---
- branch: netbsd-5
  date: Sat Jan  9 01:22:58 UTC 2010
  files:
  - new: 1.72.4.1
    old: '1.72'
    path: src/sys/fs/puffs/puffs_msgif.c
    pathrev: src/sys/fs/puffs/puffs_msgif.c@1.72.4.1
    type: modified
  - new: 1.70.20.1
    old: '1.70'
    path: src/sys/fs/puffs/puffs_sys.h
    pathrev: src/sys/fs/puffs/puffs_sys.h@1.70.20.1
    type: modified
  - new: 1.81.8.1
    old: '1.81'
    path: src/sys/fs/puffs/puffs_vfsops.c
    pathrev: src/sys/fs/puffs/puffs_vfsops.c@1.81.8.1
    type: modified
  id: 20100109T012258Z.3c23bbdd68c6af3dd3d842f7b7d37fb9f6ef43a7
  log: "Pull up following revision(s) (requested by pooka in ticket #1212):\n\tsys/fs/puffs/puffs_msgif.c:
    revision 1.76 via patch\n\tsys/fs/puffs/puffs_sys.h: revision 1.73 via patch\n\tsys/fs/puffs/puffs_vfsops.c:
    revision 1.84 via patch\nProcess flush requests from the file server in a separate
    thread\ncontext.  This fixes a long-standing but seldomly seen deadlock,\nwhere
    the kernel was holding pages busy (due to e.g. readahead\nrequest) while waiting
    for the server to respond, and the server\nmade a callback into the kernel asking
    to invalidate those pages.\n... or, well, theoretically fixes, since I didn't
    have any reliable\nway of repeating the deadlock and I think I saw it only twice.\n"
  module: src
  subject: 'CVS commit: [netbsd-5] src/sys/fs/puffs'
  unixtime: '1263000178'
  user: snj