On Wed, Jul 17, 2024 at 04:18:33PM -0700, Khazhismel Kumykov wrote:
> do_resume when loading a new map first calls dm_suspend, which could
> silently fail. When we proceeded to dm_swap_table, we would bail out
> with EINVAL. Instead, attempt to restore new_map and return ERESTARTSYS
> when signaled.
>
> Signed-off-by: Khazhismel Kumykov <khazhy@xxxxxxxxxx>
> ---
> drivers/md/dm-ioctl.c | 23 +++++++++++++++++++++--
> 1 file changed, 21 insertions(+), 2 deletions(-)
>
> v2: don't leak new_map if we can't assign it back to hc.
>
> diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c
> index c2c07bfa6471..0591455ad63c 100644
> --- a/drivers/md/dm-ioctl.c
> +++ b/drivers/md/dm-ioctl.c
> @@ -1181,8 +1181,27 @@ static int do_resume(struct dm_ioctl *param)
> suspend_flags &= ~DM_SUSPEND_LOCKFS_FLAG;
> if (param->flags & DM_NOFLUSH_FLAG)
> suspend_flags |= DM_SUSPEND_NOFLUSH_FLAG;
> - if (!dm_suspended_md(md))
> - dm_suspend(md, suspend_flags);
> + if (!dm_suspended_md(md)) {
> + r = dm_suspend(md, suspend_flags);
> + if (r == -EINTR)
> + r = -ERESTARTSYS;
> + if (r) {
> + down_write(&_hash_lock);
> + hc = dm_get_mdptr(md);
> + if (!hc)
> + r = -ENXIO;
> + if (hc && !hc->new_map) {
> + hc->new_map = new_map;
> + up_write(&_hash_lock);
> + } else {
> + up_write(&_hash_lock);
> + dm_sync_table(md);
> + dm_table_destroy(new_map);
> + }
> + dm_put(md);
> + return r;
> + }
> + }
>
> old_size = dm_get_size(md);
> old_map = dm_swap_table(md, new_map);
> --
> 2.45.2.993.g49e7a77208-goog
>
>
Thanks for the patch. The header could use more context for how this
issue has caused problems in practice (you touched on that in reply to
Mikulas for v1).
But I will review this closely starting the week of July 29. This is
a very fundamental codepath for DM so needs extended review.
Mike
