NetBSD Problem Report #23208
Received: (qmail 7771 invoked by uid 605); 20 Oct 2003 16:57:20 -0000
Message-Id: <200310201657.h9KGvFWr000555@induction.immanent.net>
Date: Mon, 20 Oct 2003 11:57:16 -0500 (CDT)
From: Frederick Bruckman <fredb@immanent.net>
Sender: gnats-bugs-owner@NetBSD.org
Reply-To: fredb@netbsd.org
To: gnats-bugs@gnats.netbsd.org
Subject: atabus changes in 1.6ZD break IDE on Quadra 630
X-Send-Pr-Version: 3.95
>Number: 23208
>Category: port-mac68k
>Synopsis: atabus changes in 1.6ZD break IDE on Quadra 630
>Confidential: no
>Severity: critical
>Priority: high
>Responsible: bouyer
>State: closed
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Mon Oct 20 16:58:00 +0000 2003
>Closed-Date: Wed Oct 29 21:45:16 +0000 2003
>Last-Modified: Wed Oct 29 21:45:16 +0000 2003
>Originator: Frederick Bruckman
>Release: NetBSD 1.6ZD
>Organization:
>Environment:
System: NetBSD induction.immanent.net 1.6U NetBSD 1.6U (INDUCTION) #0: Mon Jun 16 09:34:19 CDT 2003 fredb@tautology.immanent.net:/t/netbsd-current/obj.mac68k/s/netbsd/src-current/sys/arch/mac68k/compile/INDUCTION mac68k
Architecture: m68k
Machine: mac68k
The problem kernel was built from sources yesterday, October 19,
and reports as NetBSD 1.6ZD.
>Description:
Quadra 630/IDE drive used to work as of 1.6U. Now, the controller
and drive are still detected, but it panics after "root on ffs" w/
Stopped in pid 3.1 (atabus0)
panic: queue_freese
No, it's not a typo -- it really says "freese", with an "s".
"bt" in the debugger only shows that panic() is called from
atabus_thread(). "sync" seems to actually take a core dump
before going into a UVM fault, but later, savecore says:
Oct 20 11:42:51 induction savecore: can't find device 2365/480
This is probably not related to the problem at hand.
>How-To-Repeat:
Try to boot a recent kernel on a Mac with IDE root.
>Fix:
>Release-Note:
>Audit-Trail:
Responsible-Changed-From-To: port-mac68k-maintainer->bouyer
Responsible-Changed-By: bouyer
Responsible-Changed-When: Wed Oct 22 20:58:25 UTC 2003
Responsible-Changed-Why:
Looks liek it's my fault :)
From: Manuel Bouyer <bouyer@antioche.eu.org>
To: fredb@NetBSD.org
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: port-mac68k/23208: atabus changes in 1.6ZD break IDE on Quadra 630
Date: Wed, 22 Oct 2003 23:01:08 +0200
On Mon, Oct 20, 2003 at 11:57:16AM -0500, Frederick Bruckman wrote:
> >Description:
>
> Quadra 630/IDE drive used to work as of 1.6U. Now, the controller
> and drive are still detected, but it panics after "root on ffs" w/
>
> Stopped in pid 3.1 (atabus0)
> panic: queue_freese
>
> No, it's not a typo -- it really says "freese", with an "s".
Ops, I'll fix this with some other commits
> "bt" in the debugger only shows that panic() is called from
> atabus_thread(). "sync" seems to actually take a core dump
Can you add printfs to see
- what is the value of queue_freese when it panics
- and then track the queue_freese++/queue_freese-- (it's in sys/dev/ic/wdc.c,
sys/dev/ata/ata_wdc.c and sys/dev/scsipi/atapi_wdc.c)
--
Manuel Bouyer <bouyer@antioche.eu.org>
NetBSD: 24 ans d'experience feront toujours la difference
--
From: Frederick Bruckman <fredb@immanent.net>
To: Manuel Bouyer <bouyer@antioche.eu.org>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: port-mac68k/23208: atabus changes in 1.6ZD break IDE on Quadra
630
Date: Wed, 22 Oct 2003 18:21:38 -0500 (CDT)
On Wed, 22 Oct 2003, Manuel Bouyer wrote:
> On Mon, Oct 20, 2003 at 11:57:16AM -0500, Frederick Bruckman wrote:
> > >Description:
> >
> > Quadra 630/IDE drive used to work as of 1.6U. Now, the controller
> > and drive are still detected, but it panics after "root on ffs" w/
> >
> > Stopped in pid 3.1 (atabus0)
> > panic: queue_freese
> >
> > No, it's not a typo -- it really says "freese", with an "s".
>
> Ops, I'll fix this with some other commits
:-)
> > "bt" in the debugger only shows that panic() is called from
> > atabus_thread(). "sync" seems to actually take a core dump
>
> Can you add printfs to see
> - what is the value of queue_freese when it panics
> - and then track the queue_freese++/queue_freese-- (it's in sys/dev/ic/wdc.c,
> sys/dev/ata/ata_wdc.c and sys/dev/scsipi/atapi_wdc.c)
OK. I added one for the decrement, too..., but it never decrements.
This is typed by hand:
[...]
esp0 at obio0 (quick): address 0x87a000: NCR53C96, 16MHz, SCSI ID 7
scsibus0 at esp0: 8 targets, 8 luns per target
wdc0 at obio0 (Quadra/Performa series IDE interface)
atabus0 at wdc0 channel 0
[...]
scsibus0: waiting 2 seconds for devices to settle...
wd0 at atabus0 drive 0cd0 at scsibus0 target 5 lun 0: <MATSHITA, CD-ROM CD-8004A, 2.0a> cdrom removable
: <ST340016A>
wd0: 38166 MB, 77545 cyl, 16 head, 63 sec, 512 bytes/sect x 78165360 sectors
cd0: sync (500.00ns offset 8), 8-bit (2,000MB/s) transfers
wd0: drive supports PIO mode 4, DMA mode 3, Ultar-DMA mode 5 (Ultra/100)
boot device: wd0
root on wd0a dumps on wd0b
wdcwait: queue_freese is 1mountroot: trying msdos...
wdcwait: queue_freese is 2mountroot: trying cd9660...
wdcwait: queue_freese is 3mountroot: trying nfs...
mountroot: trying ffs...
wdcwait: queue_freese is 4time read from PRAM: 0xbbbc74cb
Date and time: October 22, 2003 17:54:19
PRAM: 0x3f96c44b, macos_boottime: 0x3f96c43c
root file system tyope: ffs
panic: atabus_thread: queue_freese is 4
Stopped in pid 3.1 (atabus0) at netbsd:cpu_Debugger+0x6: unlk a6
db>
ST340016A is the model of the IDE drive. On 1.6U, the messages don't
run together, but then wd0 is attached much earlier, right after wdc0.
Frederick
From: Manuel Bouyer <bouyer@antioche.eu.org>
To: Frederick Bruckman <fredb@immanent.net>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: port-mac68k/23208: atabus changes in 1.6ZD break IDE on Quadra 630
Date: Thu, 23 Oct 2003 21:39:14 +0200
--AqsLC8rIMeq19msA
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
On Wed, Oct 22, 2003 at 06:21:38PM -0500, Frederick Bruckman wrote:
> OK. I added one for the decrement, too..., but it never decrements.
> This is typed by hand:
>
>
> [...]
> esp0 at obio0 (quick): address 0x87a000: NCR53C96, 16MHz, SCSI ID 7
> scsibus0 at esp0: 8 targets, 8 luns per target
> wdc0 at obio0 (Quadra/Performa series IDE interface)
> atabus0 at wdc0 channel 0
> [...]
> scsibus0: waiting 2 seconds for devices to settle...
> wd0 at atabus0 drive 0cd0 at scsibus0 target 5 lun 0: <MATSHITA, CD-ROM CD-8004A, 2.0a> cdrom removable
> : <ST340016A>
> wd0: 38166 MB, 77545 cyl, 16 head, 63 sec, 512 bytes/sect x 78165360 sectors
> cd0: sync (500.00ns offset 8), 8-bit (2,000MB/s) transfers
> wd0: drive supports PIO mode 4, DMA mode 3, Ultar-DMA mode 5 (Ultra/100)
> boot device: wd0
> root on wd0a dumps on wd0b
> wdcwait: queue_freese is 1mountroot: trying msdos...
> wdcwait: queue_freese is 2mountroot: trying cd9660...
> wdcwait: queue_freese is 3mountroot: trying nfs...
OK, I suspect something is calling wdcwait() with the wrong parameters.
Can you try the attached patch and report the backtrace after the panic ?
>
> ST340016A is the model of the IDE drive. On 1.6U, the messages don't
> run together, but then wd0 is attached much earlier, right after wdc0.
Yes, with the last changes the probe/attach is done in a kernel thread context,
with interrupts enabled.
--
Manuel Bouyer <bouyer@antioche.eu.org>
NetBSD: 24 ans d'experience feront toujours la difference
--
--AqsLC8rIMeq19msA
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename=diff2
Index: wdc.c
===================================================================
RCS file: /cvsroot/src/sys/dev/ic/wdc.c,v
retrieving revision 1.144
diff -u -r1.144 wdc.c
--- wdc.c 2003/10/22 23:59:00 1.144
+++ wdc.c 2003/10/23 19:36:08
@@ -1311,6 +1311,8 @@
* we're probably in interrupt context,
* ask the thread to come back here
*/
+ if (chp->ch_queue->queue_freese > 0)
+ panic("wdcwait: queue_freese");
chp->ch_queue->queue_freese++;
wakeup(&chp->thread);
return(WDCWAIT_THR);
--AqsLC8rIMeq19msA--
From: Frederick Bruckman <fredb@immanent.net>
To: Manuel Bouyer <bouyer@antioche.eu.org>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: port-mac68k/23208: atabus changes in 1.6ZD break IDE on Quadra
630
Date: Thu, 23 Oct 2003 18:27:22 -0500 (CDT)
On Thu, 23 Oct 2003, Manuel Bouyer wrote:
> OK, I suspect something is calling wdcwait() with the wrong parameters.
> Can you try the attached patch and report the backtrace after the panic ?
root on wd0a: dumps on wd0b
mountroot: trying msdos...
panic: wdcwait: queue_freese
Stopped in pid 0.1 (swapper) at netbsd:cpu_Debugger+0x6: unlk a6
db> bt
cpu_Debugger()
panic()
wdcwait()
__wdccommand_intr()
__wdccommand_start()
wdcstart()
wdc_exec_xfer()
wdc_exec_command()
ata_get_params()
wd_get_params()
wdopen()
spec_open()
VOP_OPEN()
msdosfs_mountfs()
msdosfs_mountroot()
vfs_mountroot()
main()
?() at fee
(Numbers omitted.) I have a kernel with debugging symbols on the build
host, and tried to take a core dump, but it wouldn't do it:
db> sync
syncing disks... done
NetBSD/mac68k does not trust itself to update the RTC on shutdown.
dumping to dev 22,1 offset 147469
dump panic: wddump: polled command has been queued
Stopped in pid 0.1 (swapper) at netbsd:cpu_Debugger()
Frederick
From: Manuel Bouyer <bouyer@antioche.eu.org>
To: Frederick Bruckman <fredb@immanent.net>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: port-mac68k/23208: atabus changes in 1.6ZD break IDE on Quadra 630
Date: Fri, 24 Oct 2003 22:19:04 +0200
--CE+1k2dSO48ffgeK
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Hi,
On Thu, Oct 23, 2003 at 06:27:22PM -0500, Frederick Bruckman wrote:
> On Thu, 23 Oct 2003, Manuel Bouyer wrote:
>
> > OK, I suspect something is calling wdcwait() with the wrong parameters.
> > Can you try the attached patch and report the backtrace after the panic ?
>
> root on wd0a: dumps on wd0b
> mountroot: trying msdos...
> panic: wdcwait: queue_freese
> Stopped in pid 0.1 (swapper) at netbsd:cpu_Debugger+0x6: unlk a6
> db> bt
> cpu_Debugger()
> panic()
> wdcwait()
> __wdccommand_intr()
> __wdccommand_start()
> wdcstart()
> wdc_exec_xfer()
> wdc_exec_command()
> ata_get_params()
> wd_get_params()
> wdopen()
> spec_open()
> VOP_OPEN()
> msdosfs_mountfs()
> msdosfs_mountroot()
> vfs_mountroot()
> main()
> ?() at fee
OK, thanks.
I suspect your controller is WDC_CAPABILITY_NOIRQ.
Can you try the attached patch ? This should solve this case.
--
Manuel Bouyer <bouyer@antioche.eu.org>
NetBSD: 24 ans d'experience feront toujours la difference
--
--CE+1k2dSO48ffgeK
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename=diff
Index: wdc.c
===================================================================
RCS file: /cvsroot/src/sys/dev/ic/wdc.c,v
retrieving revision 1.144
diff -u -r1.144 wdc.c
--- wdc.c 2003/10/22 23:59:00 1.144
+++ wdc.c 2003/10/24 20:14:58
@@ -1294,9 +1294,11 @@
else {
error = __wdcwait(chp, mask, bits, WDCDELAY_POLL);
if (error != 0) {
- if (chp->ch_flags & WDCF_TH_RUN) {
+ if ((chp->ch_flags & WDCF_TH_RUN) ||
+ (flags & AT_WAIT)) {
/*
- * we're running in the channel thread context
+ * we're running in the channel thread
+ * or some userland thread context
*/
for (i = 0; i < timeout_hz; i++) {
if (__wdcwait(chp, mask, bits,
@@ -1311,6 +1313,10 @@
* we're probably in interrupt context,
* ask the thread to come back here
*/
+#ifdef DIAGNOSTIC
+ if (chp->ch_queue->queue_freese > 0)
+ panic("wdcwait: queue_freese");
+#endif
chp->ch_queue->queue_freese++;
wakeup(&chp->thread);
return(WDCWAIT_THR);
--CE+1k2dSO48ffgeK--
From: Frederick Bruckman <fredb@immanent.net>
To: Manuel Bouyer <bouyer@antioche.eu.org>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: port-mac68k/23208: atabus changes in 1.6ZD break IDE on Quadra
630
Date: Fri, 24 Oct 2003 20:25:10 -0500 (CDT)
On Fri, 24 Oct 2003, Manuel Bouyer wrote:
> On Thu, Oct 23, 2003 at 06:27:22PM -0500, Frederick Bruckman wrote:
> > On Thu, 23 Oct 2003, Manuel Bouyer wrote:
> >
> > > OK, I suspect something is calling wdcwait() with the wrong parameters.
> > > Can you try the attached patch and report the backtrace after the panic ?
> >
> > root on wd0a: dumps on wd0b
> > mountroot: trying msdos...
> > panic: wdcwait: queue_freese
> > Stopped in pid 0.1 (swapper) at netbsd:cpu_Debugger+0x6: unlk a6
> > db> bt
> > cpu_Debugger()
> > panic()
> > wdcwait()
> > __wdccommand_intr()
> > __wdccommand_start()
> > wdcstart()
> > wdc_exec_xfer()
> > wdc_exec_command()
> > ata_get_params()
> > wd_get_params()
> > wdopen()
> > spec_open()
> > VOP_OPEN()
> > msdosfs_mountfs()
> > msdosfs_mountroot()
> > vfs_mountroot()
> > main()
> > ?() at fee
>
> OK, thanks.
> I suspect your controller is WDC_CAPABILITY_NOIRQ.
Yes, I think I remember something about that.
> Can you try the attached patch ? This should solve this case.
Sadly no, it's exactly the same as before (with DIAGNOSTIC -- haven't
tried without).
Frederick
From: Manuel Bouyer <bouyer@antioche.eu.org>
To: Frederick Bruckman <fredb@immanent.net>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: port-mac68k/23208: atabus changes in 1.6ZD break IDE on Quadra 630
Date: Mon, 27 Oct 2003 22:19:50 +0100
--tThc/1wpZn/ma/RB
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
On Fri, Oct 24, 2003 at 08:25:10PM -0500, Frederick Bruckman wrote:
> > Can you try the attached patch ? This should solve this case.
>
> Sadly no, it's exactly the same as before (with DIAGNOSTIC -- haven't
> tried without).
Sorry, I forgot to change one line.
In __wdccommand_intr(), line 1798, please change
wflags = 0;
to
wflags = AT_WAIT | AT_POLL;
Or use the attached patch :)
--
Manuel Bouyer <bouyer@antioche.eu.org>
NetBSD: 24 ans d'experience feront toujours la difference
--
--tThc/1wpZn/ma/RB
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename=diff
Index: wdc.c
===================================================================
RCS file: /cvsroot/src/sys/dev/ic/wdc.c,v
retrieving revision 1.144
diff -u -r1.144 wdc.c
--- wdc.c 2003/10/22 23:59:00 1.144
+++ wdc.c 2003/10/27 21:19:05
@@ -268,11 +268,15 @@
chp->ch_flags |= WDCF_TH_RUN;
splx(s);
atabusconfig(atabus_sc);
- while (!(chp->ch_flags & WDCF_SHUTDOWN)) {
+ for(;;) {
s = splbio();
- chp->ch_flags &= ~WDCF_TH_RUN;
- tsleep(&chp->thread, PRIBIO, "atath", 0);
- chp->ch_flags |= WDCF_TH_RUN;
+ if ((chp->ch_flags & (WDCF_TH_RESET | WDCF_SHUTDOWN)) == 0 &&
+ ((chp->ch_flags & WDCF_ACTIVE) == 0 ||
+ chp->ch_queue->queue_freese == 0)) {
+ chp->ch_flags &= ~WDCF_TH_RUN;
+ tsleep(&chp->thread, PRIBIO, "atath", 0);
+ chp->ch_flags |= WDCF_TH_RUN;
+ }
splx(s);
if (chp->ch_flags & WDCF_SHUTDOWN)
break;
@@ -284,6 +288,7 @@
chp->ch_drive[drive].state = 0;
}
chp->ch_flags &= ~WDCF_TH_RESET;
+ chp->ch_queue->queue_freese--;
wdcstart(chp);
} else if ((chp->ch_flags & WDCF_ACTIVE) != 0 &&
chp->ch_queue->queue_freese == 1) {
@@ -983,8 +988,8 @@
if ((chp->ch_flags & WDCF_ACTIVE) != 0 ) {
return; /* channel aleady active */
}
- if ((chp->ch_flags & WDCF_TH_RESET) != 0) {
- return; /* a channel reset is pending */
+ if (__predict_false(chp->ch_queue->queue_freese > 0)) {
+ return; /* queue froozen */
}
#ifdef DIAGNOSTIC
if ((chp->ch_flags & WDCF_IRQ_WAIT) != 0)
@@ -1076,13 +1081,13 @@
WDCDEBUG_PRINT(("ata_reset_channel %s:%d for drive %d\n",
chp->wdc->sc_dev.dv_xname, chp->channel, drvp->drive),
DEBUG_FUNCS);
- if ((chp->ch_flags & WDCF_TH_RUN) == 0 &&
- (flags & AT_POLL) == 0) {
+ if ((flags & AT_POLL) == 0) {
chp->ch_flags |= WDCF_TH_RESET;
+ chp->ch_queue->queue_freese++;
wakeup(&chp->thread);
return;
}
- (void) wdcreset(chp, (flags & AT_POLL) ? RESET_POLL : RESET_SLEEP);
+ (void) wdcreset(chp, RESET_POLL);
for (drive = 0; drive < 2; drive++) {
chp->ch_drive[drive].state = 0;
}
@@ -1294,9 +1299,11 @@
else {
error = __wdcwait(chp, mask, bits, WDCDELAY_POLL);
if (error != 0) {
- if (chp->ch_flags & WDCF_TH_RUN) {
+ if ((chp->ch_flags & WDCF_TH_RUN) ||
+ (flags & AT_WAIT)) {
/*
- * we're running in the channel thread context
+ * we're running in the channel thread
+ * or some userland thread context
*/
for (i = 0; i < timeout_hz; i++) {
if (__wdcwait(chp, mask, bits,
@@ -1311,6 +1318,10 @@
* we're probably in interrupt context,
* ask the thread to come back here
*/
+#ifdef DIAGNOSTIC
+ if (chp->ch_queue->queue_freese > 0)
+ panic("wdcwait: queue_freese");
+#endif
chp->ch_queue->queue_freese++;
wakeup(&chp->thread);
return(WDCWAIT_THR);
@@ -1784,7 +1795,7 @@
if ((wdc_c->flags & (AT_WAIT | AT_POLL)) == (AT_WAIT | AT_POLL)) {
/* both wait and poll, we can tsleep here */
- wflags = 0;
+ wflags = AT_WAIT | AT_POLL;
} else {
wflags = AT_POLL;
}
--tThc/1wpZn/ma/RB--
From: Frederick Bruckman <fredb@immanent.net>
To: Manuel Bouyer <bouyer@antioche.eu.org>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: port-mac68k/23208: atabus changes in 1.6ZD break IDE on Quadra
630
Date: Mon, 27 Oct 2003 17:20:28 -0600 (CST)
On Mon, 27 Oct 2003, Manuel Bouyer wrote:
> On Fri, Oct 24, 2003 at 08:25:10PM -0500, Frederick Bruckman wrote:
> > > Can you try the attached patch ? This should solve this case.
> >
> > Sadly no, it's exactly the same as before (with DIAGNOSTIC -- haven't
> > tried without).
>
> Sorry, I forgot to change one line.
> In __wdccommand_intr(), line 1798, please change
> wflags = 0;
> to
> wflags = AT_WAIT | AT_POLL;
>
> Or use the attached patch :)
Yes, the patch works fine.
Thanks!
Frederick
State-Changed-From-To: open->closed
State-Changed-By: bouyer
State-Changed-When: Wed Oct 29 21:43:43 UTC 2003
State-Changed-Why:
Fixed in wdc.c 1.147.
Thanks for the report !
>Unformatted:
(Contact us)
$NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026
The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.