Re: [Bug-readline] [PATCH] mm: msync: require either MS_ASYNC or MS_SYNC [resend]

[Peter Zijlstra]

On Fri, Apr 04, 2014 at 09:12:58AM +0200, Michael Kerrisk (man-pages) wrote:
> >   * Clearer intentions.  Looking at the existing code and the code
> >     history, the fact that flags=0 behaves like flags=MS_ASYNC appears
> >     to be a coincidence, not the result of an intentional choice.
> 
> Maybe. You earlier asserted that the semantics when flags==0 may have
> been different, prior to Peter Zijstra's patch,
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=204ec841fbea3e5138168edbc3a76d46747cc987
> .
> It's not clear to me that that is the case. But, it would be wise to
> CC the developer, in case he has an insight.

Right; so before that patch there appears to have been a difference.
The code looked like:

  if (flags & MS_ASYNC) {
        balance_dirty_pages_ratelimited();
  } else if (flags & MS_SYNC) {
        do_fsync()
  } else {
        /* do nothing */
  }

Which would give the following semantics:

  msync(.flags = 0) -- scan PTEs and update dirty page accounting
  msync(.flags = MS_ASYNC) -- scan PTEs and dirty throttle
  msync(.flags = MS_SYNC) -- scan PTEs and flush dirty pages

However with the introduction of accurate dirty page accounting in
.19 we always had an accurate dirty page count and both .flags=0 and
.flags=MS_ASYNC turn into the same NO-OP.

Yielding todays state, where 0 and MS_ASYNC don't do anything much and
MS_SYNC issues the fsync() -- although I understand Willy recently
posted a patch to do a data-range-sync instead of the full fsync.