edd breaking vesafb

edd breaking vesafb

[Pascal Terjan]

Hello

I got a report of vesafb no longer working since edd is used on a
machine (Shuttle FN41SP).
It's a BIOS bug that was easily workarounded by calling set_video before
query_edd.
Does someone see some issue in doing that ?

Re: edd breaking vesafb

[H. Peter Anvin]

[-- Attachment #1: Type: text/plain, Size: 630 bytes --]

Pascal Terjan wrote:
> Hello
> 
> I got a report of vesafb no longer working since edd is used on a
> machine (Shuttle FN41SP).
> It's a BIOS bug that was easily workarounded by calling set_video before
> query_edd.
> Does someone see some issue in doing that ?
> 

Yes, it means that we won't get the status and error messages from 
query_edd.  In fact, we changed the ordering specifically for this reason.

The VESA code and EDD code both use the boot code heap.  One could guess 
that the problem is data left on the heap by the EDD code causing a 
failure when combined with a bad BIOS.

Does this patch work for you?

	-hpa

[-- Attachment #2: diff --]
[-- Type: text/plain, Size: 497 bytes --]

diff --git a/arch/x86/boot/boot.h b/arch/x86/boot/boot.h
index cc0ef13..f087238 100644
--- a/arch/x86/boot/boot.h
+++ b/arch/x86/boot/boot.h
@@ -205,7 +205,13 @@ static inline int isdigit(int ch)
 extern char _end[];
 extern char *HEAP;
 extern char *heap_end;
-#define RESET_HEAP() ((void *)( HEAP = _end ))
+
+static inline void *RESET_HEAP(void)
+{
+	HEAP = _end;
+	memset(HEAP, 0, heap_end-HEAP);
+	return HEAP;
+}
 static inline char *__get_heap(size_t s, size_t a, size_t n)
 {
 	char *tmp;

Re: edd breaking vesafb

[Pascal Terjan]

Le jeudi 18 septembre 2008 à 09:26 -0700, H. Peter Anvin a écrit :
> Pascal Terjan wrote:
> > Hello
> > 
> > I got a report of vesafb no longer working since edd is used on a
> > machine (Shuttle FN41SP).
> > It's a BIOS bug that was easily workarounded by calling set_video before
> > query_edd.
> > Does someone see some issue in doing that ?
> > 
> 
> Yes, it means that we won't get the status and error messages from 
> query_edd.  In fact, we changed the ordering specifically for this reason.

Indeed

> The VESA code and EDD code both use the boot code heap.  One could guess 
> that the problem is data left on the heap by the EDD code causing a 
> failure when combined with a bad BIOS.
> 
> Does this patch work for you?
> 

No it does not help

Re: edd breaking vesafb

[H. Peter Anvin]

Pascal Terjan wrote:
> 
> No it does not help
> 

What happens with edd=off or edd=skipmbr?

	-hpa

Re: edd breaking vesafb

[Pascal Terjan]

Le jeudi 18 septembre 2008 à 11:43 -0700, H. Peter Anvin a écrit :
> Pascal Terjan wrote:
> > 
> > No it does not help
> > 
> 
> What happens with edd=off or edd=skipmbr?
> 

Here is the answer from original reporter, adding him in Cc

edd=off : vesafb works properly
edd=skipmbr : vesafb is not initialized (just like edd=on)
edd=on : vesafb is not initialized 

Re: edd breaking vesafb

[gettinther]

Is there any news on this subject?

http://www.uwsg.iu.edu/hypermail/linux/kernel/0809.2/0890.html

I am still getting the issue with the 2.6.27.rc8 kernel.  I use
bootsplash and the only way for vesafb to work is to disable edd.