On Tue, Mar 02, 2021 at 01:04:19PM +0800, Anand Jain wrote: > On 26/02/2021 23:10, David Sterba wrote: > > On Fri, Feb 26, 2021 at 01:01:02PM +0800, Anand Jain wrote: > >> On 25/02/2021 12:39, Su Yue wrote: > >>> > >>> While playing with seed device(misc/next and v5.11), lockdep complains > >>> the following: > >>> > >>> To reproduce: > >>> > >>> dev1=/dev/sdb1 > >>> dev2=/dev/sdb2 > >>> > >>> umount /mnt > >>> > >>> mkfs.btrfs -f $dev1 > >>> > >>> btrfstune -S 1 $dev1 > >>> > >>> mount $dev1 /mnt > >>> > >>> btrfs device add $dev2 /mnt/ -f > >>> > >>> umount /mnt > >>> > >>> mount $dev2 /mnt > >>> > >>> umount /mnt > >>> > >>> > >> > >> In my understanding the commit 01d01caf19ff7c537527d352d169c4368375c0a1 > >> (btrfs: move the chunk_mutex in btrfs_read_chunk_tree > >> fixed this bug in 5.9. > >> Could you please try this [1] patch, > >> [1] > >> https://patchwork.kernel.org/project/linux-btrfs/patch/20200717100525.320697-1-anand.j...@oracle.com/ > >> Patch [1] still relevant as the device_list_mutex in clone_fs_devices() > >> is redundant. We could remove it as well. > > > > So the fix 01d01caf19ff7c was not sufficient, the lockdep splat is > > reproducible. > > Yes indeed. Except for adding another reported by, the patch[1] applies > on misc-next as it is. Do you need a resend of the patch?
Yes please resend, we had other fixes around device locking and that patch also has a nak because of other fixes so we need to put that into the right context.
