Re: [PATCH] btrfs: qgroup: Fix qgroup accounting when creating snapshot

2016-04-07 Thread Mark Fasheh
On Thu, Apr 07, 2016 at 04:21:53PM +0800, Qu Wenruo wrote: > I ran into one soft lockup with my patch only. So I assume it's not > caused by your inherit patch though. > But I didn't reproduce it once more. Not sure why. > > What's the reproduce rate in your environment? It happens every time

Re: [PATCH] btrfs: qgroup: Fix qgroup accounting when creating snapshot

2016-04-07 Thread Qu Wenruo
Hi Mark, Thanks for the test and reporting. Mark Fasheh wrote on 2016/04/06 19:43 -0700: Hi Qu, On Wed, Apr 06, 2016 at 10:04:54AM +0800, Qu Wenruo wrote: Current btrfs qgroup design implies a requirement that after calling btrfs_qgroup_account_extents() there must be a commit root switch.

Re: [PATCH] btrfs: qgroup: Fix qgroup accounting when creating snapshot

2016-04-06 Thread Mark Fasheh
Hi Qu, On Wed, Apr 06, 2016 at 10:04:54AM +0800, Qu Wenruo wrote: > Current btrfs qgroup design implies a requirement that after calling > btrfs_qgroup_account_extents() there must be a commit root switch. > > Normally this is OK, as btrfs_qgroup_accounting_extents() is only called > inside

Re: [PATCH] btrfs: qgroup: Fix qgroup accounting when creating snapshot

2016-04-06 Thread Qu Wenruo
Filipe Manana wrote on 2016/04/06 10:20 +0100: On Wed, Apr 6, 2016 at 3:04 AM, Qu Wenruo wrote: Current btrfs qgroup design implies a requirement that after calling btrfs_qgroup_account_extents() there must be a commit root switch. Normally this is OK, as

Re: [PATCH] btrfs: qgroup: Fix qgroup accounting when creating snapshot

2016-04-06 Thread Filipe Manana
On Wed, Apr 6, 2016 at 3:04 AM, Qu Wenruo wrote: > Current btrfs qgroup design implies a requirement that after calling > btrfs_qgroup_account_extents() there must be a commit root switch. > > Normally this is OK, as btrfs_qgroup_accounting_extents() is only called >

[PATCH] btrfs: qgroup: Fix qgroup accounting when creating snapshot

2016-04-05 Thread Qu Wenruo
Current btrfs qgroup design implies a requirement that after calling btrfs_qgroup_account_extents() there must be a commit root switch. Normally this is OK, as btrfs_qgroup_accounting_extents() is only called inside btrfs_commit_transaction() just be commit_cowonly_roots(). However there is a