Network doesn't work after resuming

[Joe Barnett]

network does not seem to work after resume (b44 module). If I rmmod ehci_hcd and
uhci_hcd and manually echo "mem" into /sys/power/state, network is fine after
resuming. maybe it's in the ifdown/ifup stuff done by acpi-support (I'm using
thom's NetworkManager packages, and none of my interfaces are set to auto).

If I try to ifup eth0 after resume using the acpi-support scripts: i get:

SIOCSIFFLAGS: Cannot allocate memory
SIOCSIFFLAGS: Cannot allocate memory
Listining on LPF/eth0/MAC_ADDRESS
Sending on LPF/eth0/MAC_ADDRESS
Sending on Socket/fallback
receive_packet failed on eth0: Network is down
DHCPDISCOVER on eth0 to 255.255.255.255 port 67 interval 7
send_packet: Network is down

and in kern.log, I get:
Jan 14 09:54:58 localhost kernel: [sys_ioctl+217/542] sys_ioctl+0xd9/0x21e
Jan 14 09:54:58 localhost kernel: [sysenter_past_esp+82/117]
sysenter_past_esp+0x52/0x75
Jan 14 09:54:58 localhost kernel: ifconfig: page allocation failure. order:8,
mode:0x21
Jan 14 09:54:58 localhost kernel: [__alloc_pages+439/859] __alloc_pages+0x1b7/0x35b
Jan 14 09:54:58 localhost kernel: [__get_free_pages+37/63]
__get_free_pages+0x25/0x3f
Jan 14 09:54:58 localhost kernel: [dma_alloc_coherent+204/259]
dma_alloc_coherent+0xcc/0x103
Jan 14 09:54:58 localhost kernel: [pg0+553913271/1069958144]
b44_alloc_consistent+0xc3/0x18b [b44]
Jan 14 09:54:58 localhost kernel: [pg0+553914516/1069958144] b44_open+0x21/0xeb
[b44]
Jan 14 09:54:58 localhost kernel: [__wake_up_locked+42/46]
__wake_up_locked+0x2a/0x2e
Jan 14 09:54:58 localhost kernel: [dev_open+121/134] dev_open+0x79/0x86
Jan 14 09:54:58 localhost kernel: [dev_mc_upload+37/67] dev_mc_upload+0x25/0x43
Jan 14 09:54:58 localhost kernel: [dev_change_flags+81/288]
dev_change_flags+0x51/0x120
Jan 14 09:54:58 localhost kernel: [dev_load+38/112] dev_load+0x26/0x70
Jan 14 09:54:58 localhost kernel: [devinet_ioctl+565/1412]
devinet_ioctl+0x235/0x584
Jan 14 09:54:58 localhost kernel: [inet_ioctl+94/158] inet_ioctl+0x5e/0x9e
Jan 14 09:54:58 localhost kernel: [sock_ioctl+203/573] sock_ioctl+0xcb/0x23d
Jan 14 09:54:58 localhost kernel: [sys_ioctl+217/542] sys_ioctl+0xd9/0x21e
Jan 14 09:54:58 localhost kernel: [sys_newuname+56/96] sys_newuname+0x38/0x60
Jan 14 09:54:58 localhost kernel: [sysenter_past_esp+82/117]
sysenter_past_esp+0x52/0x75
Jan 14 09:54:58 localhost kernel: ifconfig: page allocation failure. order:8,
mode:0x21
Jan 14 09:54:58 localhost kernel: [__alloc_pages+439/859] __alloc_pages+0x1b7/0x35b
Jan 14 09:54:58 localhost kernel: [__get_free_pages+37/63]
__get_free_pages+0x25/0x3f
Jan 14 09:54:58 localhost kernel: [dma_alloc_coherent+204/259]
dma_alloc_coherent+0xcc/0x103

[Joe Barnett (thejoe)]

This is on a dell Insipiron 8600

[Matthew Garrett (mjg59)]

Argh. This is due to weirdness in the b44 driver. There's a kernel-level
workaround - I'll try to make sure that appears in the next version of the kernel.

(Unrelatedly, does ACPI sleep work if you disable video POSTing, but leave VBE
and DPMS turned on?)

[Matthew Garrett (mjg59)]

I've submitted a patch to the kernel maintainer - it ought to go in in the next
couple of days.

[Joe Barnett (thejoe)]

Thanks Matt.

re: VBE stuff: see my coments in bug #11919

[Thom May (thombot)]

* More Matthew Garrett's crack:
    - Fix b44 driver DMA mask:
      . Add patch b44.dpatch

from 2.6.10-9 Changelog.

Has this fixed the problems you were seeing?

[Joe Barnett (thejoe)]

i actually haven't seen this bug since i took the steps in comment #2 of bug
#11919 (video post false, save/restore vbestate true). I'm also not currently
suspending/resuming, as i've been using the fglrx drivers, which don't support
it. I can switch to ati driver (or is it radeon driver -- and what's the
difference between the two?) and swap the videopost/vberestores sometime in the
next few days and see if i can get it to reproduce on latest 2.6.10-3 kernel if
you'd like me to.

[Matthew Garrett (mjg59)]

Ought to have been fixed with the b44 patch