Re: ping // [PATCH] mtd:cfi_cmdset_0002: fix atomic sleep bug when CONFIG_MTD_XIP=y

From: Vignesh Raghavendra <vigneshr@ti.com>
To: Xiaoming Ni <nixiaoming@huawei.com>,
	Miquel Raynal <miquel.raynal@bootlin.com>
Cc: wangle6@huawei.com, tudor.ambarus@microchip.com, richard@nod.at,
	linux-kernel@vger.kernel.org, stable@vger.kernel.org,
	linux-mtd@lists.infradead.org, gregkh@linuxfoundation.org,
	tglx@linutronix.de
Subject: Re: ping // [PATCH] mtd:cfi_cmdset_0002: fix atomic sleep bug when CONFIG_MTD_XIP=y
Date: Mon, 14 Dec 2020 12:46:50 +0530	[thread overview]
Message-ID: <f14ccb94-cb53-f495-e95c-fe657122f16c@ti.com> (raw)
In-Reply-To: <b5c52547-b3cc-4217-cc69-067cee7d536f@huawei.com>

On 12/8/20 6:53 AM, Xiaoming Ni wrote:
> On 2020/12/8 2:59, Vignesh Raghavendra wrote:
>> Hi Xiaoming,
>>
[...]
>>>>> ---
>>>>>    drivers/mtd/chips/cfi_cmdset_0002.c | 16 ++++++++++++++++
>>>>>    1 file changed, 16 insertions(+)
>>>>>
>>>>> diff --git a/drivers/mtd/chips/cfi_cmdset_0002.c
>>>>> b/drivers/mtd/chips/cfi_cmdset_0002.c
>>>>> index a1f3e1031c3d..12c3776f093a 100644
>>>>> --- a/drivers/mtd/chips/cfi_cmdset_0002.c
>>>>> +++ b/drivers/mtd/chips/cfi_cmdset_0002.c
>>>>> @@ -1682,7 +1682,11 @@ static int __xipram
>>>>> do_write_oneword_once(struct map_info *map,
>>>>>                set_current_state(TASK_UNINTERRUPTIBLE);
>>>>>                add_wait_queue(&chip->wq, &wait);
>>>>>                mutex_unlock(&chip->mutex);
>>>>> +            if (IS_ENABLED(CONFIG_MTD_XIP))
>>>>> +                local_irq_enable();
>>>>>                schedule();
>>>>> +            if (IS_ENABLED(CONFIG_MTD_XIP))
>>>>> +                local_irq_disable();
>>>
>>> The fix really seems strange to me. I will let Vignesh decide but I
>>> think we should consider updating/fixing xip_disable instead.
>>
>> Agree with Miquel. Have you done any testing
>> or is this purely based on code inspection?
>>
> I don't have the corresponding device test environment.
> I found the problem through code review.
> 

Sorry, I am not comfortable applying this patch without proper testing
that given the unknowns and legacy nature of the code.

> 
>> What about comment before xip_disable() function:
>>
>> /*
>>   * No interrupt what so ever can be serviced while the flash isn't in
>> array
>>   * mode.  This is ensured by the xip_disable() and xip_enable()
>> functions
>>   * enclosing any code path where the flash is known not to be in
>> array mode.
>>   * And within a XIP disabled code path, only functions marked with
>> __xipram
>>   * may be called and nothing else (it's a good thing to inspect
>> generated
>>   * assembly to make sure inline functions were actually inlined and
>> that gcc
>>   * didn't emit calls to its own support functions). Also configuring
>> MTD CFI
>>   * support to a single buswidth and a single interleave is also
>> recommended.
>>   */
>>
>> So, I don't think the fix is as simple as this patch.
>>
> +xip_enable();
>  schedule();
> +xip_disable();
> 
> Do I need to change it to this?
> 

This just narrows the window, but an IRQ is still possible just after
xip_enable() but before schedule().

Regards
Vignesh

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/