Skip to contentSkip to conversation

x86: acpi/sleep.c saved_magic fix overflow

PATCH NAK Jaswinder Singh Rajput 8 participants 17 messages Apr 10, 2009 – Apr 12, 2009

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

PATCH 1/1 x86: acpi/sleep.c saved_magic fix overflow

Impact: fix overflow

Constant 0x123456789abcdef0 is so big for a long. So using saved_magic 0x12345678 for both 32 and 64 bit.

---
 arch/x86/kernel/acpi/sleep.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)
arch/x86/kernel/acpi/sleep.c +1 -2
index 7c243a2..1ad317a 100644
@@ -94,7 +94,6 @@ int acpi_save_state_mem(void)
 #ifndef CONFIG_64BIT
 	header->pmode_entry = (u32)&wakeup_pmode_return;
 	header->pmode_cr3 = (u32)(swsusp_pg_dir - __PAGE_OFFSET);
-	saved_magic = 0x12345678;
 #else /* CONFIG_64BIT */
 	header->trampoline_segment = setup_trampoline() >> 4;
 #ifdef CONFIG_SMP
@@ -104,8 +103,8 @@ int acpi_save_state_mem(void)
 	initial_gs = per_cpu_offset(smp_processor_id());
 #endif
 	initial_code = (unsigned long)wakeup_long64;
-	saved_magic = 0x123456789abcdef0;
 #endif /* CONFIG_64BIT */
+	saved_magic = 0x12345678;
 
 	return 0;
 }
Signature · 1 line
1.6.0.6
#1 lore raw

[Jaswinder Singh Rajput - Fri, Apr 10, 2009 at 11:06:55PM +0530]

| Impact: fix overflow
| 
| Constant 0x123456789abcdef0 is so big for a long.
| So using saved_magic 0x12345678 for both 32 and 64 bit.
| 
| Signed-off-by: Jaswinder Singh Rajput <[email protected]>
| ---
|  arch/x86/kernel/acpi/sleep.c |    3 +--
|  1 files changed, 1 insertions(+), 2 deletions(-)
| 
| diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
| index 7c243a2..1ad317a 100644
| --- a/arch/x86/kernel/acpi/sleep.c
| +++ b/arch/x86/kernel/acpi/sleep.c
| @@ -94,7 +94,6 @@ int acpi_save_state_mem(void)
|  #ifndef CONFIG_64BIT
|  	header->pmode_entry = (u32)&wakeup_pmode_return;
|  	header->pmode_cr3 = (u32)(swsusp_pg_dir - __PAGE_OFFSET);
| -	saved_magic = 0x12345678;
|  #else /* CONFIG_64BIT */
|  	header->trampoline_segment = setup_trampoline() >> 4;
|  #ifdef CONFIG_SMP
| @@ -104,8 +103,8 @@ int acpi_save_state_mem(void)
|  	initial_gs = per_cpu_offset(smp_processor_id());
|  #endif
|  	initial_code = (unsigned long)wakeup_long64;
| -	saved_magic = 0x123456789abcdef0;
|  #endif /* CONFIG_64BIT */
| +	saved_magic = 0x12345678;
|  
|  	return 0;
|  }
| -- 
| 1.6.0.6
| 

Hi Jaswinder,

actually long on x86-64 is 8 bytes in size.

Cyrill

#2 lore raw
In reply to Jaswinder Singh Rajput
Quoted 4 lines · Jaswinder Singh Rajput

Impact: fix overflow

Constant 0x123456789abcdef0 is so big for a long. So using saved_magic 0x12345678 for both 32 and 64 bit.

Have you tested the patch? I have feeling that this is tested from assembly... Plus it is only set to long constant on 64-bit, so I don't think you are fixing anything. And if your code is correct the variable is write only...

Quoted 1 line

NAK.

Pavel

Quoted 26 lines
---
 arch/x86/kernel/acpi/sleep.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)
arch/x86/kernel/acpi/sleep.c +1 -2
index 7c243a2..1ad317a 100644
@@ -94,7 +94,6 @@ int acpi_save_state_mem(void)
 #ifndef CONFIG_64BIT
 	header->pmode_entry = (u32)&wakeup_pmode_return;
 	header->pmode_cr3 = (u32)(swsusp_pg_dir - __PAGE_OFFSET);
-	saved_magic = 0x12345678;
 #else /* CONFIG_64BIT */
 	header->trampoline_segment = setup_trampoline() >> 4;
 #ifdef CONFIG_SMP
@@ -104,8 +103,8 @@ int acpi_save_state_mem(void)
 	initial_gs = per_cpu_offset(smp_processor_id());
 #endif
 	initial_code = (unsigned long)wakeup_long64;
-	saved_magic = 0x123456789abcdef0;
 #endif /* CONFIG_64BIT */
+	saved_magic = 0x12345678;
 
 	return 0;
 }
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 11 lines · Pavel Machek
Quoted 4 lines · Jaswinder Singh Rajput

Impact: fix overflow

Constant 0x123456789abcdef0 is so big for a long. So using saved_magic 0x12345678 for both 32 and 64 bit.

Have you tested the patch? I have feeling that this is tested from assembly... Plus it is only set to long constant on 64-bit, so I don't think you are fixing anything. And if your code is correct the variable is write only...

I have tested this:
1. CHECK arch/x86/kernel/acpi/sleep.c
arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long

2. On 64 bit machine box:

	long l1;

	l1 = 0x123456789abcdef0;
	printf("0x%16x : ", l1);

Output: 0x 9abcdef0

--
JSR

#4 lore raw
Quoted 17 lines · Jaswinder Singh Rajput
Quoted 11 lines · Pavel Machek
Quoted 4 lines · Jaswinder Singh Rajput

Impact: fix overflow

Constant 0x123456789abcdef0 is so big for a long. So using saved_magic 0x12345678 for both 32 and 64 bit.

Have you tested the patch? I have feeling that this is tested from assembly... Plus it is only set to long constant on 64-bit, so I don't think you are fixing anything. And if your code is correct the variable is write only...

I have tested this:
1. CHECK arch/x86/kernel/acpi/sleep.c
arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long

Quoted 7 lines

2. On 64 bit machine box:

	long l1;

	l1 = 0x123456789abcdef0;
	printf("0x%16x : ", l1);

Output: 0x 9abcdef0

I don't have 64-bit box handy, but please make sure that suspend/resume works before and after your change.

If the variable is _not_ tested from assembly, just remove saved_magic...

Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#5 lore raw
In reply to Jaswinder Singh Rajput

[Jaswinder Singh Rajput - Fri, Apr 10, 2009 at 11:35:29PM +0530]

| On Fri, 2009-04-10 at 19:55 +0200, Pavel Machek wrote:
| > On Fri 2009-04-10 23:06:55, Jaswinder Singh Rajput wrote:
| > > Impact: fix overflow
| > 
| > > Constant 0x123456789abcdef0 is so big for a long.
| > > So using saved_magic 0x12345678 for both 32 and 64 bit.
| > 
| > Have you tested the patch? I have feeling that this is tested from
| > assembly... Plus it is only set to long constant on 64-bit, so I don't
| > think you are fixing anything. And if your code is correct the
| > variable is write only...
| > 
| 
| I have tested this:
| 1.  CHECK   arch/x86/kernel/acpi/sleep.c
| arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long
| 
| 2. On 64 bit machine box:
| 	long l1;
| 
| 	l1 = 0x123456789abcdef0;
| 	printf("0x%16x : ", l1);
| 
| Output: 0x        9abcdef0
| 
| --
| JSR
| 
| 

"0x%lx" will help

Cyrill

#6 lore raw
In reply to Jaswinder Singh Rajput
Quoted 17 lines · Jaswinder Singh Rajput
Quoted 11 lines · Pavel Machek
Quoted 4 lines · Jaswinder Singh Rajput

Impact: fix overflow

Constant 0x123456789abcdef0 is so big for a long. So using saved_magic 0x12345678 for both 32 and 64 bit.

Have you tested the patch? I have feeling that this is tested from assembly... Plus it is only set to long constant on 64-bit, so I don't think you are fixing anything. And if your code is correct the variable is write only...

I have tested this:
1. CHECK arch/x86/kernel/acpi/sleep.c
arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long

Quoted 8 lines

2. On 64 bit machine box:

	long l1;

	l1 = 0x123456789abcdef0;
	printf("0x%16x : ", l1);

Output: 0x 9abcdef0

OK printf("0x%16llx : ", l1);
Output: 0x123456789abcdef0

--
JSR

#7 lore raw
In reply to Pavel Machek
Quoted 10 lines · Pavel Machek
Quoted 4 lines · Jaswinder Singh Rajput

Impact: fix overflow

Constant 0x123456789abcdef0 is so big for a long. So using saved_magic 0x12345678 for both 32 and 64 bit.

Have you tested the patch? I have feeling that this is tested from assembly... Plus it is only set to long constant on 64-bit, so I don't think you are fixing anything. And if your code is correct the variable is write only...

$ git grep saved_magic
kernel/acpi/sleep.c:    saved_magic = 0x12345678;
kernel/acpi/sleep.c:    saved_magic = 0x123456789abcdef0;
kernel/acpi/sleep.h:extern long saved_magic;
kernel/acpi/wakeup_32.S:        movl    %cs:saved_magic, %eax
kernel/acpi/wakeup_32.S:ENTRY(saved_magic)      .long   0
kernel/acpi/wakeup_64.S:        movq    saved_magic, %rax
kernel/acpi/wakeup_64.S:ENTRY(saved_magic)      .quad   0

wakeup_32.S:
movl %cs:saved_magic, %eax
cmpl $0x12345678, %eax

wakeup_64.S:
movq saved_magic, %rax
movq $0x123456789abcdef0, %rdx

So patch is obviously bogus.

Sam

#8 lore raw
Quoted 26 lines · Sam Ravnborg
Quoted 10 lines · Pavel Machek
Quoted 4 lines · Jaswinder Singh Rajput

Impact: fix overflow

Constant 0x123456789abcdef0 is so big for a long. So using saved_magic 0x12345678 for both 32 and 64 bit.

Have you tested the patch? I have feeling that this is tested from assembly... Plus it is only set to long constant on 64-bit, so I don't think you are fixing anything. And if your code is correct the variable is write only...

$ git grep saved_magic
kernel/acpi/sleep.c:    saved_magic = 0x12345678;
kernel/acpi/sleep.c:    saved_magic = 0x123456789abcdef0;
kernel/acpi/sleep.h:extern long saved_magic;
kernel/acpi/wakeup_32.S:        movl    %cs:saved_magic, %eax
kernel/acpi/wakeup_32.S:ENTRY(saved_magic)      .long   0
kernel/acpi/wakeup_64.S:        movq    saved_magic, %rax
kernel/acpi/wakeup_64.S:ENTRY(saved_magic)      .quad   0

wakeup_32.S:
movl %cs:saved_magic, %eax
cmpl $0x12345678, %eax

Quoted 6 lines

wakeup_64.S:
movq saved_magic, %rax
movq $0x123456789abcdef0, %rdx

So patch is obviously bogus.

Sorry, I was confused by this sparse warning on 64 bit box:

CHECK arch/x86/kernel/acpi/sleep.c
arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long

atleast I found problem in sparse ;-)

--
JSR

#9 lore raw
Quoted 34 lines · Jaswinder Singh Rajput
Quoted 26 lines · Sam Ravnborg
Quoted 10 lines · Pavel Machek
Quoted 4 lines · Jaswinder Singh Rajput

Impact: fix overflow

Constant 0x123456789abcdef0 is so big for a long. So using saved_magic 0x12345678 for both 32 and 64 bit.

Have you tested the patch? I have feeling that this is tested from assembly... Plus it is only set to long constant on 64-bit, so I don't think you are fixing anything. And if your code is correct the variable is write only...

$ git grep saved_magic
kernel/acpi/sleep.c:    saved_magic = 0x12345678;
kernel/acpi/sleep.c:    saved_magic = 0x123456789abcdef0;
kernel/acpi/sleep.h:extern long saved_magic;
kernel/acpi/wakeup_32.S:        movl    %cs:saved_magic, %eax
kernel/acpi/wakeup_32.S:ENTRY(saved_magic)      .long   0
kernel/acpi/wakeup_64.S:        movq    saved_magic, %rax
kernel/acpi/wakeup_64.S:ENTRY(saved_magic)      .quad   0

wakeup_32.S:
movl %cs:saved_magic, %eax
cmpl $0x12345678, %eax

Quoted 6 lines

wakeup_64.S:
movq saved_magic, %rax
movq $0x123456789abcdef0, %rdx

So patch is obviously bogus.

Quoted 6 lines

Sorry, I was confused by this sparse warning on 64 bit box:

CHECK arch/x86/kernel/acpi/sleep.c
arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long

atleast I found problem in sparse ;-)

Sparse tells you that the constant is so big it is long - correct. So rather than being implicit appending an L would make sparse happy.

So please use the constant 0x123456789abcdef0L and I think sparse is silenced. And code/functionality is the same.

Sam

#10 lore raw
Quoted 6 lines · Sam Ravnborg
Quoted 5 lines · Jaswinder Singh Rajput

Sorry, I was confused by this sparse warning on 64 bit box:

CHECK arch/x86/kernel/acpi/sleep.c
arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long

Quoted 8 lines
Quoted 1 line

atleast I found problem in sparse ;-)

Sparse tells you that the constant is so big it is long - correct. So rather than being implicit appending an L would make sparse happy.

So please use the constant 0x123456789abcdef0L and I think sparse is silenced. And code/functionality is the same.

Yes, this fixes the sparse warning, Thanks :-)

arch/x86/kernel/acpi/sleep.c +1 -1
index 7c243a2..3445d0f 100644
@@ -104,7 +104,7 @@ int acpi_save_state_mem(void)
        initial_gs = per_cpu_offset(smp_processor_id());
 #endif
        initial_code = (unsigned long)wakeup_long64;
-       saved_magic = 0x123456789abcdef0;
+       saved_magic = 0x123456789abcdef0L;
 #endif /* CONFIG_64BIT */
 
        return 0;

--

Acked-by
#11 lore raw
Quoted 9 lines · Jaswinder Singh Rajput
Quoted 8 lines · Sam Ravnborg
Quoted 6 lines · Jaswinder Singh Rajput

Sorry, I was confused by this sparse warning on 64 bit box:

CHECK arch/x86/kernel/acpi/sleep.c
arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long

atleast I found problem in sparse ;-)

Quoted 8 lines
Quoted 6 lines

Sparse tells you that the constant is so big it is long - correct. So rather than being implicit appending an L would make sparse happy.

So please use the constant 0x123456789abcdef0L and I think sparse is silenced. And code/functionality is the same.

Yes, this fixes the sparse warning, Thanks :-)

Quoted 16 lines
arch/x86/kernel/acpi/sleep.c +1 -2
index 7c243a2..3445d0f 100644
@@ -104,7 +104,7 @@ int acpi_save_state_mem(void)
        initial_gs = per_cpu_offset(smp_processor_id());
 #endif
        initial_code = (unsigned long)wakeup_long64;
-       saved_magic = 0x123456789abcdef0;
+       saved_magic = 0x123456789abcdef0L;
 #endif /* CONFIG_64BIT */
 
        return 0;
 
--
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#12 lore raw
Quoted 13 lines · Pavel Machek
Quoted 12 lines · Jaswinder Singh Rajput
Quoted 11 lines · Sam Ravnborg
Quoted 6 lines · Jaswinder Singh Rajput

Sorry, I was confused by this sparse warning on 64 bit box:

CHECK arch/x86/kernel/acpi/sleep.c
arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long

atleast I found problem in sparse ;-)

Sparse tells you that the constant is so big it is long - correct. So rather than being implicit appending an L would make sparse happy.

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

So please use the constant 0x123456789abcdef0L and I think sparse is silenced. And code/functionality is the same.

Yes, this fixes the sparse warning, Thanks :-)

OK

Ingo, are you going to take it or should I handle it?

Rafael

Quoted 15 lines
Quoted 15 lines
arch/x86/kernel/acpi/sleep.c +1 -2
index 7c243a2..3445d0f 100644
@@ -104,7 +104,7 @@ int acpi_save_state_mem(void)
        initial_gs = per_cpu_offset(smp_processor_id());
 #endif
        initial_code = (unsigned long)wakeup_long64;
-       saved_magic = 0x123456789abcdef0;
+       saved_magic = 0x123456789abcdef0L;
 #endif /* CONFIG_64BIT */
 
        return 0;
 
--
Acked-by
#13 lore raw
Quoted 18 lines · Rafael J. Wysocki
Quoted 17 lines · Pavel Machek
Quoted 16 lines · Jaswinder Singh Rajput
Quoted 14 lines · Sam Ravnborg
Quoted 6 lines · Jaswinder Singh Rajput

Sorry, I was confused by this sparse warning on 64 bit box:

CHECK arch/x86/kernel/acpi/sleep.c
arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long

atleast I found problem in sparse ;-)

Sparse tells you that the constant is so big it is long - correct. So rather than being implicit appending an L would make sparse happy.

So please use the constant 0x123456789abcdef0L and I think sparse is silenced. And code/functionality is the same.

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

Yes, this fixes the sparse warning, Thanks :-)

OK

Ingo, are you going to take it or should I handle it?

Sure, feel free to take it - thanks.

Ingo

#14 lore raw
In reply to Jaswinder Singh Rajput
Quoted 6 lines · Jaswinder Singh Rajput

Sorry, I was confused by this sparse warning on 64 bit box:

CHECK arch/x86/kernel/acpi/sleep.c
arch/x86/kernel/acpi/sleep.c:107:16: warning: constant 0x123456789abcdef0 is so big it is long

atleast I found problem in sparse ;-)

Sigh. There is no problem in sparse. The problem is between keyboard and chair.

Care to read what the warning says ?

     warning: constant 0x123456789abcdef0 is so big it is long

It needs to be: 0x123456789abcdef0L

Seriously, this kind of mechanical "fix" warnings no matter what and without actually understanding the issues is more than counter productive.

The point is, you force people into wasting a lot of time into what should have been handled in 10 seconds. This is not efficient at all and you need to improve your ways of communicaton and be more careful about the patches you submit.

Thanks,

tglx

#15 lore raw
In reply to Jaswinder Singh Rajput
Quoted 5 lines · Jaswinder Singh Rajput

Impact: fix overflow

Constant 0x123456789abcdef0 is so big for a long. So using saved_magic 0x12345678 for both 32 and 64 bit.

Uhm, what? saved_magic is a long, which is 64 bits wide on on 64 bit systems.