提交 be78814a 编写于 作者: M Michal Privoznik

virNetSocketNewConnectUNIX: Use flocks when spawning a daemon

https://bugzilla.redhat.com/show_bug.cgi?id=1200149

Even though we have a mutex mechanism so that two clients don't spawn
two daemons, it's not strong enough. It can happen that while one
client is spawning the daemon, the other one fails to connect.
Basically two possible errors can happen:

  error: Failed to connect socket to '/home/mprivozn/.cache/libvirt/libvirt-sock': Connection refused

or:

  error: Failed to connect socket to '/home/mprivozn/.cache/libvirt/libvirt-sock': No such file or directory

The problem in both cases is, the daemon is only starting up, while we
are trying to connect (and fail). We should postpone the connecting
phase until the daemon is started (by the other thread that is
spawning it). In order to do that, create a file lock 'libvirt-lock'
in the directory where session daemon would create its socket. So even
when called from multiple processes, spawning a daemon will serialize
on the file lock. So only the first to come will spawn the daemon.
Tested-by: NRichard W. M. Jones <rjones@redhat.com>
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
上级 bb6a62d4
...@@ -123,7 +123,7 @@ VIR_ONCE_GLOBAL_INIT(virNetSocket) ...@@ -123,7 +123,7 @@ VIR_ONCE_GLOBAL_INIT(virNetSocket)
#ifndef WIN32 #ifndef WIN32
static int virNetSocketForkDaemon(const char *binary, int passfd) static int virNetSocketForkDaemon(const char *binary)
{ {
int ret; int ret;
virCommandPtr cmd = virCommandNewArgList(binary, virCommandPtr cmd = virCommandNewArgList(binary,
...@@ -136,10 +136,6 @@ static int virNetSocketForkDaemon(const char *binary, int passfd) ...@@ -136,10 +136,6 @@ static int virNetSocketForkDaemon(const char *binary, int passfd)
virCommandAddEnvPassBlockSUID(cmd, "XDG_RUNTIME_DIR", NULL); virCommandAddEnvPassBlockSUID(cmd, "XDG_RUNTIME_DIR", NULL);
virCommandClearCaps(cmd); virCommandClearCaps(cmd);
virCommandDaemonize(cmd); virCommandDaemonize(cmd);
if (passfd) {
virCommandPassFD(cmd, passfd, VIR_COMMAND_PASS_FD_CLOSE_PARENT);
virCommandPassListenFDs(cmd);
}
ret = virCommandRun(cmd, NULL); ret = virCommandRun(cmd, NULL);
virCommandFree(cmd); virCommandFree(cmd);
return ret; return ret;
...@@ -543,45 +539,26 @@ int virNetSocketNewConnectUNIX(const char *path, ...@@ -543,45 +539,26 @@ int virNetSocketNewConnectUNIX(const char *path,
const char *binary, const char *binary,
virNetSocketPtr *retsock) virNetSocketPtr *retsock)
{ {
char *binname = NULL; char *lockpath = NULL;
char *pidpath = NULL; int lockfd = -1;
int fd, passfd = -1; int fd = -1;
int pidfd = -1; int retries = 100;
virSocketAddr localAddr; virSocketAddr localAddr;
virSocketAddr remoteAddr; virSocketAddr remoteAddr;
char *rundir = NULL;
memset(&localAddr, 0, sizeof(localAddr)); memset(&localAddr, 0, sizeof(localAddr));
memset(&remoteAddr, 0, sizeof(remoteAddr)); memset(&remoteAddr, 0, sizeof(remoteAddr));
remoteAddr.len = sizeof(remoteAddr.data.un); remoteAddr.len = sizeof(remoteAddr.data.un);
if (spawnDaemon && !binary) { if (spawnDaemon) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s", const char *binname;
_("Auto-spawn of daemon requested, but no binary specified"));
return -1;
}
if ((fd = socket(PF_UNIX, SOCK_STREAM, 0)) < 0) {
virReportSystemError(errno, "%s", _("Failed to create socket"));
goto error;
}
remoteAddr.data.un.sun_family = AF_UNIX;
if (virStrcpyStatic(remoteAddr.data.un.sun_path, path) == NULL) {
virReportSystemError(ENOMEM, _("Path %s too long for unix socket"), path);
goto error;
}
if (remoteAddr.data.un.sun_path[0] == '@')
remoteAddr.data.un.sun_path[0] = '\0';
retry:
if (connect(fd, &remoteAddr.data.sa, remoteAddr.len) < 0) {
int status = 0;
pid_t pid = 0;
if (!spawnDaemon) { if (spawnDaemon && !binary) {
virReportSystemError(errno, _("Failed to connect socket to '%s'"), virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
path); _("Auto-spawn of daemon requested, "
"but no binary specified"));
goto error; goto error;
} }
...@@ -592,90 +569,63 @@ int virNetSocketNewConnectUNIX(const char *path, ...@@ -592,90 +569,63 @@ int virNetSocketNewConnectUNIX(const char *path,
goto error; goto error;
} }
if (virPidFileConstructPath(false, NULL, binname, &pidpath) < 0) if (!(rundir = virGetUserRuntimeDirectory()))
goto error; goto error;
pidfd = virPidFileAcquirePath(pidpath, false, getpid()); if (virFileMakePathWithMode(rundir, 0700) < 0) {
if (pidfd < 0) { virReportSystemError(errno,
/* _("Cannot create user runtime directory '%s'"),
* This can happen in a very rare case of two clients spawning two rundir);
* daemons at the same time, and the error in the logs that gets
* reset here can be a clue to some future debugging.
*/
virResetLastError();
spawnDaemon = false;
goto retry;
}
if ((passfd = socket(PF_UNIX, SOCK_STREAM, 0)) < 0) {
virReportSystemError(errno, "%s", _("Failed to create socket"));
goto error; goto error;
} }
/* if (virAsprintf(&lockpath, "%s/%s.lock", rundir, binname) < 0)
* We already even acquired the pidfile, so no one else should be using
* @path right now. So we're OK to unlink it and paying attention to
* the return value makes no real sense here. Only if it's not an
* abstract socket, of course.
*/
if (path[0] != '@')
unlink(path);
/*
* We have to fork() here, because umask() is set per-process, chmod()
* is racy and fchmod() has undefined behaviour on sockets according to
* POSIX, so it doesn't work outside Linux.
*/
if ((pid = virFork()) < 0)
goto error; goto error;
if (pid == 0) { if ((lockfd = open(lockpath, O_RDWR | O_CREAT, 0600)) < 0 ||
umask(0077); virSetCloseExec(lockfd) < 0) {
if (bind(passfd, &remoteAddr.data.sa, remoteAddr.len) < 0) virReportSystemError(errno, _("Unable to create lock '%s'"), lockpath);
_exit(EXIT_FAILURE);
_exit(EXIT_SUCCESS);
}
if (virProcessWait(pid, &status, false) < 0)
goto error; goto error;
if (status != EXIT_SUCCESS) {
/*
* OK, so the child failed to bind() the socket. This may mean that
* another daemon was starting at the same time and succeeded with
* its bind() (even though it should not happen because we using a
* pidfile for the race). So we'll try connecting again, but this
* time without spawning the daemon.
*/
spawnDaemon = false;
virPidFileDeletePath(pidpath);
VIR_FORCE_CLOSE(pidfd);
VIR_FORCE_CLOSE(passfd);
goto retry;
} }
if (listen(passfd, 0) < 0) { if (virFileLock(lockfd, false, 0, 1, true) < 0) {
virReportSystemError(errno, "%s", virReportSystemError(errno, _("Unable to lock '%s'"), lockpath);
_("Failed to listen on socket that's about "
"to be passed to the daemon"));
goto error; goto error;
} }
}
if ((fd = socket(PF_UNIX, SOCK_STREAM, 0)) < 0) {
virReportSystemError(errno, "%s", _("Failed to create socket"));
goto error;
}
if (connect(fd, &remoteAddr.data.sa, remoteAddr.len) < 0) { remoteAddr.data.un.sun_family = AF_UNIX;
if (virStrcpyStatic(remoteAddr.data.un.sun_path, path) == NULL) {
virReportSystemError(ENOMEM, _("Path %s too long for unix socket"), path);
goto error;
}
if (remoteAddr.data.un.sun_path[0] == '@')
remoteAddr.data.un.sun_path[0] = '\0';
while (retries &&
connect(fd, &remoteAddr.data.sa, remoteAddr.len) < 0) {
if (!(spawnDaemon && errno == ENOENT)) {
virReportSystemError(errno, _("Failed to connect socket to '%s'"), virReportSystemError(errno, _("Failed to connect socket to '%s'"),
path); path);
goto error; goto error;
} }
/* if (virNetSocketForkDaemon(binary) < 0)
* Do we need to eliminate the super-rare race here any more? It would
* need incorporating the following VIR_FORCE_CLOSE() into a
* virCommandHook inside a virNetSocketForkDaemon().
*/
VIR_FORCE_CLOSE(pidfd);
if (virNetSocketForkDaemon(binary, passfd) < 0)
goto error; goto error;
retries--;
usleep(5000);
}
if (lockfd) {
unlink(lockpath);
VIR_FORCE_CLOSE(lockfd);
VIR_FREE(lockpath);
} }
localAddr.len = sizeof(localAddr.data); localAddr.len = sizeof(localAddr.data);
...@@ -687,19 +637,15 @@ int virNetSocketNewConnectUNIX(const char *path, ...@@ -687,19 +637,15 @@ int virNetSocketNewConnectUNIX(const char *path,
if (!(*retsock = virNetSocketNew(&localAddr, &remoteAddr, true, fd, -1, 0))) if (!(*retsock = virNetSocketNew(&localAddr, &remoteAddr, true, fd, -1, 0)))
goto error; goto error;
VIR_FREE(pidpath);
return 0; return 0;
error: error:
if (pidfd >= 0) if (lockfd)
virPidFileDeletePath(pidpath); unlink(lockpath);
VIR_FREE(pidpath); VIR_FREE(lockpath);
VIR_FREE(rundir);
VIR_FORCE_CLOSE(fd); VIR_FORCE_CLOSE(fd);
VIR_FORCE_CLOSE(passfd); VIR_FORCE_CLOSE(lockfd);
VIR_FORCE_CLOSE(pidfd);
if (spawnDaemon)
unlink(path);
return -1; return -1;
} }
#else #else
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册