Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [RFC PATCH v5 0/9] Add Tegra driver for video capture
To:     Sowjanya Komatineni <skomatineni@nvidia.com>,
        thierry.reding@gmail.com, jonathanh@nvidia.com, frankc@nvidia.com,
        helen.koike@collabora.com
Cc:     digetx@gmail.com, sboyd@kernel.org, linux-media@vger.kernel.org,
        devicetree@vger.kernel.org, linux-clk@vger.kernel.org,
        linux-tegra@vger.kernel.org, linux-kernel@vger.kernel.org
References: <1584985955-19101-1-git-send-email-skomatineni@nvidia.com>
 <4bb6a3b8-3332-014b-e763-bce9076179dd@xs4all.nl>
 <5ca1583a-889e-abd0-f823-eab93f09a365@xs4all.nl>
 <28ab0071-2e04-d14b-9215-db421e71b6af@nvidia.com>
 <a8e6b937-de08-ce50-9eca-3b3a5ad2916c@nvidia.com>
From:   Hans Verkuil <hverkuil@xs4all.nl>
Message-ID: <2aa2f57c-57a3-ff43-48d9-6e1e5660cdf3@xs4all.nl>
Date:   Fri, 3 Apr 2020 09:19:37 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.5.0
MIME-Version: 1.0
In-Reply-To: <a8e6b937-de08-ce50-9eca-3b3a5ad2916c@nvidia.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 4/3/20 7:45 AM, Sowjanya Komatineni wrote:
> 
> On 3/30/20 9:16 AM, Sowjanya Komatineni wrote:
>>
>> On 3/30/20 4:02 AM, Hans Verkuil wrote:
>>> External email: Use caution opening links or attachments
>>>
>>>
>>> On 3/30/20 12:04 PM, Hans Verkuil wrote:
>>>> Hi Sowjanya,
>>>>
>>>> On 3/23/20 6:52 PM, Sowjanya Komatineni wrote:
>>>>> This series adds Tegra210 VI and CSI driver for built-in test pattern
>>>>> generator (TPG) capture.
>>>>>
>>>>> Tegra210 supports max 6 channels on VI and 6 ports on CSI where each
>>>>> CSI port is one-to-one mapped to VI channel for video capture.
>>>>>
>>>>> This series has TPG support only where it creates hard media links
>>>>> between CSI subdevice and VI video device without device graphs.
>>>>>
>>>>> v4l2-compliance results are available below the patch diff.
>>>>>
>>>>> [v5]:        Includes,
>>>>>       - v4 feedback
>>>>>       - fix for venc powergate mc reset order.
>>>>>       - fix to have unbind and bind work during v4l2-ctl sleep and 
>>>>> streaming.
>>>> Unfortunately, I still crash on this.
>>>>
>>>> I do the following:
>>>>
>>>> Run: v4l2-ctl --stream-mmap
>>>>
>>>> Then, from another shell as root:
>>>>
>>>> cd /sys/devices/platform/50000000.host1x/tegra-video/driver
>>>> echo -n tegra-video > unbind
>>>>
>>>> I get this crash:
>>>>
>>>> [  315.691971] Unable to handle kernel NULL pointer dereference at 
>>>> virtual address 00000000000000b0
>>>> [  315.700749] Mem abort info:
>>>> [  315.703536]   ESR = 0x96000004
>>>> [  315.706587]   EC = 0x25: DABT (current EL), IL = 32 bits
>>>> [  315.711886]   SET = 0, FnV = 0
>>>> [  315.714933]   EA = 0, S1PTW = 0
>>>> [  315.718064] Data abort info:
>>>> [  315.720936]   ISV = 0, ISS = 0x00000004
>>>> [  315.724763]   CM = 0, WnR = 0
>>>> [  315.727726] user pgtable: 4k pages, 48-bit VAs, 
>>>> pgdp=0000000178ee8000
>>>> [  315.734152] [00000000000000b0] pgd=0000000000000000
>>>> [  315.739024] Internal error: Oops: 96000004 [#1] PREEMPT SMP
>>>> [  315.744584] Modules linked in: r8152 nouveau lp855x_bl tegra_drm ttm
>>>> [  315.750942] CPU: 3 PID: 2206 Comm: bash Tainted: G W         
>>>> 5.6.0-rc1-arm64 #118
>>>> [  315.759017] Hardware name: NVIDIA Jetson TX1 Developer Kit (DT)
>>>> [  315.764927] pstate: 20000085 (nzCv daIf -PAN -UAO)
>>>> [  315.769718] pc : _raw_write_lock_irqsave+0xb0/0x2b8
>>>> [  315.774590] lr : ida_free+0x48/0x158
>>>> [  315.778155] sp : ffff800011d8bba0
>>>> [  315.781462] x29: ffff800011d8bba0 x28: ffff0000f4095400
>>>> [  315.786766] x27: 0000000000000000 x26: 0000000000000000
>>>> [  315.792070] x25: 0000000000000000 x24: 0000000000000000
>>>> [  315.797372] x23: ffff0000f21ad400 x22: ffff0000f5c93000
>>>> [  315.802674] x21: ffff0000f4095400 x20: ffff0000f86b5540
>>>> [  315.807975] x19: 0000000000000000 x18: 0000000000000000
>>>> [  315.813276] x17: 0000000000000001 x16: 0000000000000019
>>>> [  315.818578] x15: 000000148ccdabe2 x14: 0000000000000136
>>>> [  315.823879] x13: 0000000000000001 x12: 00000000000003f8
>>>> [  315.829180] x11: 0000000000000000 x10: 0000000000000000
>>>> [  315.834482] x9 : ffff0000ff899990 x8 : ffff0000ff899000
>>>> [  315.839784] x7 : 0000000040000000 x6 : 0000000000210d00
>>>> [  315.845085] x5 : 0000000000000001 x4 : 0000000000000000
>>>> [  315.850386] x3 : 00000000000000b0 x2 : 0000000000000001
>>>> [  315.855687] x1 : 0000000000000000 x0 : 0000000000000001
>>>> [  315.860988] Call trace:
>>>> [  315.863432]  _raw_write_lock_irqsave+0xb0/0x2b8
>>>> [  315.867956]  ida_free+0x48/0x158
>>>> [  315.871184]  __media_device_unregister_entity+0x28/0xf0
>>>> [  315.876402]  media_device_unregister+0x6c/0x148
>>>> [  315.880927]  host1x_video_remove+0x20/0x48
>>>> [  315.885021]  host1x_device_remove+0x1c/0x30
>>>> [  315.889198]  device_release_driver_internal+0xf4/0x1c0
>>>> [  315.894325]  device_driver_detach+0x14/0x20
>>>> [  315.898503]  unbind_store+0xd4/0xf8
>>>> [  315.901986]  drv_attr_store+0x20/0x30
>>>> [  315.905645]  sysfs_kf_write+0x40/0x50
>>>> [  315.909301]  kernfs_fop_write+0xf8/0x210
>>>> [  315.913219]  __vfs_write+0x18/0x40
>>>> [  315.916616]  vfs_write+0xdc/0x1c8
>>>> [  315.919926]  ksys_write+0x68/0xf0
>>>> [  315.923235]  __arm64_sys_write+0x18/0x20
>>>> [  315.927154]  el0_svc_common.constprop.0+0x68/0x160
>>>> [  315.931936]  do_el0_svc+0x20/0x80
>>>> [  315.935246]  el0_sync_handler+0x10c/0x180
>>>> [  315.939246]  el0_sync+0x140/0x180
>>>> [  315.942560] Code: 8803fc02 35ffffa3 17fffda6 f9800071 (885ffc60)
>>>> [  315.948644] ---[ end trace e42b943f3c1af06c ]---
>>>>
>>>> The following diff fixes this:
>>>>
>>>> ------------------ cut here ------------------
>>>> diff --git a/drivers/staging/media/tegra/tegra-vi.c 
>>>> b/drivers/staging/media/tegra/tegra-vi.c
>>>> index 9714152aa6a7..53cf37af9602 100644
>>>> --- a/drivers/staging/media/tegra/tegra-vi.c
>>>> +++ b/drivers/staging/media/tegra/tegra-vi.c
>>>> @@ -583,7 +583,7 @@ static int tegra_channel_init(struct 
>>>> tegra_vi_channel *chan)
>>>>        /* initialize the video_device */
>>>>        chan->video->fops = &tegra_channel_fops;
>>>>        chan->video->v4l2_dev = &vid->v4l2_dev;
>>>> -     chan->video->release = video_device_release_empty;
>>>> +     chan->video->release = video_device_release;
>>>>        chan->video->queue = &chan->queue;
>>>>        snprintf(chan->video->name, sizeof(chan->video->name), 
>>>> "%s-%s-%u",
>>>>                 dev_name(vi->dev), "output", chan->portno);
>>>> @@ -647,6 +647,7 @@ static int tegra_channel_init(struct 
>>>> tegra_vi_channel *chan)
>>>>        media_entity_cleanup(&chan->video->entity);
>>>>   release_vdev:
>>>>        video_device_release(chan->video);
>>>> +     chan->video = NULL;
>>>>        return ret;
>>>>   }
>>>>
>>>> @@ -707,7 +708,6 @@ static void tegra_vi_channels_cleanup(struct 
>>>> tegra_vi *vi)
>>>>                        mutex_lock(&chan->video_lock);
>>>>                        vb2_queue_release(&chan->queue);
>>>>                        mutex_unlock(&chan->video_lock);
>>>> -                     video_device_release(chan->video);
>>>>                }
>>>>
>>>>                if (chan->frame_start_sp)
>>>> ------------------ cut here ------------------
>>> Note: Sakari suggested to embed struct video_device into struct 
>>> tegra_vi_channel.
>>> In that case chan->video->release should remain 
>>> video_device_release_empty and
>>> all video_device_alloc()/release() calls would have to be dropped.
>>
>> Thanks Hans. Tried several unbind/unbind not sure why it did not repro 
>> during my testing.
>>
>> video device is also part of tegra_vi_channel. So, v6 will remove 
>> video_device_alloc and use video_device_release_empty like I had in v3.
>>
>> This should help fix crash during unbind.
>>
>>>
>>> Regards,
>>>
>>>          Hans
> 
> With video device being part of the channel structure, it gets allocated 
> along with vi channel allocation during host1x vi client init and is 
> removed during channel free which happens during host1x vi client exit.
> 
> Wit this during driver unbind with video node kept opened with v4l2-ctl 
> sleep, it crashes during media_device_unregister_entity.
> 
> Using, separate video device allocation and freeing allocation during 
> video device release callback,  driver unbind works  as by the time of 
> entity unregister its it still available.
> 
> So, will keep allocating video device separately with video_device_alloc 
> and will use video_device_release call back in v6.
> 

Sounds good. Add a comment before video_device_alloc() explaining why it is
used instead of embedding it in the vi channel data structure.

Regards,

	Hans