AF:
NF:0
PS:10
SRH:1
SFN:
DSR:
MID:
CFG:
PT:0
S:andy.sharp@lsi.com
RQ:
SSV:mhbs.lsil.com
NSV:
SSH:
R:<Maxim.Kozlovsky@lsi.com>,<Brian.Stark@lsi.com>
MAID:2
X-Sylpheed-Privacy-System:
X-Sylpheed-Sign:0
SCF:#mh/Mailbox/sent
RMID:#imap/LSI/INBOX	16962	861DA0537719934884B3D30A2666FECC010E3D2289@cosmail02.lsi.com
X-Sylpheed-End-Special-Headers: 1
Date: Tue, 23 Mar 2010 11:27:07 -0700
From: Andrew Sharp <andy.sharp@lsi.com>
To: "Kozlovsky, Maxim" <Maxim.Kozlovsky@lsi.com>
Cc: "Stark, Brian" <Brian.Stark@lsi.com>
Subject: Re: accessing txrx memory from qlogic
Message-ID: <20100323112707.01410c61@ripper.onstor.net>
In-Reply-To: <861DA0537719934884B3D30A2666FECC010E3D2289@cosmail02.lsi.com>
References: <861DA0537719934884B3D30A2666FECC010E3D207F@cosmail02.lsi.com>
 <20100322172544.71ed8fb5@ripper.onstor.net>
 <861DA0537719934884B3D30A2666FECC010E3D2289@cosmail02.lsi.com>
Organization: LSI
X-Mailer: Sylpheed-Claws 2.6.0 (GTK+ 2.8.20; x86_64-pc-linux-gnu)
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

The hardware addressing is set up by the PROM, and hasn't changed.
All the kernel can do, within reason, is program the TLB, which is about
virtual addresses.

On Tue, 23 Mar 2010 11:57:20 -0600 "Kozlovsky, Maxim"
<Maxim.Kozlovsky@lsi.com> wrote:

> qlogic uses physical addresses, not cached or uncached. 
> 
> On cougar every physical memory location has two physical addresses.
> Bits 36-39 of physical address constitute the node ID. If the node ID
> is 0, local memory is accessed. TXRX node id is 4, FP node id is 5.
> Address with node ID must be used to access the memory across HT. In
> a shared memory configuration all nodes and devices must use the same
> physical address to access the same memory location, which means TXRX
> must use addresses with its node ID to access its local memory.
> Currently the memory map is setup so txrx uses physical addresses
> with node id 0. It must be changed to use node specific addresses.
> You must be already doing that for 8-way support, I just need this
> part of the changes early so I can proceed with the qlogic testing.
> 
> -----Original Message-----
> From: Andrew Sharp [mailto:andy.sharp@lsi.com] 
> Sent: Monday, March 22, 2010 5:26 PM
> To: Kozlovsky, Maxim
> Cc: Stark, Brian
> Subject: Re: accessing txrx memory from qlogic
> 
> On Mon, 22 Mar 2010 16:52:53 -0600 "Kozlovsky, Maxim"
> <Maxim.Kozlovsky@lsi.com> wrote:
> 
> > Hello,
> > 
> > It looks like I'll need to have the kernel memory map changed to be
> > able to access txrx memory from qlogic.
> > 
> > Txrx node address (0x4000000000UL) should be added to the physical
> > address passed to qlogic, however the code must be accessing the
> > memory using the same physical address, otherwise caching anomalies
> > will happen.
> 
> I'm trying to follow what you're saying, but not quite getting it.
> The qlogic will have to use uncached addresses, I'm assuming, like
> 0x900000040.00000000.  Actually, I have no idea if that will work.  We
> might have to get the FP node is some state where memory works and use
> FP memory for the qlogics.   Now my head is starting to hurt.
