swapper: page allocation failure.

[GoogleCodeExporter]

Version 1.9.2.7-rtn-r2944  for WL500gpv2 

Sometimes in logs gulls I notice an error: 

May 18 14:56:06 kernel: swapper: page allocation failure. order:3, mode:0x4020
May 18 14:56:06 kernel: Call Trace:
May 18 14:56:06 kernel: [<80340000>] setup_arch+0x2e4/0x408
May 18 14:56:06 kernel: [<c01387d8>] cleanup_module+0x1e60/0x3200 [et]
May 18 14:56:06 kernel: [<80340000>] setup_arch+0x2e4/0x408
May 18 14:56:06 kernel: [<80001524>] ret_from_irq+0x0/0x4
May 18 14:56:06 kernel: [<80340000>] setup_arch+0x2e4/0x408
May 18 14:56:06 kernel: [<8033cc5c>] start_kernel+0x320/0x4c8
May 18 14:56:06 kernel: [<8033c57c>] unknown_bootoption+0x0/0x34c
May 18 14:56:06 kernel: Free swap:        69960kB

For more information:

[admin@WL500GPV2-HOME /tmp]$ free
             total       used       free     shared    buffers     cached
Mem:         29016      26092       2924          0        908       3784
-/+ buffers/cache:      21400       7616
Swap:        72252       3740      68512

[admin@WL500GPV2-HOME /tmp]$ ps -aux
Warning: bad ps syntax, perhaps a bogus '-'? See http://procps.sf.net/faq.html
USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
admin        1  0.0  0.9   1536   264 ttyS0    Ss+  May16   0:01 /sbin/init
admin        2  0.0  0.0      0     0 ?        S<   May16   0:00 [kthreadd]
admin        3  0.0  0.0      0     0 ?        S<   May16   0:00 [ksoftirqd/0]
admin        4  0.0  0.0      0     0 ?        S<   May16   0:00 [events/0]
admin        5  0.0  0.0      0     0 ?        S<   May16   0:00 [khelper]
admin       23  0.0  0.0      0     0 ?        S<   May16   0:00 [kblockd/0]
admin       45  0.0  0.0      0     0 ?        S    May16   0:00 [pdflush]
admin       46  0.0  0.0      0     0 ?        S    May16   0:00 [pdflush]
admin       47  0.0  0.0      0     0 ?        S<   May16   0:00 [kswapd0]
admin       48  0.0  0.0      0     0 ?        S<   May16   0:00 [aio/0]
admin       97  0.0  0.0      0     0 ?        S<   May16   0:01 [mtdblockd]
admin      193  0.0  0.5   1604   148 ?        Ss   May16   0:00 syslogd -m 0 
-O /tmp/syslog.log -S -D -l 7 -b 1
admin      196  0.0  0.5   1604   160 ?        Ss   May16   0:00 klogd
admin      198  0.0  0.5    976   160 ?        S    May16   0:00 eapd
admin      201  0.0  0.9   1196   284 ?        S    May16   0:01 nas
admin      210  0.0  0.3   1604    92 ?        Ss   May16   0:00 telnetd
admin      212  0.0  0.8   1216   256 ?        S    May16   0:00 dropbear -4
admin      215  0.0  1.0   1160   300 ?        S    May16   2:14 httpd vlan1
nobody     224  0.0  1.3    956   380 ?        S    May16   0:45 dnsmasq
admin      226  0.0  0.6   1056   192 ?        S    May16   0:00 lld2d br0 eth1
admin      233  0.0  0.0      0     0 ?        S<   May16   0:00 [khubd]
admin      280  0.0  0.5    844   152 ?        Ss   May16   0:00 lpd
admin      282  0.0  0.6    740   180 ?        Ss   May16   0:00 p9100d -f 
/dev/lp0 0
admin      314  0.0  0.0      0     0 ?        S<   May16   0:00 [scsi_eh_0]
admin      315  0.0  0.0      0     0 ?        S<   May16   0:00 [usb-storage]
admin      342  0.0  0.7   1616   212 ?        Ss   May16   0:00 /sbin/udhcpc 
-i vlan1 -p /var/run/udhcpc0.pid -b -O routes -O staticroutes -O msstaticroutes
admin      343  0.0  1.1   1556   332 ?        Ss   May16   1:46 watchdog
admin      366  0.0  0.0      0     0 ?        S<   May16   0:00 [kjournald]
admin      368  0.0  0.0      0     0 ?        S<   May16   0:00 [kjournald]
admin      371  0.0  0.5   1032   152 ?        Ss   May16   0:00 
/usr/sbin/vsftpd
admin      373  0.0  2.1   1820   632 ?        Ss   May16   0:06 /usr/sbin/nmbd 
-D
admin      375  0.0  1.5   2044   444 ?        Ss   May16   0:00 /usr/sbin/smbd 
-D
admin      389  0.0 31.8  12800  9244 ?        S    May16   1:16 upnp -D -L br0 
-W vlan1
admin      500  0.7  1.9   1288   580 ?        Rs   15:07   0:04 dropbear -4
admin      501  0.1  1.6   1620   492 pts/0    Ss   15:09   0:00 -sh
admin      509  2.5  1.7   1524   512 pts/0    R+   15:17   0:00 ps -aux

Original issue reported on code.google.com by Vladimir...@gmail.com on 18 May 2011 at 11:18

[GoogleCodeExporter]

1) provide output of cat /proc/meminfo

2) will turning off upnp will help?

Original comment by lly.dev on 18 May 2011 at 11:28

[GoogleCodeExporter]

Since upgrading my wl500w to 1.9.2.7-rtn-r2972 I've also been getting alot of 
these errors in my logs.

Here's a copy of my /proc/meminfo

[admin@Router root]$ free
             total       used       free     shared    buffers     cached
Mem:         29016      27872       1144          0        628       4152
-/+ buffers/cache:      23092       5924
Swap:       506008       7204     498804
[admin@Router root]$ cat /proc/meminfo
MemTotal:        29016 kB
MemFree:          1076 kB
Buffers:           588 kB
Cached:           4136 kB
SwapCached:       2440 kB
Active:          15028 kB
Inactive:          804 kB
SwapTotal:      506008 kB
SwapFree:       498816 kB
Dirty:               0 kB
Writeback:           0 kB
AnonPages:       11100 kB
Mapped:           3672 kB
Slab:             4444 kB
SReclaimable:     1220 kB
SUnreclaim:       3224 kB
PageTables:        388 kB
NFS_Unstable:        0 kB
Bounce:              0 kB
CommitLimit:    520516 kB
Committed_AS:    19336 kB
VmallocTotal:   786356 kB
VmallocUsed:      4092 kB
VmallocChunk:   780888 kB

Original comment by jange...@gmail.com on 30 May 2011 at 8:01

[GoogleCodeExporter]

Since you are not a Vladimir, info provided are useless. From which version you 
are upgraded? Is torrents active? "ps" output? and more, more ...

P.S. 500M swap on device with 32M RAM is wrong, please read wl500g.info

Original comment by lly.dev on 30 May 2011 at 8:14

[GoogleCodeExporter]

Issue 231 has been merged into this issue.

Original comment by lly.dev on 30 May 2011 at 4:39

[GoogleCodeExporter]

I moved from RT-N-1.9.2.7-rtn-r2775 which had the same problem (which I was too 
lazy to report. Before that I used releases with the old kernel, I didn't have 
this issue onboard.

Original comment by vladimir...@gmail.com on 30 May 2011 at 6:13

Attachments:

• ps.txt
• meminfo.txt

[GoogleCodeExporter]

BTW,

Torrent client is not used, I only use it as a router; 

UPNP is a must, I have some VoIP box which uses it and I also run FTP server on 
some machine which also utilizes UPNP. I can try but can't leave the VoIP off 
for a couple of days (I really use it).

WBR,

Vladimir

Original comment by vladimir...@gmail.com on 30 May 2011 at 6:27

[GoogleCodeExporter]

Same question as to Vladimir - will turning off upnp will help? If yes, try 
last night snapshot from 
http://wpte.kicks-ass.net/downloads/Oleg%20Firmware/Nightlys/1.9.2.7-rtn/

Will disable LLTD responder helps?

P.S. Old kernel 2.4-based FW has significant difference against "rtn" branch

Original comment by lly.dev on 30 May 2011 at 6:31

[GoogleCodeExporter]

>Same question as to Vladimir - will turning off upnp will help? If yes, try 
last night >snapshot from 
http://wpte.kicks-ass.net/downloads/Oleg%20Firmware/Nightlys/1.9.2.7-rtn/

I while poorly represent to myself why this demon in general worked also what 
my actions could lead to his start. Now this problem didn't repeat, on it I and 
didn't stitch a router expecting a repeated case of falling. I can start him 
and if it probably to keep core dump or a log's  file if it in general is 
possible.

Original comment by Vladimir...@gmail.com on 31 May 2011 at 4:48

[GoogleCodeExporter]

Meanwhile I run 19 hrs with TTLD and UPNP off, no problem so far. Not much but 
so is the run...

Maybe the logging itself causes memory to overflow and require allocation which 
is not possible? (wild guess)

BR,
Vladimir

Original comment by vladimir...@gmail.com on 31 May 2011 at 5:56

[GoogleCodeExporter]

I can attach USB to serial adapter (Aten UC232A) to the box and have the serial 
connected to PC for logging if needed, just need the drivers and how-to for 
patching.

The problem is reproducible but takes lots of time to.

Tell me what you need for a trace.

BR,

Vladimir

Original comment by vladimir...@gmail.com on 31 May 2011 at 6:01

[GoogleCodeExporter]

Unfortunately, libnetconf+upnp has leaks on 2.6 kernel (new iptables 1.4.x).

In last builds upnp replaced with miniupnpd - it has much more capabilities.

Original comment by lly.dev on 31 May 2011 at 6:02

[GoogleCodeExporter]

I see,

So your advice is to move on to the next release rather than continue testing 
current (to clarify which module to blame LLTD/UPNP? Currently both disabled, I 
still can wait 24HRS and start switching on one by one.
If a new build, which one? Should I use the stated night-build from post #8?

BR

Vladimir

Original comment by vladimir...@gmail.com on 31 May 2011 at 6:14

[GoogleCodeExporter]

Console adapter useful for kernel bugs only, while we have user-space memory 
leak in this issue.

Yes, you have two ways:
1) try to use r3058, for example
2) wait for new stable build (approx. one month)

Original comment by lly.dev on 31 May 2011 at 6:19

• Changed state: Accepted

[GoogleCodeExporter]

I only test it as internet GW for ethernet access with upnp/igmp/udpxy/dmz and 
static port forwarding;

I can try using usb-3g dongle, 8021.q/p, samba, NFS, speed benchmarking, and 
and other stuff, but only for short term runs as I don't have a spare piece.

If you have a qual procedure, post it.

Original comment by vladimir...@gmail.com on 31 May 2011 at 6:21

[GoogleCodeExporter]

Unfortunately, we haven't full QC procedure, due it requires too much hardware 
& complex test-cases.

Original comment by lly.dev on 31 May 2011 at 6:25

[GoogleCodeExporter]

In case you need a contributor for qual AIs on wl500gpv1,
I have the following:
1. managed dlink switch with most 802.1 onboard
2. WL500GPv1
3. 3 PCs with MS XP,7 and BT4 and also 1 MAC with SL all with both 802.11 and 
802.3
4. Huawei 3G usb datacard
5. MFU with usb and network support
6. IPTV by UDP cast from the ISP
7. ISP's Intranet for PtP stress testing

The things I miss are wimax card, adsl modem and dslam to cover all wl500gp 
features, and jtag for disaster recovery...
If you plan to implement VPN client I also have a peer to cisco and checkpoint 
sites.

The bottleneck is my free time.

Original comment by vladimir...@gmail.com on 31 May 2011 at 7:29

[GoogleCodeExporter]

Big thanks for help suggestion. It will be excellent if someone could do 
independent tests. I suppose it is better to discuss details in wl500g.info 
forum or private e-mail - me & themiron.

Original comment by lly.dev on 1 Jun 2011 at 8:21

[GoogleCodeExporter]

Fixed since r3009 (upnpd replaced with miniupnpd)

Original comment by lly.dev on 15 Jun 2011 at 4:49

• Changed state: Fixed

[GoogleCodeExporter]

Hi,

UPnP does not work correctly, application that uses UPnP returns an error while 
freeing the assigned ports and the ports get stuck.

WBR,

Vladimir

Original comment by vladimir...@gmail.com on 15 Jun 2011 at 10:27

[GoogleCodeExporter]

Only I can say - many people confirm that miniupnpd works fine.

Original comment by lly.dev on 16 Jun 2011 at 6:12

[GoogleCodeExporter]

please provide test case to reproduce the bug encountered

Original comment by themiron.ru on 16 Jun 2011 at 7:38

[GoogleCodeExporter]

Hi,

Sorry for long gaps between the posts - I've been a lil' busy these days.

I ran r3058, which is newer than r3009;

If I try to use UPnP the FTP becomes unreachable. Static port forwarding is a 
bad workaround...

Original comment by vladimir...@gmail.com on 18 Jun 2011 at 8:54

Attachments:

• scenario.vsd

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

UPnP app error log (tried for 12 hours, currently static) easily reproducible.

Original comment by vladimir...@gmail.com on 18 Jun 2011 at 9:11

Attachments:

• upnp_on.png

[GoogleCodeExporter]

BTW WAN IP reported by UPnP is non-configurable and differ from real (which is 
normal condition for such environment) so RFE to the next MRP

Original comment by vladimir...@gmail.com on 18 Jun 2011 at 9:32

[GoogleCodeExporter]

First of all, please attach syslog.log - probably may be conflict of double 
forwarded 21 port.

Is it possible to convert scenario.vsd from M$ Visio proprietary format to any 
other which can be opened with free product?

Original comment by lly.dev on 19 Jun 2011 at 9:14

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

Hi,

Scenario drawing converted;
Will reproduce the scenario and send you the log at next possible time-frame.
I have to engage another server or any app that uses port21 in order to justify 
double forwarding assumption, but I do not have any.

Original comment by vladimir...@gmail.com on 19 Jun 2011 at 7:09

Attachments:

• scenario.jpg

[GoogleCodeExporter]

Please clarify what is it "ISP" device? ADSL modem in bridge mode? Or something 
else?

Original comment by lly.dev on 20 Jun 2011 at 6:19

[GoogleCodeExporter]

ISP has routed MEN with dedicated IP and full NAT

Original comment by vladimir...@gmail.com on 20 Jun 2011 at 7:49

[GoogleCodeExporter]

i.e. your ISP uses ugly scheme(NAT 1:1) in that you hasn't real IP on wl500gpv1 
WAN interface? And your LAN is double NAT'ed?
In such case miniupnpd unable to detect real IP, it is expected behavior.

Original comment by lly.dev on 20 Jun 2011 at 8:00

[GoogleCodeExporter]

the only solution I see in this case - to add support for manual additional WAN 
address to get router aware of NAT 1:1 scheme. ddns, upnp, vserver, dmz will be 
affected.
what do you think?

Original comment by themiron.ru on 20 Jun 2011 at 10:26

[GoogleCodeExporter]

Yes, I suppose this is the only solution.
BTW ddns works fine now, it learns the wan IP using a dialogue from source 
(which is elegant and would be tasty to have using one of the million free find 
my IP services) but since wan address is static in my explicit case, I don't 
have much use of ddns.

NAT1:1 scheme is commonly used in order to implement MAN services such as ptp 
and media streaming so I think adding a possibility for manual wan ip 
definition for upnp will be widely used.
NAT options should be kept.

Original comment by vladimir...@gmail.com on 20 Jun 2011 at 1:43

[GoogleCodeExporter]

no, nat1:1 scheme is neither commonly nor widely used. it's really ugly by 
design, because external ip doesn't belong to client interface.

Original comment by themiron.ru on 21 Jun 2011 at 6:42

[GoogleCodeExporter]

Absolutely agree with theMIROn.

About ddns - agent on router(inadyn) itself doesn't know his public IP, it 
simply ask this info on remote server. In case of upnp/vserver/dmz - we hasn't 
such helper server at all.

Original comment by lly.dev on 21 Jun 2011 at 7:03

[GoogleCodeExporter]

lly, i guess folowing will be more than enought:
ip addr add <additional_ip> dev wan_if
iptables -t nat -I PREROUTING 1 -d <additional_ip> -j VSERVER
iptables -t nat -I POSTROUTING 1 -s ! <additional_ip> -j MASQUERADE
and put <additional_ip> to /etc/miniupnpd.conf, if desired

Original comment by themiron.ru on 21 Jun 2011 at 7:54