Description of problem: Setup: One gnbd exporter and two importers. Exported a single device (uncached) and import on the others. No volume manager was involved. Created a GFS on the device and mounted on all. Ran traffic on all. In this case the traffic was just a loop of bonnie++. The setup is similar to bug #127032, but the result is different, so new bug. On the first iteration the first importer was doing "Delete files in random order..." and the kernel Oopsed: Unable to handle kernel paging request at virtual address 00200214 printing eip: e0316965 *pde = 1a09e067 Oops: 0000 [#1] SMP Modules linked in: gnbd gfs lock_dlm dlm cman lock_harness ipv6 parport_pc lp parport autofs4 sunrpc e1000 floppy sg microcode dm_mod uhci_hcd ehci_hcd button battery asus_acpi ac ext3 jbd qla2xxx scsi_transport_fc sd_mod scsi_mod CPU: 0 EIP: 0060:[<e0316965>] Not tainted EFLAGS: 00010206 (2.6.8.1) EIP is at depend_sync_old+0x25/0x50 [gfs] eax: 0000ea60 ebx: e0501000 ecx: 002001f8 edx: c1000000 esi: df759200 edi: 0015817d ebp: 00000000 esp: d97f7d34 ds: 007b es: 007b ss: 0068 Process bonnie++ (pid: 6660, threadinfo=d97f6000 task=dee43830) Stack: dd4703c0 00000001 dd4703c0 e02f1ae6 0015817c 00000000 00000000 00000001 ca3ad800 e0501000 df759200 de923c4c c8db4ce0 de923c4c df759200 df759200 e0501000 e0318b7b 0015817c 00000000 00000000 00000000 de923c4c e03000d7 Call Trace: [<e02f1ae6>] gfs_wipe_buffers+0x136/0x2a0 [gfs] [<e0318b7b>] gfs_difree+0x5b/0x70 [gfs] [<e03000d7>] dinode_dealloc+0xc7/0x190 [gfs] [<e0300310>] inode_dealloc+0x170/0x260 [gfs] [<e0300486>] inode_dealloc_init+0x86/0xc0 [gfs] [<e031bd27>] gfs_unlinked_limit+0x57/0xa0 [gfs] [<e0310072>] gfs_unlink+0x32/0x190 [gfs] [<e02fcaf1>] gfs_glock_dq+0xc1/0x130 [gfs] [<e02fcd68>] gfs_glock_dq_uninit+0x8/0x10 [gfs] [<e03114b9>] gfs_permission+0x139/0x1b0 [gfs] [<c0168ab2>] permission+0x62/0x70 [<c016b836>] vfs_unlink+0x116/0x1e0 [<c016b9c6>] sys_unlink+0xc6/0x150 [<c01268d4>] __do_softirq+0xb4/0xc0 [<c0118b6c>] smp_apic_timer_interrupt+0xcc/0x130 [<c0105e4d>] sysenter_past_esp+0x52/0x71 Code: 8b 51 1c 01 d0 8b 15 00 fc 33 c0 39 c2 78 0b 89 ca 89 d8 e8 Version-Release number of selected component (if applicable): From cvs on s.r.c gnbd_import -V gnbd_import DEVEL.1094226134 (built Sep 3 2004 10:43:47) Copyright (C) Red Hat, Inc. 2004 All rights reserved. How reproducible: Have not attempted yet. Will try next without the exporter running traffic. Steps to Reproduce: 1. Listed above. 2. 3. Actual results: Expected results: Additional info:
I have reproduced this on the same setup but without running traffic locally do the device from the exporter. The two exporters were simultaneously running bonnie++ in a loop. Again, the node with the error was on the bonnie section of "Delete files in random order..." when it Oopsed, iteration #25. <1>Unable to handle kernel paging request at virtual address a3341158 printing eip: e02bb17c *pde = 00000000 Oops: 0002 [#2] SMP Modules linked in: gnbd lock_dlm dlm cman gfs lock_harness ipv6 parport_pc lp parport autofs4 sunrpc e1000 floppy sg microcode dm_mod uhci_hcd ehci_hcd button battery asus_acpi ac ext3 jbd qla2xxx scsi_transport_fc sd_mod scsi_mod CPU: 0 EIP: 0060:[<e02bb17c>] Not tainted EFLAGS: 00010202 (2.6.8.1) EIP is at gfs_holder_init+0x5c/0x80 [gfs] eax: da256b98 ebx: da256b70 ecx: 00000000 edx: da256b94 esi: a3341140 edi: da256ba0 ebp: 00000160 esp: d2e2bd70 ds: 007b es: 007b ss: 0068 Process df (pid: 6861, threadinfo=d2e2a000 task=d3d1ec70) Stack: da256b70 0000003d da256c00 00000000 e02bcd37 da256b70 da256b70 0000003d e02da559 da256b70 00000000 00000040 da256000 dec5cc00 00000001 d2e2be44 00000000 d2e2be44 d2e2be78 0000000d e02da686 00000000 00000001 e032d000 Call Trace: [<e02bcd37>] gfs_glock_nq_init+0x17/0x40 [gfs] [<e02da559>] stat_gfs_async+0xf9/0x1d0 [gfs] [<e02da686>] gfs_stat_gfs+0x56/0x180 [gfs] [<c017bcd4>] __mark_inode_dirty+0x1a4/0x1b0 [<c0106909>] error_code+0x2d/0x38 [<e02d218e>] gfs_statfs+0x2e/0xe0 [gfs] [<c0159a46>] vfs_statfs+0x66/0x80 [<c0159b44>] vfs_statfs64+0x14/0x40 [<c016a1da>] __user_walk+0x6a/0x80 [<c0159c84>] sys_statfs64+0x74/0xb0 [<c010c51c>] sys_mmap2+0x8c/0xd0 [<c011b950>] do_page_fault+0x0/0x4fc [<c0105e4d>] sysenter_past_esp+0x52/0x71 Code: f0 ff 46 18 8b 1c 24 8b 74 24 04 8b 7c 24 08 8b 6c 24 0c 83
Comment #1 was less than clear: - No traffic running on the exporter - Traffic running on two importers
I tried to reproduce this bug for a while, with no success. If anything changes, I can start trying again.
OK, I can no longer reproduce this bug either. Closing.