Re: free: regression due to a different calculation of Used memory

Hello everyone,

After some inspection, I have to stand behind Michal's opinion.
After discussion with our kernelist, Jerome Marchand, who stated that Shmem
is a subset of Cached (and not all cache pages are easily reclaimable) -
this can be tested by writing into huge file in tmpfs -, I learned that the
current calculation of Used does not seem to be valid.
I believe there were many reasons why this calculation changed to its
current form and I realize there is rather philosophical question what
'used' memory should be considered as (whether just allocated or
non-reclaimable), however, in the light of recent information, I suggest to
at least add a "--pure" (or so) switch to options which enables calculation
`used = total - free`, as all allocated memory is "somehow used", whether
reclaimable or not.

If you agree, I can send a MR with the patch (in order not to bloat gitlab
with MRs under discussion), anyway I'm certainly looking forward to reading
your opinion.
I think it's worth mentioning that since newlib's aim is to have ABI stable
library (IIRC), switching the calculation within free.c binary should be
considered, should the change be accepted. But I believe if the calculation
is really wrong, it ought to be corrected globally in the library.

Thanks.

Jan Rybar

On Tue, Jun 1, 2021 at 9:46 AM Michal Hocko <[email protected]>
wrote:

> On Thu 29-04-21 14:30:36, Michal Hocko wrote:
> > On Thu 29-04-21 22:08:31, Craig Small wrote:
> [...]
> > > proc(5) says Cached (as in the line /proc/meminfo ) is only based on
> disk,
> > > or in its words
> > >   "In-memory cache for files read from the disk (the page cache).
> Doesn't
> > > include SwapCached."
> > >
> > > Is the man page incorrect?
> >
> > yes and no. It is correct in the sense that swapcache is explicitly
> > excluded but it is not fully correct claiming it is disk backed file
> > cache. Because it includes shmem which is a filesystem which is not
> > backed by disk (except for the swapout).
> >
> > A proper wording would be: In-memory cache for filesystem content (both
> > disk based as memory based like shmem). Doesn't include Swap cache and
> > buffer cache which are presented separately (SwapCached resp. Buffers).
>
> Should I send a patch to clarify this?
>
> > > I think that's the main difference now.  Ideally this could be settled
> by
> > > the kernel exposing MemUsed in /proc/meminfo and everyone can use that
> > > (it's how available was solved).
> >
> > Why Total-Available doesn't serve that purpose already?
>
> Would be a patch to change the behavior to use Total-Available at least
> considered?
> --
> Michal Hocko
> SUSE Labs
>
>