[U-Boot-Users] a problem about rtl8139 driver on sandpoint8245 board

[U-Boot-Users] a problem about rtl8139 driver on sandpoint8245 board

[Lucas Jin]

Hi, all

I am debuging RTL8139 driver for a board based on Sandpoint8245 and I have a problem.

The CPU panic on line 529 in rtl8139.c. (outb(CmdReset, ioaddr + ChipCmd);)

Serial port displayed:
(
Bad trap at PC: fffffffc, SR: 1000, vector=800
NIP: FFFFFFFC XER: 20000000 LR: 00001288 REGS: 077ceb30 TRAP: 0800 DAR: 077CECC0
MSR: 00001000 EE: 0 PR: 0 FP: 0 ME: 1 IR/DR: 00

GPR00: 00000010 077CEC20 FFFFFF1F 00000000 0000000A 00000010 FFFFFFFB 00000000 
GPR08: 00000040 81000000 00000000 FFFFFFFF 00000020 A7AC5E24 07FFD000 080D1000 
GPR16: 31380950 8E109020 05808002 55D84040 00003032 00000001 00000000 00000001 
GPR24: 00000000 077CEEA8 077CEFC8 00000005 07FF80C8 077CEF94 07FFD6D8 00000000 
Call backtrace: 
Exception in kernel pc fffffffc signal 0
)

The value of ioaddr is 0x80000000 

The value of _IO_BASE outb uses is defined as 0

Thank you all
Jinlu

[U-Boot-Users] a problem about rtl8139 driver on sandpoint8245 board

[Shinya Kuribayashi]

Lucas Jin wrote:
> Hi, all
> 
> I am debuging RTL8139 driver for a board based on Sandpoint8245 and I have a problem.

I don't know much about Sandpoint 8245, but

> The CPU panic on line 529 in rtl8139.c. (outb(CmdReset, ioaddr + ChipCmd);)
> 
> Serial port displayed:
> (
> Bad trap at PC: fffffffc, SR: 1000, vector=800
> NIP: FFFFFFFC XER: 20000000 LR: 00001288 REGS: 077ceb30 TRAP: 0800 DAR: 077CECC0
> MSR: 00001000 EE: 0 PR: 0 FP: 0 ME: 1 IR/DR: 00
> 
> GPR00: 00000010 077CEC20 FFFFFF1F 00000000 0000000A 00000010 FFFFFFFB 00000000 
> GPR08: 00000040 81000000 00000000 FFFFFFFF 00000020 A7AC5E24 07FFD000 080D1000 
> GPR16: 31380950 8E109020 05808002 55D84040 00003032 00000001 00000000 00000001 
> GPR24: 00000000 077CEEA8 077CEFC8 00000005 07FF80C8 077CEF94 07FFD6D8 00000000 
> Call backtrace: 
> Exception in kernel pc fffffffc signal 0
> )
> 
> The value of ioaddr is 0x80000000 
> 
> The value of _IO_BASE outb uses is defined as 0

AFAICT the drivers/rtl8139.c has at least two problems:

  - First, the prospekctive target is MIPS
    because it uses KSEG1ADDR macro to obtain uncached address.

  - Moreover it only works under cache-{disabled,less} CPU
    because it lacks of flush_cache() before transmitting data or
    reclaiming rx buffers.

thanks,

    Shinya Kuribayashi

[U-Boot-Users] a problem about rtl8139 driver on sandpoint8245board

[Lucas Jin]

Shinya Kuribayashi:

Thank you for your hint. I think it will be of help in the future debugging.

But my problem occurs just in the process of rtl8139 reset.

>Lucas Jin wrote:
>> Hi, all
>> 
>> I am debuging RTL8139 driver for a board based on Sandpoint8245 and I have a problem.
>
>I don't know much about Sandpoint 8245, but
>
>> The CPU panic on line 529 in rtl8139.c. (outb(CmdReset, ioaddr + ChipCmd);)
>> 
>> Serial port displayed:
>> (
>> Bad trap at PC: fffffffc, SR: 1000, vector=800
>> NIP: FFFFFFFC XER: 20000000 LR: 00001288 REGS: 077ceb30 TRAP: 0800 DAR: 077CECC0
>> MSR: 00001000 EE: 0 PR: 0 FP: 0 ME: 1 IR/DR: 00
>> 
>> GPR00: 00000010 077CEC20 FFFFFF1F 00000000 0000000A 00000010 FFFFFFFB 00000000 
>> GPR08: 00000040 81000000 00000000 FFFFFFFF 00000020 A7AC5E24 07FFD000 080D1000 
>> GPR16: 31380950 8E109020 05808002 55D84040 00003032 00000001 00000000 00000001 
>> GPR24: 00000000 077CEEA8 077CEFC8 00000005 07FF80C8 077CEF94 07FFD6D8 00000000 
>> Call backtrace: 
>> Exception in kernel pc fffffffc signal 0
>> )
>> 
>> The value of ioaddr is 0x80000000 
>> 
>> The value of _IO_BASE outb uses is defined as 0
>
>AFAICT the drivers/rtl8139.c has at least two problems:
>
>  - First, the prospekctive target is MIPS
>    because it uses KSEG1ADDR macro to obtain uncached address.
>
>  - Moreover it only works under cache-{disabled,less} CPU
>    because it lacks of flush_cache() before transmitting data or
>    reclaiming rx buffers.
>
>thanks,
>
>    Shinya Kuribayashi
>

Yours 
Lucas Jin

[U-Boot-Users] a problem about rtl8139 driver on sandpoint8245 board

[Scott Wood]

Shinya Kuribayashi wrote:
>   - Moreover it only works under cache-{disabled,less} CPU
>     because it lacks of flush_cache() before transmitting data or
>     reclaiming rx buffers.

DMA is cache-coherent on 8245.

-Scott