Re: [PATCH v5 00/14] last set for add_disk() error handling

[Jens Axboe]

On 11/4/21 11:07 AM, Luis Chamberlain wrote:
> On Thu, Nov 04, 2021 at 06:53:34AM -0600, Jens Axboe wrote:
>> On 11/4/21 5:49 AM, Jens Axboe wrote:
>>> On Wed, 3 Nov 2021 16:04:23 -0700, Luis Chamberlain wrote:
>>>> Jens,
>>>>
>>>> as requested, I've folded all pending changes into this series. This
>>>> v5 pegs on Christoph's reviewed-by tags and since I was respinning I
>>>> modified the ataprobe and floppy driver changes as he suggested.
>>>>
>>>> I think this is it. The world of floppy has been exciting for v5.16.
>>>>
>>>> [...]
>>>
>>> Applied, thanks!
>>>
>>> [01/14] nvdimm/btt: use goto error labels on btt_blk_init()
>>>         commit: 2762ff06aa49e3a13fb4b779120f4f8c12c39fd1
>>> [02/14] nvdimm/btt: add error handling support for add_disk()
>>>         commit: 16be7974ff5d0a5cd9f345571c3eac1c3f6ba6de
>>> [03/14] nvdimm/blk: avoid calling del_gendisk() on early failures
>>>         commit: b7421afcec0c77ab58633587ddc29d53e6eb95af
>>> [04/14] nvdimm/blk: add error handling support for add_disk()
>>>         commit: dc104f4bb2d0a652dee010e47bc89c1ad2ab37c9
>>> [05/14] nvdimm/pmem: cleanup the disk if pmem_release_disk() is yet assigned
>>>         commit: accf58afb689f81daadde24080ea1164ad2db75f
>>> [06/14] nvdimm/pmem: use add_disk() error handling
>>>         commit: 5a192ccc32e2981f721343c750b8cfb4c3f41007
>>> [07/14] z2ram: add error handling support for add_disk()
>>>         commit: 15733754ccf35c49d2f36a7ac51adc8b975c1c78
>>> [08/14] block/sunvdc: add error handling support for add_disk()
>>>         commit: f583eaef0af39b792d74e39721b5ba4b6948a270
>>> [09/14] mtd/ubi/block: add error handling support for add_disk()
>>>         commit: ed73919124b2e48490adbbe48ffe885a2a4c6fee
>>> [10/14] ataflop: remove ataflop_probe_lock mutex
>>>         commit: 4ddb85d36613c45bde00d368bf9f357bd0708a0c
>>> [11/14] block: update __register_blkdev() probe documentation
>>>         commit: 26e06f5b13671d194d67ae8e2b66f524ab174153
>>> [12/14] ataflop: address add_disk() error handling on probe
>>>         commit: 46a7db492e7a27408bc164cbe6424683e79529b0
>>> [13/14] floppy: address add_disk() error handling on probe
>>>         commit: ec28fcc6cfcd418d20038ad2c492e87bf3a9f026
>>> [14/14] block: add __must_check for *add_disk*() callers
>>>         commit: 1698712d85ec2f128fc7e7c5dc2018b5ed2b7cf6
>>
>> rivers/scsi/sd.c: In function ‘sd_probe’:
>> drivers/scsi/sd.c:3573:9: warning: ignoring return value of ‘device_add_disk’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
>>  3573 |         device_add_disk(dev, gd, NULL);
>>       |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> drivers/scsi/sr.c: In function ‘sr_probe’:
>> drivers/scsi/sr.c:731:9: warning: ignoring return value of ‘device_add_disk’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
>>   731 |         device_add_disk(&sdev->sdev_gendev, disk, NULL);
>>       |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>
>>
>> Dropping the last two patches...
> 
> Martin K Peterson has the respective patches needed queued up on his tree
> for v5.16:
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/mkp/scsi.git/commit/?h=5.16/scsi-staging&id=e9d658c2175b95a8f091b12ddefb271683aeacd9
> https://git.kernel.org/pub/scm/linux/kernel/git/mkp/scsi.git/commit/?h=5.16/scsi-staging&id=2a7a891f4c406822801ecd676b076c64de072c9e
> 
> Would the last patch be sent once that gets to Linus?

But that dependency wasn't clear in the patches posted, and it leaves me
wondering if there are others? I obviously can't queue up a patch that
adds a must_check to a function, when we still have callers that don't
properly check it.

That should have been made clear, and that last patch never should've
been part of the series. Please send it once Linus's tree has all
callers checking the result.

> Also curious why drop the last two patches instead just the last one for
> now?

Sorry, meant just the last one.

-- 
Jens Axboe