Date: Sat, 30 Oct 2004 12:37:37 +0900 From: Shunsuke SHINOMIYA <shino@fornext.org> To: John Baldwin <jhb@FreeBSD.org> Cc: obrien@FreeBSD.org Subject: Re[2]: disabling interrupt storm protection Message-ID: <20041030105024.360A.SHINO@fornext.org> In-Reply-To: <200410281111.24398.jhb@FreeBSD.org> References: <20041028165604.DCF8.SHINO@fornext.org> <200410281111.24398.jhb@FreeBSD.org>
index | next in thread | previous in thread | raw e-mail
[-- Attachment #1 --]
Thank you John,
I wrote a patch(which is attached to this mail) so that the problem
doesn't occur in my environment.
This patch changes the meaning of `storm'.
But, I think use of the parameter(hw_intr_threshold) to control the
detector in this method is clearer than the present method, and it
needs no DELAY(1).
In the detector to which this patch is applied and compiled with
-DHACK2 flag, `storm' is the situation that the number of generated
interrupts per unit time(1/hz) is larger than threshold.
If kern_intr.c is compiled with -DHACK2 -DHACK3 flags, it works as
interrupt rate limiter.
What do you think about this method?
--
Shunsuke SHINOMIYA <shino@fornext.org>
[-- Attachment #2 --]
--- /home/shino/work/freebsd/src/sys/kern/kern_intr.c Thu Sep 9 19:03:19 2004
+++ /home/shino/work/packetmerge/sys/kern/kern_intr.c Sat Oct 30 11:53:59 2004
@@ -485,14 +485,23 @@ ithread_loop(void *arg)
struct intrhand *ih; /* and our interrupt handler chain */
struct thread *td;
struct proc *p;
+#ifndef HACK2
int count, warming, warned;
+#else
+ int count, curr_ticks, warned;
+#endif
td = curthread;
p = td->td_proc;
ithd = (struct ithd *)arg; /* point to myself */
KASSERT(ithd->it_td == td && td->td_ithd == ithd,
("%s: ithread and proc linkage out of sync", __func__));
+#ifndef HACK2
warming = 10 * intr_storm_threshold;
+#else
+ count = 0;
+ curr_ticks = ticks;
+#endif
warned = 0;
/*
@@ -514,7 +523,9 @@ ithread_loop(void *arg)
CTR4(KTR_INTR, "%s: pid %d: (%s) need=%d", __func__,
p->p_pid, p->p_comm, ithd->it_need);
+#ifndef HACK2
count = 0;
+#endif
while (ithd->it_need) {
/*
* Service interrupts. If another interrupt
@@ -548,6 +559,7 @@ restart:
if ((ih->ih_flags & IH_MPSAFE) == 0)
mtx_unlock(&Giant);
}
+#ifndef HACK2
if (ithd->it_enable != NULL) {
ithd->it_enable(ithd->it_vector);
@@ -567,6 +579,16 @@ restart:
}
}
+#else
+ if(intr_storm_threshold <= 0)
+ goto enable_intr;
+
+ if(curr_ticks != ticks) {
+ count = 0;
+ curr_ticks = ticks;
+ goto enable_intr;
+ }
+#endif /* HACK2 */
/*
* If we detect an interrupt storm, sleep until
* the next hardclock tick. We sleep at the
@@ -595,9 +617,21 @@ restart:
* away unless the interrupt repeats
* less often the hardclock interrupt.
*/
+#ifndef HACK2
count = INT_MAX - 1;
}
+#else
+#ifndef HACK3
+ curr_ticks = ticks;
+#endif /* HACK3 */
+ } else
+#endif /* HACK2 */
count++;
+#ifdef HACK2
+enable_intr:
+ if (ithd->it_enable != NULL)
+ ithd->it_enable(ithd->it_vector);
+#endif /* HACK2 */
}
WITNESS_WARN(WARN_PANIC, NULL, "suspending ithread");
mtx_assert(&Giant, MA_NOTOWNED);
help
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20041030105024.360A.SHINO>