Fix random memory corruption during suspend-to-RAM resume

Fix random memory corruption during suspend-to-RAM resume

[Pavel Machek]

Hi!

We really want to use wakeup_stack as stack, not as pointer to
stack... Please apply,
								Pavel

--- clean/arch/i386/kernel/acpi_wakeup.S	2002-12-18 22:20:47.000000000 +0100
+++ linux-swsusp/arch/i386/kernel/acpi_wakeup.S	2003-02-09 21:13:19.000000000 +0100
@@ -31,7 +31,7 @@
 	movw	%cs, %ax
 	movw	%ax, %ds					# Make ds:0 point to wakeup_start
 	movw	%ax, %ss
-	mov	wakeup_stack - wakeup_code, %sp			# Private stack is needed for ASUS board
+	mov	$(wakeup_stack - wakeup_code), %sp		# Private stack is needed for ASUS board
 	movw	$0x0e00 + 'S', %fs:(0x12)
 
 	pushl	$0						# Kill any dangerous flags

-- 
Worst form of spam? Adding advertisment signatures ala sourceforge.net.
What goes next? Inserting advertisment *into* email?

Re: Fix random memory corruption during suspend-to-RAM resume

[Linus Torvalds]

On Sun, 9 Feb 2003, Pavel Machek wrote:
> 
> We really want to use wakeup_stack as stack, not as pointer to
> stack... Please apply,

Hmm.. The stack grows downwards, are you sure you don't really mean

	mov $(wakeup_end-wakeup_code),%sp

(because wakeup_end is the end of the wakeup_stack area..)

		Linus

Re: Fix random memory corruption during suspend-to-RAM resume

[Pavel Machek]

Hi!

> > We really want to use wakeup_stack as stack, not as pointer to
> > stack... Please apply,
> 
> Hmm.. The stack grows downwards, are you sure you don't really mean
> 
> 	mov $(wakeup_end-wakeup_code),%sp
> 
> (because wakeup_end is the end of the wakeup_stack area..)

Yes, I'm sure:

        .long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
        .long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
        .long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
        .long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
wakeup_stack:
        .long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
        .long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
        .long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
        .long 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0

Okay, I might want to kill the second set of longs ;-).

I need to substract wakeup_code because code is being relocated in
quite a strange way.

							Pavel
-- 
Casualities in World Trade Center: ~3k dead inside the building,
cryptography in U.S.A. and free speech in Czech Republic.

Re: Fix random memory corruption during suspend-to-RAM resume

[Linus Torvalds]

On Tue, 11 Feb 2003, Pavel Machek wrote:
> > 
> > Hmm.. The stack grows downwards, are you sure you don't really mean
> > 
> > 	mov $(wakeup_end-wakeup_code),%sp
> > 
> > (because wakeup_end is the end of the wakeup_stack area..)
> 
> Yes, I'm sure:

No. The code is crap. Please send a _fixed_ patch that clearly puts the 
stack pointer at the _end_ of the stack.

What you are sure of is that is _happens_ to work by pure luck.

I'd rather have known-broken code in the kernel than code that works by 
mistake. The former at least gets fixed.

		Linus

Re: Fix random memory corruption during suspend-to-RAM resume

[Pavel Machek]

Hi!

> > > Hmm.. The stack grows downwards, are you sure you don't really mean
> > > 
> > > 	mov $(wakeup_end-wakeup_code),%sp
> > > 
> > > (because wakeup_end is the end of the wakeup_stack area..)
> > 
> > Yes, I'm sure:
> 
> No. The code is crap. Please send a _fixed_ patch that clearly puts the 
> stack pointer at the _end_ of the stack.
> 
> What you are sure of is that is _happens_ to work by pure luck.
> 
> I'd rather have known-broken code in the kernel than code that works by 
> mistake. The former at least gets fixed.

It should be already in the queue (waiting on my home machine sorry),
along with fix to make it all fit in one page.

								Pavel
-- 
Casualities in World Trade Center: ~3k dead inside the building,
cryptography in U.S.A. and free speech in Czech Republic.