On 07/08/2014 05:55 PM, H. Peter Anvin wrote:> On 07/08/2014 03:05 PM, Alexander Perlis wrote: >> >> Question: what's special about the transition from 8203KiB to 8204KiB? >> Both have CHS=8/64/32 and both involve a fractional end cylinder. > > Fractional cylinders with floppy disks is almost certainly a very bad thing. >Agreed, but that doesn't appear to be the issue here. Fractional 8203K FreeDOS does work, yet larger values, including non-fractional ones, do not. Alex
On 07/08/2014 04:42 PM, Alexander Perlis wrote:> On 07/08/2014 05:55 PM, H. Peter Anvin wrote: >> On 07/08/2014 03:05 PM, Alexander Perlis wrote: >>> >>> Question: what's special about the transition from 8203KiB to 8204KiB? >>> Both have CHS=8/64/32 and both involve a fractional end cylinder. >> >> Fractional cylinders with floppy disks is almost certainly a very bad >> thing. >> > > Agreed, but that doesn't appear to be the issue here. Fractional 8203K > FreeDOS does work, yet larger values, including non-fractional ones, do > not. >Yes, that is odd. You might be looking at the transition point from FAT12 to FAT16 maybe... -hpa
On Tue, Jul 8, 2014 at 7:45 PM, H. Peter Anvin <hpa at zytor.com> wrote:> On 07/08/2014 04:42 PM, Alexander Perlis wrote: >> On 07/08/2014 05:55 PM, H. Peter Anvin wrote: >>> On 07/08/2014 03:05 PM, Alexander Perlis wrote: >>>> >>>> Question: what's special about the transition from 8203KiB to 8204KiB? >>>> Both have CHS=8/64/32 and both involve a fractional end cylinder. >>> >>> Fractional cylinders with floppy disks is almost certainly a very bad >>> thing. >>> >> >> Agreed, but that doesn't appear to be the issue here. Fractional 8203K >> FreeDOS does work, yet larger values, including non-fractional ones, do >> not. >> > > Yes, that is odd. You might be looking at the transition point from > FAT12 to FAT16 maybe...Although I've never seen the restart loop, the size-dependence sounds like a hardware-specific memory component like memory map leading to different placement or actual memory usage not lining up with the map. Have you tried these images on other hardware? Have you tried MEMDISK from 3.86 or 4.04? Have you tried a Syslinux variant of 3.86 or 4.04 with MEMDISK 3.86, 4.04 or 6.03-pre18? -- -Gene
> On 07/08/2014 05:55 PM, H. Peter Anvin wrote: > > On 07/08/2014 03:05 PM, Alexander Perlis wrote: > >> > >> Question: what's special about the transition from 8203KiB to 8204KiB? > >> Both have CHS=8/64/32 and both involve a fractional end cylinder. > > > > Fractional cylinders with floppy disks is almost certainly a very bad thing. > > > > Agreed, but that doesn't appear to be the issue here. Fractional 8203K > FreeDOS does work, yet larger values, including non-fractional ones, do not. > > Alex >My guess is that the problem is not the total size (I have used MEMDISK with bigger superfloppy sizes), but the cluster size or the FAT size. For instance, FAT12 should not use more than 4078 data clusters for maximum compatibility. Unless we have at least the exact CHS (not the assumption of MEMDISK nor according to the total size of the image but rather from the FAT values), the amount of FAT sectors (9 for standard 1440KiB floppies, max. 12 for FAT12), and the amount of data clusters, there is no way to know whether the problem is in the superfloppy image or in MEMDISK. Regards, Ady.