Re: Severe Linux 2.4 kernel memory leakage

Chris Chabot (chabotc@reviewboard.com)
25 Nov 2001 16:39:56 +0100


Hi, It's almost nice to hear i am not the only person with this problem
;-)

After reading your email, the common factors seem to be:

- Dual P3 Setup (diff mobo, same intel chipset?)
- Software Raid
- Mixed IDE / SCSI
- Internal Adaptec AHA-29xx
(i have the 80mb, you the 160mb version)
- Multiple network cards
- Intel EtherExpress Pro 10/100
- ECC checking ram

Since you don't use weird routes, nor iptables, i think it's posible to
assume these do not cause the problem. Also since you do not have the
problem under 2.4.9 (ditto for me if i remeber correctly), it is safe to
assume the bug was introduced in kernel 2.4.10 or up.

ps, the reason why i imidiatly switched to 2.4.11 (and up) and not have
a lot of experiance with 2.4.9 is because my dell servers with adaptect
hardware raid are a LOT (> 100%) faster under those newer kernels.
Somehow a block layer change in that kernel speeded up the dells a lot..
and since i love consitent kernel versions accross all my machines, i
upgraded my own boxes as well.

-- Chris

On Sun, 2001-11-25 at 16:27, Peter T. Breuer wrote:
> "A month of sundays ago Chris Chabot wrote:"
> > The box has ran Redhat 7.1 and 7.2, with plain vanilla linux kernels
> > 2.4.9 upto 2.4.15, in all situations the same problem appeared.
> >
> > The problem is that when the box boots up, it uses about 60Mb of memory.
> > However after only 1 1/2 days, the memory usage is already around 430Mb
> > (!!). (this is ofcource used - buffers - cache, as displayed by 'free').
>
> I also have this problem. Unknown circumstances provoke it. Kernel
> 2.4.9 to 2.4.13. When it occurs I lose about 30MB a day.
>
> Dual 500MHz i686, 4 scsi disks (adaptec) under raid5 and raid0
> with 2 intelpro's and 1 IDE disk (and xfs and lvm).
>
> Right now I'm on 2.4.9 and it's NOT happening. Doing nothing different
> to any other day.
>
> > When the box keeps on running for about a month, the memory usage gets
> > so high that it turns into a swap-crazy, low-memory and slow server ;-/
> > (it does free up cache memory, and swaps stuff out, however the 'leaked'
> > memory only grows and is never re-claimed).
>
> Same.
>
> > The box runs dhcpd, bind, fetchmail (cron), pppd (to adsl modem), smb,
> > nfs, xinetd (imapd mostly) and sshd.
>
> Only thing in common with me is nfs. Running X 4.1. glibc 2.1.
>
> > based routing) for my cable modem & adsl modem. Also it has a 310Gb raid
> > 0 array on 4 IDE disks.
>
> Could be.
>
> > The hardware on the box is : Asus p2b-ds, 2x p3-600, 1Gb (ECC) ram, 3
>
> My mobo is whatever came from dell, and you also are running 2xP3. My
> ram is also ECC but there's only 128MB of it.
>
> > network cards (1x Intel EtherExpressPro, 2x 3c905 tx), Internal adaptect
>
> I have 2 network cards, both EEPRO.
>
> > 29xx u2w scsi, internal intel IDE, 2x Seagate Cheetah (u2w) 18 Gb disks
>
> Yep, I have internal adaptec too. Aic7xxx running ultra 160 at 20MHz
> on terminated cable.
>
> Adaptec AIC7xxx driver version: 6.2.1
> aic7892: Ultra160 Wide Channel A, SCSI Id=7, 32/255 SCBs
>
> 4 WD disks:
>
> Host: scsi0 Channel: 00 Id: 00 Lun: 00
> Vendor: WDIGTL Model: WDE9100 ULTRA2 Rev: 1.21
> Type: Direct-Access ANSI SCSI revision: 02
> Host: scsi0 Channel: 00 Id: 01 Lun: 00
> Vendor: WDIGTL Model: WDE9100 ULTRA2 Rev: 1.21
> Type: Direct-Access ANSI SCSI revision: 02
> Host: scsi0 Channel: 00 Id: 02 Lun: 00
> Vendor: WDIGTL Model: WDE9100 ULTRA2 Rev: 1.21
> Type: Direct-Access ANSI SCSI revision: 02
> Host: scsi0 Channel: 00 Id: 03 Lun: 00
> Vendor: WDIGTL Model: WDE9100 ULTRA2 Rev: 1.21
> Type: Direct-Access ANSI SCSI revision: 02
>
> > (/ and /var), 4x 80 Gb Maxtor IDE disks (raid 0 array) and a NVidia TNT2
> > card. This hardware
>
> Umm .. I think I run ati rage, external card, though there is one on
> the mobo.
>
> (--) PCI:*(0:16:0) ATI Mach64 GU rev 154, Mem @ 0xf5000000/24,
> 0xfe201000/12, I/O @ 0xd400/8
> (--) PCI: (1:0:0) ATI Mach64 GW rev 122, Mem @ 0xfc000000/24,
> 0xfbfff000/12, I/O @ 0xec00/8
>
> > The kernel is compiled with all network- and scsi card and raid0 drivers
> > build in, and nfs + iptables as modules. The machine currently uses ext3
>
> I have it all compiled OUT. Including iptables, which I don't use.
>
> > (also build in), however this problem was also present before i
> > converted the raid0 volume to ext3, so i do not suspect it to cause this
>
> I am using xfs on top of lvm on top of raid5.
>
> > problem. The kernel is also set for HIGHMEM (4gb) to use the last Mb's
> > of the 1Gb of ram (else 127Mb isnt detected).
>
> Mine isn't. Normal setup.
>
> > I do not know which component (iptables / route hack / raid0 / network
> > cards / highmem) cause this problem. I run several of these components
>
> Looks from this as though it might be raid5 or 0 + adaptec scsi + SMP.
>
> Peter

-
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/