msync()

by James L. McGi » Fri, 04 Aug 1995 04:00:00

I am trying to use the 'lmbench-1.0' package but it will not build...
evidently linux libraries / headers do not support the msync() call
(According to my Sytem-Vee box, that's \"synchronize memory with
physical storage\")

Should I hack this to some other system call or are my linux
installations broken?

d-e-a-d-i-c-a-t-e-d---[ http://www.conservatory.com/~fishbowl ]-----l-i-g-h-t

Honesty pays, but it doesn't seem to pay enough to suit some people.
-- F. M. Hubbard

Top

1. mmap() & msync() question

Hi All,

If I am using mmap with the MAP_SHARED flag and if a
process updates stuff in its address space, does the kernel
write back the changes to the file on disk without any
intervention by the programmer?.

If so, how often does it write back to the disk?. Is there
any way I can control this?.

If not, do I have to use msync() to force a write back everytime?.

If I use msync() and provide the address range that needs to be
written back to disk, does msync() validate which pages are dirty
before flushing them back to disk or does it not care and try to
write back each and every page within that address range?.

If I set MS_SYNC in msync(), does the process block until all write
operations are completed?. How does this reflect on the fact that I
have used MAP_SHARED in mmap() call?.

Thank you all for your time and consideration.

Regards,
Harish

2. driverfs API Updates

3. msync errors compiling GNU binutils

4. Is syslog() a time/cpu waster?

5. msync() bug

6. Ethernet and PPP

7. usage and example of mmap,msync,sbrk....

8. gated.conf, Does anyone have this puppy working on a linux box.

9. msync writing when MS_INVALIDATE set and memory locked

10. msync() does not work reliably for pages created by ftruncate()

11. Tru64 5.1a msync causes corrupt files (sometimes)