swsusp regression [Was: 2.6.17-mm1]
In-Reply-To: <[email protected]>
Quoted 28 lines · Andrew Morton
Quoted 9 lines
Quoted 2 lines
Code: 05 c4 42 43 c0 31 43 43 c0 c3 8b 2d 68 6e 54 c0 8b 1d 60 6e 54 c0 8b 35 6c 6e 54 c0 8b 3d 70 6d 54 c0 ff 35 74 6e 54 c0 9d c3 90 <e8> 6d 38 ea ff e8 a2 ff ff ff 6a 03 e8 ec b6 de ff 83 c4 04 c3 EIP: [c043431c>] do_suspend_lowlevel+0x0/0x15 SS:ESP 0068:f6cb6ea4~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ha, wait a moment, this is interesting line. Can you trace down which instruction causes this?
We recently changed pagetable handling during swsusp, perhaps thats it? It went to Linus few minutes ago...
That's a good possibility. It does appear to be oopsing at the first instruction of arch/i386/kernel/acpi/wakeup.S:do_suspend_lowlevel(). Perhaps there's enough info in that oops trace to tell us whether it was the instruction fetch which oopsed.
One wonders whether this will help...
arch/i386/kernel/acpi/wakeup.S renamed from arch/i386/kernel/acpi/wakeup.S~a +1
-0@@ -270,6 +270,7 @@ ALIGN ENTRY(saved_magic) .long 0 ENTRY(saved_eip) .long 0 +.text save_registers: leal 4(%esp), %eax movl %eax, saved_context_esp
arch/i386/kernel/acpi/wakeup.S ret_point: · line 311
Quoted 7 lines
arch/i386/kernel/acpi/wakeup.S renamed from arch/i386/kernel/acpi/wakeup.S~a +1
-0@@ -304,6 +305,7 @@ ret_point: call restore_processor_state ret +.data ALIGN # saved registers saved_gdt: .long 0,0
This is in 2.6.17-mm1 already:
From: Shaohua Li <[email protected]>
Move do_suspend_lowlevel to correct segment. If it is in the same hugepage with ro data, mark_rodata_ro will make it unexecutable.
---
arch/i386/kernel/acpi/wakeup.S | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-)
arch/i386/kernel/acpi/wakeup.S renamed from arch/i386/kernel/acpi/wakeup.S~move-do_suspend_lowlevel-to-correct-segment +4 -5
@@ -265,11 +265,6 @@ ENTRY(acpi_copy_wakeup_routine) movl $0x12345678, saved_magic ret -.data -ALIGN -ENTRY(saved_magic) .long 0 -ENTRY(saved_eip) .long 0 - save_registers: leal 4(%esp), %eax movl %eax, saved_context_esp
@@ -304,7 +299,11 @@ ret_point: call restore_processor_state ret +.data ALIGN +ENTRY(saved_magic) .long 0 +ENTRY(saved_eip) .long 0 + # saved registers saved_gdt: .long 0,0 saved_idt: .long 0,0
_
Signature · 2 lines
Chuck "You can't read a newspaper if you can't read." --George W. Bush
Quoted 44 lines · On Fri, 23 Jun 2006 21:54:00 -0400 Chuck Ebbert <[email protected]>…
In-Reply-To: <[email protected]>
Quoted 28 lines · Andrew Morton
Quoted 9 lines
Quoted 2 lines
Code: 05 c4 42 43 c0 31 43 43 c0 c3 8b 2d 68 6e 54 c0 8b 1d 60 6e 54 c0 8b 35 6c 6e 54 c0 8b 3d 70 6d 54 c0 ff 35 74 6e 54 c0 9d c3 90 <e8> 6d 38 ea ff e8 a2 ff ff ff 6a 03 e8 ec b6 de ff 83 c4 04 c3 EIP: [c043431c>] do_suspend_lowlevel+0x0/0x15 SS:ESP 0068:f6cb6ea4~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ha, wait a moment, this is interesting line. Can you trace down which instruction causes this?
We recently changed pagetable handling during swsusp, perhaps thats it? It went to Linus few minutes ago...
That's a good possibility. It does appear to be oopsing at the first instruction of arch/i386/kernel/acpi/wakeup.S:do_suspend_lowlevel(). Perhaps there's enough info in that oops trace to tell us whether it was the instruction fetch which oopsed.
One wonders whether this will help...
arch/i386/kernel/acpi/wakeup.S renamed from arch/i386/kernel/acpi/wakeup.S~a +1
-0@@ -270,6 +270,7 @@ ALIGN ENTRY(saved_magic) .long 0 ENTRY(saved_eip) .long 0 +.text save_registers: leal 4(%esp), %eax movl %eax, saved_context_esparch/i386/kernel/acpi/wakeup.S ret_point: · line 311
Quoted 7 lines
arch/i386/kernel/acpi/wakeup.S renamed from arch/i386/kernel/acpi/wakeup.S~a +1
-0@@ -304,6 +305,7 @@ ret_point: call restore_processor_state ret +.data ALIGN # saved registers saved_gdt: .long 0,0This is in 2.6.17-mm1 already:
Quoted 5 lines
From: Shaohua Li <[email protected]>
Move
do_suspend_lowlevelto correct segment. If it is in the same hugepage with ro data,mark_rodata_rowill make it unexecutable.
OK. But this bug report is against 2.6.17-mm1, isn't it?