On Wed, May 20, 2020 at 11:54:22AM -0700, Darrick J. Wong wrote: > On Wed, May 20, 2020 at 08:46:43AM +0200, Christoph Hellwig wrote: > > On Tue, May 19, 2020 at 06:46:27PM -0700, Darrick J. Wong wrote: > > > From: Darrick J. Wong <darrick.wong@xxxxxxxxxx> > > > > > > Hide the incore inode walk interface because callers outside of the > > > icache code don't need to know about iter_flags and radix tags and other > > > implementation details of the incore inode cache. > > > > I don't really see the point here. It isn't hiding much, and only from > > a single caller. I have to say I also prefer the old naming over _ici_ > > and find the _all postfix not exactly descriptive. > > This last patch is more of a prep patch for the patchsets that come > after it: cleaning up the block gc stuff and deferred inode > inactivation. It's getting kinda late so I didn't want to send 11 more > patches, but perhaps that would make it clearer where this is all > heading? I'd say drop it from this series and resend it with that series if you are going to send it out. > The quota dqrele_all code does not care about inode radix tree tags nor It only started to care about them because you merged the functions and now exposed them to the dqrele code at the beginning of this series. But with just a single user not caring and too aring I'm perfectly fine with exposing the tags in the interface. > does it need the ability to grab inodes /while/ they're in INEW state, > so there's no reason to pass those arguments around. > > OTOH I guess I could have hid XFS_AGITER_INEW_WAIT in xfs_icache.c and > left the function names unchanged. Maybe just flip the default for the flag if that makes your series easier?
