Skip to contentSkip to conversation

-mm: PM=y, VT=n doesn't compile

DISCUSSION Adrian Bunk 4 participants 8 messages Mar 17, 2006 – Mar 20, 2006

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

swsusp-pm-refuse-to-suspend-devices-if-wrong-console-is-active.patch causes the following compile error with CONFIG_PM=y, CONFIG_VT=n:

<-- snip -->

...
LD .tmp_vmlinux1
drivers/built-in.o: In function device_suspend': undefined reference to fg_console' drivers/built-in.o: In function device_suspend': undefined reference to vc_cons'
make: *** [.tmp_vmlinux1] Error 1

<-- snip -->

cu
Adrian

Signature · 4 lines
       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed
#1 lore raw
Quoted 11 lines · Adrian Bunk

swsusp-pm-refuse-to-suspend-devices-if-wrong-console-is-active.patch causes the following compile error with CONFIG_PM=y, CONFIG_VT=n:

<-- snip -->

...
LD .tmp_vmlinux1
drivers/built-in.o: In function device_suspend': undefined reference to fg_console' drivers/built-in.o: In function device_suspend': undefined reference to vc_cons'
make: *** [.tmp_vmlinux1] Error 1

Right, thanks. I'll drop that patch.

Guys, please don't go poking around in the vt subsystems's internal data structures from within power management code.

Write a nice little helper function, put that in vt.c, give it a static inline `return 0' stub in the header file, then call that.

That's Kernel Programming 101, no?

#3 lore raw
In reply to Adrian Bunk

Hi,

Quoted 5 lines · Adrian Bunk

swsusp-pm-refuse-to-suspend-devices-if-wrong-console-is-active.patch causes the following compile error with CONFIG_PM=y, CONFIG_VT=n:

<-- snip -->

Quoted 7 lines

...
LD .tmp_vmlinux1
drivers/built-in.o: In function device_suspend': undefined reference to fg_console' drivers/built-in.o: In function device_suspend': undefined reference to vc_cons'
make: *** [.tmp_vmlinux1] Error 1

<-- snip -->

Sorry, I overlooked your message yesterday. [Actually Andrew turned my attention to it by dropping the patch from -mm. ;-)]

Fixed version of the patch is appended.

Greetings,
Rafael

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

1) Remove the console-switching code from the suspend part of the swsusp userland interface and let the userland tools switch the console.

2) It is unsafe to suspend devices if the hardware is controlled by X. Add an extra check to prevent this from happening.

---
 drivers/base/power/suspend.c |   17 +++++++++++++++++
 kernel/power/user.c          |    3 ---
 2 files changed, 17 insertions(+), 3 deletions(-)
drivers/base/power/suspend.c +17 -0
Index: linux-2.6.16-rc6-mm2/drivers/base/power/suspend.c
===================================================================
@@ -8,6 +8,9 @@
  *
  */
 
+#include <linux/vt_kern.h>
+#include <linux/kbd_kern.h>
+#include <linux/console.h>
 #include <linux/device.h>
 #include <linux/kallsyms.h>
 #include <linux/pm.h>
@@ -65,6 +68,17 @@ int suspend_device(struct device * dev, 
 	return error;
 }
 
+#ifdef CONFIG_VT
+static inline int is_suspend_console_safe(void)
+{
+	/* It is unsafe to suspend devices while X has control of the
+	 * hardware. Make sure we are running on a kernel-controlled console.
+	 */
+	return vc_cons[fg_console].d->vc_mode == KD_TEXT;
+}
+#else
+#define is_suspend_console_safe()	1
+#endif
 
 /**
  *	device_suspend - Save state and stop all devices in system.
@@ -85,6 +99,9 @@ int device_suspend(pm_message_t state)
 {
 	int error = 0;
 
+	if (!is_suspend_console_safe())
+		return -EINVAL;
+
 	down(&dpm_sem);
 	down(&dpm_list_sem);
 	while (!list_empty(&dpm_active) && error == 0) {
kernel/power/user.c +0 -3
Index: linux-2.6.16-rc6-mm2/kernel/power/user.c
===================================================================
@@ -138,12 +138,10 @@ static int snapshot_ioctl(struct inode *
 		if (data->frozen)
 			break;
 		down(&pm_sem);
-		pm_prepare_console();
 		disable_nonboot_cpus();
 		if (freeze_processes()) {
 			thaw_processes();
 			enable_nonboot_cpus();
-			pm_restore_console();
 			error = -EBUSY;
 		}
 		up(&pm_sem);
@@ -157,7 +155,6 @@ static int snapshot_ioctl(struct inode *
 		down(&pm_sem);
 		thaw_processes();
 		enable_nonboot_cpus();
-		pm_restore_console();
 		up(&pm_sem);
 		data->frozen = 0;
 		break;
#2 lore raw
In reply to Andrew Morton
Quoted 13 lines · Andrew Morton
Quoted 11 lines · Adrian Bunk

swsusp-pm-refuse-to-suspend-devices-if-wrong-console-is-active.patch causes the following compile error with CONFIG_PM=y, CONFIG_VT=n:

<-- snip -->

...
LD .tmp_vmlinux1
drivers/built-in.o: In function device_suspend': undefined reference to fg_console' drivers/built-in.o: In function device_suspend': undefined reference to vc_cons'
make: *** [.tmp_vmlinux1] Error 1

Quoted 7 lines

Right, thanks. I'll drop that patch.

Guys, please don't go poking around in the vt subsystems's internal data structures from within power management code.

Write a nice little helper function, put that in vt.c, give it a static inline `return 0' stub in the header file, then call that.

OK, I will, but the changes in kernel/power/user.c in that patch are necessary for the userland suspend we're working on right now.

Could you please accept the following patch instead of that one?

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

Remove the console-switching code from the suspend part of the swsusp userland interface and let the userland tools switch the console.

---
 kernel/power/user.c |    3 ---
 1 files changed, 3 deletions(-)
kernel/power/user.c +0 -3
Index: linux-2.6.16-rc6-mm2/kernel/power/user.c
===================================================================
@@ -138,12 +138,10 @@ static int snapshot_ioctl(struct inode *
 		if (data->frozen)
 			break;
 		down(&pm_sem);
-		pm_prepare_console();
 		disable_nonboot_cpus();
 		if (freeze_processes()) {
 			thaw_processes();
 			enable_nonboot_cpus();
-			pm_restore_console();
 			error = -EBUSY;
 		}
 		up(&pm_sem);
@@ -157,7 +155,6 @@ static int snapshot_ioctl(struct inode *
 		down(&pm_sem);
 		thaw_processes();
 		enable_nonboot_cpus();
-		pm_restore_console();
 		up(&pm_sem);
 		data->frozen = 0;
 		break;
#4 lore raw
In reply to Rafael J. Wysocki
Quoted 17 lines · Rafael J. Wysocki

--- linux-2.6.16-rc6-mm2.orig/drivers/base/power/suspend.c
+++ linux-2.6.16-rc6-mm2/drivers/base/power/suspend.c @@ -8,6 +8,9 @@
*
*/

+#include <linux/vt_kern.h>
+#include <linux/kbd_kern.h>
+#include <linux/console.h>
#include <linux/device.h>
#include <linux/kallsyms.h>
#include <linux/pm.h>
@@ -65,6 +68,17 @@ int suspend_device(struct device * dev,
return error;
}

Quoted 8 lines

+#ifdef CONFIG_VT
+static inline int is_suspend_console_safe(void)
+{
+ /* It is unsafe to suspend devices while X has control of the
+ * hardware. Make sure we are running on a kernel-controlled console.
+ */
+ return vc_cons[fg_console].d->vc_mode == KD_TEXT;
+}

Please implement this inside the vt subsystem, not the pm subsystem. That way

a) It gets to be called "console_is_in_text_mode()", or "vt_not_running_X()" or something, which is something someone else might want to know.

b) People who work on vt code don't need to keep an eye on a hunk of pm code at the same time.

c) You won't need all those includes.

Quoted 2 lines

+#else
+#define is_suspend_console_safe() 1

And this can go in console.h

Quoted 1 line

+#endif

#5 lore raw
Quoted 29 lines · Andrew Morton
Quoted 17 lines · Rafael J. Wysocki

--- linux-2.6.16-rc6-mm2.orig/drivers/base/power/suspend.c
+++ linux-2.6.16-rc6-mm2/drivers/base/power/suspend.c @@ -8,6 +8,9 @@
*
*/

+#include <linux/vt_kern.h>
+#include <linux/kbd_kern.h>
+#include <linux/console.h>
#include <linux/device.h>
#include <linux/kallsyms.h>
#include <linux/pm.h>
@@ -65,6 +68,17 @@ int suspend_device(struct device * dev,
return error;
}

Quoted 8 lines

+#ifdef CONFIG_VT
+static inline int is_suspend_console_safe(void)
+{
+ /* It is unsafe to suspend devices while X has control of the
+ * hardware. Make sure we are running on a kernel-controlled console.
+ */
+ return vc_cons[fg_console].d->vc_mode == KD_TEXT;
+}

Please implement this inside the vt subsystem, not the pm subsystem. That way

Quoted 8 lines

a) It gets to be called "console_is_in_text_mode()", or "vt_not_running_X()" or something, which is something someone else might want to know.

b) People who work on vt code don't need to keep an eye on a hunk of pm code at the same time.

c) You won't need all those includes.

I'm working on this now. [For some obscure reason I received your reply to Adrian only a couple of minutes ago.]

Still I'd like to separate it from the console-related changes in kernel/power/user.c that are needed for the userland suspend, so I'd like to split the dropped patch in two.

I've already sent the kernel/power/user.c changes in a separate patch, and I'll send the drivers/base/power/suspend.c changes when I test them (in a while).

#6 lore raw
Quoted 42 lines · Rafael J. Wysocki
Quoted 29 lines · Andrew Morton
Quoted 17 lines · Rafael J. Wysocki

--- linux-2.6.16-rc6-mm2.orig/drivers/base/power/suspend.c
+++ linux-2.6.16-rc6-mm2/drivers/base/power/suspend.c @@ -8,6 +8,9 @@
*
*/

+#include <linux/vt_kern.h>
+#include <linux/kbd_kern.h>
+#include <linux/console.h>
#include <linux/device.h>
#include <linux/kallsyms.h>
#include <linux/pm.h>
@@ -65,6 +68,17 @@ int suspend_device(struct device * dev,
return error;
}

Quoted 8 lines

+#ifdef CONFIG_VT
+static inline int is_suspend_console_safe(void)
+{
+ /* It is unsafe to suspend devices while X has control of the
+ * hardware. Make sure we are running on a kernel-controlled console.
+ */
+ return vc_cons[fg_console].d->vc_mode == KD_TEXT;
+}

Please implement this inside the vt subsystem, not the pm subsystem. That way

Quoted 8 lines

a) It gets to be called "console_is_in_text_mode()", or "vt_not_running_X()" or something, which is something someone else might want to know.

b) People who work on vt code don't need to keep an eye on a hunk of pm code at the same time.

c) You won't need all those includes.

I'm working on this now. [For some obscure reason I received your reply to Adrian only a couple of minutes ago.]

Quoted 7 lines

Still I'd like to separate it from the console-related changes in kernel/power/user.c that are needed for the userland suspend, so I'd like to split the dropped patch in two.

I've already sent the kernel/power/user.c changes in a separate patch, and I'll send the drivers/base/power/suspend.c changes when I test them (in a while).

OK, here they go.

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

It is unsafe to suspend devices if the hardware is controlled by X. Add an extra check to prevent this from happening.

---
 drivers/base/power/suspend.c |    5 ++++-
 drivers/char/vt.c            |    8 ++++++++
 include/linux/vt_kern.h      |    5 +++++
 3 files changed, 17 insertions(+), 1 deletion(-)
drivers/base/power/suspend.c +4 -1
Index: linux-2.6.16-rc6-mm2/drivers/base/power/suspend.c
===================================================================
@@ -8,6 +8,7 @@
  *
  */
 
+#include <linux/vt_kern.h>
 #include <linux/device.h>
 #include <linux/kallsyms.h>
 #include <linux/pm.h>
@@ -65,7 +66,6 @@ int suspend_device(struct device * dev, 
 	return error;
 }
 
-
 /**
  *	device_suspend - Save state and stop all devices in system.
  *	@state:		Power state to put each device in.
@@ -85,6 +85,9 @@ int device_suspend(pm_message_t state)
 {
 	int error = 0;
 
+	if (!is_console_suspend_safe())
+		return -EINVAL;
+
 	down(&dpm_sem);
 	down(&dpm_list_sem);
 	while (!list_empty(&dpm_active) && error == 0) {
drivers/char/vt.c +8 -0
Index: linux-2.6.16-rc6-mm2/drivers/char/vt.c
===================================================================
@@ -3234,6 +3234,14 @@ void vcs_scr_writew(struct vc_data *vc, 
 	}
 }
 
+int is_console_suspend_safe(void)
+{
+	/* It is unsafe to suspend devices while X has control of the
+	 * hardware. Make sure we are running on a kernel-controlled console.
+	 */
+	return vc_cons[fg_console].d->vc_mode == KD_TEXT;
+}
+
 /*
  *	Visible symbols for modules
  */
include/linux/vt_kern.h +5 -0
Index: linux-2.6.16-rc6-mm2/include/linux/vt_kern.h
===================================================================
@@ -73,6 +73,11 @@ int con_copy_unimap(struct vc_data *dst_
 int vt_waitactive(int vt);
 void change_console(struct vc_data *new_vc);
 void reset_vc(struct vc_data *vc);
+#ifdef CONFIG_VT
+int is_console_suspend_safe(void);
+#else
+static inline int is_console_suspend_safe(void) { return 1; }
+#endif
 
 /*
  * vc_screen.c shares this temporary buffer with the console write code so that
#7 lore raw

Hi!

Quoted 1 line

OK, here they go.

ACK.

include/linux/vt_kern.h int con_copy_unimap(struct vc_data *dst_ · line 77

Quoted 10 lines

Index: linux-2.6.16-rc6-mm2/include/linux/vt_kern.h
===================================================================

include/linux/vt_kern.h +2 -0
@@ -73,6 +73,11 @@ int con_copy_unimap(struct vc_data *dst_
 int vt_waitactive(int vt);
 void change_console(struct vc_data *new_vc);
 void reset_vc(struct vc_data *vc);
+#ifdef CONFIG_VT
+int is_console_suspend_safe(void);

I guess I'd just do static inline ... and put code here, but both versions are okay.

Quoted 6 lines
include/linux/vt_kern.h +3 -0
+#else
+static inline int is_console_suspend_safe(void) { return 1; }
+#endif
 
 /*
  * vc_screen.c shares this temporary buffer with the console write code so that
Signature · 1 line
42:             i < SampleTable.Length;