home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!auspex-gw!guy
- From: guy@Auspex.COM (Guy Harris)
- Newsgroups: comp.unix.wizards
- Subject: Re: SUN NFS sync
- Keywords: sync
- Message-ID: <15360@auspex-gw.auspex.com>
- Date: 5 Nov 92 23:28:06 GMT
- References: <1992Oct28.221355.9933@gupta.portal.com> <chris.720593732@suite.sw.oz.au>
- Sender: news@auspex-gw.auspex.com
- Organization: Auspex Systems, Santa Clara
- Lines: 25
- Nntp-Posting-Host: auspex.auspex.com
-
- >It's unclear to me whether these
- >NFS implementations support the O_SYNC option or the fsync(2) system
- >call.
-
- Some of them may not necessarily support *any* UNIX operations! They
- aren't necessarily running UNIX....
-
- I.e., the server can't support them as an NFS server, as there's no NFS
- request that corresponds to "fsync()", nor is there a flag on the NFS
- write operation that corresponds to O_SYNC.
-
- O_SYNC, and "fsync()", are *client*-side functions - and, on many
- (probably most, if not all) UNIX implementations, they *are* necessary;
- the claim that
-
- >The answer is: "You don't need to worry because the NFS protocol
- >specifies that all server writes are synchronous." This means that
- >your client file write system call won't return until the data is
- >actually on disk on the server. No sync is required.
-
- isn't true. The client file write system call can (as has been
- mentioned elsewhere in this thread) simply queue up the write to be done
- at some later time to the server. The write in question will be
- synchronous on the server, but it's not necessarily going to be done as
- part of the client file write system call.
-