UBIFS assert failed in ubifs_set_page_dirty at 1421

[hujianyang]

On 2014/10/20 19:52, Artem Bityutskiy wrote:
> On Thu, 2014-10-09 at 06:48 +0000, Jijiagang wrote:
>> Hi Artem and Adrian,
>>
>> We encountered a “UBIFS assert failed in ubifs_set_page_dirty at 1421
>> (pid 543)” problem when running MONKEY test in android platform.
> 
> Hi, yes, this looks like a bug,
> 
> could you please re-send this e-mail in plain test format, because the
> MTD mailing list traps HTML e-mails.
> 
> Thanks!
> 

Hi Artem,

My colleagues find me and complain about this problem in September. The
previous bug I fixed, commit 691a7c6f, didn't fix the problem they
tolerate now. It has a different dump stack.

I don't have enough board or enough understanding to recreate this failure.

Do you think it is related?

line 1160, func try_to_unmap_one() in mm/rmap.c

""
        /* Move the dirty bit to the physical page now the pte is gone. */
        if (pte_dirty(pteval))
                set_page_dirty(page);
""

Is this pte_dirty() cause this error?

The recent patch you send, [PATCH] UBIFS: fix budget leak in error path,
seems not to fix this problem. They've tested this patch and find the
problem still exist.

Do you have any suggestions about this? I was busy these days so maybe
I could start to deal with this problem in November.

I should say, I don't know if it is a real problem caused by kernel. Maybe
this problem is caused by other changes outside kernel.

Any help is welcomed~!

Hu