[syslinux] memdisk load failure on HP DL360g9

Rick Miller vmiller at hostileadmin.com
Mon Feb 6 11:17:33 PST 2017


On Thu, Feb 2, 2017 at 1:26 PM, Ady Ady via Syslinux <syslinux at zytor.com>
wrote:

>
> > Hi all,
> >
> > I have observed a problem where, under legacy BIOS booting mode on the
> > DL360g9, memdisk fails to execute.  The same mechanism works fine on the
> > DL360g8.  The system displays the below on the console and ceases to
> > respond and must be power cycled.  Updating the firmware didn't help.
> > There didn't appear to be any similar issues on the mailing list unless I
> > missed them.
> >
> > Wondering if this has been observed before and whether or not there is an
> > existing resolution or workaround?
> >
> > e820: 00000000000e0000 0000000000020000 2
> > e820: 0000000000100000 000000005a6a1000 1
> > e820: 000000005a7a1000 0000000000e40000 2
> > e820: 000000005b5e1000 000000001db1e000 1
> > e820: 00000000790ff000 0000000000100000 2
> > e820: 00000000791ff000 0000000002400000 4
> > e820: 000000007b5ff000 0000000000200000 3
> > e820: 000000007b7ff000 0000000000001000 1
> > e820: 000000007b800000 0000000000800000 2
> > e820: 000000007c000000 0000000003c00000 2
> > e820: 000000007fc00000 0000000000400000 2
> > e820: 0000000080000000 0000000010000000 2
> > e820: 00000000ff800000 0000000000800000 2
> > e820: 0000000100000000 0000001f80000000 1
> > Ramdisk at 0x57ba1000, length 0x02c00000
> > command line: raw harddisk
> > MEMDISK: Image seems to have fractional end cylinder
> > Disk is hd0, 45056 K, C/H/S = 5/255/63 (MBR/MBR), EDD on, rw
> > Using raw access to high memory
> > Code 1744, meminfo 228, cmdline 13, stack 512
> > Total size needed = 2497 bytes, allocating 3K
> > Old dos memory at 0x90c00 (map says 0x94000), loading at 0x90000
> > 1588: 0xffff 15E801: 0x3c00 0x56ba
> > INT 13 08: Success, count = 3, BPT = 0000:0000
>
>
>
> Are there relevant differences between these two systems (such as type
> and amount of RAM)?
>

Both systems consist of 128GB RDIMM RAM; g8 = DDR3 and g9 = DDR4.


Are the "BIOS" settings exactly the same in both systems?
>
> Would you please post the entire syslinux.cfg?
>

iPXE is used in place of syslinux here.

Machines booting under these circumstances boot via the following iPXE
commands:

kernel http://192.168.1.2/path/to/memdisk
imgargs memdisk raw harddisk
initrd http://192.168.1.2/path/to/boot.img
boot

"harddisk" changes to "floppy" when attempting to boot a DOS or memtest
floppy image.


Have you attempted to use/set "complete" CHS values for this HDD image?
>

No


Which CHS values are detected by memdisk when the same HDD image is
> booted in the "g8" system?
>

6/256/63; these same values were detected by memdisk when attempting to
load memtestp.bin from the memtest floppy.


Is the problem happening with every-and-all (HDD) images? Have you
> tried with a simple 1440K floppy image too? (It doesn't matter what
> happens next within the floppy, but rather whether you get the same
> type of error message from memdisk.)
>

Booting to memtest86+-5.01.bin exhibits the same behavior while booting
memtestp.bin (from the floppy image) is executing something (constant
scrolling of 0X:0200-like messages on stdout).  A separate bootable floppy (
allbootdisks.com) was also successfully booted to a DOS prompt.


Which exact version of Syslinux are we talking about? Package? Official
> pre-built binaries? Your own build?
>

As previously mentioned, iPXE is used in place of syslinux.


Are the binaries (e.g. the bootloader file, c32 files, memdisk, etc.)
> originated from the same exact build?
>

Only the memdisk binary in varying versions has been tested, all of which
exhibited identical behavior.

MEMDISK 4.05 0x4ee25341
MEMDISK 6.03 0x5873ba2a
MEMDISK 6.04 pre1-20-gbb41e93


I would like to ask you to try the same Syslinux binaries to boot
> memtest86+ (please use the one from www.memtest.org ).
>
> The reason for this memetst86+ request is not to actually test your RAM
> (which might be useful anyway), but rather to test whether the
> memtest86+ kernel is correctly booting (using Syslinux) in these
> machines. There have been reports in which the memtest86+ 5.01 kernel
> fails to boot in certain hardware (e.g. Dell XPS 13 9350, BIOS version
> 1.4.10, November 22, 2016) when it is being launched from Syslinux
> 6.03+ but it runs correctly in the same hardware when Syslinux 4.xx is
> used instead. Such reports would suggest the possibility of a
> regression bug / problem in Syslinux 6.xx in comparison to 4.xx.
>

Booting to memtest86+-5.01.bin exhibits the same behavior.



-- 
Take care
Rick Miller


More information about the Syslinux mailing list