Skip to contentSkip to conversation

swsusp: Change code ordering related to ACPI

PATCH In review Rafael J. Wysocki 3 participants 11 messages May 4, 2007 – May 5, 2007

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 swsusp: Change code ordering related to ACPI

Hi,

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It has been confirmed that the appended patch fixes the problem, but it's not certain if this changes don't break some other systems. For this reason, all users of hibernation (swsusp, uswsusp) are gently requested to verify if this patch doesn't break their systems.

Greetings,
Rafael

---
From: Rafael J. Wysocki <[email protected]>

The current code ordering in the hibernation code paths causes some systems to have problems. It has been confirmed by testers that these problems do not appear if the acpi_pm_finish() function (called via platform_finish()) is executed after device_suspend().

---
 kernel/power/disk.c |    4 ++--
 kernel/power/user.c |    8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)
kernel/power/disk.c +2 -2
Index: linux-2.6.21/kernel/power/disk.c
===================================================================
@@ -195,9 +195,9 @@ int hibernate(void)
 
 	if (in_suspend) {
 		enable_nonboot_cpus();
-		platform_finish();
 		device_resume();
 		resume_console();
+		platform_finish();
 		pr_debug("PM: writing image.\n");
 		error = swsusp_write();
 		if (!error)
@@ -214,9 +214,9 @@ int hibernate(void)
  Enable_cpus:
 	enable_nonboot_cpus();
  Resume_devices:
-	platform_finish();
 	device_resume();
 	resume_console();
+	platform_finish();
  Thaw:
 	mutex_unlock(&pm_mutex);
 	unprepare_processes();
kernel/power/user.c +4 -4
Index: linux-2.6.21/kernel/power/user.c
===================================================================
@@ -169,11 +169,11 @@ static inline int snapshot_suspend(int p
 	}
 	enable_nonboot_cpus();
  Resume_devices:
+	device_resume();
+	resume_console();
 	if (platform_suspend)
 		platform_finish();
 
-	device_resume();
-	resume_console();
  Finish:
 	mutex_unlock(&pm_mutex);
 	return error;
@@ -201,11 +201,11 @@ static inline int snapshot_restore(int p
 
 	enable_nonboot_cpus();
  Resume_devices:
+	device_resume();
+	resume_console();
 	if (platform_suspend)
 		platform_finish();
 
-	device_resume();
-	resume_console();
  Finish:
 	pm_restore_console();
 	mutex_unlock(&pm_mutex);
#1 lore raw

Doing _WAK (from pm_finish) before device_resume will brake almost all ACPI devices.

Regards,
Alex.

Quoted 91 lines · Rafael J. Wysocki

Hi,

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It has been confirmed that the appended patch fixes the problem, but it's not certain if this changes don't break some other systems. For this reason, all users of hibernation (swsusp, uswsusp) are gently requested to verify if this patch doesn't break their systems.

Greetings,
Rafael

---
From: Rafael J. Wysocki <[email protected]>

The current code ordering in the hibernation code paths causes some systems to have problems. It has been confirmed by testers that these problems do not appear if the acpi_pm_finish() function (called via platform_finish()) is executed after device_suspend().

---
 kernel/power/disk.c |    4 ++--
 kernel/power/user.c |    8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)

Index: linux-2.6.21/kernel/power/disk.c
===================================================================

kernel/power/disk.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/disk.c 2007-05-04 10:31:51.000000000 +0200 +2 -2
@@ -195,9 +195,9 @@ int hibernate(void)
 
        if (in_suspend) {
                enable_nonboot_cpus();
-               platform_finish();
                device_resume();
                resume_console();
+               platform_finish();
                pr_debug("PM: writing image.\n");
                error = swsusp_write();
                if (!error)
@@ -214,9 +214,9 @@ int hibernate(void)
  Enable_cpus:
        enable_nonboot_cpus();
  Resume_devices:
-       platform_finish();
        device_resume();
        resume_console();
+       platform_finish();
  Thaw:
        mutex_unlock(&pm_mutex);
        unprepare_processes();

Index: linux-2.6.21/kernel/power/user.c
===================================================================

kernel/power/user.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/user.c 2007-05-04 10:31:51.000000000 +0200 +4 -4
@@ -169,11 +169,11 @@ static inline int snapshot_suspend(int p
        }
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        mutex_unlock(&pm_mutex);
        return error;
@@ -201,11 +201,11 @@ static inline int snapshot_restore(int p
 
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        pm_restore_console();
        mutex_unlock(&pm_mutex);

_______________________________________________
linux-pm mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/linux-pm

#2 lore raw
In reply to Rafael J. Wysocki

Doing _WAK (from pm_finish) before device_resume will brake almost all ACPI devices.

Regards,
Alex.

Quoted 91 lines · Rafael J. Wysocki

Hi,

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It has been confirmed that the appended patch fixes the problem, but it's not certain if this changes don't break some other systems. For this reason, all users of hibernation (swsusp, uswsusp) are gently requested to verify if this patch doesn't break their systems.

Greetings,
Rafael

---
From: Rafael J. Wysocki <[email protected]>

The current code ordering in the hibernation code paths causes some systems to have problems. It has been confirmed by testers that these problems do not appear if the acpi_pm_finish() function (called via platform_finish()) is executed after device_suspend().

---
 kernel/power/disk.c |    4 ++--
 kernel/power/user.c |    8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)

Index: linux-2.6.21/kernel/power/disk.c
===================================================================

kernel/power/disk.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/disk.c 2007-05-04 10:31:51.000000000 +0200 +2 -2
@@ -195,9 +195,9 @@ int hibernate(void)
 
        if (in_suspend) {
                enable_nonboot_cpus();
-               platform_finish();
                device_resume();
                resume_console();
+               platform_finish();
                pr_debug("PM: writing image.\n");
                error = swsusp_write();
                if (!error)
@@ -214,9 +214,9 @@ int hibernate(void)
  Enable_cpus:
        enable_nonboot_cpus();
  Resume_devices:
-       platform_finish();
        device_resume();
        resume_console();
+       platform_finish();
  Thaw:
        mutex_unlock(&pm_mutex);
        unprepare_processes();

Index: linux-2.6.21/kernel/power/user.c
===================================================================

kernel/power/user.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/user.c 2007-05-04 10:31:51.000000000 +0200 +4 -4
@@ -169,11 +169,11 @@ static inline int snapshot_suspend(int p
        }
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        mutex_unlock(&pm_mutex);
        return error;
@@ -201,11 +201,11 @@ static inline int snapshot_restore(int p
 
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        pm_restore_console();
        mutex_unlock(&pm_mutex);

_______________________________________________
linux-pm mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/linux-pm

#3 lore raw

Sorry, I meant executing _WAK _after_ resume of devices...

Alex.

Quoted 99 lines · Alexey Starikovskiy

Doing _WAK (from pm_finish) before device_resume will brake almost all ACPI devices.

Regards,
Alex.

Quoted 91 lines · Rafael J. Wysocki

Hi,

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It has been confirmed that the appended patch fixes the problem, but it's not certain if this changes don't break some other systems. For this reason, all users of hibernation (swsusp, uswsusp) are gently requested to verify if this patch doesn't break their systems.

Greetings,
Rafael

---
From: Rafael J. Wysocki <[email protected]>

The current code ordering in the hibernation code paths causes some systems to have problems. It has been confirmed by testers that these problems do not appear if the acpi_pm_finish() function (called via platform_finish()) is executed after device_suspend().

---
 kernel/power/disk.c |    4 ++--
 kernel/power/user.c |    8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)

Index: linux-2.6.21/kernel/power/disk.c
===================================================================

kernel/power/disk.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/disk.c 2007-05-04 10:31:51.000000000 +0200 +2 -2
@@ -195,9 +195,9 @@ int hibernate(void)
 
        if (in_suspend) {
                enable_nonboot_cpus();
-               platform_finish();
                device_resume();
                resume_console();
+               platform_finish();
                pr_debug("PM: writing image.\n");
                error = swsusp_write();
                if (!error)
@@ -214,9 +214,9 @@ int hibernate(void)
  Enable_cpus:
        enable_nonboot_cpus();
  Resume_devices:
-       platform_finish();
        device_resume();
        resume_console();
+       platform_finish();
  Thaw:
        mutex_unlock(&pm_mutex);
        unprepare_processes();

Index: linux-2.6.21/kernel/power/user.c
===================================================================

kernel/power/user.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/user.c 2007-05-04 10:31:51.000000000 +0200 +4 -4
@@ -169,11 +169,11 @@ static inline int snapshot_suspend(int p
        }
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        mutex_unlock(&pm_mutex);
        return error;
@@ -201,11 +201,11 @@ static inline int snapshot_restore(int p
 
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        pm_restore_console();
        mutex_unlock(&pm_mutex);

_______________________________________________
linux-pm mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/linux-pm

#4 lore raw
In reply to Alexey Starikovskiy

Sorry, I meant executing _WAK _after_ resume of devices...

Alex.

Quoted 99 lines · Alexey Starikovskiy

Doing _WAK (from pm_finish) before device_resume will brake almost all ACPI devices.

Regards,
Alex.

Quoted 91 lines · Rafael J. Wysocki

Hi,

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It has been confirmed that the appended patch fixes the problem, but it's not certain if this changes don't break some other systems. For this reason, all users of hibernation (swsusp, uswsusp) are gently requested to verify if this patch doesn't break their systems.

Greetings,
Rafael

---
From: Rafael J. Wysocki <[email protected]>

The current code ordering in the hibernation code paths causes some systems to have problems. It has been confirmed by testers that these problems do not appear if the acpi_pm_finish() function (called via platform_finish()) is executed after device_suspend().

---
 kernel/power/disk.c |    4 ++--
 kernel/power/user.c |    8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)

Index: linux-2.6.21/kernel/power/disk.c
===================================================================

kernel/power/disk.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/disk.c 2007-05-04 10:31:51.000000000 +0200 +2 -2
@@ -195,9 +195,9 @@ int hibernate(void)
 
        if (in_suspend) {
                enable_nonboot_cpus();
-               platform_finish();
                device_resume();
                resume_console();
+               platform_finish();
                pr_debug("PM: writing image.\n");
                error = swsusp_write();
                if (!error)
@@ -214,9 +214,9 @@ int hibernate(void)
  Enable_cpus:
        enable_nonboot_cpus();
  Resume_devices:
-       platform_finish();
        device_resume();
        resume_console();
+       platform_finish();
  Thaw:
        mutex_unlock(&pm_mutex);
        unprepare_processes();

Index: linux-2.6.21/kernel/power/user.c
===================================================================

kernel/power/user.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/user.c 2007-05-04 10:31:51.000000000 +0200 +4 -4
@@ -169,11 +169,11 @@ static inline int snapshot_suspend(int p
        }
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        mutex_unlock(&pm_mutex);
        return error;
@@ -201,11 +201,11 @@ static inline int snapshot_restore(int p
 
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        pm_restore_console();
        mutex_unlock(&pm_mutex);

_______________________________________________
linux-pm mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/linux-pm

#5 lore raw
Quoted 1 line · Alexey Starikovskiy

Sorry, I meant executing _WAK _after_ resume of devices...

Somehow, that's what we've done for several months at the very least. Moreover, my test boxes apparently don't care.

OTOH, evidently, some systems have been broken by moving the call to acpi_pm_finish() before device_resume(). This seems to implicate that there are two parts of code in acpi_pm_finish(), one that should be executed before device_resume(), including the _WAK method, and the other that should be executed after device_resume(). Now, the question is how to split it into these parts.

Greetings,
Rafael

Quoted 102 lines · Alexey Starikovskiy
Quoted 99 lines

Doing _WAK (from pm_finish) before device_resume will brake almost all ACPI devices.

Regards,
Alex.

Quoted 91 lines · Rafael J. Wysocki

Hi,

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It has been confirmed that the appended patch fixes the problem, but it's not certain if this changes don't break some other systems. For this reason, all users of hibernation (swsusp, uswsusp) are gently requested to verify if this patch doesn't break their systems.

Greetings,
Rafael

---
From: Rafael J. Wysocki <[email protected]>

The current code ordering in the hibernation code paths causes some systems to have problems. It has been confirmed by testers that these problems do not appear if the acpi_pm_finish() function (called via platform_finish()) is executed after device_suspend().

---
 kernel/power/disk.c |    4 ++--
 kernel/power/user.c |    8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)

Index: linux-2.6.21/kernel/power/disk.c
===================================================================

kernel/power/disk.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/disk.c 2007-05-04 10:31:51.000000000 +0200 +2 -2
@@ -195,9 +195,9 @@ int hibernate(void)
 
        if (in_suspend) {
                enable_nonboot_cpus();
-               platform_finish();
                device_resume();
                resume_console();
+               platform_finish();
                pr_debug("PM: writing image.\n");
                error = swsusp_write();
                if (!error)
@@ -214,9 +214,9 @@ int hibernate(void)
  Enable_cpus:
        enable_nonboot_cpus();
  Resume_devices:
-       platform_finish();
        device_resume();
        resume_console();
+       platform_finish();
  Thaw:
        mutex_unlock(&pm_mutex);
        unprepare_processes();

Index: linux-2.6.21/kernel/power/user.c
===================================================================

kernel/power/user.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/user.c 2007-05-04 10:31:51.000000000 +0200 +4 -4
@@ -169,11 +169,11 @@ static inline int snapshot_suspend(int p
        }
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        mutex_unlock(&pm_mutex);
        return error;
@@ -201,11 +201,11 @@ static inline int snapshot_restore(int p
 
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        pm_restore_console();
        mutex_unlock(&pm_mutex);

_______________________________________________
linux-pm mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/linux-pm

Signature · 3 lines
If you don't have the time to read,
you don't have the time or the tools to write.
		- Stephen King
#6 lore raw
In reply to Alexey Starikovskiy
Quoted 1 line · Alexey Starikovskiy

Sorry, I meant executing _WAK _after_ resume of devices...

Somehow, that's what we've done for several months at the very least. Moreover, my test boxes apparently don't care.

OTOH, evidently, some systems have been broken by moving the call to acpi_pm_finish() before device_resume(). This seems to implicate that there are two parts of code in acpi_pm_finish(), one that should be executed before device_resume(), including the _WAK method, and the other that should be executed after device_resume(). Now, the question is how to split it into these parts.

Greetings,
Rafael

Quoted 102 lines · Alexey Starikovskiy
Quoted 99 lines

Doing _WAK (from pm_finish) before device_resume will brake almost all ACPI devices.

Regards,
Alex.

Quoted 91 lines · Rafael J. Wysocki

Hi,

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It has been confirmed that the appended patch fixes the problem, but it's not certain if this changes don't break some other systems. For this reason, all users of hibernation (swsusp, uswsusp) are gently requested to verify if this patch doesn't break their systems.

Greetings,
Rafael

---
From: Rafael J. Wysocki <[email protected]>

The current code ordering in the hibernation code paths causes some systems to have problems. It has been confirmed by testers that these problems do not appear if the acpi_pm_finish() function (called via platform_finish()) is executed after device_suspend().

---
 kernel/power/disk.c |    4 ++--
 kernel/power/user.c |    8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)

Index: linux-2.6.21/kernel/power/disk.c
===================================================================

kernel/power/disk.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/disk.c 2007-05-04 10:31:51.000000000 +0200 +2 -2
@@ -195,9 +195,9 @@ int hibernate(void)
 
        if (in_suspend) {
                enable_nonboot_cpus();
-               platform_finish();
                device_resume();
                resume_console();
+               platform_finish();
                pr_debug("PM: writing image.\n");
                error = swsusp_write();
                if (!error)
@@ -214,9 +214,9 @@ int hibernate(void)
  Enable_cpus:
        enable_nonboot_cpus();
  Resume_devices:
-       platform_finish();
        device_resume();
        resume_console();
+       platform_finish();
  Thaw:
        mutex_unlock(&pm_mutex);
        unprepare_processes();

Index: linux-2.6.21/kernel/power/user.c
===================================================================

kernel/power/user.c 2007-05-04 10:32:12.000000000 +0200 renamed from kernel/power/user.c 2007-05-04 10:31:51.000000000 +0200 +4 -4
@@ -169,11 +169,11 @@ static inline int snapshot_suspend(int p
        }
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        mutex_unlock(&pm_mutex);
        return error;
@@ -201,11 +201,11 @@ static inline int snapshot_restore(int p
 
        enable_nonboot_cpus();
  Resume_devices:
+       device_resume();
+       resume_console();
        if (platform_suspend)
                platform_finish();
 
-       device_resume();
-       resume_console();
  Finish:
        pm_restore_console();
        mutex_unlock(&pm_mutex);

_______________________________________________
linux-pm mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/linux-pm

Signature · 3 lines
If you don't have the time to read,
you don't have the time or the tools to write.
		- Stephen King
#7 lore raw
In reply to Rafael J. Wysocki
Quoted 3 lines · Rafael J. Wysocki

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It seems that somewhere between 2.6.21-rc4 and 2.6.21 final my laptop stopped being able to come out of suspend to RAM. Before I start bisecting (again, sigh), is this ringing any bells for anyone? In particular your, patch (snipped) that deals with hibernation, would it also affect suspend to RAM?

Ray

#8 lore raw
In reply to Rafael J. Wysocki
Quoted 3 lines · Rafael J. Wysocki

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It seems that somewhere between 2.6.21-rc4 and 2.6.21 final my laptop stopped being able to come out of suspend to RAM. Before I start bisecting (again, sigh), is this ringing any bells for anyone? In particular your, patch (snipped) that deals with hibernation, would it also affect suspend to RAM?

Ray

#9 lore raw
Quoted 10 lines · Ray Lee
Quoted 3 lines · Rafael J. Wysocki

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It seems that somewhere between 2.6.21-rc4 and 2.6.21 final my laptop stopped being able to come out of suspend to RAM. Before I start bisecting (again, sigh), is this ringing any bells for anyone? In particular your, patch (snipped) that deals with hibernation, would it also affect suspend to RAM?

Not this particular one, but you may try to move pm_finish() after resume_console() in kernel/power/main.c:suspend_prepare() and see if that helps.

If it doesn't help, then try to compile the kernel with NO_HZ and HIGH_RES_TIMERS unset. If that doesn't help, you may try with HPET_TIMER unset additionally.

Greetings,
Rafael

#10 lore raw
Quoted 12 lines · Rafael J. Wysocki
Quoted 10 lines · Ray Lee
Quoted 3 lines · Rafael J. Wysocki

The change of the hibernation/suspend code ordering made before 2.6.21 has caused some systems to have problems related to ACPI. In particular, the 'platform' hibernation mode doesn't work any more on some systems.

It seems that somewhere between 2.6.21-rc4 and 2.6.21 final my laptop stopped being able to come out of suspend to RAM. Before I start bisecting (again, sigh), is this ringing any bells for anyone? In particular your, patch (snipped) that deals with hibernation, would it also affect suspend to RAM?

Quoted 7 lines

Not this particular one, but you may try to move pm_finish() after resume_console() in kernel/power/main.c:suspend_prepare() and see if that helps.

If it doesn't help, then try to compile the kernel with NO_HZ and HIGH_RES_TIMERS unset. If that doesn't help, you may try with HPET_TIMER unset additionally.

Okay, my system is doing it with a known good kernel as well (2.6.21-rc4), so this is something else. Please ignore my report for now. In the meantime, I'll be downgrading my hal and drm to what they were a week ago to see if that fixes the problem.

(I knew better than to change more than one thing at a time, really I did. Sigh.)

Thanks for your help regardless,

Ray