lkml.org 
[lkml]   [2002]   [Jul]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] VM accounting 1/3 trivial
On 23 Jul 2002, Alan Cox wrote:
> On Tue, 2002-07-23 at 18:27, Hugh Dickins wrote:
> > First of three patches against 2.4.19-rc3-ac3 fixing some VM accounting.
> > Could be split further, but this one too trivial to need much thought.
> >
> > 1. do_munmap doesn't need an extra acct arg (and rc3-ac3 still leaves
> > arch files without it): just clear VM_ACCOUNT in mremap's move_vma.
>
> Are you sure that is correct. I started off on that basis but never got
> it to work reliably when mremap changes multiple vmas ?

Thank you, it is surely incorrect (in the case where the do_munmap does
not cover the whole vma, leaving one or two pieces behind: I think that
must be the case you're remembering). Would a patch which (if necessary)
reapplies VM_ACCOUNT to the leftover piece(s) be welcome, or would it
just look like an ugly face-saving exercise?

> Can you split out items #2, #4 first of all and submit those alone, then
> I can review each item on its own and run vm_validate tests

Okay, will do, thanks. And you will also ignore patches 2/3, 3/3 now,
since David has given a clear vote for more MAP_NORESERVE consistency,
and VM_NORESERVE would be more natural to pass from do_mmap_pgoff to
shmem_file_setup than my VM_SHARED|VM_ACCOUNT abuse.

But I'll still have a consistency problem with MAP_NORESERVE versus
sysctl_overcommit_memory, when the latter is changed (> 1 or <= 1).
The closest I can get to a consistent position is that do_mmap_pgoff
only sets VM_NORESERVE if MAP_NORESERVE and sysctl_overcommit_memory
<= 1 (as you wish), but that thereafter (in mprotect and in mremap,
even when extending the mapping) VM_NORESERVE will be respected (and
propagated when splitting) even while sysctl_overcommit_memory > 1.
Mirroring the vice versa situation, of the reservations made when
MAP_NORESERVE is ignored, and thereafter.

You might prefer more draconian enforcment, but I fear it
could behave strangely. Does the above sound fair to you?

Hugh

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:22    [W:0.061 / U:0.516 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site