On Apr 25, 2010, at 23:00 , Andrew Deason wrote:
On Sun, 25 Apr 2010 21:19:24 -0400 Jeffrey Altman <jalt...@secure-endpoints.com> wrote:When a whole file lock is write-held, all of the dirty data in the cache must be written back to the file server before the lock is released. This is currently not being done and as a result, the database becomes corrupted.If I understand Simon correctly, clients are also not getting file updates while the file is opened O_RDWR. That would still seem to be a problem even if we flush on unlock.
Flushing will update the server, which will break callbacks as necessary. The default behavior is to only flush changes to the server on close, but there's no good reason to change that in the situation where there is no locking; the specced behavior remains "last closer wins". (I think fsync() updates the server in all cases, though.)
-- brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allb...@kf8nh.com system administrator [openafs,heimdal,too many hats] allb...@ece.cmu.edu electrical and computer engineering, carnegie mellon university KF8NH
PGP.sig
Description: This is a digitally signed message part