#6 Ethernet dies on large file transfer

[Anonymous]

Hi.

I have a laptop which uses the et131x driver, when I transfer a large file across the LAN the transfer stalls and the connection is lost. I am now running Debian SID with all the latest updates, this problem persists on ubuntu from 7.04 to 8.04 as well as Debian Etch to Sid, I am using the latest driver version as well. I finally managed to catch this in my log files.

Jul 17 13:08:53 Prometheus kernel: [ 726.773925] ------------[ cut here ]------------
Jul 17 13:08:53 Prometheus kernel: [ 726.773930] WARNING: at kernel/irq/manage.c:425 free_irq+0x2b/0xd7()
Jul 17 13:08:53 Prometheus kernel: [ 726.773936] Modules linked in: binfmt_misc fglrx(P) rfcomm l2cap bluetooth ppdev parport_pc lp parport video
output powernow_k8 cpufreq_powersave cpufreq_stats cpufreq_userspace cpufreq_ondemand cpufreq_conservative freq_table ipv6 nls_utf8
ntfs nls_base ext3 jbd mbcache ndiswrapper loop firewire_sbp2 pcmcia firmware_class snd_hda_intel snd_pcm_oss
snd_mixer_oss snd_pcm snd_timer rtc pcspkr yenta_socket rsrc_nonstatic pcmcia_core psmouse serio_raw snd
k8temp soundcore snd_page_alloc i2c_piix4 i2c_core battery ac button et131x(P) shpchp pci_hotplug ati_agp agpgart
evdev reiserfs ide_cd_mod cdrom ide_disk ide_pci_generic atiixp ide_core firewire_ohci firewire_core sdhci crc_itu_t
mmc_core ata_generic libata scsi_mod dock ehci_hcd ohci_hcd usbcore thermal processor fan
Jul 17 13:08:53 Prometheus kernel: [ 726.774031] Pid: 0, comm: swapper Tainted: P 2.6.25-2-686 #1
Jul 17 13:08:53 Prometheus kernel: [ 726.774046] [<c0123f9f>] warn_on_slowpath+0x40/0x4f
Jul 17 13:08:53 Prometheus kernel: [ 726.774118] [<c0124745>] release_console_sem+0x18a/0x1a3
Jul 17 13:08:53 Prometheus kernel: [ 726.774232] [<c0124b95>] printk+0x14/0x18
Jul 17 13:08:53 Prometheus kernel: [ 726.774253] [<f8a48922>] DumpDeviceBlock+0xb9/0xca [et131x]
Jul 17 13:08:53 Prometheus kernel: [ 726.774295] [<c015414b>] free_irq+0x2b/0xd7
Jul 17 13:08:53 Prometheus kernel: [ 726.774323] [<f8a4786b>] et131x_close+0x8e/0xd2 [et131x]
Jul 17 13:08:53 Prometheus kernel: [ 726.774351] [<f8a4809f>] et131x_tx_timeout+0x23f/0x25d [et131x]
Jul 17 13:08:53 Prometheus kernel: [ 726.774521] [<c0262ea2>] dev_watchdog+0x0/0xae
Jul 17 13:08:53 Prometheus kernel: [ 726.774530] [<c0262f0b>] dev_watchdog+0x69/0xae
Jul 17 13:08:53 Prometheus kernel: [ 726.774550] [<c012b338>] run_timer_softirq+0x11a/0x17c
Jul 17 13:08:53 Prometheus kernel: [ 726.774562] [<c0262ea2>] dev_watchdog+0x0/0xae
Jul 17 13:08:53 Prometheus kernel: [ 726.774595] [<c0128309>] __do_softirq+0x66/0xd3
Jul 17 13:08:53 Prometheus kernel: [ 726.774615] [<c01283bb>] do_softirq+0x45/0x53
Jul 17 13:08:53 Prometheus kernel: [ 726.774625] [<c012865e>] irq_exit+0x35/0x67
Jul 17 13:08:53 Prometheus kernel: [ 726.774632] [<c01135aa>] smp_apic_timer_interrupt+0x6b/0x75
Jul 17 13:08:53 Prometheus kernel: [ 726.774650] [<c0108314>] apic_timer_interrupt+0x28/0x30
Jul 17 13:08:53 Prometheus kernel: [ 726.774694] [<c0117a1c>] native_safe_halt+0x2/0x3
Jul 17 13:08:53 Prometheus kernel: [ 726.774709] [<c01066c9>] default_idle+0x43/0x78
Jul 17 13:08:53 Prometheus kernel: [ 726.774718] [<c0106686>] default_idle+0x0/0x78
Jul 17 13:08:53 Prometheus kernel: [ 726.774730] [<c01065c9>] cpu_idle+0xa1/0xc1
Jul 17 13:08:53 Prometheus kernel: [ 726.774773] =======================
Jul 17 13:08:53 Prometheus kernel: [ 726.774778] ---[ end trace 0a22819bb5f578cd ]---

Any help would be appreciated on this. Oh, I'm running kernel 2.6.25-2

Chris.

[W Unruh]

Logged In: YES
user_id=2177620
Originator: NO

Chris: Totally on a wild guess, this might be another symptom of the bug I just reported with the add_timer funtion in the routine et131x_initpci.c

Change the line
add_timer( &pAdapter->ErrorTimer);
to something like
mod_timer( &pAdapter->ErrorTimer, jiffies + HZ );

This is definitely a bug in the program and the kernel statements that
Jul 17 13:08:53 Prometheus kernel: [ 726.774615] [<c01283bb>] do_softirq+0x45/0x53
makes me suspect that your problem is a timer problem.
Anyway, it cannot hurt (and fixes a bug which makes the driver swamp the cpu) and may help.

Bill