tech-kern archive

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

Re: UVM/genfs review



On Wed, Jan 14, 2009 at 10:18:40PM +0100, Joerg Sonnenberger wrote:
> On Wed, Jan 14, 2009 at 02:43:53PM -0600, Greg Oster wrote:
> > So where are we at on this?  Does the deadlock that Juergen refers to 
> > occur only with fss+wapbl or with fss+other as well?
> 
> The patch only changes the way the WAPBL journal is locked, so fss on
> non-WAPBL doesn't change. I simply don't know how easy it is to hit that
> deadlock. I do have issues with snapacct panicing during the suspend of
> an active filesystem. Once the filesystem was suspended I could tar the
> suspended filesystem up while a bulk build was running.

I propose that we either:

A) disallow creation of snapshots on filesystems mounted with log, or
B) cause options WAPBL to imply options FFS_NO_SNAPSHOT (there are
   already other FFS-related options, like APPLE_UFS, which do this).

As temporary workarounds to get WAPBL to perform well enough for common
workloads in 5.0 to be generally useful.

Thor


Home | Main Index | Thread Index | Old Index