Re: [crash, -git] NULL pointer dereference, IP: [<c0190884>] page_cgroup_zoneinfo+0xf/0x1b

[Balbir Singh <balbir@linux.vnet.ibm.com>]

Ingo Molnar wrote:
> * KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> 
>>> BUG: unable to handle kernel NULL pointer dereference at 00000000
>>> IP: [<c018faa4>] page_cgroup_zoneinfo+0xf/0x1b
> [...]
>>> bad:   a9b6148: USB: Fix unused label warnings in drivers/usb/host/ehci-hcd.c
>>> good:  c813b4e: Merge git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/
>>>
>> yes, now chasing. and have found
>>
>>  page_cgroup->page is NULL because of unknown reason.
> 
> i've attempted to bisect it - but the most critical part of this series 
> is not bisectable, the bootup hangs at this tree:
> 
>   40b6a76: cpuset.c: remove extra variable
> 
> It hangs in the mm/cgroups code at:
> 
>  [<c01895fe>] ? constant_test_bit+0xe/0x1b
>  [<c01898a1>] lock_page_cgroup+0x13/0x25
>  [<c0189e5f>] mem_cgroup_charge_common+0xfd/0x1bf
>  [<c018a56f>] mem_cgroup_cache_charge+0x7c/0x8a
>  [<c016bceb>] add_to_page_cache_locked+0x2f/0xc4
>  [<c016bda3>] add_to_page_cache_lru+0x23/0x5e
>  [<c016c3e9>] find_or_create_page+0x3f/0x62
>  [<c01a4709>] __getblk+0xfa/0x1cc
>  [<c01ff4fb>] jread+0x110/0x1ed
>  [<c01ff6d4>] do_one_pass+0xfc/0x476
>  [<c01408b0>] ? wake_bit_function+0x0/0x43
>  [<c01ffac8>] journal_recover+0x7a/0x13d
>  [<c0200d8e>] ? buffer_uptodate+0xc/0xe
>  [<c0201bd9>] journal_load+0x42/0x76
>  [<c01f810b>] ext3_fill_super+0xd9c/0x1491
> 
> My bisection got this far:
> 
>  # good: [c813b4e1] Merge git://git.kernel.org/pub/scm/linux/kernel/gi
>  # bad:  [a9b6148d] USB: Fix unused label warnings in drivers/usb/host
>  # bad:  [656eb2cd] add CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
>  # good: [bdbf0ac7] Merge branch 'hwmon-for-linus' of git://jdelvare.p
>  # good: [f7ea4a4b] Merge branch 'drm-next' of git://git.kernel.org/pu
>  # good: [556adecb] vmscan: second chance replacement for anonymous pa
>  # good: [28f74e71] hwmon/pc87360 separate alarm files: define some co
>  # good: [46d01a22] ext3: fix ext3 block reservation early ENOSPC issu
>  # good: [47c59803] devcgroup: remove spin_lock()
> 
> The remaining bisection window of uncategorized commits is:
> 
>  nr of bisection commits: 14
>  ----------------------------
>  656eb2c: add CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
>  6409324: coredump: format_corename: don't append .%pid if  multi-threaded
>  b747c8c: make ptrace_untrace() static
>  c459643: bitmask: remove bitmap_scnprintf_len()
>  30e8e13: cpuset: use seq_*mask_* to print masks
>  3eda201: seq_file: add seq_cpumask_list(), seq_nodemask_list()
>  85dd030: seq_file: don't call bitmap_scnprintf_len()
>  40b6a76: cpuset.c: remove extra variable
>  52d4b9a: memcg: allocate all page_cgroup at boot
>  c05555b: memcg: atomic ops for page_cgroup->flags
>  addb9ef: memcg: optimize per-cpu statistics
>  5b4e655: memcg: avoid accounting special pages
>  b7abea9: memcg: make page->mapping NULL before uncharge
>  073e587: memcg: move charge swapin under lock
> 
> i'll poke in the middle - maybe i'm lucky and can find a 
> working/crashing tree to nudge the bisection back into life.

Hi, Ingo,

Li Zefan has reported this problem and Kamezawa is looking into it. I am unwell,
I hope to look into the problem when I feel better or at-least in the intervals
of time that I feel better.

-- 
	Balbir