Page MenuHomeFreeBSD
Files F163250043

D51349.id158615.diff
No OneTemporary
Actions

D51349.id158615.diff
View Options

diff --git a/sys/conf/files b/sys/conf/files
--- a/sys/conf/files
+++ b/sys/conf/files
@@ -3755,6 +3755,7 @@
kern/bus_if.m standard
kern/clock_if.m standard
+kern/coredump_vnode.c standard
kern/cpufreq_if.m standard
kern/device_if.m standard
kern/imgact_binmisc.c optional imgact_binmisc
diff --git a/sys/kern/coredump_vnode.c b/sys/kern/coredump_vnode.c
new file mode 100644
--- /dev/null
+++ b/sys/kern/coredump_vnode.c
@@ -0,0 +1,520 @@
+/*-
+ * SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (c) 1993, David Greenman
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/acct.h>
+#include <sys/compressor.h>
+#include <sys/coredump.h>
+#include <sys/devctl.h>
+#include <sys/fcntl.h>
+#include <sys/jail.h>
+#include <sys/limits.h>
+#include <sys/namei.h>
+#include <sys/proc.h>
+#include <sys/sbuf.h>
+#include <sys/stat.h>
+#include <sys/sysctl.h>
+#include <sys/sysent.h>
+#include <sys/syslog.h>
+#include <sys/unistd.h>
+#include <sys/vnode.h>
+
+#include <security/audit/audit.h>
+
+#define GZIP_SUFFIX ".gz"
+#define ZSTD_SUFFIX ".zst"
+
+#define MAX_NUM_CORE_FILES 100000
+#ifndef NUM_CORE_FILES
+#define NUM_CORE_FILES 5
+#endif
+
+CTASSERT(NUM_CORE_FILES >= 0 && NUM_CORE_FILES <= MAX_NUM_CORE_FILES);
+static int num_cores = NUM_CORE_FILES;
+
+static int capmode_coredump;
+SYSCTL_INT(_kern, OID_AUTO, capmode_coredump, CTLFLAG_RWTUN,
+ &capmode_coredump, 0, "Allow processes in capability mode to dump core");
+
+static int set_core_nodump_flag = 0;
+SYSCTL_INT(_kern, OID_AUTO, nodump_coredump, CTLFLAG_RW, &set_core_nodump_flag,
+ 0, "Enable setting the NODUMP flag on coredump files");
+
+static int coredump_devctl = 0;
+SYSCTL_INT(_kern, OID_AUTO, coredump_devctl, CTLFLAG_RW, &coredump_devctl,
+ 0, "Generate a devctl notification when processes coredump");
+
+/*
+ * Protect the access to corefilename[] by allproc_lock.
+ */
+#define corefilename_lock allproc_lock
+
+static char corefilename[MAXPATHLEN] = {"%N.core"};
+TUNABLE_STR("kern.corefile", corefilename, sizeof(corefilename));
+
+static int
+sysctl_kern_corefile(SYSCTL_HANDLER_ARGS)
+{
+ int error;
+
+ sx_xlock(&corefilename_lock);
+ error = sysctl_handle_string(oidp, corefilename, sizeof(corefilename),
+ req);
+ sx_xunlock(&corefilename_lock);
+
+ return (error);
+}
+SYSCTL_PROC(_kern, OID_AUTO, corefile, CTLTYPE_STRING | CTLFLAG_RW |
+ CTLFLAG_MPSAFE, 0, 0, sysctl_kern_corefile, "A",
+ "Process corefile name format string");
+
+static int
+sysctl_debug_num_cores_check (SYSCTL_HANDLER_ARGS)
+{
+ int error;
+ int new_val;
+
+ new_val = num_cores;
+ error = sysctl_handle_int(oidp, &new_val, 0, req);
+ if (error != 0 || req->newptr == NULL)
+ return (error);
+ if (new_val > MAX_NUM_CORE_FILES)
+ new_val = MAX_NUM_CORE_FILES;
+ if (new_val < 0)
+ new_val = 0;
+ num_cores = new_val;
+ return (0);
+}
+SYSCTL_PROC(_debug, OID_AUTO, ncores,
+ CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, sizeof(int),
+ sysctl_debug_num_cores_check, "I",
+ "Maximum number of generated process corefiles while using index format");
+
+static void
+vnode_close_locked(struct thread *td, struct vnode *vp)
+{
+
+ VOP_UNLOCK(vp);
+ vn_close(vp, FWRITE, td->td_ucred, td);
+}
+
+int
+core_vn_write(const struct coredump_writer *cdw, const void *base, size_t len,
+ off_t offset, enum uio_seg seg, struct ucred *cred, size_t *resid,
+ struct thread *td)
+{
+ struct coredump_vnode_ctx *ctx = cdw->ctx;
+
+ return (vn_rdwr_inchunks(UIO_WRITE, ctx->vp, __DECONST(void *, base),
+ len, offset, seg, IO_UNIT | IO_DIRECT | IO_RANGELOCKED,
+ cred, ctx->fcred, resid, td));
+}
+
+int
+core_vn_extend(const struct coredump_writer *cdw, off_t newsz,
+ struct ucred *cred)
+{
+ struct coredump_vnode_ctx *ctx = cdw->ctx;
+ struct mount *mp;
+ int error;
+
+ error = vn_start_write(ctx->vp, &mp, V_WAIT);
+ if (error != 0)
+ return (error);
+ vn_lock(ctx->vp, LK_EXCLUSIVE | LK_RETRY);
+ error = vn_truncate_locked(ctx->vp, newsz, false, cred);
+ VOP_UNLOCK(ctx->vp);
+ vn_finished_write(mp);
+ return (error);
+}
+
+/*
+ * If the core format has a %I in it, then we need to check
+ * for existing corefiles before defining a name.
+ * To do this we iterate over 0..ncores to find a
+ * non-existing core file name to use. If all core files are
+ * already used we choose the oldest one.
+ */
+static int
+corefile_open_last(struct thread *td, char *name, int indexpos,
+ int indexlen, int ncores, struct vnode **vpp)
+{
+ struct vnode *oldvp, *nextvp, *vp;
+ struct vattr vattr;
+ struct nameidata nd;
+ int error, i, flags, oflags, cmode;
+ char ch;
+ struct timespec lasttime;
+
+ nextvp = oldvp = NULL;
+ cmode = S_IRUSR | S_IWUSR;
+ oflags = VN_OPEN_NOAUDIT | VN_OPEN_NAMECACHE |
+ (capmode_coredump ? VN_OPEN_NOCAPCHECK : 0);
+
+ for (i = 0; i < ncores; i++) {
+ flags = O_CREAT | FWRITE | O_NOFOLLOW;
+
+ ch = name[indexpos + indexlen];
+ (void)snprintf(name + indexpos, indexlen + 1, "%.*u", indexlen,
+ i);
+ name[indexpos + indexlen] = ch;
+
+ NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, name);
+ error = vn_open_cred(&nd, &flags, cmode, oflags, td->td_ucred,
+ NULL);
+ if (error != 0)
+ break;
+
+ vp = nd.ni_vp;
+ NDFREE_PNBUF(&nd);
+ if ((flags & O_CREAT) == O_CREAT) {
+ nextvp = vp;
+ break;
+ }
+
+ error = VOP_GETATTR(vp, &vattr, td->td_ucred);
+ if (error != 0) {
+ vnode_close_locked(td, vp);
+ break;
+ }
+
+ if (oldvp == NULL ||
+ lasttime.tv_sec > vattr.va_mtime.tv_sec ||
+ (lasttime.tv_sec == vattr.va_mtime.tv_sec &&
+ lasttime.tv_nsec >= vattr.va_mtime.tv_nsec)) {
+ if (oldvp != NULL)
+ vn_close(oldvp, FWRITE, td->td_ucred, td);
+ oldvp = vp;
+ VOP_UNLOCK(oldvp);
+ lasttime = vattr.va_mtime;
+ } else {
+ vnode_close_locked(td, vp);
+ }
+ }
+
+ if (oldvp != NULL) {
+ if (nextvp == NULL) {
+ if ((td->td_proc->p_flag & P_SUGID) != 0) {
+ error = EFAULT;
+ vn_close(oldvp, FWRITE, td->td_ucred, td);
+ } else {
+ nextvp = oldvp;
+ error = vn_lock(nextvp, LK_EXCLUSIVE);
+ if (error != 0) {
+ vn_close(nextvp, FWRITE, td->td_ucred,
+ td);
+ nextvp = NULL;
+ }
+ }
+ } else {
+ vn_close(oldvp, FWRITE, td->td_ucred, td);
+ }
+ }
+ if (error != 0) {
+ if (nextvp != NULL)
+ vnode_close_locked(td, oldvp);
+ } else {
+ *vpp = nextvp;
+ }
+
+ return (error);
+}
+
+/*
+ * corefile_open(comm, uid, pid, td, compress, vpp, namep)
+ * Expand the name described in corefilename, using name, uid, and pid
+ * and open/create core file.
+ * corefilename is a printf-like string, with three format specifiers:
+ * %N name of process ("name")
+ * %P process id (pid)
+ * %U user id (uid)
+ * For example, "%N.core" is the default; they can be disabled completely
+ * by using "/dev/null", or all core files can be stored in "/cores/%U/%N-%P".
+ * This is controlled by the sysctl variable kern.corefile (see above).
+ */
+static int
+corefile_open(const char *comm, uid_t uid, pid_t pid, struct thread *td,
+ int compress, int signum, struct vnode **vpp, char **namep)
+{
+ struct sbuf sb;
+ struct nameidata nd;
+ const char *format;
+ char *hostname, *name;
+ int cmode, error, flags, i, indexpos, indexlen, oflags, ncores;
+
+ hostname = NULL;
+ format = corefilename;
+ name = malloc(MAXPATHLEN, M_TEMP, M_WAITOK | M_ZERO);
+ indexlen = 0;
+ indexpos = -1;
+ ncores = num_cores;
+ (void)sbuf_new(&sb, name, MAXPATHLEN, SBUF_FIXEDLEN);
+ sx_slock(&corefilename_lock);
+ for (i = 0; format[i] != '\0'; i++) {
+ switch (format[i]) {
+ case '%': /* Format character */
+ i++;
+ switch (format[i]) {
+ case '%':
+ sbuf_putc(&sb, '%');
+ break;
+ case 'H': /* hostname */
+ if (hostname == NULL) {
+ hostname = malloc(MAXHOSTNAMELEN,
+ M_TEMP, M_WAITOK);
+ }
+ getcredhostname(td->td_ucred, hostname,
+ MAXHOSTNAMELEN);
+ sbuf_cat(&sb, hostname);
+ break;
+ case 'I': /* autoincrementing index */
+ if (indexpos != -1) {
+ sbuf_printf(&sb, "%%I");
+ break;
+ }
+
+ indexpos = sbuf_len(&sb);
+ sbuf_printf(&sb, "%u", ncores - 1);
+ indexlen = sbuf_len(&sb) - indexpos;
+ break;
+ case 'N': /* process name */
+ sbuf_printf(&sb, "%s", comm);
+ break;
+ case 'P': /* process id */
+ sbuf_printf(&sb, "%u", pid);
+ break;
+ case 'S': /* signal number */
+ sbuf_printf(&sb, "%i", signum);
+ break;
+ case 'U': /* user id */
+ sbuf_printf(&sb, "%u", uid);
+ break;
+ default:
+ log(LOG_ERR,
+ "Unknown format character %c in "
+ "corename `%s'\n", format[i], format);
+ break;
+ }
+ break;
+ default:
+ sbuf_putc(&sb, format[i]);
+ break;
+ }
+ }
+ sx_sunlock(&corefilename_lock);
+ free(hostname, M_TEMP);
+ if (compress == COMPRESS_GZIP)
+ sbuf_cat(&sb, GZIP_SUFFIX);
+ else if (compress == COMPRESS_ZSTD)
+ sbuf_cat(&sb, ZSTD_SUFFIX);
+ if (sbuf_error(&sb) != 0) {
+ log(LOG_ERR, "pid %ld (%s), uid (%lu): corename is too "
+ "long\n", (long)pid, comm, (u_long)uid);
+ sbuf_delete(&sb);
+ free(name, M_TEMP);
+ return (ENOMEM);
+ }
+ sbuf_finish(&sb);
+ sbuf_delete(&sb);
+
+ if (indexpos != -1) {
+ error = corefile_open_last(td, name, indexpos, indexlen, ncores,
+ vpp);
+ if (error != 0) {
+ log(LOG_ERR,
+ "pid %d (%s), uid (%u): Path `%s' failed "
+ "on initial open test, error = %d\n",
+ pid, comm, uid, name, error);
+ }
+ } else {
+ cmode = S_IRUSR | S_IWUSR;
+ oflags = VN_OPEN_NOAUDIT | VN_OPEN_NAMECACHE |
+ (capmode_coredump ? VN_OPEN_NOCAPCHECK : 0);
+ flags = O_CREAT | FWRITE | O_NOFOLLOW;
+ if ((td->td_proc->p_flag & P_SUGID) != 0)
+ flags |= O_EXCL;
+
+ NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, name);
+ error = vn_open_cred(&nd, &flags, cmode, oflags, td->td_ucred,
+ NULL);
+ if (error == 0) {
+ *vpp = nd.ni_vp;
+ NDFREE_PNBUF(&nd);
+ }
+ }
+
+ if (error != 0) {
+#ifdef AUDIT
+ audit_proc_coredump(td, name, error);
+#endif
+ free(name, M_TEMP);
+ return (error);
+ }
+ *namep = name;
+ return (0);
+}
+
+/*
+ * The vnode dumper is the traditional coredump handler. Our policy and limits
+ * are generally checked already, so it creates the coredump name and passes on
+ * a vnode and a size limit to the process-specific coredump routine if there is
+ * one. If there _is not_ one, it returns ENOSYS; otherwise it returns the
+ * error from the process-specific routine.
+ */
+int
+coredump_vnode(struct thread *td, off_t limit)
+{
+ struct proc *p = td->td_proc;
+ struct ucred *cred = td->td_ucred;
+ struct vnode *vp;
+ struct coredump_vnode_ctx wctx;
+ struct coredump_writer cdw;
+ struct flock lf;
+ struct vattr vattr;
+ size_t fullpathsize;
+ int error, error1, jid, locked, ppid, sig;
+ char *name; /* name of corefile */
+ void *rl_cookie;
+ char *fullpath, *freepath = NULL;
+ struct sbuf *sb;
+
+ PROC_LOCK_ASSERT(p, MA_OWNED);
+
+ ppid = p->p_oppid;
+ sig = p->p_sig;
+ jid = p->p_ucred->cr_prison->pr_id;
+ PROC_UNLOCK(p);
+
+ error = corefile_open(p->p_comm, cred->cr_uid, p->p_pid, td,
+ compress_user_cores, sig, &vp, &name);
+ if (error != 0)
+ return (error);
+
+ /*
+ * Don't dump to non-regular files or files with links.
+ * Do not dump into system files. Effective user must own the corefile.
+ */
+ if (vp->v_type != VREG || VOP_GETATTR(vp, &vattr, cred) != 0 ||
+ vattr.va_nlink != 1 || (vp->v_vflag & VV_SYSTEM) != 0 ||
+ vattr.va_uid != cred->cr_uid) {
+ VOP_UNLOCK(vp);
+ error = EFAULT;
+ goto out;
+ }
+
+ VOP_UNLOCK(vp);
+
+ /* Postpone other writers, including core dumps of other processes. */
+ rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
+
+ lf.l_whence = SEEK_SET;
+ lf.l_start = 0;
+ lf.l_len = 0;
+ lf.l_type = F_WRLCK;
+ locked = (VOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &lf, F_FLOCK) == 0);
+
+ VATTR_NULL(&vattr);
+ vattr.va_size = 0;
+ if (set_core_nodump_flag)
+ vattr.va_flags = UF_NODUMP;
+ vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
+ VOP_SETATTR(vp, &vattr, cred);
+ VOP_UNLOCK(vp);
+ PROC_LOCK(p);
+ p->p_acflag |= ACORE;
+ PROC_UNLOCK(p);
+
+ wctx.vp = vp;
+ wctx.fcred = NOCRED;
+
+ cdw.ctx = &wctx;
+ cdw.write_fn = core_vn_write;
+ cdw.extend_fn = core_vn_extend;
+
+ if (p->p_sysent->sv_coredump != NULL) {
+ error = p->p_sysent->sv_coredump(td, &cdw, limit, 0);
+ } else {
+ error = ENOSYS;
+ }
+
+ if (locked) {
+ lf.l_type = F_UNLCK;
+ VOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_FLOCK);
+ }
+ vn_rangelock_unlock(vp, rl_cookie);
+
+ /*
+ * Notify the userland helper that a process triggered a core dump.
+ * This allows the helper to run an automated debugging session.
+ */
+ if (error != 0 || coredump_devctl == 0)
+ goto out;
+ sb = sbuf_new_auto();
+ if (vn_fullpath_global(p->p_textvp, &fullpath, &freepath) != 0)
+ goto out2;
+ sbuf_cat(sb, "comm=\"");
+ devctl_safe_quote_sb(sb, fullpath);
+ free(freepath, M_TEMP);
+ sbuf_cat(sb, "\" core=\"");
+
+ /*
+ * We can't lookup core file vp directly. When we're replacing a core, and
+ * other random times, we flush the name cache, so it will fail. Instead,
+ * if the path of the core is relative, add the current dir in front if it.
+ */
+ if (name[0] != '/') {
+ fullpathsize = MAXPATHLEN;
+ freepath = malloc(fullpathsize, M_TEMP, M_WAITOK);
+ if (vn_getcwd(freepath, &fullpath, &fullpathsize) != 0) {
+ free(freepath, M_TEMP);
+ goto out2;
+ }
+ devctl_safe_quote_sb(sb, fullpath);
+ free(freepath, M_TEMP);
+ sbuf_putc(sb, '/');
+ }
+ devctl_safe_quote_sb(sb, name);
+ sbuf_putc(sb, '"');
+
+ sbuf_printf(sb, " jid=%d pid=%d ppid=%d signo=%d",
+ jid, p->p_pid, ppid, sig);
+ if (sbuf_finish(sb) == 0)
+ devctl_notify("kernel", "signal", "coredump", sbuf_data(sb));
+out2:
+ sbuf_delete(sb);
+out:
+ error1 = vn_close(vp, FWRITE, cred, td);
+ if (error == 0)
+ error = error1;
+#ifdef AUDIT
+ audit_proc_coredump(td, name, error);
+#endif
+ free(name, M_TEMP);
+ return (error);
+}
diff --git a/sys/kern/imgact_elf.c b/sys/kern/imgact_elf.c
--- a/sys/kern/imgact_elf.c
+++ b/sys/kern/imgact_elf.c
@@ -36,6 +36,7 @@
#include <sys/param.h>
#include <sys/capsicum.h>
#include <sys/compressor.h>
+#include <sys/coredump.h>
#include <sys/exec.h>
#include <sys/fcntl.h>
#include <sys/imgact.h>
@@ -1562,9 +1563,6 @@
TAILQ_HEAD(note_info_list, note_info);
-extern int compress_user_cores;
-extern int compress_user_cores_level;
-
static void cb_put_phdr(vm_map_entry_t, void *);
static void cb_size_segment(vm_map_entry_t, void *);
static void each_dumpable_segment(struct thread *, segment_callback, void *,
diff --git a/sys/kern/kern_exec.c b/sys/kern/kern_exec.c
--- a/sys/kern/kern_exec.c
+++ b/sys/kern/kern_exec.c
@@ -39,6 +39,7 @@
#include <sys/asan.h>
#include <sys/capsicum.h>
#include <sys/compressor.h>
+#include <sys/coredump.h>
#include <sys/eventhandler.h>
#include <sys/exec.h>
#include <sys/fcntl.h>
@@ -1998,18 +1999,6 @@
return (error);
}
-int
-core_vn_write(const struct coredump_writer *cdw, const void *base, size_t len,
- off_t offset, enum uio_seg seg, struct ucred *cred, size_t *resid,
- struct thread *td)
-{
- struct coredump_vnode_ctx *ctx = cdw->ctx;
-
- return (vn_rdwr_inchunks(UIO_WRITE, ctx->vp, __DECONST(void *, base),
- len, offset, seg, IO_UNIT | IO_DIRECT | IO_RANGELOCKED,
- cred, ctx->fcred, resid, td));
-}
-
int
core_write(struct coredump_params *cp, const void *base, size_t len,
off_t offset, enum uio_seg seg, size_t *resid)
@@ -2018,24 +2007,6 @@
cp->active_cred, resid, cp->td));
}
-int
-core_vn_extend(const struct coredump_writer *cdw, off_t newsz,
- struct ucred *cred)
-{
- struct coredump_vnode_ctx *ctx = cdw->ctx;
- struct mount *mp;
- int error;
-
- error = vn_start_write(ctx->vp, &mp, V_WAIT);
- if (error != 0)
- return (error);
- vn_lock(ctx->vp, LK_EXCLUSIVE | LK_RETRY);
- error = vn_truncate_locked(ctx->vp, newsz, false, cred);
- VOP_UNLOCK(ctx->vp);
- vn_finished_write(mp);
- return (error);
-}
-
static int
core_extend(struct coredump_params *cp, off_t newsz)
{
diff --git a/sys/kern/kern_sig.c b/sys/kern/kern_sig.c
--- a/sys/kern/kern_sig.c
+++ b/sys/kern/kern_sig.c
@@ -39,6 +39,7 @@
#include <sys/param.h>
#include <sys/capsicum.h>
+#include <sys/coredump.h>
#include <sys/ctype.h>
#include <sys/systm.h>
#include <sys/signalvar.h>
@@ -198,22 +199,10 @@
SYSCTL_INT(_kern, OID_AUTO, sugid_coredump, CTLFLAG_RWTUN,
&sugid_coredump, 0, "Allow setuid and setgid processes to dump core");
-static int capmode_coredump;
-SYSCTL_INT(_kern, OID_AUTO, capmode_coredump, CTLFLAG_RWTUN,
- &capmode_coredump, 0, "Allow processes in capability mode to dump core");
-
static int do_coredump = 1;
SYSCTL_INT(_kern, OID_AUTO, coredump, CTLFLAG_RW,
&do_coredump, 0, "Enable/Disable coredumps");
-static int set_core_nodump_flag = 0;
-SYSCTL_INT(_kern, OID_AUTO, nodump_coredump, CTLFLAG_RW, &set_core_nodump_flag,
- 0, "Enable setting the NODUMP flag on coredump files");
-
-static int coredump_devctl = 0;
-SYSCTL_INT(_kern, OID_AUTO, coredump_devctl, CTLFLAG_RW, &coredump_devctl,
- 0, "Generate a devctl notification when processes coredump");
-
/*
* Signal properties and actions.
* The array below categorizes the signals and their default actions
@@ -3813,38 +3802,6 @@
sigparent(p, reason, status);
}
-#define MAX_NUM_CORE_FILES 100000
-#ifndef NUM_CORE_FILES
-#define NUM_CORE_FILES 5
-#endif
-CTASSERT(NUM_CORE_FILES >= 0 && NUM_CORE_FILES <= MAX_NUM_CORE_FILES);
-static int num_cores = NUM_CORE_FILES;
-
-static int
-sysctl_debug_num_cores_check (SYSCTL_HANDLER_ARGS)
-{
- int error;
- int new_val;
-
- new_val = num_cores;
- error = sysctl_handle_int(oidp, &new_val, 0, req);
- if (error != 0 || req->newptr == NULL)
- return (error);
- if (new_val > MAX_NUM_CORE_FILES)
- new_val = MAX_NUM_CORE_FILES;
- if (new_val < 0)
- new_val = 0;
- num_cores = new_val;
- return (0);
-}
-SYSCTL_PROC(_debug, OID_AUTO, ncores,
- CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, sizeof(int),
- sysctl_debug_num_cores_check, "I",
- "Maximum number of generated process corefiles while using index format");
-
-#define GZIP_SUFFIX ".gz"
-#define ZSTD_SUFFIX ".zst"
-
int compress_user_cores = 0;
static int
@@ -3873,264 +3830,6 @@
&compress_user_cores_level, 0,
"Corefile compression level");
-/*
- * Protect the access to corefilename[] by allproc_lock.
- */
-#define corefilename_lock allproc_lock
-
-static char corefilename[MAXPATHLEN] = {"%N.core"};
-TUNABLE_STR("kern.corefile", corefilename, sizeof(corefilename));
-
-static int
-sysctl_kern_corefile(SYSCTL_HANDLER_ARGS)
-{
- int error;
-
- sx_xlock(&corefilename_lock);
- error = sysctl_handle_string(oidp, corefilename, sizeof(corefilename),
- req);
- sx_xunlock(&corefilename_lock);
-
- return (error);
-}
-SYSCTL_PROC(_kern, OID_AUTO, corefile, CTLTYPE_STRING | CTLFLAG_RW |
- CTLFLAG_MPSAFE, 0, 0, sysctl_kern_corefile, "A",
- "Process corefile name format string");
-
-static void
-vnode_close_locked(struct thread *td, struct vnode *vp)
-{
-
- VOP_UNLOCK(vp);
- vn_close(vp, FWRITE, td->td_ucred, td);
-}
-
-/*
- * If the core format has a %I in it, then we need to check
- * for existing corefiles before defining a name.
- * To do this we iterate over 0..ncores to find a
- * non-existing core file name to use. If all core files are
- * already used we choose the oldest one.
- */
-static int
-corefile_open_last(struct thread *td, char *name, int indexpos,
- int indexlen, int ncores, struct vnode **vpp)
-{
- struct vnode *oldvp, *nextvp, *vp;
- struct vattr vattr;
- struct nameidata nd;
- int error, i, flags, oflags, cmode;
- char ch;
- struct timespec lasttime;
-
- nextvp = oldvp = NULL;
- cmode = S_IRUSR | S_IWUSR;
- oflags = VN_OPEN_NOAUDIT | VN_OPEN_NAMECACHE |
- (capmode_coredump ? VN_OPEN_NOCAPCHECK : 0);
-
- for (i = 0; i < ncores; i++) {
- flags = O_CREAT | FWRITE | O_NOFOLLOW;
-
- ch = name[indexpos + indexlen];
- (void)snprintf(name + indexpos, indexlen + 1, "%.*u", indexlen,
- i);
- name[indexpos + indexlen] = ch;
-
- NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, name);
- error = vn_open_cred(&nd, &flags, cmode, oflags, td->td_ucred,
- NULL);
- if (error != 0)
- break;
-
- vp = nd.ni_vp;
- NDFREE_PNBUF(&nd);
- if ((flags & O_CREAT) == O_CREAT) {
- nextvp = vp;
- break;
- }
-
- error = VOP_GETATTR(vp, &vattr, td->td_ucred);
- if (error != 0) {
- vnode_close_locked(td, vp);
- break;
- }
-
- if (oldvp == NULL ||
- lasttime.tv_sec > vattr.va_mtime.tv_sec ||
- (lasttime.tv_sec == vattr.va_mtime.tv_sec &&
- lasttime.tv_nsec >= vattr.va_mtime.tv_nsec)) {
- if (oldvp != NULL)
- vn_close(oldvp, FWRITE, td->td_ucred, td);
- oldvp = vp;
- VOP_UNLOCK(oldvp);
- lasttime = vattr.va_mtime;
- } else {
- vnode_close_locked(td, vp);
- }
- }
-
- if (oldvp != NULL) {
- if (nextvp == NULL) {
- if ((td->td_proc->p_flag & P_SUGID) != 0) {
- error = EFAULT;
- vn_close(oldvp, FWRITE, td->td_ucred, td);
- } else {
- nextvp = oldvp;
- error = vn_lock(nextvp, LK_EXCLUSIVE);
- if (error != 0) {
- vn_close(nextvp, FWRITE, td->td_ucred,
- td);
- nextvp = NULL;
- }
- }
- } else {
- vn_close(oldvp, FWRITE, td->td_ucred, td);
- }
- }
- if (error != 0) {
- if (nextvp != NULL)
- vnode_close_locked(td, oldvp);
- } else {
- *vpp = nextvp;
- }
-
- return (error);
-}
-
-/*
- * corefile_open(comm, uid, pid, td, compress, vpp, namep)
- * Expand the name described in corefilename, using name, uid, and pid
- * and open/create core file.
- * corefilename is a printf-like string, with three format specifiers:
- * %N name of process ("name")
- * %P process id (pid)
- * %U user id (uid)
- * For example, "%N.core" is the default; they can be disabled completely
- * by using "/dev/null", or all core files can be stored in "/cores/%U/%N-%P".
- * This is controlled by the sysctl variable kern.corefile (see above).
- */
-static int
-corefile_open(const char *comm, uid_t uid, pid_t pid, struct thread *td,
- int compress, int signum, struct vnode **vpp, char **namep)
-{
- struct sbuf sb;
- struct nameidata nd;
- const char *format;
- char *hostname, *name;
- int cmode, error, flags, i, indexpos, indexlen, oflags, ncores;
-
- hostname = NULL;
- format = corefilename;
- name = malloc(MAXPATHLEN, M_TEMP, M_WAITOK | M_ZERO);
- indexlen = 0;
- indexpos = -1;
- ncores = num_cores;
- (void)sbuf_new(&sb, name, MAXPATHLEN, SBUF_FIXEDLEN);
- sx_slock(&corefilename_lock);
- for (i = 0; format[i] != '\0'; i++) {
- switch (format[i]) {
- case '%': /* Format character */
- i++;
- switch (format[i]) {
- case '%':
- sbuf_putc(&sb, '%');
- break;
- case 'H': /* hostname */
- if (hostname == NULL) {
- hostname = malloc(MAXHOSTNAMELEN,
- M_TEMP, M_WAITOK);
- }
- getcredhostname(td->td_ucred, hostname,
- MAXHOSTNAMELEN);
- sbuf_cat(&sb, hostname);
- break;
- case 'I': /* autoincrementing index */
- if (indexpos != -1) {
- sbuf_printf(&sb, "%%I");
- break;
- }
-
- indexpos = sbuf_len(&sb);
- sbuf_printf(&sb, "%u", ncores - 1);
- indexlen = sbuf_len(&sb) - indexpos;
- break;
- case 'N': /* process name */
- sbuf_printf(&sb, "%s", comm);
- break;
- case 'P': /* process id */
- sbuf_printf(&sb, "%u", pid);
- break;
- case 'S': /* signal number */
- sbuf_printf(&sb, "%i", signum);
- break;
- case 'U': /* user id */
- sbuf_printf(&sb, "%u", uid);
- break;
- default:
- log(LOG_ERR,
- "Unknown format character %c in "
- "corename `%s'\n", format[i], format);
- break;
- }
- break;
- default:
- sbuf_putc(&sb, format[i]);
- break;
- }
- }
- sx_sunlock(&corefilename_lock);
- free(hostname, M_TEMP);
- if (compress == COMPRESS_GZIP)
- sbuf_cat(&sb, GZIP_SUFFIX);
- else if (compress == COMPRESS_ZSTD)
- sbuf_cat(&sb, ZSTD_SUFFIX);
- if (sbuf_error(&sb) != 0) {
- log(LOG_ERR, "pid %ld (%s), uid (%lu): corename is too "
- "long\n", (long)pid, comm, (u_long)uid);
- sbuf_delete(&sb);
- free(name, M_TEMP);
- return (ENOMEM);
- }
- sbuf_finish(&sb);
- sbuf_delete(&sb);
-
- if (indexpos != -1) {
- error = corefile_open_last(td, name, indexpos, indexlen, ncores,
- vpp);
- if (error != 0) {
- log(LOG_ERR,
- "pid %d (%s), uid (%u): Path `%s' failed "
- "on initial open test, error = %d\n",
- pid, comm, uid, name, error);
- }
- } else {
- cmode = S_IRUSR | S_IWUSR;
- oflags = VN_OPEN_NOAUDIT | VN_OPEN_NAMECACHE |
- (capmode_coredump ? VN_OPEN_NOCAPCHECK : 0);
- flags = O_CREAT | FWRITE | O_NOFOLLOW;
- if ((td->td_proc->p_flag & P_SUGID) != 0)
- flags |= O_EXCL;
-
- NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, name);
- error = vn_open_cred(&nd, &flags, cmode, oflags, td->td_ucred,
- NULL);
- if (error == 0) {
- *vpp = nd.ni_vp;
- NDFREE_PNBUF(&nd);
- }
- }
-
- if (error != 0) {
-#ifdef AUDIT
- audit_proc_coredump(td, name, error);
-#endif
- free(name, M_TEMP);
- return (error);
- }
- *namep = name;
- return (0);
-}
-
/*
* Dump a process' core. The main routine does some
* policy checking, and creates the name of the coredump;
@@ -4143,19 +3842,8 @@
coredump(struct thread *td)
{
struct proc *p = td->td_proc;
- struct ucred *cred = td->td_ucred;
- struct vnode *vp;
- struct coredump_vnode_ctx wctx;
- struct coredump_writer cdw;
- struct flock lf;
- struct vattr vattr;
- size_t fullpathsize;
- int error, error1, jid, locked, ppid, sig;
- char *name; /* name of corefile */
- void *rl_cookie;
off_t limit;
- char *fullpath, *freepath = NULL;
- struct sbuf *sb;
+ int error;
PROC_LOCK_ASSERT(p, MA_OWNED);
MPASS((p->p_flag & P_HADTHREADS) == 0 || p->p_singlethread == td);
@@ -4180,116 +3868,9 @@
return (EFBIG);
}
- ppid = p->p_oppid;
- sig = p->p_sig;
- jid = p->p_ucred->cr_prison->pr_id;
- PROC_UNLOCK(p);
-
- error = corefile_open(p->p_comm, cred->cr_uid, p->p_pid, td,
- compress_user_cores, p->p_sig, &vp, &name);
- if (error != 0)
- return (error);
-
- /*
- * Don't dump to non-regular files or files with links.
- * Do not dump into system files. Effective user must own the corefile.
- */
- if (vp->v_type != VREG || VOP_GETATTR(vp, &vattr, cred) != 0 ||
- vattr.va_nlink != 1 || (vp->v_vflag & VV_SYSTEM) != 0 ||
- vattr.va_uid != cred->cr_uid) {
- VOP_UNLOCK(vp);
- error = EFAULT;
- goto out;
- }
-
- VOP_UNLOCK(vp);
-
- /* Postpone other writers, including core dumps of other processes. */
- rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
-
- lf.l_whence = SEEK_SET;
- lf.l_start = 0;
- lf.l_len = 0;
- lf.l_type = F_WRLCK;
- locked = (VOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &lf, F_FLOCK) == 0);
-
- VATTR_NULL(&vattr);
- vattr.va_size = 0;
- if (set_core_nodump_flag)
- vattr.va_flags = UF_NODUMP;
- vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
- VOP_SETATTR(vp, &vattr, cred);
- VOP_UNLOCK(vp);
- PROC_LOCK(p);
- p->p_acflag |= ACORE;
- PROC_UNLOCK(p);
-
- wctx.vp = vp;
- wctx.fcred = NOCRED;
-
- cdw.ctx = &wctx;
- cdw.write_fn = core_vn_write;
- cdw.extend_fn = core_vn_extend;
-
- if (p->p_sysent->sv_coredump != NULL) {
- error = p->p_sysent->sv_coredump(td, &cdw, limit, 0);
- } else {
- error = ENOSYS;
- }
+ error = coredump_vnode(td, limit);
+ PROC_LOCK_ASSERT(p, MA_NOTOWNED);
- if (locked) {
- lf.l_type = F_UNLCK;
- VOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_FLOCK);
- }
- vn_rangelock_unlock(vp, rl_cookie);
-
- /*
- * Notify the userland helper that a process triggered a core dump.
- * This allows the helper to run an automated debugging session.
- */
- if (error != 0 || coredump_devctl == 0)
- goto out;
- sb = sbuf_new_auto();
- if (vn_fullpath_global(p->p_textvp, &fullpath, &freepath) != 0)
- goto out2;
- sbuf_cat(sb, "comm=\"");
- devctl_safe_quote_sb(sb, fullpath);
- free(freepath, M_TEMP);
- sbuf_cat(sb, "\" core=\"");
-
- /*
- * We can't lookup core file vp directly. When we're replacing a core, and
- * other random times, we flush the name cache, so it will fail. Instead,
- * if the path of the core is relative, add the current dir in front if it.
- */
- if (name[0] != '/') {
- fullpathsize = MAXPATHLEN;
- freepath = malloc(fullpathsize, M_TEMP, M_WAITOK);
- if (vn_getcwd(freepath, &fullpath, &fullpathsize) != 0) {
- free(freepath, M_TEMP);
- goto out2;
- }
- devctl_safe_quote_sb(sb, fullpath);
- free(freepath, M_TEMP);
- sbuf_putc(sb, '/');
- }
- devctl_safe_quote_sb(sb, name);
- sbuf_putc(sb, '"');
-
- sbuf_printf(sb, " jid=%d pid=%d ppid=%d signo=%d",
- jid, p->p_pid, ppid, sig);
- if (sbuf_finish(sb) == 0)
- devctl_notify("kernel", "signal", "coredump", sbuf_data(sb));
-out2:
- sbuf_delete(sb);
-out:
- error1 = vn_close(vp, FWRITE, cred, td);
- if (error == 0)
- error = error1;
-#ifdef AUDIT
- audit_proc_coredump(td, name, error);
-#endif
- free(name, M_TEMP);
return (error);
}
diff --git a/sys/sys/exec.h b/sys/sys/coredump.h
copy from sys/sys/exec.h
copy to sys/sys/coredump.h
--- a/sys/sys/exec.h
+++ b/sys/sys/coredump.h
@@ -34,31 +34,11 @@
* SUCH DAMAGE.
*/
-#ifndef _SYS_EXEC_H_
-#define _SYS_EXEC_H_
+#ifndef _SYS_COREDUMP_H_
+#define _SYS_COREDUMP_H_
#include <sys/_uio.h>
-/*
- * Before ps_args existed, the following structure, found at the top of
- * the user stack of each user process, was used by ps(1) to locate
- * environment and argv strings. Normally ps_argvstr points to the
- * argv vector, and ps_nargvstr is the same as the program's argc. The
- * fields ps_envstr and ps_nenvstr are the equivalent for the environment.
- *
- * Programs should now use setproctitle(3) to change ps output.
- * setproctitle() always informs the kernel with sysctl and sets the
- * pointers in ps_strings. The kern.proc.args sysctl first tries p_args.
- * If p_args is NULL, it then falls back to reading ps_strings and following
- * the pointers.
- */
-struct ps_strings {
- char **ps_argvstr; /* first of 0 or more argument strings */
- unsigned int ps_nargvstr; /* the number of argument strings */
- char **ps_envstr; /* first of 0 or more environment strings */
- unsigned int ps_nenvstr; /* the number of environment strings */
-};
-
/* Coredump output parameters. */
struct coredump_writer;
struct thread;
@@ -75,6 +55,7 @@
coredump_write_fn core_vn_write;
coredump_extend_fn core_vn_extend;
+int coredump_vnode(struct thread *, off_t);
struct coredump_writer {
void *ctx;
@@ -90,44 +71,6 @@
struct compressor *comp;
};
-struct image_params;
-
-struct execsw {
- int (*ex_imgact)(struct image_params *);
- const char *ex_name;
-};
-
-#include <machine/exec.h>
-
-#ifdef _KERNEL
-#include <sys/cdefs.h>
-
-/*
- * Address of ps_strings structure (in user space).
- * Prefer the kern.ps_strings or kern.proc.ps_strings sysctls to this constant.
- */
-#define PS_STRINGS (USRSTACK - sizeof(struct ps_strings))
-#define PROC_PS_STRINGS(p) \
- ((p)->p_vmspace->vm_stacktop - (p)->p_sysent->sv_psstringssz)
-
-/*
- * Address of signal trampoline (in user space).
- * This assumes that the sigcode resides in the shared page.
- */
-#define PROC_SIGCODE(p) \
- ((p)->p_vmspace->vm_shp_base + (p)->p_sysent->sv_sigcode_offset)
-
-#define PROC_HAS_SHP(p) \
- ((p)->p_sysent->sv_shared_page_obj != NULL)
-
-int exec_map_first_page(struct image_params *);
-void exec_unmap_first_page(struct image_params *);
-
-int exec_register(const struct execsw *);
-int exec_unregister(const struct execsw *);
-
-enum uio_seg;
-
#define CORE_BUF_SIZE (16 * 1024)
int core_write(struct coredump_params *, const void *, size_t, off_t,
@@ -138,46 +81,7 @@
extern int coredump_pack_fileinfo;
extern int coredump_pack_vmmapinfo;
-/*
- * note: name##_mod cannot be const storage because the
- * linker_file_sysinit() function modifies _file in the
- * moduledata_t.
- */
-
-#include <sys/module.h>
-
-#define EXEC_SET(name, execsw_arg) \
- static int __CONCAT(name,_modevent)(module_t mod, int type, \
- void *data) \
- { \
- struct execsw *exec = (struct execsw *)data; \
- int error = 0; \
- switch (type) { \
- case MOD_LOAD: \
- /* printf(#name " module loaded\n"); */ \
- error = exec_register(exec); \
- if (error) \
- printf(__XSTRING(name) "register failed\n"); \
- break; \
- case MOD_UNLOAD: \
- /* printf(#name " module unloaded\n"); */ \
- error = exec_unregister(exec); \
- if (error) \
- printf(__XSTRING(name) " unregister failed\n");\
- break; \
- default: \
- error = EOPNOTSUPP; \
- break; \
- } \
- return error; \
- } \
- static moduledata_t __CONCAT(name,_mod) = { \
- __XSTRING(name), \
- __CONCAT(name,_modevent), \
- (void *)& execsw_arg \
- }; \
- DECLARE_MODULE_TIED(name, __CONCAT(name,_mod), SI_SUB_EXEC, \
- SI_ORDER_ANY)
-#endif
+extern int compress_user_cores;
+extern int compress_user_cores_level;
-#endif
+#endif /* _SYS_COREDUMP_H_ */
diff --git a/sys/sys/exec.h b/sys/sys/exec.h
--- a/sys/sys/exec.h
+++ b/sys/sys/exec.h
@@ -37,8 +37,6 @@
#ifndef _SYS_EXEC_H_
#define _SYS_EXEC_H_
-#include <sys/_uio.h>
-
/*
* Before ps_args existed, the following structure, found at the top of
* the user stack of each user process, was used by ps(1) to locate
@@ -59,37 +57,6 @@
unsigned int ps_nenvstr; /* the number of environment strings */
};
-/* Coredump output parameters. */
-struct coredump_writer;
-struct thread;
-struct ucred;
-typedef int coredump_write_fn(const struct coredump_writer *, const void *, size_t,
- off_t, enum uio_seg, struct ucred *, size_t *, struct thread *);
-typedef int coredump_extend_fn(const struct coredump_writer *, off_t,
- struct ucred *);
-
-struct coredump_vnode_ctx {
- struct vnode *vp;
- struct ucred *fcred;
-};
-
-coredump_write_fn core_vn_write;
-coredump_extend_fn core_vn_extend;
-
-struct coredump_writer {
- void *ctx;
- coredump_write_fn *write_fn;
- coredump_extend_fn *extend_fn;
-};
-
-struct coredump_params {
- off_t offset;
- struct ucred *active_cred;
- struct thread *td;
- const struct coredump_writer *cdw;
- struct compressor *comp;
-};
-
struct image_params;
struct execsw {
@@ -128,16 +95,6 @@
enum uio_seg;
-#define CORE_BUF_SIZE (16 * 1024)
-
-int core_write(struct coredump_params *, const void *, size_t, off_t,
- enum uio_seg, size_t *);
-int core_output(char *, size_t, off_t, struct coredump_params *, void *);
-int sbuf_drain_core_output(void *, const char *, int);
-
-extern int coredump_pack_fileinfo;
-extern int coredump_pack_vmmapinfo;
-
/*
* note: name##_mod cannot be const storage because the
* linker_file_sysinit() function modifies _file in the

File Metadata

Mime Type
text/plain
Expires
Wed, Jul 22, 10:20 AM (13 h, 48 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
35360899
Default Alt Text
D51349.id158615.diff (34 KB)

Event Timeline