by Damien Le Moal

[permalink] [raw]

Subject: Re: [PATCH 6/6] scsi: mvsas: Use sas_task_find_rq() for tagging

On 9/29/22 16:49, John Garry wrote:
> On 29/09/2022 03:22, Damien Le Moal wrote:
>> On 9/28/22 21:27, John Garry wrote:
>>> The request associated with a scsi command coming from the block layer
>>> has a unique tag, so use that when possible for getting a slot.
>>>
>>> Unfortunately we don't support reserved commands in the SCSI midlayer yet.
>>> As such, SMP tasks - as an example - will not have a request associated, so
>>> in the interim continue to manage those tags for that type of sas_task
>>> internally.
>>>
>>> We reserve an arbitrary 4 tags for these internal tags. Indeed, we already
>>> decrement MVS_RSVD_SLOTS by 2 for the shost can_queue when flag
>>> MVF_FLAG_SOC is set. This change was made in commit 20b09c2992fef
>>> ("[PATCH] [SCSI] mvsas: add support for 94xx; layout change; bug fixes"),
>>> but what those 2 slots are used for is not obvious.
>>>
>>> Signed-off-by: John Garry <[email protected]>
>>> ---
>>> drivers/scsi/mvsas/mv_defs.h | 1 +
>>> drivers/scsi/mvsas/mv_init.c | 4 ++--
>>> drivers/scsi/mvsas/mv_sas.c | 22 +++++++++++++++++-----
>>> drivers/scsi/mvsas/mv_sas.h | 1 -
>>> 4 files changed, 20 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/drivers/scsi/mvsas/mv_defs.h b/drivers/scsi/mvsas/mv_defs.h
>>> index 7123a2efbf58..8ef174cd4d37 100644
>>> --- a/drivers/scsi/mvsas/mv_defs.h
>>> +++ b/drivers/scsi/mvsas/mv_defs.h
>>> @@ -40,6 +40,7 @@ enum driver_configuration {
>>> MVS_ATA_CMD_SZ = 96, /* SATA command table buffer size */
>>> MVS_OAF_SZ = 64, /* Open address frame buffer size */
>>> MVS_QUEUE_SIZE = 64, /* Support Queue depth */
>>> + MVS_RSVD_SLOTS = 4,
>>> MVS_SOC_CAN_QUEUE = MVS_SOC_SLOTS - 2,
>>> };
>>>
>>> diff --git a/drivers/scsi/mvsas/mv_init.c b/drivers/scsi/mvsas/mv_init.c
>>> index c85fb812ad43..d834ed9e8e4a 100644
>>> --- a/drivers/scsi/mvsas/mv_init.c
>>> +++ b/drivers/scsi/mvsas/mv_init.c
>>> @@ -284,7 +284,7 @@ static int mvs_alloc(struct mvs_info *mvi, struct Scsi_Host *shost)
>>> printk(KERN_DEBUG "failed to create dma pool %s.\n", pool_name);
>>> goto err_out;
>>> }
>>> - mvi->tags_num = slot_nr;
>>> + mvi->tags_num = MVS_RSVD_SLOTS;
>>
>> Same comment as for pm8001: do you really need this field if the value
>> is always MVS_RSVD_SLOTS ?
>
> Right, I don't need this struct member. Again I can just use this macro
> directly.
>
>>
>>>
>>> return 0;
>>> err_out:
>>> @@ -367,7 +367,7 @@ static struct mvs_info *mvs_pci_alloc(struct pci_dev *pdev,
>>> mvi->sas = sha;
>>> mvi->shost = shost;
>>>
>>> - mvi->tags = kzalloc(MVS_CHIP_SLOT_SZ>>3, GFP_KERNEL);
>>> + mvi->tags = kzalloc(MVS_RSVD_SLOTS, GFP_KERNEL);
>>
>> Field name ? reserved_tags ?
>> Also, the alloc seems wrong. This will allocate 4 bytes, but you only
>> need 4 bits. You could make this an unsigned long and not allocate
>> anything.
>
> Well spotted. I should have questioned more why they had >>3 previously.
>
> But I would rather keep as a bitmap, i.e. *unsigned long for simplicity.>
>> Same remark for pm8001 by the way.
>
> I think it's ok as it uses bitmap_zalloc()

Yes !

>
>>
>> That would cap MVS_RSVD_SLOTS to BITS_PER_LONG maximum, but that is easy
>> to check at compile time with a #if/#error.
>>
>
> As above, I'd rather keep as a bitmap. It's a little inefficient, but is
> a one off in the driver.
>
> Thanks,
> John
>
>

--
Damien Le Moal
Western Digital Research