BAD MCP validity signature

[sham pavman]

Hi All,

I'm modifying the broadcom 57712 driver and i'm facing some weird
inconsistent issues with this card.
Here is what i've found out thus far and need your advice on this.

I'm running a custom OS over a hypervisor and this OS is configured to use
the 57712 card for FE connectivity. The problem is that every alternate
shutdown causes a panic. (mind you i'm shutting down only the VM ) .
To be more clear, when i first boot the VM everything is fine but when i
try to shutdown the VM and then reboot it, i see that i get a 'BAD MCP
validity signature' error and the 57712 firmware dumps core.

Through my investigation into this issue what i've found out is that
the *shared
memory regions which are are used to find out the MCP value are getting
changed when i reboot the VM.*

The surprising part is that the addr location is obtained correctly on
every alternate reboots.
First reboot fails, second reboot works, third fails and so on.
(if not anything at least this is a consistent behavior).

As a fix to this problem whenever the MCP is returned as bad i tried to
re-initialize MCP , and when i ran that code i get another error stating
"MCP not up" and again the firmware dumps .
(Should i increase the delay in waiting for the the register writes to
happen as i'm in a virtual env?).

So as a test, i obtained the addr that was being returned on a successful
boot and tried to hard code it in case i'm not looking at that register .
This time the "BAD MCP" issue didn't occur however later down in the boot
sequence things got messed up totally becuase of my hard coding i guess and
i had a totally different failure.

Can anyone point me in the right direction, or if what i've done (by re
initializing MCP) was correct but may be it needs more work ?

Thanks
Shampavman
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20140214/8711df80/attachment-0001.html