[Devel] [PATCH hci-8.1 v11 09/10] dm-qcow2: retry R1/R2 writeback that failed with ENOSPC

[Konstantin Khorenko]

I won't apply this version, please send v12 with the changes below.

On 11/9/26 15:01, Alexander Ivanov wrote:
> A failed R1/R2 writeback sets MD_WRITEBACK_ERROR and drops MD_DIRTY.

> which gives the page one more try. A page that still fails then is not
> redirtied, so no timer is left armed after the flush, before the image
> is freed, and md_writeback_error is set as before.

> +	/* Retry R1/R2 writeback failed with ENOSPC once, then give up */

Item 3 of the cover letter reply: say "up to two attempts", and that
MD_DIRTY is dropped at submit.

Why: a page that failed with ENOSPC before the flush is redirtied. Pass
0 fires the timer and writes it, and since md_wb_no_retry is still
false there, a second failure redirties it again. Pass 1 writes it once
more. MD_DIRTY is cleared in submit_metadata_writeback(), before the
write, not by its failure.

Item 4 of the cover letter reply: mention the leak per resubmission.

Why: with this patch a request whose L2 write fails with ENOSPC ends
with BLK_STS_NOSPC and is resubmitted after ENOSPC_TIMEOUT_JI. The L2
entry was reverted to 0, so the resubmitted request allocates a new
data cluster, and patch 8 keeps the old one. One sentence in the
message is enough; it lasts only as long as the host is out of space.

-- 
Konstantin Khorenko,
Virtuozzo Linux Kernel Team