Date: Fri, 22 Nov 1996 20:50:46 -0700 From: Steve Passe <smp@csn.net> To: freebsd-smp@freefall.freebsd.org Cc: "John S. Dyson" <toor@dyson.iquest.net> Subject: Re: SMP -current merge Message-ID: <199611230350.UAA13108@clem.systemsix.com>
index | next in thread | raw e-mail
Hi,
Additional clues on SMP brokeness.
I added this patch to kern/init_main.c:smp_idle():
----------------------------------- cut --------------------------------------
*** init_main.c 1996/11/22 18:04:29 1.1
--- init_main.c 1996/11/22 20:30:00
***************
*** 812,817 ****
--- 812,835 ----
rel_mplock();
}
+ #if 1
+ if (smp_active == 2 && cpunumber() != 0) {
+ get_mplock();
+ printf("cpu#%d spinning; mp_lock: %08x\n", cpunumber(),
+ mp_lock);
+ wakeup((caddr_t)&smp_active);
+ rel_mplock();
+
+ while (smp_active == 2) {
+ __asm __volatile("" : : : "memory");
+ }
+ get_mplock();
+ printf("cpu#%d waking up! mp_lock: %08x\n",
cpunumber(),
+ mp_lock);
+ rel_mplock();
+ }
+ #endif /** 1 */
+
if (whichqs || whichrtqs || whichidqs) {
/* grab lock for kernel "entry" */
----------------------------------- cut --------------------------------------
What it does is "freeze" the 2nd CPU after it has initialized itself and is
about to run its first proc. When I start it via "sysctl kern.smp_active=2"
it gets to this point without hurting the system. The first CPU returns
from the sysctl, and I can continue to use the system without reboot. If I
allow the 2nd CPU to exit this freeze via "sysctl kern.smp_active=3" I
get the prompt back, but, as before, the first action causes a REBOOT.
If I change:
+ if (smp_active == 2 && cpunumber() != 0) {
to:
+ if (smp_active == 2 && cpunumber() == 0) {
it works differently. This change causes the 1st CPU to be the one frozen,
ie the 2nd CPU is now the only one running. I cannot do anything else after
the sysctl returns without REBOOTing. Ie, its NOT that 2 CPUs are running,
its the running of the 2nd CPU that REBOOTs.
So it appears that the 2nd CPU is not getting a correct environment in which
to run anymore. Could this be something different in the VM system?
It appears to have changed alot since the last merge.
--
Steve Passe | powered by
smp@csn.net | FreeBSD
-----BEGIN PGP PUBLIC KEY BLOCK-----
Version: 2.6.2
mQCNAzHe7tEAAAEEAM274wAEEdP+grIrV6UtBt54FB5ufifFRA5ujzflrvlF8aoE
04it5BsUPFi3jJLfvOQeydbegexspPXL6kUejYt2OeptHuroIVW5+y2M2naTwqtX
WVGeBP6s2q/fPPAS+g+sNZCpVBTbuinKa/C4Q6HJ++M9AyzIq5EuvO0a8Rr9AAUR
tBlTdGV2ZSBQYXNzZSA8c21wQGNzbi5uZXQ+
=ds99
-----END PGP PUBLIC KEY BLOCK-----
home |
help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199611230350.UAA13108>