Crash Dump Region

Crash Dump Region

[Horms]

Hi,

I am currently looking over the code that places the crashdump
region into /proc/iomem, and the code that determines its base
address if it is not passed on the kernel comamnd. It seems to me that 
the current code allows the crashkernel to be placed incide a
/proc/iomem region of any type. Is this behaviour correct?
If not, should it be restricted to "System RAM" regions?

-- 
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/

RE: Crash Dump Region

[Zou, Nanhai]

> -----Original Message-----
> From: Horms [mailto:horms@verge.net.au]
> Sent: 2007Äê3ÔÂ6ÈÕ 9:57
> To: Zou, Nanhai
> Cc: Linux-IA64; fastboot
> Subject: Crash Dump Region
> 
> Hi,
> 
> I am currently looking over the code that places the crashdump
> region into /proc/iomem, and the code that determines its base
> address if it is not passed on the kernel comamnd. It seems to me that
> the current code allows the crashkernel to be placed incide a
> /proc/iomem region of any type. Is this behaviour correct?
> If not, should it be restricted to "System RAM" regions?
> 
 I not sure if I understand your question.
 Kernel will find a big enough region inside efi memmap with WB attribute, excluding all the other reserved regions.

 Zou Nanhai.

> --
> Horms
>   H: http://www.vergenet.net/~horms/
>   W: http://www.valinux.co.jp/en/

RE: Crash Dump Region

[Zou, Nanhai]

> -----Original Message-----
> From: Zou, Nanhai
> Sent: 2007年3月6日 10:11
> To: 'Horms'
> Cc: Linux-IA64; fastboot
> Subject: RE: Crash Dump Region
> 
> > -----Original Message-----
> > From: Horms [mailto:horms@verge.net.au]
> > Sent: 2007年3月6日 9:57
> > To: Zou, Nanhai
> > Cc: Linux-IA64; fastboot
> > Subject: Crash Dump Region
> >
> > Hi,
> >
> > I am currently looking over the code that places the crashdump
> > region into /proc/iomem, and the code that determines its base
> > address if it is not passed on the kernel comamnd. It seems to me that
> > the current code allows the crashkernel to be placed incide a
> > /proc/iomem region of any type. Is this behaviour correct?
> > If not, should it be restricted to "System RAM" regions?
> >
>  I not sure if I understand your question.
>  Kernel will find a big enough region inside efi memmap with WB attribute,
> excluding all the other reserved regions.
> 
Oh, 
  You mean we should check is_memory_available(md) instead of only check efi_wb(md) in kdump_find_rsvd_region? 
 Yes, I think that is better.
 
 Thanks
 Zou Nanhai

>  Zou Nanhai.
> 
> > --
> > Horms
> >   H: http://www.vergenet.net/~horms/
> >   W: http://www.valinux.co.jp/en/
-
To unsubscribe from this list: send the line "unsubscribe linux-ia64" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Crash Dump Region

[Zou Nan hai]

On Tue, 2007-03-06 at 09:56, Horms wrote:
> Hi,
> 
> I am currently looking over the code that places the crashdump
> region into /proc/iomem, and the code that determines its base
> address if it is not passed on the kernel comamnd. It seems to me that
> the current code allows the crashkernel to be placed incide a
> /proc/iomem region of any type. Is this behaviour correct?
> If not, should it be restricted to "System RAM" regions?
> 
> -- 
> Horms
>   H: http://www.vergenet.net/~horms/
>   W: http://www.valinux.co.jp/en/
> 

Please review the patch.

Thanks
Zou Nan hai

Signed-off-by: Zou Nan hai <nanhai.zou@intel.com>

--- a/arch/ia64/kernel/efi.c	2007-02-27 00:42:06.000000000 -0500
+++ b/arch/ia64/kernel/efi.c	2007-03-05 22:09:40.000000000 -0500
@@ -1154,7 +1154,7 @@ kdump_find_rsvd_region (unsigned long si
 
   for (p = efi_map_start; p < efi_map_end; p += efi_desc_size) {
 	  md = p;
-	  if (!efi_wb(md))
+	  if (!is_memory_available(md))
 		  continue;
 	  start = ALIGN(md->phys_addr, alignment);
 	  end = efi_md_end(md);

Re: Crash Dump Region

[Horms]

On Tue, Mar 06, 2007 at 10:18:59AM +0800, Zou, Nanhai wrote:
> > -----Original Message-----
> > From: Zou, Nanhai
> > Sent: 2007年3月6日 10:11
> > To: 'Horms'
> > Cc: Linux-IA64; fastboot
> > Subject: RE: Crash Dump Region
> > 
> > > -----Original Message-----
> > > From: Horms [mailto:horms@verge.net.au]
> > > Sent: 2007年3月6日 9:57
> > > To: Zou, Nanhai
> > > Cc: Linux-IA64; fastboot
> > > Subject: Crash Dump Region
> > >
> > > Hi,
> > >
> > > I am currently looking over the code that places the crashdump
> > > region into /proc/iomem, and the code that determines its base
> > > address if it is not passed on the kernel comamnd. It seems to me that
> > > the current code allows the crashkernel to be placed incide a
> > > /proc/iomem region of any type. Is this behaviour correct?
> > > If not, should it be restricted to "System RAM" regions?
> > >
> >  I not sure if I understand your question.
> >  Kernel will find a big enough region inside efi memmap with WB attribute,
> > excluding all the other reserved regions.
> > 
> Oh, 
>   You mean we should check is_memory_available(md) instead of only check efi_wb(md) in kdump_find_rsvd_region? 
>  Yes, I think that is better.

Thanks.

I had missed the efi_wb(), which was the cause of some confusion on my
part. But as you suggest, is_memory_available() probaly is better, I'll
comment on your patch separately.

However, what I am more worried about is the case where the base address
is passed by the end-user, in which case it seems that some kind of
check should be added to efi_initialize_iomem_resources().

A quick check on my tiger2 box shows that the following is possible. I
know its a silly example, but I think it does demonstrate the problem
that I was trying to explain in my intial email.  And I think that you
have answered my question - this is not correct

crashkernel=1k@520k

00082000-00083fff : reserved
  00082000-000823ff : Crash kernel

I'll make a patch to fix this up if you have no objections.

-- 
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/

Re: Crash Dump Region

[Horms]

On Tue, Mar 06, 2007 at 10:32:09AM +0800, Zou Nan hai wrote:
> On Tue, 2007-03-06 at 09:56, Horms wrote:
> > Hi,
> > 
> > I am currently looking over the code that places the crashdump
> > region into /proc/iomem, and the code that determines its base
> > address if it is not passed on the kernel comamnd. It seems to me that
> > the current code allows the crashkernel to be placed incide a
> > /proc/iomem region of any type. Is this behaviour correct?
> > If not, should it be restricted to "System RAM" regions?
> 
> Please review the patch.

This looks good to me.

Acked-by: Simon Horman <horms@verge.net.au>

> Thanks
> Zou Nan hai
> 
> Signed-off-by: Zou Nan hai <nanhai.zou@intel.com>
> 
> --- a/arch/ia64/kernel/efi.c	2007-02-27 00:42:06.000000000 -0500
> +++ b/arch/ia64/kernel/efi.c	2007-03-05 22:09:40.000000000 -0500
> @@ -1154,7 +1154,7 @@ kdump_find_rsvd_region (unsigned long si
>  
>    for (p = efi_map_start; p < efi_map_end; p += efi_desc_size) {
>  	  md = p;
> -	  if (!efi_wb(md))
> +	  if (!is_memory_available(md))
>  		  continue;
>  	  start = ALIGN(md->phys_addr, alignment);
>  	  end = efi_md_end(md);
> 
> 

-- 
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/