LZMA magic number

[Myles Watson]

Is there a magic number for LZMA?  Should there be?

I accidentally used an uncompressed payload with v2 when it expected a
compressed payload, and it gave me the message:

       Decoder scratchpad too small!!
       Decoding error = 1

I think it would be nicer to have an error like:

       Payload not compressed with lzma!

Myles

[Peter Stuge]

On Fri, Jan 25, 2008 at 10:55:24AM -0700, Myles Watson wrote:
> I accidentally used an uncompressed payload with v2 when it
> expected a compressed payload, and it gave me the message:
>
> Decoder scratchpad too small!!
>        Decoding error = 1

I got this too.
> I think it would be nicer to have an error like:
>
> Payload not compressed with lzma!

And we can check it already at build time.


//Peter

[Carl-Daniel Hailfinger]

On 25.01.2008 18:58, Peter Stuge wrote:
> On Fri, Jan 25, 2008 at 10:55:24AM -0700, Myles Watson wrote:
> > I accidentally used an uncompressed payload with v2 when it
> > expected a compressed payload, and it gave me the message:
> >
> > Decoder scratchpad too small!!
> >        Decoding error = 1
>
> I got this too.
> > I think it would be nicer to have an error like:
> >
> > Payload not compressed with lzma!
>
> And we can check it already at build time.

LZMA compressed payloads to not have a magic number, all you can do is
check if the header parameters are in a reasonable range. Sorry.
This problem fortunately does not exist in v3 because LAR does the
compression itself and notes the compression type in th header.

Regards,
Carl-Daniel

[PATCH] Re: LZMA magic number

[Ward Vandewege]

On Sat, Jan 26, 2008 at 02:18:39AM +0100, Carl-Daniel Hailfinger wrote:
> On 25.01.2008 18:58, Peter Stuge wrote:
> > On Fri, Jan 25, 2008 at 10:55:24AM -0700, Myles Watson wrote:
> > > I accidentally used an uncompressed payload with v2 when it
> > > expected a compressed payload, and it gave me the message:
> > >
> > > Decoder scratchpad too small!!
> > >        Decoding error = 1
> >
> > I got this too.
> > > I think it would be nicer to have an error like:
> > >
> > > Payload not compressed with lzma!
> >
> > And we can check it already at build time.
>
> LZMA compressed payloads to not have a magic number, all you can do is
> check if the header parameters are in a reasonable range. Sorry.
> This problem fortunately does not exist in v3 because LAR does the
> compression itself and notes the compression type in th header.

I ran into this runtime error as well today.

Please find a patch attached that makes coreboot a bit smarter in dealing
with this situation, and avoids the fatal runtime error.

Thanks,
Ward.

-- 
Ward Vandewege <[e-mail]>
Free Software Foundation - Senior Systems Administrator

[PATCH] Re: LZMA magic number

[Myles Watson]

> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]]
> On Behalf Of Ward Vandewege
> Sent: Wednesday, August 27, 2008 3:02 PM
> To: Carl-Daniel Hailfinger
> Cc: [e-mail]
> Subject: [coreboot] [PATCH] Re: LZMA magic number
>
> On Sat, Jan 26, 2008 at 02:18:39AM +0100, Carl-Daniel Hailfinger wrote:
> > On 25.01.2008 18:58, Peter Stuge wrote:
> > > On Fri, Jan 25, 2008 at 10:55:24AM -0700, Myles Watson wrote:
> > > > I accidentally used an uncompressed payload with v2 when it
> > > > expected a compressed payload, and it gave me the message:
> > > >
> > > > Decoder scratchpad too small!!
> > > >        Decoding error = 1
> > >
> > > I got this too.
> > > > I think it would be nicer to have an error like:
> > > >
> > > > Payload not compressed with lzma!
> > >
> > > And we can check it already at build time.
> >
> > LZMA compressed payloads to not have a magic number, all you can do is
> > check if the header parameters are in a reasonable range. Sorry.
> > This problem fortunately does not exist in v3 because LAR does the
> > compression itself and notes the compression type in th header.
>
> I ran into this runtime error as well today.
>
> Please find a patch attached that makes coreboot a bit smarter in dealing
> with this situation, and avoids the fatal runtime error.

Ward,

I like the idea of assuming no compression if uncompressing it fails.  I
wonder if it would be better to copy it to dest instead of running it in
place.

Thanks,
Myles

[PATCH] Re: LZMA magic number

[Ward Vandewege]

On Wed, Aug 27, 2008 at 03:22:38PM -0600, Myles Watson wrote:
> I like the idea of assuming no compression if uncompressing it fails.  I
> wonder if it would be better to copy it to dest instead of running it in
> place.

If compression is disabled in Config.lb the current code runs it in place as
well. So I guess it's no problem?

Thanks,
Ward.

-- 
Ward Vandewege <[e-mail]>
Free Software Foundation - Senior Systems Administrator

[PATCH] Re: LZMA magic number

[Myles Watson]

> -----Original Message-----
> From: Ward Vandewege [mailto:[e-mail]]
> Sent: Wednesday, August 27, 2008 3:31 PM
> To: Myles Watson
> Cc: 'Carl-Daniel Hailfinger'; [e-mail]
> Subject: Re: [coreboot] [PATCH] Re: LZMA magic number
>
> On Wed, Aug 27, 2008 at 03:22:38PM -0600, Myles Watson wrote:
> > I like the idea of assuming no compression if uncompressing it fails.  I
> > wonder if it would be better to copy it to dest instead of running it in
> > place.
>
> If compression is disabled in Config.lb the current code runs it in place
> as
> well. So I guess it's no problem?

Ward,

I should have checked that.

Acked-by: Myles Watson <[e-mail]>

Thanks,
Myles
> Thanks,
> Ward.

[PATCH] Re: LZMA magic number

[Carl-Daniel Hailfinger]

On 27.08.2008 23:39, Myles Watson wrote:
> > -----Original Message-----
> > From: Ward Vandewege [mailto:[e-mail]]
> > Sent: Wednesday, August 27, 2008 3:31 PM
> > To: Myles Watson
> > Cc: 'Carl-Daniel Hailfinger'; [e-mail]
> > Subject: Re: [coreboot] [PATCH] Re: LZMA magic number
> >
> > On Wed, Aug 27, 2008 at 03:22:38PM -0600, Myles Watson wrote:
> > > I like the idea of assuming no compression if uncompressing it fails.  I
> > > wonder if it would be better to copy it to dest instead of running it in
> > > place.
> >
> > If compression is disabled in Config.lb the current code runs it in place
> > as
> > well. So I guess it's no problem?
>
> Ward,
>
> I should have checked that.
>
> Acked-by: Myles Watson <[e-mail]>

The change looks sensible to me. Since it is tested, here is my ack.
Acked-by: Carl-Daniel Hailfinger <[e-mail]>

Regards,
Carl-Daniel

-- 
http://www.hailfinger.org/

[PATCH] Re: LZMA magic number

[Ward Vandewege]

On Wed, Aug 27, 2008 at 11:48:26PM +0200, Carl-Daniel Hailfinger wrote:
> > Acked-by: Myles Watson <[e-mail]>
>
> The change looks sensible to me. Since it is tested, here is my ack.
> Acked-by: Carl-Daniel Hailfinger <[e-mail]>

r3542.

Thanks,
Ward.

-- 
Ward Vandewege <[e-mail]>
Free Software Foundation - Senior Systems Administrator

[PATCH] Re: LZMA magic number

[Peter Stuge]

Ward Vandewege wrote:
> r3542.

Fantastic improvement. Thank you so much Ward.


//Peter