Badness in 2.6.15-rc1 on 8xx

[Ingo Hornberger]

Hi Joakim,

I was recently faced with the same problem. It's right that it boots
fine into user land. But the bad 'consistent' start address would cause
trouble as soon as you try to initiate some dma transfer. 

It is configurable under [Advanced setup].
I now use a value, seen on another 8xx board (it was some siemens board)
which was 0xf0000000. It works just fine.
As far as I understood the exact address isn't important, only that it
is unused in the kernel.


regards,
	Ingo

On Fri, 2005-11-25 at 22:41 +0100, Joakim Tjernlund wrote:
>  > 
> > > -----Original Message----- 
> > > From: linuxppc-embedded-bounces at ozlabs.org 
> > > [mailto:linuxppc-embedded-bounces at ozlabs.org] On Behalf Of 
> > > Joakim Tjernlund 
> > > Sent: Freitag, 25. November 2005 14:28 
> > > To: linuxppc-embedded at ozlabs.org 
> > > Subject: Badness in 2.6.15-rc1 on 8xx 
> > > 
> > > Anyone seen this when booting 2.6.15-rc1 on 8xx? 
> > >   Mount-cache hash table entries: 512 
> > >   Badness in dma_alloc_init at arch/ppc/kernel/dma-mapping.c:346 
> > >   Call trace: 
> > >    [c00039e8] check_bug_trap+0x80/0xa8 
> > >    [c0003c1c] program_check_exception+0x20c/0x480 
> > >    [c00031e0] ret_from_except_full+0x0/0x4c 
> > >    [c01b86b8] dma_alloc_init+0x40/0xcc 
> > >    [c000225c] init+0x8c/0x288 
> > >    [c00050ac] kernel_thread+0x44/0x60 
> > >   NET: Registered protocol family 16 
> > > The kernel boots just fine into user space so it seems 
> > > harmless, but I suspect it will bite me later. 
> > > 
> > > Something anoying: 
> > > Why did the new cpm_uart driver change major and minor number 
> > > for the tty? 
> > > As it is now I can't boot my 2.4 rootfs as init think it 
> > > should find the console on ttyS0. 
> > > Would be great if major and minor could be configurable. 
> > Because it's a new driver...with a new name (ttyCPM0) and a new 
> > device number. It shared the device number and name in 2.4 
> > with the "standard" device driver (8250/16550), and that made 
> > it very difficult to use both in 2.4. 
> 
> To me it makes more sense to let a major number represent a function, not a driver.
> Shouldn't be that hard to make these drivers cooperate wrt. minor number.
> 
> To allow for easy customization one could instead do:
> #ifndef SERIAL_CPM_MAJOR
> #define SERIAL_CPM_MAJOR        204
> #endif
> #ifndef SERIAL_CPM_MINOR
> #define SERIAL_CPM_MINOR        46
> #endif
> 
> Each platform could then define major and minor as they like.
> 
>  Jocke
>   
> > Maybe you should use console=ttyCPM0 in your boot parameters for 2.6. 
> > 
> > Regards, 
> > Torsten 
> > > 
> > >   Jocke 
> 
> _______________________________________________
> Linuxppc-embedded mailing list
> Linuxppc-embedded at ozlabs.org
> https://ozlabs.org/mailman/listinfo/linuxppc-embedded
>