Re: [PATCH 7/11] lpfc 8.1.1 : Add support for more members of the Light Pulse 11xxx (4Gb) family

[Matthew Wilcox <matthew@wil.cx>]

On Mon, Nov 28, 2005 at 10:49:36AM -0700, Matthew Wilcox wrote:
> On Mon, Nov 28, 2005 at 12:08:46PM -0500, James.Smart@Emulex.Com wrote:
> > Please note: there will be 2 other areas where we will still be accessing
> > config space:
> >   - We need the bar value, as the hardware sees it, to program a base
> >     address it can decode to keep ring accesses on-chip. The current pci
> >     infrastructure doesn't provide this. (we've talked about this on this
> >     reflector before).
> 
> sym2 has the same need for pretty much the same reason.  It has a
> private function called pci_get_base_address() right now.  I believe the
> correct thing to do (but haven't got round to trying it yet) is to call
> pcibios_resource_to_bus().

Done, and tested on ia64 and parisc.  Here's the patch that does it for
sym2, hopefully your code will be similar:

http://hera.kernel.org/git/?p=linux/kernel/git/willy/sym2.git;a=blobdiff;h=e6ff89a8524c74f93971715a762b8c64d18176e1;hp=ec9d93aa160d89a256aee7e92b17493205149fca;hb=b5d641b0098e2ae2263f6fa8f24fa18fcbe831ef;f=drivers/scsi/sym53c8xx_2/sym_glue.c

> >   - If we reset the card, there were hardware-isms that may erroneously
> >     drive perr/serr. So, we need to temporarily turn them off while we
> >     reset. (I believe we went through this as well on this list early on).
> 
> That's certainly a reasonable need.

I wonder if it's functionality that should be provided by the PCI layer.
I don't think yours is the only card with this issue.