On Tue, Dec 04, 2018 at 05:22:19PM +0200, Nikolay Borisov wrote:
> > @@ -3874,16 +3882,9 @@ int btrfs_scrub_dev(struct btrfs_fs_info *fs_info, u64 devid, u64 start,
> > if (ret) {
> > mutex_unlock(&fs_info->scrub_lock);
> > mutex_unlock(&fs_info->fs_devices->device_list_mutex);
> > - return ret;
> > + goto out_free_ctx;
>
> Don't we suffer the same issue when calling scrub_workers_get since in
> it we do btrfs_alloc_workqueue which also calls kzalloc with GFP_KERNEL?
Yes, that's right. I instrumented only the allocations in scrub.c to see
if the nofs and lock_not_held assertions work at all so this one did not
get caught directly.
As scrub_workers_get still needs the scrub_lock, fixing it by moving
does not work and would need more restructuring.
