[LinuxBIOS] [PATCH] LAR compression in xip not supported

[Alex Beregszaszi]

Hi,

attached patch returns if someone tries to execute in place a compressed
entry.

[Stefan Reinauer]

* Alex Beregszaszi  [070904 13:19]:
> Hi,
>
> attached patch returns if someone tries to execute in place a compressed
> entry.

r492

[ron minnich]

On 9/4/07, Alex Beregszaszi  wrote:
> Hi,
>
> attached patch returns if someone tries to execute in place a compressed
> entry.

This is a must never happen. If the lar tool ever creates this, it has
seriously screwed up. We need to make sure the lar tool can not set
both compression and XIP.

thanks

ron

[Alex Beregszaszi]

Hi,

On Tue, 2007-09-04 at 20:19 -0700, ron minnich wrote:
> On 9/4/07, Alex Beregszaszi  wrote:
> > Hi,
> >
> > attached patch returns if someone tries to execute in place a compressed
> > entry.
>
> This is a must never happen. If the lar tool ever creates this, it has
> seriously screwed up. We need to make sure the lar tool can not set
> both compression and XIP.

LAR doesn't has a field for telling that an entry will be XIP or not.
Whether we run it in place or move it is only our _runtime_ decision,
thus this check is needed.

If you mean that the LAR tool should check for mandatory names (like
initram) which can't be compressed, well, that is handled in the
makefiles (see nocompress:normal/initram).

I think such an extra check wouldn't hurt (maybe if you care about
space, you could merge some checks to print only one message).

[ron minnich]

On 9/5/07, Alex Beregszaszi  wrote:
> LAR doesn't has a field for telling that an entry will be XIP or not.
> Whether we run it in place or move it is only our _runtime_ decision,
> thus this check is needed.

So here is a good point. Why not adopt the existing convention that a
load address of 0xffffffff means "XIP"?
> I think such an extra check wouldn't hurt (maybe if you care about
> space, you could merge some checks to print only one message).

We should do all possible checks at LAR time. Any runtime check that
fails means we had a failure in how we built the lar.

ron

[Marc Jones]

ron minnich wrote:
> So here is a good point. Why not adopt the existing convention that a
> load address of 0xffffffff means "XIP"?

Yes agreed, That is how I did it in my testing.

Marc

[Carl-Daniel Hailfinger]

On 05.09.2007 19:26, Marc Jones wrote:
> ron minnich wrote:
> > So here is a good point. Why not adopt the existing convention that a
> > load address of 0xffffffff means "XIP"?
>
> Yes agreed, That is how I did it in my testing.

Was there a followon patch for lar?

Carl-Daniel