[PATCH] safety check in claim_heap()

[PATCH] safety check in claim_heap()

[Robert Millan]

[-- Attachment #1: Type: text/plain, Size: 356 bytes --]


Hi!

What do you think of adding this kind of checks?  I know it's not a
perfect safegard, but it can save your day in some situations (it helped
me during i386/ieee1275 port, when link address was 0x100000).

-- 
Robert Millan

<GPLv2> I know my rights; I want my phone call!
<DRM> What use is a phone call… if you are unable to speak?
(as seen on /.)

[-- Attachment #2: claim_safety.diff --]
[-- Type: text/x-diff, Size: 884 bytes --]

--- ../../grub2/kern/powerpc/ieee1275/init.c	2008-01-14 15:30:59.000000000 +0100
+++ kern/powerpc/ieee1275/init.c	2008-01-20 15:20:23.000000000 +0100
@@ -143,6 +143,17 @@
         (total + (HEAP_MAX_ADDR - addr) > HEAP_MIN_SIZE))	/* only limit ourselves when we can afford to */
        len = HEAP_MAX_ADDR - addr;
 
+    /* In theory, firmware should already prevent this from happening by not
+       listing our own image in /memory/available.  The check below is intended
+       as a safegard in case that doesn't happen.  It does, however, not protect
+       us from corrupting our module area, which extends up to a
+       yet-undetermined region above _end.  */
+    if ((addr < _end) && ((addr + len) > _start))
+      {
+        grub_printf ("Warning: attempt to claim over our own code!\n");
+        len = 0;
+      }
+
     if (len)
       {
 	/* Claim and use it.  */

Re: [PATCH] safety check in claim_heap()

[Marco Gerards]

Robert Millan <rmh@aybabtu.com> writes:

Hi,

> What do you think of adding this kind of checks?  I know it's not a
> perfect safegard, but it can save your day in some situations (it helped
> me during i386/ieee1275 port, when link address was 0x100000).

If this fixes the i386-ieee1275 port, this is fine to me.

--
Marco

Re: [PATCH] safety check in claim_heap()

[Robert Millan]

On Mon, Jan 21, 2008 at 07:07:37PM +0100, Marco Gerards wrote:
> Robert Millan <rmh@aybabtu.com> writes:
> 
> Hi,
> 
> > What do you think of adding this kind of checks?  I know it's not a
> > perfect safegard, but it can save your day in some situations (it helped
> > me during i386/ieee1275 port, when link address was 0x100000).
> 
> If this fixes the i386-ieee1275 port, this is fine to me.

It isn't currently needed by either of the i386-ieee1275 targets (qemu and
xo).

However, it was during an earlier stage of the port in which link address
was 0x100000 rather than 0x10000.

My question is whether it makes sense as a general safety check.

-- 
Robert Millan

<GPLv2> I know my rights; I want my phone call!
<DRM> What use is a phone call… if you are unable to speak?
(as seen on /.)

Re: [PATCH] safety check in claim_heap()

[Marco Gerards]

Robert Millan <rmh@aybabtu.com> writes:

> On Mon, Jan 21, 2008 at 07:07:37PM +0100, Marco Gerards wrote:
>> Robert Millan <rmh@aybabtu.com> writes:
>> 
>> Hi,
>> 
>> > What do you think of adding this kind of checks?  I know it's not a
>> > perfect safegard, but it can save your day in some situations (it helped
>> > me during i386/ieee1275 port, when link address was 0x100000).
>> 
>> If this fixes the i386-ieee1275 port, this is fine to me.
>
> It isn't currently needed by either of the i386-ieee1275 targets (qemu and
> xo).
>
> However, it was during an earlier stage of the port in which link address
> was 0x100000 rather than 0x10000.
>
> My question is whether it makes sense as a general safety check.

Well, if it isn't going to occur with the current firmware
implementations, I do not see the use.  But it will not hurt either.
Simply, I do not care either way...  I leave this up to you ;-)

--
Marco

Re: [PATCH] safety check in claim_heap()

[Robert Millan]

On Mon, Jan 21, 2008 at 10:11:42PM +0100, Marco Gerards wrote:
> > My question is whether it makes sense as a general safety check.
> 
> Well, if it isn't going to occur with the current firmware
> implementations, I do not see the use.  But it will not hurt either.
> Simply, I do not care either way...  I leave this up to you ;-)

Ok, I checked that in.

-- 
Robert Millan

<GPLv2> I know my rights; I want my phone call!
<DRM> What use is a phone call… if you are unable to speak?
(as seen on /.)