[RFC] Handling of errors for AMD NOR (cfi_cmdset_0002) chips

[Norbert van Bolhuis]

On 03/24/11 10:27, Detlev Zundel wrote:
...
>> But the unstable bit problem is different from software timeout
>> related issues regarding UBI and NOR flash the OP mentioned.
>> It is anyway good to hear work is in progress here.
>
> It is never good to ignore errors reported by the hardware if you are
> looking for problems in higher software layers.  That's why we decided
> to implement such error checks which can then be helpful in our further
> testing.  We are hoping that we can better differentiate between
> software and hardware problems this way.
>

ok, I see.

>> I asked about the software timeout issue because I've seen false
>> timeouts with powerpc + CONFIG_NO_HZ a while ago.
>> In my case the low-level do_write_buffer (cfi_cmdset_0002.c) timed
>> out too early. See
>> http://lkml.org/lkml/2009/9/3/84
>
> Ok, I see, thanks for the pointer.  The problems you experienced then
> were unneccessary software timeouts, but there was no data corruption,
> correct?
>

yes that is correct.
It happened only when system was 'stressed' a bit and it caused false
software timeouts which made UBIFS remount the fs in read-only mode.

---
N. van Bolhuis.