dirkmueller/systemd
SHA256
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
b68246ecb3
systemd/handle-racy-daemon.patch

1067 lines
36 KiB
Diff
Raw Normal View History

- Remove storage-after-cryptsetup.service, add storage-after-cryptsetup.patch instead to prevent dependency cycle (bnc#722539). - Add delay-fsck-cryptsetup-after-md-lvm-dmraid.patch: ensure fsck/cryptsetup is run after lvm/md/dmraid have landed (bnc#724912). - Add cron-tty-pam.patch: Fix cron filling logs (bnc#731358). - Add do_not_warn_pidfile.patch: Fix PID warning in logs (bnc#732912). - Add mount-swap-log.patch: Ensure swap and mount output is redirected to default log target (rhb#750032). - Add color-on-boot.patch: ensure colored status are displayed at boot time. - Update modules_on_boot.patch to fix bnc#732041. - Replace private_tmp_crash.patch with log_on_close.patch, better upstream fix for bnc#699829 and fix bnc#731719. - Update vconsole patch to fix memleaks and crash (bnc#734527). - Add handle-racy-daemon.patch: fix warnings with sendmail (bnc#732912). - Add new-lsb-headers.patch: support PIDFile: and X-Systemd-RemainAfterExit: header in initscript (bnc#727771). - Update bootsplash services to not start if vga= is missing from cmdline (bnc#727771) - Add lock-opensuse.patch: disable /var/lock/{subsys,lockdev} and change default permissions on /var/lock (bnc#733523). - Add garbage_collect_units: ensure error units are correctly garbage collected (rhb#680122). - Add crypt-loop-file.patch: add support for crypt file loop (bnc#730496). OBS-URL: https://build.opensuse.org/package/show/Base:System/systemd?expand=0&rev=228
2011-12-09 15:44:59 +01:00
From 35d50f55f346c71fd5e957d35ebcae1c50b1f9ce Mon Sep 17 00:00:00 2001
From: Michal Schmidt <mschmidt@redhat.com>
Date: Sat, 3 Dec 2011 00:41:34 +0100
Subject: [PATCH 1/5] util: fix error checking after fgets()
fgets() does not set errno on EOF.
---
src/util.c | 8 ++++----
1 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/src/util.c b/src/util.c
index e93e6f6..da71e4d 100644
--- a/src/util.c
+++ b/src/util.c
@@ -516,7 +516,7 @@ int get_parent_of_pid(pid_t pid, pid_t *_ppid) {
return -errno;
if (!(fgets(line, sizeof(line), f))) {
- r = -errno;
+ r = feof(f) ? -EIO : -errno;
fclose(f);
return r;
}
@@ -561,7 +561,7 @@ int get_starttime_of_pid(pid_t pid, unsigned long long *st) {
return -errno;
if (!(fgets(line, sizeof(line), f))) {
- r = -errno;
+ r = feof(f) ? -EIO : -errno;
fclose(f);
return r;
}
@@ -708,7 +708,7 @@ int read_one_line_file(const char *fn, char **line) {
return -errno;
if (!(fgets(t, sizeof(t), f))) {
- r = -errno;
+ r = feof(f) ? -EIO : -errno;
goto finish;
}
@@ -3266,7 +3266,7 @@ int get_ctty_devnr(pid_t pid, dev_t *d) {
return -errno;
if (!fgets(line, sizeof(line), f)) {
- k = -errno;
+ k = feof(f) ? -EIO : -errno;
fclose(f);
return k;
}
--
1.7.7
From 768147d13d0877a4c3e5f6f986c3064de62ff4f1 Mon Sep 17 00:00:00 2001
From: Michal Schmidt <mschmidt@redhat.com>
Date: Sat, 3 Dec 2011 01:36:05 +0100
Subject: [PATCH 2/5] path: use %m instead of strerror(errno)
and strerror(-errno) was just wrong.
---
src/path.c | 4 ++--
1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/path.c b/src/path.c
index f15c921..142fd2d 100644
--- a/src/path.c
+++ b/src/path.c
@@ -556,7 +556,7 @@ static void path_fd_event(Unit *u, int fd, uint32_t events, Watch *w) {
}
if (ioctl(fd, FIONREAD, &l) < 0) {
- log_error("FIONREAD failed: %s", strerror(errno));
+ log_error("FIONREAD failed: %m");
goto fail;
}
@@ -568,7 +568,7 @@ static void path_fd_event(Unit *u, int fd, uint32_t events, Watch *w) {
}
if ((k = read(fd, buf, l)) < 0) {
- log_error("Failed to read inotify event: %s", strerror(-errno));
+ log_error("Failed to read inotify event: %m");
goto fail;
}
--
1.7.7
From 4b562198c79e4ebfc3d84b69a1dae374bc6cf9f5 Mon Sep 17 00:00:00 2001
From: Michal Schmidt <mschmidt@redhat.com>
Date: Sat, 3 Dec 2011 01:38:30 +0100
Subject: [PATCH 3/5] path: refactor PathSpec usage
path_*() functions operate on "Path *p" and they do not touch PathSpec
internals directly.
pathspec_*() functions operate on "PathSpec *s". The PathSpec class will
be useful outside of path.c.
---
src/path.c | 356 +++++++++++++++++++++++++++++++++---------------------------
src/path.h | 8 ++
2 files changed, 203 insertions(+), 161 deletions(-)
diff --git a/src/path.c b/src/path.c
index 142fd2d..db6f873 100644
--- a/src/path.c
+++ b/src/path.c
@@ -39,26 +39,202 @@ static const UnitActiveState state_translation_table[_PATH_STATE_MAX] = {
[PATH_FAILED] = UNIT_FAILED
};
-static void path_init(Unit *u) {
- Path *p = PATH(u);
+int pathspec_watch(PathSpec *s, Unit *u) {
+ static const int flags_table[_PATH_TYPE_MAX] = {
+ [PATH_EXISTS] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB,
+ [PATH_EXISTS_GLOB] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB,
+ [PATH_CHANGED] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CLOSE_WRITE|IN_CREATE|IN_DELETE|IN_MOVED_FROM|IN_MOVED_TO,
+ [PATH_DIRECTORY_NOT_EMPTY] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CREATE|IN_MOVED_TO
+ };
+
+ bool exists = false;
+ char *k, *slash;
+ int r;
assert(u);
- assert(u->meta.load_state == UNIT_STUB);
+ assert(s);
- p->directory_mode = 0755;
+ pathspec_unwatch(s, u);
+
+ if (!(k = strdup(s->path)))
+ return -ENOMEM;
+
+ if ((s->inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC)) < 0) {
+ r = -errno;
+ goto fail;
+ }
+
+ if (unit_watch_fd(u, s->inotify_fd, EPOLLIN, &s->watch) < 0) {
+ r = -errno;
+ goto fail;
+ }
+
+ if ((s->primary_wd = inotify_add_watch(s->inotify_fd, k, flags_table[s->type])) >= 0)
+ exists = true;
+
+ do {
+ int flags;
+
+ /* This assumes the path was passed through path_kill_slashes()! */
+ if (!(slash = strrchr(k, '/')))
+ break;
+
+ /* Trim the path at the last slash. Keep the slash if it's the root dir. */
+ slash[slash == k] = 0;
+
+ flags = IN_MOVE_SELF;
+ if (!exists)
+ flags |= IN_DELETE_SELF | IN_ATTRIB | IN_CREATE | IN_MOVED_TO;
+
+ if (inotify_add_watch(s->inotify_fd, k, flags) >= 0)
+ exists = true;
+ } while (slash != k);
+
+ return 0;
+
+fail:
+ free(k);
+
+ pathspec_unwatch(s, u);
+ return r;
}
-static void path_unwatch_one(Path *p, PathSpec *s) {
+void pathspec_unwatch(PathSpec *s, Unit *u) {
if (s->inotify_fd < 0)
return;
- unit_unwatch_fd(UNIT(p), &s->watch);
+ unit_unwatch_fd(u, &s->watch);
close_nointr_nofail(s->inotify_fd);
s->inotify_fd = -1;
}
+int pathspec_fd_event(PathSpec *s, uint32_t events) {
+ uint8_t *buf = NULL;
+ struct inotify_event *e;
+ ssize_t k;
+ int l;
+ int r = 0;
+
+ if (events != EPOLLIN) {
+ log_error("Got Invalid poll event on inotify.");
+ r = -EINVAL;
+ goto out;
+ }
+
+ if (ioctl(s->inotify_fd, FIONREAD, &l) < 0) {
+ log_error("FIONREAD failed: %m");
+ r = -errno;
+ goto out;
+ }
+
+ assert(l > 0);
+
+ if (!(buf = malloc(l))) {
+ log_error("Failed to allocate buffer: %m");
+ r = -errno;
+ goto out;
+ }
+
+ if ((k = read(s->inotify_fd, buf, l)) < 0) {
+ log_error("Failed to read inotify event: %m");
+ r = -errno;
+ goto out;
+ }
+
+ e = (struct inotify_event*) buf;
+
+ while (k > 0) {
+ size_t step;
+
+ if (s->type == PATH_CHANGED && s->primary_wd == e->wd)
+ r = 1;
+
+ step = sizeof(struct inotify_event) + e->len;
+ assert(step <= (size_t) k);
+
+ e = (struct inotify_event*) ((uint8_t*) e + step);
+ k -= step;
+ }
+out:
+ free(buf);
+ return r;
+}
+
+static bool pathspec_check_good(PathSpec *s, bool initial) {
+ bool good = false;
+
+ switch (s->type) {
+
+ case PATH_EXISTS:
+ good = access(s->path, F_OK) >= 0;
+ break;
+
+ case PATH_EXISTS_GLOB:
+ good = glob_exists(s->path) > 0;
+ break;
+
+ case PATH_DIRECTORY_NOT_EMPTY: {
+ int k;
+
+ k = dir_is_empty(s->path);
+ good = !(k == -ENOENT || k > 0);
+ break;
+ }
+
+ case PATH_CHANGED: {
+ bool b;
+
+ b = access(s->path, F_OK) >= 0;
+ good = !initial && b != s->previous_exists;
+ s->previous_exists = b;
+ break;
+ }
+
+ default:
+ ;
+ }
+
+ return good;
+}
+
+static bool pathspec_startswith(PathSpec *s, const char *what) {
+ return path_startswith(s->path, what);
+}
+
+static void pathspec_mkdir(PathSpec *s, mode_t mode) {
+ int r;
+
+ if (s->type == PATH_EXISTS || s->type == PATH_EXISTS_GLOB)
+ return;
+
+ if ((r = mkdir_p(s->path, mode)) < 0)
+ log_warning("mkdir(%s) failed: %s", s->path, strerror(-r));
+}
+
+static void pathspec_dump(PathSpec *s, FILE *f, const char *prefix) {
+ fprintf(f,
+ "%s%s: %s\n",
+ prefix,
+ path_type_to_string(s->type),
+ s->path);
+}
+
+void pathspec_done(PathSpec *s) {
+ assert(s->inotify_fd == -1);
+ free(s->path);
+}
+
+static void path_init(Unit *u) {
+ Path *p = PATH(u);
+
+ assert(u);
+ assert(u->meta.load_state == UNIT_STUB);
+
+ p->directory_mode = 0755;
+}
+
static void path_done(Unit *u) {
Path *p = PATH(u);
PathSpec *s;
@@ -66,9 +242,9 @@ static void path_done(Unit *u) {
assert(p);
while ((s = p->specs)) {
- path_unwatch_one(p, s);
+ pathspec_unwatch(s, u);
LIST_REMOVE(PathSpec, spec, p->specs, s);
- free(s->path);
+ pathspec_done(s);
free(s);
}
}
@@ -86,7 +262,7 @@ int path_add_one_mount_link(Path *p, Mount *m) {
LIST_FOREACH(spec, s, p->specs) {
- if (!path_startswith(s->path, m->where))
+ if (!pathspec_startswith(s, m->where))
continue;
if ((r = unit_add_two_dependencies(UNIT(p), UNIT_AFTER, UNIT_REQUIRES, UNIT(m), true)) < 0)
@@ -187,71 +363,7 @@ static void path_dump(Unit *u, FILE *f, const char *prefix) {
prefix, p->directory_mode);
LIST_FOREACH(spec, s, p->specs)
- fprintf(f,
- "%s%s: %s\n",
- prefix,
- path_type_to_string(s->type),
- s->path);
-}
-
-static int path_watch_one(Path *p, PathSpec *s) {
- static const int flags_table[_PATH_TYPE_MAX] = {
- [PATH_EXISTS] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB,
- [PATH_EXISTS_GLOB] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB,
- [PATH_CHANGED] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CLOSE_WRITE|IN_CREATE|IN_DELETE|IN_MOVED_FROM|IN_MOVED_TO,
- [PATH_DIRECTORY_NOT_EMPTY] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CREATE|IN_MOVED_TO
- };
-
- bool exists = false;
- char *k, *slash;
- int r;
-
- assert(p);
- assert(s);
-
- path_unwatch_one(p, s);
-
- if (!(k = strdup(s->path)))
- return -ENOMEM;
-
- if ((s->inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC)) < 0) {
- r = -errno;
- goto fail;
- }
-
- if (unit_watch_fd(UNIT(p), s->inotify_fd, EPOLLIN, &s->watch) < 0) {
- r = -errno;
- goto fail;
- }
-
- if ((s->primary_wd = inotify_add_watch(s->inotify_fd, k, flags_table[s->type])) >= 0)
- exists = true;
-
- do {
- int flags;
-
- /* This assumes the path was passed through path_kill_slashes()! */
- if (!(slash = strrchr(k, '/')))
- break;
-
- /* Trim the path at the last slash. Keep the slash if it's the root dir. */
- slash[slash == k] = 0;
-
- flags = IN_MOVE_SELF;
- if (!exists)
- flags |= IN_DELETE_SELF | IN_ATTRIB | IN_CREATE | IN_MOVED_TO;
-
- if (inotify_add_watch(s->inotify_fd, k, flags) >= 0)
- exists = true;
- } while (slash != k);
-
- return 0;
-
-fail:
- free(k);
-
- path_unwatch_one(p, s);
- return r;
+ pathspec_dump(s, f, prefix);
}
static void path_unwatch(Path *p) {
@@ -260,7 +372,7 @@ static void path_unwatch(Path *p) {
assert(p);
LIST_FOREACH(spec, s, p->specs)
- path_unwatch_one(p, s);
+ pathspec_unwatch(s, UNIT(p));
}
static int path_watch(Path *p) {
@@ -270,7 +382,7 @@ static int path_watch(Path *p) {
assert(p);
LIST_FOREACH(spec, s, p->specs)
- if ((r = path_watch_one(p, s)) < 0)
+ if ((r = pathspec_watch(s, UNIT(p))) < 0)
return r;
return 0;
@@ -361,37 +473,7 @@ static bool path_check_good(Path *p, bool initial) {
assert(p);
LIST_FOREACH(spec, s, p->specs) {
-
- switch (s->type) {
-
- case PATH_EXISTS:
- good = access(s->path, F_OK) >= 0;
- break;
-
- case PATH_EXISTS_GLOB:
- good = glob_exists(s->path) > 0;
- break;
-
- case PATH_DIRECTORY_NOT_EMPTY: {
- int k;
-
- k = dir_is_empty(s->path);
- good = !(k == -ENOENT || k > 0);
- break;
- }
-
- case PATH_CHANGED: {
- bool b;
-
- b = access(s->path, F_OK) >= 0;
- good = !initial && b != s->previous_exists;
- s->previous_exists = b;
- break;
- }
-
- default:
- ;
- }
+ good = pathspec_check_good(s, initial);
if (good)
break;
@@ -440,15 +522,8 @@ static void path_mkdir(Path *p) {
if (!p->make_directory)
return;
- LIST_FOREACH(spec, s, p->specs) {
- int r;
-
- if (s->type == PATH_EXISTS || s->type == PATH_EXISTS_GLOB)
- continue;
-
- if ((r = mkdir_p(s->path, p->directory_mode)) < 0)
- log_warning("mkdir(%s) failed: %s", s->path, strerror(-r));
- }
+ LIST_FOREACH(spec, s, p->specs)
+ pathspec_mkdir(s, p->directory_mode);
}
static int path_start(Unit *u) {
@@ -525,12 +600,8 @@ static const char *path_sub_state_to_string(Unit *u) {
static void path_fd_event(Unit *u, int fd, uint32_t events, Watch *w) {
Path *p = PATH(u);
- int l;
- ssize_t k;
- uint8_t *buf = NULL;
- struct inotify_event *e;
PathSpec *s;
- bool changed;
+ int changed;
assert(p);
assert(fd >= 0);
@@ -541,13 +612,8 @@ static void path_fd_event(Unit *u, int fd, uint32_t events, Watch *w) {
/* log_debug("inotify wakeup on %s.", u->meta.id); */
- if (events != EPOLLIN) {
- log_error("Got Invalid poll event on inotify.");
- goto fail;
- }
-
LIST_FOREACH(spec, s, p->specs)
- if (s->inotify_fd == fd)
+ if (pathspec_owns_inotify_fd(s, fd))
break;
if (!s) {
@@ -555,55 +621,23 @@ static void path_fd_event(Unit *u, int fd, uint32_t events, Watch *w) {
goto fail;
}
- if (ioctl(fd, FIONREAD, &l) < 0) {
- log_error("FIONREAD failed: %m");
- goto fail;
- }
-
- assert(l > 0);
-
- if (!(buf = malloc(l))) {
- log_error("Failed to allocate buffer: %s", strerror(ENOMEM));
- goto fail;
- }
-
- if ((k = read(fd, buf, l)) < 0) {
- log_error("Failed to read inotify event: %m");
+ changed = pathspec_fd_event(s, events);
+ if (changed < 0)
goto fail;
- }
/* If we are already running, then remember that one event was
* dispatched so that we restart the service only if something
* actually changed on disk */
p->inotify_triggered = true;
- e = (struct inotify_event*) buf;
-
- changed = false;
- while (k > 0) {
- size_t step;
-
- if (s->type == PATH_CHANGED && s->primary_wd == e->wd)
- changed = true;
-
- step = sizeof(struct inotify_event) + e->len;
- assert(step <= (size_t) k);
-
- e = (struct inotify_event*) ((uint8_t*) e + step);
- k -= step;
- }
-
if (changed)
path_enter_running(p);
else
path_enter_waiting(p, false, true);
- free(buf);
-
return;
fail:
- free(buf);
path_enter_dead(p, false);
}
diff --git a/src/path.h b/src/path.h
index 116fc63..4e6ccf5 100644
--- a/src/path.h
+++ b/src/path.h
@@ -60,6 +60,14 @@ typedef struct PathSpec {
} PathSpec;
+int pathspec_watch(PathSpec *s, Unit *u);
+void pathspec_unwatch(PathSpec *s, Unit *u);
+int pathspec_fd_event(PathSpec *s, uint32_t events);
+void pathspec_done(PathSpec *s);
+static inline bool pathspec_owns_inotify_fd(PathSpec *s, int fd) {
+ return s->inotify_fd == fd;
+}
+
struct Path {
Meta meta;
--
1.7.7
From e92238567b9fc83ef77e359588d7b005ecae3d70 Mon Sep 17 00:00:00 2001
From: Michal Schmidt <mschmidt@redhat.com>
Date: Sat, 3 Dec 2011 10:22:26 +0100
Subject: [PATCH 4/5] path: add PathModified (= PathChanged + IN_MODIFY)
---
man/systemd.path.xml | 17 ++++++++++-------
src/path.c | 2 ++
src/path.h | 1 +
3 files changed, 13 insertions(+), 7 deletions(-)
diff --git a/man/systemd.path.xml b/man/systemd.path.xml
index 10d8f73..5b1ff75 100644
--- a/man/systemd.path.xml
+++ b/man/systemd.path.xml
@@ -113,6 +113,7 @@
<term><varname>PathExists=</varname></term>
<term><varname>PathExistsGlob=</varname></term>
<term><varname>PathChanged=</varname></term>
+ <term><varname>PathModified=</varname></term>
<term><varname>DirectoryNotEmpty=</varname></term>
<listitem><para>Defines paths to
@@ -129,8 +130,14 @@
specified. <varname>PathChanged=</varname>
may be used to watch a file or
directory and activate the configured
- unit whenever it changes or is
- modified. <varname>DirectoryNotEmpty=</varname>
+ unit whenever it changes. It is not activated
+ on every write to the watched file but it is
+ activated if the file which was open for writing
+ gets closed. <varname>PathModified=</varname>
+ is similar, but additionally it is activated
+ also on simple writes to the watched file.
+
+ <varname>DirectoryNotEmpty=</varname>
may be used to watch a directory and
activate the configured unit whenever
it contains at least one file.</para>
@@ -154,11 +161,7 @@
activated, then the configured unit is
immediately activated as
well. Something similar does not apply
- to
- <varname>PathChanged=</varname>. The
- latter is not activated on simple
- writes but only if files with were
- opened for writing are closed.
+ to <varname>PathChanged=</varname>.
</para></listitem>
</varlistentry>
<varlistentry>
diff --git a/src/path.c b/src/path.c
index db6f873..1e5d825 100644
--- a/src/path.c
+++ b/src/path.c
@@ -44,6 +44,7 @@ int pathspec_watch(PathSpec *s, Unit *u) {
[PATH_EXISTS] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB,
[PATH_EXISTS_GLOB] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB,
[PATH_CHANGED] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CLOSE_WRITE|IN_CREATE|IN_DELETE|IN_MOVED_FROM|IN_MOVED_TO,
+ [PATH_MODIFIED] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CLOSE_WRITE|IN_CREATE|IN_DELETE|IN_MOVED_FROM|IN_MOVED_TO|IN_MODIFY,
[PATH_DIRECTORY_NOT_EMPTY] = IN_DELETE_SELF|IN_MOVE_SELF|IN_ATTRIB|IN_CREATE|IN_MOVED_TO
};
@@ -713,6 +714,7 @@ static const char* const path_type_table[_PATH_TYPE_MAX] = {
[PATH_EXISTS] = "PathExists",
[PATH_EXISTS_GLOB] = "PathExistsGlob",
[PATH_CHANGED] = "PathChanged",
+ [PATH_MODIFIED] = "PathModified",
[PATH_DIRECTORY_NOT_EMPTY] = "DirectoryNotEmpty"
};
diff --git a/src/path.h b/src/path.h
index 4e6ccf5..1d78fe4 100644
--- a/src/path.h
+++ b/src/path.h
@@ -41,6 +41,7 @@ typedef enum PathType {
PATH_EXISTS_GLOB,
PATH_DIRECTORY_NOT_EMPTY,
PATH_CHANGED,
+ PATH_MODIFIED,
_PATH_TYPE_MAX,
_PATH_TYPE_INVALID = -1
} PathType;
--
1.7.7
From 3a11183858af30bc9b4e9dac430dd7541deec19b Mon Sep 17 00:00:00 2001
From: Michal Schmidt <mschmidt@redhat.com>
Date: Sat, 3 Dec 2011 02:13:30 +0100
Subject: [PATCH 5/5] service: handle services with racy daemonization
gracefully
There are a lot of forking daemons that do not exactly follow the
initialization steps as described in daemon(7). It is common that they
do not bother waiting in the parent process for the child to write the
PID file before exiting. The daemons' developers often do not perceive
this as a bug and they're unwilling to change.
Currently systemd warns about the missing PID file and falls back to
guessing the main PID. Being not quite deterministic, the guess can be
wrong with bad consequences. If the guessing is disabled, determinism is
achieved at the cost of losing the ability of noticing when the main
process of the service dies.
As long as it does not negatively affect properly written services,
systemd should strive for compatibility even with services with racy
daemonization. It is possible to provide determinism _and_ main process
supervision to them.
If the PID file is not there, rather than guessing and considering the
service running immediately after getting the SIGCHLD from the ExecStart
(or ExecStartPost) process, we can keep the service in the activating
state for a bit longer. We can use inotify to wait for the PID file to
appear. Only when it finally does appear and we read a valid PID from
it, we'll move the service to the running state. If the PID file never
appears, the usual timeout kicks in and the service fails.
---
src/service.c | 175 ++++++++++++++++++++++++++++++++++++++++++++++++++-------
src/service.h | 3 +
2 files changed, 157 insertions(+), 21 deletions(-)
diff --git a/src/service.c b/src/service.c
index 6fc2484..07137d2 100644
--- a/src/service.c
+++ b/src/service.c
@@ -1302,8 +1302,8 @@ static int service_load_pid_file(Service *s, bool may_warn) {
if ((r = read_one_line_file(s->pid_file, &k)) < 0) {
if (may_warn)
- log_warning("Failed to read PID file %s after %s. The service might be broken.",
- s->pid_file, service_state_to_string(s->state));
+ log_info("PID file %s not readable (yet?) after %s.",
+ s->pid_file, service_state_to_string(s->state));
return r;
}
@@ -1315,8 +1315,8 @@ static int service_load_pid_file(Service *s, bool may_warn) {
if (kill(pid, 0) < 0 && errno != EPERM) {
if (may_warn)
- log_warning("PID %lu read from file %s does not exist. Your service or init script might be broken.",
- (unsigned long) pid, s->pid_file);
+ log_info("PID %lu read from file %s does not exist.",
+ (unsigned long) pid, s->pid_file);
return -ESRCH;
}
@@ -1328,7 +1328,8 @@ static int service_load_pid_file(Service *s, bool may_warn) {
(unsigned long) s->main_pid, (unsigned long) pid);
service_unwatch_main_pid(s);
s->main_pid_known = false;
- }
+ } else
+ log_debug("Main PID loaded: %lu", (unsigned long) pid);
if ((r = service_set_main_pid(s, pid)) < 0)
return r;
@@ -1359,6 +1360,7 @@ static int service_search_main_pid(Service *s) {
if ((pid = cgroup_bonding_search_main_pid_list(s->meta.cgroup_bondings)) <= 0)
return -ENOENT;
+ log_debug("Main PID guessed: %lu", (unsigned long) pid);
if ((r = service_set_main_pid(s, pid)) < 0)
return r;
@@ -1451,6 +1453,17 @@ static int service_notify_sockets_dead(Service *s) {
return 0;
}
+static void service_unwatch_pid_file(Service *s) {
+ if (!s->pid_file_pathspec)
+ return;
+
+ log_debug("Stopping watch for %s's PID file %s", s->meta.id, s->pid_file_pathspec->path);
+ pathspec_unwatch(s->pid_file_pathspec, UNIT(s));
+ pathspec_done(s->pid_file_pathspec);
+ free(s->pid_file_pathspec);
+ s->pid_file_pathspec = NULL;
+}
+
static void service_set_state(Service *s, ServiceState state) {
ServiceState old_state;
assert(s);
@@ -1458,6 +1471,8 @@ static void service_set_state(Service *s, ServiceState state) {
old_state = s->state;
s->state = state;
+ service_unwatch_pid_file(s);
+
if (state != SERVICE_START_PRE &&
state != SERVICE_START &&
state != SERVICE_START_POST &&
@@ -2602,6 +2617,95 @@ static bool service_check_snapshot(Unit *u) {
return !s->got_socket_fd;
}
+static int service_retry_pid_file(Service *s) {
+ int r;
+
+ assert(s->pid_file);
+ assert(s->state == SERVICE_START || s->state == SERVICE_START_POST);
+
+ r = service_load_pid_file(s, false);
+ if (r < 0)
+ return r;
+
+ service_unwatch_pid_file(s);
+
+ service_enter_running(s, true);
+ return 0;
+}
+
+static int service_watch_pid_file(Service *s) {
+ int r;
+
+ log_debug("Setting watch for %s's PID file %s", s->meta.id, s->pid_file_pathspec->path);
+ r = pathspec_watch(s->pid_file_pathspec, UNIT(s));
+ if (r < 0)
+ goto fail;
+
+ /* the pidfile might have appeared just before we set the watch */
+ service_retry_pid_file(s);
+
+ return 0;
+fail:
+ log_error("Failed to set a watch for %s's PID file %s: %s",
+ s->meta.id, s->pid_file_pathspec->path, strerror(-r));
+ service_unwatch_pid_file(s);
+ return r;
+}
+
+static int service_demand_pid_file(Service *s) {
+ PathSpec *ps;
+
+ assert(s->pid_file);
+ assert(!s->pid_file_pathspec);
+
+ ps = new0(PathSpec, 1);
+ if (!ps)
+ return -ENOMEM;
+
+ ps->path = strdup(s->pid_file);
+ if (!ps->path) {
+ free(ps);
+ return -ENOMEM;
+ }
+
+ path_kill_slashes(ps->path);
+
+ /* PATH_CHANGED would not be enough. There are daemons (sendmail) that
+ * keep their PID file open all the time. */
+ ps->type = PATH_MODIFIED;
+ ps->inotify_fd = -1;
+
+ s->pid_file_pathspec = ps;
+
+ return service_watch_pid_file(s);
+}
+
+static void service_fd_event(Unit *u, int fd, uint32_t events, Watch *w) {
+ Service *s = SERVICE(u);
+
+ assert(s);
+ assert(fd >= 0);
+ assert(s->state == SERVICE_START || s->state == SERVICE_START_POST);
+ assert(s->pid_file_pathspec);
+ assert(pathspec_owns_inotify_fd(s->pid_file_pathspec, fd));
+
+ log_debug("inotify event for %s", u->meta.id);
+
+ if (pathspec_fd_event(s->pid_file_pathspec, events) < 0)
+ goto fail;
+
+ if (service_retry_pid_file(s) == 0)
+ return;
+
+ if (service_watch_pid_file(s) < 0)
+ goto fail;
+
+ return;
+fail:
+ service_unwatch_pid_file(s);
+ service_enter_signal(s, SERVICE_STOP_SIGTERM, false);
+}
+
static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) {
Service *s = SERVICE(u);
bool success;
@@ -2707,7 +2811,7 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) {
success = true;
}
- log_full(success ? LOG_DEBUG : LOG_NOTICE,
+ log_full(success ? LOG_DEBUG : LOG_NOTICE,
"%s: control process exited, code=%s status=%i", u->meta.id, sigchld_code_to_string(code), status);
s->failure = s->failure || !success;
@@ -2742,27 +2846,41 @@ static void service_sigchld_event(Unit *u, pid_t pid, int code, int status) {
case SERVICE_START:
assert(s->type == SERVICE_FORKING);
- /* Let's try to load the pid
- * file here if we can. We
- * ignore the return value,
- * since the PID file might
- * actually be created by a
- * START_POST script */
-
- if (success) {
- service_load_pid_file(s, !s->exec_command[SERVICE_EXEC_START_POST]);
- service_search_main_pid(s);
+ if (!success) {
+ service_enter_signal(s, SERVICE_FINAL_SIGTERM, false);
+ break;
+ }
- service_enter_start_post(s);
+ if (s->pid_file) {
+ /* Let's try to load the pid file here if we can.
+ * The PID file might actually be created by a START_POST
+ * script. In that case don't worry if the loading fails. */
+ bool has_start_post = !!s->exec_command[SERVICE_EXEC_START_POST];
+ int r = service_load_pid_file(s, !has_start_post);
+ if (!has_start_post && r < 0) {
+ r = service_demand_pid_file(s);
+ if (r < 0 || !cgroup_good(s))
+ service_enter_signal(s, SERVICE_FINAL_SIGTERM, false);
+ break;
+ }
} else
- service_enter_signal(s, SERVICE_FINAL_SIGTERM, false);
+ service_search_main_pid(s);
+ service_enter_start_post(s);
break;
case SERVICE_START_POST:
if (success) {
- service_load_pid_file(s, true);
- service_search_main_pid(s);
+ if (s->pid_file) {
+ int r = service_load_pid_file(s, true);
+ if (r < 0) {
+ r = service_demand_pid_file(s);
+ if (r < 0 || !cgroup_good(s))
+ service_enter_stop(s, false);
+ break;
+ }
+ } else
+ service_search_main_pid(s);
}
s->reload_failure = !success;
@@ -2907,6 +3025,20 @@ static void service_cgroup_notify_event(Unit *u) {
* except when we don't know pid which to expect the
* SIGCHLD for. */
+ case SERVICE_START:
+ case SERVICE_START_POST:
+ /* If we were hoping for the daemon to write its PID file,
+ * we can give up now. */
+ if (s->pid_file_pathspec) {
+ log_warning("%s never wrote its PID file. Failing.", s->meta.id);
+ service_unwatch_pid_file(s);
+ if (s->state == SERVICE_START)
+ service_enter_signal(s, SERVICE_FINAL_SIGTERM, false);
+ else
+ service_enter_stop(s, false);
+ }
+ break;
+
case SERVICE_RUNNING:
service_enter_running(s, true);
break;
@@ -3216,7 +3348,7 @@ static int service_enumerate(Manager *m) {
r = 0;
#ifdef TARGET_SUSE
- sysv_facility_in_insserv_conf (m);
+ sysv_facility_in_insserv_conf (m);
#endif
finish:
@@ -3511,6 +3643,7 @@ const UnitVTable service_vtable = {
.sigchld_event = service_sigchld_event,
.timer_event = service_timer_event,
+ .fd_event = service_fd_event,
.reset_failed = service_reset_failed,
diff --git a/src/service.h b/src/service.h
index e28f74b..15d58cc 100644
--- a/src/service.h
+++ b/src/service.h
@@ -86,6 +86,8 @@ typedef enum NotifyAccess {
_NOTIFY_ACCESS_INVALID = -1
} NotifyAccess;
+typedef struct PathSpec PathSpec;
+
struct Service {
Meta meta;
@@ -157,6 +159,7 @@ struct Service {
Set *configured_sockets;
Watch timer_watch;
+ PathSpec *pid_file_pathspec;
NotifyAccess notify_access;
};
--
1.7.7
From 714d943f72417f53bcb98ed45d002aa270e793c4 Mon Sep 17 00:00:00 2001
From: Michal Schmidt <mschmidt@redhat.com>
Date: Thu, 8 Dec 2011 12:09:10 +0100
Subject: [PATCH] path: add missing pieces for PathModified
PATH_MODIFIED worked internally for PID files detection, but was unusable
in units.
---
src/load-fragment-gperf.gperf.m4 | 1 +
src/path.c | 6 ++++--
2 files changed, 5 insertions(+), 2 deletions(-)
diff --git a/src/load-fragment-gperf.gperf.m4 b/src/load-fragment-gperf.gperf.m4
index 84ae28c..35ec005 100644
--- a/src/load-fragment-gperf.gperf.m4
+++ b/src/load-fragment-gperf.gperf.m4
@@ -210,6 +210,7 @@ m4_dnl
Path.PathExists, config_parse_path_spec, 0, 0
Path.PathExistsGlob, config_parse_path_spec, 0, 0
Path.PathChanged, config_parse_path_spec, 0, 0
+Path.PathModified, config_parse_path_spec, 0, 0
Path.DirectoryNotEmpty, config_parse_path_spec, 0, 0
Path.Unit, config_parse_path_unit, 0, 0
Path.MakeDirectory, config_parse_bool, 0, offsetof(Path, make_directory)
diff --git a/src/path.c b/src/path.c
index 1e5d825..3fee247 100644
--- a/src/path.c
+++ b/src/path.c
@@ -149,7 +149,8 @@ int pathspec_fd_event(PathSpec *s, uint32_t events) {
while (k > 0) {
size_t step;
- if (s->type == PATH_CHANGED && s->primary_wd == e->wd)
+ if ((s->type == PATH_CHANGED || s->type == PATH_MODIFIED) &&
+ s->primary_wd == e->wd)
r = 1;
step = sizeof(struct inotify_event) + e->len;
@@ -184,7 +185,8 @@ static bool pathspec_check_good(PathSpec *s, bool initial) {
break;
}
- case PATH_CHANGED: {
+ case PATH_CHANGED:
+ case PATH_MODIFIED: {
bool b;
b = access(s->path, F_OK) >= 0;
--
1.7.7
Reference in New Issue Copy Permalink