[PATCH v2] UBIFS: Fix assert failed in ubifs_set_page_dirty

[Dolev Raviv]

Hi Hujianyang and Laurence,
I'm hitting an assertion in very similar code during the callback
ubifs_releasepage(). I'm quite new to this code area, and currently diving
in to understanding the race you described (before I attempt to look at
the patch).

I was wondering if this assertion is related?

[  862.695278] UBIFS assert failed in ubifs_releasepage at 1434 (pid 11088)
[  862.700952] CPU: 0 PID: 11088 Comm: busybox Tainted: G        W   
3.10.0+ #1
[  862.714460] [<c00147f4>] (unwind_backtrace+0x0/0x11c) from [<c0011bc0>]
(show_stack+0x10/0x14)
[  862.722047] [<c0011bc0>] (show_stack+0x10/0x14) from [<c0187fe8>]
(ubifs_releasepage+0x70/0xb8)
[  862.746573] [<c0187fe8>] (ubifs_releasepage+0x70/0xb8) from
[<c00c7228>] (try_to_release_page+0x44/0x5c)
[  862.755865] [<c00c7228>] (try_to_release_page+0x44/0x5c) from
[<c00d3b58>] (invalidate_inode_page+0x60/0x94)
[  862.764955] vbat_low_handler: vbat low
[  862.768728] [<c00d3b58>] (invalidate_inode_page+0x60/0x94) from
[<c00d3c00>] (invalidate_mapping_pages+0x74/0x114)
[  862.779244] [<c00d3c00>] (invalidate_mapping_pages+0x74/0x114) from
[<c0145004>] (drop_pagecache_sb+0x7c/0xbc)
[  862.789079] [<c0145004>] (drop_pagecache_sb+0x7c/0xbc) from
[<c0104ab4>] (iterate_supers+0x74/0xc8)
[  862.798092] [<c0104ab4>] (iterate_supers+0x74/0xc8) from [<c0145088>]
(drop_caches_sysctl_handler+0x44/0x90)
[  862.808907] [<c0145088>] (drop_caches_sysctl_handler+0x44/0x90) from
[<c014f280>] (proc_sys_call_handler+0x84/0xa0)
[  862.821868] [<c014f280>] (proc_sys_call_handler+0x84/0xa0) from
[<c014f2ac>] (proc_sys_write+0x10/0x14)
[  862.830349] [<c014f2ac>] (proc_sys_write+0x10/0x14) from [<c01022fc>]
(vfs_write+0xd4/0x16c)
[  862.838719] [<c01022fc>] (vfs_write+0xd4/0x16c) from [<c010263c>]
(SyS_write+0x3c/0x60)
[  862.846738] [<c010263c>] (SyS_write+0x3c/0x60) from [<c000e440>]
(ret_fast_syscall+0x0/0x30)


> On Wed, Apr 30, 2014 at 02:06:06PM +0800, hujianyang wrote:
>> According to this situation, my v2 fix returns from page_mkwrite
>> without performing unlock_page. We return VM_FAULT_LOCKED instead
>> of just return 0. After doing this, the race above will not happen.
>
> Hi,
>
> I have been testing patch v1 on our hardware. Normally, at higher data
> rates,
> we can hit the assert in under 8 hours. So far I have been running for 48
> hours on one system without seeing the assert message.
>
> I will now switch to running v2 and see if I can't find a few more systems
> to
> run in parallel.
>
> Many thanks, and bye for now,
> --
> Laurence Withers, <lwithers at guralp.com>
> http://www.guralp.com/
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
>


-- 
QUALCOMM ISRAEL, on behalf of Qualcomm Innovation Center, Inc. is a member
of Code Aurora Forum, hosted by The Linux Foundation