At 05/30/2012 09:17 PM, Eric Blake Wrote:
On 05/30/2012 03:20 AM, Wen Congyang wrote:
fdstream: If fd is fds[0] or fds[1], we should set to -1 if we meet some error.
childfd is fds[0] or fds[1], so we should close it only when virFDStreamOpenFileInternal() successes.
qemu_migration: If we migrate to fd, spec->fwdType is not MIGRATION_FWD_DIRECT, we will close spec->dest.fd.local in qemuMigrationRun(). So we should set spec->dest.fd.local to -1 in qemuMigrationRun().
command: we should not set *outfd or *errfd if virExecWithHook() failed because the caller may close these fds.
We should split this into three separate patches, to aid backporting each patch across appropriate versions. Needs a v2 for this reason; as I want these bugs fixed sooner rather than later, I'll probably help by reposting things myself.
--- src/fdstream.c | 15 ++++++++++----- src/qemu/qemu_migration.c | 4 +++- src/util/command.c | 8 ++++---- 3 files changed, 17 insertions(+), 10 deletions(-)
diff --git a/src/fdstream.c b/src/fdstream.c index 32d386d..d0ea0ee 100644 --- a/src/fdstream.c +++ b/src/fdstream.c @@ -581,6 +581,7 @@ virFDStreamOpenFileInternal(virStreamPtr st, struct stat sb; virCommandPtr cmd = NULL; int errfd = -1; + int childfd = -1;
VIR_DEBUG("st=%p path=%s oflags=%x offset=%llu length=%llu mode=%o", st, path, oflags, offset, length, mode); @@ -619,7 +620,6 @@ virFDStreamOpenFileInternal(virStreamPtr st, if ((st->flags & VIR_STREAM_NONBLOCK) && (!S_ISCHR(sb.st_mode) && !S_ISFIFO(sb.st_mode))) { - int childfd;
if ((oflags & O_ACCMODE) == O_RDWR) { streamsReportError(VIR_ERR_INTERNAL_ERROR, @@ -652,15 +652,20 @@ virFDStreamOpenFileInternal(virStreamPtr st, } virCommandSetErrorFD(cmd, &errfd);
- if (virCommandRunAsync(cmd, NULL) < 0) + if (virCommandRunAsync(cmd, NULL) < 0) { + /* donot close fd twice if we meet some error */
s/donot/don't/
+ fd = -1; goto error; - - VIR_FORCE_CLOSE(childfd); + } }
- if (virFDStreamOpenInternal(st, fd, cmd, errfd, length) < 0) + if (virFDStreamOpenInternal(st, fd, cmd, errfd, length) < 0) { + /* donot close fd twice if we meet some error */
and again.
+ fd = -1; goto error; + }
+ VIR_FORCE_CLOSE(childfd); return 0;
Doesn't this leak childfd on error?
No, childfd is fds[0] or fds[1]. We have closed fds[0] and fds[1] on error, so we should not close childfd on error.
Maybe a better solution here would be that when we assign fd and childfd to elements of fds[], then we also assign fds to -1, as in:
Good idea. Thanks Wen Congyang
diff --git i/src/fdstream.c w/src/fdstream.c index 32d386d..b797a05 100644 --- i/src/fdstream.c +++ w/src/fdstream.c @@ -581,6 +581,7 @@ virFDStreamOpenFileInternal(virStreamPtr st, struct stat sb; virCommandPtr cmd = NULL; int errfd = -1; + int childfd = -1;
VIR_DEBUG("st=%p path=%s oflags=%x offset=%llu length=%llu mode=%o", st, path, oflags, offset, length, mode); @@ -619,7 +620,6 @@ virFDStreamOpenFileInternal(virStreamPtr st, if ((st->flags & VIR_STREAM_NONBLOCK) && (!S_ISCHR(sb.st_mode) && !S_ISFIFO(sb.st_mode))) { - int childfd;
if ((oflags & O_ACCMODE) == O_RDWR) { streamsReportError(VIR_ERR_INTERNAL_ERROR, @@ -650,6 +650,7 @@ virFDStreamOpenFileInternal(virStreamPtr st, fd = fds[1]; virCommandSetInputFD(cmd, childfd); } + fds[0] = fds[1] = -1; virCommandSetErrorFD(cmd, &errfd);
if (virCommandRunAsync(cmd, NULL) < 0) @@ -668,6 +669,7 @@ error: VIR_FORCE_CLOSE(fds[0]); VIR_FORCE_CLOSE(fds[1]); VIR_FORCE_CLOSE(fd); + VIR_FORCE_CLOSE(childfd); if (oflags & O_CREAT) unlink(path); return -1;
+++ b/src/qemu/qemu_migration.c @@ -1910,8 +1910,10 @@ qemuMigrationRun(struct qemud_driver *driver, break;
case MIGRATION_DEST_FD: - if (spec->fwdType != MIGRATION_FWD_DIRECT) + if (spec->fwdType != MIGRATION_FWD_DIRECT) { fd = spec->dest.fd.local; + spec->dest.fd.local = -1; + } ret = qemuMonitorMigrateToFd(priv->mon, migrate_flags, spec->dest.fd.qemu); VIR_FORCE_CLOSE(spec->dest.fd.qemu);
This hunk is probably okay on its own.
diff --git a/src/util/command.c b/src/util/command.c index eaa9f16..2723fde 100644 --- a/src/util/command.c +++ b/src/util/command.c @@ -493,6 +493,10 @@ virExecWithHook(const char *const*argv, }
if (pid) { /* parent */ + if (forkRet < 0) { + goto cleanup; + } + VIR_FORCE_CLOSE(null); if (outfd && *outfd == -1) { VIR_FORCE_CLOSE(pipeout[1]); @@ -503,10 +507,6 @@ virExecWithHook(const char *const*argv, *errfd = pipeerr[0]; }
- if (forkRet < 0) { - goto cleanup; - } - *retpid = pid;
if (binary != argv[0])
And this one is probably okay as well (I'll have to spend more time reading it); again a reason for splitting these into independent patches.