[dm-devel] [EXT] Re: a potential issue for bow_target after "dm: delay registering the gendisk" applied

Faqiang Zhu faqiang.zhu at nxp.com
Tue Feb 22 03:06:07 UTC 2022


Thank you. I just noticed that this is an out-of-tree module.

+ Paul Lawrence & Eric Biggers

Best Regards,
Zhu Faqiang.
-----Original Message-----
From: Mike Snitzer <snitzer at redhat.com> 
Sent: 2022年2月22日 0:43
To: Faqiang Zhu <faqiang.zhu at nxp.com>
Cc: hch at lst.de; agk at redhat.com; dm-devel at redhat.com
Subject: [EXT] Re: a potential issue for bow_target after "dm: delay
registering the gendisk" applied

Caution: EXT Email

On Mon, Feb 21 2022 at  5:36P -0500,
Faqiang Zhu <faqiang.zhu at nxp.com> wrote:

> Hi Christoph,
>
>
>
> Recently when I use kernel5.15 with android, an issue related to
bow_target
> is found with below log:
>
>
>
> [    8.935429][    T1] kobject_add_internal failed for bow (error: -2
> parent: (null))
>
>
>
> The call trace to the preceding log is as below:
>
> dm_ctl_ioctl
>
>  -> ctl_ioctl
>
>    -> lookup_ioctl    // returns the table_load function pointer
>
>       table_load
>
>         -> populate_table
>
>              -> dm_table_add_target
>
>                   -> dm_bow_ctr
>
>                        -> kobject_init_and_add
>
>                             -> kobject_add_varg
>
>                                  -> kobject_add_internal
>
>
>
>
>
> Compared to kernel5.10, then I found it's related to below patch. The
> gendisk, it's not registered yet, its kobject is not in sysfs, so its
child
> "bow" kobject cannot be added to the kernel.
>
> commit 89f871af1b26d98d983cba7ed0e86effa45ba5f8
>
> Author: Christoph Hellwig  <mailto:hch at lst.de> hch at lst.de
>
> Date:   Wed Aug 4 11:41:46 2021 +0200
>
>
>
>     dm: delay registering the gendisk
>
>
>
>
>
> Is this an issue? if it is, is there any patch to fix it? if it is not,
then
> what is the right way to use the bow_target in userspace?
>

It is an issue for this out-of-tree dm-bow target.  Likely needs to be
fixed by adding a new DM target operation to 'struct target_type' that
would get called by DM core at end of dm_setup_md_queue().

Or dm-bow discontinues using sysfs for whatever and convert to using
DM messages (with .message) and exposing associated state via .status

Mike

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 9555 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/dm-devel/attachments/20220222/0c4b4eb7/attachment.p7s>


More information about the dm-devel mailing list