On Tue, Dec 21, 2010 at 06:28:56PM +1100, Dave Chinner wrote:
> I'm sending the entire series of scalability patches in a single
> patchbomb because I'm tired and it's too much like hard work to send
> it out in multiple patchsets (i.e. I'm being lazy). Overall there
> are relatively few changes:
> - new patch for iolock lockdep annotations
> - new patch for allocations under ilock
> rcu inode freeing and lookup:
> - reworked reclaim to use rcu read locking
> - removed synchronise_rcu() from lookup failure
> - cleaned up validity checks, added comments and rcu_read_lock_held
> AIL locking
> - fixed aild sleep to use TASK_INTERRUPTABLE
> Log grant scaling
> - made reserveq/writeq tracing just indicate if there are queued
> - cleaned up some minor formatting nitpicks suggested by Christoph
> - split xlog_space_left() into __xlog_space_left() for AIl tail
> pushing to work off a single tail lsn value.
> I'm mainly concerned with getting reviews for the few remaining
> patches that don't currently have reviewed-by tags. Christoph, I
> think I've fixed all the things your last round of comments covered,
> so there should be relatively little remaining to be fixed up.
> The series is in the following git tree which is based on the
> current OSS xfs tree. Alex, once I get the remaining reviews
> complete I'll update the branch and send you a pull request.
I've just pushed a new version out with all the new reviewed-by tags
and fixups noticed.
kernel.org is pretty slow right now, so it might take a while for
it to propagate through.
The patches I still need reviews for are:
lib: percpu counter add unless less than functionality
xfs: dynamic speculative EOF preallocation
xfs: convert l_tail_lsn to an atomic variable