cvs commit: src/lib/libc/sys msync.2
alc at cs.rice.edu
Sun Sep 5 12:46:34 PDT 2004
On Sat, Sep 04, 2004 at 09:30:43PM -0400, David Schultz wrote:
> On Fri, Sep 03, 2004, Peter Pentchev wrote:
> > roam 2004-09-03 06:24:25 UTC
> > FreeBSD src repository (doc,ports committer)
> > Modified files:
> > lib/libc/sys msync.2
> > Log:
> > Add a BUGS section and copy the wording from mmap(2)'s MAP_NOSYNC,
> > documenting the obsoleteness of the msync(2) syscall and its single
> > remaining purpose.
> I'm not nitpicking at you since you didn't write the original
> text, but if msync(2) still has a purpose, then it isn't really
> obsolete, is it? (Moreover, the text only describes the purpose of
> msync(2) with the MS_ASYNC flag.) Applications such as databases
> that want greater control over the flushing of dirty data may
> still find msync(2) very useful.
I agree. This sentence should be removed from both msync.2 and mmap.2
for precisely the reason you state.
It is, however, worth mentioning that FreeBSD has a unified buffer and
virtual memory page cache in these man pages. In fact, that does make
most uses of msync()'s MS_INVALIDATE flag unnecessary. (The useful cases
being things like the extension used by the Nvidia driver.)
On a related note, where do we stand with respect to the use of X/Open
man pages? The X/Open man page for msync(2) is much better than ours.
More information about the cvs-all