Re: [REGRESSION] tg3 dead after s2ram

On Wednesday 01 August 2007 23:00:23 Michael Chan wrote:

On Wed, 2007-08-01 at 10:47 -0700, Michael Chan wrote:

quoted

You have 2 Broadcom devices in your system.  07:00.0 is a wireless
device, I think.  8:4.0 is the tg3 device.

It's clear that the tg3 device is still in D3 state after resume and
that explains why all register accesses fail.  tg3_resume() should put
the device back in D0 state in a very straight forward way and I don't
see how that can fail.  It worked for me when I tested it last night.
Can you add some printk() to tg3_resume() to see what's happening?  Let
me know if you want me to send you some debug patches to do that.

I misread the PCI registers below.  The power state was ok.

The problem is that memory enable and bus master were not set in PCI
register 4 after resume.  This also explains the register access
failures.

In tg3_resume(), we call pci_restore_state() which should re-enable
those 2 bits in PCI register 4.  Can you add some printk() to see why
those bits are not restored after pci_restore_state()?

Reading pci_restore_state() it looks already instrumented.  Sorry about the 
wrong pci device, looking at my BCM5788 it is pci device 08:04.0 and looking 
at the log from my first post there is nothing restored in the first 64 
bytes!  Otherwise it would have said "PM: Writing back..."

Is this what you are looking for or should I do other printk instrumentation?

-Joachim