AF:
NF:0
PS:10
SRH:1
SFN:
DSR:
MID:<20081017181929.6a24dfee@ripper.onstor.net>
CFG:
PT:0
S:andy.sharp@onstor.com
RQ:
SSV:onstor-exch02.onstor.net
NSV:
SSH:
R:<maxim.kozlovsky@onstor.com>,<bfisher@onstor.com>
MAID:1
X-Sylpheed-Privacy-System:
X-Sylpheed-Sign:0
SCF:#mh/Mailbox/sent
RMID:#imap/andys@onstor.net@onstor-exch02.onstor.net/INBOX	0	BB375AF679D4A34E9CA8DFA650E2B04E0C19BE5E@onstor-exch02.onstor.net
X-Sylpheed-End-Special-Headers: 1
Date: Fri, 17 Oct 2008 18:45:36 -0700
From: Andrew Sharp <andy.sharp@onstor.com>
To: "Maxim Kozlovsky" <maxim.kozlovsky@onstor.com>
Cc: "Bill Fisher" <bfisher@onstor.com>
Subject: Re: ACPU/NCPU Linux Proposal and Task List; Take 2
Message-ID: <20081017184536.2cffd439@ripper.onstor.net>
In-Reply-To: <BB375AF679D4A34E9CA8DFA650E2B04E0C19BE5E@onstor-exch02.onstor.net>
References: <48F7E198.4080502@onstor.com>
	<BB375AF679D4A34E9CA8DFA650E2B04E0C19BE5E@onstor-exch02.onstor.net>
Organization: Onstor
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: quoted-printable

On Fri, 17 Oct 2008 11:57:38 -0700 "Maxim Kozlovsky"
<maxim.kozlovsky@onstor.com> wrote:

> I think this gets most or all of the major tasks right. A few
> additions:
>=20
> 2.2. We probably will need to have the NFS client in the NCPU kernel
> for other reasons even if we chose to boot with root on ramdisk. For
> example we need to support tcpdump on NCPU, which will need to store
> the capture somewhere. The easiest way to implement that is to mount
> one of our own volumes via NFS, which requires NFS client code.

=46rom now on, the "NCPU kernel" shall simply be referred to as "TUXRX".
Please modify this document to reflect this.

We can easily use sockets/network pipes for this.  But we will want NFS
client all over the place for development.  NFS root filesystem, etc.

> Exception handling - in the current code all cores establish the
> "passive crash" interrupt handlers which are activated by mailbox
> interrupt that is posted by the core that crashes. This allows
> stopping all cores in the event of the crash. We'll need similar or
> better functionality. The code will have to be modified because we'll
> have to use the mailbox interrupts for more tasks (the "passive
> crash" is the only use of the mailboxes currently). For example we'll
> need some modifications to ACPU and FP code to interrupt the NCPU
> when the messages are placed in the IPC queues, or when the number of
> TXRX buffers/descriptors available for allocation on FP falls below
> certain threshold.

I'm not seeing why we would need any functionality similar to the
"passive crash" thing you describe, but maybe I'm not fully getting
what that is.  Are you saying you want the ACPU thread(s) to "stop"
when an FP core crashes?  We can do the equivalent of that.

I'm also not seeing why we would need interrupts.  Is polling not
good enough for the kernel?

> Profiler support - gprof and sibyte profilers will need to be
> supported for NCPU and ACPU.

Hmm.  Are you sure top(1) won't be good enough?  There are various
kernel profiling "subsystems" that we can employ.

grof is needed for the SSC too but for some reason we aren't building
the profiler C libraries, or if we are, they are being held back
from us.  Laaaaarry?!

> Access the NCPU mailbox register from SSC so we can interrupt the NCPU
> when posting a message - need to check if this is possible. This might
> require some PROM changes. Brian/Warren should know the answer.

Again I don't see why the TUXRX can't poll like everyone else.  But I
suspect that it may already set up this way since I asked Brian about it
a couple of weeks ago.

> Watchdog support - I assume this will have to change since Linux has
> its own watchdog driver. The ACPU should be able to start/stop/kick
> its watchdog.

Well there's the Sibyte WD driver I wrote as well as the softdog
driver.

> >-----Original Message-----
> >From: Bill Fisher
> >Sent: Thursday, October 16, 2008 5:52 PM
> >To: Andy Sharp; Maxim Kozlovsky
> >Cc: Bill Fisher
> >Subject: ACPU/NCPU Linux Proposal and Task List; Take 2
> >
> >Guys:
> >
> >Attached is take 2 of the proposal. I have re-ordered the tasks
> >as per Max's suggestion and incorporated the changes as
> >per his review comments.
> >
> >As before, let me know what I am missing or have
> >gotten wrong.
> >
> >Thanks,
> >
> >-- Bill
>=20
