tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: rename(), wapbl and deadlock

On Wed, Jan 13, 2010 at 09:46:27AM -0800, Brian Buhrow wrote:
 >      Hello.  We were seeing this issue with FFS and softdep, no wapbl under
 > 5.x.  Would these patches help that?  It looks like not, but I'm a litle
 > unclear about that.

It should - there should only be one ufs_rename implementation, and
wapbl has its own only because the original locking was too broken to
wedge wapbl into (semi-)correctly. The private wapbl copy of
ufs_rename is going to be going away after this stuff gets fully
debugged and committed.

You may find it blows up (with the non-wapbl case assertion Manuel
posted, or with the v_size thing I posted, or with something else) but
it's certainly worth testing if you have a box to test it on. I'd say
the patches aren't ready for production use yet though even with
yesterday's fixes.

David A. Holland

Home | Main Index | Thread Index | Old Index