PERFORCE change 102970 for review
John Birrell
jb at FreeBSD.org
Tue Aug 1 20:42:26 UTC 2006
http://perforce.freebsd.org/chv.cgi?CH=102970
Change 102970 by jb at jb_freebsd2 on 2006/08/01 20:42:20
When heading for a panic due to an unacked IPI, we really only need the
first CPU which decides to panic to print put the trace buffer.
Also report the pad fields from the per-cpu structure which we'll try
to poke stuff in when we're at tl2 to see what is going on.
Affected files ...
.. //depot/projects/kmacy_sun4v_stable/src/sys/sun4v/sun4v/pmap.c#11 edit
Differences ...
==== //depot/projects/kmacy_sun4v_stable/src/sys/sun4v/sun4v/pmap.c#11 (text+ko) ====
@@ -1218,6 +1218,10 @@
m->md.pv_list_count++;
}
+#ifdef TRAP_TRACING
+static int trap_trace_report_done;
+#endif
+
#ifdef SMP
static cpumask_t
pmap_ipi(pmap_t pmap, char *func, uint64_t arg1, uint64_t arg2)
@@ -1288,9 +1292,17 @@
curactive & ~ackmask);
#ifdef TRAP_TRACING
- for (j = 0; j < MAXCPU; j++)
- if (((1 << j) & curactive & ~ackmask) != 0)
- trap_trace_report(j);
+ if (!trap_trace_report_done) {
+ trap_trace_report_done = 1;
+ for (j = 0; j < MAXCPU; j++)
+ if (((1 << j) & curactive & ~ackmask) != 0) {
+ struct pcpu *pc = pcpu_find(j);
+ printf("pcpu pad 0x%jx 0x%jx 0x%jx 0x%jx 0x%jx 0x%jx 0x%jx\n",
+ pc->pad[0], pc->pad[1], pc->pad[2], pc->pad[3],
+ pc->pad[4], pc->pad[5], pc->pad[6]);
+ trap_trace_report(j);
+ }
+ }
#endif
hv_cpu_state((uint64_t)ffs64(curactive & ~ackmask), &cpu_state);
More information about the p4-projects
mailing list