Driver getting into wierd state... stress test of 2002-10-12 version

Michael Milligan milli at acmeps.com
Sat Oct 26 02:56:44 EDT 2002


Hello,

Got around to stress testing the 2002-10-12 driver, which had been working 
flawlessly for a few weeks now.  Ran into some trouble with the driver 
freezing up on my server (hostap_pci) running in Master mode.  Client is 
also running hostap, but under PCMCIA (hostap_cs) and in Managed mode.  The 
driver didn't crash the server, just hosed up wlan0.  I'm using a 40-bit WEP 
key.

The server...  Dual-Pentium III box (SMP), kernel version 2.4.19, Debian 
"sarge" (a.k.a., testing), compiled Hostap by patching kernel.  Only change 
I made was to enable download support in the hostap_config.h file.  Only 
running hostap_pci.  The card...  Intel 2011B PCI card (which is a MiniPCI 
card in a carrier, Prism 2.5 chipset).

# lsmod
Module                  Size  Used by    Not tainted
hostap_crypt_wep        3020   1 (autoclean)
hostap_pci             43028   1
hostap                 54852   0 [hostap_pci]
hostap_crypt            1776   0 [hostap_crypt_wep hostap]
...

# dmesg
...
hostap_pci: hostap_pci.c 0.0.0 2002-10-12 (SSH Communications Security Corp, 
Jouni Malinen)
hostap_pci: (c) Jouni Malinen <jkmaline at cc.hut.fi>
hostap_pci: Registered netdevice wlan0
prism2_hw_init()
prism2_hw_config: initialized in 17316 iterations
wlan0: trying to read PDA from 0x007f0000: failed
wlan0: trying to read PDA from 0x003f0000: failed
wlan0: trying to read PDA from 0x00390000: failed
wlan0: NIC: id=0x8013 v1.0.0
wlan0: PRI: id=0x15 v1.0.7
wlan0: STA: id=0x1f v1.3.5
wlan0: defaulting to host-based encryption as a workaround for firmware bug 
in Host AP mode WEP
wlan0: LinkStatus=2 (Disconnected)
wlan0: Intersil Prism2.5 PCI: mem=0xfc0ff000, irq=18
wlan0: dropped received packet from non-associated STA 00:05:5d:da:c5:0c 
(type=0x02, subtype=0x04)
handle_ap_item - data frame
wlan0: prism2_send_mgmt - device is not UP - cannot send frame
wlan0: LinkStatus=2 (Disconnected)
wlan0: LinkStatus=2 (Disconnected)
hostap_crypt: registered algorithm 'WEP'
wlan0: LinkStatus=2 (Disconnected)
wlan0: prism2_open
wlan0: dropped received packet from non-associated STA 00:05:5d:da:c5:0c 
(type=0x02, subtype=0x00)
handle_ap_item - data frame
wlan0: 00:05:5d:da:c5:0c auth (alg=0 trans#=1 stat=0 len=6fc=00b0) ==> 0 
(authenticated)
wlan0: 00:05:5d:da:c5:0c assoc (len=22 prev_ap=00:00:00:00:00:00) => 0(0) 
(OK, new AID)
wlan0: STA 00:05:5d:da:c5:0c TX rate lowered to 55
wlan0: STA 00:05:5d:da:c5:0c TX rate raised to 110
wlan0: STA 00:05:5d:da:c5:0c TX rate lowered to 55
wlan0: STA 00:05:5d:da:c5:0c TX rate raised to 110
wlan0: STA 00:05:5d:da:c5:0c TX rate lowered to 55
wlan0: STA 00:05:5d:da:c5:0c TX rate raised to 110
wlan0: INFDROP event
wlan0: INFDROP event
...

The client... laptop running same kernel (but not SMP).  The card... Dlink 
DWL-650 card.  The laptop never froze up.  Resetting the server side always 
resumed connectivity.

# lsmod
Module                  Size  Used by    Tainted: PF
hostap_cs              43696   1
hostap                 49156   0 [hostap_cs]
hostap_crypt_wep        2956   1 (autoclean)
hostap_crypt            1360   0 [hostap hostap_crypt_wep]
openafs               410592   2
opl3sa2                 7856   2 (autoclean)
mpu401                 18404   0 (autoclean) [opl3sa2]
ad1848                 20684   0 (autoclean) [opl3sa2]
sound                  52332   2 (autoclean) [opl3sa2 mpu401 ad1848]
soundcore               3460   6 (autoclean) [sound]
xircom_tulip_cb        11512   0
ds                      6408   2 [hostap_cs]
yenta_socket            8704   2
pcmcia_core            38656   0 [hostap_cs ds yenta_socket]
ntfs                   48736   0 (unused)
vfat                    9276   0 (unused)
fat                    29112   0 [vfat]

# dmesg
...
wlan0: LinkStatus=4 (Access point out of range)
wlan0: LinkStatus: BSSID=00:05:3c:00:fb:c9
afs: Lost contact with file server 172.20.21.1 in cell ...
afs: Lost contact with file server 172.20.21.1 in cell ...
wlan0: LinkStatus=5 (Access point in range)
wlan0: LinkStatus: BSSID=00:05:3c:00:fb:c9
wlan0: LinkStatus=2 (Disconnected)
wlan0: LinkStatus: BSSID=00:05:3c:00:fb:c9
wlan0: LinkStatus=1 (Connected)
wlan0: LinkStatus: BSSID=00:05:3c:00:fb:c9
...
(repeated many times, corresponding to server resets)

# (from /var/log/messages)
Oct 25 21:17:45 chief cardmgr[341]: socket 1: Intersil PRISM2 11 Mbps 
Wireless Adapter
Oct 25 21:17:45 chief cardmgr[341]: executing: 'modprobe hostap_cs'
Oct 25 21:17:46 chief kernel: hostap_cs: hostap_cs.c 0.0.0 2002-10-12 (SSH 
Communications Security Corp, Jouni Malinen)
Oct 25 21:17:46 chief kernel: hostap_cs: (c) Jouni Malinen <jkmaline at cc.hut.fi>
Oct 25 21:17:46 chief kernel: hostap_cs: index 0x01: Vcc 3.3, irq 7, io 
0x0100-0x013f
Oct 25 21:17:46 chief kernel: hostap_cs: Registered netdevice wlan0
Oct 25 21:17:46 chief kernel: wlan0: NIC: id=0x800c v1.0.0
Oct 25 21:17:46 chief kernel: wlan0: PRI: id=0x15 v1.0.7
Oct 25 21:17:46 chief kernel: wlan0: STA: id=0x1f v1.3.5
Oct 25 21:17:46 chief /etc/hotplug/net.agent: invoke ifup wlan0


Stess testing...  ripping and enconding CDs (to ogg of course ;-) using a 
directory within the AFS file system on the server for temporary (.wav) and 
final storage (.ogg).  Also downloading files from the Internet (http) and 
reading mail (imap) all at the same time.  Found no correlation of any of 
these activities to the driver failures.  Also tried wget (http) directly 
between server and the client, saw between 550 and 600K/s transfer rates. 
That causes the problem after about 20Meg of a 100Meg file.

Log messages on server while under load:

Oct 25 21:56:01 phantom kernel: NETDEV WATCHDOG: wlan0: transmit timed out
Oct 25 21:56:01 phantom kernel: wlan0 Tx timed out! Resetting card
Oct 25 21:56:01 phantom kernel: hostap_pci: wlan0: resetting card
Oct 25 22:02:03 phantom kernel: NETDEV WATCHDOG: wlan0: transmit timed out
Oct 25 22:02:03 phantom kernel: wlan0 Tx timed out! Resetting card
Oct 25 22:02:03 phantom kernel: hostap_pci: wlan0: resetting card
Oct 25 22:02:24 phantom kernel: Already released txfid found at idx 5
Oct 25 22:02:24 phantom last message repeated 5 times
Oct 25 22:02:24 phantom kernel: hostap_pci: wlan0: resetting card
Oct 25 22:02:38 phantom ntpd[686]: synchronisation lost
Oct 25 22:06:32 phantom kernel: Already released txfid found at idx 0
Oct 25 22:06:33 phantom last message repeated 2 times
Oct 25 22:06:33 phantom kernel: hostap_pci: wlan0: resetting card
Oct 25 22:06:33 phantom kernel: prism2_enable_aux_port: was not disabled!?
Oct 25 22:06:33 phantom kernel: hostap_pci: Initialization failed

at this point, turning down the interface and rmmod'ing the hostap* modules, 
then bringing it back up clears things:

Oct 25 22:08:19 phantom kernel: hostap_pci: Netdevice wlan0 unregistered
Oct 25 22:08:19 phantom kernel: hostap_pci: Driver unloaded
Oct 25 22:08:19 phantom /etc/hotplug/net.agent: invoke ifdown wlan0
Oct 25 22:08:24 phantom kernel: hostap_pci: hostap_pci.c 0.0.0 2002-10-12 
(SSH Communications Security Corp, Jouni Malinen)
Oct 25 22:08:24 phantom kernel: hostap_pci: (c) Jouni Malinen 
<jkmaline at cc.hut.fi>
Oct 25 22:08:25 phantom kernel: hostap_pci: Registered netdevice wlan0
Oct 25 22:08:25 phantom kernel: wlan0: NIC: id=0x8013 v1.0.0
Oct 25 22:08:25 phantom kernel: wlan0: PRI: id=0x15 v1.0.7
Oct 25 22:08:25 phantom kernel: wlan0: STA: id=0x1f v1.3.5
Oct 25 22:08:25 phantom kernel: wlan0: Intersil Prism2.5 PCI: 
mem=0xfc0ff000, irq=18
Oct 25 22:08:25 phantom /etc/hotplug/net.agent: invoke ifup wlan0

Any ideas?  Need more info, lemme know.

Regards,
Mike

-- 
Michael Milligan  --  Free Agent  --  milli at acmeps.com




More information about the HostAP mailing list