linux-scsi.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Damien Le Moal <damien.lemoal@opensource.wdc.com>
To: Bart Van Assche <bvanassche@acm.org>,
	"linux-scsi @ vger . kernel . org" <linux-scsi@vger.kernel.org>,
	"Martin K . Petersen" <martin.petersen@oracle.com>,
	"linux-ide@vger.kernel.org" <linux-ide@vger.kernel.org>
Subject: Re: Problem with scsi device sysfs attributes
Date: Thu, 11 Nov 2021 17:46:40 +0900	[thread overview]
Message-ID: <9c20c383-90fe-186c-86d4-e2b381dfbd8c@opensource.wdc.com> (raw)
In-Reply-To: <93498302-c4c6-de7f-f177-e55de2828db0@opensource.wdc.com>

On 11/11/21 14:27, Damien Le Moal wrote:
> On 11/11/21 12:31, Damien Le Moal wrote:
>> Bart,
>>
>> Your patch 92c4b58b15c5 ("scsi: core: Register sysfs attributes earlier")
>> modified the scsi device sysfs attributes initialization to use the scsi
>> template shost_groups and sdev_groups for adding attributes using groups instead
>> of arrays of attrs. However, this patch completely removed the
>> sysfs_create_groups() call to actually create the attributes listed in the groups.
>>
>> As a result, I see many missing sysfs device attributes for at least ahci (e.g.
>> ncq_prio_enavle, ncq_prio_supported), but I suspect other device types may have
>> similar problems.
>>
>> I do not see where the attribute groups in the arrray sdev->gendev_attr_groups
>> are registered with sysfs. In fact, it looks like sdev->gendev_attr_groups is
>> referenced only in scsi_sysfs.c but only for initializing it. It is never used
>> to actually register the attr groups... Am I missing something ?
>>
>> This is at least breaking NCQ priority support right now. Did your patch
>> 92c4b58b15c5 remove too much code ? Shouldn't we have a call to
>> sysfs_create_groups() somewhere ? I think that should be in
>> scsi_sysfs_device_initialize() but I am not 100% sure.
> 
> Adding this:
> 
> diff --git a/drivers/scsi/scsi_sysfs.c b/drivers/scsi/scsi_sysfs.c
> index d3d362289ecc..a1a30af96e17 100644
> --- a/drivers/scsi/scsi_sysfs.c
> +++ b/drivers/scsi/scsi_sysfs.c
> @@ -1339,6 +1339,8 @@ int scsi_sysfs_add_sdev(struct scsi_device *sdev)
>  {
>         int error;
>         struct scsi_target *starget = sdev->sdev_target;
> +       struct Scsi_Host *shost = sdev->host;
> +       struct scsi_host_template *hostt = shost->hostt;
> 
>         error = scsi_target_add(starget);
>         if (error)
> @@ -1365,6 +1367,17 @@ int scsi_sysfs_add_sdev(struct scsi_device *sdev)
>                 return error;
>         }
> 
> +       if (hostt->sdev_groups) {
> +               error = sysfs_create_groups(&sdev->sdev_gendev.kobj,
> +                                           hostt->sdev_groups);
> +               if (error) {
> +                       sdev_printk(KERN_INFO, sdev,
> +                                   "failed to create device attributes:
> %d\n",
> +                                   error);
> +                       return error;
> +               }
> +       }
> +
>         device_enable_async_suspend(&sdev->sdev_dev);
>         error = device_add(&sdev->sdev_dev);
>         if (error) {
> 
> I can see the AHCI NCQ attributes are visible under
> /sysfs/block/sdX/device/.
> 
> So it looks like the LLD attribute groups added to
> sdev->gendev_attr_groups by scsi_sysfs_device_initialize() using
> hostt->sdev_groups are never registered with sysfs, but the default
> common scsi attributes defined by scsi_sdev_attr_group are registered...
> This is very weird.
> 
> I do not understand why. I cannot find where sdev->gendev_attr_groups is
> used to create the attributes in sysfs...

I figured it out: scsi_sdev_attr_groups is set in scsi_dev_type (struct
device_type) and so the attributes defined in this group are created in
sysfs when the scis device sdev_gendev is added. However, the additional
attribute groups in sdev->gendev_attr_groups added by the LLD are never
processed since scsi_dev_type->groups is scsi_sdev_attr_groups and not
sdev->gendev_attr_groups.

I sent a patch. Please let me know what you think.


-- 
Damien Le Moal
Western Digital Research

      reply	other threads:[~2021-11-11  8:46 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-11  3:31 Problem with scsi device sysfs attributes Damien Le Moal
2021-11-11  5:27 ` Damien Le Moal
2021-11-11  8:46   ` Damien Le Moal [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=9c20c383-90fe-186c-86d4-e2b381dfbd8c@opensource.wdc.com \
    --to=damien.lemoal@opensource.wdc.com \
    --cc=bvanassche@acm.org \
    --cc=linux-ide@vger.kernel.org \
    --cc=linux-scsi@vger.kernel.org \
    --cc=martin.petersen@oracle.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).