[pve-devel] rbd: don't assert on writes to snapshots

[Thomas Lamprecht]

Hi,

On 7/16/19 10:50 PM, Kitt wrote:
> Hello.
> 
> Would it be possible to add this patch to the bionic and dingo kernels and
> fix the ceph snapshot hang issue?

Yes, it would. The commit[0] from the diff you linked, looks OK.

[0]: https://git.kernel.org/torvalds/c/b91a7bdca4439e286f26cdd6c15ed338e6a9fda2

> 
> Diff:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/diff/drivers/block/rbd.c?id=b91a7bdca4439e286f26cdd6c15ed338e6a9fda2
> 
> diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
> index 99de7166bf89..e5009a34f9c2 100644
> --- a/drivers/block/rbd.c
> <https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/block/rbd.c?id=a32e414325c2f0d430436e4708a33c756b082fd8>
> +++ b/drivers/block/rbd.c
> <https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/block/rbd.c?id=b91a7bdca4439e286f26cdd6c15ed338e6a9fda2>
> @@ -3842,8 +3842,12 @@ static void rbd_queue_workfn(struct work_struct
> *work)
> goto err_rq;
> }
> - rbd_assert(op_type == OBJ_OP_READ ||
> - rbd_dev->spec->snap_id == CEPH_NOSNAP);
> + if (op_type != OBJ_OP_READ && rbd_dev->spec->snap_id != CEPH_NOSNAP) {
> + rbd_warn(rbd_dev, "%s on read-only snapshot",
> + obj_op_name(op_type));
> + result = -EIO;
> + goto err;
> + }
> /*
> * Quit early if the mapped snapshot no longer exists. It's
> 
> 5.2 kernel commit:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b91a7bdca4439e286f26cdd6c15ed338e6a9fda2
> 
> Ceph bug notes: https://tracker.ceph.com/issues/39010
> 
> Proxmox bugzilla: https://bugzilla.proxmox.com/show_bug.cgi?id=1911
> 
> 
> Thank you.
> 

Much thanks for the hint here, and the detailed information you posted!

I'll queue this patch up for the next kernel release (probably early
next week).

cheers,
Thomas