Re: [PATCH] pm-hibernate: flush block device cache when hibernating

[Askar Safin <safinaskar@gmail.com>]

Mikulas Patocka <mpatocka@redhat.com>:
> Hi
> 
> Does this patch fix it?
> 
> Mikulas
> 
> 
> From: Mikulas Patocka <mpatocka@redhat.com>
> 
> There was reported failure that hibernation doesn't work with 
> dm-integrity. The reason for the failure is that the hibernation code 
> doesn't issue the FLUSH bio - the data still sits in the dm-integrity 
> cache and they are lost when poweroff happens.

I tested this patch in Qemu on current master (43e9ad0c55a3). Also I
applied Mario's patch
https://lore.kernel.org/linux-pm/20251026033115.436448-1-superm1@kernel.org/ .
It is needed, otherwise you get WARNING when you try to hibernate.

The patch doesn't work.

Here is script I used for reproduction:
https://zerobin.net/?66669be7d2404586#xWufhCq7zCoOk3LJcJCj7W4k3vYT3U4vhGutTN3p8m0= .
It is the same script as in previous letter. I just added some
"integritysetup status /dev/mapper/..." calls.

Here are results:
https://zerobin.net/?2331637d633d20c5#EmyhxiHLDmoZT1jBVbe/q9iJKhDEw4n+Bwr5mAcaOpM= .

File names mean the same as in previous letter, i. e.:

> "log-def-1" is output of first Qemu invocation (i. e. first boot) with
> default integritysetup options. "log-def-2" is second Qemu invocation
> (i. e. when we try to resume).
> 
> log-bit-{1,2} is same thing, but with "--integrity-bitmap-mode" added to
> "integritysetup format" and "integritysetup open".
> 
> log-no-{1,2} is same, but with "--integrity-no-journal".
> 
> log-nodm-{1,2} is same, but without dm-integrity at all, i. e. we create
> swap directly on partition.

Results are somewhat better than without the patch. Without the patch
we don't even try to resume in default mode. "blkid" simply reports
"swap" instead of "swsuspend". With patch "blkid" reports "swsuspend", and
so we try to resume. But then in the middle of resuming we get this:

[    1.008223] PM: Image loading progress:  70%
[    1.017478] PM: Image loading progress:  80%
[    1.027069] PM: Image loading progress:  90%
[    1.029653] PM: hibernation: Read 36196 kbytes in 0.49 seconds (73.86 MB/s)
[    1.030146] PM: Error -1 resuming
[    1.030322] PM: hibernation: Failed to load image, recovering.

(See link above for full logs.)

Very similar thing happens in "--integrity-no-journal" mode in the middle of
resuming:

[    0.531245] device-mapper: integrity: dm-0: Checksum failed at sector 0x6e70
[    0.531600] PM: Error -84 resuming
[    0.531799] PM: hibernation: Failed to load image, recovering.

The patch doesn't change anything in "--integrity-bitmap-mode" mode:
we still are able to resume, but then get integrity errors when we do
"cat /dev/mapper/swap > /dev/null".

-- 
Askar Safin