Re: [PATCH] wifi: ath9k: Don't mark channelmap stack variable read-only in ath9k_mci_update_wlan_channels()

["Toke Høiland-Jørgensen" <toke@toke.dk>]

Kalle Valo <kvalo@kernel.org> writes:

> (dropping stable from cc)
>
> Toke Høiland-Jørgensen <toke@toke.dk> writes:
>
>> Kalle Valo <kvalo@kernel.org> writes:
>>
>>> Toke Høiland-Jørgensen <toke@toke.dk> writes:
>>>
>>>> This partially reverts commit e161d4b60ae3a5356e07202e0bfedb5fad82c6aa.
>>>>
>>>> Turns out the channelmap variable is not actually read-only, it's modified
>>>> through the MCI_GPM_CLR_CHANNEL_BIT() macro further down in the function,
>>>> so making it read-only causes page faults when that code is hit.
>>>>
>>>> Link: https://bugzilla.kernel.org/show_bug.cgi?id=217183
>>>> Fixes: e161d4b60ae3 ("wifi: ath9k: Make arrays prof_prio and
>>>> channelmap static const")
>>>> Cc: stable@vger.kernel.org
>>>> Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
>>>
>>> I guess the casting in MCI_GPM_CLR_CHANNEL_BIT() hide this and made it
>>> impossible for the compiler to detect it? A perfect example why I hate
>>> casting :)
>>
>> Yup, exactly. I was also assuming the compiler would catch it, but yay, C! :/
>
> We have so many static checkers that I wonder if those would be able to
> catch these kind of buggy casts? We had a similar bug in rtw89 something
> like a year ago.

No idea. Would be nice, yeah... :)

>> Anyway, cf the bugzilla this was a pretty bad regression for 6.2, so
>> would be good to move this along reasonably quickly (although I guess we
>> just missed the -net PR for rc7)...
>
> I'm not planning to send anymore stuff to v6.3 so my plan is to take
> this to -next. The merge window is very close anyway so this shouldn't
> cause too much delay.

Hmm, okay, a bit unfortunate that we'll ship 6.3 with the same bug, but
if it goes in during the merge window, I guess we'll get the fix into
6.3.1 (or something close to that) via stable? I can live with that...

-Toke