2010-03-31 12:19:19     Kernel Freeze maybe due to bfin_mac?

Document created by Aaronwu Employee on Aug 21, 2013
Version 1Show Document
  • View in full screen mode

2010-03-31 12:19:19     Kernel Freeze maybe due to bfin_mac?

Stefan Wanja (GERMANY)

Message: 87929   

 

Hello,

 

I opened a new thread for an old problem because I fear that at the old thread noone looks anymore. It was this thread: https://blackfin.uclinux.org/gf/project/uclinux-dist/forum/?action=ForumBrowse&forum_id=39&thread_id=39883

 

Now we got us a gnICE+ debugger and now I can see a bit more.

 

Looking at the frozen board with bfin-elf-gdb the stack consists only of three entries (I repeated this on the bf527 EZKit a couple of times now):

 

    evt_timer() at /..../linux-2.6.x/arch/blackfin/include/asm/current.h:17

    bfin_mac_hard_start_xmit(skb=0x457288, dev=0x3d8b0a0) at drivers/net/bfin_mac.c:685

    ??()

 

Looking at the memory there are zeros until 0x20400000, where the kernel begins, from there on its all 0xadadadad (or different at other times)

 

Looking at the registers there are only zeros except for

 

r7=0x3d8b0a0

p4=0x457288

sp=0x42fb10

fp=0x42fb10

rets=0xd4eec <bfin_mac_hard_start_xmit+388>

pc= 0xd4eec <bfin_mac_hard_start_xmit+388>

 

Within bfin_mac_hard_start_xmit it stopped at the function call to blackfin_dcache_flush_range with parameter data being 0x0 (problably the value after the flush).

 

I don't know how the caching works, but we've had problems in the bfin_mac driver with caching before... I think there might be a bug in there.

 

Due to the yet unsolved older bug we (still) have WRITE_THROUGH cache policy in use as a work around, might be that the problem only occurs with that setting.

 

Kind Regards,

 

Stefan

QuoteReplyEditDelete

 

 

2010-03-31 17:23:47     Re: Kernel Freeze maybe due to bfin_mac?

Mike Frysinger (UNITED STATES)

Message: 87939   

 

dont post the same thing twice please

Attachments

    Outcomes