On Mon, Feb 25, 2008 at 11:05:13AM +0900, YONETANI Tomokazu wrote: > On Sun, Feb 24, 2008 at 02:03:45PM -0800, Matthew Dillon wrote: > > Ok, I found it. It's a live-loop in the deadlock handling code. > > I'm tracking it down now and should get it fixed today. > > Hehe, as always -- you fix things before I could even get a chance > to reply; I'll try the committed version later today.
Ok, it's pretty stable now! I only tried on an SMP kernel running on VMware, but at least several runs of the following tests can't crash HAMMER anymore: - blogbench -d /mnt/HAMMER - cpdup /usr/pkgsrc/. /mnt/HAMMER/. The only thing I noticed was messages like below appeared all over the console: Warning: vfsync_bp skipping dirty buffer 0xc13ab924 but I see this one on a UFS filesystem too, so I think I don't need to worry much about it. Cheers.
