2002-10-25 18:45:23

by Dave Jones

[permalink] [raw]
Subject: e100 doing bad things in 2.5.44.

Oct 25 18:38:12 tetrachloride kernel: Debug: sleeping function called from illegal context at mm/slab.c:1384
Oct 25 18:38:12 tetrachloride kernel: Call Trace:
Oct 25 18:38:12 tetrachloride kernel: [<c011dd94>] __might_sleep+0x54/0x58
Oct 25 18:38:12 tetrachloride kernel: [<c01403ae>] kmalloc+0x5a/0x314
Oct 25 18:38:12 tetrachloride kernel: [<c017e2da>] proc_create+0x76/0xcc
Oct 25 18:38:12 tetrachloride kernel: [<c017e3fb>] proc_mkdir+0x17/0x40
Oct 25 18:38:12 tetrachloride kernel: [<c010a4c3>] register_irq_proc+0x6b/0xb0
Oct 25 18:38:12 tetrachloride kernel: [<c010a2ba>] setup_irq+0x166/0x174
Oct 25 18:38:12 tetrachloride kernel: [<c02b9424>] e100intr+0x0/0x308
Oct 25 18:38:12 tetrachloride kernel: [<c0109b48>] request_irq+0x88/0xa4
Oct 25 18:38:12 tetrachloride kernel: [<c02b84fa>] e100_open+0x106/0x188
Oct 25 18:38:12 tetrachloride kernel: [<c02b9424>] e100intr+0x0/0x308
Oct 25 18:38:12 tetrachloride kernel: [<c03a5080>] dev_open+0x50/0xb0
Oct 25 18:38:12 tetrachloride kernel: [<c03a65e1>] dev_change_flags+0x51/0x104
Oct 25 18:38:12 tetrachloride kernel: [<c03decc5>] devinet_ioctl+0x331/0x6d8
Oct 25 18:38:12 tetrachloride kernel: [<c03e1bc7>] inet_ioctl+0xab/0xf0
Oct 25 18:38:12 tetrachloride kernel: [<c039e6a1>] sock_ioctl+0xfd/0x160
Oct 25 18:38:12 tetrachloride kernel: [<c0162cfb>] sys_ioctl+0x27f/0x2f5
Oct 25 18:38:12 tetrachloride kernel: [<c0108029>] error_code+0x2d/0x38
Oct 25 18:38:12 tetrachloride kernel: [<c01075c7>] syscall_call+0x7/0xb

Another weirdo.. Check out the Speed..

Oct 25 18:38:12 tetrachloride kernel: Intel(R) PRO/100 Network Driver - version 2.1.24-k1
Oct 25 18:38:12 tetrachloride kernel: Copyright (c) 2002 Intel Corporation
Oct 25 18:38:12 tetrachloride kernel:
Oct 25 18:38:12 tetrachloride kernel: e100: selftest OK.
Oct 25 18:38:12 tetrachloride kernel: e100: eth0: Intel(R) PRO/100 VE Network Connection
Oct 25 18:38:12 tetrachloride kernel: Mem:0xfeafc000 IRQ:20 Speed:0 Mbps Dx:N/A
Oct 25 18:38:12 tetrachloride kernel: Hardware receive checksums enabled


--
| Dave Jones. http://www.codemonkey.org.uk


2002-10-26 21:11:44

by Jeff Garzik

[permalink] [raw]
Subject: Re: e100 doing bad things in 2.5.44.

Dave Jones wrote:

>Oct 25 18:38:12 tetrachloride kernel: Debug: sleeping function called from illegal context at mm/slab.c:1384
>Oct 25 18:38:12 tetrachloride kernel: Call Trace:
>Oct 25 18:38:12 tetrachloride kernel: [<c011dd94>] __might_sleep+0x54/0x58
>Oct 25 18:38:12 tetrachloride kernel: [<c01403ae>] kmalloc+0x5a/0x314
>Oct 25 18:38:12 tetrachloride kernel: [<c017e2da>] proc_create+0x76/0xcc
>Oct 25 18:38:12 tetrachloride kernel: [<c017e3fb>] proc_mkdir+0x17/0x40
>Oct 25 18:38:12 tetrachloride kernel: [<c010a4c3>] register_irq_proc+0x6b/0xb0
>Oct 25 18:38:12 tetrachloride kernel: [<c010a2ba>] setup_irq+0x166/0x174
>Oct 25 18:38:12 tetrachloride kernel: [<c02b9424>] e100intr+0x0/0x308
>Oct 25 18:38:12 tetrachloride kernel: [<c0109b48>] request_irq+0x88/0xa4
>Oct 25 18:38:12 tetrachloride kernel: [<c02b84fa>] e100_open+0x106/0x188
>
>
Well, it holds bdp->isolate_lock for an incredibly long time, so that
might trigger this. At least interrupts aren't disabled.

Another bug: e100_close doesn't get the lock.


>Another weirdo.. Check out the Speed..
>
>Oct 25 18:38:12 tetrachloride kernel: Intel(R) PRO/100 Network Driver - version 2.1.24-k1
>Oct 25 18:38:12 tetrachloride kernel: Copyright (c) 2002 Intel Corporation
>Oct 25 18:38:12 tetrachloride kernel:
>Oct 25 18:38:12 tetrachloride kernel: e100: selftest OK.
>Oct 25 18:38:12 tetrachloride kernel: e100: eth0: Intel(R) PRO/100 VE Network Connection
>Oct 25 18:38:12 tetrachloride kernel: Mem:0xfeafc000 IRQ:20 Speed:0 Mbps Dx:N/A
>Oct 25 18:38:12 tetrachloride kernel: Hardware receive checksums enabled
>

Cosmetic or real, that's indeed another bug...

Jeff




2002-10-27 04:25:57

by Feldman, Scott

[permalink] [raw]
Subject: RE: e100 doing bad things in 2.5.44.

> Well, it holds bdp->isolate_lock for an incredibly long time, so that
> might trigger this. At least interrupts aren't disabled.
>
> Another bug: e100_close doesn't get the lock.

This came up earlier and we're working on getting rid of isolate_lock
altogether. It's a bunch of complication for no real benefit, really.

> > Oct 25 18:38:12 tetrachloride kernel: Mem:0xfeafc000
> > IRQ:20 Speed:0 Mbps Dx:N/A
>
> Cosmetic or real, that's indeed another bug...

We'll I guess no link would give you a speed of zero Mbps. ;)

I'm inclined to strike this message line because it's 1) misleading, 2)
redundant.

-scott