lkml.org 
[lkml]   [2014]   [Apr]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH v2] ipc,shm: disable shmmax and shmall by default
From
Date
On Thu, 2014-04-17 at 22:23 +0200, Michael Kerrisk (man-pages) wrote:
> Hi Manfred!
>
> On Thu, Apr 17, 2014 at 6:22 PM, Manfred Spraul
> <manfred@colorfullife.com> wrote:
> > Hi Michael,
> >
> >
> > On 04/17/2014 12:53 PM, Michael Kerrisk wrote:
> >>
> >> On Sat, Apr 12, 2014 at 5:22 AM, Davidlohr Bueso <davidlohr@hp.com> wrote:
> >>>
> >>> From: Davidlohr Bueso <davidlohr@hp.com>
> >>>
> >>> The default size for shmmax is, and always has been, 32Mb.
> >>> Today, in the XXI century, it seems that this value is rather small,
> >>> making users have to increase it via sysctl, which can cause
> >>> unnecessary work and userspace application workarounds[1].
> >>>
> >>> Instead of choosing yet another arbitrary value, larger than 32Mb,
> >>> this patch disables the use of both shmmax and shmall by default,
> >>> allowing users to create segments of unlimited sizes. Users and
> >>> applications that already explicitly set these values through sysctl
> >>> are left untouched, and thus does not change any of the behavior.
> >>>
> >>> So a value of 0 bytes or pages, for shmmax and shmall, respectively,
> >>> implies unlimited memory, as opposed to disabling sysv shared memory.
> >>> This is safe as 0 cannot possibly be used previously as SHMMIN is
> >>> hardcoded to 1 and cannot be modified.
> >>>
> >>> This change allows Linux to treat shm just as regular anonymous memory.
> >>> One important difference between them, though, is handling out-of-memory
> >>> conditions: as opposed to regular anon memory, the OOM killer will not
> >>> free the memory as it is shm, allowing users to potentially abuse this.
> >>> To overcome this situation, the shm_rmid_forced option must be enabled.
> >>>
> >>> [1]: http://rhaas.blogspot.com/2012/06/absurd-shared-memory-limits.html
> >>>
> >>> Signed-off-by: Davidlohr Bueso <davidlohr@hp.com>
> >>> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> >>> Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
> >>
> >> Of the two proposed approaches (the other being
> >> marc.info/?l=linux-kernel&m=139730332306185), this looks preferable to
> >> me, since it allows strange users to maintain historical behavior
> >> (i.e., the ability to set a limit) if they really want it, so:
> >>
> >> Acked-by: Michael Kerrisk <mtk.manpages@gmail.com>
> >>
> >> One or two comments below, that you might consider for your v3 patch.
> >
> > I don't understand what you mean.
>
> As noted in the other mail, you don't understand, because I was being
> dense (and misled a little by the commit message).
>
> > After a
> > # echo 33554432 > /proc/sys/kernel/shmmax
> > # echo 2097152 > /proc/sys/kernel/shmmax
> >
> > both patches behave exactly identical.
>
> Yes.
>
> > There are only two differences:
> > - Davidlohr's patch handles
> > # echo <really huge number that doesn't fit into 64-bit> >
> > /proc/sys/kernel/shmmax
> > With my patch, shmmax would end up as 0 and all allocations fail.
> >
> > - My patch handles the case if some startup code/installer checks
> > shmmax and complains if it is below the requirement of the application.
>
> Thanks for that clarification. I withdraw my Ack.

:(

> In fact, maybe I
> even like your approach a little more, because of that last point.

And it is a fair point. However, this is my counter argument: if users
are checking shmmax then they sure better be checking shmmin as well! So
if my patch causes shmctl(,IPC_INFO,) to return shminfo.shmmax = 0 and a
user only checks this value and breaks the application, then *he's*
doing it wrong. Checking shmmin is just as important... 0 value is
*bogus*, heck it even says so in shmctl's manpage.

> Did
> one of you not yet manage to persuade the other to his point of view
> yet?

I think we've left that up to akpm.





\
 
 \ /
  Last update: 2014-04-18 01:01    [W:0.075 / U:0.172 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site