Date: Tue, 19 Aug 1997 23:49:18 -0600 From: Steve Passe <smp@csn.net> To: smp@freebsd.org Cc: current@freebsd.org Subject: HEADS UP: new FAST_INTR() code. Message-ID: <199708200549.XAA18105@Ilsa.StevesCafe.com>
next in thread | raw e-mail | index | archive | help
Hi,
I just committed a bunch o' code that does several new things.
- re-organization in preparation for making spl/cpl MP-safe.
- 2 new variations in FAST_INTR() ISRs:
- skip the check of ipending and cpl, I can't see a good reason for it
and its absense makes life simpler in SMP. This might go back in
if it doesn't work out...
- the "giant lock" is no longer used, FAST_INTR routines can enter the
kernel while other ISRs or traps are active. A simplelock is used to
serialize FAST_INTR()s, but that may also go away someday.
Let me know how sio code "feels", any improvements? or is it worse?
silo overflows?
If you get a system freeze, try to get to ddb via CNTL-PRINTSCREEN and
get me a trace. I ran the system for 2-3 hours with a continuous 115200
stream in/out COM1 while rebuilding a kernel in another xterm with '-j12'
and top in a 3rd xterm, all xterms on a remote machine. Ie. a good
mix of sio/scsi/net INTs and syscalls. No signs of any problems.
--
Steve Passe | powered by
smp@csn.net | Symmetric MultiProcessor FreeBSD
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199708200549.XAA18105>