[BUG] pxa3xx: wait time out when scanning for bb

Ezequiel Garcia ezequiel at vanguardiasur.com.ar
Sat Dec 9 15:04:02 PST 2017


Hi Sean,

On 28 November 2017 at 10:02, Miquel RAYNAL
<miquel.raynal at free-electrons.com> wrote:
> Hi Sean,
>
>> I'm currently in the process of doing bringup on a custom armada 38x
>> board. During the firsts boots of the Linux kernel we saw that the
>> NAND driver throws a lot of timeout messages.
>>
>> Like this:
>> [ 28.321054] pxa3xx-nand f10d0000.flash: Wait time out!!!
>> [ 28.521054] pxa3xx-nand f10d0000.flash: Wait time out!!!
>> [ 28.721054] pxa3xx-nand f10d0000.flash: Wait time out!!!
>
> Could you paste the entire dmesg please ?
>
> Also, what kernel are you using ?
>
>>
>> That lead me to investigate the issue and found this thread (quite
>> old): https://www.spinics.net/lists/arm-kernel/msg488581.html
>>
>> In the thread some is describing a possible race condition when the
>> NAND framework is scanning for bad blocks.
>
> Not sure this is still relevant, but, who knows.
>

As far as I can recall, there isn't any possible race condition.
The pxa3xx-nand driver should be stable.

See: https://www.spinics.net/lists/arm-kernel/msg476955.html,
where I explain this.

The timeout described by that thread was caused by not using
CONFIG_PREEMPT. Just like I have described on that thread,
not using a preemptable kernel increases latency, causing
issues on drivers that are not prepared for it.

You can find the analysis of that particular timeout here:

https://www.spinics.net/lists/arm-kernel/msg487956.html

BTW, you need a very good, a *very* good, reason
not to use CONFIG_PREEMPT.
-- 
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar



More information about the linux-mtd mailing list