Re: [RFC]disable msi mode in pci_disable_device

[Linas Vepstas]

On Fri, May 26, 2006 at 10:26:57PM +0200, Brice Goglin wrote:
> Andrew Morton wrote:
> >> In
> >> his usage, pci_save_state will be called at runtime, and later (after
> >> the device operates for some time and has an error) pci_restore_state
> >> will be called.
> >
> > Is that a sane thing for a driver to be doing? (Not relevant to this issue
> > though).
> 
> The aim is to be able to recover from a memory parity error in the NIC.
> Such errors happen sometimes, especially when a cosmic ray comes by. To
> recover, we restore the state that we saved at the end of the
> initialization. As saving currently disables MSI, we currently have to
> restore the state right after saving it at the end of the initialization
> (see the end of
> myri10ge_probe in http://lkml.org/lkml/2006/5/23/24).

My experience dealing with a similar thing suggests that its usually
easier to restore the state to where it was after a cold boot, but
before the device driver touched the h/w.  

--linas

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/