UBIFS assert failed in ubifs_set_page_dirty at 1421

[Jijiagang]

Hi Kirill,
I test your patch, bus there's no dump_vma definition.
The log is here, hope it will be helpful.

 page:817fd7e0 count:3 mapcount:1 mapping:a318bb8c index:0x4
 page flags: 0xa19(locked|uptodate|dirty|arch_1|private)
 pte_write: 1
 page:81441a80 count:3 mapcount:1 mapping:a318bb8c index:0x5
 page flags: 0x209(locked|uptodate|arch_1)
 pte_write: 1
 UBIFS assert failed in ubifs_set_page_dirty at 1422 (pid 545)
 CPU: 2 PID: 545 Comm: kswapd0 Tainted: P           O 3.10.0_s40 #19
 [<8001d8a0>] (unwind_backtrace+0x0/0x108) from [<80019f44>] (show_stack+0x20/0x24)
 [<80019f44>] (show_stack+0x20/0x24) from [<80acfa18>] (dump_stack+0x24/0x2c)
 [<80acfa18>] (dump_stack+0x24/0x2c) from [<8029766c>] (ubifs_set_page_dirty+0x54/0x5c)
 [<8029766c>] (ubifs_set_page_dirty+0x54/0x5c) from [<800cea60>] (set_page_dirty+0x50/0x78)
 [<800cea60>] (set_page_dirty+0x50/0x78) from [<800f4b10>] (try_to_unmap_one+0x124/0x410)
 [<800f4b10>] (try_to_unmap_one+0x124/0x410) from [<800f4f84>] (try_to_unmap_file+0x9c/0x740)
 [<800f4f84>] (try_to_unmap_file+0x9c/0x740) from [<800f56b8>] (try_to_unmap+0x40/0x78)
 [<800f56b8>] (try_to_unmap+0x40/0x78) from [<800d6a04>] (shrink_page_list+0x23c/0x884)
 [<800d6a04>] (shrink_page_list+0x23c/0x884) from [<800d76c8>] (shrink_inactive_list+0x21c/0x3c8)
 [<800d76c8>] (shrink_inactive_list+0x21c/0x3c8) from [<800d7c20>] (shrink_lruvec+0x3ac/0x524)
 [<800d7c20>] (shrink_lruvec+0x3ac/0x524) from [<800d8970>] (kswapd+0x854/0xdc0)
 [<800d8970>] (kswapd+0x854/0xdc0) from [<80051e28>] (kthread+0xc8/0xcc)
 [<80051e28>] (kthread+0xc8/0xcc) from [<80015198>] (ret_from_fork+0x14/0x20)
 UBIFS assert failed in ubifs_release_budget at 567 (pid 6)
 CPU: 3 PID: 6 Comm: kworker/u8:0 Tainted: P           O 3.10.0_s40 #19

> -----Original Message-----
> From: Kirill A. Shutemov [mailto:kirill at shutemov.name]
> Sent: Monday, November 24, 2014 5:10 PM
> To: Jijiagang
> Cc: Hujianyang; dedekind1 at gmail.com; Caizhiyong;
> linux-fsdevel at vger.kernel.org; linux-mm at kvack.org; Wanli (welly);
> linux-mtd at lists.infradead.org; adrian.hunter at intel.com
> Subject: Re: UBIFS assert failed in ubifs_set_page_dirty at 1421
> 
> On Mon, Nov 24, 2014 at 02:59:51AM +0000, Jijiagang wrote:
> > Hi Kirill,
> >
> > I add dump_page(page) in function ubifs_set_page_dirty.
> > And get this log when ubifs assert fail. Is it helpful for this problem?
> 
> Not really. It seems you called dump_page() after
> __set_page_dirty_nobuffers() in ubifs_set_page_dirty().
> 
> Could you try something like patch below. It assumes ubifs to compiled in (not
> module).
> 
> diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c index b5b593c45270..7b4386dd174e
> 100644
> --- a/fs/ubifs/file.c
> +++ b/fs/ubifs/file.c
> @@ -1531,7 +1531,7 @@ out_unlock:
>         return err;
>  }
> 
> -static const struct vm_operations_struct ubifs_file_vm_ops = {
> +const struct vm_operations_struct ubifs_file_vm_ops = {
>         .fault        = filemap_fault,
>         .map_pages = filemap_map_pages,
>         .page_mkwrite = ubifs_vm_page_mkwrite, diff --git a/mm/rmap.c
> b/mm/rmap.c index 19886fb2f13a..343c4571df68 100644
> --- a/mm/rmap.c
> +++ b/mm/rmap.c
> @@ -1171,8 +1171,15 @@ static int try_to_unmap_one(struct page *page,
> struct vm_area_struct *vma,
>         pteval = ptep_clear_flush(vma, address, pte);
> 
>         /* Move the dirty bit to the physical page now the pte is gone. */
> -       if (pte_dirty(pteval))
> +       if (pte_dirty(pteval)) {
> +               extern const struct vm_operations_struct
> ubifs_file_vm_ops;
> +               if (vma->vm_ops == &ubifs_file_vm_ops) {
> +                       dump_vma(vma);
> +                       dump_page(page, __func__);
> +                       pr_emerg("pte_write: %d\n", pte_write(pteval));
> +               }
>                 set_page_dirty(page);
> +       }
> 
>         /* Update high watermark before we lower rss */
>         update_hiwater_rss(mm);
> --
>  Kirill A. Shutemov