z_origin : 1; /* Process rpath and soname tokens */ bool z_nodelete : 1; /* Do not unload the object and dependencies */ bool z_noopen : 1; /* Do not load on dlopen */ bool z_loadfltr : 1; /* Immediately load filtees */ bool z_interpose : 1; /* Interpose all objects but main */ bool z_nodeflib : 1; /* Don't search default library path */ bool z_global : 1; /* Make the object global */ bool ref_nodel : 1; /* Refcount increased to prevent dlclose */ bool init_scanned: 1; /* Object is already on init list. */ bool on_fini_list: 1; /* Object is already on fini list. */ bool dag_inited : 1; /* Object has its DAG initialized. */ bool filtees_loaded : 1; /* Filtees loaded */ bool irelative : 1; /* Object has R_MACHDEP_IRELATIVE relocs */ bool gnu_ifunc : 1; /* Object has references to STT_GNU_IFUNC */ bool non_plt_gnu_ifunc : 1; /* Object has non-plt IFUNC references */ bool crt_no_init : 1; /* Object' crt does not call _init/_fini */ bool valid_hash_sysv : 1; /* A valid System V hash hash tag is available */ bool valid_hash_gnu : 1; /* A valid GNU hash tag is available */ bool dlopened : 1; /* dlopen()-ed (vs. load statically) */ bool marker : 1; /* marker on the global obj list */ struct link_map linkmap; /* For GDB and dlinfo() */ Objlist dldags; /* Object belongs to these dlopened DAGs (%) */ Objlist dagmembers; /* DAG has these members (%) */ dev_t dev; /* Object's filesystem's device */ ino_t ino; /* Object's inode number */ void *priv; /* Platform-dependent */ } Obj_Entry; #define RTLD_MAGIC 0xd550b87a #define RTLD_VERSION 1 TAILQ_HEAD(obj_entry_q, Struct_Obj_Entry); #define RTLD_STATIC_TLS_EXTRA 128 /* Flags to be passed into symlook_ family of functions. */ #define SYMLOOK_IN_PLT 0x01 /* Lookup for PLT symbol */ #define SYMLOOK_DLSYM 0x02 /* Return newest versioned symbol. Used by dlsym. */ #define SYMLOOK_EARLY 0x04 /* Symlook is done during initialization. */ #define SYMLOOK_IFUNC 0x08 /* Allow IFUNC processing in reloc_non_plt(). */ /* Flags for load_object(). */ #define RTLD_LO_NOLOAD 0x01 /* dlopen() specified RTLD_NOLOAD. */ #define RTLD_LO_DLOPEN 0x02 /* Load_object() called from dlopen(). */ #define RTLD_LO_TRACE 0x04 /* Only tracing. */ #define RTLD_LO_NODELETE 0x08 /* Loaded object cannot be closed. */ #define RTLD_LO_FILTEES 0x10 /* Loading filtee. */ #define RTLD_LO_EARLY 0x20 /* Do not call ctors, postpone it to the initialization during the image start. */ /* * Symbol cache entry used during relocation to avoid multiple lookups * of the same symbol. */ typedef struct Struct_SymCache { const Elf_Sym *sym; /* Symbol table entry */ const Obj_Entry *obj; /* Shared object which defines it */ } SymCache; /* * This structure provides a reentrant way to keep a list of objects and * check which ones have already been processed in some way. */ typedef struct Struct_DoneList { const Obj_Entry **objs; /* Array of object pointers */ unsigned int num_alloc; /* Allocated size of the array */ unsigned int num_used; /* Number of array slots used */ } DoneList; struct Struct_RtldLockState { int lockstate; sigjmp_buf env; }; struct fill_search_info_args { int request; unsigned int flags; struct dl_serinfo *serinfo; struct dl_serpath *serpath; char *strspace; }; /* * The pack of arguments and results for the symbol lookup functions. */ typedef struct Struct_SymLook { const char *name; unsigned long hash; uint32_t hash_gnu; const Ver_Entry *ventry; int flags; const Obj_Entry *defobj_out; const Elf_Sym *sym_out; struct Struct_RtldLockState *lockstate; } SymLook; void _rtld_error(const char *, ...) __printflike(1, 2) __exported; void rtld_die(void) __dead2; const char *rtld_strerror(int); Obj_Entry *map_object(int, const char *, const struct stat *); void *xcalloc(size_t, size_t); void *xmalloc(size_t); char *xstrdup(const char *); void *malloc_aligned(size_t size, size_t align); void free_aligned(void *ptr); extern Elf_Addr _GLOBAL_OFFSET_TABLE_[]; extern Elf_Sym sym_zero; /* For resolving undefined weak refs. */ void dump_relocations(Obj_Entry *); void dump_obj_relocations(Obj_Entry *); void dump_Elf_Rel(Obj_Entry *, const Elf_Rel *, u_long); void dump_Elf_Rela(Obj_Entry *, const Elf_Rela *, u_long); /* * Function declarations. */ unsigned long elf_hash(const char *); const Elf_Sym *find_symdef(unsigned long, const Obj_Entry *, const Obj_Entry **, int, SymCache *, struct Struct_RtldLockState *); void init_pltgot(Obj_Entry *); void lockdflt_init(void); void digest_notes(Obj_Entry *, Elf_Addr, Elf_Addr); Obj_Entry *globallist_curr(const Obj_Entry *obj); Obj_Entry *globallist_next(const Obj_Entry *obj); void obj_free(Obj_Entry *); Obj_Entry *obj_new(void); void _rtld_bind_start(void); void *rtld_resolve_ifunc(const Obj_Entry *obj, const Elf_Sym *def); void symlook_init(SymLook *, const char *); int symlook_obj(SymLook *, const Obj_Entry *); void *tls_get_addr_common(Elf_Addr** dtvp, int index, size_t offset); void *allocate_tls(Obj_Entry *, void *, size_t, size_t); void free_tls(void *, size_t, size_t); void *allocate_module_tls(int index); bool allocate_tls_offset(Obj_Entry *obj); void free_tls_offset(Obj_Entry *obj); const Ver_Entry *fetch_ventry(const Obj_Entry *obj, unsigned long); +int convert_prot(int elfflags); /* * MD function declarations. */ int do_copy_relocations(Obj_Entry *); int reloc_non_plt(Obj_Entry *, Obj_Entry *, int flags, struct Struct_RtldLockState *); int reloc_plt(Obj_Entry *); int reloc_jmpslots(Obj_Entry *, int flags, struct Struct_RtldLockState *); int reloc_iresolve(Obj_Entry *, struct Struct_RtldLockState *); int reloc_gnu_ifunc(Obj_Entry *, int flags, struct Struct_RtldLockState *); void allocate_initial_tls(Obj_Entry *); #endif /* } */ Index: projects/release-pkg/share/man/man9/make_dev.9 =================================================================== --- projects/release-pkg/share/man/man9/make_dev.9 (revision 296327) +++ projects/release-pkg/share/man/man9/make_dev.9 (revision 296328) @@ -1,493 +1,492 @@ .\" Copyright (c) 1999 Chris Costello .\" All rights reserved. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .\" -.Dd Jan 3, 2016 +.Dd March 2, 2016 .Dt MAKE_DEV 9 .Os .Sh NAME .Nm make_dev , .Nm make_dev_cred , .Nm make_dev_credf , .Nm make_dev_p , .Nm make_dev_s , .Nm make_dev_alias , .Nm make_dev_alias_p , .Nm destroy_dev , .Nm destroy_dev_sched , .Nm destroy_dev_sched_cb , .Nm destroy_dev_drain , .Nm dev_depends .Nd manage .Vt cdev Ns 's and DEVFS registration for devices .Sh SYNOPSIS .In sys/param.h .In sys/conf.h .Ft void .Fn make_dev_args_init "struct make_dev_args *args" .Ft int .Fn make_dev_s "struct make_dev_args *args" "struct cdev **cdev" "const char *fmt" ... .Ft int .Fn make_dev_alias_p "int flags" "struct cdev **cdev" "struct cdev *pdev" "const char *fmt" ... .Ft void .Fn destroy_dev "struct cdev *dev" .Ft void .Fn destroy_dev_sched "struct cdev *dev" .Ft void .Fn destroy_dev_sched_cb "struct cdev *dev" "void (*cb)(void *)" "void *arg" .Ft void .Fn destroy_dev_drain "struct cdevsw *csw" .Ft void .Fn dev_depends "struct cdev *pdev" "struct cdev *cdev" .Pp LEGACY INTERFACES .Ft struct cdev * .Fn make_dev "struct cdevsw *cdevsw" "int unit" "uid_t uid" "gid_t gid" "int perms" "const char *fmt" ... .Ft struct cdev * .Fn make_dev_cred "struct cdevsw *cdevsw" "int unit" "struct ucred *cr" "uid_t uid" "gid_t gid" "int perms" "const char *fmt" ... .Ft struct cdev * .Fn make_dev_credf "int flags" "struct cdevsw *cdevsw" "int unit" "struct ucred *cr" "uid_t uid" "gid_t gid" "int perms" "const char *fmt" ... .Ft int .Fn make_dev_p "int flags" "struct cdev **cdev" "struct cdevsw *devsw" "struct ucred *cr" "uid_t uid" "gid_t gid" "int mode" "const char *fmt" ... .Ft struct cdev * .Fn make_dev_alias "struct cdev *pdev" "const char *fmt" ... .Sh DESCRIPTION The .Fn make_dev_s function creates a .Fa cdev structure for a new device, which is returned into the .Fa cdev argument. It also notifies .Xr devfs 5 of the presence of the new device, that causes corresponding nodes to be created. Besides this, a .Xr devctl 4 notification is sent. The function takes the structure .Va struct make_dev_args args , which specifies the parameters for the device creation: .Pp .Bd -literal -offset indent -compact struct make_dev_args { size_t mda_size; int mda_flags; struct cdevsw *mda_devsw; struct ucred *mda_cr; uid_t mda_uid; gid_t mda_gid; int mda_mode; int mda_unit; void *mda_si_drv1; void *mda_si_drv2; }; .Ed Before use and filling with the desired values, the structure must be initialized by the .Fn make_dev_args_init function, which ensures that future kernel interface expansion does not affect driver source code or binary interface. .Pp The created device will be owned by .Va args.mda_uid , with the group ownership as .Va args.mda_gid . The name is the expansion of .Va fmt and following arguments as .Xr printf 9 would print it. The name determines its path under .Pa /dev or other .Xr devfs 5 mount point and may contain slash .Ql / characters to denote subdirectories. The permissions of the file specified in .Va args.mda_mode are defined in .In sys/stat.h : .Pp .Bd -literal -offset indent -compact #define S_IRWXU 0000700 /* RWX mask for owner */ #define S_IRUSR 0000400 /* R for owner */ #define S_IWUSR 0000200 /* W for owner */ #define S_IXUSR 0000100 /* X for owner */ #define S_IRWXG 0000070 /* RWX mask for group */ #define S_IRGRP 0000040 /* R for group */ #define S_IWGRP 0000020 /* W for group */ #define S_IXGRP 0000010 /* X for group */ #define S_IRWXO 0000007 /* RWX mask for other */ #define S_IROTH 0000004 /* R for other */ #define S_IWOTH 0000002 /* W for other */ #define S_IXOTH 0000001 /* X for other */ #define S_ISUID 0004000 /* set user id on execution */ #define S_ISGID 0002000 /* set group id on execution */ #define S_ISVTX 0001000 /* sticky bit */ #ifndef _POSIX_SOURCE #define S_ISTXT 0001000 #endif .Ed .Pp The .Va args.mda_cr argument specifies credentials that will be stored in the .Fa si_cred member of the initialized .Fa struct cdev . .Pp The .Va args.mda_flags argument alters the operation of .Fn make_dev_s. The following values are currently accepted: .Pp .Bl -tag -width "It Dv MAKEDEV_CHECKNAME" -compact -offset indent .It Dv MAKEDEV_REF reference the created device .It Dv MAKEDEV_NOWAIT do not sleep, the call may fail .It Dv MAKEDEV_WAITOK allow the function to sleep to satisfy malloc .It Dv MAKEDEV_ETERNAL created device will be never destroyed .It Dv MAKEDEV_CHECKNAME return an error if the device name is invalid or already exists .El .Pp Only .Dv MAKEDEV_NOWAIT , .Dv MAKEDEV_WAITOK and .Dv MAKEDEV_CHECKNAME values are accepted for the .Fn make_dev_alias_p function. .Pp The .Dv MAKEDEV_WAITOK flag is assumed if none of .Dv MAKEDEV_WAITOK , .Dv MAKEDEV_NOWAIT is specified. .Pp The .Xr dev_clone 9 event handler shall specify .Dv MAKEDEV_REF flag when creating a device in response to lookup, to avoid race where the device created is destroyed immediately after .Xr devfs_lookup 9 drops his reference to cdev. .Pp The .Dv MAKEDEV_ETERNAL flag allows the kernel to not acquire some locks when translating system calls into the cdevsw methods calls. It is responsibility of the driver author to make sure that .Fn destroy_dev is never called on the returned cdev. For the convenience, use the .Dv MAKEDEV_ETERNAL_KLD flag for the code that can be compiled into kernel or loaded (and unloaded) as loadable module. .Pp A panic will occur if the .Dv MAKEDEV_CHECKNAME flag is not specified and the device name is invalid or already exists. .Pp The .Fn make_dev_p use of the form .Bd -literal -offset indent struct cdev *dev; int res; res = make_dev_p(flags, &dev, cdevsw, cred, uid, gid, perms, name); .Ed is equivalent to the code .Bd -literal -offset indent struct cdev *dev; struct make_dev_args args; int res; make_dev_args_init(&args); args.mda_flags = flags; args.mda_devsw = cdevsw; args.mda_cred = cred; args.mda_uid = uid; args.mda_gid = gid; args.mda_mode = perms; res = make_dev_s(&args, &dev, name); .Ed .Pp Similarly, the .Fn make_dev_credf function call is equivalent to .Bd -literal -offset indent (void) make_dev_s(&args, &dev, name); .Ed In other words, .Fn make_dev_credf does not allow the caller to obtain the return value, and in kernels compiled with the .Va INVARIANTS options, the function asserts that the device creation succeeded. .Pp The .Fn make_dev_cred function is equivalent to the call .Bd -literal -offset indent make_dev_credf(0, cdevsw, unit, cr, uid, gid, perms, fmt, ...); .Ed .Pp The .Fn make_dev function call is the same as .Bd -literal -offset indent make_dev_credf(0, cdevsw, unit, NULL, uid, gid, perms, fmt, ...); .Ed .Pp The .Fn make_dev_alias_p function takes the returned .Ft cdev from .Fn make_dev and makes another (aliased) name for this device. It is an error to call .Fn make_dev_alias_p prior to calling .Fn make_dev . .Pp The .Fn make_dev_alias function is similar to .Fn make_dev_alias but it returns the resulting aliasing .Ft *cdev and may not return an error. .Pp The .Fa cdev returned by .Fn make_dev_s and .Fn make_dev_alias_p has two fields, .Fa si_drv1 and .Fa si_drv2 , that are available to store state. Both fields are of type .Ft void * , and can be initialized simultaneously with the .Va cdev allocation by filling .Va args.mda_si_drv1 and .Va args.mda_si_drv2 members of the .Fn make_dev_s argument structure, or filled after the .Va cdev is allocated, if using legacy interfaces. In the latter case, the driver should handle the race of accessing uninitialized .Va si_drv1 and .Va si_drv2 itself. These are designed to replace the .Fa unit argument to .Fn make_dev , which can be obtained with .Fn dev2unit . .Pp The .Fn destroy_dev function takes the returned .Fa cdev from .Fn make_dev and destroys the registration for that device. The notification is sent to .Xr devctl 4 about the destruction event. Do not call .Fn destroy_dev on devices that were created with .Fn make_dev_alias . .Pp The .Fn dev_depends function establishes a parent-child relationship between two devices. The net effect is that a .Fn destroy_dev of the parent device will also result in the destruction of the child device(s), if any exist. A device may simultaneously be a parent and a child, so it is possible to build a complete hierarchy. .Pp The .Fn destroy_dev_sched_cb function schedules execution of the .Fn destroy_dev for the specified .Fa cdev in the safe context. After .Fn destroy_dev is finished, and if the supplied .Fa cb is not .Dv NULL , the callback .Fa cb is called, with argument .Fa arg . The .Fn destroy_dev_sched function is the same as .Bd -literal -offset indent destroy_dev_sched_cb(cdev, NULL, NULL); .Ed .Pp The .Fn d_close driver method cannot call .Fn destroy_dev directly. Doing so causes deadlock when .Fn destroy_dev waits for all threads to leave the driver methods. Also, because .Fn destroy_dev sleeps, no non-sleepable locks may be held over the call. The .Fn destroy_dev_sched family of functions overcome these issues. .Pp The device driver may call the .Fn destroy_dev_drain function to wait until all devices that have supplied .Fa csw as cdevsw, are destroyed. This is useful when driver knows that .Fn destroy_dev_sched is called for all instantiated devices, but need to postpone module unload until .Fn destroy_dev is actually finished for all of them. .Sh RETURN VALUES If successful, .Fn make_dev_s and .Fn make_dev_p will return 0, otherwise they will return an error. If successful, .Fn make_dev_credf will return a valid .Fa cdev pointer, otherwise it will return .Dv NULL . .Sh ERRORS The .Fn make_dev_s , .Fn make_dev_p and .Fn make_dev_alias_p calls will fail and the device will be not registered if: .Bl -tag -width Er .It Bq Er ENOMEM The .Dv MAKEDEV_NOWAIT flag was specified and a memory allocation request could not be satisfied. .It Bq Er ENAMETOOLONG The .Dv MAKEDEV_CHECKNAME flag was specified and the provided device name is longer than .Dv SPECNAMELEN . .It Bq Er EINVAL The .Dv MAKEDEV_CHECKNAME flag was specified and the provided device name is empty, contains a .Qq \&. or .Qq .. path component or ends with .Ql / . .It Bq Er EINVAL The .Dv MAKEDEV_CHECKNAME flag was specified and the provided device name contains invalid characters. .It Bq Er EEXIST The .Dv MAKEDEV_CHECKNAME flag was specified and the provided device name already exists. .El .Sh SEE ALSO .Xr devctl 4 , .Xr devfs 5 , -.Xr destroy_dev_drain 9 , .Xr dev_clone 9 .Sh HISTORY The .Fn make_dev and .Fn destroy_dev functions first appeared in .Fx 4.0 . The function .Fn make_dev_alias first appeared in .Fx 4.1 . The function .Fn dev_depends first appeared in .Fx 5.0 . The functions .Fn make_dev_credf , .Fn destroy_dev_sched , .Fn destroy_dev_sched_cb first appeared in .Fx 7.0 . The function .Fn make_dev_p first appeared in .Fx 8.2 . The function .Fn make_dev_s first appeared in .Fx 11.0 . Index: projects/release-pkg/share/mk/bsd.lib.mk =================================================================== --- projects/release-pkg/share/mk/bsd.lib.mk (revision 296327) +++ projects/release-pkg/share/mk/bsd.lib.mk (revision 296328) @@ -1,467 +1,472 @@ # from: @(#)bsd.lib.mk 5.26 (Berkeley) 5/2/91 # $FreeBSD$ # .include .if defined(LIB_CXX) LIB= ${LIB_CXX} _LD= ${CXX} .else _LD= ${CC} .endif LIB_PRIVATE= ${PRIVATELIB:Dprivate} # Set up the variables controlling shared libraries. After this section, # SHLIB_NAME will be defined only if we are to create a shared library. # SHLIB_LINK will be defined only if we are to create a link to it. # INSTALL_PIC_ARCHIVE will be defined only if we are to create a PIC archive. .if defined(NO_PIC) .undef SHLIB_NAME .undef INSTALL_PIC_ARCHIVE .else .if !defined(SHLIB) && defined(LIB) SHLIB= ${LIB} .endif .if !defined(SHLIB_NAME) && defined(SHLIB) && defined(SHLIB_MAJOR) SHLIB_NAME= lib${LIB_PRIVATE}${SHLIB}.so.${SHLIB_MAJOR} .endif .if defined(SHLIB_NAME) && !empty(SHLIB_NAME:M*.so.*) SHLIB_LINK?= ${SHLIB_NAME:R} .endif SONAME?= ${SHLIB_NAME} .endif .if defined(CRUNCH_CFLAGS) CFLAGS+= ${CRUNCH_CFLAGS} .endif .if ${MK_ASSERT_DEBUG} == "no" CFLAGS+= -DNDEBUG NO_WERROR= .endif .if defined(DEBUG_FLAGS) CFLAGS+= ${DEBUG_FLAGS} .if ${MK_CTF} != "no" && ${DEBUG_FLAGS:M-g} != "" CTFFLAGS+= -g .endif .else STRIP?= -s .endif .if ${SHLIBDIR:M*lib32*} TAGS+= lib32 .endif .if defined(NO_ROOT) .if !defined(TAGS) || ! ${TAGS:Mpackage=*} TAGS+= package=${PACKAGE:Uruntime} .endif TAG_ARGS= -T ${TAGS:[*]:S/ /,/g} .endif .if ${MK_DEBUG_FILES} != "no" && empty(DEBUG_FLAGS:M-g) && \ empty(DEBUG_FLAGS:M-gdwarf*) SHARED_CFLAGS+= -g SHARED_CXXFLAGS+= -g CTFFLAGS+= -g .endif .include # prefer .s to a .c, add .po, remove stuff not used in the BSD libraries # .So used for PIC object files .SUFFIXES: .SUFFIXES: .out .o .po .So .S .asm .s .c .cc .cpp .cxx .C .f .y .l .ln .if !defined(PICFLAG) .if ${MACHINE_CPUARCH} == "sparc64" PICFLAG=-fPIC .else PICFLAG=-fpic .endif .endif PO_FLAG=-pg .c.o: ${CC} ${STATIC_CFLAGS} ${CFLAGS} -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .c.po: ${CC} ${PO_FLAG} ${STATIC_CFLAGS} ${PO_CFLAGS} -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .c.So: ${CC} ${PICFLAG} -DPIC ${SHARED_CFLAGS} ${CFLAGS} -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .cc.o .C.o .cpp.o .cxx.o: ${CXX} ${STATIC_CXXFLAGS} ${CXXFLAGS} -c ${.IMPSRC} -o ${.TARGET} .cc.po .C.po .cpp.po .cxx.po: ${CXX} ${PO_FLAG} ${STATIC_CXXFLAGS} ${PO_CXXFLAGS} -c ${.IMPSRC} -o ${.TARGET} .cc.So .C.So .cpp.So .cxx.So: ${CXX} ${PICFLAG} -DPIC ${SHARED_CXXFLAGS} ${CXXFLAGS} -c ${.IMPSRC} -o ${.TARGET} .f.po: ${FC} -pg ${FFLAGS} -o ${.TARGET} -c ${.IMPSRC} ${CTFCONVERT_CMD} .f.So: ${FC} ${PICFLAG} -DPIC ${FFLAGS} -o ${.TARGET} -c ${.IMPSRC} ${CTFCONVERT_CMD} .s.po .s.So: ${AS} ${AFLAGS} -o ${.TARGET} ${.IMPSRC} ${CTFCONVERT_CMD} .asm.po: ${CC:N${CCACHE_BIN}} -x assembler-with-cpp -DPROF ${PO_CFLAGS} \ ${ACFLAGS} -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .asm.So: ${CC:N${CCACHE_BIN}} -x assembler-with-cpp ${PICFLAG} -DPIC \ ${CFLAGS} ${ACFLAGS} -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} .S.po: ${CC:N${CCACHE_BIN}} -DPROF ${PO_CFLAGS} ${ACFLAGS} -c ${.IMPSRC} \ -o ${.TARGET} ${CTFCONVERT_CMD} .S.So: ${CC:N${CCACHE_BIN}} ${PICFLAG} -DPIC ${CFLAGS} ${ACFLAGS} \ -c ${.IMPSRC} -o ${.TARGET} ${CTFCONVERT_CMD} _LIBDIR:=${LIBDIR} _SHLIBDIR:=${SHLIBDIR} .if defined(SHLIB_NAME) .if ${MK_DEBUG_FILES} != "no" SHLIB_NAME_FULL=${SHLIB_NAME}.full # Use ${DEBUGDIR} for base system debug files, else .debug subdirectory .if ${_SHLIBDIR} == "/boot" ||\ ${SHLIBDIR:C%/lib(/.*)?$%/lib%} == "/lib" ||\ ${SHLIBDIR:C%/usr/(tests/)?lib(32|exec)?(/.*)?%/usr/lib%} == "/usr/lib" DEBUGFILEDIR=${DEBUGDIR}${_SHLIBDIR} .else DEBUGFILEDIR=${_SHLIBDIR}/.debug DEBUGMKDIR= .endif .else SHLIB_NAME_FULL=${SHLIB_NAME} .endif .endif .include # Allow libraries to specify their own version map or have it # automatically generated (see bsd.symver.mk above). .if ${MK_SYMVER} == "yes" && !empty(VERSION_MAP) ${SHLIB_NAME_FULL}: ${VERSION_MAP} LDFLAGS+= -Wl,--version-script=${VERSION_MAP} .endif .if defined(LIB) && !empty(LIB) || defined(SHLIB_NAME) OBJS+= ${SRCS:N*.h:R:S/$/.o/} CLEANFILES+= ${OBJS} ${STATICOBJS} .endif .if defined(LIB) && !empty(LIB) _LIBS= lib${LIB_PRIVATE}${LIB}.a lib${LIB_PRIVATE}${LIB}.a: ${OBJS} ${STATICOBJS} @${ECHO} building static ${LIB} library @rm -f ${.TARGET} ${AR} ${ARFLAGS} ${.TARGET} `NM='${NM}' NMFLAGS='${NMFLAGS}' lorder ${OBJS} ${STATICOBJS} | tsort -q` ${ARADD} ${RANLIB} ${RANLIBFLAGS} ${.TARGET} .endif .if !defined(INTERNALLIB) .if ${MK_PROFILE} != "no" && defined(LIB) && !empty(LIB) _LIBS+= lib${LIB_PRIVATE}${LIB}_p.a POBJS+= ${OBJS:.o=.po} ${STATICOBJS:.o=.po} DEPENDOBJS+= ${POBJS} CLEANFILES+= ${POBJS} lib${LIB_PRIVATE}${LIB}_p.a: ${POBJS} @${ECHO} building profiled ${LIB} library @rm -f ${.TARGET} ${AR} ${ARFLAGS} ${.TARGET} `NM='${NM}' NMFLAGS='${NMFLAGS}' lorder ${POBJS} | tsort -q` ${ARADD} ${RANLIB} ${RANLIBFLAGS} ${.TARGET} .endif .if defined(SHLIB_NAME) || \ defined(INSTALL_PIC_ARCHIVE) && defined(LIB) && !empty(LIB) SOBJS+= ${OBJS:.o=.So} DEPENDOBJS+= ${SOBJS} CLEANFILES+= ${SOBJS} .endif .if defined(SHLIB_NAME) _LIBS+= ${SHLIB_NAME} SOLINKOPTS= -shared -Wl,-x .if !defined(ALLOW_SHARED_TEXTREL) .if defined(LD_FATAL_WARNINGS) && ${LD_FATAL_WARNINGS} == "no" SOLINKOPTS+= -Wl,--no-fatal-warnings .else SOLINKOPTS+= -Wl,--fatal-warnings .endif SOLINKOPTS+= -Wl,--warn-shared-textrel .endif .if target(beforelinking) beforelinking: ${SOBJS} ${SHLIB_NAME_FULL}: beforelinking .endif .if defined(SHLIB_LINK) # ${_LDSCRIPTROOT} is needed when cross-building # and when building 32 bits library shims. # # ${_LDSCRIPTROOT} is the directory prefix that will be used when generating # ld(1) scripts. The crosstools' ld is configured to lookup libraries in an # alternative directory which is called "sysroot", so during buildworld binaries # won't be linked against the running system libraries but against the ones of # the current source tree. ${_LDSCRIPTROOT} behavior is twisted because of # the location where we store them: # - 64 bits libs are located under sysroot, so ${_LDSCRIPTROOT} must be empty # because ld(1) will manage to find them from sysroot; # - 32 bits shims are not, so ${_LDSCRIPTROOT} is used to specify their full # path, outside of sysroot. # Note that ld(1) scripts are generated both during buildworld and # installworld; in the later case ${_LDSCRIPTROOT} must be obviously empty # because on the target system, libraries are meant to be looked up from /. .if defined(SHLIB_LDSCRIPT) && !empty(SHLIB_LDSCRIPT) && exists(${.CURDIR}/${SHLIB_LDSCRIPT}) ${SHLIB_LINK:R}.ld: ${.CURDIR}/${SHLIB_LDSCRIPT} sed -e 's,@@SHLIB@@,${_LDSCRIPTROOT}${_SHLIBDIR}/${SHLIB_NAME},g' \ -e 's,@@LIBDIR@@,${_LDSCRIPTROOT}${_LIBDIR},g' \ -e 's,/[^ ]*/,,g' \ ${.ALLSRC} > ${.TARGET} ${SHLIB_NAME_FULL}: ${SHLIB_LINK:R}.ld CLEANFILES+= ${SHLIB_LINK:R}.ld .endif CLEANFILES+= ${SHLIB_LINK} .endif ${SHLIB_NAME_FULL}: ${SOBJS} @${ECHO} building shared library ${SHLIB_NAME} @rm -f ${SHLIB_NAME} ${SHLIB_LINK} .if defined(SHLIB_LINK) && !commands(${SHLIB_LINK:R}.ld) @${INSTALL_SYMLINK} ${TAG_ARGS:D${TAG_ARGS},development} ${SHLIB_NAME} ${SHLIB_LINK} .endif ${_LD:N${CCACHE_BIN}} ${LDFLAGS} ${SSP_CFLAGS} ${SOLINKOPTS} \ -o ${.TARGET} -Wl,-soname,${SONAME} \ `NM='${NM}' NMFLAGS='${NMFLAGS}' lorder ${SOBJS} | tsort -q` ${LDADD} .if ${MK_CTF} != "no" ${CTFMERGE} ${CTFFLAGS} -o ${.TARGET} ${SOBJS} .endif .if ${MK_DEBUG_FILES} != "no" CLEANFILES+= ${SHLIB_NAME_FULL} ${SHLIB_NAME}.debug ${SHLIB_NAME}: ${SHLIB_NAME_FULL} ${SHLIB_NAME}.debug ${OBJCOPY} --strip-debug --add-gnu-debuglink=${SHLIB_NAME}.debug \ ${SHLIB_NAME_FULL} ${.TARGET} ${SHLIB_NAME}.debug: ${SHLIB_NAME_FULL} ${OBJCOPY} --only-keep-debug ${SHLIB_NAME_FULL} ${.TARGET} .endif .endif #defined(SHLIB_NAME) .if defined(INSTALL_PIC_ARCHIVE) && defined(LIB) && !empty(LIB) && ${MK_TOOLCHAIN} != "no" _LIBS+= lib${LIB_PRIVATE}${LIB}_pic.a lib${LIB_PRIVATE}${LIB}_pic.a: ${SOBJS} @${ECHO} building special pic ${LIB} library @rm -f ${.TARGET} ${AR} ${ARFLAGS} ${.TARGET} ${SOBJS} ${ARADD} ${RANLIB} ${RANLIBFLAGS} ${.TARGET} .endif .if defined(WANT_LINT) && !defined(NO_LINT) && defined(LIB) && !empty(LIB) LINTLIB= llib-l${LIB}.ln _LIBS+= ${LINTLIB} LINTOBJS+= ${SRCS:M*.c:.c=.ln} CLEANFILES+= ${LINTOBJS} ${LINTLIB}: ${LINTOBJS} @${ECHO} building lint library ${.TARGET} @rm -f ${.TARGET} ${LINT} ${LINTLIBFLAGS} ${CFLAGS:M-[DIU]*} ${.ALLSRC} .endif .endif # !defined(INTERNALLIB) .if defined(_SKIP_BUILD) all: .else .if defined(_LIBS) && !empty(_LIBS) all: ${_LIBS} CLEANFILES+= ${_LIBS} .endif .if ${MK_MAN} != "no" && !defined(LIBRARIES_ONLY) all: all-man .endif .endif _EXTRADEPEND: .if ${MK_FAST_DEPEND} == "no" @TMP=_depend$$$$; \ sed -e 's/^\([^\.]*\).o[ ]*:/\1.o \1.po \1.So:/' < ${DEPENDFILE} \ > $$TMP; \ mv $$TMP ${DEPENDFILE} .endif .if !defined(NO_EXTRADEPEND) && defined(SHLIB_NAME) .if defined(DPADD) && !empty(DPADD) echo ${SHLIB_NAME_FULL}: ${DPADD} >> ${DEPENDFILE} .endif .endif .if !target(install) .if defined(PRECIOUSLIB) .if !defined(NO_FSCHG) SHLINSTALLFLAGS+= -fschg .endif SHLINSTALLFLAGS+= -S .endif _INSTALLFLAGS:= ${INSTALLFLAGS} .for ie in ${INSTALLFLAGS_EDIT} _INSTALLFLAGS:= ${_INSTALLFLAGS${ie}} .endfor _SHLINSTALLFLAGS:= ${SHLINSTALLFLAGS} .for ie in ${INSTALLFLAGS_EDIT} _SHLINSTALLFLAGS:= ${_SHLINSTALLFLAGS${ie}} .endfor .if !defined(INTERNALLIB) realinstall: _libinstall .ORDER: beforeinstall _libinstall _libinstall: .if defined(LIB) && !empty(LIB) && ${MK_INSTALLLIB} != "no" ${INSTALL} ${TAG_ARGS:D${TAG_ARGS},development} -C -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} lib${LIB_PRIVATE}${LIB}.a ${DESTDIR}${_LIBDIR}/ .endif .if ${MK_PROFILE} != "no" && defined(LIB) && !empty(LIB) ${INSTALL} ${TAG_ARGS:D${TAG_ARGS},profile} -C -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} lib${LIB_PRIVATE}${LIB}_p.a ${DESTDIR}${_LIBDIR}/ .endif .if defined(SHLIB_NAME) ${INSTALL} ${TAG_ARGS} ${STRIP} -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} ${_SHLINSTALLFLAGS} \ ${SHLIB_NAME} ${DESTDIR}${_SHLIBDIR}/ .if ${MK_DEBUG_FILES} != "no" .if defined(DEBUGMKDIR) ${INSTALL} ${TAG_ARGS:D${TAG_ARGS},debug} -d ${DESTDIR}${DEBUGFILEDIR}/ .endif ${INSTALL} ${TAG_ARGS:D${TAG_ARGS},debug} -o ${LIBOWN} -g ${LIBGRP} -m ${DEBUGMODE} \ ${_INSTALLFLAGS} \ ${SHLIB_NAME}.debug ${DESTDIR}${DEBUGFILEDIR}/ .endif .if defined(SHLIB_LINK) .if commands(${SHLIB_LINK:R}.ld) ${INSTALL} ${TAG_ARGS:D${TAG_ARGS},development} -S -C -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} ${SHLIB_LINK:R}.ld \ ${DESTDIR}${_LIBDIR}/${SHLIB_LINK} .for _SHLIB_LINK_LINK in ${SHLIB_LDSCRIPT_LINKS} ${INSTALL_SYMLINK} ${SHLIB_LINK} ${DESTDIR}${_LIBDIR}/${_SHLIB_LINK_LINK} .endfor .else .if ${_SHLIBDIR} == ${_LIBDIR} .if ${SHLIB_LINK:Mlib*} ${INSTALL_RSYMLINK} ${TAG_ARGS:D${TAG_ARGS},development} ${SHLIB_NAME} ${DESTDIR}${_LIBDIR}/${SHLIB_LINK} .else ${INSTALL_RSYMLINK} ${TAG_ARGS} ${DESTDIR}${_SHLIBDIR}/${SHLIB_NAME} \ ${DESTDIR}${_LIBDIR}/${SHLIB_LINK} .endif .else .if ${SHLIB_LINK:Mlib*} ${INSTALL_RSYMLINK} ${TAG_ARGS:D${TAG_ARGS},development} ${DESTDIR}${_SHLIBDIR}/${SHLIB_NAME} \ ${DESTDIR}${_LIBDIR}/${SHLIB_LINK} .else ${INSTALL_RSYMLINK} ${TAG_ARGS} ${DESTDIR}${_SHLIBDIR}/${SHLIB_NAME} \ ${DESTDIR}${_LIBDIR}/${SHLIB_LINK} .endif .if exists(${DESTDIR}${_LIBDIR}/${SHLIB_NAME}) -chflags noschg ${DESTDIR}${_LIBDIR}/${SHLIB_NAME} rm -f ${DESTDIR}${_LIBDIR}/${SHLIB_NAME} .endif .endif .endif # SHLIB_LDSCRIPT .endif # SHLIB_LINK .endif # SHIB_NAME .if defined(INSTALL_PIC_ARCHIVE) && defined(LIB) && !empty(LIB) && ${MK_TOOLCHAIN} != "no" ${INSTALL} ${TAG_ARGS:D${TAG_ARGS},development} -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} lib${LIB}_pic.a ${DESTDIR}${_LIBDIR}/ .endif .if defined(WANT_LINT) && !defined(NO_LINT) && defined(LIB) && !empty(LIB) ${INSTALL} ${TAG_ARGS:D${TAG_ARGS},development} -o ${LIBOWN} -g ${LIBGRP} -m ${LIBMODE} \ ${_INSTALLFLAGS} ${LINTLIB} ${DESTDIR}${LINTLIBDIR}/ .endif .endif # !defined(INTERNALLIB) .if !defined(LIBRARIES_ONLY) .include .include .include .include .endif .include .if ${MK_MAN} != "no" && !defined(LIBRARIES_ONLY) realinstall: maninstall .ORDER: beforeinstall maninstall .endif .endif .if !target(lint) lint: ${SRCS:M*.c} ${LINT} ${LINTFLAGS} ${CFLAGS:M-[DIU]*} ${.ALLSRC} .endif .if ${MK_MAN} != "no" && !defined(LIBRARIES_ONLY) .include .endif .if defined(LIB) && !empty(LIB) OBJS_DEPEND_GUESS+= ${SRCS:M*.h} -.if ${MK_FAST_DEPEND} == "no" && !exists(${.OBJDIR}/${DEPENDFILE}) -${OBJS} ${STATICOBJS} ${POBJS}: ${OBJS_DEPEND_GUESS} -.endif .for _S in ${SRCS:N*.[hly]} OBJS_DEPEND_GUESS.${_S:R}.po= ${_S} -.if ${MK_FAST_DEPEND} == "no" && !exists(${.OBJDIR}/${DEPENDFILE}) -${_S:R}.po: ${OBJS_DEPEND_GUESS.${_S:R}.po} -.endif .endfor .endif .if defined(SHLIB_NAME) || \ defined(INSTALL_PIC_ARCHIVE) && defined(LIB) && !empty(LIB) -.if ${MK_FAST_DEPEND} == "no" && !exists(${.OBJDIR}/${DEPENDFILE}) -${SOBJS}: ${OBJS_DEPEND_GUESS} -.endif .for _S in ${SRCS:N*.[hly]} OBJS_DEPEND_GUESS.${_S:R}.So= ${_S} +.endfor +.endif + +.include + +.if defined(LIB) && !empty(LIB) .if ${MK_FAST_DEPEND} == "no" && !exists(${.OBJDIR}/${DEPENDFILE}) +${OBJS} ${STATICOBJS} ${POBJS}: ${OBJS_DEPEND_GUESS} +.for _S in ${SRCS:N*.[hly]} +${_S:R}.po: ${OBJS_DEPEND_GUESS.${_S:R}.po} +.endfor +.if defined(SHLIB_NAME) || \ + defined(INSTALL_PIC_ARCHIVE) && defined(LIB) && !empty(LIB) +${SOBJS}: ${OBJS_DEPEND_GUESS} +.for _S in ${SRCS:N*.[hly]} ${_S:R}.So: ${OBJS_DEPEND_GUESS.${_S:R}.So} -.endif .endfor .endif +.endif +.endif -.include .include .include .include Index: projects/release-pkg/share/mk/bsd.prog.mk =================================================================== --- projects/release-pkg/share/mk/bsd.prog.mk (revision 296327) +++ projects/release-pkg/share/mk/bsd.prog.mk (revision 296328) @@ -1,295 +1,299 @@ # from: @(#)bsd.prog.mk 5.26 (Berkeley) 6/25/91 # $FreeBSD$ .include .include .SUFFIXES: .out .o .c .cc .cpp .cxx .C .m .y .l .ln .s .S .asm # XXX The use of COPTS in modern makefiles is discouraged. .if defined(COPTS) CFLAGS+=${COPTS} .endif .if ${MK_ASSERT_DEBUG} == "no" CFLAGS+= -DNDEBUG NO_WERROR= .endif .if defined(DEBUG_FLAGS) CFLAGS+=${DEBUG_FLAGS} CXXFLAGS+=${DEBUG_FLAGS} .if ${MK_CTF} != "no" && ${DEBUG_FLAGS:M-g} != "" CTFFLAGS+= -g .endif .endif .if defined(PROG_CXX) PROG= ${PROG_CXX} .endif .if !empty(LDFLAGS:M-Wl,*--oformat,*) || !empty(LDFLAGS:M-static) MK_DEBUG_FILES= no .endif .if defined(CRUNCH_CFLAGS) CFLAGS+=${CRUNCH_CFLAGS} .else .if ${MK_DEBUG_FILES} != "no" && empty(DEBUG_FLAGS:M-g) && \ empty(DEBUG_FLAGS:M-gdwarf-*) CFLAGS+= -g CTFFLAGS+= -g .endif .endif .if !defined(DEBUG_FLAGS) STRIP?= -s .endif .if defined(NO_ROOT) .if !defined(TAGS) || ! ${TAGS:Mpackage=*} TAGS+= package=${PACKAGE:Uruntime} .endif TAG_ARGS= -T ${TAGS:[*]:S/ /,/g} .endif .if defined(NO_SHARED) && (${NO_SHARED} != "no" && ${NO_SHARED} != "NO") LDFLAGS+= -static .endif .if ${MK_DEBUG_FILES} != "no" PROG_FULL=${PROG}.full # Use ${DEBUGDIR} for base system debug files, else .debug subdirectory .if defined(BINDIR) && (\ ${BINDIR} == "/bin" ||\ ${BINDIR:C%/libexec(/.*)?%/libexec%} == "/libexec" ||\ ${BINDIR} == "/sbin" ||\ ${BINDIR:C%/usr/(bin|bsdinstall|libexec|lpr|sendmail|sm.bin|sbin|tests)(/.*)?%/usr/bin%} == "/usr/bin"\ ) DEBUGFILEDIR= ${DEBUGDIR}${BINDIR} .else DEBUGFILEDIR?= ${BINDIR}/.debug DEBUGMKDIR= .endif .else PROG_FULL= ${PROG} .endif .if defined(PROG) PROGNAME?= ${PROG} .if defined(SRCS) OBJS+= ${SRCS:N*.h:R:S/$/.o/g} .if target(beforelinking) beforelinking: ${OBJS} ${PROG_FULL}: beforelinking .endif ${PROG_FULL}: ${OBJS} .if defined(PROG_CXX) ${CXX:N${CCACHE_BIN}} ${CXXFLAGS:N-M*} ${LDFLAGS} -o ${.TARGET} \ ${OBJS} ${LDADD} .else ${CC:N${CCACHE_BIN}} ${CFLAGS:N-M*} ${LDFLAGS} -o ${.TARGET} ${OBJS} \ ${LDADD} .endif .if ${MK_CTF} != "no" ${CTFMERGE} ${CTFFLAGS} -o ${.TARGET} ${OBJS} .endif .else # !defined(SRCS) .if !target(${PROG}) .if defined(PROG_CXX) SRCS= ${PROG}.cc .else SRCS= ${PROG}.c .endif # Always make an intermediate object file because: # - it saves time rebuilding when only the library has changed # - the name of the object gets put into the executable symbol table instead of # the name of a variable temporary object. # - it's useful to keep objects around for crunching. OBJS+= ${PROG}.o .if target(beforelinking) beforelinking: ${OBJS} ${PROG_FULL}: beforelinking .endif ${PROG_FULL}: ${OBJS} .if defined(PROG_CXX) ${CXX:N${CCACHE_BIN}} ${CXXFLAGS:N-M*} ${LDFLAGS} -o ${.TARGET} \ ${OBJS} ${LDADD} .else ${CC:N${CCACHE_BIN}} ${CFLAGS:N-M*} ${LDFLAGS} -o ${.TARGET} ${OBJS} \ ${LDADD} .endif .if ${MK_CTF} != "no" ${CTFMERGE} ${CTFFLAGS} -o ${.TARGET} ${OBJS} .endif .endif # !target(${PROG}) .endif # !defined(SRCS) .if ${MK_DEBUG_FILES} != "no" ${PROG}: ${PROG_FULL} ${PROGNAME}.debug ${OBJCOPY} --strip-debug --add-gnu-debuglink=${PROGNAME}.debug \ ${PROG_FULL} ${.TARGET} ${PROGNAME}.debug: ${PROG_FULL} ${OBJCOPY} --only-keep-debug ${PROG_FULL} ${.TARGET} .endif .if ${MK_MAN} != "no" && !defined(MAN) && \ !defined(MAN1) && !defined(MAN2) && !defined(MAN3) && \ !defined(MAN4) && !defined(MAN5) && !defined(MAN6) && \ !defined(MAN7) && !defined(MAN8) && !defined(MAN9) MAN= ${PROG}.1 MAN1= ${MAN} .endif .endif # defined(PROG) .if defined(_SKIP_BUILD) all: .else all: ${PROG} ${SCRIPTS} .if ${MK_MAN} != "no" all: all-man .endif .endif .if defined(PROG) CLEANFILES+= ${PROG} .if ${MK_DEBUG_FILES} != "no" CLEANFILES+= ${PROG_FULL} ${PROGNAME}.debug .endif .endif .if defined(OBJS) CLEANFILES+= ${OBJS} .endif .include .if defined(PROG) _EXTRADEPEND: .if defined(LDFLAGS) && !empty(LDFLAGS:M-nostdlib) .if defined(DPADD) && !empty(DPADD) echo ${PROG_FULL}: ${DPADD} >> ${DEPENDFILE} .endif .else echo ${PROG_FULL}: ${LIBC} ${DPADD} >> ${DEPENDFILE} .if defined(PROG_CXX) .if ${COMPILER_TYPE} == "clang" && empty(CXXFLAGS:M-stdlib=libstdc++) echo ${PROG_FULL}: ${LIBCPLUSPLUS} >> ${DEPENDFILE} .else echo ${PROG_FULL}: ${LIBSTDCPLUSPLUS} >> ${DEPENDFILE} .endif .endif .endif .endif .if !target(install) .if defined(PRECIOUSPROG) .if !defined(NO_FSCHG) INSTALLFLAGS+= -fschg .endif INSTALLFLAGS+= -S .endif _INSTALLFLAGS:= ${INSTALLFLAGS} .for ie in ${INSTALLFLAGS_EDIT} _INSTALLFLAGS:= ${_INSTALLFLAGS${ie}} .endfor .if !target(realinstall) && !defined(INTERNALPROG) realinstall: _proginstall .ORDER: beforeinstall _proginstall _proginstall: .if defined(PROG) ${INSTALL} ${TAG_ARGS} ${STRIP} -o ${BINOWN} -g ${BINGRP} -m ${BINMODE} \ ${_INSTALLFLAGS} ${PROG} ${DESTDIR}${BINDIR}/${PROGNAME} .if ${MK_DEBUG_FILES} != "no" .if defined(DEBUGMKDIR) ${INSTALL} ${TAG_ARGS:D${TAG_ARGS},debug} -d ${DESTDIR}${DEBUGFILEDIR}/ .endif ${INSTALL} ${TAG_ARGS:D${TAG_ARGS},debug} -o ${BINOWN} -g ${BINGRP} -m ${DEBUGMODE} \ ${PROGNAME}.debug ${DESTDIR}${DEBUGFILEDIR}/${PROGNAME}.debug .endif .endif .endif # !target(realinstall) .if defined(SCRIPTS) && !empty(SCRIPTS) realinstall: _scriptsinstall .ORDER: beforeinstall _scriptsinstall SCRIPTSDIR?= ${BINDIR} SCRIPTSOWN?= ${BINOWN} SCRIPTSGRP?= ${BINGRP} SCRIPTSMODE?= ${BINMODE} STAGE_AS_SETS+= scripts stage_as.scripts: ${SCRIPTS} FLAGS.stage_as.scripts= -m ${SCRIPTSMODE} STAGE_FILES_DIR.scripts= ${STAGE_OBJTOP} .for script in ${SCRIPTS} .if defined(SCRIPTSNAME) SCRIPTSNAME_${script:T}?= ${SCRIPTSNAME} .else SCRIPTSNAME_${script:T}?= ${script:T:R} .endif SCRIPTSDIR_${script:T}?= ${SCRIPTSDIR} SCRIPTSOWN_${script:T}?= ${SCRIPTSOWN} SCRIPTSGRP_${script:T}?= ${SCRIPTSGRP} SCRIPTSMODE_${script:T}?= ${SCRIPTSMODE} STAGE_AS_${script:T}= ${SCRIPTSDIR_${script:T}}/${SCRIPTSNAME_${script:T}} _scriptsinstall: _SCRIPTSINS_${script:T} _SCRIPTSINS_${script:T}: ${script} ${INSTALL} ${TAG_ARGS} -o ${SCRIPTSOWN_${.ALLSRC:T}} \ -g ${SCRIPTSGRP_${.ALLSRC:T}} -m ${SCRIPTSMODE_${.ALLSRC:T}} \ ${.ALLSRC} \ ${DESTDIR}${SCRIPTSDIR_${.ALLSRC:T}}/${SCRIPTSNAME_${.ALLSRC:T}} .endfor .endif NLSNAME?= ${PROG} .include .include .include .include .include .if ${MK_MAN} != "no" realinstall: maninstall .ORDER: beforeinstall maninstall .endif .endif # !target(install) .if !target(lint) lint: ${SRCS:M*.c} .if defined(PROG) ${LINT} ${LINTFLAGS} ${CFLAGS:M-[DIU]*} ${.ALLSRC} .endif .endif .if ${MK_MAN} != "no" .include .endif .if defined(PROG) OBJS_DEPEND_GUESS+= ${SRCS:M*.h} +.endif + +.include + +.if defined(PROG) .if ${MK_FAST_DEPEND} == "no" && !exists(${.OBJDIR}/${DEPENDFILE}) ${OBJS}: ${OBJS_DEPEND_GUESS} .endif .endif -.include .include .include .include Index: projects/release-pkg/share/mk/meta2deps.sh =================================================================== --- projects/release-pkg/share/mk/meta2deps.sh (revision 296327) +++ projects/release-pkg/share/mk/meta2deps.sh (revision 296328) @@ -1,404 +1,404 @@ # $FreeBSD$ #!/bin/sh # NAME: # meta2deps.sh - extract useful info from .meta files # # SYNOPSIS: # meta2deps.sh SB="SB" "meta" ... # # DESCRIPTION: # This script looks each "meta" file and extracts the # information needed to deduce build and src dependencies. # # To do this, we extract the 'CWD' record as well as all the # syscall traces which describe 'R'ead, 'C'hdir and 'E'xec # syscalls. # # The typical meta file looks like:: #.nf # # # Meta data file "path" # CMD "command-line" # CWD "cwd" # TARGET "target" # -- command output -- # -- filemon acquired metadata -- # # buildmon version 2 # V 2 # E "pid" "path" # R "pid" "path" # C "pid" "cwd" # R "pid" "path" # X "pid" "status" #.fi # # The fact that all the syscall entry lines start with a single # character make these files quite easy to process using sed(1). # # To simplify the logic the 'CWD' line is made to look like a # normal 'C'hdir entry, and "cwd" is remembered so that it can # be prefixed to any "path" which is not absolute. # # If the "path" being read ends in '.srcrel' it is the content # of (actually the first line of) that file that we are # interested in. # # Any "path" which lies outside of the sandbox "SB" is generally # not of interest and is ignored. # # The output, is a set of absolute paths with "SB" like: #.nf # # $SB/obj-i386/bsd/gnu/lib/csu # $SB/obj-i386/bsd/gnu/lib/libgcc # $SB/obj-i386/bsd/include # $SB/obj-i386/bsd/lib/csu/i386-elf # $SB/obj-i386/bsd/lib/libc # $SB/src/bsd/include # $SB/src/bsd/sys/i386/include # $SB/src/bsd/sys/sys # $SB/src/pan-release/rtsock # $SB/src/pfe-shared/include/jnx #.fi # # Which can then be further processed by 'gendirdeps.mk' # # If we are passed 'DPDEPS='"dpdeps", then for each src file # outside of "CURDIR" we read, we output a line like: #.nf # # DPDEPS_$path += $RELDIR #.fi # # with "$path" geting turned into reldir's, so that we can end # up with a list of all the directories which depend on each src # file in another directory. This can allow for efficient yet # complete testing of changes. # RCSid: # $Id: meta2deps.sh,v 1.9 2015/04/03 18:23:25 sjg Exp $ # Copyright (c) 2010-2013, Juniper Networks, Inc. # 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 COPYRIGHT HOLDERS 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 COPYRIGHT # OWNER 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. meta2src() { cat /dev/null "$@" | sed -n '/^R .*\.[chyl]$/s,^..[0-9]* ,,p' | sort -u } meta2dirs() { cat /dev/null "$@" | sed -n '/^R .*\/.*\.[a-z0-9][^\/]*$/s,^..[0-9]* \(.*\)/[^/]*$,\1,p' | sort -u } add_list() { sep=' ' suffix= while : do case "$1" in "|") sep="$1"; shift;; -s) suffix="$2"; shift 2;; *) break;; esac done name=$1 shift eval list="\$$name" for top in "$@" do case "$sep$list$sep" in *"$sep$top$suffix$sep"*) continue;; esac list="${list:+$list$sep}$top$suffix" done eval "$name=\"$list\"" } _excludes_f() { egrep -v "$EXCLUDES" } meta2deps() { DPDEPS= SRCTOPS=$SRCTOP OBJROOTS= EXCLUDES= while : do case "$1" in *=*) eval export "$1"; shift;; -a) MACHINE_ARCH=$2; shift 2;; -m) MACHINE=$2; shift 2;; -C) CURDIR=$2; shift 2;; -H) HOST_TARGET=$2; shift 2;; -S) add_list SRCTOPS $2; shift 2;; -O) add_list OBJROOTS $2; shift 2;; -X) add_list EXCLUDES '|' $2; shift 2;; -R) RELDIR=$2; shift 2;; -T) TARGET_SPEC=$2; shift 2;; *) break;; esac done _th= _o= case "$MACHINE" in host) _ht=$HOST_TARGET;; esac for o in $OBJROOTS do case "$MACHINE,/$o/" in host,*$HOST_TARGET*) ;; *$MACHINE*|*${TARGET_SPEC:-$MACHINE}*) ;; *) add_list _o $o; continue;; esac for x in $_ht $TARGET_SPEC $MACHINE do case "$o" in "") continue;; */$x/) add_list _o ${o%$x/}; o=;; */$x) add_list _o ${o%$x}; o=;; *$x/) add_list _o ${o%$x/}; o=;; *$x) add_list _o ${o%$x}; o=;; esac done done OBJROOTS="$_o" case "$OBJTOP" in "") for o in $OBJROOTS do OBJTOP=$o${TARGET_SPEC:-$MACHINE} break done ;; esac src_re= obj_re= add_list '|' -s '/*' src_re $SRCTOPS add_list '|' -s '*' obj_re $OBJROOTS [ -z "$RELDIR" ] && unset DPDEPS tf=/tmp/m2d$$-$USER rm -f $tf.* trap 'rm -f $tf.*; trap 0' 0 > $tf.dirdep > $tf.qual > $tf.srcdep > $tf.srcrel > $tf.dpdeps seenit= seensrc= lpid= case "$EXCLUDES" in "") _excludes=cat;; *) _excludes=_excludes_f;; esac # handle @list files case "$@" in *@[!.]*) for f in "$@" do case "$f" in *.meta) cat $f;; @*) xargs cat < ${f#@};; *) cat $f;; esac done ;; *) cat /dev/null "$@";; esac 2> /dev/null | sed -e 's,^CWD,C C,;/^[CREFLM] /!d' -e "s,',,g" | $_excludes | while read op pid path junk do : op=$op pid=$pid path=$path # we track cwd and ldir (of interest) per pid # CWD is bmake's cwd case "$lpid,$pid" in ,C) CWD=$path cwd=$path ldir=$path if [ -z "$SB" ]; then SB=`echo $CWD | sed 's,/obj.*,,'` fi SRCTOP=${SRCTOP:-$SB/src} continue ;; $pid,$pid) ;; *) case "$lpid" in "") ;; *) eval ldir_$lpid=$ldir cwd_$lpid=$cwd;; esac eval ldir=\${ldir_$pid:-$CWD} cwd=\${cwd_$pid:-$CWD} lpid=$pid ;; esac case "$op,$path" in W,*srcrel|*.dirdep) continue;; C,*) case "$path" in /*) cwd=$path;; *) cwd=`cd $cwd/$path 2> /dev/null && /bin/pwd`;; esac # watch out for temp dirs that no longer exist test -d ${cwd:-/dev/null/no/such} || cwd=$CWD continue ;; F,*) eval cwd_$path=$cwd ldir_$path=$ldir continue ;; *) dir=${path%/*} case "$path" in $src_re|$obj_re) ;; /*/stage/*) ;; /*) continue;; *) for path in $ldir/$path $cwd/$path do test -e $path && break done dir=${path%/*} ;; esac ;; esac # avoid repeating ourselves... case "$DPDEPS,$seensrc," in ,*) case ",$seenit," in *,$dir,*) continue;; esac ;; *,$path,*) continue;; esac # canonicalize if needed case "/$dir/" in */../*|*/./*) rdir=$dir dir=`cd $dir 2> /dev/null && /bin/pwd` seen="$rdir,$dir" ;; *) seen=$dir;; esac case "$dir" in - ${CURDIR:-.}|${CURDIR:-.}/*|"") continue;; + ${CURDIR:-.}|"") continue;; $src_re) # avoid repeating ourselves... case "$DPDEPS,$seensrc," in ,*) case ",$seenit," in *,$dir,*) continue;; esac ;; esac ;; *) case ",$seenit," in *,$dir,*) continue;; esac ;; esac if [ -d $path ]; then case "$path" in */..) ldir=${dir%/*};; *) ldir=$path;; esac continue fi [ -f $path ] || continue case "$dir" in $CWD) continue;; # ignore $src_re) seenit="$seenit,$seen" echo $dir >> $tf.srcdep case "$DPDEPS,$reldir,$seensrc," in ,*) ;; *) seensrc="$seensrc,$path" echo "DPDEPS_$dir/${path##*/} += $RELDIR" >> $tf.dpdeps ;; esac continue ;; esac # if there is a .dirdep we cannot skip # just because we've seen the dir before. if [ -s $path.dirdep ]; then # this file contains: # '# ${RELDIR}.' echo $path.dirdep >> $tf.qual continue elif [ -s $dir.dirdep ]; then echo $dir.dirdep >> $tf.qual seenit="$seenit,$seen" continue fi seenit="$seenit,$seen" case "$dir" in $obj_re) echo $dir;; esac done > $tf.dirdep _nl=echo for f in $tf.dirdep $tf.qual $tf.srcdep do [ -s $f ] || continue case $f in *qual) # a list of .dirdep files # we can prefix everything with $OBJTOP to # tell gendirdeps.mk that these are # DIRDEP entries, since they are already # qualified with . as needed. # We strip .$MACHINE though xargs cat < $f | sort -u | sed "s,^# ,,;s,^,$OBJTOP/,;s,\.${TARGET_SPEC:-$MACHINE}\$,,;s,\.$MACHINE\$,," ;; *) sort -u $f;; esac _nl=: done if [ -s $tf.dpdeps ]; then case "$DPDEPS" in */*) ;; *) echo > $DPDEPS;; # the echo is needed! esac sort -u $tf.dpdeps | sed "s,${SRCTOP}/,,;s,${SB_BACKING_SB:-$SB}/src/,," >> $DPDEPS fi # ensure we produce _something_ else egrep -v gets upset $_nl } case /$0 in */meta2dep*) meta2deps "$@";; */meta2dirs*) meta2dirs "$@";; */meta2src*) meta2src "$@";; esac Index: projects/release-pkg/share =================================================================== --- projects/release-pkg/share (revision 296327) +++ projects/release-pkg/share (revision 296328) Property changes on: projects/release-pkg/share ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/share:r296315-296327 Index: projects/release-pkg/sys/arm64/include/param.h =================================================================== --- projects/release-pkg/sys/arm64/include/param.h (revision 296327) +++ projects/release-pkg/sys/arm64/include/param.h (revision 296328) @@ -1,125 +1,127 @@ /*- * 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. * * 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 _MACHINE_PARAM_H_ #define _MACHINE_PARAM_H_ /* * Machine dependent constants for arm64. */ #include #define STACKALIGNBYTES (16 - 1) #define STACKALIGN(p) ((uint64_t)(p) & ~STACKALIGNBYTES) +#define __PCI_REROUTE_INTERRUPT + #ifndef MACHINE #define MACHINE "arm64" #endif #ifndef MACHINE_ARCH #define MACHINE_ARCH "aarch64" #endif #if defined(SMP) || defined(KLD_MODULE) #ifndef MAXCPU #define MAXCPU 96 #endif #else #define MAXCPU 1 #endif /* SMP || KLD_MODULE */ #ifndef MAXMEMDOM #define MAXMEMDOM 1 #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) ((((u_long)(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 6 #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 PAGE_SHIFT_16K 14 #define PAGE_SIZE_16K (1 << PAGE_SHIFT_16K) #define PAGE_MASK_16K (PAGE_SIZE_16K - 1) #define PAGE_SHIFT_64K 16 #define PAGE_SIZE_64K (1 << PAGE_SHIFT_64K) #define PAGE_MASK_64K (PAGE_SIZE_64K - 1) #define MAXPAGESIZES 1 /* maximum number of supported page sizes */ #ifndef KSTACK_PAGES #define KSTACK_PAGES 4 /* pages of kernel stack (with pcb) */ #endif #define KSTACK_GUARD_PAGES 1 /* pages of kstack guard; 0 disables */ #define PCPU_PAGES 1 /* * Ceiling on size of buffer cache (really only effects write queueing, * the VM page cache is not effected), can be changed via * the kern.maxbcache /boot/loader.conf variable. */ #ifndef VM_BCACHE_SIZE_MAX #define VM_BCACHE_SIZE_MAX (400 * 1024 * 1024) #endif /* * Mach derived conversion macros */ #define round_page(x) (((unsigned long)(x) + PAGE_MASK) & ~PAGE_MASK) #define trunc_page(x) ((unsigned long)(x) & ~PAGE_MASK) #define atop(x) ((unsigned long)(x) >> PAGE_SHIFT) #define ptoa(x) ((unsigned long)(x) << PAGE_SHIFT) #define arm64_btop(x) ((unsigned long)(x) >> PAGE_SHIFT) #define arm64_ptob(x) ((unsigned long)(x) << PAGE_SHIFT) #define pgtok(x) ((unsigned long)(x) * (PAGE_SIZE / 1024)) #endif /* !_MACHINE_PARAM_H_ */ Index: projects/release-pkg/sys/conf/kmod.mk =================================================================== --- projects/release-pkg/sys/conf/kmod.mk (revision 296327) +++ projects/release-pkg/sys/conf/kmod.mk (revision 296328) @@ -1,466 +1,468 @@ # From: @(#)bsd.prog.mk 5.26 (Berkeley) 6/25/91 # $FreeBSD$ # # The include file handles building and installing loadable # kernel modules. # # # +++ variables +++ # # CLEANFILES Additional files to remove for the clean and cleandir targets. # # EXPORT_SYMS A list of symbols that should be exported from the module, # or the name of a file containing a list of symbols, or YES # to export all symbols. If not defined, no symbols are # exported. # # KMOD The name of the kernel module to build. # # KMODDIR Base path for kernel modules (see kld(4)). [/boot/kernel] # # KMODOWN Module file owner. [${BINOWN}] # # KMODGRP Module file group. [${BINGRP}] # # KMODMODE Module file mode. [${BINMODE}] # # KMODLOAD Command to load a kernel module [/sbin/kldload] # # KMODUNLOAD Command to unload a kernel module [/sbin/kldunload] # # KMODISLOADED Command to check whether a kernel module is # loaded [/sbin/kldstat -q -n] # # PROG The name of the kernel module to build. # If not supplied, ${KMOD}.ko is used. # # SRCS List of source files. # # FIRMWS List of firmware images in format filename:shortname:version # # FIRMWARE_LICENSE # Set to the name of the license the user has to agree on in # order to use this firmware. See /usr/share/doc/legal # # DESTDIR The tree where the module gets installed. [not set] # # +++ targets +++ # # install: # install the kernel module; if the Makefile # does not itself define the target install, the targets # beforeinstall and afterinstall may also be used to cause # actions immediately before and after the install target # is executed. # # load: # Load a module. # # unload: # Unload a module. # # reload: # Unload if loaded, then load. # AWK?= awk KMODLOAD?= /sbin/kldload KMODUNLOAD?= /sbin/kldunload KMODISLOADED?= /sbin/kldstat -q -n OBJCOPY?= objcopy .include # Grab all the options for a kernel build. For backwards compat, we need to # do this after bsd.own.mk. .include "kern.opts.mk" .include .include "config.mk" .SUFFIXES: .out .o .c .cc .cxx .C .y .l .s .S .m # amd64 and mips use direct linking for kmod, all others use shared binaries .if ${MACHINE_CPUARCH} != amd64 && ${MACHINE_CPUARCH} != mips __KLD_SHARED=yes .else __KLD_SHARED=no .endif .if !empty(CFLAGS:M-O[23s]) && empty(CFLAGS:M-fno-strict-aliasing) CFLAGS+= -fno-strict-aliasing .endif WERROR?= -Werror CFLAGS+= ${WERROR} CFLAGS+= -D_KERNEL CFLAGS+= -DKLD_MODULE # Don't use any standard or source-relative include directories. NOSTDINC= -nostdinc CFLAGS:= ${CFLAGS:N-I*} ${NOSTDINC} ${INCLMAGIC} ${CFLAGS:M-I*} .if defined(KERNBUILDDIR) CFLAGS+= -DHAVE_KERNEL_OPTION_HEADERS -include ${KERNBUILDDIR}/opt_global.h .endif # Add -I paths for system headers. Individual module makefiles don't # need any -I paths for this. Similar defaults for .PATH can't be # set because there are no standard paths for non-headers. CFLAGS+= -I. -I${SYSDIR} CFLAGS.gcc+= -finline-limit=${INLINE_LIMIT} CFLAGS.gcc+= -fms-extensions CFLAGS.gcc+= --param inline-unit-growth=100 CFLAGS.gcc+= --param large-function-growth=1000 # Disallow common variables, and if we end up with commons from # somewhere unexpected, allocate storage for them in the module itself. CFLAGS+= -fno-common LDFLAGS+= -d -warn-common CFLAGS+= ${DEBUG_FLAGS} .if ${MACHINE_CPUARCH} == amd64 CFLAGS+= -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer .endif .if ${MACHINE_CPUARCH} == "aarch64" CFLAGS+= -fPIC .endif # Temporary workaround for PR 196407, which contains the fascinating details. # Don't allow clang to use fpu instructions or registers in kernel modules. .if ${MACHINE_CPUARCH} == arm CFLAGS.clang+= -mllvm -arm-use-movt=0 CFLAGS.clang+= -mfpu=none CFLAGS+= -funwind-tables .endif .if ${MACHINE_CPUARCH} == powerpc CFLAGS+= -mlongcall -fno-omit-frame-pointer .endif .if ${MACHINE_CPUARCH} == mips CFLAGS+= -G0 -fno-pic -mno-abicalls -mlong-calls .endif .if defined(DEBUG) || defined(DEBUG_FLAGS) CTFFLAGS+= -g .endif .if defined(FIRMWS) ${KMOD:S/$/.c/}: ${SYSDIR}/tools/fw_stub.awk ${AWK} -f ${SYSDIR}/tools/fw_stub.awk ${FIRMWS} -m${KMOD} -c${KMOD:S/$/.c/g} \ ${FIRMWARE_LICENSE:C/.+/-l/}${FIRMWARE_LICENSE} SRCS+= ${KMOD:S/$/.c/} CLEANFILES+= ${KMOD:S/$/.c/} .for _firmw in ${FIRMWS} ${_firmw:C/\:.*$/.fwo/}: ${_firmw:C/\:.*$//} @${ECHO} ${_firmw:C/\:.*$//} ${.ALLSRC:M*${_firmw:C/\:.*$//}} @if [ -e ${_firmw:C/\:.*$//} ]; then \ ${LD} -b binary --no-warn-mismatch ${_LDFLAGS} \ -r -d -o ${.TARGET} ${_firmw:C/\:.*$//}; \ else \ ln -s ${.ALLSRC:M*${_firmw:C/\:.*$//}} ${_firmw:C/\:.*$//}; \ ${LD} -b binary --no-warn-mismatch ${_LDFLAGS} \ -r -d -o ${.TARGET} ${_firmw:C/\:.*$//}; \ rm ${_firmw:C/\:.*$//}; \ fi OBJS+= ${_firmw:C/\:.*$/.fwo/} .endfor .endif # Conditionally include SRCS based on kernel config options. .for _o in ${KERN_OPTS} SRCS+=${SRCS.${_o}} .endfor OBJS+= ${SRCS:N*.h:R:S/$/.o/g} .if !defined(PROG) PROG= ${KMOD}.ko .endif .if !defined(DEBUG_FLAGS) FULLPROG= ${PROG} .else FULLPROG= ${PROG}.full ${PROG}: ${FULLPROG} ${PROG}.debug ${OBJCOPY} --strip-debug --add-gnu-debuglink=${PROG}.debug \ ${FULLPROG} ${.TARGET} ${PROG}.debug: ${FULLPROG} ${OBJCOPY} --only-keep-debug ${FULLPROG} ${.TARGET} .endif .if ${__KLD_SHARED} == yes ${FULLPROG}: ${KMOD}.kld .if ${MACHINE_CPUARCH} != "aarch64" ${LD} -Bshareable ${_LDFLAGS} -o ${.TARGET} ${KMOD}.kld .else #XXXKIB Relocatable linking in aarch64 ld from binutils 2.25.1 does # not work. The linker corrupts the references to the external # symbols which are defined by other object in the linking set # and should therefore loose the GOT entry. The problem seems # to be fixed in the binutils-gdb git HEAD as of 2015-10-04. Hack # below allows to get partially functioning modules for now. ${LD} -Bshareable ${_LDFLAGS} -o ${.TARGET} ${OBJS} .endif .if !defined(DEBUG_FLAGS) ${OBJCOPY} --strip-debug ${.TARGET} .endif .endif EXPORT_SYMS?= NO .if ${EXPORT_SYMS} != YES CLEANFILES+= export_syms .endif .if ${__KLD_SHARED} == yes ${KMOD}.kld: ${OBJS} .else ${FULLPROG}: ${OBJS} .endif ${LD} ${_LDFLAGS} -r -d -o ${.TARGET} ${OBJS} .if ${MK_CTF} != "no" ${CTFMERGE} ${CTFFLAGS} -o ${.TARGET} ${OBJS} .endif .if defined(EXPORT_SYMS) .if ${EXPORT_SYMS} != YES .if ${EXPORT_SYMS} == NO :> export_syms .elif !exists(${.CURDIR}/${EXPORT_SYMS}) echo ${EXPORT_SYMS} > export_syms .else grep -v '^#' < ${EXPORT_SYMS} > export_syms .endif ${AWK} -f ${SYSDIR}/conf/kmod_syms.awk ${.TARGET} \ export_syms | xargs -J% ${OBJCOPY} % ${.TARGET} .endif .endif .if !defined(DEBUG_FLAGS) && ${__KLD_SHARED} == no ${OBJCOPY} --strip-debug ${.TARGET} .endif _ILINKS=machine .if ${MACHINE} != ${MACHINE_CPUARCH} && ${MACHINE} != "arm64" _ILINKS+=${MACHINE_CPUARCH} .endif .if ${MACHINE_CPUARCH} == "i386" || ${MACHINE_CPUARCH} == "amd64" _ILINKS+=x86 .endif CLEANFILES+=${_ILINKS} all: ${PROG} beforedepend: ${_ILINKS} beforebuild: ${_ILINKS} # Ensure that the links exist without depending on it when it exists which # causes all the modules to be rebuilt when the directory pointed to changes. .for _link in ${_ILINKS} .if !exists(${.OBJDIR}/${_link}) ${OBJS}: ${_link} .endif .endfor # Search for kernel source tree in standard places. .for _dir in ${.CURDIR}/../.. ${.CURDIR}/../../.. /sys /usr/src/sys .if !defined(SYSDIR) && exists(${_dir}/kern/) SYSDIR= ${_dir} .endif .endfor .if !defined(SYSDIR) || !exists(${SYSDIR}/kern/) .error "can't find kernel source tree" .endif .NOPATH: ${_ILINKS} ${_ILINKS}: @case ${.TARGET} in \ machine) \ path=${SYSDIR}/${MACHINE}/include ;; \ *) \ path=${SYSDIR}/${.TARGET:T}/include ;; \ esac ; \ path=`(cd $$path && /bin/pwd)` ; \ ${ECHO} ${.TARGET:T} "->" $$path ; \ ln -sf $$path ${.TARGET:T} CLEANFILES+= ${PROG} ${KMOD}.kld ${OBJS} .if defined(DEBUG_FLAGS) CLEANFILES+= ${FULLPROG} ${PROG}.debug .endif .if !target(install) _INSTALLFLAGS:= ${INSTALLFLAGS} .for ie in ${INSTALLFLAGS_EDIT} _INSTALLFLAGS:= ${_INSTALLFLAGS${ie}} .endfor .if !target(realinstall) KERN_DEBUGDIR?= ${DEBUGDIR} realinstall: _kmodinstall .ORDER: beforeinstall _kmodinstall _kmodinstall: ${INSTALL} -T release -o ${KMODOWN} -g ${KMODGRP} -m ${KMODMODE} \ ${_INSTALLFLAGS} ${PROG} ${DESTDIR}${KMODDIR}/ .if defined(DEBUG_FLAGS) && !defined(INSTALL_NODEBUG) && ${MK_KERNEL_SYMBOLS} != "no" ${INSTALL} -T debug -o ${KMODOWN} -g ${KMODGRP} -m ${KMODMODE} \ ${_INSTALLFLAGS} ${PROG}.debug ${DESTDIR}${KERN_DEBUGDIR}${KMODDIR}/ .endif .include .if !defined(NO_XREF) afterinstall: _kldxref .ORDER: realinstall _kldxref .ORDER: _installlinks _kldxref _kldxref: @if type kldxref >/dev/null 2>&1; then \ ${ECHO} kldxref ${DESTDIR}${KMODDIR}; \ kldxref ${DESTDIR}${KMODDIR}; \ fi .endif .endif # !target(realinstall) .endif # !target(install) .if !target(load) load: ${PROG} ${KMODLOAD} -v ${.OBJDIR}/${PROG} .endif .if !target(unload) unload: if ${KMODISLOADED} ${PROG} ; then ${KMODUNLOAD} -v ${PROG} ; fi .endif .if !target(reload) reload: unload load .endif .if defined(KERNBUILDDIR) .PATH: ${KERNBUILDDIR} CFLAGS+= -I${KERNBUILDDIR} .for _src in ${SRCS:Mopt_*.h} CLEANFILES+= ${_src} .if !target(${_src}) ${_src}: ln -sf ${KERNBUILDDIR}/${_src} ${.TARGET} .endif .endfor .else .for _src in ${SRCS:Mopt_*.h} CLEANFILES+= ${_src} .if !target(${_src}) ${_src}: :> ${.TARGET} .endif .endfor .endif # Respect configuration-specific C flags. CFLAGS+= ${CONF_CFLAGS} .if !empty(SRCS:Mvnode_if.c) CLEANFILES+= vnode_if.c vnode_if.c: ${SYSDIR}/tools/vnode_if.awk ${SYSDIR}/kern/vnode_if.src ${AWK} -f ${SYSDIR}/tools/vnode_if.awk ${SYSDIR}/kern/vnode_if.src -c .endif .if !empty(SRCS:Mvnode_if.h) CLEANFILES+= vnode_if.h vnode_if_newproto.h vnode_if_typedef.h vnode_if.h vnode_if_newproto.h vnode_if_typedef.h: ${SYSDIR}/tools/vnode_if.awk \ ${SYSDIR}/kern/vnode_if.src vnode_if.h: vnode_if_newproto.h vnode_if_typedef.h ${AWK} -f ${SYSDIR}/tools/vnode_if.awk ${SYSDIR}/kern/vnode_if.src -h vnode_if_newproto.h: ${AWK} -f ${SYSDIR}/tools/vnode_if.awk ${SYSDIR}/kern/vnode_if.src -p vnode_if_typedef.h: ${AWK} -f ${SYSDIR}/tools/vnode_if.awk ${SYSDIR}/kern/vnode_if.src -q .endif # Build _if.[ch] from _if.m, and clean them when we're done. # This is duplicated in sys/modules/Makefile. .if !defined(__MPATH) __MPATH!=find ${SYSDIR:tA}/ -name \*_if.m .export __MPATH .endif _MFILES=${__MPATH:T:O} _MPATH=${__MPATH:H:O:u} .PATH.m: ${_MPATH} .for _i in ${SRCS:M*_if.[ch]} _MATCH=M${_i:R:S/$/.m/} _MATCHES=${_MFILES:${_MATCH}} .if !empty(_MATCHES) CLEANFILES+= ${_i} .endif .endfor # _i .m.c: ${SYSDIR}/tools/makeobjops.awk ${AWK} -f ${SYSDIR}/tools/makeobjops.awk ${.IMPSRC} -c .m.h: ${SYSDIR}/tools/makeobjops.awk ${AWK} -f ${SYSDIR}/tools/makeobjops.awk ${.IMPSRC} -h .for _i in mii pccard .if !empty(SRCS:M${_i}devs.h) CLEANFILES+= ${_i}devs.h ${_i}devs.h: ${SYSDIR}/tools/${_i}devs2h.awk ${SYSDIR}/dev/${_i}/${_i}devs ${AWK} -f ${SYSDIR}/tools/${_i}devs2h.awk ${SYSDIR}/dev/${_i}/${_i}devs .endif .endfor # _i .if !empty(SRCS:Musbdevs.h) CLEANFILES+= usbdevs.h usbdevs.h: ${SYSDIR}/tools/usbdevs2h.awk ${SYSDIR}/dev/usb/usbdevs ${AWK} -f ${SYSDIR}/tools/usbdevs2h.awk ${SYSDIR}/dev/usb/usbdevs -h .endif .if !empty(SRCS:Musbdevs_data.h) CLEANFILES+= usbdevs_data.h usbdevs_data.h: ${SYSDIR}/tools/usbdevs2h.awk ${SYSDIR}/dev/usb/usbdevs ${AWK} -f ${SYSDIR}/tools/usbdevs2h.awk ${SYSDIR}/dev/usb/usbdevs -d .endif .if !empty(SRCS:Macpi_quirks.h) CLEANFILES+= acpi_quirks.h acpi_quirks.h: ${SYSDIR}/tools/acpi_quirks2h.awk ${SYSDIR}/dev/acpica/acpi_quirks ${AWK} -f ${SYSDIR}/tools/acpi_quirks2h.awk ${SYSDIR}/dev/acpica/acpi_quirks .endif .if !empty(SRCS:Massym.s) CLEANFILES+= assym.s genassym.o assym.s: genassym.o .if defined(KERNBUILDDIR) genassym.o: opt_global.h .endif assym.s: ${SYSDIR}/kern/genassym.sh sh ${SYSDIR}/kern/genassym.sh genassym.o > ${.TARGET} genassym.o: ${SYSDIR}/${MACHINE}/${MACHINE}/genassym.c genassym.o: ${SRCS:Mopt_*.h} ${CC} -c ${CFLAGS:N-fno-common} \ ${SYSDIR}/${MACHINE}/${MACHINE}/genassym.c .endif lint: ${SRCS} ${LINT} ${LINTKERNFLAGS} ${CFLAGS:M-[DILU]*} ${.ALLSRC:M*.c} .if defined(KERNBUILDDIR) ${OBJS}: opt_global.h .endif CLEANDEPENDFILES+= ${_ILINKS} # .depend needs include links so we remove them only together. cleanilinks: rm -f ${_ILINKS} OBJS_DEPEND_GUESS+= ${SRCS:M*.h} + +.include + .if ${MK_FAST_DEPEND} == "no" && !exists(${.OBJDIR}/${DEPENDFILE}) ${OBJS}: ${OBJS_DEPEND_GUESS} .endif -.include .include .include .include "kern.mk" Index: projects/release-pkg/sys/conf =================================================================== --- projects/release-pkg/sys/conf (revision 296327) +++ projects/release-pkg/sys/conf (revision 296328) Property changes on: projects/release-pkg/sys/conf ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/sys/conf:r296315-296327 Index: projects/release-pkg/sys/dev/pci/pci_host_generic.c =================================================================== --- projects/release-pkg/sys/dev/pci/pci_host_generic.c (revision 296327) +++ projects/release-pkg/sys/dev/pci/pci_host_generic.c (revision 296328) @@ -1,815 +1,818 @@ /*- * Copyright (c) 2015 Ruslan Bukin * Copyright (c) 2014 The FreeBSD Foundation * All rights reserved. * * This software was developed by Semihalf under * the sponsorship of the FreeBSD Foundation. * * 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. */ /* Generic ECAM PCIe driver */ #include __FBSDID("$FreeBSD$"); #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 "pcib_if.h" /* Assembling ECAM Configuration Address */ #define PCIE_BUS_SHIFT 20 #define PCIE_SLOT_SHIFT 15 #define PCIE_FUNC_SHIFT 12 #define PCIE_BUS_MASK 0xFF #define PCIE_SLOT_MASK 0x1F #define PCIE_FUNC_MASK 0x07 #define PCIE_REG_MASK 0xFFF #define PCIE_ADDR_OFFSET(bus, slot, func, reg) \ ((((bus) & PCIE_BUS_MASK) << PCIE_BUS_SHIFT) | \ (((slot) & PCIE_SLOT_MASK) << PCIE_SLOT_SHIFT) | \ (((func) & PCIE_FUNC_MASK) << PCIE_FUNC_SHIFT) | \ ((reg) & PCIE_REG_MASK)) #define PCI_IO_WINDOW_OFFSET 0x1000 #define SPACE_CODE_SHIFT 24 #define SPACE_CODE_MASK 0x3 #define SPACE_CODE_IO_SPACE 0x1 #define PROPS_CELL_SIZE 1 #define PCI_ADDR_CELL_SIZE 2 /* OFW bus interface */ struct generic_pcie_ofw_devinfo { struct ofw_bus_devinfo di_dinfo; struct resource_list di_rl; }; /* Forward prototypes */ static int generic_pcie_probe(device_t dev); static int parse_pci_mem_ranges(struct generic_pcie_softc *sc); static uint32_t generic_pcie_read_config(device_t dev, u_int bus, u_int slot, u_int func, u_int reg, int bytes); static void generic_pcie_write_config(device_t dev, u_int bus, u_int slot, u_int func, u_int reg, uint32_t val, int bytes); static int generic_pcie_maxslots(device_t dev); static int generic_pcie_read_ivar(device_t dev, device_t child, int index, uintptr_t *result); static int generic_pcie_write_ivar(device_t dev, device_t child, int index, uintptr_t value); static struct resource *generic_pcie_alloc_resource_ofw(device_t, device_t, int, int *, rman_res_t, rman_res_t, rman_res_t, u_int); static struct resource *generic_pcie_alloc_resource_pcie(device_t dev, device_t child, int type, int *rid, rman_res_t start, rman_res_t end, rman_res_t count, u_int flags); static int generic_pcie_release_resource(device_t dev, device_t child, int type, int rid, struct resource *res); static int generic_pcie_release_resource_ofw(device_t, device_t, int, int, struct resource *); static int generic_pcie_release_resource_pcie(device_t, device_t, int, int, struct resource *); static int generic_pcie_ofw_bus_attach(device_t); static const struct ofw_bus_devinfo *generic_pcie_ofw_get_devinfo(device_t, device_t); static __inline void get_addr_size_cells(phandle_t node, pcell_t *addr_cells, pcell_t *size_cells) { *addr_cells = 2; /* Find address cells if present */ OF_getencprop(node, "#address-cells", addr_cells, sizeof(*addr_cells)); *size_cells = 2; /* Find size cells if present */ OF_getencprop(node, "#size-cells", size_cells, sizeof(*size_cells)); } static int generic_pcie_probe(device_t dev) { if (!ofw_bus_status_okay(dev)) return (ENXIO); if (ofw_bus_is_compatible(dev, "pci-host-ecam-generic")) { device_set_desc(dev, "Generic PCI host controller"); return (BUS_PROBE_GENERIC); } + if (ofw_bus_is_compatible(dev, "arm,gem5_pcie")) { + device_set_desc(dev, "GEM5 PCIe host controller"); + return (BUS_PROBE_DEFAULT); + } return (ENXIO); } int pci_host_generic_attach(device_t dev) { struct generic_pcie_softc *sc; uint64_t phys_base; uint64_t pci_base; uint64_t size; int error; int tuple; int rid; sc = device_get_softc(dev); sc->dev = dev; /* Retrieve 'ranges' property from FDT */ if (bootverbose) device_printf(dev, "parsing FDT for ECAM%d:\n", sc->ecam); if (parse_pci_mem_ranges(sc)) return (ENXIO); /* Attach OFW bus */ if (generic_pcie_ofw_bus_attach(dev) != 0) return (ENXIO); rid = 0; sc->res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid, RF_ACTIVE); if (sc->res == NULL) { device_printf(dev, "could not map memory.\n"); return (ENXIO); } sc->bst = rman_get_bustag(sc->res); sc->bsh = rman_get_bushandle(sc->res); sc->mem_rman.rm_type = RMAN_ARRAY; sc->mem_rman.rm_descr = "PCIe Memory"; sc->io_rman.rm_type = RMAN_ARRAY; sc->io_rman.rm_descr = "PCIe IO window"; /* Initialize rman and allocate memory regions */ error = rman_init(&sc->mem_rman); if (error) { device_printf(dev, "rman_init() failed. error = %d\n", error); return (error); } error = rman_init(&sc->io_rman); if (error) { device_printf(dev, "rman_init() failed. error = %d\n", error); return (error); } for (tuple = 0; tuple < MAX_RANGES_TUPLES; tuple++) { phys_base = sc->ranges[tuple].phys_base; pci_base = sc->ranges[tuple].pci_base; size = sc->ranges[tuple].size; if (phys_base == 0 || size == 0) continue; /* empty range element */ if (sc->ranges[tuple].flags & FLAG_MEM) { error = rman_manage_region(&sc->mem_rman, - phys_base, - phys_base + size); + phys_base, phys_base + size - 1); } else if (sc->ranges[tuple].flags & FLAG_IO) { error = rman_manage_region(&sc->io_rman, - pci_base + PCI_IO_WINDOW_OFFSET, - pci_base + PCI_IO_WINDOW_OFFSET + size); + pci_base + PCI_IO_WINDOW_OFFSET, + pci_base + PCI_IO_WINDOW_OFFSET + size - 1); } else continue; if (error) { device_printf(dev, "rman_manage_region() failed." "error = %d\n", error); rman_fini(&sc->mem_rman); return (error); } } ofw_bus_setup_iinfo(ofw_bus_get_node(dev), &sc->pci_iinfo, sizeof(cell_t)); device_add_child(dev, "pci", -1); return (bus_generic_attach(dev)); } static int parse_pci_mem_ranges(struct generic_pcie_softc *sc) { pcell_t pci_addr_cells, parent_addr_cells; pcell_t attributes, size_cells; cell_t *base_ranges; int nbase_ranges; phandle_t node; int i, j, k; int tuple; node = ofw_bus_get_node(sc->dev); OF_getencprop(node, "#address-cells", &pci_addr_cells, sizeof(pci_addr_cells)); OF_getencprop(node, "#size-cells", &size_cells, sizeof(size_cells)); OF_getencprop(OF_parent(node), "#address-cells", &parent_addr_cells, sizeof(parent_addr_cells)); if (parent_addr_cells != 2 || pci_addr_cells != 3 || size_cells != 2) { device_printf(sc->dev, "Unexpected number of address or size cells in FDT\n"); return (ENXIO); } nbase_ranges = OF_getproplen(node, "ranges"); sc->nranges = nbase_ranges / sizeof(cell_t) / (parent_addr_cells + pci_addr_cells + size_cells); base_ranges = malloc(nbase_ranges, M_DEVBUF, M_WAITOK); OF_getencprop(node, "ranges", base_ranges, nbase_ranges); for (i = 0, j = 0; i < sc->nranges; i++) { attributes = (base_ranges[j++] >> SPACE_CODE_SHIFT) & \ SPACE_CODE_MASK; if (attributes == SPACE_CODE_IO_SPACE) { sc->ranges[i].flags |= FLAG_IO; } else { sc->ranges[i].flags |= FLAG_MEM; } sc->ranges[i].pci_base = 0; for (k = 0; k < (pci_addr_cells - 1); k++) { sc->ranges[i].pci_base <<= 32; sc->ranges[i].pci_base |= base_ranges[j++]; } sc->ranges[i].phys_base = 0; for (k = 0; k < parent_addr_cells; k++) { sc->ranges[i].phys_base <<= 32; sc->ranges[i].phys_base |= base_ranges[j++]; } sc->ranges[i].size = 0; for (k = 0; k < size_cells; k++) { sc->ranges[i].size <<= 32; sc->ranges[i].size |= base_ranges[j++]; } } for (; i < MAX_RANGES_TUPLES; i++) { /* zero-fill remaining tuples to mark empty elements in array */ sc->ranges[i].pci_base = 0; sc->ranges[i].phys_base = 0; sc->ranges[i].size = 0; } if (bootverbose) { for (tuple = 0; tuple < MAX_RANGES_TUPLES; tuple++) { device_printf(sc->dev, "\tPCI addr: 0x%jx, CPU addr: 0x%jx, Size: 0x%jx\n", sc->ranges[tuple].pci_base, sc->ranges[tuple].phys_base, sc->ranges[tuple].size); } } free(base_ranges, M_DEVBUF); return (0); } static uint32_t generic_pcie_read_config(device_t dev, u_int bus, u_int slot, u_int func, u_int reg, int bytes) { struct generic_pcie_softc *sc; bus_space_handle_t h; bus_space_tag_t t; uint64_t offset; uint32_t data; if ((bus > PCI_BUSMAX) || (slot > PCI_SLOTMAX) || (func > PCI_FUNCMAX) || (reg > PCIE_REGMAX)) return (~0U); sc = device_get_softc(dev); offset = PCIE_ADDR_OFFSET(bus, slot, func, reg); t = sc->bst; h = sc->bsh; switch (bytes) { case 1: data = bus_space_read_1(t, h, offset); break; case 2: data = le16toh(bus_space_read_2(t, h, offset)); break; case 4: data = le32toh(bus_space_read_4(t, h, offset)); break; default: return (~0U); } return (data); } static void generic_pcie_write_config(device_t dev, u_int bus, u_int slot, u_int func, u_int reg, uint32_t val, int bytes) { struct generic_pcie_softc *sc; bus_space_handle_t h; bus_space_tag_t t; uint64_t offset; if ((bus > PCI_BUSMAX) || (slot > PCI_SLOTMAX) || (func > PCI_FUNCMAX) || (reg > PCIE_REGMAX)) return; sc = device_get_softc(dev); offset = PCIE_ADDR_OFFSET(bus, slot, func, reg); t = sc->bst; h = sc->bsh; switch (bytes) { case 1: bus_space_write_1(t, h, offset, val); break; case 2: bus_space_write_2(t, h, offset, htole16(val)); break; case 4: bus_space_write_4(t, h, offset, htole32(val)); break; default: return; } } static int generic_pcie_maxslots(device_t dev) { return (31); /* max slots per bus acc. to standard */ } static int generic_pcie_route_interrupt(device_t bus, device_t dev, int pin) { struct generic_pcie_softc *sc; struct ofw_pci_register reg; uint32_t pintr, mintr[2]; phandle_t iparent; int intrcells; sc = device_get_softc(bus); pintr = pin; bzero(®, sizeof(reg)); reg.phys_hi = (pci_get_bus(dev) << OFW_PCI_PHYS_HI_BUSSHIFT) | (pci_get_slot(dev) << OFW_PCI_PHYS_HI_DEVICESHIFT) | (pci_get_function(dev) << OFW_PCI_PHYS_HI_FUNCTIONSHIFT); intrcells = ofw_bus_lookup_imap(ofw_bus_get_node(dev), &sc->pci_iinfo, ®, sizeof(reg), &pintr, sizeof(pintr), mintr, sizeof(mintr), &iparent); if (intrcells) { pintr = ofw_bus_map_intr(dev, iparent, intrcells, mintr); return (pintr); } device_printf(bus, "could not route pin %d for device %d.%d\n", pin, pci_get_slot(dev), pci_get_function(dev)); return (PCI_INVALID_IRQ); } static int generic_pcie_read_ivar(device_t dev, device_t child, int index, uintptr_t *result) { struct generic_pcie_softc *sc; int secondary_bus; sc = device_get_softc(dev); if (index == PCIB_IVAR_BUS) { /* this pcib adds only pci bus 0 as child */ secondary_bus = 0; *result = secondary_bus; return (0); } if (index == PCIB_IVAR_DOMAIN) { *result = sc->ecam; return (0); } if (bootverbose) device_printf(dev, "ERROR: Unknown index %d.\n", index); return (ENOENT); } static int generic_pcie_write_ivar(device_t dev, device_t child, int index, uintptr_t value) { return (ENOENT); } static struct rman * generic_pcie_rman(struct generic_pcie_softc *sc, int type) { switch (type) { case SYS_RES_IOPORT: return (&sc->io_rman); case SYS_RES_MEMORY: return (&sc->mem_rman); default: break; } return (NULL); } static int generic_pcie_release_resource_pcie(device_t dev, device_t child, int type, int rid, struct resource *res) { struct generic_pcie_softc *sc; struct rman *rm; sc = device_get_softc(dev); rm = generic_pcie_rman(sc, type); if (rm != NULL) { KASSERT(rman_is_region_manager(res, rm), ("rman mismatch")); rman_release_resource(res); } return (bus_generic_release_resource(dev, child, type, rid, res)); } static int generic_pcie_release_resource(device_t dev, device_t child, int type, int rid, struct resource *res) { /* For PCIe devices that do not have FDT nodes, use PCIB method */ if ((int)ofw_bus_get_node(child) <= 0) { return (generic_pcie_release_resource_pcie(dev, child, type, rid, res)); } /* For other devices use OFW method */ return (generic_pcie_release_resource_ofw(dev, child, type, rid, res)); } struct resource * pci_host_generic_alloc_resource(device_t dev, device_t child, int type, int *rid, rman_res_t start, rman_res_t end, rman_res_t count, u_int flags) { /* For PCIe devices that do not have FDT nodes, use PCIB method */ if ((int)ofw_bus_get_node(child) <= 0) return (generic_pcie_alloc_resource_pcie(dev, child, type, rid, start, end, count, flags)); /* For other devices use OFW method */ return (generic_pcie_alloc_resource_ofw(dev, child, type, rid, start, end, count, flags)); } static struct resource * generic_pcie_alloc_resource_pcie(device_t dev, device_t child, int type, int *rid, rman_res_t start, rman_res_t end, rman_res_t count, u_int flags) { struct generic_pcie_softc *sc; struct resource *res; struct rman *rm; sc = device_get_softc(dev); rm = generic_pcie_rman(sc, type); if (rm == NULL) return (BUS_ALLOC_RESOURCE(device_get_parent(dev), dev, type, rid, start, end, count, flags)); if (bootverbose) { device_printf(dev, "rman_reserve_resource: start=%#lx, end=%#lx, count=%#lx\n", start, end, count); } res = rman_reserve_resource(rm, start, end, count, flags, child); if (res == NULL) goto fail; rman_set_rid(res, *rid); if (flags & RF_ACTIVE) if (bus_activate_resource(child, type, *rid, res)) { rman_release_resource(res); goto fail; } return (res); fail: device_printf(dev, "%s FAIL: type=%d, rid=%d, " "start=%016lx, end=%016lx, count=%016lx, flags=%x\n", __func__, type, *rid, start, end, count, flags); return (NULL); } static int generic_pcie_adjust_resource(device_t dev, device_t child, int type, struct resource *res, rman_res_t start, rman_res_t end) { struct generic_pcie_softc *sc; struct rman *rm; sc = device_get_softc(dev); rm = generic_pcie_rman(sc, type); if (rm != NULL) return (rman_adjust_resource(res, start, end)); return (bus_generic_adjust_resource(dev, child, type, res, start, end)); } static int generic_pcie_activate_resource(device_t dev, device_t child, int type, int rid, struct resource *r) { struct generic_pcie_softc *sc; uint64_t phys_base; uint64_t pci_base; uint64_t size; int found; int res; int i; sc = device_get_softc(dev); if ((res = rman_activate_resource(r)) != 0) return (res); switch(type) { case SYS_RES_IOPORT: found = 0; for (i = 0; i < MAX_RANGES_TUPLES; i++) { pci_base = sc->ranges[i].pci_base; phys_base = sc->ranges[i].phys_base; size = sc->ranges[i].size; if ((rid > pci_base) && (rid < (pci_base + size))) { found = 1; break; } } if (found) { rman_set_start(r, rman_get_start(r) + phys_base); BUS_ACTIVATE_RESOURCE(device_get_parent(dev), child, type, rid, r); } else { device_printf(dev, "Failed to activate IOPORT resource\n"); res = 0; } break; case SYS_RES_MEMORY: BUS_ACTIVATE_RESOURCE(device_get_parent(dev), child, type, rid, r); break; default: break; } return (res); } static int generic_pcie_deactivate_resource(device_t dev, device_t child, int type, int rid, struct resource *r) { struct generic_pcie_softc *sc; vm_offset_t vaddr; int res; sc = device_get_softc(dev); if ((res = rman_deactivate_resource(r)) != 0) return (res); switch(type) { case SYS_RES_IOPORT: case SYS_RES_MEMORY: vaddr = (vm_offset_t)rman_get_virtual(r); pmap_unmapdev(vaddr, rman_get_size(r)); break; default: break; } return (res); } static device_method_t generic_pcie_methods[] = { DEVMETHOD(device_probe, generic_pcie_probe), DEVMETHOD(device_attach, pci_host_generic_attach), DEVMETHOD(bus_read_ivar, generic_pcie_read_ivar), DEVMETHOD(bus_write_ivar, generic_pcie_write_ivar), DEVMETHOD(bus_alloc_resource, pci_host_generic_alloc_resource), DEVMETHOD(bus_adjust_resource, generic_pcie_adjust_resource), DEVMETHOD(bus_release_resource, generic_pcie_release_resource), DEVMETHOD(bus_activate_resource, generic_pcie_activate_resource), DEVMETHOD(bus_deactivate_resource, generic_pcie_deactivate_resource), DEVMETHOD(bus_setup_intr, bus_generic_setup_intr), DEVMETHOD(bus_teardown_intr, bus_generic_teardown_intr), /* pcib interface */ DEVMETHOD(pcib_maxslots, generic_pcie_maxslots), DEVMETHOD(pcib_route_interrupt, generic_pcie_route_interrupt), DEVMETHOD(pcib_read_config, generic_pcie_read_config), DEVMETHOD(pcib_write_config, generic_pcie_write_config), #if defined(__aarch64__) DEVMETHOD(pcib_alloc_msi, arm_alloc_msi), DEVMETHOD(pcib_release_msi, arm_release_msi), DEVMETHOD(pcib_alloc_msix, arm_alloc_msix), DEVMETHOD(pcib_release_msix, arm_release_msix), DEVMETHOD(pcib_map_msi, arm_map_msi), #endif /* ofw_bus interface */ DEVMETHOD(ofw_bus_get_devinfo, generic_pcie_ofw_get_devinfo), DEVMETHOD(ofw_bus_get_compat, ofw_bus_gen_get_compat), DEVMETHOD(ofw_bus_get_model, ofw_bus_gen_get_model), DEVMETHOD(ofw_bus_get_name, ofw_bus_gen_get_name), DEVMETHOD(ofw_bus_get_node, ofw_bus_gen_get_node), DEVMETHOD(ofw_bus_get_type, ofw_bus_gen_get_type), DEVMETHOD_END }; static const struct ofw_bus_devinfo * generic_pcie_ofw_get_devinfo(device_t bus __unused, device_t child) { struct generic_pcie_ofw_devinfo *di; di = device_get_ivars(child); return (&di->di_dinfo); } static struct resource * generic_pcie_alloc_resource_ofw(device_t bus, device_t child, int type, int *rid, rman_res_t start, rman_res_t end, rman_res_t count, u_int flags) { struct generic_pcie_softc *sc; struct generic_pcie_ofw_devinfo *di; struct resource_list_entry *rle; int i; sc = device_get_softc(bus); if (RMAN_IS_DEFAULT_RANGE(start, end)) { if ((di = device_get_ivars(child)) == NULL) return (NULL); if (type == SYS_RES_IOPORT) type = SYS_RES_MEMORY; /* Find defaults for this rid */ rle = resource_list_find(&di->di_rl, type, *rid); if (rle == NULL) return (NULL); start = rle->start; end = rle->end; count = rle->count; } if (type == SYS_RES_MEMORY) { /* Remap through ranges property */ for (i = 0; i < MAX_RANGES_TUPLES; i++) { if (start >= sc->ranges[i].phys_base && end < sc->ranges[i].pci_base + sc->ranges[i].size) { start -= sc->ranges[i].phys_base; start += sc->ranges[i].pci_base; end -= sc->ranges[i].phys_base; end += sc->ranges[i].pci_base; break; } } if (i == MAX_RANGES_TUPLES) { device_printf(bus, "Could not map resource " "%#lx-%#lx\n", start, end); return (NULL); } } return (bus_generic_alloc_resource(bus, child, type, rid, start, end, count, flags)); } static int generic_pcie_release_resource_ofw(device_t bus, device_t child, int type, int rid, struct resource *res) { return (bus_generic_release_resource(bus, child, type, rid, res)); } /* Helper functions */ static int generic_pcie_ofw_bus_attach(device_t dev) { struct generic_pcie_ofw_devinfo *di; device_t child; phandle_t parent, node; pcell_t addr_cells, size_cells; parent = ofw_bus_get_node(dev); if (parent > 0) { get_addr_size_cells(parent, &addr_cells, &size_cells); /* Iterate through all bus subordinates */ for (node = OF_child(parent); node > 0; node = OF_peer(node)) { /* Allocate and populate devinfo. */ di = malloc(sizeof(*di), M_DEVBUF, M_WAITOK | M_ZERO); if (ofw_bus_gen_setup_devinfo(&di->di_dinfo, node) != 0) { free(di, M_DEVBUF); continue; } /* Initialize and populate resource list. */ resource_list_init(&di->di_rl); ofw_bus_reg_to_rl(dev, node, addr_cells, size_cells, &di->di_rl); ofw_bus_intr_to_rl(dev, node, &di->di_rl, NULL); /* Add newbus device for this FDT node */ child = device_add_child(dev, NULL, -1); if (child == NULL) { resource_list_free(&di->di_rl); ofw_bus_gen_destroy_devinfo(&di->di_dinfo); free(di, M_DEVBUF); continue; } device_set_ivars(child, di); } } return (0); } DEFINE_CLASS_0(pcib, generic_pcie_driver, generic_pcie_methods, sizeof(struct generic_pcie_softc)); devclass_t generic_pcie_devclass; DRIVER_MODULE(pcib, simplebus, generic_pcie_driver, generic_pcie_devclass, 0, 0); DRIVER_MODULE(pcib, ofwbus, generic_pcie_driver, generic_pcie_devclass, 0, 0); Index: projects/release-pkg/sys/kern/kern_timeout.c =================================================================== --- projects/release-pkg/sys/kern/kern_timeout.c (revision 296327) +++ projects/release-pkg/sys/kern/kern_timeout.c (revision 296328) @@ -1,1617 +1,1617 @@ /*- * Copyright (c) 1982, 1986, 1991, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 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: @(#)kern_clock.c 8.5 (Berkeley) 1/21/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_callout_profiling.h" #if defined(__arm__) #include "opt_timer.h" #endif #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef SMP #include #endif #ifndef NO_EVENTTIMERS DPCPU_DECLARE(sbintime_t, hardclocktime); #endif SDT_PROVIDER_DEFINE(callout_execute); SDT_PROBE_DEFINE1(callout_execute, , , callout__start, "struct callout *"); SDT_PROBE_DEFINE1(callout_execute, , , callout__end, "struct callout *"); #ifdef CALLOUT_PROFILING static int avg_depth; SYSCTL_INT(_debug, OID_AUTO, to_avg_depth, CTLFLAG_RD, &avg_depth, 0, "Average number of items examined per softclock call. Units = 1/1000"); static int avg_gcalls; SYSCTL_INT(_debug, OID_AUTO, to_avg_gcalls, CTLFLAG_RD, &avg_gcalls, 0, "Average number of Giant callouts made per softclock call. Units = 1/1000"); static int avg_lockcalls; SYSCTL_INT(_debug, OID_AUTO, to_avg_lockcalls, CTLFLAG_RD, &avg_lockcalls, 0, "Average number of lock callouts made per softclock call. Units = 1/1000"); static int avg_mpcalls; SYSCTL_INT(_debug, OID_AUTO, to_avg_mpcalls, CTLFLAG_RD, &avg_mpcalls, 0, "Average number of MP callouts made per softclock call. Units = 1/1000"); static int avg_depth_dir; SYSCTL_INT(_debug, OID_AUTO, to_avg_depth_dir, CTLFLAG_RD, &avg_depth_dir, 0, "Average number of direct callouts examined per callout_process call. " "Units = 1/1000"); static int avg_lockcalls_dir; SYSCTL_INT(_debug, OID_AUTO, to_avg_lockcalls_dir, CTLFLAG_RD, &avg_lockcalls_dir, 0, "Average number of lock direct callouts made per " "callout_process call. Units = 1/1000"); static int avg_mpcalls_dir; SYSCTL_INT(_debug, OID_AUTO, to_avg_mpcalls_dir, CTLFLAG_RD, &avg_mpcalls_dir, 0, "Average number of MP direct callouts made per callout_process call. " "Units = 1/1000"); #endif static int ncallout; SYSCTL_INT(_kern, OID_AUTO, ncallout, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &ncallout, 0, "Number of entries in callwheel and size of timeout() preallocation"); #ifdef RSS static int pin_default_swi = 1; static int pin_pcpu_swi = 1; #else static int pin_default_swi = 0; static int pin_pcpu_swi = 0; #endif SYSCTL_INT(_kern, OID_AUTO, pin_default_swi, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &pin_default_swi, 0, "Pin the default (non-per-cpu) swi (shared with PCPU 0 swi)"); SYSCTL_INT(_kern, OID_AUTO, pin_pcpu_swi, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &pin_pcpu_swi, 0, "Pin the per-CPU swis (except PCPU 0, which is also default"); /* * TODO: * allocate more timeout table slots when table overflows. */ u_int callwheelsize, callwheelmask; /* * The callout cpu exec entities represent informations necessary for * describing the state of callouts currently running on the CPU and the ones * necessary for migrating callouts to the new callout cpu. In particular, * the first entry of the array cc_exec_entity holds informations for callout * running in SWI thread context, while the second one holds informations * for callout running directly from hardware interrupt context. * The cached informations are very important for deferring migration when * the migrating callout is already running. */ struct cc_exec { struct callout *cc_curr; void (*cc_drain)(void *); #ifdef SMP void (*ce_migration_func)(void *); void *ce_migration_arg; int ce_migration_cpu; sbintime_t ce_migration_time; sbintime_t ce_migration_prec; #endif bool cc_cancel; bool cc_waiting; }; /* * There is one struct callout_cpu per cpu, holding all relevant * state for the callout processing thread on the individual CPU. */ struct callout_cpu { struct mtx_padalign cc_lock; struct cc_exec cc_exec_entity[2]; struct callout *cc_next; struct callout *cc_callout; struct callout_list *cc_callwheel; struct callout_tailq cc_expireq; struct callout_slist cc_callfree; sbintime_t cc_firstevent; sbintime_t cc_lastscan; void *cc_cookie; u_int cc_bucket; u_int cc_inited; char cc_ktr_event_name[20]; }; #define callout_migrating(c) ((c)->c_iflags & CALLOUT_DFRMIGRATION) #define cc_exec_curr(cc, dir) cc->cc_exec_entity[dir].cc_curr #define cc_exec_drain(cc, dir) cc->cc_exec_entity[dir].cc_drain #define cc_exec_next(cc) cc->cc_next #define cc_exec_cancel(cc, dir) cc->cc_exec_entity[dir].cc_cancel #define cc_exec_waiting(cc, dir) cc->cc_exec_entity[dir].cc_waiting #ifdef SMP #define cc_migration_func(cc, dir) cc->cc_exec_entity[dir].ce_migration_func #define cc_migration_arg(cc, dir) cc->cc_exec_entity[dir].ce_migration_arg #define cc_migration_cpu(cc, dir) cc->cc_exec_entity[dir].ce_migration_cpu #define cc_migration_time(cc, dir) cc->cc_exec_entity[dir].ce_migration_time #define cc_migration_prec(cc, dir) cc->cc_exec_entity[dir].ce_migration_prec struct callout_cpu cc_cpu[MAXCPU]; #define CPUBLOCK MAXCPU #define CC_CPU(cpu) (&cc_cpu[(cpu)]) #define CC_SELF() CC_CPU(PCPU_GET(cpuid)) #else struct callout_cpu cc_cpu; #define CC_CPU(cpu) &cc_cpu #define CC_SELF() &cc_cpu #endif #define CC_LOCK(cc) mtx_lock_spin(&(cc)->cc_lock) #define CC_UNLOCK(cc) mtx_unlock_spin(&(cc)->cc_lock) #define CC_LOCK_ASSERT(cc) mtx_assert(&(cc)->cc_lock, MA_OWNED) static int timeout_cpu; static void callout_cpu_init(struct callout_cpu *cc, int cpu); static void softclock_call_cc(struct callout *c, struct callout_cpu *cc, #ifdef CALLOUT_PROFILING int *mpcalls, int *lockcalls, int *gcalls, #endif int direct); static MALLOC_DEFINE(M_CALLOUT, "callout", "Callout datastructures"); /** * Locked by cc_lock: * cc_curr - If a callout is in progress, it is cc_curr. * If cc_curr is non-NULL, threads waiting in * callout_drain() will be woken up as soon as the * relevant callout completes. * cc_cancel - Changing to 1 with both callout_lock and cc_lock held * guarantees that the current callout will not run. * The softclock() function sets this to 0 before it * drops callout_lock to acquire c_lock, and it calls * the handler only if curr_cancelled is still 0 after * cc_lock is successfully acquired. * cc_waiting - If a thread is waiting in callout_drain(), then * callout_wait is nonzero. Set only when * cc_curr is non-NULL. */ /* * Resets the execution entity tied to a specific callout cpu. */ static void cc_cce_cleanup(struct callout_cpu *cc, int direct) { cc_exec_curr(cc, direct) = NULL; cc_exec_cancel(cc, direct) = false; cc_exec_waiting(cc, direct) = false; #ifdef SMP cc_migration_cpu(cc, direct) = CPUBLOCK; cc_migration_time(cc, direct) = 0; cc_migration_prec(cc, direct) = 0; cc_migration_func(cc, direct) = NULL; cc_migration_arg(cc, direct) = NULL; #endif } /* * Checks if migration is requested by a specific callout cpu. */ static int cc_cce_migrating(struct callout_cpu *cc, int direct) { #ifdef SMP return (cc_migration_cpu(cc, direct) != CPUBLOCK); #else return (0); #endif } /* * Kernel low level callwheel initialization * called on cpu0 during kernel startup. */ static void callout_callwheel_init(void *dummy) { struct callout_cpu *cc; /* * Calculate the size of the callout wheel and the preallocated * timeout() structures. * XXX: Clip callout to result of previous function of maxusers * maximum 384. This is still huge, but acceptable. */ memset(CC_CPU(0), 0, sizeof(cc_cpu)); ncallout = imin(16 + maxproc + maxfiles, 18508); TUNABLE_INT_FETCH("kern.ncallout", &ncallout); /* * Calculate callout wheel size, should be next power of two higher * than 'ncallout'. */ callwheelsize = 1 << fls(ncallout); callwheelmask = callwheelsize - 1; /* * Fetch whether we're pinning the swi's or not. */ TUNABLE_INT_FETCH("kern.pin_default_swi", &pin_default_swi); TUNABLE_INT_FETCH("kern.pin_pcpu_swi", &pin_pcpu_swi); /* * Only cpu0 handles timeout(9) and receives a preallocation. * * XXX: Once all timeout(9) consumers are converted this can * be removed. */ timeout_cpu = PCPU_GET(cpuid); cc = CC_CPU(timeout_cpu); cc->cc_callout = malloc(ncallout * sizeof(struct callout), M_CALLOUT, M_WAITOK); callout_cpu_init(cc, timeout_cpu); } SYSINIT(callwheel_init, SI_SUB_CPU, SI_ORDER_ANY, callout_callwheel_init, NULL); /* * Initialize the per-cpu callout structures. */ static void callout_cpu_init(struct callout_cpu *cc, int cpu) { struct callout *c; int i; mtx_init(&cc->cc_lock, "callout", NULL, MTX_SPIN | MTX_RECURSE); SLIST_INIT(&cc->cc_callfree); cc->cc_inited = 1; cc->cc_callwheel = malloc(sizeof(struct callout_list) * callwheelsize, M_CALLOUT, M_WAITOK); for (i = 0; i < callwheelsize; i++) LIST_INIT(&cc->cc_callwheel[i]); TAILQ_INIT(&cc->cc_expireq); cc->cc_firstevent = SBT_MAX; for (i = 0; i < 2; i++) cc_cce_cleanup(cc, i); snprintf(cc->cc_ktr_event_name, sizeof(cc->cc_ktr_event_name), "callwheel cpu %d", cpu); if (cc->cc_callout == NULL) /* Only cpu0 handles timeout(9) */ return; for (i = 0; i < ncallout; i++) { c = &cc->cc_callout[i]; callout_init(c, 0); c->c_iflags = CALLOUT_LOCAL_ALLOC; SLIST_INSERT_HEAD(&cc->cc_callfree, c, c_links.sle); } } #ifdef SMP /* * Switches the cpu tied to a specific callout. * The function expects a locked incoming callout cpu and returns with * locked outcoming callout cpu. */ static struct callout_cpu * callout_cpu_switch(struct callout *c, struct callout_cpu *cc, int new_cpu) { struct callout_cpu *new_cc; MPASS(c != NULL && cc != NULL); CC_LOCK_ASSERT(cc); /* * Avoid interrupts and preemption firing after the callout cpu * is blocked in order to avoid deadlocks as the new thread * may be willing to acquire the callout cpu lock. */ c->c_cpu = CPUBLOCK; spinlock_enter(); CC_UNLOCK(cc); new_cc = CC_CPU(new_cpu); CC_LOCK(new_cc); spinlock_exit(); c->c_cpu = new_cpu; return (new_cc); } #endif /* * Start standard softclock thread. */ static void start_softclock(void *dummy) { struct callout_cpu *cc; char name[MAXCOMLEN]; #ifdef SMP int cpu; struct intr_event *ie; #endif cc = CC_CPU(timeout_cpu); snprintf(name, sizeof(name), "clock (%d)", timeout_cpu); if (swi_add(&clk_intr_event, name, softclock, cc, SWI_CLOCK, INTR_MPSAFE, &cc->cc_cookie)) panic("died while creating standard software ithreads"); if (pin_default_swi && (intr_event_bind(clk_intr_event, timeout_cpu) != 0)) { printf("%s: timeout clock couldn't be pinned to cpu %d\n", __func__, timeout_cpu); } #ifdef SMP CPU_FOREACH(cpu) { if (cpu == timeout_cpu) continue; cc = CC_CPU(cpu); cc->cc_callout = NULL; /* Only cpu0 handles timeout(9). */ callout_cpu_init(cc, cpu); snprintf(name, sizeof(name), "clock (%d)", cpu); ie = NULL; if (swi_add(&ie, name, softclock, cc, SWI_CLOCK, INTR_MPSAFE, &cc->cc_cookie)) panic("died while creating standard software ithreads"); if (pin_pcpu_swi && (intr_event_bind(ie, cpu) != 0)) { printf("%s: per-cpu clock couldn't be pinned to " "cpu %d\n", __func__, cpu); } } #endif } SYSINIT(start_softclock, SI_SUB_SOFTINTR, SI_ORDER_FIRST, start_softclock, NULL); #define CC_HASH_SHIFT 8 static inline u_int callout_hash(sbintime_t sbt) { return (sbt >> (32 - CC_HASH_SHIFT)); } static inline u_int callout_get_bucket(sbintime_t sbt) { return (callout_hash(sbt) & callwheelmask); } void callout_process(sbintime_t now) { struct callout *tmp, *tmpn; struct callout_cpu *cc; struct callout_list *sc; sbintime_t first, last, max, tmp_max; uint32_t lookahead; u_int firstb, lastb, nowb; #ifdef CALLOUT_PROFILING int depth_dir = 0, mpcalls_dir = 0, lockcalls_dir = 0; #endif cc = CC_SELF(); mtx_lock_spin_flags(&cc->cc_lock, MTX_QUIET); /* Compute the buckets of the last scan and present times. */ firstb = callout_hash(cc->cc_lastscan); cc->cc_lastscan = now; nowb = callout_hash(now); /* Compute the last bucket and minimum time of the bucket after it. */ if (nowb == firstb) lookahead = (SBT_1S / 16); else if (nowb - firstb == 1) lookahead = (SBT_1S / 8); else lookahead = (SBT_1S / 2); first = last = now; first += (lookahead / 2); last += lookahead; last &= (0xffffffffffffffffLLU << (32 - CC_HASH_SHIFT)); lastb = callout_hash(last) - 1; max = last; /* * Check if we wrapped around the entire wheel from the last scan. * In case, we need to scan entirely the wheel for pending callouts. */ if (lastb - firstb >= callwheelsize) { lastb = firstb + callwheelsize - 1; if (nowb - firstb >= callwheelsize) nowb = lastb; } /* Iterate callwheel from firstb to nowb and then up to lastb. */ do { sc = &cc->cc_callwheel[firstb & callwheelmask]; tmp = LIST_FIRST(sc); while (tmp != NULL) { /* Run the callout if present time within allowed. */ if (tmp->c_time <= now) { /* * Consumer told us the callout may be run * directly from hardware interrupt context. */ if (tmp->c_iflags & CALLOUT_DIRECT) { #ifdef CALLOUT_PROFILING ++depth_dir; #endif cc_exec_next(cc) = LIST_NEXT(tmp, c_links.le); cc->cc_bucket = firstb & callwheelmask; LIST_REMOVE(tmp, c_links.le); softclock_call_cc(tmp, cc, #ifdef CALLOUT_PROFILING &mpcalls_dir, &lockcalls_dir, NULL, #endif 1); tmp = cc_exec_next(cc); cc_exec_next(cc) = NULL; } else { tmpn = LIST_NEXT(tmp, c_links.le); LIST_REMOVE(tmp, c_links.le); TAILQ_INSERT_TAIL(&cc->cc_expireq, tmp, c_links.tqe); tmp->c_iflags |= CALLOUT_PROCESSED; tmp = tmpn; } continue; } /* Skip events from distant future. */ if (tmp->c_time >= max) goto next; /* * Event minimal time is bigger than present maximal * time, so it cannot be aggregated. */ if (tmp->c_time > last) { lastb = nowb; goto next; } /* Update first and last time, respecting this event. */ if (tmp->c_time < first) first = tmp->c_time; tmp_max = tmp->c_time + tmp->c_precision; if (tmp_max < last) last = tmp_max; next: tmp = LIST_NEXT(tmp, c_links.le); } /* Proceed with the next bucket. */ firstb++; /* * Stop if we looked after present time and found * some event we can't execute at now. * Stop if we looked far enough into the future. */ } while (((int)(firstb - lastb)) <= 0); cc->cc_firstevent = last; #ifndef NO_EVENTTIMERS cpu_new_callout(curcpu, last, first); #endif #ifdef CALLOUT_PROFILING avg_depth_dir += (depth_dir * 1000 - avg_depth_dir) >> 8; avg_mpcalls_dir += (mpcalls_dir * 1000 - avg_mpcalls_dir) >> 8; avg_lockcalls_dir += (lockcalls_dir * 1000 - avg_lockcalls_dir) >> 8; #endif mtx_unlock_spin_flags(&cc->cc_lock, MTX_QUIET); /* * swi_sched acquires the thread lock, so we don't want to call it * with cc_lock held; incorrect locking order. */ if (!TAILQ_EMPTY(&cc->cc_expireq)) swi_sched(cc->cc_cookie, 0); } static struct callout_cpu * callout_lock(struct callout *c) { struct callout_cpu *cc; int cpu; for (;;) { cpu = c->c_cpu; #ifdef SMP if (cpu == CPUBLOCK) { while (c->c_cpu == CPUBLOCK) cpu_spinwait(); continue; } #endif cc = CC_CPU(cpu); CC_LOCK(cc); if (cpu == c->c_cpu) break; CC_UNLOCK(cc); } return (cc); } static void callout_cc_add(struct callout *c, struct callout_cpu *cc, sbintime_t sbt, sbintime_t precision, void (*func)(void *), void *arg, int cpu, int flags) { int bucket; CC_LOCK_ASSERT(cc); if (sbt < cc->cc_lastscan) sbt = cc->cc_lastscan; c->c_arg = arg; c->c_iflags |= CALLOUT_PENDING; c->c_iflags &= ~CALLOUT_PROCESSED; c->c_flags |= CALLOUT_ACTIVE; if (flags & C_DIRECT_EXEC) c->c_iflags |= CALLOUT_DIRECT; c->c_func = func; c->c_time = sbt; c->c_precision = precision; bucket = callout_get_bucket(c->c_time); CTR3(KTR_CALLOUT, "precision set for %p: %d.%08x", c, (int)(c->c_precision >> 32), (u_int)(c->c_precision & 0xffffffff)); LIST_INSERT_HEAD(&cc->cc_callwheel[bucket], c, c_links.le); if (cc->cc_bucket == bucket) cc_exec_next(cc) = c; #ifndef NO_EVENTTIMERS /* * Inform the eventtimers(4) subsystem there's a new callout * that has been inserted, but only if really required. */ if (SBT_MAX - c->c_time < c->c_precision) c->c_precision = SBT_MAX - c->c_time; sbt = c->c_time + c->c_precision; if (sbt < cc->cc_firstevent) { cc->cc_firstevent = sbt; cpu_new_callout(cpu, sbt, c->c_time); } #endif } static void callout_cc_del(struct callout *c, struct callout_cpu *cc) { if ((c->c_iflags & CALLOUT_LOCAL_ALLOC) == 0) return; c->c_func = NULL; SLIST_INSERT_HEAD(&cc->cc_callfree, c, c_links.sle); } static void softclock_call_cc(struct callout *c, struct callout_cpu *cc, #ifdef CALLOUT_PROFILING int *mpcalls, int *lockcalls, int *gcalls, #endif int direct) { struct rm_priotracker tracker; void (*c_func)(void *); void *c_arg; struct lock_class *class; struct lock_object *c_lock; uintptr_t lock_status; int c_iflags; #ifdef SMP struct callout_cpu *new_cc; void (*new_func)(void *); void *new_arg; int flags, new_cpu; sbintime_t new_prec, new_time; #endif #if defined(DIAGNOSTIC) || defined(CALLOUT_PROFILING) sbintime_t sbt1, sbt2; struct timespec ts2; static sbintime_t maxdt = 2 * SBT_1MS; /* 2 msec */ static timeout_t *lastfunc; #endif KASSERT((c->c_iflags & CALLOUT_PENDING) == CALLOUT_PENDING, ("softclock_call_cc: pend %p %x", c, c->c_iflags)); KASSERT((c->c_flags & CALLOUT_ACTIVE) == CALLOUT_ACTIVE, ("softclock_call_cc: act %p %x", c, c->c_flags)); class = (c->c_lock != NULL) ? LOCK_CLASS(c->c_lock) : NULL; lock_status = 0; if (c->c_flags & CALLOUT_SHAREDLOCK) { if (class == &lock_class_rm) lock_status = (uintptr_t)&tracker; else lock_status = 1; } c_lock = c->c_lock; c_func = c->c_func; c_arg = c->c_arg; c_iflags = c->c_iflags; if (c->c_iflags & CALLOUT_LOCAL_ALLOC) c->c_iflags = CALLOUT_LOCAL_ALLOC; else c->c_iflags &= ~CALLOUT_PENDING; cc_exec_curr(cc, direct) = c; cc_exec_cancel(cc, direct) = false; cc_exec_drain(cc, direct) = NULL; CC_UNLOCK(cc); if (c_lock != NULL) { class->lc_lock(c_lock, lock_status); /* * The callout may have been cancelled * while we switched locks. */ if (cc_exec_cancel(cc, direct)) { class->lc_unlock(c_lock); goto skip; } /* The callout cannot be stopped now. */ cc_exec_cancel(cc, direct) = true; if (c_lock == &Giant.lock_object) { #ifdef CALLOUT_PROFILING (*gcalls)++; #endif CTR3(KTR_CALLOUT, "callout giant %p func %p arg %p", c, c_func, c_arg); } else { #ifdef CALLOUT_PROFILING (*lockcalls)++; #endif CTR3(KTR_CALLOUT, "callout lock %p func %p arg %p", c, c_func, c_arg); } } else { #ifdef CALLOUT_PROFILING (*mpcalls)++; #endif CTR3(KTR_CALLOUT, "callout %p func %p arg %p", c, c_func, c_arg); } KTR_STATE3(KTR_SCHED, "callout", cc->cc_ktr_event_name, "running", "func:%p", c_func, "arg:%p", c_arg, "direct:%d", direct); #if defined(DIAGNOSTIC) || defined(CALLOUT_PROFILING) sbt1 = sbinuptime(); #endif THREAD_NO_SLEEPING(); SDT_PROBE1(callout_execute, , , callout__start, c); c_func(c_arg); SDT_PROBE1(callout_execute, , , callout__end, c); THREAD_SLEEPING_OK(); #if defined(DIAGNOSTIC) || defined(CALLOUT_PROFILING) sbt2 = sbinuptime(); sbt2 -= sbt1; if (sbt2 > maxdt) { if (lastfunc != c_func || sbt2 > maxdt * 2) { ts2 = sbttots(sbt2); printf( "Expensive timeout(9) function: %p(%p) %jd.%09ld s\n", c_func, c_arg, (intmax_t)ts2.tv_sec, ts2.tv_nsec); } maxdt = sbt2; lastfunc = c_func; } #endif KTR_STATE0(KTR_SCHED, "callout", cc->cc_ktr_event_name, "idle"); CTR1(KTR_CALLOUT, "callout %p finished", c); if ((c_iflags & CALLOUT_RETURNUNLOCKED) == 0) class->lc_unlock(c_lock); skip: CC_LOCK(cc); KASSERT(cc_exec_curr(cc, direct) == c, ("mishandled cc_curr")); cc_exec_curr(cc, direct) = NULL; if (cc_exec_drain(cc, direct)) { void (*drain)(void *); drain = cc_exec_drain(cc, direct); cc_exec_drain(cc, direct) = NULL; CC_UNLOCK(cc); drain(c_arg); CC_LOCK(cc); } if (cc_exec_waiting(cc, direct)) { /* * There is someone waiting for the * callout to complete. * If the callout was scheduled for * migration just cancel it. */ if (cc_cce_migrating(cc, direct)) { cc_cce_cleanup(cc, direct); /* * It should be assert here that the callout is not * destroyed but that is not easy. */ c->c_iflags &= ~CALLOUT_DFRMIGRATION; } cc_exec_waiting(cc, direct) = false; CC_UNLOCK(cc); wakeup(&cc_exec_waiting(cc, direct)); CC_LOCK(cc); } else if (cc_cce_migrating(cc, direct)) { KASSERT((c_iflags & CALLOUT_LOCAL_ALLOC) == 0, ("Migrating legacy callout %p", c)); #ifdef SMP /* * If the callout was scheduled for * migration just perform it now. */ new_cpu = cc_migration_cpu(cc, direct); new_time = cc_migration_time(cc, direct); new_prec = cc_migration_prec(cc, direct); new_func = cc_migration_func(cc, direct); new_arg = cc_migration_arg(cc, direct); cc_cce_cleanup(cc, direct); /* * It should be assert here that the callout is not destroyed * but that is not easy. * * As first thing, handle deferred callout stops. */ if (!callout_migrating(c)) { CTR3(KTR_CALLOUT, "deferred cancelled %p func %p arg %p", c, new_func, new_arg); callout_cc_del(c, cc); return; } c->c_iflags &= ~CALLOUT_DFRMIGRATION; new_cc = callout_cpu_switch(c, cc, new_cpu); flags = (direct) ? C_DIRECT_EXEC : 0; callout_cc_add(c, new_cc, new_time, new_prec, new_func, new_arg, new_cpu, flags); CC_UNLOCK(new_cc); CC_LOCK(cc); #else panic("migration should not happen"); #endif } /* * If the current callout is locally allocated (from * timeout(9)) then put it on the freelist. * * Note: we need to check the cached copy of c_iflags because * if it was not local, then it's not safe to deref the * callout pointer. */ KASSERT((c_iflags & CALLOUT_LOCAL_ALLOC) == 0 || c->c_iflags == CALLOUT_LOCAL_ALLOC, ("corrupted callout")); if (c_iflags & CALLOUT_LOCAL_ALLOC) callout_cc_del(c, cc); } /* * The callout mechanism is based on the work of Adam M. Costello and * George Varghese, published in a technical report entitled "Redesigning * the BSD Callout and Timer Facilities" and modified slightly for inclusion * in FreeBSD by Justin T. Gibbs. The original work on the data structures * used in this implementation was published by G. Varghese and T. Lauck in * the paper "Hashed and Hierarchical Timing Wheels: Data Structures for * the Efficient Implementation of a Timer Facility" in the Proceedings of * the 11th ACM Annual Symposium on Operating Systems Principles, * Austin, Texas Nov 1987. */ /* * Software (low priority) clock interrupt. * Run periodic events from timeout queue. */ void softclock(void *arg) { struct callout_cpu *cc; struct callout *c; #ifdef CALLOUT_PROFILING int depth = 0, gcalls = 0, lockcalls = 0, mpcalls = 0; #endif cc = (struct callout_cpu *)arg; CC_LOCK(cc); while ((c = TAILQ_FIRST(&cc->cc_expireq)) != NULL) { TAILQ_REMOVE(&cc->cc_expireq, c, c_links.tqe); softclock_call_cc(c, cc, #ifdef CALLOUT_PROFILING &mpcalls, &lockcalls, &gcalls, #endif 0); #ifdef CALLOUT_PROFILING ++depth; #endif } #ifdef CALLOUT_PROFILING avg_depth += (depth * 1000 - avg_depth) >> 8; avg_mpcalls += (mpcalls * 1000 - avg_mpcalls) >> 8; avg_lockcalls += (lockcalls * 1000 - avg_lockcalls) >> 8; avg_gcalls += (gcalls * 1000 - avg_gcalls) >> 8; #endif CC_UNLOCK(cc); } /* * timeout -- * Execute a function after a specified length of time. * * untimeout -- * Cancel previous timeout function call. * * callout_handle_init -- * Initialize a handle so that using it with untimeout is benign. * * See AT&T BCI Driver Reference Manual for specification. This * implementation differs from that one in that although an * identification value is returned from timeout, the original * arguments to timeout as well as the identifier are used to * identify entries for untimeout. */ struct callout_handle timeout(timeout_t *ftn, void *arg, int to_ticks) { struct callout_cpu *cc; struct callout *new; struct callout_handle handle; cc = CC_CPU(timeout_cpu); CC_LOCK(cc); /* Fill in the next free callout structure. */ new = SLIST_FIRST(&cc->cc_callfree); if (new == NULL) /* XXX Attempt to malloc first */ panic("timeout table full"); SLIST_REMOVE_HEAD(&cc->cc_callfree, c_links.sle); callout_reset(new, to_ticks, ftn, arg); handle.callout = new; CC_UNLOCK(cc); return (handle); } void untimeout(timeout_t *ftn, void *arg, struct callout_handle handle) { struct callout_cpu *cc; /* * Check for a handle that was initialized * by callout_handle_init, but never used * for a real timeout. */ if (handle.callout == NULL) return; cc = callout_lock(handle.callout); if (handle.callout->c_func == ftn && handle.callout->c_arg == arg) callout_stop(handle.callout); CC_UNLOCK(cc); } void callout_handle_init(struct callout_handle *handle) { handle->callout = NULL; } /* * New interface; clients allocate their own callout structures. * * callout_reset() - establish or change a timeout * callout_stop() - disestablish a timeout * callout_init() - initialize a callout structure so that it can * safely be passed to callout_reset() and callout_stop() * * defines three convenience macros: * * callout_active() - returns truth if callout has not been stopped, * drained, or deactivated since the last time the callout was * reset. * callout_pending() - returns truth if callout is still waiting for timeout * callout_deactivate() - marks the callout as having been serviced */ int callout_reset_sbt_on(struct callout *c, sbintime_t sbt, sbintime_t precision, void (*ftn)(void *), void *arg, int cpu, int flags) { sbintime_t to_sbt, pr; struct callout_cpu *cc; int cancelled, direct; int ignore_cpu=0; cancelled = 0; if (cpu == -1) { ignore_cpu = 1; } else if ((cpu >= MAXCPU) || ((CC_CPU(cpu))->cc_inited == 0)) { /* Invalid CPU spec */ panic("Invalid CPU in callout %d", cpu); } if (flags & C_ABSOLUTE) { to_sbt = sbt; } else { if ((flags & C_HARDCLOCK) && (sbt < tick_sbt)) sbt = tick_sbt; if ((flags & C_HARDCLOCK) || #ifdef NO_EVENTTIMERS sbt >= sbt_timethreshold) { to_sbt = getsbinuptime(); /* Add safety belt for the case of hz > 1000. */ to_sbt += tc_tick_sbt - tick_sbt; #else sbt >= sbt_tickthreshold) { /* * Obtain the time of the last hardclock() call on * this CPU directly from the kern_clocksource.c. * This value is per-CPU, but it is equal for all * active ones. */ #ifdef __LP64__ to_sbt = DPCPU_GET(hardclocktime); #else spinlock_enter(); to_sbt = DPCPU_GET(hardclocktime); spinlock_exit(); #endif #endif if ((flags & C_HARDCLOCK) == 0) to_sbt += tick_sbt; } else to_sbt = sbinuptime(); if (SBT_MAX - to_sbt < sbt) to_sbt = SBT_MAX; else to_sbt += sbt; pr = ((C_PRELGET(flags) < 0) ? sbt >> tc_precexp : sbt >> C_PRELGET(flags)); if (pr > precision) precision = pr; } /* * This flag used to be added by callout_cc_add, but the * first time you call this we could end up with the * wrong direct flag if we don't do it before we add. */ if (flags & C_DIRECT_EXEC) { direct = 1; } else { direct = 0; } KASSERT(!direct || c->c_lock == NULL, ("%s: direct callout %p has lock", __func__, c)); cc = callout_lock(c); /* * Don't allow migration of pre-allocated callouts lest they * become unbalanced or handle the case where the user does * not care. */ if ((c->c_iflags & CALLOUT_LOCAL_ALLOC) || ignore_cpu) { cpu = c->c_cpu; } if (cc_exec_curr(cc, direct) == c) { /* * We're being asked to reschedule a callout which is * currently in progress. If there is a lock then we * can cancel the callout if it has not really started. */ if (c->c_lock != NULL && !cc_exec_cancel(cc, direct)) cancelled = cc_exec_cancel(cc, direct) = true; if (cc_exec_waiting(cc, direct)) { /* * Someone has called callout_drain to kill this * callout. Don't reschedule. */ CTR4(KTR_CALLOUT, "%s %p func %p arg %p", cancelled ? "cancelled" : "failed to cancel", c, c->c_func, c->c_arg); CC_UNLOCK(cc); return (cancelled); } #ifdef SMP if (callout_migrating(c)) { /* * This only occurs when a second callout_reset_sbt_on * is made after a previous one moved it into * deferred migration (below). Note we do *not* change * the prev_cpu even though the previous target may * be different. */ cc_migration_cpu(cc, direct) = cpu; cc_migration_time(cc, direct) = to_sbt; cc_migration_prec(cc, direct) = precision; cc_migration_func(cc, direct) = ftn; cc_migration_arg(cc, direct) = arg; cancelled = 1; CC_UNLOCK(cc); return (cancelled); } #endif } if (c->c_iflags & CALLOUT_PENDING) { if ((c->c_iflags & CALLOUT_PROCESSED) == 0) { if (cc_exec_next(cc) == c) cc_exec_next(cc) = LIST_NEXT(c, c_links.le); LIST_REMOVE(c, c_links.le); } else { TAILQ_REMOVE(&cc->cc_expireq, c, c_links.tqe); } cancelled = 1; c->c_iflags &= ~ CALLOUT_PENDING; c->c_flags &= ~ CALLOUT_ACTIVE; } #ifdef SMP /* * If the callout must migrate try to perform it immediately. * If the callout is currently running, just defer the migration * to a more appropriate moment. */ if (c->c_cpu != cpu) { if (cc_exec_curr(cc, direct) == c) { /* * Pending will have been removed since we are * actually executing the callout on another * CPU. That callout should be waiting on the * lock the caller holds. If we set both * active/and/pending after we return and the * lock on the executing callout proceeds, it * will then see pending is true and return. * At the return from the actual callout execution * the migration will occur in softclock_call_cc * and this new callout will be placed on the * new CPU via a call to callout_cpu_switch() which * will get the lock on the right CPU followed * by a call callout_cc_add() which will add it there. * (see above in softclock_call_cc()). */ cc_migration_cpu(cc, direct) = cpu; cc_migration_time(cc, direct) = to_sbt; cc_migration_prec(cc, direct) = precision; cc_migration_func(cc, direct) = ftn; cc_migration_arg(cc, direct) = arg; c->c_iflags |= (CALLOUT_DFRMIGRATION | CALLOUT_PENDING); c->c_flags |= CALLOUT_ACTIVE; CTR6(KTR_CALLOUT, "migration of %p func %p arg %p in %d.%08x to %u deferred", c, c->c_func, c->c_arg, (int)(to_sbt >> 32), (u_int)(to_sbt & 0xffffffff), cpu); CC_UNLOCK(cc); return (cancelled); } cc = callout_cpu_switch(c, cc, cpu); } #endif callout_cc_add(c, cc, to_sbt, precision, ftn, arg, cpu, flags); CTR6(KTR_CALLOUT, "%sscheduled %p func %p arg %p in %d.%08x", cancelled ? "re" : "", c, c->c_func, c->c_arg, (int)(to_sbt >> 32), (u_int)(to_sbt & 0xffffffff)); CC_UNLOCK(cc); return (cancelled); } /* * Common idioms that can be optimized in the future. */ int callout_schedule_on(struct callout *c, int to_ticks, int cpu) { return callout_reset_on(c, to_ticks, c->c_func, c->c_arg, cpu); } int callout_schedule(struct callout *c, int to_ticks) { return callout_reset_on(c, to_ticks, c->c_func, c->c_arg, c->c_cpu); } int -_callout_stop_safe(struct callout *c, int safe, void (*drain)(void *)) +_callout_stop_safe(struct callout *c, int flags, void (*drain)(void *)) { struct callout_cpu *cc, *old_cc; struct lock_class *class; int direct, sq_locked, use_lock; int not_on_a_list; - if (safe) + if ((flags & CS_DRAIN) != 0) WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, c->c_lock, "calling %s", __func__); /* * Some old subsystems don't hold Giant while running a callout_stop(), * so just discard this check for the moment. */ - if (!safe && c->c_lock != NULL) { + if ((flags & CS_DRAIN) == 0 && c->c_lock != NULL) { if (c->c_lock == &Giant.lock_object) use_lock = mtx_owned(&Giant); else { use_lock = 1; class = LOCK_CLASS(c->c_lock); class->lc_assert(c->c_lock, LA_XLOCKED); } } else use_lock = 0; if (c->c_iflags & CALLOUT_DIRECT) { direct = 1; } else { direct = 0; } sq_locked = 0; old_cc = NULL; again: cc = callout_lock(c); if ((c->c_iflags & (CALLOUT_DFRMIGRATION | CALLOUT_PENDING)) == (CALLOUT_DFRMIGRATION | CALLOUT_PENDING) && ((c->c_flags & CALLOUT_ACTIVE) == CALLOUT_ACTIVE)) { /* * Special case where this slipped in while we * were migrating *as* the callout is about to * execute. The caller probably holds the lock * the callout wants. * * Get rid of the migration first. Then set * the flag that tells this code *not* to * try to remove it from any lists (its not * on one yet). When the callout wheel runs, * it will ignore this callout. */ c->c_iflags &= ~CALLOUT_PENDING; c->c_flags &= ~CALLOUT_ACTIVE; not_on_a_list = 1; } else { not_on_a_list = 0; } /* * If the callout was migrating while the callout cpu lock was * dropped, just drop the sleepqueue lock and check the states * again. */ if (sq_locked != 0 && cc != old_cc) { #ifdef SMP CC_UNLOCK(cc); sleepq_release(&cc_exec_waiting(old_cc, direct)); sq_locked = 0; old_cc = NULL; goto again; #else panic("migration should not happen"); #endif } /* * If the callout isn't pending, it's not on the queue, so * don't attempt to remove it from the queue. We can try to * stop it by other means however. */ if (!(c->c_iflags & CALLOUT_PENDING)) { /* * If it wasn't on the queue and it isn't the current * callout, then we can't stop it, so just bail. * It probably has already been run (if locking * is properly done). You could get here if the caller * calls stop twice in a row for example. The second * call would fall here without CALLOUT_ACTIVE set. */ c->c_flags &= ~CALLOUT_ACTIVE; if (cc_exec_curr(cc, direct) != c) { CTR3(KTR_CALLOUT, "failed to stop %p func %p arg %p", c, c->c_func, c->c_arg); CC_UNLOCK(cc); if (sq_locked) sleepq_release(&cc_exec_waiting(cc, direct)); return (-1); } - if (safe) { + if ((flags & CS_DRAIN) != 0) { /* * The current callout is running (or just * about to run) and blocking is allowed, so * just wait for the current invocation to * finish. */ while (cc_exec_curr(cc, direct) == c) { /* * Use direct calls to sleepqueue interface * instead of cv/msleep in order to avoid * a LOR between cc_lock and sleepqueue * chain spinlocks. This piece of code * emulates a msleep_spin() call actually. * * If we already have the sleepqueue chain * locked, then we can safely block. If we * don't already have it locked, however, * we have to drop the cc_lock to lock * it. This opens several races, so we * restart at the beginning once we have * both locks. If nothing has changed, then * we will end up back here with sq_locked * set. */ if (!sq_locked) { CC_UNLOCK(cc); sleepq_lock( &cc_exec_waiting(cc, direct)); sq_locked = 1; old_cc = cc; goto again; } /* * Migration could be cancelled here, but * as long as it is still not sure when it * will be packed up, just let softclock() * take care of it. */ cc_exec_waiting(cc, direct) = true; DROP_GIANT(); CC_UNLOCK(cc); sleepq_add( &cc_exec_waiting(cc, direct), &cc->cc_lock.lock_object, "codrain", SLEEPQ_SLEEP, 0); sleepq_wait( &cc_exec_waiting(cc, direct), 0); sq_locked = 0; old_cc = NULL; /* Reacquire locks previously released. */ PICKUP_GIANT(); CC_LOCK(cc); } } else if (use_lock && !cc_exec_cancel(cc, direct) && (drain == NULL)) { /* * The current callout is waiting for its * lock which we hold. Cancel the callout * and return. After our caller drops the * lock, the callout will be skipped in * softclock(). This *only* works with a * callout_stop() *not* callout_drain() or * callout_async_drain(). */ cc_exec_cancel(cc, direct) = true; CTR3(KTR_CALLOUT, "cancelled %p func %p arg %p", c, c->c_func, c->c_arg); KASSERT(!cc_cce_migrating(cc, direct), ("callout wrongly scheduled for migration")); if (callout_migrating(c)) { c->c_iflags &= ~CALLOUT_DFRMIGRATION; #ifdef SMP cc_migration_cpu(cc, direct) = CPUBLOCK; cc_migration_time(cc, direct) = 0; cc_migration_prec(cc, direct) = 0; cc_migration_func(cc, direct) = NULL; cc_migration_arg(cc, direct) = NULL; #endif } CC_UNLOCK(cc); KASSERT(!sq_locked, ("sleepqueue chain locked")); return (1); } else if (callout_migrating(c)) { /* * The callout is currently being serviced * and the "next" callout is scheduled at * its completion with a migration. We remove * the migration flag so it *won't* get rescheduled, * but we can't stop the one thats running so * we return 0. */ c->c_iflags &= ~CALLOUT_DFRMIGRATION; #ifdef SMP /* * We can't call cc_cce_cleanup here since * if we do it will remove .ce_curr and * its still running. This will prevent a * reschedule of the callout when the * execution completes. */ cc_migration_cpu(cc, direct) = CPUBLOCK; cc_migration_time(cc, direct) = 0; cc_migration_prec(cc, direct) = 0; cc_migration_func(cc, direct) = NULL; cc_migration_arg(cc, direct) = NULL; #endif CTR3(KTR_CALLOUT, "postponing stop %p func %p arg %p", c, c->c_func, c->c_arg); if (drain) { cc_exec_drain(cc, direct) = drain; } CC_UNLOCK(cc); - return (0); + return ((flags & CS_MIGRBLOCK) != 0); } CTR3(KTR_CALLOUT, "failed to stop %p func %p arg %p", c, c->c_func, c->c_arg); if (drain) { cc_exec_drain(cc, direct) = drain; } CC_UNLOCK(cc); KASSERT(!sq_locked, ("sleepqueue chain still locked")); return (0); } if (sq_locked) sleepq_release(&cc_exec_waiting(cc, direct)); c->c_iflags &= ~CALLOUT_PENDING; c->c_flags &= ~CALLOUT_ACTIVE; CTR3(KTR_CALLOUT, "cancelled %p func %p arg %p", c, c->c_func, c->c_arg); if (not_on_a_list == 0) { if ((c->c_iflags & CALLOUT_PROCESSED) == 0) { if (cc_exec_next(cc) == c) cc_exec_next(cc) = LIST_NEXT(c, c_links.le); LIST_REMOVE(c, c_links.le); } else { TAILQ_REMOVE(&cc->cc_expireq, c, c_links.tqe); } } callout_cc_del(c, cc); CC_UNLOCK(cc); return (1); } void callout_init(struct callout *c, int mpsafe) { bzero(c, sizeof *c); if (mpsafe) { c->c_lock = NULL; c->c_iflags = CALLOUT_RETURNUNLOCKED; } else { c->c_lock = &Giant.lock_object; c->c_iflags = 0; } c->c_cpu = timeout_cpu; } void _callout_init_lock(struct callout *c, struct lock_object *lock, int flags) { bzero(c, sizeof *c); c->c_lock = lock; KASSERT((flags & ~(CALLOUT_RETURNUNLOCKED | CALLOUT_SHAREDLOCK)) == 0, ("callout_init_lock: bad flags %d", flags)); KASSERT(lock != NULL || (flags & CALLOUT_RETURNUNLOCKED) == 0, ("callout_init_lock: CALLOUT_RETURNUNLOCKED with no lock")); KASSERT(lock == NULL || !(LOCK_CLASS(lock)->lc_flags & (LC_SPINLOCK | LC_SLEEPABLE)), ("%s: invalid lock class", __func__)); c->c_iflags = flags & (CALLOUT_RETURNUNLOCKED | CALLOUT_SHAREDLOCK); c->c_cpu = timeout_cpu; } #ifdef APM_FIXUP_CALLTODO /* * Adjust the kernel calltodo timeout list. This routine is used after * an APM resume to recalculate the calltodo timer list values with the * number of hz's we have been sleeping. The next hardclock() will detect * that there are fired timers and run softclock() to execute them. * * Please note, I have not done an exhaustive analysis of what code this * might break. I am motivated to have my select()'s and alarm()'s that * have expired during suspend firing upon resume so that the applications * which set the timer can do the maintanence the timer was for as close * as possible to the originally intended time. Testing this code for a * week showed that resuming from a suspend resulted in 22 to 25 timers * firing, which seemed independant on whether the suspend was 2 hours or * 2 days. Your milage may vary. - Ken Key */ void adjust_timeout_calltodo(struct timeval *time_change) { register struct callout *p; unsigned long delta_ticks; /* * How many ticks were we asleep? * (stolen from tvtohz()). */ /* Don't do anything */ if (time_change->tv_sec < 0) return; else if (time_change->tv_sec <= LONG_MAX / 1000000) delta_ticks = (time_change->tv_sec * 1000000 + time_change->tv_usec + (tick - 1)) / tick + 1; else if (time_change->tv_sec <= LONG_MAX / hz) delta_ticks = time_change->tv_sec * hz + (time_change->tv_usec + (tick - 1)) / tick + 1; else delta_ticks = LONG_MAX; if (delta_ticks > INT_MAX) delta_ticks = INT_MAX; /* * Now rip through the timer calltodo list looking for timers * to expire. */ /* don't collide with softclock() */ CC_LOCK(cc); for (p = calltodo.c_next; p != NULL; p = p->c_next) { p->c_time -= delta_ticks; /* Break if the timer had more time on it than delta_ticks */ if (p->c_time > 0) break; /* take back the ticks the timer didn't use (p->c_time <= 0) */ delta_ticks = -p->c_time; } CC_UNLOCK(cc); return; } #endif /* APM_FIXUP_CALLTODO */ static int flssbt(sbintime_t sbt) { sbt += (uint64_t)sbt >> 1; if (sizeof(long) >= sizeof(sbintime_t)) return (flsl(sbt)); if (sbt >= SBT_1S) return (flsl(((uint64_t)sbt) >> 32) + 32); return (flsl(sbt)); } /* * Dump immediate statistic snapshot of the scheduled callouts. */ static int sysctl_kern_callout_stat(SYSCTL_HANDLER_ARGS) { struct callout *tmp; struct callout_cpu *cc; struct callout_list *sc; sbintime_t maxpr, maxt, medpr, medt, now, spr, st, t; int ct[64], cpr[64], ccpbk[32]; int error, val, i, count, tcum, pcum, maxc, c, medc; #ifdef SMP int cpu; #endif val = 0; error = sysctl_handle_int(oidp, &val, 0, req); if (error != 0 || req->newptr == NULL) return (error); count = maxc = 0; st = spr = maxt = maxpr = 0; bzero(ccpbk, sizeof(ccpbk)); bzero(ct, sizeof(ct)); bzero(cpr, sizeof(cpr)); now = sbinuptime(); #ifdef SMP CPU_FOREACH(cpu) { cc = CC_CPU(cpu); #else cc = CC_CPU(timeout_cpu); #endif CC_LOCK(cc); for (i = 0; i < callwheelsize; i++) { sc = &cc->cc_callwheel[i]; c = 0; LIST_FOREACH(tmp, sc, c_links.le) { c++; t = tmp->c_time - now; if (t < 0) t = 0; st += t / SBT_1US; spr += tmp->c_precision / SBT_1US; if (t > maxt) maxt = t; if (tmp->c_precision > maxpr) maxpr = tmp->c_precision; ct[flssbt(t)]++; cpr[flssbt(tmp->c_precision)]++; } if (c > maxc) maxc = c; ccpbk[fls(c + c / 2)]++; count += c; } CC_UNLOCK(cc); #ifdef SMP } #endif for (i = 0, tcum = 0; i < 64 && tcum < count / 2; i++) tcum += ct[i]; medt = (i >= 2) ? (((sbintime_t)1) << (i - 2)) : 0; for (i = 0, pcum = 0; i < 64 && pcum < count / 2; i++) pcum += cpr[i]; medpr = (i >= 2) ? (((sbintime_t)1) << (i - 2)) : 0; for (i = 0, c = 0; i < 32 && c < count / 2; i++) c += ccpbk[i]; medc = (i >= 2) ? (1 << (i - 2)) : 0; printf("Scheduled callouts statistic snapshot:\n"); printf(" Callouts: %6d Buckets: %6d*%-3d Bucket size: 0.%06ds\n", count, callwheelsize, mp_ncpus, 1000000 >> CC_HASH_SHIFT); printf(" C/Bk: med %5d avg %6d.%06jd max %6d\n", medc, count / callwheelsize / mp_ncpus, (uint64_t)count * 1000000 / callwheelsize / mp_ncpus % 1000000, maxc); printf(" Time: med %5jd.%06jds avg %6jd.%06jds max %6jd.%06jds\n", medt / SBT_1S, (medt & 0xffffffff) * 1000000 >> 32, (st / count) / 1000000, (st / count) % 1000000, maxt / SBT_1S, (maxt & 0xffffffff) * 1000000 >> 32); printf(" Prec: med %5jd.%06jds avg %6jd.%06jds max %6jd.%06jds\n", medpr / SBT_1S, (medpr & 0xffffffff) * 1000000 >> 32, (spr / count) / 1000000, (spr / count) % 1000000, maxpr / SBT_1S, (maxpr & 0xffffffff) * 1000000 >> 32); printf(" Distribution: \tbuckets\t time\t tcum\t" " prec\t pcum\n"); for (i = 0, tcum = pcum = 0; i < 64; i++) { if (ct[i] == 0 && cpr[i] == 0) continue; t = (i != 0) ? (((sbintime_t)1) << (i - 1)) : 0; tcum += ct[i]; pcum += cpr[i]; printf(" %10jd.%06jds\t 2**%d\t%7d\t%7d\t%7d\t%7d\n", t / SBT_1S, (t & 0xffffffff) * 1000000 >> 32, i - 1 - (32 - CC_HASH_SHIFT), ct[i], tcum, cpr[i], pcum); } return (error); } SYSCTL_PROC(_kern, OID_AUTO, callout_stat, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0, sysctl_kern_callout_stat, "I", "Dump immediate statistic snapshot of the scheduled callouts"); Index: projects/release-pkg/sys/kern/subr_sleepqueue.c =================================================================== --- projects/release-pkg/sys/kern/subr_sleepqueue.c (revision 296327) +++ projects/release-pkg/sys/kern/subr_sleepqueue.c (revision 296328) @@ -1,1247 +1,1248 @@ /*- * Copyright (c) 2004 John Baldwin * 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. */ /* * Implementation of sleep queues used to hold queue of threads blocked on * a wait channel. Sleep queues different from turnstiles in that wait * channels are not owned by anyone, so there is no priority propagation. * Sleep queues can also provide a timeout and can also be interrupted by * signals. That said, there are several similarities between the turnstile * and sleep queue implementations. (Note: turnstiles were implemented * first.) For example, both use a hash table of the same size where each * bucket is referred to as a "chain" that contains both a spin lock and * a linked list of queues. An individual queue is located by using a hash * to pick a chain, locking the chain, and then walking the chain searching * for the queue. This means that a wait channel object does not need to * embed it's queue head just as locks do not embed their turnstile queue * head. Threads also carry around a sleep queue that they lend to the * wait channel when blocking. Just as in turnstiles, the queue includes * a free list of the sleep queues of other threads blocked on the same * wait channel in the case of multiple waiters. * * Some additional functionality provided by sleep queues include the * ability to set a timeout. The timeout is managed using a per-thread * callout that resumes a thread if it is asleep. A thread may also * catch signals while it is asleep (aka an interruptible sleep). The * signal code uses sleepq_abort() to interrupt a sleeping thread. Finally, * sleep queues also provide some extra assertions. One is not allowed to * mix the sleep/wakeup and cv APIs for a given wait channel. Also, one * must consistently use the same lock to synchronize with a wait channel, * though this check is currently only a warning for sleep/wakeup due to * pre-existing abuse of that API. The same lock must also be held when * awakening threads, though that is currently only enforced for condition * variables. */ #include __FBSDID("$FreeBSD$"); #include "opt_sleepqueue_profiling.h" #include "opt_ddb.h" #include "opt_sched.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #endif /* * Constants for the hash table of sleep queue chains. * SC_TABLESIZE must be a power of two for SC_MASK to work properly. */ #define SC_TABLESIZE 256 /* Must be power of 2. */ #define SC_MASK (SC_TABLESIZE - 1) #define SC_SHIFT 8 #define SC_HASH(wc) ((((uintptr_t)(wc) >> SC_SHIFT) ^ (uintptr_t)(wc)) & \ SC_MASK) #define SC_LOOKUP(wc) &sleepq_chains[SC_HASH(wc)] #define NR_SLEEPQS 2 /* * There two different lists of sleep queues. Both lists are connected * via the sq_hash entries. The first list is the sleep queue chain list * that a sleep queue is on when it is attached to a wait channel. The * second list is the free list hung off of a sleep queue that is attached * to a wait channel. * * Each sleep queue also contains the wait channel it is attached to, the * list of threads blocked on that wait channel, flags specific to the * wait channel, and the lock used to synchronize with a wait channel. * The flags are used to catch mismatches between the various consumers * of the sleep queue API (e.g. sleep/wakeup and condition variables). * The lock pointer is only used when invariants are enabled for various * debugging checks. * * Locking key: * c - sleep queue chain lock */ struct sleepqueue { TAILQ_HEAD(, thread) sq_blocked[NR_SLEEPQS]; /* (c) Blocked threads. */ u_int sq_blockedcnt[NR_SLEEPQS]; /* (c) N. of blocked threads. */ LIST_ENTRY(sleepqueue) sq_hash; /* (c) Chain and free list. */ LIST_HEAD(, sleepqueue) sq_free; /* (c) Free queues. */ void *sq_wchan; /* (c) Wait channel. */ int sq_type; /* (c) Queue type. */ #ifdef INVARIANTS struct lock_object *sq_lock; /* (c) Associated lock. */ #endif }; struct sleepqueue_chain { LIST_HEAD(, sleepqueue) sc_queues; /* List of sleep queues. */ struct mtx sc_lock; /* Spin lock for this chain. */ #ifdef SLEEPQUEUE_PROFILING u_int sc_depth; /* Length of sc_queues. */ u_int sc_max_depth; /* Max length of sc_queues. */ #endif }; #ifdef SLEEPQUEUE_PROFILING u_int sleepq_max_depth; static SYSCTL_NODE(_debug, OID_AUTO, sleepq, CTLFLAG_RD, 0, "sleepq profiling"); static SYSCTL_NODE(_debug_sleepq, OID_AUTO, chains, CTLFLAG_RD, 0, "sleepq chain stats"); SYSCTL_UINT(_debug_sleepq, OID_AUTO, max_depth, CTLFLAG_RD, &sleepq_max_depth, 0, "maxmimum depth achieved of a single chain"); static void sleepq_profile(const char *wmesg); static int prof_enabled; #endif static struct sleepqueue_chain sleepq_chains[SC_TABLESIZE]; static uma_zone_t sleepq_zone; /* * Prototypes for non-exported routines. */ static int sleepq_catch_signals(void *wchan, int pri); static int sleepq_check_signals(void); static int sleepq_check_timeout(void); #ifdef INVARIANTS static void sleepq_dtor(void *mem, int size, void *arg); #endif static int sleepq_init(void *mem, int size, int flags); static int sleepq_resume_thread(struct sleepqueue *sq, struct thread *td, int pri); static void sleepq_switch(void *wchan, int pri); static void sleepq_timeout(void *arg); SDT_PROBE_DECLARE(sched, , , sleep); SDT_PROBE_DECLARE(sched, , , wakeup); /* * Initialize SLEEPQUEUE_PROFILING specific sysctl nodes. * Note that it must happen after sleepinit() has been fully executed, so * it must happen after SI_SUB_KMEM SYSINIT() subsystem setup. */ #ifdef SLEEPQUEUE_PROFILING static void init_sleepqueue_profiling(void) { char chain_name[10]; struct sysctl_oid *chain_oid; u_int i; for (i = 0; i < SC_TABLESIZE; i++) { snprintf(chain_name, sizeof(chain_name), "%u", i); chain_oid = SYSCTL_ADD_NODE(NULL, SYSCTL_STATIC_CHILDREN(_debug_sleepq_chains), OID_AUTO, chain_name, CTLFLAG_RD, NULL, "sleepq chain stats"); SYSCTL_ADD_UINT(NULL, SYSCTL_CHILDREN(chain_oid), OID_AUTO, "depth", CTLFLAG_RD, &sleepq_chains[i].sc_depth, 0, NULL); SYSCTL_ADD_UINT(NULL, SYSCTL_CHILDREN(chain_oid), OID_AUTO, "max_depth", CTLFLAG_RD, &sleepq_chains[i].sc_max_depth, 0, NULL); } } SYSINIT(sleepqueue_profiling, SI_SUB_LOCK, SI_ORDER_ANY, init_sleepqueue_profiling, NULL); #endif /* * Early initialization of sleep queues that is called from the sleepinit() * SYSINIT. */ void init_sleepqueues(void) { int i; for (i = 0; i < SC_TABLESIZE; i++) { LIST_INIT(&sleepq_chains[i].sc_queues); mtx_init(&sleepq_chains[i].sc_lock, "sleepq chain", NULL, MTX_SPIN | MTX_RECURSE); } sleepq_zone = uma_zcreate("SLEEPQUEUE", sizeof(struct sleepqueue), #ifdef INVARIANTS NULL, sleepq_dtor, sleepq_init, NULL, UMA_ALIGN_CACHE, 0); #else NULL, NULL, sleepq_init, NULL, UMA_ALIGN_CACHE, 0); #endif thread0.td_sleepqueue = sleepq_alloc(); } /* * Get a sleep queue for a new thread. */ struct sleepqueue * sleepq_alloc(void) { return (uma_zalloc(sleepq_zone, M_WAITOK)); } /* * Free a sleep queue when a thread is destroyed. */ void sleepq_free(struct sleepqueue *sq) { uma_zfree(sleepq_zone, sq); } /* * Lock the sleep queue chain associated with the specified wait channel. */ void sleepq_lock(void *wchan) { struct sleepqueue_chain *sc; sc = SC_LOOKUP(wchan); mtx_lock_spin(&sc->sc_lock); } /* * Look up the sleep queue associated with a given wait channel in the hash * table locking the associated sleep queue chain. If no queue is found in * the table, NULL is returned. */ struct sleepqueue * sleepq_lookup(void *wchan) { struct sleepqueue_chain *sc; struct sleepqueue *sq; KASSERT(wchan != NULL, ("%s: invalid NULL wait channel", __func__)); sc = SC_LOOKUP(wchan); mtx_assert(&sc->sc_lock, MA_OWNED); LIST_FOREACH(sq, &sc->sc_queues, sq_hash) if (sq->sq_wchan == wchan) return (sq); return (NULL); } /* * Unlock the sleep queue chain associated with a given wait channel. */ void sleepq_release(void *wchan) { struct sleepqueue_chain *sc; sc = SC_LOOKUP(wchan); mtx_unlock_spin(&sc->sc_lock); } /* * Places the current thread on the sleep queue for the specified wait * channel. If INVARIANTS is enabled, then it associates the passed in * lock with the sleepq to make sure it is held when that sleep queue is * woken up. */ void sleepq_add(void *wchan, struct lock_object *lock, const char *wmesg, int flags, int queue) { struct sleepqueue_chain *sc; struct sleepqueue *sq; struct thread *td; td = curthread; sc = SC_LOOKUP(wchan); mtx_assert(&sc->sc_lock, MA_OWNED); MPASS(td->td_sleepqueue != NULL); MPASS(wchan != NULL); MPASS((queue >= 0) && (queue < NR_SLEEPQS)); /* If this thread is not allowed to sleep, die a horrible death. */ KASSERT(td->td_no_sleeping == 0, ("%s: td %p to sleep on wchan %p with sleeping prohibited", __func__, td, wchan)); /* Look up the sleep queue associated with the wait channel 'wchan'. */ sq = sleepq_lookup(wchan); /* * If the wait channel does not already have a sleep queue, use * this thread's sleep queue. Otherwise, insert the current thread * into the sleep queue already in use by this wait channel. */ if (sq == NULL) { #ifdef INVARIANTS int i; sq = td->td_sleepqueue; for (i = 0; i < NR_SLEEPQS; i++) { KASSERT(TAILQ_EMPTY(&sq->sq_blocked[i]), ("thread's sleep queue %d is not empty", i)); KASSERT(sq->sq_blockedcnt[i] == 0, ("thread's sleep queue %d count mismatches", i)); } KASSERT(LIST_EMPTY(&sq->sq_free), ("thread's sleep queue has a non-empty free list")); KASSERT(sq->sq_wchan == NULL, ("stale sq_wchan pointer")); sq->sq_lock = lock; #endif #ifdef SLEEPQUEUE_PROFILING sc->sc_depth++; if (sc->sc_depth > sc->sc_max_depth) { sc->sc_max_depth = sc->sc_depth; if (sc->sc_max_depth > sleepq_max_depth) sleepq_max_depth = sc->sc_max_depth; } #endif sq = td->td_sleepqueue; LIST_INSERT_HEAD(&sc->sc_queues, sq, sq_hash); sq->sq_wchan = wchan; sq->sq_type = flags & SLEEPQ_TYPE; } else { MPASS(wchan == sq->sq_wchan); MPASS(lock == sq->sq_lock); MPASS((flags & SLEEPQ_TYPE) == sq->sq_type); LIST_INSERT_HEAD(&sq->sq_free, td->td_sleepqueue, sq_hash); } thread_lock(td); TAILQ_INSERT_TAIL(&sq->sq_blocked[queue], td, td_slpq); sq->sq_blockedcnt[queue]++; td->td_sleepqueue = NULL; td->td_sqqueue = queue; td->td_wchan = wchan; td->td_wmesg = wmesg; if (flags & SLEEPQ_INTERRUPTIBLE) { td->td_flags |= TDF_SINTR; td->td_flags &= ~TDF_SLEEPABORT; } thread_unlock(td); } /* * Sets a timeout that will remove the current thread from the specified * sleep queue after timo ticks if the thread has not already been awakened. */ void sleepq_set_timeout_sbt(void *wchan, sbintime_t sbt, sbintime_t pr, int flags) { struct sleepqueue_chain *sc; struct thread *td; td = curthread; sc = SC_LOOKUP(wchan); mtx_assert(&sc->sc_lock, MA_OWNED); MPASS(TD_ON_SLEEPQ(td)); MPASS(td->td_sleepqueue == NULL); MPASS(wchan != NULL); callout_reset_sbt_on(&td->td_slpcallout, sbt, pr, sleepq_timeout, td, PCPU_GET(cpuid), flags | C_DIRECT_EXEC); } /* * Return the number of actual sleepers for the specified queue. */ u_int sleepq_sleepcnt(void *wchan, int queue) { struct sleepqueue *sq; KASSERT(wchan != NULL, ("%s: invalid NULL wait channel", __func__)); MPASS((queue >= 0) && (queue < NR_SLEEPQS)); sq = sleepq_lookup(wchan); if (sq == NULL) return (0); return (sq->sq_blockedcnt[queue]); } /* * Marks the pending sleep of the current thread as interruptible and * makes an initial check for pending signals before putting a thread * to sleep. Enters and exits with the thread lock held. Thread lock * may have transitioned from the sleepq lock to a run lock. */ static int sleepq_catch_signals(void *wchan, int pri) { struct sleepqueue_chain *sc; struct sleepqueue *sq; struct thread *td; struct proc *p; struct sigacts *ps; int sig, ret; td = curthread; p = curproc; sc = SC_LOOKUP(wchan); mtx_assert(&sc->sc_lock, MA_OWNED); MPASS(wchan != NULL); if ((td->td_pflags & TDP_WAKEUP) != 0) { td->td_pflags &= ~TDP_WAKEUP; ret = EINTR; thread_lock(td); goto out; } /* * See if there are any pending signals for this thread. If not * we can switch immediately. Otherwise do the signal processing * directly. */ thread_lock(td); if ((td->td_flags & (TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK)) == 0) { sleepq_switch(wchan, pri); return (0); } thread_unlock(td); mtx_unlock_spin(&sc->sc_lock); CTR3(KTR_PROC, "sleepq catching signals: thread %p (pid %ld, %s)", (void *)td, (long)p->p_pid, td->td_name); PROC_LOCK(p); ps = p->p_sigacts; mtx_lock(&ps->ps_mtx); sig = cursig(td); if (sig == 0) { mtx_unlock(&ps->ps_mtx); ret = thread_suspend_check(1); MPASS(ret == 0 || ret == EINTR || ret == ERESTART); } else { if (SIGISMEMBER(ps->ps_sigintr, sig)) ret = EINTR; else ret = ERESTART; mtx_unlock(&ps->ps_mtx); } /* * Lock the per-process spinlock prior to dropping the PROC_LOCK * to avoid a signal delivery race. PROC_LOCK, PROC_SLOCK, and * thread_lock() are currently held in tdsendsignal(). */ PROC_SLOCK(p); mtx_lock_spin(&sc->sc_lock); PROC_UNLOCK(p); thread_lock(td); PROC_SUNLOCK(p); if (ret == 0) { sleepq_switch(wchan, pri); return (0); } out: /* * There were pending signals and this thread is still * on the sleep queue, remove it from the sleep queue. */ if (TD_ON_SLEEPQ(td)) { sq = sleepq_lookup(wchan); if (sleepq_resume_thread(sq, td, 0)) { #ifdef INVARIANTS /* * This thread hasn't gone to sleep yet, so it * should not be swapped out. */ panic("not waking up swapper"); #endif } } mtx_unlock_spin(&sc->sc_lock); MPASS(td->td_lock != &sc->sc_lock); return (ret); } /* * Switches to another thread if we are still asleep on a sleep queue. * Returns with thread lock. */ static void sleepq_switch(void *wchan, int pri) { struct sleepqueue_chain *sc; struct sleepqueue *sq; struct thread *td; td = curthread; sc = SC_LOOKUP(wchan); mtx_assert(&sc->sc_lock, MA_OWNED); THREAD_LOCK_ASSERT(td, MA_OWNED); /* * If we have a sleep queue, then we've already been woken up, so * just return. */ if (td->td_sleepqueue != NULL) { mtx_unlock_spin(&sc->sc_lock); return; } /* * If TDF_TIMEOUT is set, then our sleep has been timed out * already but we are still on the sleep queue, so dequeue the * thread and return. */ if (td->td_flags & TDF_TIMEOUT) { MPASS(TD_ON_SLEEPQ(td)); sq = sleepq_lookup(wchan); if (sleepq_resume_thread(sq, td, 0)) { #ifdef INVARIANTS /* * This thread hasn't gone to sleep yet, so it * should not be swapped out. */ panic("not waking up swapper"); #endif } mtx_unlock_spin(&sc->sc_lock); return; } #ifdef SLEEPQUEUE_PROFILING if (prof_enabled) sleepq_profile(td->td_wmesg); #endif MPASS(td->td_sleepqueue == NULL); sched_sleep(td, pri); thread_lock_set(td, &sc->sc_lock); SDT_PROBE0(sched, , , sleep); TD_SET_SLEEPING(td); mi_switch(SW_VOL | SWT_SLEEPQ, NULL); KASSERT(TD_IS_RUNNING(td), ("running but not TDS_RUNNING")); CTR3(KTR_PROC, "sleepq resume: thread %p (pid %ld, %s)", (void *)td, (long)td->td_proc->p_pid, (void *)td->td_name); } /* * Check to see if we timed out. */ static int sleepq_check_timeout(void) { struct thread *td; td = curthread; THREAD_LOCK_ASSERT(td, MA_OWNED); /* * If TDF_TIMEOUT is set, we timed out. */ if (td->td_flags & TDF_TIMEOUT) { td->td_flags &= ~TDF_TIMEOUT; return (EWOULDBLOCK); } /* * If TDF_TIMOFAIL is set, the timeout ran after we had * already been woken up. */ if (td->td_flags & TDF_TIMOFAIL) td->td_flags &= ~TDF_TIMOFAIL; /* * If callout_stop() fails, then the timeout is running on * another CPU, so synchronize with it to avoid having it * accidentally wake up a subsequent sleep. */ - else if (callout_stop(&td->td_slpcallout) == 0) { + else if (_callout_stop_safe(&td->td_slpcallout, CS_MIGRBLOCK, NULL) + == 0) { td->td_flags |= TDF_TIMEOUT; TD_SET_SLEEPING(td); mi_switch(SW_INVOL | SWT_SLEEPQTIMO, NULL); } return (0); } /* * Check to see if we were awoken by a signal. */ static int sleepq_check_signals(void) { struct thread *td; td = curthread; THREAD_LOCK_ASSERT(td, MA_OWNED); /* We are no longer in an interruptible sleep. */ if (td->td_flags & TDF_SINTR) td->td_flags &= ~TDF_SINTR; if (td->td_flags & TDF_SLEEPABORT) { td->td_flags &= ~TDF_SLEEPABORT; return (td->td_intrval); } return (0); } /* * Block the current thread until it is awakened from its sleep queue. */ void sleepq_wait(void *wchan, int pri) { struct thread *td; td = curthread; MPASS(!(td->td_flags & TDF_SINTR)); thread_lock(td); sleepq_switch(wchan, pri); thread_unlock(td); } /* * Block the current thread until it is awakened from its sleep queue * or it is interrupted by a signal. */ int sleepq_wait_sig(void *wchan, int pri) { int rcatch; int rval; rcatch = sleepq_catch_signals(wchan, pri); rval = sleepq_check_signals(); thread_unlock(curthread); if (rcatch) return (rcatch); return (rval); } /* * Block the current thread until it is awakened from its sleep queue * or it times out while waiting. */ int sleepq_timedwait(void *wchan, int pri) { struct thread *td; int rval; td = curthread; MPASS(!(td->td_flags & TDF_SINTR)); thread_lock(td); sleepq_switch(wchan, pri); rval = sleepq_check_timeout(); thread_unlock(td); return (rval); } /* * Block the current thread until it is awakened from its sleep queue, * it is interrupted by a signal, or it times out waiting to be awakened. */ int sleepq_timedwait_sig(void *wchan, int pri) { int rcatch, rvalt, rvals; rcatch = sleepq_catch_signals(wchan, pri); rvalt = sleepq_check_timeout(); rvals = sleepq_check_signals(); thread_unlock(curthread); if (rcatch) return (rcatch); if (rvals) return (rvals); return (rvalt); } /* * Returns the type of sleepqueue given a waitchannel. */ int sleepq_type(void *wchan) { struct sleepqueue *sq; int type; MPASS(wchan != NULL); sleepq_lock(wchan); sq = sleepq_lookup(wchan); if (sq == NULL) { sleepq_release(wchan); return (-1); } type = sq->sq_type; sleepq_release(wchan); return (type); } /* * Removes a thread from a sleep queue and makes it * runnable. */ static int sleepq_resume_thread(struct sleepqueue *sq, struct thread *td, int pri) { struct sleepqueue_chain *sc; MPASS(td != NULL); MPASS(sq->sq_wchan != NULL); MPASS(td->td_wchan == sq->sq_wchan); MPASS(td->td_sqqueue < NR_SLEEPQS && td->td_sqqueue >= 0); THREAD_LOCK_ASSERT(td, MA_OWNED); sc = SC_LOOKUP(sq->sq_wchan); mtx_assert(&sc->sc_lock, MA_OWNED); SDT_PROBE2(sched, , , wakeup, td, td->td_proc); /* Remove the thread from the queue. */ sq->sq_blockedcnt[td->td_sqqueue]--; TAILQ_REMOVE(&sq->sq_blocked[td->td_sqqueue], td, td_slpq); /* * Get a sleep queue for this thread. If this is the last waiter, * use the queue itself and take it out of the chain, otherwise, * remove a queue from the free list. */ if (LIST_EMPTY(&sq->sq_free)) { td->td_sleepqueue = sq; #ifdef INVARIANTS sq->sq_wchan = NULL; #endif #ifdef SLEEPQUEUE_PROFILING sc->sc_depth--; #endif } else td->td_sleepqueue = LIST_FIRST(&sq->sq_free); LIST_REMOVE(td->td_sleepqueue, sq_hash); td->td_wmesg = NULL; td->td_wchan = NULL; td->td_flags &= ~TDF_SINTR; CTR3(KTR_PROC, "sleepq_wakeup: thread %p (pid %ld, %s)", (void *)td, (long)td->td_proc->p_pid, td->td_name); /* Adjust priority if requested. */ MPASS(pri == 0 || (pri >= PRI_MIN && pri <= PRI_MAX)); if (pri != 0 && td->td_priority > pri && PRI_BASE(td->td_pri_class) == PRI_TIMESHARE) sched_prio(td, pri); /* * Note that thread td might not be sleeping if it is running * sleepq_catch_signals() on another CPU or is blocked on its * proc lock to check signals. There's no need to mark the * thread runnable in that case. */ if (TD_IS_SLEEPING(td)) { TD_CLR_SLEEPING(td); return (setrunnable(td)); } return (0); } #ifdef INVARIANTS /* * UMA zone item deallocator. */ static void sleepq_dtor(void *mem, int size, void *arg) { struct sleepqueue *sq; int i; sq = mem; for (i = 0; i < NR_SLEEPQS; i++) { MPASS(TAILQ_EMPTY(&sq->sq_blocked[i])); MPASS(sq->sq_blockedcnt[i] == 0); } } #endif /* * UMA zone item initializer. */ static int sleepq_init(void *mem, int size, int flags) { struct sleepqueue *sq; int i; bzero(mem, size); sq = mem; for (i = 0; i < NR_SLEEPQS; i++) { TAILQ_INIT(&sq->sq_blocked[i]); sq->sq_blockedcnt[i] = 0; } LIST_INIT(&sq->sq_free); return (0); } /* * Find the highest priority thread sleeping on a wait channel and resume it. */ int sleepq_signal(void *wchan, int flags, int pri, int queue) { struct sleepqueue *sq; struct thread *td, *besttd; int wakeup_swapper; CTR2(KTR_PROC, "sleepq_signal(%p, %d)", wchan, flags); KASSERT(wchan != NULL, ("%s: invalid NULL wait channel", __func__)); MPASS((queue >= 0) && (queue < NR_SLEEPQS)); sq = sleepq_lookup(wchan); if (sq == NULL) return (0); KASSERT(sq->sq_type == (flags & SLEEPQ_TYPE), ("%s: mismatch between sleep/wakeup and cv_*", __func__)); /* * Find the highest priority thread on the queue. If there is a * tie, use the thread that first appears in the queue as it has * been sleeping the longest since threads are always added to * the tail of sleep queues. */ besttd = NULL; TAILQ_FOREACH(td, &sq->sq_blocked[queue], td_slpq) { if (besttd == NULL || td->td_priority < besttd->td_priority) besttd = td; } MPASS(besttd != NULL); thread_lock(besttd); wakeup_swapper = sleepq_resume_thread(sq, besttd, pri); thread_unlock(besttd); return (wakeup_swapper); } /* * Resume all threads sleeping on a specified wait channel. */ int sleepq_broadcast(void *wchan, int flags, int pri, int queue) { struct sleepqueue *sq; struct thread *td, *tdn; int wakeup_swapper; CTR2(KTR_PROC, "sleepq_broadcast(%p, %d)", wchan, flags); KASSERT(wchan != NULL, ("%s: invalid NULL wait channel", __func__)); MPASS((queue >= 0) && (queue < NR_SLEEPQS)); sq = sleepq_lookup(wchan); if (sq == NULL) return (0); KASSERT(sq->sq_type == (flags & SLEEPQ_TYPE), ("%s: mismatch between sleep/wakeup and cv_*", __func__)); /* Resume all blocked threads on the sleep queue. */ wakeup_swapper = 0; TAILQ_FOREACH_SAFE(td, &sq->sq_blocked[queue], td_slpq, tdn) { thread_lock(td); if (sleepq_resume_thread(sq, td, pri)) wakeup_swapper = 1; thread_unlock(td); } return (wakeup_swapper); } /* * Time sleeping threads out. When the timeout expires, the thread is * removed from the sleep queue and made runnable if it is still asleep. */ static void sleepq_timeout(void *arg) { struct sleepqueue_chain *sc; struct sleepqueue *sq; struct thread *td; void *wchan; int wakeup_swapper; td = arg; wakeup_swapper = 0; CTR3(KTR_PROC, "sleepq_timeout: thread %p (pid %ld, %s)", (void *)td, (long)td->td_proc->p_pid, (void *)td->td_name); /* * First, see if the thread is asleep and get the wait channel if * it is. */ thread_lock(td); if (TD_IS_SLEEPING(td) && TD_ON_SLEEPQ(td)) { wchan = td->td_wchan; sc = SC_LOOKUP(wchan); THREAD_LOCKPTR_ASSERT(td, &sc->sc_lock); sq = sleepq_lookup(wchan); MPASS(sq != NULL); td->td_flags |= TDF_TIMEOUT; wakeup_swapper = sleepq_resume_thread(sq, td, 0); thread_unlock(td); if (wakeup_swapper) kick_proc0(); return; } /* * If the thread is on the SLEEPQ but isn't sleeping yet, it * can either be on another CPU in between sleepq_add() and * one of the sleepq_*wait*() routines or it can be in * sleepq_catch_signals(). */ if (TD_ON_SLEEPQ(td)) { td->td_flags |= TDF_TIMEOUT; thread_unlock(td); return; } /* * Now check for the edge cases. First, if TDF_TIMEOUT is set, * then the other thread has already yielded to us, so clear * the flag and resume it. If TDF_TIMEOUT is not set, then the * we know that the other thread is not on a sleep queue, but it * hasn't resumed execution yet. In that case, set TDF_TIMOFAIL * to let it know that the timeout has already run and doesn't * need to be canceled. */ if (td->td_flags & TDF_TIMEOUT) { MPASS(TD_IS_SLEEPING(td)); td->td_flags &= ~TDF_TIMEOUT; TD_CLR_SLEEPING(td); wakeup_swapper = setrunnable(td); } else td->td_flags |= TDF_TIMOFAIL; thread_unlock(td); if (wakeup_swapper) kick_proc0(); } /* * Resumes a specific thread from the sleep queue associated with a specific * wait channel if it is on that queue. */ void sleepq_remove(struct thread *td, void *wchan) { struct sleepqueue *sq; int wakeup_swapper; /* * Look up the sleep queue for this wait channel, then re-check * that the thread is asleep on that channel, if it is not, then * bail. */ MPASS(wchan != NULL); sleepq_lock(wchan); sq = sleepq_lookup(wchan); /* * We can not lock the thread here as it may be sleeping on a * different sleepq. However, holding the sleepq lock for this * wchan can guarantee that we do not miss a wakeup for this * channel. The asserts below will catch any false positives. */ if (!TD_ON_SLEEPQ(td) || td->td_wchan != wchan) { sleepq_release(wchan); return; } /* Thread is asleep on sleep queue sq, so wake it up. */ thread_lock(td); MPASS(sq != NULL); MPASS(td->td_wchan == wchan); wakeup_swapper = sleepq_resume_thread(sq, td, 0); thread_unlock(td); sleepq_release(wchan); if (wakeup_swapper) kick_proc0(); } /* * Abort a thread as if an interrupt had occurred. Only abort * interruptible waits (unfortunately it isn't safe to abort others). */ int sleepq_abort(struct thread *td, int intrval) { struct sleepqueue *sq; void *wchan; THREAD_LOCK_ASSERT(td, MA_OWNED); MPASS(TD_ON_SLEEPQ(td)); MPASS(td->td_flags & TDF_SINTR); MPASS(intrval == EINTR || intrval == ERESTART); /* * If the TDF_TIMEOUT flag is set, just leave. A * timeout is scheduled anyhow. */ if (td->td_flags & TDF_TIMEOUT) return (0); CTR3(KTR_PROC, "sleepq_abort: thread %p (pid %ld, %s)", (void *)td, (long)td->td_proc->p_pid, (void *)td->td_name); td->td_intrval = intrval; td->td_flags |= TDF_SLEEPABORT; /* * If the thread has not slept yet it will find the signal in * sleepq_catch_signals() and call sleepq_resume_thread. Otherwise * we have to do it here. */ if (!TD_IS_SLEEPING(td)) return (0); wchan = td->td_wchan; MPASS(wchan != NULL); sq = sleepq_lookup(wchan); MPASS(sq != NULL); /* Thread is asleep on sleep queue sq, so wake it up. */ return (sleepq_resume_thread(sq, td, 0)); } #ifdef SLEEPQUEUE_PROFILING #define SLEEPQ_PROF_LOCATIONS 1024 #define SLEEPQ_SBUFSIZE 512 struct sleepq_prof { LIST_ENTRY(sleepq_prof) sp_link; const char *sp_wmesg; long sp_count; }; LIST_HEAD(sqphead, sleepq_prof); struct sqphead sleepq_prof_free; struct sqphead sleepq_hash[SC_TABLESIZE]; static struct sleepq_prof sleepq_profent[SLEEPQ_PROF_LOCATIONS]; static struct mtx sleepq_prof_lock; MTX_SYSINIT(sleepq_prof_lock, &sleepq_prof_lock, "sleepq_prof", MTX_SPIN); static void sleepq_profile(const char *wmesg) { struct sleepq_prof *sp; mtx_lock_spin(&sleepq_prof_lock); if (prof_enabled == 0) goto unlock; LIST_FOREACH(sp, &sleepq_hash[SC_HASH(wmesg)], sp_link) if (sp->sp_wmesg == wmesg) goto done; sp = LIST_FIRST(&sleepq_prof_free); if (sp == NULL) goto unlock; sp->sp_wmesg = wmesg; LIST_REMOVE(sp, sp_link); LIST_INSERT_HEAD(&sleepq_hash[SC_HASH(wmesg)], sp, sp_link); done: sp->sp_count++; unlock: mtx_unlock_spin(&sleepq_prof_lock); return; } static void sleepq_prof_reset(void) { struct sleepq_prof *sp; int enabled; int i; mtx_lock_spin(&sleepq_prof_lock); enabled = prof_enabled; prof_enabled = 0; for (i = 0; i < SC_TABLESIZE; i++) LIST_INIT(&sleepq_hash[i]); LIST_INIT(&sleepq_prof_free); for (i = 0; i < SLEEPQ_PROF_LOCATIONS; i++) { sp = &sleepq_profent[i]; sp->sp_wmesg = NULL; sp->sp_count = 0; LIST_INSERT_HEAD(&sleepq_prof_free, sp, sp_link); } prof_enabled = enabled; mtx_unlock_spin(&sleepq_prof_lock); } static int enable_sleepq_prof(SYSCTL_HANDLER_ARGS) { int error, v; v = prof_enabled; error = sysctl_handle_int(oidp, &v, v, req); if (error) return (error); if (req->newptr == NULL) return (error); if (v == prof_enabled) return (0); if (v == 1) sleepq_prof_reset(); mtx_lock_spin(&sleepq_prof_lock); prof_enabled = !!v; mtx_unlock_spin(&sleepq_prof_lock); return (0); } static int reset_sleepq_prof_stats(SYSCTL_HANDLER_ARGS) { int error, v; v = 0; error = sysctl_handle_int(oidp, &v, 0, req); if (error) return (error); if (req->newptr == NULL) return (error); if (v == 0) return (0); sleepq_prof_reset(); return (0); } static int dump_sleepq_prof_stats(SYSCTL_HANDLER_ARGS) { struct sleepq_prof *sp; struct sbuf *sb; int enabled; int error; int i; error = sysctl_wire_old_buffer(req, 0); if (error != 0) return (error); sb = sbuf_new_for_sysctl(NULL, NULL, SLEEPQ_SBUFSIZE, req); sbuf_printf(sb, "\nwmesg\tcount\n"); enabled = prof_enabled; mtx_lock_spin(&sleepq_prof_lock); prof_enabled = 0; mtx_unlock_spin(&sleepq_prof_lock); for (i = 0; i < SC_TABLESIZE; i++) { LIST_FOREACH(sp, &sleepq_hash[i], sp_link) { sbuf_printf(sb, "%s\t%ld\n", sp->sp_wmesg, sp->sp_count); } } mtx_lock_spin(&sleepq_prof_lock); prof_enabled = enabled; mtx_unlock_spin(&sleepq_prof_lock); error = sbuf_finish(sb); sbuf_delete(sb); return (error); } SYSCTL_PROC(_debug_sleepq, OID_AUTO, stats, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0, dump_sleepq_prof_stats, "A", "Sleepqueue profiling statistics"); SYSCTL_PROC(_debug_sleepq, OID_AUTO, reset, CTLTYPE_INT | CTLFLAG_RW, NULL, 0, reset_sleepq_prof_stats, "I", "Reset sleepqueue profiling statistics"); SYSCTL_PROC(_debug_sleepq, OID_AUTO, enable, CTLTYPE_INT | CTLFLAG_RW, NULL, 0, enable_sleepq_prof, "I", "Enable sleepqueue profiling"); #endif #ifdef DDB DB_SHOW_COMMAND(sleepq, db_show_sleepqueue) { struct sleepqueue_chain *sc; struct sleepqueue *sq; #ifdef INVARIANTS struct lock_object *lock; #endif struct thread *td; void *wchan; int i; if (!have_addr) return; /* * First, see if there is an active sleep queue for the wait channel * indicated by the address. */ wchan = (void *)addr; sc = SC_LOOKUP(wchan); LIST_FOREACH(sq, &sc->sc_queues, sq_hash) if (sq->sq_wchan == wchan) goto found; /* * Second, see if there is an active sleep queue at the address * indicated. */ for (i = 0; i < SC_TABLESIZE; i++) LIST_FOREACH(sq, &sleepq_chains[i].sc_queues, sq_hash) { if (sq == (struct sleepqueue *)addr) goto found; } db_printf("Unable to locate a sleep queue via %p\n", (void *)addr); return; found: db_printf("Wait channel: %p\n", sq->sq_wchan); db_printf("Queue type: %d\n", sq->sq_type); #ifdef INVARIANTS if (sq->sq_lock) { lock = sq->sq_lock; db_printf("Associated Interlock: %p - (%s) %s\n", lock, LOCK_CLASS(lock)->lc_name, lock->lo_name); } #endif db_printf("Blocked threads:\n"); for (i = 0; i < NR_SLEEPQS; i++) { db_printf("\nQueue[%d]:\n", i); if (TAILQ_EMPTY(&sq->sq_blocked[i])) db_printf("\tempty\n"); else TAILQ_FOREACH(td, &sq->sq_blocked[0], td_slpq) { db_printf("\t%p (tid %d, pid %d, \"%s\")\n", td, td->td_tid, td->td_proc->p_pid, td->td_name); } db_printf("(expected: %u)\n", sq->sq_blockedcnt[i]); } } /* Alias 'show sleepqueue' to 'show sleepq'. */ DB_SHOW_ALIAS(sleepqueue, db_show_sleepqueue); #endif Index: projects/release-pkg/sys/sys/callout.h =================================================================== --- projects/release-pkg/sys/sys/callout.h (revision 296327) +++ projects/release-pkg/sys/sys/callout.h (revision 296328) @@ -1,129 +1,135 @@ /*- * Copyright (c) 1990, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 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. * * @(#)callout.h 8.2 (Berkeley) 1/21/94 * $FreeBSD$ */ #ifndef _SYS_CALLOUT_H_ #define _SYS_CALLOUT_H_ #include #define CALLOUT_LOCAL_ALLOC 0x0001 /* was allocated from callfree */ #define CALLOUT_ACTIVE 0x0002 /* callout is currently active */ #define CALLOUT_PENDING 0x0004 /* callout is waiting for timeout */ #define CALLOUT_MPSAFE 0x0008 /* deprecated */ #define CALLOUT_RETURNUNLOCKED 0x0010 /* handler returns with mtx unlocked */ #define CALLOUT_SHAREDLOCK 0x0020 /* callout lock held in shared mode */ #define CALLOUT_DFRMIGRATION 0x0040 /* callout in deferred migration mode */ #define CALLOUT_PROCESSED 0x0080 /* callout in wheel or processing list? */ #define CALLOUT_DIRECT 0x0100 /* allow exec from hw int context */ #define C_DIRECT_EXEC 0x0001 /* direct execution of callout */ #define C_PRELBITS 7 #define C_PRELRANGE ((1 << C_PRELBITS) - 1) #define C_PREL(x) (((x) + 1) << 1) #define C_PRELGET(x) (int)((((x) >> 1) & C_PRELRANGE) - 1) #define C_HARDCLOCK 0x0100 /* align to hardclock() calls */ #define C_ABSOLUTE 0x0200 /* event time is absolute. */ struct callout_handle { struct callout *callout; }; +/* Flags for callout_stop_safe() */ +#define CS_DRAIN 0x0001 /* callout_drain(), wait allowed */ +#define CS_MIGRBLOCK 0x0002 /* Block migration, return value + indicates that the callout was + executing */ + #ifdef _KERNEL /* * Note the flags field is actually *two* fields. The c_flags * field is the one that caller operations that may, or may not have * a lock touches i.e. callout_deactivate(). The other, the c_iflags, * is the internal flags that *must* be kept correct on which the * callout system depend on e.g. callout_pending(). * The c_iflag is used internally by the callout system to determine which * list the callout is on and track internal state. Callers *should not* * use the c_flags field directly but should use the macros provided. * * The c_iflags field holds internal flags that are protected by internal * locks of the callout subsystem. The c_flags field holds external flags. * The caller must hold its own lock while manipulating or reading external * flags via callout_active(), callout_deactivate(), callout_reset*(), or * callout_stop() to avoid races. */ #define callout_active(c) ((c)->c_flags & CALLOUT_ACTIVE) #define callout_deactivate(c) ((c)->c_flags &= ~CALLOUT_ACTIVE) -#define callout_drain(c) _callout_stop_safe(c, 1, NULL) +#define callout_drain(c) _callout_stop_safe(c, CS_DRAIN, NULL) void callout_init(struct callout *, int); void _callout_init_lock(struct callout *, struct lock_object *, int); #define callout_init_mtx(c, mtx, flags) \ _callout_init_lock((c), ((mtx) != NULL) ? &(mtx)->lock_object : \ NULL, (flags)) #define callout_init_rm(c, rm, flags) \ _callout_init_lock((c), ((rm) != NULL) ? &(rm)->lock_object : \ NULL, (flags)) #define callout_init_rw(c, rw, flags) \ _callout_init_lock((c), ((rw) != NULL) ? &(rw)->lock_object : \ NULL, (flags)) #define callout_pending(c) ((c)->c_iflags & CALLOUT_PENDING) int callout_reset_sbt_on(struct callout *, sbintime_t, sbintime_t, void (*)(void *), void *, int, int); #define callout_reset_sbt(c, sbt, pr, fn, arg, flags) \ callout_reset_sbt_on((c), (sbt), (pr), (fn), (arg), -1, (flags)) #define callout_reset_sbt_curcpu(c, sbt, pr, fn, arg, flags) \ callout_reset_sbt_on((c), (sbt), (pr), (fn), (arg), PCPU_GET(cpuid),\ (flags)) #define callout_reset_on(c, to_ticks, fn, arg, cpu) \ callout_reset_sbt_on((c), tick_sbt * (to_ticks), 0, (fn), (arg), \ (cpu), C_HARDCLOCK) #define callout_reset(c, on_tick, fn, arg) \ callout_reset_on((c), (on_tick), (fn), (arg), -1) #define callout_reset_curcpu(c, on_tick, fn, arg) \ callout_reset_on((c), (on_tick), (fn), (arg), PCPU_GET(cpuid)) #define callout_schedule_sbt_on(c, sbt, pr, cpu, flags) \ callout_reset_sbt_on((c), (sbt), (pr), (c)->c_func, (c)->c_arg, \ (cpu), (flags)) #define callout_schedule_sbt(c, sbt, pr, flags) \ callout_schedule_sbt_on((c), (sbt), (pr), -1, (flags)) #define callout_schedule_sbt_curcpu(c, sbt, pr, flags) \ callout_schedule_sbt_on((c), (sbt), (pr), PCPU_GET(cpuid), (flags)) int callout_schedule(struct callout *, int); int callout_schedule_on(struct callout *, int, int); #define callout_schedule_curcpu(c, on_tick) \ callout_schedule_on((c), (on_tick), PCPU_GET(cpuid)) #define callout_stop(c) _callout_stop_safe(c, 0, NULL) int _callout_stop_safe(struct callout *, int, void (*)(void *)); void callout_process(sbintime_t now); #define callout_async_drain(c, d) \ _callout_stop_safe(c, 0, d) #endif #endif /* _SYS_CALLOUT_H_ */ Index: projects/release-pkg/sys =================================================================== --- projects/release-pkg/sys (revision 296327) +++ projects/release-pkg/sys (revision 296328) Property changes on: projects/release-pkg/sys ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/sys:r296315-296327 Index: projects/release-pkg/usr.sbin/daemon/daemon.8 =================================================================== --- projects/release-pkg/usr.sbin/daemon/daemon.8 (revision 296327) +++ projects/release-pkg/usr.sbin/daemon/daemon.8 (revision 296328) @@ -1,163 +1,167 @@ .\" Copyright (c) 1999 Berkeley Software Design, Inc. 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. Berkeley Software Design Inc's name may not be used to endorse or .\" promote products derived from this software without specific prior .\" written permission. .\" .\" THIS SOFTWARE IS PROVIDED BY BERKELEY SOFTWARE DESIGN INC ``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 BERKELEY SOFTWARE DESIGN INC 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. .\" .\" $FreeBSD$ .\" -.Dd September 13, 2013 +.Dd March 2, 2016 .Dt DAEMON 8 .Os .Sh NAME .Nm daemon .Nd run detached from the controlling terminal .Sh SYNOPSIS .Nm .Op Fl cfr .Op Fl p Ar child_pidfile .Op Fl P Ar supervisor_pidfile +.Op Fl t Ar title .Op Fl u Ar user .Ar command arguments ... .Sh DESCRIPTION The .Nm utility detaches itself from the controlling terminal and executes the program specified by its arguments. Privileges may be lowered to the specified user. .Pp The options are as follows: .Bl -tag -width indent .It Fl c Change the current working directory to the root .Pq Dq Pa / . .It Fl f Redirect standard input, standard output and standard error to .Pa /dev/null . .It Fl p Ar child_pidfile Write the ID of the created process into the .Ar child_pidfile using the .Xr pidfile 3 functionality. The program is executed in a spawned child process while the .Nm waits until it terminates to keep the .Ar child_pidfile locked and removes it after the process exits. The .Ar child_pidfile owner is the user who runs the .Nm regardless of whether the .Fl u option is used or not. .It Fl P Ar supervisor_pidfile Write the ID of the .Nm process into the .Ar supervisor_pidfile using the .Xr pidfile 3 functionality. The program is executed in a spawned child process while the .Nm waits until it terminates to keep the .Ar supervisor_pidfile locked and removes it after the process exits. The .Ar supervisor_pidfile owner is the user who runs the .Nm regardless of whether the .Fl u option is used or not. .It Fl r Supervise and restart the program if it has been terminated. +.It Fl t Ar title +Process title for the daemon to make it easily identifiable. .It Fl u Ar user Login name of the user to execute the program under. Requires adequate superuser privileges. .El .Pp If the .Fl p , .Fl P or .Fl r option is specified the program is executed in a spawned child process. The .Nm waits until it terminates to keep the pid file(s) locked and removes them after the process exits or restarts the program. In this case if the monitoring .Nm receives software termination signal (SIGTERM) it forwards it to the spawned process. Normally it will cause the child to exit, remove the pidfile(s) and then terminate. .Pp The .Fl P option is useful combined with the .Fl r option as .Ar supervisor_pidfile contains the ID of the supervisor -not the child. This is especially important if you use +not the child. +This is especially important if you use .Fl r in an rc script as the .Fl p option will give you the child's ID to signal when you attempt to stop the service, causing .Nm to restart the child. .Sh EXIT STATUS The .Nm utility exits 1 if an error is returned by the .Xr daemon 3 library routine, 2 if .Ar child_pidfile or .Ar supervisor_pidfile is requested, but cannot be opened, 3 if process is already running (pidfile exists and is locked), otherwise 0. .Sh DIAGNOSTICS If the command cannot be executed, an error message is displayed on standard error unless the .Fl f flag is specified. .Sh SEE ALSO .Xr setregid 2 , .Xr setreuid 2 , .Xr daemon 3 , .Xr exec 3 , .Xr pidfile 3 , .Xr termios 4 , .Xr tty 4 .Sh HISTORY The .Nm utility first appeared in .Fx 4.7 . Index: projects/release-pkg/usr.sbin/daemon/daemon.c =================================================================== --- projects/release-pkg/usr.sbin/daemon/daemon.c (revision 296327) +++ projects/release-pkg/usr.sbin/daemon/daemon.c (revision 296328) @@ -1,276 +1,282 @@ /*- * Copyright (c) 1999 Berkeley Software Design, Inc. 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. Berkeley Software Design Inc's name may not be used to endorse or * promote products derived from this software without specific prior * written permission. * * THIS SOFTWARE IS PROVIDED BY BERKELEY SOFTWARE DESIGN INC ``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 BERKELEY SOFTWARE DESIGN INC 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 BSDI: daemon.c,v 1.2 1996/08/15 01:11:09 jch Exp */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include static void dummy_sighandler(int); static void restrict_process(const char *); static int wait_child(pid_t pid, sigset_t *mask); static void usage(void); int main(int argc, char *argv[]) { struct pidfh *ppfh, *pfh; sigset_t mask, oldmask; int ch, nochdir, noclose, restart, serrno; - const char *pidfile, *ppidfile, *user; + const char *pidfile, *ppidfile, *title, *user; pid_t otherpid, pid; nochdir = noclose = 1; restart = 0; - ppidfile = pidfile = user = NULL; - while ((ch = getopt(argc, argv, "cfp:P:ru:")) != -1) { + ppidfile = pidfile = title = user = NULL; + while ((ch = getopt(argc, argv, "cfp:P:rt:u:")) != -1) { switch (ch) { case 'c': nochdir = 0; break; case 'f': noclose = 0; break; case 'p': pidfile = optarg; break; case 'P': ppidfile = optarg; break; case 'r': restart = 1; break; + case 't': + title = optarg; + break; case 'u': user = optarg; break; default: usage(); } } argc -= optind; argv += optind; if (argc == 0) usage(); ppfh = pfh = NULL; /* * Try to open the pidfile before calling daemon(3), * to be able to report the error intelligently */ if (pidfile != NULL) { pfh = pidfile_open(pidfile, 0600, &otherpid); if (pfh == NULL) { if (errno == EEXIST) { errx(3, "process already running, pid: %d", otherpid); } err(2, "pidfile ``%s''", pidfile); } } /* Do the same for actual daemon process. */ if (ppidfile != NULL) { ppfh = pidfile_open(ppidfile, 0600, &otherpid); if (ppfh == NULL) { serrno = errno; pidfile_remove(pfh); errno = serrno; if (errno == EEXIST) { errx(3, "process already running, pid: %d", otherpid); } err(2, "ppidfile ``%s''", ppidfile); } } if (daemon(nochdir, noclose) == -1) { warn("daemon"); goto exit; } /* Write out parent pidfile if needed. */ pidfile_write(ppfh); /* * If the pidfile or restart option is specified the daemon * executes the command in a forked process and wait on child * exit to remove the pidfile or restart the command. Normally * we don't want the monitoring daemon to be terminated * leaving the running process and the stale pidfile, so we * catch SIGTERM and forward it to the children expecting to * get SIGCHLD eventually. */ pid = -1; if (pidfile != NULL || ppidfile != NULL || restart) { /* * Restore default action for SIGTERM in case the * parent process decided to ignore it. */ if (signal(SIGTERM, SIG_DFL) == SIG_ERR) { warn("signal"); goto exit; } /* * Because SIGCHLD is ignored by default, setup dummy handler * for it, so we can mask it. */ if (signal(SIGCHLD, dummy_sighandler) == SIG_ERR) { warn("signal"); goto exit; } /* * Block interesting signals. */ sigemptyset(&mask); sigaddset(&mask, SIGTERM); sigaddset(&mask, SIGCHLD); if (sigprocmask(SIG_SETMASK, &mask, &oldmask) == -1) { warn("sigprocmask"); goto exit; } /* * Try to protect against pageout kill. Ignore the * error, madvise(2) will fail only if a process does * not have superuser privileges. */ (void)madvise(NULL, 0, MADV_PROTECT); restart: /* * Spawn a child to exec the command, so in the parent * we could wait for it to exit and remove pidfile. */ pid = fork(); if (pid == -1) { warn("fork"); goto exit; } } if (pid <= 0) { if (pid == 0) { /* Restore old sigmask in the child. */ if (sigprocmask(SIG_SETMASK, &oldmask, NULL) == -1) err(1, "sigprocmask"); } /* Now that we are the child, write out the pid. */ pidfile_write(pfh); if (user != NULL) restrict_process(user); execvp(argv[0], argv); /* * execvp() failed -- report the error. The child is * now running, so the exit status doesn't matter. */ err(1, "%s", argv[0]); } - setproctitle("%s[%d]", argv[0], pid); + if (title != NULL) + setproctitle("%s[%d]", title, pid); + else + setproctitle("%s[%d]", argv[0], pid); if (wait_child(pid, &mask) == 0 && restart) { sleep(1); goto restart; } exit: pidfile_remove(pfh); pidfile_remove(ppfh); exit(1); /* If daemon(3) succeeded exit status does not matter. */ } static void dummy_sighandler(int sig __unused) { /* Nothing to do. */ } static void restrict_process(const char *user) { struct passwd *pw = NULL; pw = getpwnam(user); if (pw == NULL) errx(1, "unknown user: %s", user); if (setusercontext(NULL, pw, pw->pw_uid, LOGIN_SETALL) != 0) errx(1, "failed to set user environment"); } static int wait_child(pid_t pid, sigset_t *mask) { int terminate, signo; terminate = 0; for (;;) { if (sigwait(mask, &signo) == -1) { warn("sigwaitinfo"); return (-1); } switch (signo) { case SIGCHLD: if (waitpid(pid, NULL, WNOHANG) == -1) { warn("waitpid"); return (-1); } return (terminate); case SIGTERM: terminate = 1; if (kill(pid, signo) == -1) { warn("kill"); return (-1); } continue; default: warnx("sigwaitinfo: invalid signal: %d", signo); return (-1); } } } static void usage(void) { (void)fprintf(stderr, "usage: daemon [-cfr] [-p child_pidfile] [-P supervisor_pidfile] " "[-u user]\n command arguments ...\n"); exit(1); } Index: projects/release-pkg =================================================================== --- projects/release-pkg (revision 296327) +++ projects/release-pkg (revision 296328) Property changes on: projects/release-pkg ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head:r296315-296327