Hello, FYI. This problem looks similar to the problem I reported in Februari http://www.syslinux.org/archives/2013-February/019517.html Kind regards, Wim. On 06/28/2013 10:02 AM, Matt Fleming wrote:> On Fri, 28 Jun, at 08:13:03AM, N?meth P?ter wrote: >> Hi guys! >> I tried to boot memtest86+ v4.2 (renamed memtest.bin to memtest), and >> i have issues to boot it with pxelinux 5.x and greater.Pxelinux 4.06 >> works fine both memtest 4.1 and 4.2. >> With version 6.00 no error message caught, only writes out dots. Any >> keypress does a reboot.With version 5.01 the error message is the >> follow:Loading images/memtest/memtest86... okBooting kernel failed: >> Bad file number >> The pxelinux.cfg/default file contents:DEFAULT memtest2 >> LABEL memtest KERNEL images/memtest/memtest.original APPEND console=ttyS0,9600n8 >> LABEL memtest2 KERNEL images/memtest/memtest86 APPEND console=ttyS0,9600n8 >> PROMPT 1TIMEOUT 100 >> Any ide how to solve this? > > This is the infamous PXE-stack-smasher-bug. It's a regression in the > 5.xx series (which means it's also in the 6.xx series). This has been > one of those difficult to reproduce bugs, because it's very much > dependent on the BIOS memory map. > > Is anyone that is hitting this issue willing/able to build PXELINUX from > a git branch to help with debugging this? >-- Wim Janssen Email: WPJ.Janssen at science.ru.nl Radboud University Nijmegen Faculty of Science, room: HG03.055 C&CZ tel: +31243652744
> Hello, > > FYI. This problem looks similar to the problem I reported in Februari > http://www.syslinux.org/archives/2013-February/019517.html > > Kind regards, > Wim. > > On 06/28/2013 10:02 AM, Matt Fleming wrote: > > On Fri, 28 Jun, at 08:13:03AM, N?meth P?ter wrote: > >> Hi guys! > >> I tried to boot memtest86+ v4.2 (renamed memtest.bin to memtest), and > >> i have issues to boot it with pxelinux 5.x and greater.Pxelinux 4.06 > >> works fine both memtest 4.1 and 4.2. > >> With version 6.00 no error message caught, only writes out dots. Any > >> keypress does a reboot.With version 5.01 the error message is the > >> follow:Loading images/memtest/memtest86... okBooting kernel failed: > >> Bad file number > >> The pxelinux.cfg/default file contents:DEFAULT memtest2 > >> LABEL memtest KERNEL images/memtest/memtest.original APPEND console=ttyS0,9600n8 > >> LABEL memtest2 KERNEL images/memtest/memtest86 APPEND console=ttyS0,9600n8 > >> PROMPT 1TIMEOUT 100 > >> Any ide how to solve this? > > > > This is the infamous PXE-stack-smasher-bug. It's a regression in the > > 5.xx series (which means it's also in the 6.xx series). This has been > > one of those difficult to reproduce bugs, because it's very much > > dependent on the BIOS memory map. > > > > Is anyone that is hitting this issue willing/able to build PXELINUX from > > a git branch to help with debugging this? > > > > -- > Wim Janssen Email: WPJ.Janssen at science.ru.nl > Radboud University Nijmegen > Faculty of Science, room: HG03.055 > C&CZ tel: +31243652744 > _______________________________________________@Wim and @All, FWIW, in 2013-February there still were other additional issues with some kernels, including memory testers, so having a problem to boot memtest(86|86+|other) might or might not be related to the "PXE-stack-smasher" bug. For example, in v.4.06, the cfg entry could be: LINUX memtest and it would work OK, but if the entry was: COM32 linux.c32 APPEND memtest it could / would fail, even using SYSLINUX (or PXELINUX). This behavior with linux.c32 seemed to indicate that there were already potential problems ahead in Syslinux 5.xx with certain kernels. Also, as a remainder, memtest is for BIOS only (AFAIK, at least current versions). So, if someone was having problems with memtest before, I would suggest trying again with the latest Syslinux build(s), booting with SYSLINUX and with PXELINUX. Comparing the results might help clarify whether the problem is actually related to the "PXE-stack-smasher" bug, or to something else (which might have been already solved in the last few months / days / hours). Best Regards, Ady.
On 06/28/2013 04:25 AM, Ady wrote:> > @Wim and @All, > FWIW, in 2013-February there still were other additional issues with > some kernels, including memory testers, so having a problem to boot > memtest(86|86+|other) might or might not be related to the > "PXE-stack-smasher" bug. > > For example, in v.4.06, the cfg entry could be: > LINUX memtest > > and it would work OK, but if the entry was: > COM32 linux.c32 > APPEND memtest > > it could / would fail, even using SYSLINUX (or PXELINUX). This > behavior with linux.c32 seemed to indicate that there were already > potential problems ahead in Syslinux 5.xx with certain kernels. > > Also, as a remainder, memtest is for BIOS only (AFAIK, at least > current versions). > > So, if someone was having problems with memtest before, I would > suggest trying again with the latest Syslinux build(s), booting with > SYSLINUX and with PXELINUX. Comparing the results might help clarify > whether the problem is actually related to the "PXE-stack-smasher" > bug, or to something else (which might have been already solved in > the last few months / days / hours). >This problem is understood, it just isn't entirely trivial to fix. The problem is that memtest is an "ancient" kernel which has to be loaded at a fixed address, but the PXE stack also wants to use the same address. This causes issues if one also wants sane error behavior. Incidentally, a member of my team at Intel did a patch to update this bit of memtest; I'm should check to see if it went upstream yet. -hpa