Skip to contentSkip to conversation

prevent swsusp with PAE

PATCH Pavel Machek 5 participants 11 messages Aug 31, 2006 – Sep 2, 2006

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

If HIGHMEM64G and swsusp are used at the same time, nasty random crashes happen during resume. Cause is known; prevent that combination.

I guess I'd like to see this one in 2.6.18...

---
commit acb3b411ec93f827b25b8481d53670c5c9195d89 tree f52cd5518e34af16fe5ae28064717bcc95929f28 parent cd03e183c58e6e7073e054a7fe335cf50c61fe2f author <[email protected]> Thu, 31 Aug 2006 15:52:34 +0200 committer <[email protected]> Thu, 31 Aug 2006 15:52:34 +0200

 include/asm-i386/suspend.h |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)
include/asm-i386/suspend.h +9 -0
index 08be1e5..01cd812 100644
@@ -16,6 +16,15 @@ arch_prepare_suspend(void)
 		printk(KERN_ERR "PSE is required for swsusp.\n");
 		return -EPERM;
 	}
+
+#ifdef CONFIG_X86_PAE
+	printk(KERN_ERR "swsusp is incompatible with PAE.\n");
+	/* This is actually instance of the same problem. We need
+	   identity mapping self-contained in swsusp_pg_dir, and PAE
+	   prevents that. Solution could be copied from x86_64. */
+	return -EPERM;
+#endif
+
 	return 0;
 }
 
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#1 lore raw
Quoted 27 lines · On Thu, 31 Aug 2006 15:53:36 +0200 Pavel Machek <[email protected]> wrote:

If HIGHMEM64G and swsusp are used at the same time, nasty random crashes happen during resume. Cause is known; prevent that combination.

I guess I'd like to see this one in 2.6.18...

---
commit acb3b411ec93f827b25b8481d53670c5c9195d89 tree f52cd5518e34af16fe5ae28064717bcc95929f28 parent cd03e183c58e6e7073e054a7fe335cf50c61fe2f author <[email protected]> Thu, 31 Aug 2006 15:52:34 +0200 committer <[email protected]> Thu, 31 Aug 2006 15:52:34 +0200

 include/asm-i386/suspend.h |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)
include/asm-i386/suspend.h +1 -0
index 08be1e5..01cd812 100644
@@ -16,6 +16,15 @@ arch_prepare_suspend(void)
 		printk(KERN_ERR "PSE is required for swsusp.\n");
 		return -EPERM;
 	}
+

include/asm-i386/suspend.h arch_prepare_suspend(void) · line 29

Quoted 10 lines
include/asm-i386/suspend.h +8 -0
+#ifdef CONFIG_X86_PAE
+	printk(KERN_ERR "swsusp is incompatible with PAE.\n");
+	/* This is actually instance of the same problem. We need
+	   identity mapping self-contained in swsusp_pg_dir, and PAE
+	   prevents that. Solution could be copied from x86_64. */
+	return -EPERM;
+#endif
+
 	return 0;
 }

Why not do this in Kconfig??

#2 lore raw

Hi!

Quoted 33 lines
Quoted 31 lines · Pavel Machek

If HIGHMEM64G and swsusp are used at the same time, nasty random crashes happen during resume. Cause is known; prevent that combination.

I guess I'd like to see this one in 2.6.18...

---
commit acb3b411ec93f827b25b8481d53670c5c9195d89 tree f52cd5518e34af16fe5ae28064717bcc95929f28 parent cd03e183c58e6e7073e054a7fe335cf50c61fe2f author <[email protected]> Thu, 31 Aug 2006 15:52:34 +0200 committer <[email protected]> Thu, 31 Aug 2006 15:52:34 +0200

 include/asm-i386/suspend.h |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)
include/asm-i386/suspend.h +5 -0
index 08be1e5..01cd812 100644
@@ -16,6 +16,15 @@ arch_prepare_suspend(void)
 		printk(KERN_ERR "PSE is required for swsusp.\n");
 		return -EPERM;
 	}
+
+#ifdef CONFIG_X86_PAE
+	printk(KERN_ERR "swsusp is incompatible with PAE.\n");
+	/* This is actually instance of the same problem. We need
+	   identity mapping self-contained in swsusp_pg_dir, and PAE
Quoted 8 lines

include/asm-i386/suspend.h arch_prepare_suspend(void) · line 29

Quoted 6 lines
include/asm-i386/suspend.h +4 -0
+	   prevents that. Solution could be copied from x86_64. */
+	return -EPERM;
+#endif
+
 	return 0;
 }

Why not do this in Kconfig??

Well, Kconfig does not provide natural place for comments, and disappearing config option is sure to confuse people. But of course I can do it.
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#3 lore raw
Quoted 44 lines · On Fri, 1 Sep 2006 00:35:21 +0200 Pavel Machek wrote:

Hi!

Quoted 33 lines
Quoted 31 lines · Pavel Machek

If HIGHMEM64G and swsusp are used at the same time, nasty random crashes happen during resume. Cause is known; prevent that combination.

I guess I'd like to see this one in 2.6.18...

---
commit acb3b411ec93f827b25b8481d53670c5c9195d89 tree f52cd5518e34af16fe5ae28064717bcc95929f28 parent cd03e183c58e6e7073e054a7fe335cf50c61fe2f author <[email protected]> Thu, 31 Aug 2006 15:52:34 +0200 committer <[email protected]> Thu, 31 Aug 2006 15:52:34 +0200

 include/asm-i386/suspend.h |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)
include/asm-i386/suspend.h +5 -0
index 08be1e5..01cd812 100644
@@ -16,6 +16,15 @@ arch_prepare_suspend(void)
 		printk(KERN_ERR "PSE is required for swsusp.\n");
 		return -EPERM;
 	}
+
+#ifdef CONFIG_X86_PAE
+	printk(KERN_ERR "swsusp is incompatible with PAE.\n");
+	/* This is actually instance of the same problem. We need
+	   identity mapping self-contained in swsusp_pg_dir, and PAE
Quoted 8 lines

include/asm-i386/suspend.h arch_prepare_suspend(void) · line 29

Quoted 6 lines
include/asm-i386/suspend.h +4 -0
+	   prevents that. Solution could be copied from x86_64. */
+	return -EPERM;
+#endif
+
 	return 0;
 }

Why not do this in Kconfig??

Quoted 3 lines

Well, Kconfig does not provide natural place for comments, and disappearing config option is sure to confuse people. But of course I can do it.

Documentation/kbuild/kconfig-language.txt includes a "comment" keyword.

USB-storage uses it to suggest that SCSI + SCSI disk support need to be enabled for usb-storage.

---
~Randy

#4 lore raw
In reply to Pavel Machek
Quoted 22 lines · On Fri, 1 Sep 2006 00:35:21 +0200 Pavel Machek <[email protected]> wrote:
Quoted 13 lines
Quoted 13 lines
include/asm-i386/suspend.h +5 -0
index 08be1e5..01cd812 100644
@@ -16,6 +16,15 @@ arch_prepare_suspend(void)
 		printk(KERN_ERR "PSE is required for swsusp.\n");
 		return -EPERM;
 	}
+
+#ifdef CONFIG_X86_PAE
+	printk(KERN_ERR "swsusp is incompatible with PAE.\n");
+	/* This is actually instance of the same problem. We need
+	   identity mapping self-contained in swsusp_pg_dir, and PAE
Quoted 8 lines

include/asm-i386/suspend.h arch_prepare_suspend(void) · line 29

Quoted 6 lines
include/asm-i386/suspend.h +4 -0
+	   prevents that. Solution could be copied from x86_64. */
+	return -EPERM;
+#endif
+
 	return 0;
 }

Why not do this in Kconfig??

Quoted 3 lines

Well, Kconfig does not provide natural place for comments, and disappearing config option is sure to confuse people. But of course I can do it.

It would be more conventional.

I think what this really points at is a weakness in the menuconfig/xconfig/etc user interfaces. It should be possible to navigate to the presently-disabled config option and ask it "why can't I turn you on?".

#5 lore raw
Quoted 25 lines · Andrew Morton
Quoted 22 lines · On Fri, 1 Sep 2006 00:35:21 +0200 Pavel Machek <[email protected]> wrote:
Quoted 21 lines
Quoted 9 lines
include/asm-i386/suspend.h +1 -0
index 08be1e5..01cd812 100644
@@ -16,6 +16,15 @@ arch_prepare_suspend(void)
 		printk(KERN_ERR "PSE is required for swsusp.\n");
 		return -EPERM;
 	}
+

include/asm-i386/suspend.h arch_prepare_suspend(void) · line 29

Quoted 10 lines
include/asm-i386/suspend.h +8 -0
+#ifdef CONFIG_X86_PAE
+	printk(KERN_ERR "swsusp is incompatible with PAE.\n");
+	/* This is actually instance of the same problem. We need
+	   identity mapping self-contained in swsusp_pg_dir, and PAE
+	   prevents that. Solution could be copied from x86_64. */
+	return -EPERM;
+#endif
+
 	return 0;
 }

Why not do this in Kconfig??

Quoted 5 lines
Quoted 3 lines

Well, Kconfig does not provide natural place for comments, and disappearing config option is sure to confuse people. But of course I can do it.

It would be more conventional.

Well, I have very similar check few lines above, and this is both i386 specific, so I slightly prefer to do it in the code, but...

Quoted 3 lines

I think what this really points at is a weakness in the menuconfig/xconfig/etc user interfaces. It should be possible to navigate to the presently-disabled config option and ask it "why can't I turn you on?".

Yes, but I'll still have users asking me "why I can't turn it on" ;-).
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#6 lore raw
Quoted 26 lines · On Fri, 1 Sep 2006 00:52:34 +0200 Pavel Machek wrote:
Quoted 25 lines · Andrew Morton
Quoted 22 lines · On Fri, 1 Sep 2006 00:35:21 +0200 Pavel Machek <[email protected]> wrote:
Quoted 21 lines
Quoted 9 lines
include/asm-i386/suspend.h +1 -0
index 08be1e5..01cd812 100644
@@ -16,6 +16,15 @@ arch_prepare_suspend(void)
 		printk(KERN_ERR "PSE is required for swsusp.\n");
 		return -EPERM;
 	}
+

include/asm-i386/suspend.h arch_prepare_suspend(void) · line 29

Quoted 10 lines
include/asm-i386/suspend.h +8 -0
+#ifdef CONFIG_X86_PAE
+	printk(KERN_ERR "swsusp is incompatible with PAE.\n");
+	/* This is actually instance of the same problem. We need
+	   identity mapping self-contained in swsusp_pg_dir, and PAE
+	   prevents that. Solution could be copied from x86_64. */
+	return -EPERM;
+#endif
+
 	return 0;
 }

Why not do this in Kconfig??

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

Well, Kconfig does not provide natural place for comments, and disappearing config option is sure to confuse people. But of course I can do it.

It would be more conventional.

Well, I have very similar check few lines above, and this is both i386 specific, so I slightly prefer to do it in the code, but...

If we can prevent a non-working build in Kconfig, that's what we should do.

Quoted 5 lines
Quoted 3 lines

I think what this really points at is a weakness in the menuconfig/xconfig/etc user interfaces. It should be possible to navigate to the presently-disabled config option and ask it "why can't I turn you on?".

Yes, but I'll still have users asking me "why I can't turn it on" ;-).

menuconfig and xconfig both have Help and Search that can aid with that, but I would still use the "comment" keyword also.

---
~Randy

#7 lore raw
Quoted 8 lines · On Thu, 31 Aug 2006 16:05:46 -0700 "Randy.Dunlap" <[email protected]> wrote:
Quoted 5 lines
Quoted 3 lines

I think what this really points at is a weakness in the menuconfig/xconfig/etc user interfaces. It should be possible to navigate to the presently-disabled config option and ask it "why can't I turn you on?".

Yes, but I'll still have users asking me "why I can't turn it on" ;-).

menuconfig and xconfig both have Help and Search that can aid with that, but I would still use the "comment" keyword also.

Help and Search options aren't much use if you cannot see the option and if you do not know that it exists.

A better UI design would be to show the unselectable option in some dimmed-out fashion and then provide an interface which permits the user to

a) query the item to find out why it's not selectable and

b) turn on the depends-on option(s) at the present site, thus making the present item selectable (ie: not dimmed any more).

I dont't think any of that would require any Kconfig changes - it's purely a UI thing?

#8 lore raw
Quoted 18 lines · On Thu, 31 Aug 2006 17:16:15 -0700 Andrew Morton wrote:
Quoted 8 lines · On Thu, 31 Aug 2006 16:05:46 -0700 "Randy.Dunlap" <[email protected]> wrote:
Quoted 5 lines
Quoted 3 lines

I think what this really points at is a weakness in the menuconfig/xconfig/etc user interfaces. It should be possible to navigate to the presently-disabled config option and ask it "why can't I turn you on?".

Yes, but I'll still have users asking me "why I can't turn it on" ;-).

menuconfig and xconfig both have Help and Search that can aid with that, but I would still use the "comment" keyword also.

Help and Search options aren't much use if you cannot see the option and if you do not know that it exists.

A better UI design would be to show the unselectable option in some dimmed-out fashion and then provide an interface which permits the user to

Quoted 7 lines

a) query the item to find out why it's not selectable and

b) turn on the depends-on option(s) at the present site, thus making the present item selectable (ie: not dimmed any more).

I dont't think any of that would require any Kconfig changes - it's purely a UI thing?

Probably. xconfig already has this ability/feature. I enable/use it all the time. menuconfig could probably be taught how to do it also. (and gconfig seems to be orphaned)

---
~Randy

#10 lore raw
In reply to Pavel Machek
Quoted 6 lines
Quoted 4 lines

I think what this really points at is a weakness in the menuconfig/xconfig/etc user interfaces. It should be possible to navigate to the presently-disabled config option and ask it "why can't I turn you on?".

Yes, but I'll still have users asking me "why I can't turn it on" ;-).

That would probably be handy, just as http://lkml.org/lkml/2006/8/25/25

Jan Engelhardt

#9 lore raw
In reply to Randy.Dunlap

Next version of prevent-swsusp-with-PAE, this time I disable it in Kconfig.

PAE + swsusp results in hard-to-debug crash about 50% of time during resume. Cause is known, fix needs to be ported from x86-64 (but we can't make it to 2.6.18, and I'd like this to be worked around in 2.6.18).

kernel/power/Kconfig +5 -1
index ae44a70..619ecab 100644
@@ -56,7 +56,7 @@ config PM_TRACE
 
 config SOFTWARE_SUSPEND
 	bool "Software Suspend"
-	depends on PM && SWAP && (X86 && (!SMP || SUSPEND_SMP)) || ((FRV || PPC32) && !SMP)
+	depends on PM && SWAP && ((X86 && (!SMP || SUSPEND_SMP) && !X86_PAE) || ((FRV || PPC32) && !SMP))
 	---help---
 	  Enable the possibility of suspending the machine.
 	  It doesn't need ACPI or APM.
@@ -78,6 +78,10 @@ config SOFTWARE_SUSPEND
 
 	  For more information take a look at <file:Documentation/power/swsusp.txt>.
 
+	  (For now, swsusp is incompatible with PAE aka HIGHMEM_64G on i386.
+	  we need identity mapping for resume to work, and that is trivial
+	  to get with 4MB pages, but less than trivial on PAE).
+
 config PM_STD_PARTITION
 	string "Default resume partition"
 	depends on SOFTWARE_SUSPEND
Signature · 5 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

-- 
VGER BF report: U 0.489608