[linux-dvb] Re: dvb-bt8xx and net device

[Holger Waechtler <holger@convergence.de>]

Hi Steffen, André,

can you please try again and report if this oops is fixed?

Holger



André Stein wrote:
> Hi,
>
> > >     First of all, I always set the MAC address manually because after 
> > > ifconfig -a dvb0_0 the mac address was always inited to 0.. Then the 
> > > driver actually never oops. It simply "fucks" up the whole system and 
> > > I don't know why..
> > >
> > > I made a similar approach as you. I tuned and brought the net device 
> > > up with dvbtune
> > >  # dvbtune -f 10773250 -p H -s 22000 -n 251 -i
> > > I get a long list of service available at the Astranet frequence and 
> > > I try to use the T-DSL proxy service (pid 251). Finally the net 
> > > device is successfully setup.
> > >  # ifconfig dvb0_0 10.0.1.1
> > > and the net device is running.. The mac is also set correctly as I 
> > > hardcoded it in the add_if net code.
> > >  # iptraf -d dvb0_0
> > > gets me lots of UDP packages but before I can actually start the 
> > > proxy service, the system hangs and nothing can't be done anymore. If 
> > > you're lucky you get data lost because the hd buffers weren't flushed..
> > >
> > > So after that I am pretty sure that this nasty bug isn't related with 
> > > setting the mac address. As I said before, I always set the mac 
> > > before using the net device and ifconfig -a dvb0_0 showed me that 
> > > everything went fine. It must be in some way related to the hardware 
> > > as you're nova s seems to work and Stefan reported the same problem 
> > > as me with a pinnacle card.
> > > As you're familiar with the dvb interface, is there a possibilty that 
> > > a package buffer is flooded with data that actually makes the kernel 
> > > go nuts?
> >
> >
> > well, I fear that the actual bug is in the bt8xx driver, somewhere in 
> > DMA setup and interrupt handling.
> >
> > Is your system responding on SysRq-Keys after the lockup? Might it 
> > hang in a busy-loop somewhere, maybe in interrupt context? Try e.g. 
> > alt-sysrq-t or alt-sysrq-p to get an idea where the kernel is actually 
> > hanging. Remove the SA_INTERRUPT flag in request_irq() to enable other 
> > interrupts while processing bt878_irq().
> >
> > In any case you might want to set up a serial console on a second box 
> > in order to save the logs.
>     
>     Things seem to become stranger and stranger while I try the dvb 
> drivers out.. Because I had added some printk's in the net code, I 
> thought it would be a good idea to get a new, vanilla dvb tree. I did 
> that but when I tried the latest drivers out, I couldn't even bring the 
> netdevice up. I got that typical seg fault.
> Steffen, could you please try the latest dvb-kernel out and check 
> whether you get the same problem as me?
>
> cpu: 0, clocks: 1993363, slice: 996681
> ac97_codec: AC97 Audio codec, id: ??v9(SigmaTel STAC9721/23)
> Unable to handle kernel paging request at virtual address 2d343369
> c021ddac
> *pde = 00000000
> Oops: 0000
> CPU:    0
> EIP:    0010:[<c021ddac>]    Tainted: P
> Using defaults from ksymoops -t elf32-i386 -a i386
> EFLAGS: 00010212
> eax: 0000000a   ebx: 00000004   ecx: 00000001   edx: 00000008
> esi: 2d343369   edi: ca98f05c   ebp: 00000008   esp: cab5bdd4
> ds: 0018   es: 0018   ss: 0018
> Process dvbtune (pid: 374, stackpage=cab5b000)
> Stack: cbde4300 00000000 c02f1c2d c91f2288 c021e235 cbde4300 0000000a 
> 00000004
>        2d343369 cbde4300 c91f2288 00000005 00000000 c02f1c28 00000202 
> c02f1c60
>        00000f14 ca98f000 ca98f000 00001000 c021e41a cbde4300 c91f2288 
> 00000010
> Call Trace:    [<c021e235>] [<c021e41a>] [<c021e892>] [<c0121e5a>] 
> [<c0219ccd>]
>   [<c01b7bca>] [<cca271c2>] [<cca24611>] [<cca246cc>] [<cca2019b>] 
> [<c0140e08>]
>   [<c01370c6>] [<cca2665c>] [<c0137398>] [<c0135ff9>] [<c0135f0e>] 
> [<cca2474b>]
>   [<cca24680>] [<c0144967>] [<c0106f17>]
> Code: f3 a5 f6 c3 02 74 02 66 a5 f6 c3 01 74 01 a4 5b 5e 5f 5d c3
>
>
>  >>EIP; c021ddac <__rta_fill+70/84>   <=====
>
>  >>edi; ca98f05c <_end+a61d7d8/c4dc7dc>
>  >>esp; cab5bdd4 <_end+a7ea550/c4dc7dc>
>
> Trace; c021e235 <rtnetlink_fill_ifinfo+31d/3b8>
> Trace; c021e41a <rtmsg_ifinfo+32/7c>
> Trace; c021e892 <rtnetlink_event+32/38>
> Trace; c0121e5a <notifier_call_chain+1e/38>
> Trace; c0219ccd <register_netdevice+151/168>
> Trace; c01b7bca <register_netdev+5e/70>
> Trace; cca271c2 <[dvb-core]__kstrtab_dvb_ringbuffer_avail+d/1d>
> Trace; cca24611 <[dvb-core]dvb_net_add_if+139/150>
> Trace; cca246cc <[dvb-core]dvb_net_do_ioctl+4c/ac>
> Trace; cca2019b <[dvb-core]dvb_usercopy+b3/129>
> Trace; c0140e08 <link_path_walk+888/99c>
> Trace; c01370c6 <get_chrfops+5a/15c>
> Trace; cca2665c <[dvb-core]dvb_device_open+68/dc>
> Trace; c0137398 <chrdev_open+64/98>
> Trace; c0135ff9 <dentry_open+e1/188>
> Trace; c0135f0e <filp_open+52/5c>
> Trace; cca2474b <[dvb-core]dvb_net_ioctl+1f/24>
> Trace; cca24680 <[dvb-core]dvb_net_do_ioctl+0/ac>
> Trace; c0144967 <sys_ioctl+1bb/1f6>
> Trace; c0106f17 <system_call+33/38>
>
> Code;  c021ddac <__rta_fill+70/84>
> 00000000 <_EIP>:
> Code;  c021ddac <__rta_fill+70/84>   <=====
>    0:   f3 a5                     repz movsl %ds:(%esi),%es:(%edi)   <=====
> Code;  c021ddae <__rta_fill+72/84>
>    2:   f6 c3 02                  test   $0x2,%bl
> Code;  c021ddb1 <__rta_fill+75/84>
>    5:   74 02                     je     9 <_EIP+0x9>
> Code;  c021ddb3 <__rta_fill+77/84>
>    7:   66 a5                     movsw  %ds:(%esi),%es:(%edi)
> Code;  c021ddb5 <__rta_fill+79/84>
>    9:   f6 c3 01                  test   $0x1,%bl
> Code;  c021ddb8 <__rta_fill+7c/84>
>    c:   74 01                     je     f <_EIP+0xf>
> Code;  c021ddba <__rta_fill+7e/84>
>    e:   a4                        movsb  %ds:(%esi),%es:(%edi)
> Code;  c021ddbb <__rta_fill+7f/84>
>    f:   5b                        pop    %ebx
> Code;  c021ddbc <__rta_fill+80/84>
>   10:   5e                        pop    %esi
> Code;  c021ddbd <__rta_fill+81/84>
>   11:   5f                        pop    %edi
> Code;  c021ddbe <__rta_fill+82/84>
>   12:   5d                        pop    %ebp
> Code;  c021ddbf <__rta_fill+83/84>
>   13:   c3                        ret
>
>
> In the case that I get the drivers partly working again, I will put some 
> printk in all dvb net functions so I can trace the last called 
> functions.  Unfortunately I could never see any meaningful log output 
> after a system  "hang".
>
> Thanks,
> -André.


--
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe linux-dvb" as subject.