On Wed, Aug 17, 2011 at 03:24:25PM +0000, Emmanuel Dreyfus wrote: > I backported this change, and the problem occurs much less often, > and now always at offset 0. Do we have two different bugs? > http://mail-index.netbsd.org/source-changes/2009/11/05/msg002710.html
Here is a trace below (netbsd-5, not using PUFFS_KFLAG_WTCACHE, with yamt's patch and the fix above backported. zeroed chunk appeared 953 bytes @ 8192) reqid: 0, opclass 2 (FAF), optype: PUFFS_VN_WRITE, cookie: 0xbb90f340, aux: 0xbb91f02c, auxlen: 9289, pid: 0, lwpid: 27 offset: 0, resid 9145, ioflag 0x0 since previous call: 0.128994 reqid: 2703, opclass 2, optype: PUFFS_VN_SETATTR, cookie: 0xbb90f340, aux: 0xbb91e02c, auxlen: 240, pid: 0, lwpid: 27 size: 9145/0x23b9 since previous call: 0.000512 reqid: 2704, opclass 2, optype: PUFFS_VN_GETATTR, cookie: 0xbb90f340, aux: 0xbb91f02c, auxlen: 240, pid: 19324, lwpid: 1 since previous call: 0.005915 RV reqid: 2704, result: 0 RV reqid: 2703, result: 0 reqid: 2705, opclass 2, optype: PUFFS_VN_FSYNC, cookie: 0xbb90f340, aux: 0xbb92002c, auxlen: 148, pid: 0, lwpid: 27 since previous call: 0.083257 RV reqid: 2705, result: 0 reqid: 2706, opclass 1, optype: PUFFS_VFS_SYNC, cookie: 0x0, aux: 0xbb91e02c, auxlen: 132, pid: 0, lwpid: 27 since previous call: 0.009297 RV reqid: 2706, result: 0 That seems to be triggered by fsync. -- Emmanuel Dreyfus m...@netbsd.org