Skip to contentSkip to conversation

Disable swsusp on CONFIG_HIGHMEM64

PATCH In review Jan Spitalnik 2 participants 9 messages Jan 6, 2006 – Jan 11, 2006

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Disable swsusp on CONFIG_HIGHMEM64

Hello,

suspending to disk is not supported on CONFIG_HIGHMEM64G setups
(http://suspend2.net/features). Also suspend to ram doesn't work. This patch fixes Kconfig to disallow such combination. I'm not 100% sure about the ACPI_SLEEP part, as it might be disabling some working setup - but i think that s2r and s2d are the only acpi sleeps allowed, no?

Bye,
spity

PS: I didn't know that this is not supported so I had some nice oops after resume (from ram) and s2d didn't resume at all :-)

Signature · 2 lines
Jan Spitalnik
[email protected]
drivers/acpi/Kconfig +1 -1
diff --exclude=CVS --exclude=.svn -up --new-file --recursive linux-2.6.old/drivers/acpi/Kconfig linux-2.6/drivers/acpi/Kconfig
@@ -46,7 +46,7 @@ if ACPI
 
 config ACPI_SLEEP
 	bool "Sleep States"
-	depends on X86 && (!SMP || SUSPEND_SMP)
+	depends on X86 && (!SMP || SUSPEND_SMP) && !HIGHMEM64G
 	depends on PM
 	default y
 	---help---
kernel/power/Kconfig +1 -1
diff --exclude=CVS --exclude=.svn -up --new-file --recursive linux-2.6.old/kernel/power/Kconfig linux-2.6/kernel/power/Kconfig
@@ -38,7 +38,7 @@ config PM_DEBUG
 
 config SOFTWARE_SUSPEND
 	bool "Software Suspend"
-	depends on PM && SWAP && (X86 && (!SMP || SUSPEND_SMP)) || ((FVR || PPC32) && !SMP)
+	depends on PM && SWAP && (X86 && (!SMP || SUSPEND_SMP) && !HIGHMEM64G) || ((FVR || PPC32) && !SMP)
 	---help---
 	  Enable the possibility of suspending the machine.
 	  It doesn't need APM.
#1 lore raw
Quoted 4 lines · Jan Spitalnik

Hello,

suspending to disk is not supported on CONFIG_HIGHMEM64G setups
(http://suspend2.net/features). Also suspend to ram doesn't work. This patch

NAK. suspend2.net describes very different code.

Quoted 3 lines

fixes Kconfig to disallow such combination. I'm not 100% sure about the ACPI_SLEEP part, as it might be disabling some working setup - but i think that s2r and s2d are the only acpi sleeps allowed, no?

s2ram probably works. Try getting it working without highmem64, then turn it on.

#2 lore raw

Dne pátek 06 leden 2006 00:43 Pavel Machek napsal(a):

Quoted 8 lines · Jan Spitalnik
Quoted 5 lines

Hello,

suspending to disk is not supported on CONFIG_HIGHMEM64G setups
(http://suspend2.net/features). Also suspend to ram doesn't work. This patch

NAK. suspend2.net describes very different code.

Well, I was using suspend2.net's page just as reference, to point out the fact that HIGHMEM is on both suspend "platforms" supported only up to 4G. I was not refering to suspend2's actual features, but rather swsusp's (or what's the proper name for suspend1 code). So i guess the patch still holds, no?

Quoted 7 lines
Quoted 4 lines

fixes Kconfig to disallow such combination. I'm not 100% sure about the ACPI_SLEEP part, as it might be disabling some working setup - but i think that s2r and s2d are the only acpi sleeps allowed, no?

s2ram probably works. Try getting it working without highmem64, then turn it on.

It works with HIGHMEM but not HIGHMEM64G. You can find the oops from HIGHMEM64G below. It crashes very reliably on little stress after resume.

Jan  5 20:12:47 hathor Unable to handle kernel paging request at virtual address dc025000
Jan  5 20:12:47 hathor printing eip:
Jan  5 20:12:47 hathor c013edcc
Jan  5 20:12:47 hathor *pde = 00000000
Jan  5 20:12:47 hathor Oops: 0002 [#3]
Jan  5 20:12:47 hathor Modules linked in: usbmouse i915 drm autofs4 bridge snd_seq_oss snd_seq_midi_event snd_seq snd_seq_device snd_pcm_oss snd_mixer_oss pcspkr parport_pc\
 parport irtty_sir sir_dev irda crc_ccitt 8250_pnp snd_intel8x0m snd_intel8x0 snd_ac97_codec snd_ac97_bus snd_pcm snd_timer snd soundcore snd_page_alloc ehci_hcd usbhid uhc\
i_hcd intel_agp agpgart usbcore binfmt_misc nls_iso8859_2 nls_cp852 vfat fat nls_base tg3 evdev thermal fan 8250_acpi 8250 serial_core tun loop yenta_socket rsrc_nonstatic \
pcmcia pcmcia_core firmware_class rfcomm l2cap bluetooth
Jan  5 20:12:47 hathor CPU:    0
Jan  5 20:12:47 hathor EIP:    0060:[<c013edcc>]    Not tainted VLI
Jan  5 20:12:47 hathor EFLAGS: 00010296   (2.6.15)
Jan  5 20:12:47 hathor EIP is at buffered_rmqueue+0x1bc/0x230
Jan  5 20:12:47 hathor eax: 00000000   ebx: c13804a0   ecx: 00000400   edx: dc025000
Jan  5 20:12:47 hathor esi: 00000000   edi: dc025000   ebp: 00000000   esp: de021e64
Jan  5 20:12:47 hathor ds: 007b   es: 007b   ss: 0068
Jan  5 20:12:47 hathor Process konqueror (pid: 11756, threadinfo=de020000 task=ed4600b0)
Jan  5 20:12:47 hathor Stack: c13804a0 00000003 0000001f c0319690 000081a4 00000001 00000000 c13804a0
Jan  5 20:12:47 hathor c03198ec 00000044 00000000 00000003 c013efda c0319660 00000000 000280d2
Jan  5 20:12:47 hathor 00000003 00000044 c03198e8 000280d2 ed4600b0 df0b4320 c013f05f 000280d2
Jan  5 20:12:47 hathor Call Trace:
Jan  5 20:12:47 hathor [<c013efda>] get_page_from_freelist+0xba/0xe0
Jan  5 20:12:47 hathor [<c013f05f>] __alloc_pages+0x5f/0x320
Jan  5 20:12:47 hathor [<c0150557>] anon_vma_prepare+0x17/0xa0
Jan  5 20:12:47 hathor [<c014b4e4>] do_anonymous_page+0x54/0x1e0
Jan  5 20:12:47 hathor [<c014baec>] __handle_mm_fault+0x12c/0x330
Jan  5 20:12:47 hathor [<c02cf5f5>] notifier_call_chain+0x25/0x50
Jan  5 20:12:47 hathor [<c02cf0c3>] do_page_fault+0x1c3/0x6d0
Jan  5 20:12:47 hathor [<c0159f33>] sys_close+0x53/0x70
Jan  5 20:12:47 hathor [<c02cef00>] do_page_fault+0x0/0x6d0
Jan  5 20:12:47 hathor [<c0103b03>] error_code+0x4f/0x54
Jan  5 20:12:47 hathor Code: fb 00 7e 44 89 5c 24 14 8b 5c 24 1c 31 f6 90 89 1c 24 ba 03 00 00 00 89 54 24 04 e8 6f 8a fd ff 89 c2 89 c7 b9 00 04 00 00 89 f0 <f3> ab 89 14 \
24 b8 03 00 00 00 45 89 44 24 04 e8 10 8b fd ff 83
Jan  5 20:12:47 hathor <6>note: konqueror[11756] exited with preempt_count 1
Signature · 2 lines
Jan Spitalnik
[email protected]
#3 lore raw
Quoted 8 lines · Jan Spitalnik

Dne pátek 06 leden 2006 00:43 Pavel Machek napsal(a):

Quoted 7 lines · Jan Spitalnik
Quoted 5 lines

Hello,

suspending to disk is not supported on CONFIG_HIGHMEM64G setups
(http://suspend2.net/features). Also suspend to ram doesn't work. This patch

Quoted 7 lines
Quoted 1 line

NAK. suspend2.net describes very different code.

Well, I was using suspend2.net's page just as reference, to point out the fact that HIGHMEM is on both suspend "platforms" supported only up to 4G. I was not refering to suspend2's actual features, but rather swsusp's (or what's the proper name for suspend1 code). So i guess the patch still holds, no?

No.

Quoted 10 lines
Quoted 6 lines
Quoted 3 lines

fixes Kconfig to disallow such combination. I'm not 100% sure about the ACPI_SLEEP part, as it might be disabling some working setup - but i think that s2r and s2d are the only acpi sleeps allowed, no?

s2ram probably works. Try getting it working without highmem64, then turn it on.

It works with HIGHMEM but not HIGHMEM64G. You can find the oops from HIGHMEM64G below. It crashes very reliably on little stress after resume.

s2ram should not depend on ammount of memory. Try debugging it, but do not disable feature just because it does not work for you. I'd start with minimum drivers...

Signature · 1 line
Thanks, Sharp!
#4 lore raw

Dne pátek 06 leden 2006 05:30 Pavel Machek napsal(a):

Quoted 11 lines · Jan Spitalnik
Quoted 10 lines

Dne pátek 06 leden 2006 00:43 Pavel Machek napsal(a):

Quoted 8 lines · Jan Spitalnik
Quoted 5 lines

Hello,

suspending to disk is not supported on CONFIG_HIGHMEM64G setups
(http://suspend2.net/features). Also suspend to ram doesn't work. This patch

NAK. suspend2.net describes very different code.

Quoted 7 lines
Quoted 5 lines

Well, I was using suspend2.net's page just as reference, to point out the fact that HIGHMEM is on both suspend "platforms" supported only up to 4G. I was not refering to suspend2's actual features, but rather swsusp's (or what's the proper name for suspend1 code). So i guess the patch still holds, no?

No.

Could you be please more specific? Is there some list of swsusp's features? swsusp.txt says that it "A: It should work okay with highmem." Does that mean both possible highmem configurations?

Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
Quoted 4 lines

fixes Kconfig to disallow such combination. I'm not 100% sure about the ACPI_SLEEP part, as it might be disabling some working setup - but i think that s2r and s2d are the only acpi sleeps allowed, no?

s2ram probably works. Try getting it working without highmem64, then turn it on.

Quoted 6 lines
Quoted 2 lines

It works with HIGHMEM but not HIGHMEM64G. You can find the oops from HIGHMEM64G below. It crashes very reliably on little stress after resume.

s2ram should not depend on ammount of memory. Try debugging it, but do not disable feature just because it does not work for you. I'd start with minimum drivers...

Well, I've tried it with the bare minimum that was needed to run the system, but it did the same. I'm sorry but i lack the knowledge to properly debug it on source level. Do you see something that perhaps i don't see in the oops? Maybe some clues as what might be going wrong?

Thanks,

Signature · 2 lines
Jan Spitalnik
[email protected]
#5 lore raw
Quoted 11 lines
Quoted 11 lines
Quoted 10 lines
Quoted 7 lines
Quoted 4 lines

fixes Kconfig to disallow such combination. I'm not 100% sure about the ACPI_SLEEP part, as it might be disabling some working setup - but i think that s2r and s2d are the only acpi sleeps allowed, no?

s2ram probably works. Try getting it working without highmem64, then turn it on.

It works with HIGHMEM but not HIGHMEM64G. You can find the oops from HIGHMEM64G below. It crashes very reliably on little stress after resume.

Quoted 8 lines
Quoted 3 lines

s2ram should not depend on ammount of memory. Try debugging it, but do not disable feature just because it does not work for you. I'd start with minimum drivers...

Well, I've tried it with the bare minimum that was needed to run the system, but it did the same. I'm sorry but i lack the knowledge to properly debug it on source level. Do you see something that perhaps i don't see in the oops? Maybe some clues as what might be going wrong?

I tried highmem64 on -mm2 here, and machine did not even boot :-(. I may try it again on latest -git a bit later.
Pavel

.config +5 -4
@@ -1,7 +1,7 @@
 #
 # Automatically generated make config: don't edit
 # Linux kernel version: 2.6.15-mm2
-# Sun Jan  8 13:55:53 2006
+# Sun Jan  8 14:18:22 2006
 #
 CONFIG_X86_32=y
 CONFIG_GENERIC_TIME=y
@@ -165,9 +165,10 @@
 # CONFIG_DELL_RBU is not set
 # CONFIG_DCDBAS is not set
 # CONFIG_NOHIGHMEM is not set
-CONFIG_HIGHMEM4G=y
-# CONFIG_HIGHMEM64G is not set
+# CONFIG_HIGHMEM4G is not set
+CONFIG_HIGHMEM64G=y
 CONFIG_HIGHMEM=y
+CONFIG_X86_PAE=y
 CONFIG_ARCH_FLATMEM_ENABLE=y
 CONFIG_ARCH_SPARSEMEM_ENABLE=y
 CONFIG_ARCH_SELECT_MEMORY_MODEL=y
@@ -179,7 +180,7 @@
 CONFIG_FLAT_NODE_MEM_MAP=y
 CONFIG_SPARSEMEM_STATIC=y
 CONFIG_SPLIT_PTLOCK_CPUS=4
-# CONFIG_HIGHPTE is not set
+CONFIG_HIGHPTE=y
 # CONFIG_MATH_EMULATION is not set
 CONFIG_MTRR=y
 # CONFIG_EFI is not set
Signature · 1 line
Thanks, Sharp!
#6 lore raw
In reply to Jan Spitalnik
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
Quoted 2 lines

s2ram probably works. Try getting it working without highmem64, then turn it on.

It works with HIGHMEM but not HIGHMEM64G. You can find the oops from HIGHMEM64G below. It crashes very reliably on little stress after resume.

Quoted 8 lines
Quoted 3 lines

s2ram should not depend on ammount of memory. Try debugging it, but do not disable feature just because it does not work for you. I'd start with minimum drivers...

Well, I've tried it with the bare minimum that was needed to run the system, but it did the same. I'm sorry but i lack the knowledge to properly debug it on source level. Do you see something that perhaps i don't see in the oops? Maybe some clues as what might be going wrong?

HIGHMEM64G breaks boot on my thinkpad (pretty recent git). I guess I'm not solving this one.
Pavel

Signature · 1 line
Thanks, Sharp!
#8 lore raw
In reply to Jan Spitalnik
Quoted 11 lines
Quoted 7 lines
Quoted 5 lines

Well, I was using suspend2.net's page just as reference, to point out the fact that HIGHMEM is on both suspend "platforms" supported only up to 4G. I was not refering to suspend2's actual features, but rather swsusp's (or what's the proper name for suspend1 code). So i guess the patch still holds, no?

No.

Could you be please more specific? Is there some list of swsusp's features? swsusp.txt says that it "A: It should work okay with highmem." Does that mean both possible highmem configurations?

It should work in all configurations. I'd like to try fixing it before disabling it in config.

Quoted 8 lines
Quoted 3 lines

s2ram should not depend on ammount of memory. Try debugging it, but do not disable feature just because it does not work for you. I'd start with minimum drivers...

Well, I've tried it with the bare minimum that was needed to run the system, but it did the same. I'm sorry but i lack the knowledge to properly debug it on source level. Do you see something that perhaps i don't see in the oops? Maybe some clues as what might be going wrong?

No clues :-(. I'll try reproducing it locally.

Signature · 1 line
Thanks, Sharp!