Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 8 Sep 2000 00:01:54 -0700 (PDT)
From:      John Baldwin <jhb@pike.osd.bsdi.com>
To:        Warner Losh <imp@village.org>
Cc:        current@FreeBSD.ORG
Subject:   Re: SMPNG kernel on UP
Message-ID:  <200009080701.AAA36477@pike.osd.bsdi.com>
In-Reply-To: <200009080636.e886aCG18672@billy-club.village.org> from Warner Losh at "Sep 8, 2000 00:36:12 am"

next in thread | previous in thread | raw e-mail | index | archive | help
Warner Losh wrote:
> 
> I've noticed a few things wrong with the SMPNG code.  These are mostly
> impressions.  As I learn more about them, I'll file more formal bug
> reports.  This is an FYI to fellow travelers along the path.
> 	1) You can't have I386_CPU at all.  This is likely bad in
> 	   the long run, but OK for now.

Err, this should be fixed now.  Do you have the latest i386/include/smp.h?

> 	2) APM is now broken.  It worked after the ACPI integration,
> 	   but after the SMPNG stuff neither apm -z nor the BIOS keys
> 	   seems to suspend.

Hmm.

> 	3) Linux emuation is panics the machine in linux_open on
> 	   a normal boot.

Hate to ask, but are you sure it is loading the right module?  With the
kernel.ko change, modules don't live in the same place anymore, and I haven't
seen any commits to kldload, /etc/rc, or linux.sh that takes this into account.

> 	5) USB ethernet is more likely to hang the machine than it was
> 	   before.  This hand is transient, however.  I have seen
> 	   one crash removing the USB adapter that I didn't used to
> 	   see, but wasn't in a position to look at it in detail.

Hmm, ok.

> 	7) SSH to a machine on my local network is dog slow
> 	   sometimes.  I can type about one line or two lines ahead of
> 	   it in email when it happens.  It feels like a network pause
> 	   of about 1-2 seconds.  Local windows behave well during
> 	   these episodes.

Hmm, haven't seen this.

> 	8) The machine seems to pause more often than it used to for
> 	   reasons totally unknown at this time.  These pauses last
> 	   for several seconds and then things are good again.  This
> 	   might be related to #7, but the total machine pauses happen
> 	   w/o any network connections.  Next time this happens, I'll
> 	   hit capslock to see if the interrupts are blocked or not.

Haven't seen this either. :(  Here are some of the known issues that are
gotchas:

1) sio is broken.  It doesn't receieve any data, although Greg has just
   fixed this and should be committing a fix shortly.
2) There seem to be some problems with possibly the ahc controller (haven't
   tried ata yet) where after several hours of heavy load (-j 256 buildworlds
   and the like) the disk driver will hang.  The machine will run fine, top
   still runs over ssh, for example.  However, any process that accesses the
   disk hangs since it blocks forever waiting on the disk, thus rendering
   the machine useless.  Note that this only happens with SMP.  UP seems to
   be very stable.  I haven't had a UP kernel panic or freeze in weeks.
3) The alpha is not entirely stable.  I'm working on adding interrupt threads
   to the alpha right now, which will hopefully help.  I think there are
   some areas in the kernel that are assuming we have the type of locking we
   have on x86 with ithreads, and am hoping that alpha ithreads will fix
   those assumptions.  Also, there are many changes in dfr's alpha patches
   that I at least do not understand, and it possibly some of them are
   incomplete/bogus.

> Now is a bad time to be thinking about cuting a production
> machine/snapshot from the -current branch.

Are there good times for this?  ;-)

> Warner

-- 

John Baldwin <jhb@bsdi.com> -- http://www.FreeBSD.org/~jhb/
PGP Key: http://www.cslab.vt.edu/~jobaldwi/pgpkey.asc
"Power Users Use the Power to Serve!"  -  http://www.FreeBSD.org/


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200009080701.AAA36477>