FS#31493 - [syslinux] 4.05-7 Intermittent failure to boot with RAID1
Attached to Project:
Arch Linux
Opened by Peter Hardman (shetland_breeder) - Tuesday, 11 September 2012, 08:30 GMT
Last edited by Thomas Bächler (brain0) - Monday, 29 October 2012, 15:32 GMT
Opened by Peter Hardman (shetland_breeder) - Tuesday, 11 September 2012, 08:30 GMT
Last edited by Thomas Bächler (brain0) - Monday, 29 October 2012, 15:32 GMT
|
Details
I have encountered a problem similar to that of
If I reboot the PC it will almost always fail to load the menu (sits for several seconds with a blank screen) and then reboot. If I power the machine off for a few seconds and then power on it will mostly boot normally, but sometimes just keep restarting as above. Judging from the time delay between the POST completing and the machine rebooting _something_ is running in a loop. I have only once had SYSLINUX hang at the copyright notice. This machine runs software RAID1 with two hard drives. I have / in the first partition, the swap (also raid), /home and several other partitions all also RAID1. The metadata is 0.9. and the partitions start at block 63. This array has been in use for 3 years or more now in a Dell PE840, hence the old partitioning and mdadm metadata. I built a new machine a fortnight ago with a Gigabyte GB-H61M-S2PV motherboard and moved the RAID array to the new machine (very smooth - just installed the correct video driver and moved the drives). When I built the machine and before removing the RAID array I installed a single drive and did a clean Arch install. This was using syslinux and there were no boot problems. Syslinux was installed on the RAID array with the syslinux-install_update script, and that gave me the approriate messages for installing on a raid array. I've not done any debugging apart from adding the syntax hilighter lines to syslinux.cfg to make sure there's no bad syntax. The machine is now back to booting with grub-legacy (just restored the MBR bootloader). An obvious step is to swap the array back into the Dell (and the single drive back to the new machine). I shall report back once this is done. Additional info: * package version(s) * config and/or log files etc. Steps to reproduce: |
This task depends upon
Closed by Thomas Bächler (brain0)
Monday, 29 October 2012, 15:32 GMT
Reason for closing: Fixed
Additional comments about closing: _Probably_ fixed, reopen when the problem surfaces again.
Monday, 29 October 2012, 15:32 GMT
Reason for closing: Fixed
Additional comments about closing: _Probably_ fixed, reopen when the problem surfaces again.
In my new hardware this fails to boot every time. If I run the Arch live CD to the point where it logs in and then reboot it will get to the syslinux menu and boot OK.
If I downgrade syslinux to 4.05-6 then again it boots every time.
Seems there is a problem with 4.05-7 on this specific hardware - I have two other PCs (with old hardware) as well as the Dell and a Thinkpad T61 that boot fine with 4.05-7.
Note that syslinux 4.05-7 is now built the way upstream recommends: We do NOT use our own compiler to rebuild the lowlevel bootloader files, but instead we use the files shipped with syslinux ('make installer' target).
Bootloader are incredibly complex, so I suggest you do the following:
1) Download the syslinux tarball from the homepage and install syslinux using their extlinux installer - this should yield the same result as the installer from Arch - if not, we have a bug I cannot explain.
2) Join the #syslinux IRC channel on freenode or the syslinux mailing list. There are experts there who know how to debug these kinds of problems - I sadly don't.
Also, as your issue seems partially fixed by powering on/off the machine, there could be a memory problem that wasn't visible with a different syslinux build. You can verify that your memory is fine with memtest, but that is time-consuming.
You are right, there's plenty of lines of attack yet, but you might have known what the problem was already :)
memtest86+ throws no errors (over 4 runs) BTW.
I probably shan't be able to work on this again until October - away on holiday. If you leave this open I'll update it when I've done some more work on it.
Pete
It works flawlessly.
However - every single file apart from poweroff.com (including extlinux) has a different md5 sum from those in the Arch package. So I suppose our toolchain somehow does something different from Syslinux' toolchain.
Pete
Can I just install syslinux from testing? I'm very reluctant to install everything from testing as this is my 'production' PC.