linux-kernel - Re: [PATCHv4 01/11] videobuf2: add cache management members

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Sat, 7 Mar 2020 12:47:00 +0100
From:   Hans Verkuil <hverkuil@...all.nl>
To:     Sergey Senozhatsky <senozhatsky@...omium.org>
Cc:     Hans Verkuil <hans.verkuil@...co.com>,
        Tomasz Figa <tfiga@...omium.org>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        Kyungmin Park <kyungmin.park@...sung.com>,
        Marek Szyprowski <m.szyprowski@...sung.com>,
        Sakari Ailus <sakari.ailus@....fi>,
        Laurent Pinchart <laurent.pinchart@...asonboard.com>,
        Pawel Osciak <posciak@...omium.org>,
        linux-media@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCHv4 01/11] videobuf2: add cache management members

On 07/03/2020 12:28, Sergey Senozhatsky wrote:
> On (20/03/07 11:10), Hans Verkuil wrote:
> [..]
>>>>> @@ -564,6 +573,7 @@ struct vb2_queue {
>>>>>  	unsigned			requires_requests:1;
>>>>>  	unsigned			uses_qbuf:1;
>>>>>  	unsigned			uses_requests:1;
>>>>> +	unsigned			allow_cache_hints:1;
>>>
>>> Shall I use "unsigned int" here instead of "unsigned"?
>>
>> The vb2_queue bitfields are the only places in that header were 'unsigned' is
>> used. I think that that should be fixed in a separate patch. It's nice to have
>> it consistent.
>>
>> Put that patch in the beginning of the series, that way I can pick it up in the
>> next pull request.
> 
> OK, done.
> 
> For the time being the series has moved to github public repo [0],
> I'll try to run more 'twisty' cases and re-submit once it survives
> beating.

Create those tests in v4l2-compliance: that's where they belong.

You need these tests:

For non-MMAP modes:

1) test that V4L2_BUF_CAP_SUPPORTS_CACHE_HINTS is never set.

If V4L2_BUF_CAP_SUPPORTS_CACHE_HINTS is not set, then:

1) attempting to use V4L2_FLAG_MEMORY_NON_CONSISTENT will clear the flag
   upon return (test with both reqbufs and create_bufs).
2) attempting to use V4L2_BUF_FLAG_NO_CACHE_INVALIDATE or V4L2_BUF_FLAG_NO_CACHE_CLEAN
   will clear those flags upon return (do we actually do that in the patch series?).

If V4L2_BUF_CAP_SUPPORTS_CACHE_HINTS is set, then:

1) set V4L2_FLAG_MEMORY_NON_CONSISTENT in reqbufs, but clear in create_bufs:
   this should fail.
2) clear V4L2_FLAG_MEMORY_NON_CONSISTENT in reqbufs, but set in create_bufs:
   this should fail.
3) set V4L2_FLAG_MEMORY_NON_CONSISTENT in both reqbufs and create_bufs: this should
   work.
4) clear V4L2_FLAG_MEMORY_NON_CONSISTENT in both reqbufs and create_bufs: this should
   work.
5) you can use V4L2_BUF_FLAG_NO_CACHE_INVALIDATE or V4L2_BUF_FLAG_NO_CACHE_CLEAN
   without these flags being cleared in v4l2_buffer.

All these tests can be done in testReqBufs().

Regards,

	Hans

> 
> [0] https://github.com/sergey-senozhatsky/v4l2-mmap-cache-flags
> 
> 	-ss
>