by Hans Verkuil

[permalink] [raw]

Subject: Re: [PATCH v10 0/8] Add StarFive Camera Subsystem driver

On 18/10/2023 11:56, Hans Verkuil wrote:
> On 18/10/2023 11:52, Jack Zhu wrote:
>>
>>
>> On 2023/10/18 17:31, Hans Verkuil wrote:
>>> On 18/10/2023 11:25, Jack Zhu wrote:
>>>>
>>>>
>>>> On 2023/10/18 16:50, Hans Verkuil wrote:
>>>>> Hi Jack,
>>>>>
>>>>> On 18/10/2023 04:37, Jack Zhu wrote:
>>>>>
>>>>> <snip>
>>>>>
>>>>>>>> --------------------------------------------------------------------------------
>>>>>>>> Compliance test for device /dev/v4l-subdev1:
>>>>>>>>
>>>>>>>> Driver Info:
>>>>>>>> Driver version : 6.6.0
>>>>>>>> Capabilities : 0x00000000
>>>>>>>
>>>>>>> But this does not appear for v4l-subdev1.
>>>>>>>
>>>>>>> I can't really tell why it doesn't show that. Can you debug a little bit?
>>>>>>> The code is in v4l2-compliance.cpp, line 1086:
>>>>>>>
>>>>>>> ent_id = mi_media_info_for_fd(media_fd, node.g_fd(), &is_invalid, &node.function);
>>>>>>>
>>>>>>> The mi_media_info_for_fd() function calls ioctl(media_fd, MEDIA_IOC_DEVICE_INFO, &mdinfo),
>>>>>>> and that fails for some reason. It could be that media_fd is invalid (would be weird).
>>>>>>>
>>>>>>> This could well be a v4l2-compliance bug that you hit with this driver.
>>>>>>>
>>>>>>
>>>>>> On the test board, /dev/v4l-subdev1 is imx219, and the corresponding directory is
>>>>>> /sys/dev/char/81:3/device. Media0 does not exist in this directory. Therefore, the media_fd
>>>>>> obtained through mi_get_media_fd(node.g_fd(), node.bus_info) is invalid.
>>>>>>
>>>>>> I don't know why media0 does not exist in /sys/dev/char/81:3/device?
>>>>>>
>>>>>
>>>>> Can you try again with this v4l2-compliance patch?
>>>>>
>>>>> I need to dig a bit deeper as to why media0 is missing, but for now try this.
>>>>>
>>>>> Regards,
>>>>>
>>>>> Hans
>>>>>
>>>>> diff --git a/utils/v4l2-compliance/v4l2-compliance.cpp b/utils/v4l2-compliance/v4l2-compliance.cpp
>>>>> index 7169eefe..29475d6b 100644
>>>>> --- a/utils/v4l2-compliance/v4l2-compliance.cpp
>>>>> +++ b/utils/v4l2-compliance/v4l2-compliance.cpp
>>>>> @@ -968,7 +968,7 @@ err:
>>>>> }
>>>>>
>>>>> void testNode(struct node &node, struct node &node_m2m_cap, struct node &expbuf_node, media_type type,
>>>>> - unsigned frame_count, unsigned all_fmt_frame_count)
>>>>> + unsigned frame_count, unsigned all_fmt_frame_count, int parent_media_fd)
>>>>> {
>>>>> struct node node2;
>>>>> struct v4l2_capability vcap = {};
>>>>> @@ -997,8 +997,12 @@ void testNode(struct node &node, struct node &node_m2m_cap, struct node &expbuf_
>>>>> memset(&vcap, 0, sizeof(vcap));
>>>>> }
>>>>>
>>>>> - if (!node.is_media())
>>>>> - media_fd = mi_get_media_fd(node.g_fd(), node.bus_info);
>>>>> + if (!node.is_media()) {
>>>>> + if (parent_media_fd >= 0)
>>>>> + media_fd = parent_media_fd;
>>>>> + else
>>>>> + media_fd = mi_get_media_fd(node.g_fd(), node.bus_info);
>>>>> + }
>>>>>
>>>>> int fd = node.is_media() ? node.g_fd() : media_fd;
>>>>> if (fd >= 0) {
>>>>> diff --git a/utils/v4l2-compliance/v4l2-compliance.h b/utils/v4l2-compliance/v4l2-compliance.h
>>>>> index 7caf254b..c47f25f5 100644
>>>>> --- a/utils/v4l2-compliance/v4l2-compliance.h
>>>>> +++ b/utils/v4l2-compliance/v4l2-compliance.h
>>>>> @@ -308,7 +308,7 @@ int check_ustring(const __u8 *s, int len);
>>>>> int check_0(const void *p, int len);
>>>>> int restoreFormat(struct node *node);
>>>>> void testNode(struct node &node, struct node &node_m2m_cap, struct node &expbuf_node, media_type type,
>>>>> - unsigned frame_count, unsigned all_fmt_frame_count);
>>>>> + unsigned frame_count, unsigned all_fmt_frame_count, int parent_media_fd = -1);
>>>>> std::string stream_from(const std::string &pixelformat, bool &use_hdr);
>>>>>
>>>>> // Media Controller ioctl tests
>>>>>
>>>>
>>>> From the log, there is no change.
>>>
>>> Oops, my mistake. Also apply this change:
>>>
>>> diff --git a/utils/v4l2-compliance/v4l2-test-media.cpp b/utils/v4l2-compliance/v4l2-test-media.cpp
>>> index 0195ac58..52ab7fb8 100644
>>> --- a/utils/v4l2-compliance/v4l2-test-media.cpp
>>> +++ b/utils/v4l2-compliance/v4l2-test-media.cpp
>>> @@ -612,7 +612,7 @@ void walkTopology(struct node &node, struct node &expbuf_node,
>>> }
>>>
>>> testNode(test_node, test_node, expbuf_node, type,
>>> - frame_count, all_fmt_frame_count);
>>> + frame_count, all_fmt_frame_count, node.g_fd());
>>> test_node.close();
>>> }
>>> }
>>>
>>
>> Can see relevant Info in the log.
>
> Great! Can you do one more thing? Please run 'v4l2-compliance -m /dev/media0 --verbose'
> and mail the output to me. It's pretty big, so just email it to me, without CCs.
>
> I want to take a closer look at the output to see why this patch is needed.

Thank you for your help. This v4l2-compliance patch is in fact needed, and I have just
pushed the fix.

Regards,

Hans