Software Suspend: Fix suspend when console is in VT_AUTO/KD_GRAPHICS mode

PATCH In review Andrew Johnson 4 participants 14 messages Mar 8, 2007 – Mar 10, 2007 Char and misc drivers

lore

Share & tools

lore

Andrew Johnson

author

2007-03-08 23:09 UTC #0 lore raw

PATCH 1/1 Software Suspend: Fix suspend when console is in VT_AUTO/KD_GRAPHICS mode

Hi Pavel,

When the console is in VT_AUTO/KD_GRAPHICS mode, switching to the SUSPEND_CONSOLE fails, resulting in vt_waitactive() waiting indefinately or until the task is interrupted. The following patch tests if a console switch can occur in set_console() and returns early if a console switch is not possible.

Signed-off-by Andrew Johnson <[email protected]>

drivers/char/vt.c +6 -1

diff -rup linux-2.6.20.1/drivers/char/vt.c linux/drivers/char/vt.c
@@ -2188,10 +2188,20 @@ static void console_callback(struct work
 	release_console_sem();
 }
 
-void set_console(int nr)
+extern char vt_dont_switch;
+
+int set_console(int nr)
 {
+	struct vc_data *vc = vc_cons[fg_console].d;
+
+	if(!vc_cons_allocated(nr) || vt_dont_switch || vc->vc_mode ==

KD_GRAPHICS) {
+ return -EINVAL;
+ }
+
want_console = nr;
schedule_console_callback();
+
+ return 0;
}

struct tty_driver *console_driver;
diff -rup linux-2.6.20.1/drivers/char/vt_ioctl.c
linux/drivers/char/vt_ioctl.c
--- linux-2.6.20.1/drivers/char/vt_ioctl.c 2007-02-19 22:34:32.000000000
-0800
+++ linux/drivers/char/vt_ioctl.c 2007-03-08 14:15:41.000000000 -0800
@@ -34,7 +34,7 @@
#include <linux/kbd_diacr.h>
#include <linux/selection.h>

-static char vt_dont_switch;
+char vt_dont_switch;
extern struct tty_driver *console_driver;

#define VT_IS_IN_USE(i) (console_driver->ttys[i] &&
console_driver->ttys[i]->count)
diff -rup linux-2.6.20.1/include/linux/kbd_kern.h linux/include/linux/kbd_kern.h
--- linux-2.6.20.1/include/linux/kbd_kern.h 2007-02-19
22:34:32.000000000 -0800
+++ linux/include/linux/kbd_kern.h 2007-03-08 14:15:41.000000000 -0800
@@ -75,7 +75,7 @@ extern int do_poke_blanked_console;

extern void (*kbd_ledfunc)(unsigned int led);

-extern void set_console(int nr);
+extern int set_console(int nr);
extern void schedule_console_callback(void);

static inline void set_leds(void)
diff -rup linux-2.6.20.1/kernel/power/console.c
linux/kernel/power/console.c
--- linux-2.6.20.1/kernel/power/console.c 2007-02-19 22:34:32.000000000
-0800
+++ linux/kernel/power/console.c 2007-03-08 14:15:41.000000000 -0800
@@ -27,7 +27,11 @@ int pm_prepare_console(void)
return 1;
}

- set_console(SUSPEND_CONSOLE);
+ if (set_console(SUSPEND_CONSOLE)) {
+ /* Unable to change to the new console */
+ release_console_sem();
+ return 1;
+ }
release_console_sem();

if (vt_waitactive(SUSPEND_CONSOLE)) {

Daniel Drake

2007-03-08 23:36 UTC #1 lore raw

Quoted 22 lines · Andrew Johnson

When the console is in VT_AUTO/KD_GRAPHICS mode, switching to the SUSPEND_CONSOLE fails, resulting in vt_waitactive() waiting indefinately or until the task is interrupted. The following patch tests if a console switch can occur in set_console() and returns early if a console switch is not possible.
Signed-off-by Andrew Johnson <[email protected]>
diff -rup linux-2.6.20.1/drivers/char/vt.c linux/drivers/char/vt.c
drivers/char/vt.c +5 -1
@@ -2188,10 +2188,20 @@ static void console_callback(struct work
 	release_console_sem();
 }
 
-void set_console(int nr)
+extern char vt_dont_switch;
+
+int set_console(int nr)
 {
+	struct vc_data *vc = vc_cons[fg_console].d;
+

drivers/char/vt.c static void console_callback(struct work · line 2206

Quoted 10 lines

drivers/char/vt.c +6 -0

+	if(!vc_cons_allocated(nr) || vt_dont_switch || vc->vc_mode ==
 KD_GRAPHICS) {
+		return -EINVAL;
+	}
+
 	want_console = nr;
 	schedule_console_callback();
+
+	return 0;
 }

I haven't tested, but I think the above -EINVAL return will break chvt. chvt uses the VT_ACTIVATE ioctl which calls set_console(), and it is valid for chvt to be used to change away from a graphics-mode console -- that shouldn't be an error condition.

Daniel

Andrew Johnson

author

2007-03-09 05:01 UTC #2 lore raw

Quoted 3 lines · Daniel Drake

Quoted 2 lines · Andrew Johnson
When the console is in VT_AUTO/KD_GRAPHICS mode, switching to the SUSPEND_CONSOLE fails, resulting in vt_waitactive() waiting

indefinately

Quoted 2 lines

Quoted 2 lines
or until the task is interrupted. The following patch tests if a console switch can occur in set_console() and returns early if a

console

Quoted 6 lines

Quoted 6 lines
switch is not possible.
Signed-off-by Andrew Johnson <[email protected]>
diff -rup linux-2.6.20.1/drivers/char/vt.c linux/drivers/char/vt.c
--- linux-2.6.20.1/drivers/char/vt.c 2007-02-19

22:34:32.000000000 -

Quoted 17 lines

0800

Quoted 16 lines

+++ linux/drivers/char/vt.c 2007-03-08 14:15:41.000000000 -0800

drivers/char/vt.c +8 -1

@@ -2188,10 +2188,20 @@ static void console_callback(struct work
 	release_console_sem();
 }
 
-void set_console(int nr)
+extern char vt_dont_switch;
+
+int set_console(int nr)
 {
+	struct vc_data *vc = vc_cons[fg_console].d;
+
+	if(!vc_cons_allocated(nr) || vt_dont_switch || vc->vc_mode ==
 KD_GRAPHICS) {
+		return -EINVAL;
+	}

Quoted 8 lines

Review static void console_callback(struct work · line 2206
Quoted 6 lines
(quoted diff) +3 -0
+
 	want_console = nr;
 	schedule_console_callback();
+
+	return 0;
 }
I haven't tested, but I think the above -EINVAL return will break

chvt.

Quoted 2 lines

chvt uses the VT_ACTIVATE ioctl which calls set_console(), and it is valid for chvt to be used to change away from a graphics-mode console

Quoted 1 line

that shouldn't be an error condition.

Currently the VT_ACTIVATE ioctl will eventually result in a call to change_console() which will ignore the change if the console is in KD_GRAPHICS+VT_AUTO mode. Thus, chvt should fail. However, the return code from set_console() is ignored in vt_ioctl so no error code will result (which is the same as current behavior).

-- Andrew

Pavel Machek

2007-03-09 09:08 UTC #3 lore raw

In reply to Andrew Johnson

Hi!

Quoted 5 lines

When the console is in VT_AUTO/KD_GRAPHICS mode, switching to the SUSPEND_CONSOLE fails, resulting in vt_waitactive() waiting indefinately or until the task is interrupted. The following patch tests if a console switch can occur in set_console() and returns early if a console switch is not possible.

Your mailer wraps lines...

How do I reproduce this? Suspending from X works for me...?

Quoted 11 lines

Signed-off-by Andrew Johnson <[email protected]>
diff -rup linux-2.6.20.1/drivers/char/vt.c linux/drivers/char/vt.c
drivers/char/vt.c +1 -1
@@ -2188,10 +2188,20 @@ static void console_callback(struct work
 	release_console_sem();
 }
 
-void set_console(int nr)
+extern char vt_dont_switch;

drivers/char/vt.c static void console_callback(struct work · line 2201

Quoted 10 lines

drivers/char/vt.c +8 -0

+
+int set_console(int nr)
 {
+	struct vc_data *vc = vc_cons[fg_console].d;
+
+	if(!vc_cons_allocated(nr) || vt_dont_switch || vc->vc_mode ==
 KD_GRAPHICS) {
+		return -EINVAL;
+	}
+

So... if current console is graphical, we leave X accessing the console... That's bad, because video state is not going to be restored...?
Pavel

Signature · 2 lines

(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Matthew Garrett

2007-03-09 15:35 UTC #4 lore raw

Quoted 3 lines · Pavel Machek

So... if current console is graphical, we leave X accessing the console... That's bad, because video state is not going to be restored...?

A graphical console is not necessarily X. Is there any requirement for there to be a single VT that isn't in text mode? The vt switching is a hack, we shouldn't make life difficult for people who have their own userspace code that's entirely capable of restoring video state on its own.

Signature · 1 line

Matthew Garrett | [email protected]

Andrew Johnson

author

2007-03-09 16:10 UTC #5 lore raw

Quoted 11 lines · Matthew Garrett

Quoted 3 lines · Pavel Machek
So... if current console is graphical, we leave X accessing the console... That's bad, because video state is not going to be restored...?
A graphical console is not necessarily X. Is there any requirement for there to be a single VT that isn't in text mode? The vt switching is a hack, we shouldn't make life difficult for people who have their own userspace code that's entirely capable of restoring video state on its own.

The problem actually comes about when using Qtopia Phone Edition (QPE) on a PXA270. QPE puts the console into VT_AUTO+KD_GRAPHICS mode and writes directly to the framebuffer from then on. In this mode the kernel correctly disallows a console change, as QPE is not getting notification of a console change and thus does not know when to repaint the screen.

AFAIK, X uses VT_PROCESS+KD_GRAPHICS mode, so it gets notification of a change to and from the X console, thus it knows when to repaint the screen.

I think you can test this by changing the mode of a text console to KD_GRAPHICS using the KDSETMODE ioctl, then attempting to change to another text console using chvt.

-- Andrew

Pavel Machek

2007-03-09 21:14 UTC #6 lore raw

Hi!

Quoted 10 lines

Quoted 9 lines
Quoted 3 lines
So... if current console is graphical, we leave X accessing the console... That's bad, because video state is not going to be restored...?
A graphical console is not necessarily X. Is there any requirement for there to be a single VT that isn't in text mode? The vt switching is a hack, we shouldn't make life difficult for people who have their own userspace code that's entirely capable of restoring video state on its own.

Quoted 6 lines

The problem actually comes about when using Qtopia Phone Edition (QPE) on a PXA270. QPE puts the console into VT_AUTO+KD_GRAPHICS mode and writes directly to the framebuffer from then on. In this mode the kernel correctly disallows a console change, as QPE is not getting notification of a console change and thus does not know when to repaint the screen.

Fix qpe to use vt_process, instead?

...how does qpe know when to repaint the screen, anyway?

Pavel

Signature · 2 lines

(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Andrew Johnson

author

2007-03-09 21:59 UTC #7 lore raw

Quoted 12 lines · Pavel Machek

Hi!
Quoted 10 lines
Quoted 9 lines
Quoted 3 lines
So... if current console is graphical, we leave X accessing the console... That's bad, because video state is not going to be restored...?
A graphical console is not necessarily X. Is there any requirement for there to be a single VT that isn't in text mode? The vt switching is a hack, we shouldn't make life difficult for people who have their own userspace code that's entirely capable of restoring video state on its own.

Quoted 8 lines

Quoted 6 lines
The problem actually comes about when using Qtopia Phone Edition (QPE) on a PXA270. QPE puts the console into VT_AUTO+KD_GRAPHICS mode and writes directly to the framebuffer from then on. In this mode the kernel correctly disallows a console change, as QPE is not getting notification of a console change and thus does not know when to repaint the screen.
Fix qpe to use vt_process, instead?

The real issue seems to be the assumption in pm_prepare_console() that set_console() will succeed. set_console() may fail (resulting in vt_waitactive waiting for a console switch that can never happen or a break) for two currently unchecked reasons:

- console is in VT_AUTO+KD_GRAPHICS mode
- vt_dont_switch is set

Quoted 1 line

...how does qpe know when to repaint the screen, anyway?

QPE doesn't need to repaint the screen after wake-up - the framebuffer memory is retained so the PXA270 lcd controller simply displays what was last on the screen when it is re-enabled.

-- Andrew

Pavel Machek

2007-03-10 22:26 UTC #8 lore raw

Hi!

Quoted 5 lines

Quoted 1 line
...how does qpe know when to repaint the screen, anyway?
QPE doesn't need to repaint the screen after wake-up - the framebuffer memory is retained so the PXA270 lcd controller simply displays what was last on the screen when it is re-enabled.

That probably means QPE is broken on machines that do not preserve framebuffer over suspend :-(.
Pavel

Signature · 2 lines

(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Andrew Johnson

author

2007-03-09 19:03 UTC #9 lore raw

In reply to Matthew Garrett

Quoted 11 lines · Matthew Garrett

Quoted 3 lines · Pavel Machek
So... if current console is graphical, we leave X accessing the console... That's bad, because video state is not going to be restored...?
A graphical console is not necessarily X. Is there any requirement for there to be a single VT that isn't in text mode? The vt switching is a hack, we shouldn't make life difficult for people who have their own userspace code that's entirely capable of restoring video state on its own.

I realised that the previous patch would disallow a console switch while running X. Attached is an updated patch with this scenario fixed.

Another approach might be to fail in vt_waitactive() if a console switch is not going to occur.

-- Andrew

Signed-off-by Andrew Johnson <[email protected]>

---

drivers/char/vt.c +13 -1

diff -rup linux-2.6.20.1/drivers/char/vt.c linux/drivers/char/vt.c
@@ -2188,10 +2188,22 @@ static void console_callback(struct work
 	release_console_sem();
 }
 
-void set_console(int nr)
+extern char vt_dont_switch;
+
+int set_console(int nr)
 {
+	struct vc_data *vc = vc_cons[fg_console].d;
+
+	if(!vc_cons_allocated(nr) || vt_dont_switch || 
+		(vc->vt_mode.mode != VT_PROCESS && vc->vc_mode == KD_GRAPHICS)) {
+
+		return -EINVAL;
+	}
+
 	want_console = nr;
 	schedule_console_callback();
+
+	return 0;
 }
 
 struct tty_driver *console_driver;

diff -rup linux-2.6.20.1/drivers/char/vt_ioctl.c
linux/drivers/char/vt_ioctl.c
--- linux-2.6.20.1/drivers/char/vt_ioctl.c 2007-02-19 22:34:32.000000000
-0800
+++ linux/drivers/char/vt_ioctl.c 2007-03-08 14:15:41.000000000 -0800
@@ -34,7 +34,7 @@
#include <linux/kbd_diacr.h>
#include <linux/selection.h>

-static char vt_dont_switch;
+char vt_dont_switch;
extern struct tty_driver *console_driver;

extern void (*kbd_ledfunc)(unsigned int led);

-extern void set_console(int nr);
+extern int set_console(int nr);
extern void schedule_console_callback(void);

- set_console(SUSPEND_CONSOLE);
+ if (set_console(SUSPEND_CONSOLE)) {
+ /* Unable to change to the new console */
+ release_console_sem();
+ return 1;
+ }
release_console_sem();

if (vt_waitactive(SUSPEND_CONSOLE)) {

Pavel Machek

2007-03-09 21:19 UTC #10 lore raw

Hi!

Quoted 20 lines

Quoted 9 lines
Quoted 3 lines
So... if current console is graphical, we leave X accessing the console... That's bad, because video state is not going to be restored...?
A graphical console is not necessarily X. Is there any requirement for there to be a single VT that isn't in text mode? The vt switching is a hack, we shouldn't make life difficult for people who have their own userspace code that's entirely capable of restoring video state on its own.
I realised that the previous patch would disallow a console switch while running X. Attached is an updated patch with this scenario fixed.
Another approach might be to fail in vt_waitactive() if a console switch is not going to occur.
-- Andrew

drivers/char/vt.c static void console_callback(struct work · line 2192

Quoted 3 lines

drivers/char/vt.c +2 -1

-void set_console(int nr)
+extern char vt_dont_switch;
+

What does this variable do and why do we want to use it here?

drivers/char/vt.c static void console_callback(struct work · line 2197

Quoted 5 lines

drivers/char/vt.c +4 -0

+int set_console(int nr)
 {
+	struct vc_data *vc = vc_cons[fg_console].d;
+
+	if(!vc_cons_allocated(nr) || vt_dont_switch ||

'if ('

drivers/char/vt.c static void console_callback(struct work · line 2202

Quoted 5 lines

drivers/char/vt.c +5 -0

+		(vc->vt_mode.mode != VT_PROCESS && vc->vc_mode == KD_GRAPHICS)) {
+
+		return -EINVAL;
+	}
+

I assume you want ...mode == VT_AUTO here?

And big comment explaining why we want this behaviour?

And another big comment explaining why this will not break existing set_console() users?

Quoted 42 lines

drivers/char/vt.c +2 -0
 	want_console = nr;
 	schedule_console_callback();
+
+	return 0;
 }
 
 struct tty_driver *console_driver;
diff -rup linux-2.6.20.1/drivers/char/vt_ioctl.c
linux/drivers/char/vt_ioctl.c
--- linux-2.6.20.1/drivers/char/vt_ioctl.c 2007-02-19 22:34:32.000000000
-0800
+++ linux/drivers/char/vt_ioctl.c 2007-03-08 14:15:41.000000000 -0800
(quoted diff) +1 -1
@@ -34,7 +34,7 @@
 #include <linux/kbd_diacr.h>
 #include <linux/selection.h>
 
-static char vt_dont_switch;
+char vt_dont_switch;
 extern struct tty_driver *console_driver;
 
 #define VT_IS_IN_USE(i)	(console_driver->ttys[i] &&
console_driver->ttys[i]->count)
diff -rup linux-2.6.20.1/include/linux/kbd_kern.h
linux/include/linux/kbd_kern.h
--- linux-2.6.20.1/include/linux/kbd_kern.h 2007-02-19
22:34:32.000000000 -0800
+++ linux/include/linux/kbd_kern.h 2007-03-08 14:15:41.000000000 -0800
(quoted diff) +1 -1
@@ -75,7 +75,7 @@ extern int do_poke_blanked_console;
 
 extern void (*kbd_ledfunc)(unsigned int led);
 
-extern void set_console(int nr);
+extern int set_console(int nr);
 extern void schedule_console_callback(void);
 
 static inline void set_leds(void)
diff -rup linux-2.6.20.1/kernel/power/console.c
linux/kernel/power/console.c
--- linux-2.6.20.1/kernel/power/console.c 2007-02-19 22:34:32.000000000
-0800
+++ linux/kernel/power/console.c 2007-03-08 14:15:41.000000000 -0800

Review int pm_prepare_console(void) · line 31

Quoted 6 lines

(quoted diff) +2 -1

@@ -27,7 +27,11 @@ int pm_prepare_console(void)
 		return 1;
 	}
 
-	set_console(SUSPEND_CONSOLE);
+	if (set_console(SUSPEND_CONSOLE)) {
+		/* Unable to change to the new console */

That's not what the comment should say.

It should explain why it is okay to proceed when we can't change to text console.

Signature · 2 lines

(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Andrew Johnson

author

2007-03-10 00:06 UTC #11 lore raw

Quoted 9 lines · Pavel Machek

Hi!
Review
Quoted 4 lines
(quoted diff) +2 -1
-void set_console(int nr)
+extern char vt_dont_switch;
+
What does this variable do and why do we want to use it here?

It's needed in set_console(). Console switch will fail if this is true.

Quoted 11 lines

'if ('
Quoted 1 line
+ (vc->vt_mode.mode != VT_PROCESS && vc->vc_mode ==
KD_GRAPHICS)) {
Review
Quoted 4 lines
(quoted diff) +4 -0
+
+             return -EINVAL;
+     }
+
I assume you want ...mode == VT_AUTO here?

Quoted 5 lines

And big comment explaining why we want this behaviour?
And another big comment explaining why this will not break existing set_console() users?

See updated comment in attached patch.

Quoted 9 lines

Review
Quoted 3 lines
(quoted diff) +2 -1
-     set_console(SUSPEND_CONSOLE);
+     if (set_console(SUSPEND_CONSOLE)) {
+             /* Unable to change to the new console */
That's not what the comment should say.
It should explain why it is okay to proceed when we can't change to text console.

See updated comment in attached patch. It's really up to the caller to decide what to do if we can't switch the console - currently all callers ignore the return code so I assume that it's okay to proceed anyway.

-- Andrew

Signed-off-by Andrew Johnson <[email protected]>

---

drivers/char/vt.c +21 -1

diff -rup linux-2.6.20.1/drivers/char/vt.c linux/drivers/char/vt.c
@@ -2188,10 +2188,30 @@ static void console_callback(struct work
 	release_console_sem();
 }
 
-void set_console(int nr)
+extern char vt_dont_switch;
+
+int set_console(int nr)
 {
+	struct vc_data *vc = vc_cons[fg_console].d;
+
+	if(!vc_cons_allocated(nr) || vt_dont_switch || 
+		(vc->vt_mode.mode == VT_AUTO && vc->vc_mode == KD_GRAPHICS)) {
+
+		/* 
+		 * Console switch will fail in console_callback() or 
+		 * change_console() so there is no point scheduling 
+		 * the callback
+		 *
+		 * Existing set_console() users don't check the return
+		 * value so this shouldn't break anything 
+		 */ 
+		return -EINVAL;
+	}
+
 	want_console = nr;
 	schedule_console_callback();
+
+	return 0;
 }
 
 struct tty_driver *console_driver;

-static char vt_dont_switch;
+char vt_dont_switch;
extern struct tty_driver *console_driver;

extern void (*kbd_ledfunc)(unsigned int led);

-extern void set_console(int nr);
+extern int set_console(int nr);
extern void schedule_console_callback(void);

static inline void set_leds(void)
diff -rup linux-2.6.20.1/kernel/power/console.c
linux/kernel/power/console.c
--- linux-2.6.20.1/kernel/power/console.c 2007-02-19 22:34:32.000000000
-0800
+++ linux/kernel/power/console.c 2007-03-09 15:52:32.000000000 -0800
@@ -27,7 +27,15 @@ int pm_prepare_console(void)
return 1;
}

- set_console(SUSPEND_CONSOLE);
+ if (set_console(SUSPEND_CONSOLE)) {
+ /*
+ * We're unable to switch to the SUSPEND_CONSOLE.
+ * Let the calling function know so it can decide
+ * what to do.
+ */
+ release_console_sem();
+ return 1;
+ }
release_console_sem();

if (vt_waitactive(SUSPEND_CONSOLE)) {

Pavel Machek

2007-03-10 22:28 UTC #12 lore raw

Hi1

Quoted 6 lines

Quoted 3 lines
It should explain why it is okay to proceed when we can't change to text console.
See updated comment in attached patch. It's really up to the caller to decide what to do if we can't switch the console - currently all callers ignore the return code so I assume that it's okay to proceed anyway.

Ok, I guess the patch is right thing to do after all. Fix issues below, append a changelog, and send a patch to lmkl, cc Andrew Morton and me. Oh and you have my ACK.
Pavel

Quoted 7 lines

Signed-off-by Andrew Johnson <[email protected]>
---
diff -rup linux-2.6.20.1/drivers/char/vt.c linux/drivers/char/vt.c
drivers/char/vt.c +0 -0
@@ -2188,10 +2188,30 @@ static void console_callback(struct work
 	release_console_sem();

drivers/char/vt.c static void console_callback(struct work · line 2197

Quoted 10 lines

drivers/char/vt.c +6 -1

 }
 
-void set_console(int nr)
+extern char vt_dont_switch;
+
+int set_console(int nr)
 {
+	struct vc_data *vc = vc_cons[fg_console].d;
+
+	if(!vc_cons_allocated(nr) || vt_dont_switch ||

there should be space between "if" and "(".

Quoted 5 lines

diff -rup linux-2.6.20.1/drivers/char/vt_ioctl.c
linux/drivers/char/vt_ioctl.c
--- linux-2.6.20.1/drivers/char/vt_ioctl.c 2007-02-19 22:34:32.000000000
-0800
+++ linux/drivers/char/vt_ioctl.c 2007-03-08 14:15:41.000000000

And your mailer still wordwraps.

Signature · 2 lines

(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Pavel Machek

2007-03-09 21:20 UTC #13 lore raw

In reply to Matthew Garrett

Hi!

Quoted 9 lines

Quoted 3 lines
So... if current console is graphical, we leave X accessing the console... That's bad, because video state is not going to be restored...?
A graphical console is not necessarily X. Is there any requirement for there to be a single VT that isn't in text mode? The vt switching is a hack, we shouldn't make life difficult for people who have their own userspace code that's entirely capable of restoring video state on its own.

How does their wonderful userspace code know that it needs to restore video?

vt switch *does* mean 'kernel now takes over console'.

Pavel

Signature · 2 lines

(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html