[PATCH] alauda: do not use stack for URB transfer_buffers

[Brian Norris]

On Wed, Aug 21, 2013 at 2:59 AM, Jussi Kivilinna <jussi.kivilinna at iki.fi> wrote:
> On 21.08.2013 11:41, Brian Norris wrote:
>> On Wed, Aug 21, 2013 at 09:59:27AM +0200, Bjørn Mork wrote:
>>> Brian Norris <computersforpeace at gmail.com> writes:
>>>> On Tue, Aug 06, 2013 at 03:03:29PM +0300, Jussi Kivilinna wrote:
>>>>> Patch fixes alauda not to use stack as URB transfer_buffer. URB buffers need to
>>>>> be DMA-able, which stack is not.
>>>>>
>>>>> Patch is only compile tested.
>>>>>
>>>>> Cc: stable at vger.kernel.org
>>>>> Signed-off-by: Jussi Kivilinna <jussi.kivilinna at iki.fi>
>>>>> ---
>>>>>  drivers/mtd/nand/alauda.c |   74 ++++++++++++++++++++++++++++++++++-----------
>>>>>  1 file changed, 56 insertions(+), 18 deletions(-)
>>>>
>>>> Just FYI, this driver is being removed, so I'm obviously not taking this
>>>> patch :)
>>>
>>> I think you should apply it anyway. The driver is still in v3.11 AFAICS,
>>> and the patch should also go to the maintained stable kernels. You
>>> cannot remove the driver from them, and I don't see a later driver
>>> removal as a valid reason not to fix a known bug with a patch.
>>
>> Seriously?
>>
>> The reasons given for removal:
>>
>>   "The driver has very low utility.  Devices in question are limited to
>>   about 400kB/s and the only known user (me) discarded the hardware
>>   several years back."
>>
>> And:
>>
>>   "Maybe we should just remove the driver and not spend any more time on
>>   it?"
>>
>> So you're suggesting applying an untested (compile-only) fix for an
>> unobserved bug for the theoretical user of old, slow hardware who wants
>> to use a recent stable kernel, when the last known user has given up on
>> the driver entirely?
>>
>> Anyway, just because you complained, I rebased and applied this to
>> l2-mtd.git before the driver removal. I'll defer to dwmw2 whether this
>> patch gets squashed out of existence.
>
> If this is the case, please ignore this patch. I should not have marked it for stable in first place, since it is just compile-tested.

Yes, that's a good point. Quoting Documentation/stable_kernel_rules.txt:

"It must be obviously correct and tested."

Seeing as it was not tested, I am dropping the patch entirely (it is
not stable material, and there is no point including it along with the
driver removal).

Thanks,
Brian