linux-kernel - Re: [PATCH 0/6] libsas and drivers: NCQ error handling

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Fri, 12 Aug 2022 08:39:21 -0700
From:   Damien Le Moal <damien.lemoal@...nsource.wdc.com>
To:     John Garry <john.garry@...wei.com>, jejb@...ux.ibm.com,
        martin.petersen@...cle.com, jinpu.wang@...ud.ionos.com,
        yangxingui@...wei.com, chenxiang66@...ilicon.com, hare@...e.de
Cc:     linux-scsi@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 0/6] libsas and drivers: NCQ error handling

On 2022/08/12 1:06, John Garry wrote:
> On 11/08/2022 19:54, Damien Le Moal wrote:
>> On 2022/07/22 4:24, John Garry wrote:
>>> As reported in [0], the pm8001 driver NCQ error handling more or less
>>> duplicates what libata does in link error handling, as follows:
>>> - abort all commands
>>> - do autopsy with read log ext 10 command
>>> - reset the target to recover
>>>
>>> Indeed for the hisi_sas driver we want to add similar handling for NCQ
>>> errors.
>>>
>>> This series add a new libsas API - sas_ata_link_abort() - to handle host
>>> NCQ errors, and fixes up pm8001 and hisi_sas drivers to use it. As
>>> mentioned in the pm8001 changeover patch, I would prefer a better place to
>>> locate the SATA ABORT command (rather that nexus reset callback).
>>>
>>> I would appreciate some testing of the pm8001 change as the read log ext10
>>> command mostly hangs on my arm64 machine - these arm64 hangs are a known
>>> issue.
>>
> 
> Thanks for this!
> 
>> I applied this series on top of the current Linus tree and ran some tests: a
>> bunch of fio runs and also ran libzbc test suites on a SATA SMR drive as that
>> generates many command failures. No problems detected, the tests all pass.
>> FYI, messages for failed commands look like this:
>>
>> pm80xx0:: mpi_sata_event 2685: SATA EVENT 0x23
>> sas: Enter sas_scsi_recover_host busy: 1 failed: 1
>> sas: sas_scsi_find_task: aborting task 0x00000000ba62a907
>> pm80xx0:: mpi_sata_completion 2292: task null, freeing CCB tag 2
>> sas: sas_scsi_find_task: task 0x00000000ba62a907 is aborted
>> sas: sas_eh_handle_sas_errors: task 0x00000000ba62a907 is aborted
>> ata21.00: exception Emask 0x0 SAct 0x20000000 SErr 0x0 action 0x0
>> ata21.00: failed command: WRITE FPDMA QUEUED
>> ata21.00: cmd 61/02:00:ff:ff:ea/00:00:02:00:00/40 tag 29 ncq dma 8192 out
>> res 43/04:02:ff:ff:ea/00:00:02:00:00/00 Emask 0x400 (NCQ error) <F>
>> ata21.00: status: { DRDY SENSE ERR }
>> ata21.00: error: { ABRT }
>> ata21.00: configured for UDMA/133
>> ata21: EH complete
>> sas: --- Exit sas_scsi_recover_host: busy: 0 failed: 1 tries: 1
>>
> 
> For this specific test we don't seem to run a hardreset after the 
> autopsy, but we do seem to be getting an NCQ error. That's interesting.
> 
> We have noticed this scenario for hisi_sas NCQ error, whereby the 
> autopsy decided a reset is not required or useful, such as a medium 
> error. Anyway the pm8001 driver relies on the reset being run always for 
> the NCQ error. So I am thinking of tweaking sas_ata_link_abort() as follows:
> 
> void sas_ata_link_abort(struct domain_device *device)
> {
> 	struct ata_port *ap = device->sata_dev.ap;
> 	struct ata_link *link = &ap->link;
> 
> 	link->eh_info.err_mask |= AC_ERR_DEV;
> +	link->eh_info.action |= ATA_EH_RESET;
> 	ata_link_abort(link);
> }
> 
> This should force a reset.

This is an unaligned write to a sequential write required zone on SMR. So
definitely not worth a reset. Forcing hard resetting the link for such error is
an overkill. I think it is better to let ata_link_abort() -> ... -> scsi & ata
EH decide on the disposition.

Note that patch 3 did not apply cleanly to the current Linus tree. So a rebase
for the series is needed.

> 
> Thanks,
> John
> 
>> Seems all good to me.
>>
>>>
>>> Finally with these changes we can make the libsas task alloc/free APIs
>>> private, which they should always have been.
>>>
>>> Based on v5.19-rc6
>>>
>>> [0] https://lore.kernel.org/linux-scsi/8fb3b093-55f0-1fab-81f4-e8519810a978@huawei.com/
>>>
>>> John Garry (5):
>>>    scsi: pm8001: Modify task abort handling for SATA task
>>>    scsi: libsas: Add sas_ata_link_abort()
>>>    scsi: pm8001: Use sas_ata_link_abort() to handle NCQ errors
>>>    scsi: hisi_sas: Don't issue ATA softreset in hisi_sas_abort_task()
>>>    scsi: libsas: Make sas_{alloc, alloc_slow, free}_task() private
>>>
>>> Xingui Yang (1):
>>>    scsi: hisi_sas: Add SATA_DISK_ERR bit handling for v3 hw
>>>
>>>   drivers/scsi/hisi_sas/hisi_sas_main.c  |   5 +-
>>>   drivers/scsi/hisi_sas/hisi_sas_v3_hw.c |  22 ++-
>>>   drivers/scsi/libsas/sas_ata.c          |  10 ++
>>>   drivers/scsi/libsas/sas_init.c         |   3 -
>>>   drivers/scsi/libsas/sas_internal.h     |   4 +
>>>   drivers/scsi/pm8001/pm8001_hwi.c       | 194 +++++++------------------
>>>   drivers/scsi/pm8001/pm8001_sas.c       |  13 ++
>>>   drivers/scsi/pm8001/pm8001_sas.h       |   8 +-
>>>   drivers/scsi/pm8001/pm80xx_hwi.c       | 177 ++--------------------
>>>   include/scsi/libsas.h                  |   4 -
>>>   include/scsi/sas_ata.h                 |   5 +
>>>   11 files changed, 132 insertions(+), 313 deletions(-)
>>>
>>
>>
> 


-- 
Damien Le Moal
Western Digital Research