Dear Joe,Wolfgang, and anyone reading this mail,
> 2013/6/7 Wolfgang Denk <w...@denx.de>:
> Dear jinghui shi,
...
>> Why we do not call ethernet driver's "halt" function if there was a
>> "Cannot umount" error?
>
> It appears you have discovered a bug. As this happens only in the
> error handling branch of the code, it is no big surprise that this has
> not been detected before. But yes, this should be fixed.
Is there a plan to fix this?
Now I am planning to fix it myself(If you have a plan, that's a
better choice, I think),
but something still puzzles me. There is a very simple(but maybe not a
solution) method
to avoid this, that is call eth_halt() after "puts("*** ERR: Cannot
umount\n")" and
"net_set_state(NET_LOOPFAIL)", but the problem is if we really receive
a "umount call"
reply but with a previous "rpc id", should we treat it as a error, and
"puts("*** ERR: Cannot umount\n")"?
Another method that leads the calling of eth_halt() is to treat the
["umount call"
reply but with a previous "rpc id"] as a right reply, that means do
not care about the
"rpc id", but what I worry about is that may cause some other
problems, I not clear about
NFS, so it maybe not a good idea to do so.
Can anyone give any idea/advice about this?
Best regards,
Shi Jinghui
_______________________________________________
U-Boot mailing list
U-Boot@lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot
