Syslinux - Jul 2013 - FW: pxelinux 5.x, 6.x memtest problem

It depends on the exact memory map of the system.

Gene Cumm <gene.cumm at gmail.com> wrote:
>On Fri, Jun 28, 2013 at 7:31 PM, H. Peter Anvin <hpa at zytor.com>
wrote:
>> On 06/28/2013 04:07 PM, Gene Cumm wrote:
>>> On Fri, Jun 28, 2013 at 12:53 PM, Gene Cumm <gene.cumm at
gmail.com>
>wrote:
>>>> On Fri, Jun 28, 2013 at 10:13 AM, Ady <ady-sf at
hotmail.com> wrote:
>>>>
>>>>> But, in Syslinux 4.06, there is linux.c32, which is a sort
of
>>>>> introduction to Syslinux 5.xx (please allow me to be not
strictly
>>>>> accurate in this sentence in this context).
>>>>>
>>>>> Booting with any variant of Syslinux 4.06, but using
linux.c32,
>>>>> memtest would fail, which indicated some kind of problem
related
>to
>>>>> the status of Syslinux 5.xx at the time v4.06 was released.
>>>>
>>>>
>>>> PXELINUX 4.06 with Memtest86-4.20 and Memtest86+-4.20 binaries
>>>> (renamed) using LINUX directly works.  Using linux.c32:
>>>> "syslinux_boot_linux() failed: Error 0"
>>>>
>>>> PXELINUX 5.10, 5.11-pre3, 6.01-pre5 with binaries LINUX
directly:
>>>> "Booting kernel failed: Invalid argument"
>>>
>>> N?meth P?ter :
>>> Would you happen to have a link for the Memtest86+ binaries that
you
>>> attempted to use with PXELINUX?
>>>
>>
>> It is presumably the stock memtest-86+ 4.20 binaries.  As I wrote
>> elsewhere in the thread, the problem is understood.
>
>I posted my results from what I thought was "the stock memtest-86+
>4.20 binaries" which I presume to be Syslinux core safety measures.
>What was different?  Why was Nemeth able to load something that should
>be expected to fail?  Was it packed with something like upx (which
>failed to compress Memtest binaries for me)?
>
>--
>-Gene
-- 
Sent from my mobile phone. Please excuse brevity and lack of formatting.

On Sat, 29 Jun, at 08:13:32AM, H. Peter Anvin wrote:
> It depends on the exact memory map of the system.
 
Would anyone in this thread care to try out 6.02-pre6? It contains a
bunch of changes that should hopefully get us on the path to fixing this
regression.

If you're feeling adventurous, you could even try,

 debug.c32 -e pxelinux_scan_memory

(either from the boot: prompt or via a config option) before invoking
memtest to see whether the PXE code is marking the memory regions as
available for use.

-- 
Matt Fleming, Intel Open Source Technology Center

Hi,

if you can provide a prepared src tarball like you do for releases, I can test.
But I still cannot build from git source (as you may remember).


-- 
Regards,
Igor