Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 1 Sep 2009 09:55:33 -0400
From:      John Baldwin <jhb@freebsd.org>
To:        freebsd-current@freebsd.org
Cc:        Attilio Rao <attilio@freebsd.org>, Marc UBM <ubm@u-boot-man.de>
Subject:   Re: panic: apic_free_vector: Thread already bound.
Message-ID:  <200909010955.33747.jhb@freebsd.org>
In-Reply-To: <3bbf2fe10908311331k3e871898tc36a9846651cc53@mail.gmail.com>
References:  <20090822115958.f93fcf29.ubm.freebsd@gmail.com> <20090831222650.45b6c806.ubm@u-boot-man.de> <3bbf2fe10908311331k3e871898tc36a9846651cc53@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Monday 31 August 2009 4:31:50 pm Attilio Rao wrote:
> 2009/8/31 Marc UBM <ubm@u-boot-man.de>:
> > On Mon, 24 Aug 2009 16:31:05 -0400
> > John Baldwin <jhb@freebsd.org> wrote:
> >
> >> On Saturday 22 August 2009 5:59:58 am Marc UBM wrote:
> >> >
> >> > Hiho! :-)
> >> >
> >> > I'm seeing above panic in the final stages of the shutdown process.
> >> > Coredump is available, bt as follows, textdump attached.
> >> >
> >> >
> >> > FreeBSD hamstor 8.0-BETA2 FreeBSD 8.0-BETA2 #14: Wed Aug 19 22:43:17
> >> > CEST 2009     oni0@hamstor:/usr/obj/usr/src/sys/HAMSTOR  amd64
> >> >
> >> >
> >> > (kgdb) bt
> >> > #0  doadump () at pcpu.h:223
> >> > #1  0xffffffff801b561c in db_fncall (dummy1=Variable "dummy1" is not
> >> > #available.
> >> > ) at /usr/src/sys/ddb/db_command.c:548
> >> > #2  0xffffffff801b5951 in db_command (last_cmdp=0xffffffff808594a0,
> >> > #cmd_table=Variable "cmd_table" is not available.
> >> > ) at /usr/src/sys/ddb/db_command.c:445
> >> > #3  0xffffffff801b5ba0 in db_command_loop ()
> >> > #at /usr/src/sys/ddb/db_command.c:498 4  0xffffffff801b7b69 in
> >> > #db_trap (type=Variable "type" is not available.
> >> > ) at /usr/src/sys/ddb/db_main.c:229
> >> > #5  0xffffffff8038d775 in kdb_trap (type=3, code=0,
> >> > #tf=0xffffff8000017760) at /usr/src/sys/kern/subr_kdb.c:535 6
> >> > #0xffffffff8060e170 in trap (frame=0xffffff8000017760)
> >> > #at /usr/src/sys/amd64/amd64/trap.c:611 7  0xffffffff805f3b23 in
> >> > #calltrap () at /usr/src/sys/amd64/amd64/exception.S:224 8
> >> > #0xffffffff8038d94d in kdb_enter (why=0xffffffff80681954 "panic",
> >> > #msg=0xa <Address 0xa out of bounds>) at cpufunc.h:63 9
> >> > #0xffffffff8035e04b in panic (fmt=Variable "fmt" is not available.
> >> > ) at /usr/src/sys/kern/kern_shutdown.c:562
> >> > #10 0xffffffff805fafbb in apic_free_vector (apic_id=Variable
> >> > #"apic_id" is not available.
> >> > ) at /usr/src/sys/amd64/amd64/local_apic.c:995
> >> > #11 0xffffffff805f7560 in intr_remove_handler (cookie=Variable
> >> > #"cookie" is not available.
> >> > ) at /usr/src/sys/amd64/amd64/intr_machdep.c:203
> >> > #12 0xffffffff806250d4 in hpt_shutdown_vbus
> >> > #(vbus_ext=0xffffff8000254000, howto=Variable "howto" is not
> >> > #available.
> >> > ) at /usr/src/sys/dev/hptrr/hptrr_osm_bsd.c:361
> >> > #13 0xffffffff8035da8b in boot (howto=16392)
> >> > #at /usr/src/sys/kern/kern_shutdown.c:419 14 0xffffffff8035e126 in
> >> > #reboot (td=Variable "td" is not available.
> >> > ) at /usr/src/sys/kern/kern_shutdown.c:173
> >> > #15 0xffffffff8060dbba in syscall (frame=0xffffff8000017c80)
> >> > #at /usr/src/sys/amd64/amd64/trap.c:982 16 0xffffffff805f3e01 in
> >> > #Xfast_syscall () at /usr/src/sys/amd64/amd64/exception.S:373 17
> >> > #0x000000000040892c in ?? ()
> >> > Previous frame inner to this frame (corrupt stack?)
> >> >
> >> > frame #12 & #13 make me suspect the hptrr driver (again :-)), but
> >> > I'm not sure.
> >>
> >> Ah, any driver shutting off its interrupt handler during shutdown
> >> would actually trigger this.  The hptrr driver doesn't really need to
> >> do this during shutdown, but I think we don't want to panic in this
> >> case. Unfortunately we don't currently have a "in_shutdown" type of
> >> global flag that we can check in apic_free_vector() to not worry
> >> about the thread already being bound.
> >
> > Is there any chance this fix might make it into 8.0? It's not a real
> > problem for me since I can just power-off by force, but I imagine it
> > might be a real pita for people who only have remote access and want
> > to reboot, etc.
> 
> Maybe we can ship 8 with a printf() instrad than a panic()?

Actually, it looks like 'rebooting' is what we want.  Try this:

--- //depot/vendor/freebsd/src/sys/amd64/amd64/local_apic.c	2009/08/14 21:10:13
+++ //depot/user/jhb/acpipci/amd64/amd64/local_apic.c	2009/09/01 13:54:23
@@ -990,18 +990,21 @@
 	 * we don't lose an interrupt delivery race.
 	 */
 	td = curthread;
-	thread_lock(td);
-	if (sched_is_bound(td))
-		panic("apic_free_vector: Thread already bound.\n");
-	sched_bind(td, apic_cpuid(apic_id));
-	thread_unlock(td);
+	if (!rebooting) {
+		thread_lock(td);
+		if (sched_is_bound(td))
+			panic("apic_free_vector: Thread already bound.\n");
+		sched_bind(td, apic_cpuid(apic_id));
+		thread_unlock(td);
+	}
 	mtx_lock_spin(&icu_lock);
 	lapics[apic_id].la_ioint_irqs[vector - APIC_IO_INTS] = -1;
 	mtx_unlock_spin(&icu_lock);
-	thread_lock(td);
-	sched_unbind(td);
-	thread_unlock(td);
-
+	if (!rebooting) {
+		thread_lock(td);
+		sched_unbind(td);
+		thread_unlock(td);
+	}
 }
 
 /* Map an IDT vector (APIC) to an IRQ (interrupt source). */
--- //depot/vendor/freebsd/src/sys/i386/i386/local_apic.c	2009/08/14 21:10:13
+++ //depot/user/jhb/acpipci/i386/i386/local_apic.c	2009/09/01 13:53:14
@@ -994,18 +994,21 @@
 	 * we don't lose an interrupt delivery race.
 	 */
 	td = curthread;
-	thread_lock(td);
-	if (sched_is_bound(td))
-		panic("apic_free_vector: Thread already bound.\n");
-	sched_bind(td, apic_cpuid(apic_id));
-	thread_unlock(td);
+	if (!rebooting) {
+		thread_lock(td);
+		if (sched_is_bound(td))
+			panic("apic_free_vector: Thread already bound.\n");
+		sched_bind(td, apic_cpuid(apic_id));
+		thread_unlock(td);
+	}
 	mtx_lock_spin(&icu_lock);
 	lapics[apic_id].la_ioint_irqs[vector - APIC_IO_INTS] = -1;
 	mtx_unlock_spin(&icu_lock);
-	thread_lock(td);
-	sched_unbind(td);
-	thread_unlock(td);
-
+	if (!rebooting) {
+		thread_lock(td);
+		sched_unbind(td);
+		thread_unlock(td);
+	}
 }
 
 /* Map an IDT vector (APIC) to an IRQ (interrupt source). */

-- 
John Baldwin



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200909010955.33747.jhb>