<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="Generator" CONTENT="MS Exchange Server version 6.5.7653.38">
<TITLE>RE: [PATCH] [SCSI] mpt fusion: Fix 32 bit platforms with 64 bit resources.</TITLE>
</HEAD>
<BODY>
<!-- Converted from text/plain format -->
<P><FONT SIZE=2>-----Original Message-----<BR>
From: Benjamin Herrenschmidt [<A HREF="mailto:benh@kernel.crashing.org">mailto:benh@kernel.crashing.org</A>]<BR>
Sent: Thu 11/5/2009 12:00 PM<BR>
To: James Bottomley<BR>
Cc: Josh Boyer; Eric.Moore@lsi.com; Pravin Bathija; linux-scsi@vger.kernel.org; linuxppc-dev@ozlabs.org<BR>
Subject: Re: [PATCH] [SCSI] mpt fusion: Fix 32 bit platforms with 64 bit resources.<BR>
<BR>
On Thu, 2009-11-05 at 10:07 -0600, James Bottomley wrote:<BR>
<BR>
> > > ioc->memmap = mem;<BR>
> > >- dinitprintk(ioc, printk(MYIOC_s_INFO_FMT "mem = %p, mem_phys = %lx\n",<BR>
> > >- ioc->name, mem, mem_phys));<BR>
> > >+ dinitprintk(ioc, printk(MYIOC_s_INFO_FMT "mem = %p, mem_phys = %llx\n",<BR>
> > >+ ioc->name, mem, (u64)mem_phys));<BR>
> > ><BR>
> > > ioc->mem_phys = mem_phys;<BR>
> > > ioc->chip = (SYSIF_REGS __iomem *)mem;<BR>
> > ><BR>
> > > /* Save Port IO values in case we need to do downloadboot */<BR>
> > >- ioc->pio_mem_phys = port;<BR>
> > >+ port = ioremap(port_phys, psize);<BR>
> > >+ if (port == NULL) {<BR>
> > >+ printk(MYIOC_s_ERR_FMT " : ERROR - Unable to map adapter"<BR>
> > >+ " port !\n", ioc->name);<BR>
> > >+ return -EINVAL;<BR>
><BR>
> So this looks problematic on a few platforms ... what happens to<BR>
> platforms that have no IO space? They automatically fail here and it<BR>
> looks like the adapter never attaches.<BR>
<BR>
> Yup, that part of the patch looks wrong.<BR>
<BR>
> However, a mechanical replacement of unsigned long's with<BR>
> resource_size_t to hold physical addresses should be fine despite the<BR>
> lack of feedback from LSI.<BR>
<BR>
> Pravin, that ioremap definitely seems like it has nothing to do there,<BR>
> port IO is already remapped for you by the core PCI code and should work<BR>
> "as is". Please respin without that change.<BR>
<BR>
> Cheers,<BR>
>Ben.<BR>
<BR>
Thanks for the input. Will make the suggested changes and re-submit the patch.<BR>
<BR>
Regards,<BR>
Pravin<BR>
<BR>
<BR>
<BR>
</FONT>
</P>
</BODY>
</HTML>