swsusp fails to suspend if CONFIG_DEBUG_PAGEALLOC is also enabled
A message that is not in the archive (sent off-list or not yet received): [email protected]
Hi!
Quoted 5 lines
Problem Description:
swsusp normally works, but if I enableCONFIG_DEBUG_PAGEALLOC, it breaks -- after "PM: snapshotting memory", swsusp never gets to the "critical section" and it kind of bounces back from the suspend attempt, for lack of a better way for me to describe the effect.
Okay, that is because low-level assembly requires PSE (4mb pages for kernel) and DEBUG_PAGEALLOC disables that capability.
If you feel like rewriting assembly code to turn off paging (and thus working with PSE), go ahead, but I do not think it is worth the trouble.
OTOH we should at least tell people what went wrong, some people seen same problem on VIA cpus... Please apply,
Pavel
include/asm-i386/suspend.h +4 -2
@@ -10,10 +10,12 @@arch_prepare_suspend (void) {/* If you want to make non-PSE machine work, turn off paging -in do_magic. swsusp_pg_dir should have identity mapping, so +in swsusp_arch_suspend. swsusp_pg_dir should have identity mapping, so it could work... */ -if (!cpu_has_pse) +if (!cpu_has_pse) { +printk (KERN_ERR "PSE is required for swsusp.\n" );return -EPERM ; + }return 0 ; }
Signature · 2 lines
People were complaining that M$ turns users into beta-testers... ...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!
Quoted 12 lines · Pavel Machek
Hi!
Quoted 5 lines
Problem Description:
swsusp normally works, but if I enableCONFIG_DEBUG_PAGEALLOC, it breaks -- after "PM: snapshotting memory", swsusp never gets to the "critical section" and it kind of bounces back from the suspend attempt, for lack of a better way for me to describe the effect.Okay, that is because low-level assembly requires PSE (4mb pages for kernel) and
DEBUG_PAGEALLOCdisables that capability.
Quoted 7 lines
If you feel like rewriting assembly code to turn off paging (and thus working with PSE), go ahead, but I do not think it is worth the trouble.
OTOH we should at least tell people what went wrong, some people seen same problem on VIA cpus... Please apply,
Isn't some Kconfig solution appropriate here?
Quoted 21 lines
include/asm-i386/suspend.h +4
-2@@ -10,10 +10,12 @@arch_prepare_suspend (void) {/* If you want to make non-PSE machine work, turn off paging -in do_magic. swsusp_pg_dir should have identity mapping, so +in swsusp_arch_suspend. swsusp_pg_dir should have identity mapping, so it could work... */ -if (!cpu_has_pse) +if (!cpu_has_pse) { +printk (KERN_ERR "PSE is required for swsusp.\n" );return -EPERM ; + }return 0 ; }Signature · 2 lines
People were complaining that M$ turns users into beta-testers... ...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!
PATCH 1/1 kconfig: DEBUG_PAGEALLOC and SOFTWARE_SUSPEND are incompatible on i386
On i386, SOFTWARE_SUSPEND requires the CPU to have PSE support, but DEBUG_PAGEALLOC disables PSE. Thus, allowing both options to be enabled simultaneously makes no sense. This patch disables DEBUG_PAGEALLOC if SOFTWARE_SUSPEND is enabled; it also displays a comment to briefly explain why DEBUG_PAGEALLOC is missing in that case.
I have tested this patch against oldconfig and menuconfig on 2.6.11-bk2.
arch/i386/Kconfig.debug +4 -1
@@ -38,9 +38,12 @@ This option will slow down process creation somewhat. +comment "Page alloc debug is incompatible with Software Suspend on i386" + depends on DEBUG_KERNEL && SOFTWARE_SUSPEND + config DEBUG_PAGEALLOC bool "Page alloc debugging" - depends on DEBUG_KERNEL + depends on DEBUG_KERNEL && !SOFTWARE_SUSPEND help Unmap pages from the kernel linear mapping after free_pages(). This results in a large slowdown, but helps to find certain types
Re: [PATCH] kconfig: DEBUG_PAGEALLOC and SOFTWARE_SUSPEND are incompatible on i386
Quoted 5 lines · Barry K. Nathan
On i386,
SOFTWARE_SUSPENDrequires the CPU to have PSE support, butDEBUG_PAGEALLOCdisables PSE. Thus, allowing both options to be enabled simultaneously makes no sense. This patch disablesDEBUG_PAGEALLOCifSOFTWARE_SUSPENDis enabled; it also displays a comment to briefly explain whyDEBUG_PAGEALLOCis missing in that case.
ACK.
Pavel
Quoted 16 lines
arch/i386/Kconfig.debug +4
-1@@ -38,9 +38,12 @@ This option will slow down process creation somewhat. +comment "Page alloc debug is incompatible with Software Suspend on i386" + depends on DEBUG_KERNEL && SOFTWARE_SUSPEND + config DEBUG_PAGEALLOC bool "Page alloc debugging" - depends on DEBUG_KERNEL + depends on DEBUG_KERNEL && !SOFTWARE_SUSPEND help Unmap pages from the kernel linear mapping after free_pages(). This results in a large slowdown, but helps to find certain types
Signature · 2 lines
People were complaining that M$ turns users into beta-testers... ...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!
Quoted 9 lines · Andrew Morton
Quoted 7 lines · Pavel Machek
Okay, that is because low-level assembly requires PSE (4mb pages for kernel) and
DEBUG_PAGEALLOCdisables that capability.If you feel like rewriting assembly code to turn off paging (and thus working with PSE), go ahead, but I do not think it is worth the trouble.
Quoted 5 lines
Quoted 3 lines
OTOH we should at least tell people what went wrong, some people seen same problem on VIA cpus... Please apply,
Isn't some
Kconfigsolution appropriate here?
Yes, but only for the CONFIG_DEBUG_PAGEALLOC case, it does not solve the "cpu has no PSE" case for VIA CPUs. So the Kconfig solution is an extra bonus.
Stefan
Hi,
Quoted 14 lines · Stefan Seyfried
Quoted 13 lines · Andrew Morton
Quoted 10 lines · Pavel Machek
Okay, that is because low-level assembly requires PSE (4mb pages for kernel) and
DEBUG_PAGEALLOCdisables that capability.If you feel like rewriting assembly code to turn off paging (and thus working with PSE), go ahead, but I do not think it is worth the trouble.
OTOH we should at least tell people what went wrong, some people seen same problem on VIA cpus... Please apply,
Quoted 5 lines
Quoted 1 line
Isn't some
Kconfigsolution appropriate here?Yes, but only for the
CONFIG_DEBUG_PAGEALLOCcase, it does not solve the "cpu has no PSE" case for VIA CPUs. So theKconfigsolution is an extra bonus.
Yeah - separate issues. Suspend2 has worked with CONFIG_DEBUG_PAGEALLOC. You just have to map in all the pages before doing the atomic copies.
Regards,
Nigel
Signature · 6 lines
Nigel Cunningham Software Engineer, Canberra, Australia http://www.cyclades.com Bus: +61 (2) 6291 9554; Hme: +61 (2) 6292 8028; Mob: +61 (417) 100 574 Maintainer of Suspend2 Kernel Patches http://softwaresuspend.berlios.de
Quoted 1 line · Stefan Seyfried
Andrew Morton wrote:
[snip]
Quoted 5 lines
Quoted 1 line
Isn't some
Kconfigsolution appropriate here?Yes, but only for the
CONFIG_DEBUG_PAGEALLOCcase, it does not solve the "cpu has no PSE" case for VIA CPUs. So theKconfigsolution is an extra bonus.
Note that I've posted a Kconfig solution here:
http://marc.theaimsgroup.com/?l=linux-kernel&m=111017249931972&w=2
Regarding Pavel's patch, it seems to me that it might be better to print the message at boot time, instead of (or in addition to?) his patch. Maybe we should be disabling swsusp altogether at boot in that case, if that's not unreasonably hard to implement.
-Barry K. Nathan <[email protected]>
Hi!
Quoted 6 lines
Quoted 5 lines
Quoted 1 line
Isn't some
Kconfigsolution appropriate here?Yes, but only for the
CONFIG_DEBUG_PAGEALLOCcase, it does not solve the "cpu has no PSE" case for VIA CPUs. So theKconfigsolution is an extra bonus.
Quoted 7 lines
Note that I've posted a
Kconfigsolution here:
http://marc.theaimsgroup.com/?l=linux-kernel&m=111017249931972&w=2Regarding Pavel's patch, it seems to me that it might be better to print the message at boot time, instead of (or in addition to?) his patch. Maybe we should be disabling swsusp altogether at boot in that case, if that's not unreasonably hard to implement.
Hmm, yes, that would certainly be better. It would need new per-architecture hook... Feel free to implement it.
Pavel
Signature · 2 lines
People were complaining that M$ turns users into beta-testers... ...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!
Quoted 7 lines · Pavel Machek
Quoted 4 lines
Regarding Pavel's patch, it seems to me that it might be better to print the message at boot time, instead of (or in addition to?) his patch. Maybe we should be disabling swsusp altogether at boot in that case, if that's not unreasonably hard to implement.
Hmm, yes, that would certainly be better. It would need new per-architecture hook... Feel free to implement it.
It looks like it's going to be more difficult than I was hoping for.
I would recommend accepting Pavel's patch. It's a definite improvement over the existing code, and it could be a long while until I have a boot-time message patch that is both functional and clean enough to be incorporated into mainline.
-Barry K. Nathan <[email protected]>