Re: [fuse-devel] hung_task trigger on Kernel 3.10

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 454-5900

Hi,

thanks for the help. I will try a back port. Few structures and 
APIs have changed, but the errors are not so numerous. 

In the first step I will focus on the changes in file.c. In this context, 
there were changes. I need to understand the problem even more.
Is the condition or the missing wake-up the reason for the hung.

fs/fuse/dir.c:28:9: error: dereferencing pointer to incomplete type
fs/fuse/dir.c:277:2: error: implicit declaration of function
‘check_submounts_and_drop’ [-Werror=implicit-function-declaration]
fs/fuse/dir.c:596:2: error: implicit declaration of function
‘d_instantiate_no_diralias’ [-Werror=implicit-function-declaration]
fs/fuse/dir.c:788:15: error: ‘RENAME_EXCHANGE’ undeclared (first use in
this function)
fs/fuse/dir.c:831:16: error: ‘RENAME_NOREPLACE’ undeclared (first use in
this function)
fs/fuse/dir.c:831:35: error: ‘RENAME_EXCHANGE’ undeclared (first use in
this function)
fs/fuse/dir.c:1227:3: error: implicit declaration of function ‘dir_emit’
[-Werror=implicit-function-declaration]
fs/fuse/dir.c:1233:6: error: dereferencing pointer to incomplete type
fs/fuse/dir.c:1378:7: error: dereferencing pointer to incomplete type
fs/fuse/dir.c:1422:32: error: dereferencing pointer to incomplete type
fs/fuse/dir.c:1425:32: error: dereferencing pointer to incomplete type
fs/fuse/dir.c:1799:3: error: too few arguments to function
‘truncate_pagecache’
fs/fuse/dir.c:2021:2: error: unknown field ‘rename2’ specified in
initializer
fs/fuse/dir.c:2038:2: error: unknown field ‘iterate’ specified in
initializer
fs/fuse/file.c:1239:11: error: too few arguments to function
‘generic_file_direct_write’
fs/fuse/file.c:2140:2: error: unknown field ‘map_pages’ specified in
initializer
fs/fuse/file.c:2140:15: error: ‘filemap_map_pages’ undeclared here (not in
a function)
fs/fuse/inode.c:126:2: error: implicit declaration of function
‘truncate_inode_pages_final’ [-Werror=implicit-function-declaration]
fs/fuse/inode.c:232:4: error: too few arguments to function
‘truncate_pagecache’
fs/fuse/inode.c:957:46: error: ‘BDI_CAP_STRICTLIMIT’ undeclared (first use
in this function)
fs/fuse/cuse.c:593:12: error: ‘CUSE_MINOR’ undeclared here (not in a
function)
fs/fuse/cuse.c:619:12: error: ‘struct class’ has no member named
‘dev_groups’
fs/fuse/cuse.c:619:27: error: ‘cuse_class_dev_groups’ undeclared (first
use in this function)

Thx
Harald

On Wed, 23 Jul 2014 18:44:56 +0400, Maxim Patlasov
<mpa...@pa...> wrote:
> On 07/23/2014 06:27 PM, hk wrote:
>> Hi,
>>
>> I'll try to switch to 3.15, but this is not so easy. I use board vendor
>> patches to get my system up and running.
>>
>> Would a backport also be an option?
> 
> Yes, I can try to port kernel fuse from 3.15 to 3.10.40, but this will 
> take a while and I can start only in a few days. Either you can do it 
> yourself if you have time and desire - copy fs/fuse/*.[ch] and 
> include/uapi/linux/fuse.h from 3.15 to your 3.10.40 kernel build 
> location and polish angles until compiler gets happy.
> 
> Thanks,
> Maxim
> 
>>
>> thx
>> Harald
>>
>>
>> On Wed, 23 Jul 2014 16:23:58 +0400, Maxim Patlasov
>> <mpa...@pa...> wrote:
>>> Hi,
>>>
>>> On 07/23/2014 03:48 PM, hk wrote:
>>>> Hi,
>>>> thx for the hint.
>>>>
>>>> I have installed the patch but without success. The same problem
occurs
>>>> again.
>>> Sorry for that. Can you try to reproduce the problem on a Linus'
modern
>>> kernel (e.g. v3.15)? There was a huge rework of fuse writeback
subsystem
>>> since 3.10 and some fixes might slip of 3.10.40.
>>>
>>> Thanks,
>>> Maxim
>>>
>>>> INFO: task webfuse_server:1714 blocked for more than 60 seconds.
>>>> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this
>> message.
>>>> webfuse_server D c03cd980     0  1714   1637 0x00000001
>>>> Backtrace:
>>>> [<c03cd6f4>] (__schedule+0x0/0x328) from [<c03cddfc>]
>>>> (schedule+0xa0/0xb8)
>>>> [<c03cdd5c>] (schedule+0x0/0xb8) from [<c016cb88>]
>>>> (fuse_wait_on_page_writeback+0x8c/0xac)
>>>>    r4:c6fdbc0c r3:c6fdbc18
>>>> [<c016cafc>] (fuse_wait_on_page_writeback+0x0/0xac) from [<c016cc1c>]
>>>> (fuse_page_mkwrite+0x74/0x7c)
>>>>    r7:c016cba8 r6:b6b0d000 r5:c69da780 r4:c067f280
>>>> [<c016cba8>] (fuse_page_mkwrite+0x0/0x7c) from [<c007e76c>]
>>>> (handle_pte_fault+0x500/0xab4)
>>>>    r5:c6e413c8 r4:c067f280
>>>> [<c007e26c>] (handle_pte_fault+0x0/0xab4) from [<c007edcc>]
>>>> (handle_mm_fault+0xac/0xd4)
>>>> [<c007ed20>] (handle_mm_fault+0x0/0xd4) from [<c001448c>]
>>>> (do_page_fault+0x12c/0x288)
>>>> [<c0014360>] (do_page_fault+0x0/0x288) from [<c0008600>]
>>>> (do_DataAbort+0x3c/0xa0)
>>>> [<c00085c4>] (do_DataAbort+0x0/0xa0) from [<c000e618>]
>>>> (__dabt_svc+0x38/0x60)
>>>> Exception stack(0xc6fdbdd8 to 0xc6fdbe20)
>>>> bdc0:                                                       b6b0d60f
>>>> 00000000
>>>> bde0: c6fdbfb0 4056f489 0005d275 c6fdbfb0 c0015ca8 0002c8a0 00000801
>>>> 00000000
>>>> be00: b6b0d60f c6fdbe3c 00000002 c6fdbe24 c0016114 c0015d70 40000013
>>>> ffffffff
>>>>    r7:c6fdbe0c r6:ffffffff r5:40000013 r4:c0015d70
>>>> [<c0015ca8>] (do_alignment_ldrstr+0x0/0xf8) from [<c0016114>]
>>>> (do_alignment+0x24c/0x3f8)
>>>>    r6:c0015ca8 r5:c6fdbfb0 r4:00000028
>>>> [<c0015ec8>] (do_alignment+0x0/0x3f8) from [<c0008600>]
>>>> (do_DataAbort+0x3c/0xa0)
>>>> [<c00085c4>] (do_DataAbort+0x0/0xa0) from [<c000e820>]
>>>> (__dabt_usr+0x40/0x60)
>>>> Exception stack(0xc6fdbfb0 to 0xc6fdbff8)
>>>> bfa0:                                     00000000 00000000 4056f489
>>>> 0017b658
>>>> bfc0: 00000001 00000000 00000028 00f08cbe 00000000 00000003 0000000f
>>>> 0000000f
>>>> bfe0: b6b0d5e7 bee24968 41206f55 0002c8a4 80000010 ffffffff
>>>>    r7:00f08cbe r6:ffffffff r5:80000010 r4:0002c8a0
>>>>
>>>>
>>>> Any other suggestions?
>>>> Harald
>>>>
>>>>
>>>> On Tue, 22 Jul 2014 14:33:21 +0400, Maxim Patlasov
>>>> <mpa...@pa...> wrote:
>>>>> Hi,
>>>>>
>>>>> On 07/22/2014 11:30 AM, hk wrote:
>>>>>> Hello,
>>>>>> I ported my fuse application from 2.6.20 to 3.10 kernel.
>>>>>> Now the application hangs from  time to time. It seems
>>>>>> a kernel problem happens.
>>>>> This may be due to the lack of Miklos' fix
>>>>> cca2437045dda994d23bd65891b71e091fa35b5a in 3.10.40.
>>>>>
>>>>> Thanks,
>>>>> Maxim
>>>>>
>>>>>> Here is the trace:
>>>>>>
>>>>>> INFO: task webfuse_server:1713 blocked for more than 60 seconds.
>>>>>> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this
>>>> message.
>>>>>> webfuse_server D c03cd940     0  1713   1637 0x00000001
>>>>>> Backtrace:
>>>>>> [<c03cd6b4>] (__schedule+0x0/0x328) from [<c03cddbc>]
>>>>>> (schedule+0xa0/0xb8)
>>>>>> [<c03cdd1c>] (schedule+0x0/0xb8) from [<c016cb88>]
>>>>>> (fuse_wait_on_page_writeback+0x8c/0xac)
>>>>>>     r4:c686bc14 r3:c686bc20
>>>>>> [<c016cafc>] (fuse_wait_on_page_writeback+0x0/0xac) from
[<c016cbcc>]
>>>>>> (fuse_page_mkwrite+0x24/0x2c)
>>>>>>     r7:c016cba8 r6:b6ae6000 r5:c683bd68 r4:c0686980
>>>>>> [<c016cba8>] (fuse_page_mkwrite+0x0/0x2c) from [<c007e76c>]
>>>>>> (handle_pte_fault+0x500/0xab4)
>>>>>> [<c007e26c>] (handle_pte_fault+0x0/0xab4) from [<c007edcc>]
>>>>>> (handle_mm_fault+0xac/0xd4)
>>>>>> [<c007ed20>] (handle_mm_fault+0x0/0xd4) from [<c001448c>]
>>>>>> (do_page_fault+0x12c/0x288)
>>>>>> [<c0014360>] (do_page_fault+0x0/0x288) from [<c0008600>]
>>>>>> (do_DataAbort+0x3c/0xa0)
>>>>>> [<c00085c4>] (do_DataAbort+0x0/0xa0) from [<c000e618>]
>>>>>> (__dabt_svc+0x38/0x60)
>>>>>> Exception stack(0xc686bdd8 to 0xc686be20)
>>>>>> bdc0:                                                      
b6ae660f
>>>>>> 00000000
>>>>>> bde0: c686bfb0 4056f489 02388468 c686bfb0 c0015ca8 0002c8a0
00000801
>>>>>> 00000000
>>>>>> be00: b6ae660f c686be3c 00000002 c686be24 c0016114 c0015d70
40000013
>>>>>> ffffffff
>>>>>>     r7:c686be0c r6:ffffffff r5:40000013 r4:c0015d70
>>>>>> [<c0015ca8>] (do_alignment_ldrstr+0x0/0xf8) from [<c0016114>]
>>>>>> (do_alignment+0x24c/0x3f8)
>>>>>>     r6:c0015ca8 r5:c686bfb0 r4:00000028
>>>>>> [<c0015ec8>] (do_alignment+0x0/0x3f8) from [<c0008600>]
>>>>>> (do_DataAbort+0x3c/0xa0)
>>>>>> [<c00085c4>] (do_DataAbort+0x0/0xa0) from [<c000e820>]
>>>>>> (__dabt_usr+0x40/0x60)
>>>>>> Exception stack(0xc686bfb0 to 0xc686bff8)
>>>>>> bfa0:                                     00000000 00000000
4056f489
>>>>>> 0017b658
>>>>>> bfc0: 00000001 00000000 00000028 00b3acbe 00000000 00000003
0000000f
>>>>>> 0000000f
>>>>>> bfe0: b6ae65e7 beda7968 40088777 0002c8a4 20000010 ffffffff
>>>>>>     r7:00b3acbe r6:ffffffff r5:20000010 r4:0002c8a0
>>>>>>
>>>>>>
>>>>>> System:
>>>>>> Arm Linux EABI with 3.10.40
>>>>>> arm926ej-s (Freescale MX28 / gcc 4.7.3
>>>>>> Tested with few libfuse versions (current 2.9.3)
>>>>>> uclibc 0.9.33.2
>>>>>>
>>>>>> Any hints are welcome
>>>>>> Harald
>>>>>>
>>>>>>
>>>>>>
>>
------------------------------------------------------------------------------
>>>>>> Want fast and easy access to all the code in your enterprise? Index
>> and
>>>>>> search up to 200,000 lines of code with a free copy of Black Duck
>>>>>> Code Sight - the same software that powers the world's largest code
>>>>>> search on Ohloh, the Black Duck Open Hub! Try it now.
>>>>>> http://p.sf.net/sfu/bds
>>>>>> _______________________________________________
>>>>>> fuse-devel mailing list
>>>>>> fus...@li...
>>>>>> https://lists.sourceforge.net/lists/listinfo/fuse-devel
>>>>>>
>>>>>>
>>>
>>>
>>
------------------------------------------------------------------------------
>>> Want fast and easy access to all the code in your enterprise? Index
and
>>> search up to 200,000 lines of code with a free copy of Black Duck
>>> Code Sight - the same software that powers the world's largest code
>>> search on Ohloh, the Black Duck Open Hub! Try it now.
>>> http://p.sf.net/sfu/bds
>>> _______________________________________________
>>> fuse-devel mailing list
>>> fus...@li...
>>> https://lists.sourceforge.net/lists/listinfo/fuse-devel
>>
> 
> 
>
------------------------------------------------------------------------------
> Want fast and easy access to all the code in your enterprise? Index and
> search up to 200,000 lines of code with a free copy of Black Duck
> Code Sight - the same software that powers the world's largest code
> search on Ohloh, the Black Duck Open Hub! Try it now.
> http://p.sf.net/sfu/bds
> _______________________________________________
> fuse-devel mailing list
> fus...@li...
> https://lists.sourceforge.net/lists/listinfo/fuse-devel