Most recent kernel where this bug did not occur: 2.6.14-rc2-mm2 ----------------------------------------------------- Distribution: SLES 9 SP2 ----------------------------------------------------- Hardware Environment: 2 way, Intel(R) Xeon(TM) CPU 2.80GHz, 1.5 GB SCSI driver: IBM ServeRAID Controller Network Driver: Broadcom Corporation NetXtreme BCM5703X Gigabit Ethernet ----------------------------------------------------- Software Environment: Linux x235b 2.6.14-rc5-I #1 SMP PREEMPT Mon Oct 31 08:46:32 PDT 2005 i686 i686 i386 GNU/Linux ----------------------------------------------------- Problem Description: On running the File System stress test, over Reiserfs encountered the OOPs message. x235b:~ # Oops: 0002 [#1] PREEMPT SMP DEBUG_PAGEALLOC Modules linked in: CPU: 2 EIP: 0060:[<c01c67f0>] Not tainted VLI EFLAGS: 00010286 (2.6.14-rc5-I) EIP is at reiserfs_free_jh+0x30/0x70 eax: d63eb930 ebx: d63eb928 ecx: 00000000 edx: c5a86100 esi: f189fed8 edi: c5861540 ebp: f7d31da8 esp: f7d31d9c ds: 007b es: 007b ss: 0068 Process kswapd0 (pid: 210, threadinfo=f7d30000 task=c5d7e530) Stack: c04b07d9 f189fed8 f189fed8 f7d31dd4 c01b01ab f189fed8 c5be64e0 c5aff4c0 c5a788a0 f8d110ec 00000001 f4313a44 c5861540 f4313a44 f7d31de8 c016775d c5861540 000000d0 f4313a44 f7d31e60 c014fc8f c5861540 000000d0 00000000 Call Trace: [<c010404b>] show_stack+0xab/0xc0 [<c01041e6>] show_registers+0x166/0x1e0 [<c010440d>] die+0x10d/0x1b0 [<c04b17ec>] do_page_fault+0x3dc/0x61d [<c0103cdb>] error_code+0x4f/0x54 [<c01b01ab>] reiserfs_releasepage+0x6b/0x100 [<c016775d>] try_to_release_page+0x4d/0x60 [<c014fc8f>] shrink_list+0x35f/0x480 [<c014ff6b>] shrink_cache+0x10b/0x2c0 [<c01505de>] shrink_zone+0xbe/0xf0 [<c0150aad>] balance_pgdat+0x24d/0x3d0 [<c0150d11>] kswapd+0xe1/0x140 [<c01012a9>] kernel_thread_helper+0x5/0xc Code: 0c 89 75 fc 8b 75 08 89 5d f8 8b 5e 28 85 db 74 3e c7 46 28 00 00 00 00 8d 43 08 8b 53 08 c7 43 0 <6>note: kswapd0[210] exited with preempt_count 1 Bad page state at free_hot_cold_page (in process 'fsstress', page c5a86100) flags:0x80010008 mapping:d0ea511c mapcount:0 count:0 Backtrace: [<c010407e>] dump_stack+0x1e/0x20 [<c01476d2>] bad_page+0x82/0xc0 [<c0147fc0>] free_hot_cold_page+0x90/0x150 [<c014889c>] __pagevec_free+0x1c/0x30 [<c014e80c>] release_pages+0x17c/0x190 [<c014e847>] __pagevec_release+0x27/0x40 [<c0150109>] shrink_cache+0x2a9/0x2c0 [<c01505de>] shrink_zone+0xbe/0xf0 [<c015067a>] shrink_caches+0x6a/0x90 [<c0150782>] try_to_free_pages+0xe2/0x1c0 [<c01485e1>] __alloc_pages+0x241/0x460 [<c01454b5>] generic_file_buffered_write+0x125/0x660 [<c02bc586>] xfs_write+0x886/0xc80 [<c02b7ad0>] linvfs_aio_write+0x90/0xa0 [<c01646ea>] do_sync_write+0xba/0x110 [<c01647f8>] vfs_write+0xb8/0x190 [<c016499b>] sys_write+0x4b/0x80 [<c010317f>] sysenter_past_esp+0x54/0x75 Trying to fix it up, but a reboot is needed
would you, please, share your stress test?
File System stress test is a collection of "bonnie, dbench, fs_inode, fsstress , fsx_linux,IOZone, postmark, tiobench" run parallely. You should be able to get the source code from the web.
*** Bug 5536 has been marked as a duplicate of this bug. ***
would you, please, provide the information about the tests parameters?
These are the parameters Tests Parameters bonnie -u root -d $TESTPLACE -s 8192:256 -n 10 dbench 30 fs_inode $TESTPLACE 100 200 100 fsstress -d $TESTPLACE -n 100 -p 5000 -X fsx_linux -N 100000 $TESTPLACE/fsx-linux.$FS_TYPE IOZone -a postmark set size 10 100 set number 100 set transactions 95 set location $TESTPLACE show run quit tiobench --block 4096 --block 8192 --threads $THREAD --size 3066 --numruns 10 Note: $THREAD will take values 2,4,8 in the order $TESTPLACE will point to location where the executables are located
I see XFS in the traces, so the patch to init bh->b_private on allocating buffer heads should fix this one. http://www.kernel.org/hg/linux-2.6/?cs=589b23656d52