PageReserved removal woes: vbetool, suspend-to-ram breakage

PageReserved removal woes: vbetool, suspend-to-ram breakage

[Dominik Brodowski]

Hi!

Just wanted to let you know that the warning introduced in
[PATCH] core remove PageReserved
triggers when using vbetool, which is needed on several systems for proper
suspend-to-mem support:

[4301282.746000] program vbetool is using MAP_PRIVATE, PROT_WRITE mmap of
		 VM_RESERVED memory, which is deprecated. Please report
		 this to linux-kernel@vger.kernel.org

Reports of this haven't generated many results yet, that's why I'm
addressing this with a subject line which startles more people ;)
http://lkml.org/lkml/2005/11/6/72
http://lkml.org/lkml/2005/11/6/76
http://lkml.org/lkml/2005/11/10/280

	Dominik

Re: PageReserved removal woes: vbetool, suspend-to-ram breakage

[Matthew Garrett]

Dominik Brodowski <linux@dominikbrodowski.net> wrote:

> Just wanted to let you know that the warning introduced in
> [PATCH] core remove PageReserved

It's not much of a warning, it seems to stop vbetool from working (which
may explain a few complaints about ACPI suspend being broken in 2.6.15 -
it's not, but vbetool is...)

vbetool (well, strictly it's LRMI, but...) needs to access real-mode
memory. On the other hand, it's probably a bad idea to let it actually
scribble over RAM that the kernel may be using. So
MAP_PRIVATE|PROT_WRITE seems to be the correct thing to do. This
certainly /seemed/ to work in older kernels, but doesn't any more. What
should I be doing instead?

(I'm also not quite sure why the error claims that it's deprecated,
whereas in fact it doesn't actually work at all. Breaking userspace
without warning isn't terribly nice)
-- 
Matthew Garrett | mjg59-chiark.mail.linux-rutgers.kernel@srcf.ucam.org

Re: PageReserved removal woes: vbetool, suspend-to-ram breakage

[Hugh Dickins]

On Tue, 22 Nov 2005, Matthew Garrett wrote:
> Dominik Brodowski <linux@dominikbrodowski.net> wrote:
> 
> > Just wanted to let you know that the warning introduced in
> > [PATCH] core remove PageReserved
> 
> It's not much of a warning, it seems to stop vbetool from working (which
> may explain a few complaints about ACPI suspend being broken in 2.6.15 -
> it's not, but vbetool is...)
> 
> vbetool (well, strictly it's LRMI, but...) needs to access real-mode
> memory. On the other hand, it's probably a bad idea to let it actually
> scribble over RAM that the kernel may be using. So
> MAP_PRIVATE|PROT_WRITE seems to be the correct thing to do. This
> certainly /seemed/ to work in older kernels, but doesn't any more. What
> should I be doing instead?
> 
> (I'm also not quite sure why the error claims that it's deprecated,
> whereas in fact it doesn't actually work at all. Breaking userspace
> without warning isn't terribly nice)

Sorry about that, we really didn't expect that it was being used.
And sorry I didn't get the fixes ready in time for 2.6.15-rc2.
But they've gone into Linus' git tree a couple of hours ago.
Please let me know if tomorrow's 2.6.15-rc2-git3 does not work for you.

Hugh