Hello Wang, On 03/31/13 03:44, Wang Shilong wrote:> Hello Arne, Jan, > > When i run stresstest with quota enabled, i hit a panic where > btrfs_find_all_roots() is called. Would you please double check it. >On what kernel version do you hit this (git commit id preferred)? Can you send us the message from dmesg? Thanks, arne> stresstest -n 10000 -t 4. > > Maybe there is still race condition when walking backref trees, Jan > would you please check the backref.c .. > btrfs_find_all_roots() is called in btrfs_qgroup_account_ref () to find > all roots that reference a extent. I think this is called because > > we introduce referenced/exclusive concepts. The point is that i think > this concept''s disadvantages outweigh the advantages: > <1> > walking backref tree is really time-consuming. > <2> referenced/exclusive makes it really complex when snapshot happens. > <3> quota has a high coupling with other module. > I''d prefer to just keep referenced value, it is not be elegant, but > it helps more. > > Thanks, > Wang > > > > >-- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hello Arne,> Hello Wang, > > On 03/31/13 03:44, Wang Shilong wrote: >> Hello Arne, Jan, >> >> When i run stresstest with quota enabled, i hit a panic where >> btrfs_find_all_roots() is called. Would you please double check it. >> > > On what kernel version do you hit this (git commit id preferred)? > Can you send us the message from dmesg? >Maybe it is because my patch [Btrfs: kill some BUG_ONs() in the find_parent_nodes()] would you please review the patch V3 for me? Thanks, Wang> Thanks, > arne > >> stresstest -n 10000 -t 4. >> >> Maybe there is still race condition when walking backref trees, Jan >> would you please check the backref.c .. >> btrfs_find_all_roots() is called in btrfs_qgroup_account_ref () to find >> all roots that reference a extent. I think this is called because >> >> we introduce referenced/exclusive concepts. The point is that i think >> this concept''s disadvantages outweigh the advantages: >> <1> >> walking backref tree is really time-consuming. >> <2> referenced/exclusive makes it really complex when snapshot happens. >> <3> quota has a high coupling with other module. >> I''d prefer to just keep referenced value, it is not be elegant, but >> it helps more. >> >> Thanks, >> Wang >> >> >> >> >> >-- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hello Arne, However, I remember that i have hit this panic in several kernel version. And i will double check it tomorrow when i am in the work.. If possible, i will send you the message from dmesg. Thanks, Wang> Hello Wang, > > On 03/31/13 03:44, Wang Shilong wrote: >> Hello Arne, Jan, >> >> When i run stresstest with quota enabled, i hit a panic where >> btrfs_find_all_roots() is called. Would you please double check it. >> > > On what kernel version do you hit this (git commit id preferred)? > Can you send us the message from dmesg? > > Thanks, > arne > >> stresstest -n 10000 -t 4. >> >> Maybe there is still race condition when walking backref trees, Jan >> would you please check the backref.c .. >> btrfs_find_all_roots() is called in btrfs_qgroup_account_ref () to find >> all roots that reference a extent. I think this is called because >> >> we introduce referenced/exclusive concepts. The point is that i think >> this concept''s disadvantages outweigh the advantages: >> <1> >> walking backref tree is really time-consuming. >> <2> referenced/exclusive makes it really complex when snapshot happens. >> <3> quota has a high coupling with other module. >> I''d prefer to just keep referenced value, it is not be elegant, but >> it helps more. >> >> Thanks, >> Wang >> >> >> >> >> >-- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html