[ECOS] redboot network crash with rltk8139

[jerzy dyrda]

Hi Andrew,

I had problem with rltk8139 driver too. But reason was in rltk8139_deliver() - 
chip flag - buff empty was incorrect. Driver told eth_drv_recv(), read frame 
but descriptor for this frame was dummy. At end driver lost synchronization.

Best Regards 
Jerzy
  
On Tuesday 12 October 2004 19:34, Andrew Dyer wrote:
> Hi all,
>
> I'm trying to figure out what happened with a crash on my target.  The
> target is a custom platform based on a Toshiba TX4925 MIPS part
> running redboot and using the rltk8139 driver.  The crash occured while
> the target was just sitting at the redboot prompt.
>
> The crash occured because memcpy got passed an address of 0 from
> rltk8139_recv().  rltk8139_recv() never checks for a 0 in the sg_list
> buffer addresses.   eth_drv_recv() decided to use 0 based on the return
> from eth_drv_msg_get().
>
> The comments in eth_drv_recv() indicate that a driver should know that a
> NULL in the sg_list buffer address means discard, so is it a bug to not
> check for a that or
> is there some reason why it should never occur?
>
> Looking at some other drivers, the dp83816 driver doesn't check for NULL,
> but the LAN91C111 does.
>
>
>
> gdb backtrace is below:
>
> #0  _memcpy (s1=0x0, s2=0xa0023af4, n=44)
>     at /ecos-e/Projects/ecos/packages/infra/current/src/memcpy.c:165
> #1  0xffffffff800030c8 in rltk8139_recv (sc=0x0, sg_list=0x800233d8,
> sg_len=44) at
> /ecos-e/Projects/ecos/packages/devs/eth/rltk/8139/current/src/if_8139.c:100
>9 #2  0xffffffff8001b204 in eth_drv_recv (sc=0x800213c8, total_len=60) at
> /ecos-e/Projects/ecos/packages/io/eth/current/src/stand_alone/eth_drv.c:489
> #3  0xffffffff80003340 in rltk8139_deliver (sc=0x800213c8)
>     at
> /ecos-e/Projects/ecos/packages/devs/eth/rltk/8139/current/src/if_8139.c:120
>6 #4  0xffffffff800033d0 in rltk8139_poll (sc=0xffffffff)
>     at
> /ecos-e/Projects/ecos/packages/devs/eth/rltk/8139/current/src/if_8139.c:128
>1 #5  0xffffffff8001afc8 in eth_drv_read (eth_hdr=0x80023598 "ÿÿÿÿÿÿ",
> buf=0x8002e5a8 "E", len=1526)
>     at
> /ecos-e/Projects/ecos/packages/io/eth/current/src/stand_alone/eth_drv.c:398
> #6  0xffffffff800166c8 in __enet_poll ()
>     at /ecos-e/Projects/ecos/packages/redboot/current/src/net/enet.c:158
> #7  0xffffffff80017568 in __tcp_poll ()
>     at /ecos-e/Projects/ecos/packages/redboot/current/src/net/tcp.c:645
> #8  0xffffffff80007008 in net_io_test (is_idle=0)
>     at /ecos-e/Projects/ecos/packages/redboot/current/src/net/net_io.c:551
> #9  0xffffffff80012a1c in do_idle (is_idle=1)
>     at /ecos-e/Projects/ecos/packages/redboot/current/src/main.c:211
> #10 0xffffffff8001448c in _rb_gets_preloaded (buf=0x8002c240 "",
> buflen=256, timeout=0) at
> /ecos-e/Projects/ecos/packages/redboot/current/src/io.c:383 #11
> 0xffffffff800144d8 in _rb_gets (
>     buf=0xffffffff <Address 0xffffffff out of bounds>, buflen=-1610466572,
>     timeout=44) at
> /ecos-e/Projects/ecos/packages/redboot/current/src/io.c:637 #12
> 0xffffffff80012cf4 in cyg_start ()
>     at /ecos-e/Projects/ecos/packages/redboot/current/src/main.c:371
> --
> Hardware, n.:
>         The parts of a computer system that can be kicked.

--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss