Re: [PATCH] scsi: sd: Keep the discard mode stable
From: Li Feng
Date: Mon Jun 17 2024 - 23:06:41 EST
> 2024年6月18日 00:26,Benjamin Block <bblock@xxxxxxxxxxxxx> 写道:
>
> Hey,
>
> On Sat, Jun 15, 2024 at 12:03:47AM +0800, Li Feng wrote:
>> There is a scenario where a large number of discard commands
>> are issued when the iscsi initiator connects to the target
>> and then performs a session rescan operation.
>
> Is this with just one specific target implementation? This sounds like a
> broken/buggy target, or is there a reason why this happens in general?
>
> And broken target sounds like device quirk, rather than impacting every
> possible target.
(resend due to GMail HTML bounce)
This is a common problem. Before sending a rescan, discard has been
negotiated to UNMAP. After the rescan, there will be a short window for
it to become WS16, and then it will immediately become UNMAP.
However, during this period, a small amount of discard commands
may become WS16, resulting in a strange problem.
>
>> There is a time
>> window, most of the commands are in UNMAP mode, and some
>> discard commands become WRITE SAME with UNMAP.
>>
>> The discard mode has been negotiated during the SCSI probe. If
>> the mode is temporarily changed from UNMAP to WRITE SAME with
>> UNMAP, IO ERROR may occur because the target may not implement
>> WRITE SAME with UNMAP. Keep the discard mode stable to fix this
>> issue.
>>
>> Signed-off-by: Li Feng <fengli@xxxxxxxxxx>
>> ---
>> drivers/scsi/sd.c | 7 ++++++-
>> 1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
>> index f6c822c9cbd2..0165dc70a99b 100644
>> --- a/drivers/scsi/sd.c
>> +++ b/drivers/scsi/sd.c
>> @@ -2598,7 +2598,12 @@ static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
>> if (buffer[14] & 0x40) /* LBPRZ */
>> sdkp->lbprz = 1;
>>
>> - sd_config_discard(sdkp, SD_LBP_WS16);
>> + /*
>> + * When the discard mode has been set to UNMAP, it should not be set to
>> + * WRITE SAME with UNMAP.
>> + */
>> + if (!sdkp->max_unmap_blocks)
>> + sd_config_discard(sdkp, SD_LBP_WS16);
>> }
>>
>> sdkp->capacity = lba + 1;
>
> --
> Best Regards, Benjamin Block / Linux on IBM Z Kernel Development
> IBM Deutschland Research & Development GmbH / https://www.ibm.com/privacy
> Vors. Aufs.-R.: Wolfgang Wendt / Gesch?ftsf?hrung: David Faller
> Sitz der Ges.: B?blingen / Registergericht: AmtsG Stuttgart, HRB 243294