On 07/21/2017 05:20 AM, Fam Zheng wrote: > Build time check of OFD lock is not sufficient and can cause image open > errors when the runtime environment doesn't support it. > > Add a helper function to probe it at runtime, additionally. Also provide > a qemu_has_ofd_lock() for callers to check the status. > > Signed-off-by: Fam Zheng <f...@redhat.com> > --- > include/qemu/osdep.h | 1 + > util/osdep.c | 72 > +++++++++++++++++++++++++++++++++++++++++++++------- > 2 files changed, 64 insertions(+), 9 deletions(-)
>
> +static void qemu_probe_lock_ops(void)
> +{
> + if (fcntl_op_setlk == -1) {
> +#ifdef F_OFD_SETLK
> + int fd;
> + int ret;
> + const char *tmpdir;
> + char *filename;
> + struct flock fl = {
> + .l_whence = SEEK_SET,
> + .l_start = 0,
> + .l_len = 0,
> + .l_type = F_WRLCK,
Do we actually need to grab a lock over the entire file,...
> + };
> +
> + tmpdir = getenv("TMPDIR");
> + if (!tmpdir) {
> + tmpdir = "/var/tmp";
> + }
> + filename = g_strdup_printf("%s/qemu_lock_probe.XXXXXX", tmpdir);
> + fd = mkstemp(filename);
Can we skip the temporary file, and just open("/dev/null", O_RDWR), for
slightly less work?
> + if (fd < 0) {
> + fprintf(stderr, "Failed to create temporary file '%s': %s\n",
> + filename, strerror(errno));
> + fcntl_op_setlk = F_SETLK;
> + fcntl_op_getlk = F_GETLK;
> + goto out;
> + }
> + ret = fcntl(fd, F_OFD_SETLK, &fl);
...or should we change this to the much-weaker F_OFD_GETLK? I guess it
doesn't matter if we use a temporary file, but does if we are trying to
simplify by using /dev/null.
At any rate, the rest of the patch looks sane.
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
signature.asc
Description: OpenPGP digital signature
