On Tue, Feb 05, 2019 at 11:03:02AM -0500, Dennis Zhou wrote: > On Tue, Feb 05, 2019 at 03:57:53PM +0100, David Sterba wrote: > > On Mon, Feb 04, 2019 at 03:19:56PM -0500, Dennis Zhou wrote: > > > Hi everyone, > > > > > > V2 had only a handful of changes outside of minor feedback. > > > 0001: > > > - use functions over macros > > > 0003: > > > - BTRFS_NR_WORKSPACE_MANAGERS is added instead of overriding > > > BTRFS_COMPRESS_TYPES > > > 0011 (new): > > > - address monotonic memory requirement for zstd workspaces > > > 0012: > > > - increase reclaim timer to 307s from 67s > > > - move from keeping track of time in ns to jiffies > > > - remove timer in cleanup code > > > - use min_t() instead of if statements in .set_level() > > > - add header text to describe how workspaces are managed > > > - nofs_flag type -> unsigned long to unsigned int > > > > Something is wrong, the patchset on top of 5.0-rc5 hangs in test > > btrfs/007, without a stacktrace. V1 was fine and I double checked that > > rc5 itself is fine. > > Hmmm, well that's awkward. I ran xfstests and that test passed on my > machine. I'll figure out the delta and submit a v3. Thanks David!
It failed on a VM and 2 other physical machines, so it's not somethig related to the testing setup.
