On Sat, Nov 09, 2019 at 07:00:20AM +0100, Andreas Färber wrote: > Am 09.11.19 um 05:01 schrieb Qu Wenruo: > > On 2019/11/9 上午11:45, Andreas Färber wrote: > >> On arm64 I'm seeing a regression between next-20191031 and next-20191105 > >> that breaks boot from my btrfs rootfs: next-20191105 and later oopses on > >> found->lock, or with CONFIG_BTRFS_ASSERT asserts on a NULL "found" > >> variable in btrfs_update_space_info(). > >> > >> According to git-blame that code hasn't changed in months, and I didn't > >> spot an obvious cause among the fs/btrfs/ commis between those two tags. > > It looks like caused by "btrfs: block-group: Refactor > > btrfs_read_block_groups()". > > Thanks for the quick pointer - I've reverted all fs/btrfs/ commits down > to that one and it was indeed the culprit. > > > Due to another refactor, there are conflicts in that patch and is not > > resolved properly. > > > > Please try David's latest misc-next branch, which includes the proper > > rebased refactor. > > Unfortunately that won't work as I'm developing some 100+ patches for my > linux-realtek.git, so I'll have to carry the reverts until the fix makes > it into linux-next.git. Fixed branch for linux-next will be pushed today, sorry for the inconvenience.
