Re: [PATCH v4] media: verisilicon: Fix kernel panic due to __initconst misuse

[Thorsten Leemhuis]

On 3/16/26 16:25, Nicolas Dufresne wrote:
> Le jeudi 12 mars 2026 à 18:34 +0100, Francesco Dolcini a écrit :
>> On Fri, Mar 06, 2026 at 11:10:57AM +0800, ming.qian@xxxxxxxxxxx wrote:
>>> From: Ming Qian <ming.qian@xxxxxxxxxxx>
>>>
>>> Fix a kernel panic when probing the driver as a module:
>>>
>>>   Unable to handle kernel paging request at virtual address
>>>   ffffd9c18eb05000
>>>   of_find_matching_node_and_match+0x5c/0x1a0
>>>   hantro_probe+0x2f4/0x7d0 [hantro_vpu]
>>>
>>> The imx8mq_vpu_shared_resources array is referenced by variant
>>> structures through their shared_devices field. When built as a
>>> module, __initconst causes this data to be freed after module
>>> init, but it's later accessed during probe, causing a page fault.
>>>
>>> The imx8mq_vpu_shared_resources is referenced from non-init code,
>>> so keeping __initconst or __initconst_or_module here is wrong.
>>>
>>> Drop the __initconst annotation and let it live in the normal .rodata
>>> section.
>>>
>>> A bug of __initconst called from regular non-init probe code
>>> leading to bugs during probe deferrals or during unbind-bind cycles.
>>>
>>> Reported-by: Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxxxxxxxx>
>>> Closes: https://lore.kernel.org/all/68ef934f-baa0-4bf6-93d8-834bbc441e66@xxxxxxxxxx/
>>> Reported-by: Franz Schnyder <franz.schnyder@xxxxxxxxxxx>
>>> Closes: https://lore.kernel.org/all/n3qmcb62tepxltoskpf7ws6yiirc2so62ia23b42rj3wlmpl67@rvkbuirx7kkp/
>>> Fixes: e0203ddf9af7 ("media: verisilicon: Avoid G2 bus error while decoding H.264 and HEVC")
>>> Suggested-by: Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxxxxxxxx>
>>> Suggested-by: Marco Felsch <m.felsch@xxxxxxxxxxxxxx>
>>> Reviewed-by: Marco Felsch <m.felsch@xxxxxxxxxxxxxx>
>>> Signed-off-by: Ming Qian <ming.qian@xxxxxxxxxxx>
>>
>> What's the plan to merge this? It fixes a quite severe regression,
>> a boot failure.
> 
> To be decided this week. The commit message does not say if it was released, or
> came in RCs (and I didn't check myself yet). I'd say, if its the first one, it
> will go through next and backports, otherwise its is really tight to get that
> into the RC series, but serious enough. Please fill the gap if you have time,
> and I'll handle it later, probably tomorrow.

Well, FYI: the culprit according to the fixes tag quoted above is
e0203ddf9af7c8 ("media: verisilicon: Avoid G2 bus error while decoding
H.264 and HEVC") [v7.0-rc1, v6.19.6 (286d629d10640b)]. Given that this was
* reported 11 days ago
* reported at least two times
* made it into a stable series
it is something that Linus would like to see fixed by now in mainline
afaics, as ""the rule of thumb should generally be "within a week",
preferably before the next rc."":
https://www.kernel.org/doc/html/next/process/handling-regressions.html#on-how-quickly-regressions-should-be-fixed

If that fix is too dangerous a revert is of course an option, too.

BTW, given that the culprit made it to a stable series, it would be a
wise to add a stable tag to ensure it's backported (it otherweise might
be silent dropped and/or applying might be deferred).

Ciao, Thorsten