br2684: convert to net_device_ops broke it

[Mikko Vinni <mmvinni@yahoo.com>]

Hi,

my ADSL connection works well with kernel 2.6.29, but since commit
0ba25ff4c669e5395110ba6ab4958a97a9f96922 (br2684: convert to net_device_ops)
entered the mainline using br2684ctl fails with the following bug:

kernel: [  321.015643] ATM dev 0: ADSL line: training
kernel: [  323.011315] ATM dev 0: ADSL line: channel analysis
kernel: [  327.019647] ATM dev 0: ADSL line: exchange
kernel: [  328.013485] ATM dev 0: ADSL line: up (8000 kb/s down | 896 kb/s up)
br2684ctl[5606]: Interface "nas0" could not be created, reason: File exists 
br2684ctl[5606]: Communicating over ATM 0.0.33, encapsulation: LLC 
kernel: [  334.414469] br2684_create: register_netdev failed
kernel: [  334.421276] BUG: unable to handle kernel NULL pointer dereference at (null)
kernel: [  334.421298] IP: [<c0201544>] strncmp+0x14/0x30
kernel: [  334.421325] *pde = 00000000 
kernel: [  334.421336] Oops: 0000 [#1] 
kernel: [  334.421344] last sysfs file: /sys/module/br2684/initstate
kernel: [  334.421355] Modules linked in: cxacru usbatm br2684 binfmt_misc ipv6 radeon drm bridge stp llc bnep sco rfcomm l2cap sbs sbshc powernow_k7 lp joydev ppdev pcmcia snd_ali5451 snd_ac97_codec ac97_bus snd_pcm_oss snd_mixer_oss snd_pcm snd_seq_dummy snd_seq_oss snd_seq_midi snd_rawmidi snd_seq_midi_event snd_seq snd_timer snd_seq_device i2c_ali15x3 i2c_ali1535 snd parport_pc parport psmouse i2c_core soundcore snd_page_alloc yenta_socket rsrc_nonstatic pcmcia_core serio_raw container video shpchp pci_hotplug ati_agp agpgart evdev ext4 mbcache jbd2 crc16 usbhid hid natsemi ide_cd_mod cdrom ohci1394 ide_gd_mod ieee1394 ohci_hcd usbcore fuse
kernel: [  334.421493] 
kernel: [  334.421505] Pid: 5606, comm: br2684ctl Not tainted (2.6.29-rc2-702-ge0463f5 #17) HP nx9005 (DG832A)       
kernel: [  334.421518] EIP: 0060:[<c0201544>] EFLAGS: 00210016 CPU: 0
kernel: [  334.421529] EIP is at strncmp+0x14/0x30
kernel: [  334.421537] EAX: 00000000 EBX: dcc31b84 ECX: 0000000f EDX: d6658e78
kernel: [  334.421547] ESI: 00000000 EDI: d6658e78 EBP: d6658e54 ESP: d6658e4c
kernel: [  334.421556]  DS: 007b ES: 007b FS: 0000 GS: 0033 SS: 0068
kernel: [  334.421567] Process br2684ctl (pid: 5606, ti=d6658000 task=ccc48000 task.ti=d6658000)
kernel: [  334.421575] Stack:
kernel: [  334.421580]  d6658e78 400261f2 d6658eb8 de7e95dc d677c000 d677c000 00000000 d73acd80
kernel: [  334.421600]  d6658eb0 00000002 00000002 3073616e b7fe3f00 0804ac80 08048ac0 00000000
kernel: [  334.421619]  00000000 00000000 00000001 00000000 00000000 00000000 00000000 00000000
kernel: [  334.421637] Call Trace:
kernel: [  334.421646]  [<de7e95dc>] ? br2684_ioctl+0x20c/0x410 [br2684]
kernel: [  334.421675]  [<c030b9f7>] ? vcc_ioctl+0xa7/0x380
kernel: [  334.421701]  [<c0297af2>] ? sock_ioctl+0x52/0x240
kernel: [  334.421723]  [<c0297aa0>] ? sock_ioctl+0x0/0x240
kernel: [  334.421736]  [<c0193260>] ? vfs_ioctl+0x20/0x70
kernel: [  334.421758]  [<c0193732>] ? do_vfs_ioctl+0x72/0x580
kernel: [  334.421771]  [<c029a83c>] ? release_sock+0x7c/0x90
kernel: [  334.421785]  [<c030a044>] ? pvc_setsockopt+0x44/0x60
kernel: [  334.421798]  [<c029818f>] ? sys_setsockopt+0x5f/0xb0
kernel: [  334.421812]  [<c0299cc1>] ? sys_socketcall+0x231/0x270
kernel: [  334.421824]  [<c0193ca3>] ? sys_ioctl+0x63/0x70
kernel: [  334.421837]  [<c01034c5>] ? sysenter_do_call+0x12/0x25
kernel: [  334.421853] Code: 31 c0 eb 04 19 c0 0c 01 8b 34 24 8b 7c 24 04 c9 c3 8d b6 00 00 00 00 55 89 e5 83 ec 08 89 34 24 89 7c 24 04 89 c6 89 d7 49 78 08 <ac> ae 75 08 84 c0 75 f5 31 c0 eb 04 19 c0 0c 01 8b 34 24 8b 7c 
kernel: [  334.421936] EIP: [<c0201544>] strncmp+0x14/0x30 SS:ESP 0068:d6658e4c
kernel: [  334.421956] ---[ end trace c26391b26aac614a ]---
/USR/SBIN/CRON[5723]: (root) CMD ([ -x /usr/sbin/update-motd ] && /usr/sbin/update-motd 2>/dev/null)
kernel: [  557.287581] ACPI: EC: GPE storm detected, transactions will use polling mode


I traced the failing strncmp call to be in the br2684_find_dev function. With the following patch
the messages change a little bit, but the connection still doesn't work. It seems
that net_dev->name is not initialized, but I have no idea where it should be set.

diff --git a/net/atm/br2684.c b/net/atm/br2684.c
index 334fcd4..6bf59c7 100644
--- a/net/atm/br2684.c
+++ b/net/atm/br2684.c
@@ -133,8 +133,15 @@ static struct net_device *br2684_find_dev(const struct br2684_if_spec *s)
         break;
     case BR2684_FIND_BYIFNAME:
         list_for_each(lh, &br2684_devs) {
+            int skip = 0;
             net_dev = list_entry_brdev(lh);
-            if (!strncmp(net_dev->name, s->spec.ifname, IFNAMSIZ))
+            if(net_dev->name == NULL || s->spec.ifname == NULL) {
+                skip = 1;
+                printk(KERN_EMERG 
+                "br2684_find_dev: net_dev->name: %s, s->spec.ifname: %s\n",
+                net_dev->name, s->spec.ifname);
+            }
+            if (!skip && !strncmp(net_dev->name, s->spec.ifname, IFNAMSIZ))
                 return net_dev;
         }
         break;

And the messages:
kernel: [  143.369252] ATM dev 0: ADSL line: attempting to activate
kernel: [  147.589978] br2684_find_dev: net_dev->name: <NULL>, s->spec.ifname: nas0
kernel: [  147.590030] br2684: tried to attach to non-existant device
kernel: [  151.369922] ATM dev 0: ADSL line: down

Reverting 0ba25ff4c669e5395110ba6ab4958a97a9f96922 "fixes" the issue also
in v2.6.30-rc4 (output with some debug messages):

br2684ctl[3559]: Interface "nas0" created sucessfully 
br2684ctl[3559]: Communicating over ATM 0.0.33, encapsulation: LLC 
kernel: [  171.240807] br2684_ioctl cmd: 1073897971
kernel: [  171.240826] nas0 (): not using net_device_ops yet
kernel: [  171.246543] br2684_ioctl cmd: 1073897970
kernel: [  171.246553] br2684_regvcc entered
kernel: [  171.246556] br2684_regvcc going to find_dev
kernel: [  171.246561] br2684_regvcc before loop
kernel: [  171.246563] br2684_regvcc returning success
br2684ctl[3559]: Interface configured
br2684ctl[3566]: RFC 1483/2684 bridge daemon started 


I'll be glad to test any patches.

Mikko