4.0 parisc regression: memory leak?

4.0 parisc regression: memory leak?

[Meelis Roos]

I noticed that with 4.0 I'm running out of memory on HP A500 (1G RAM) 
and HP RP3410 (2G RAM) parisc machines during gentoo "revdep-rebuild 
-ip". Each time free shows less free memory until OOM killer visits us 
(and first revdep-rebuild after nontrivial emerge compilation loop kills 
the smaller box).

3.19 did not seem to have this problem.

Nothing in the dmesg besides OOM killer activating.

I reran it, leaving cat /prov/vmstat, cat /proc/meminfo and slabtop -o 
running in loop (each 5 s). revdep-rebuild -ip was also running in 
the loop, just checking the binaries and library linking info.

The results are at

http://kodu.ut.ee/~mroos/slabtop-log
http://kodu.ut.ee/~mroos/meminfo-log
http://kodu.ut.ee/~mroos/vmstat-log

Anything else I can provide to pinpoint it?

-- 
Meelis Roos (mroos@linux.ee)

Re: 4.0 parisc regression: memory leak?

[John David Anglin]

On 2015-04-25, at 3:19 PM, Meelis Roos wrote:

> I noticed that with 4.0 I'm running out of memory on HP A500 (1G RAM) 
> and HP RP3410 (2G RAM) parisc machines during gentoo "revdep-rebuild 
> -ip". Each time free shows less free memory until OOM killer visits us 
> (and first revdep-rebuild after nontrivial emerge compilation loop kills 
> the smaller box).

I've also had problems with 4.0.  I had HPMC after a few hours on rp3440.

> 
> 3.19 did not seem to have this problem.

--
John David Anglin	dave.anglin@bell.net

Re: 4.0 parisc regression: memory leak?

[Meelis Roos]

> > I noticed that with 4.0 I'm running out of memory on HP A500 (1G RAM) 
> > and HP RP3410 (2G RAM) parisc machines during gentoo "revdep-rebuild 
> > -ip". Each time free shows less free memory until OOM killer visits us 
> > (and first revdep-rebuild after nontrivial emerge compilation loop kills 
> > the smaller box).
> 
> I've also had problems with 4.0.  I had HPMC after a few hours on rp3440.

My rp3440 worked fine for emerge world - maybe an hour or some hours. 
Because of 12G RAM, it did not run out of memory.

-- 
Meelis Roos (mroos@linux.ee)