Hi, I've created a rPath appliance which boots from a USB memory stick. Today I upgraded extlinux on my rPath appliance and the memory stick froze the system during POST. After a lot of trial and errors I found out this could be fixed by wiping the USB memory stick and reproduced it by writing extlinux (3.72) to the stick. I finally found out how to fix this issue, by entering the correct values for --sectors and --heads when installing extlinux. So by using the default settings it froze my BIOS, and now it works by using the values from "fdisk -l" (which are: 253 heads, 62 sectors/track) The BIOS is Phoenix Technologies, LTD 6.00 PG in a Shuttle Inc FG33. The USB memory stick is a SanDisk Cruzer Contour 8GB. Not related but might be useful, to boot from an USB stick using this Shuttle disable 1st, 2nd, 3rd etc Boot Device from "Advanced BIOS Features" (this is optional) and enable (enabled by default) Legacy USB Boot (legacy??) in Integrated Peripherals / USB Devices. And, select USB Harddisk mode. I hope this issue can be fixed or someone could tell me what I'm doing wrong. Thanks in advance! Dick
Dick wrote:> I hope this issue can be fixed or someone could tell me what I'm doing > wrong.What you're doing wrong? Using a broken BIOS and complaining here rather than to your vendor.> I finally found out how to fix this issue, by entering the correct > values for --sectors and --heads when installing extlinux. > So by using the default settings it froze my BIOS, and now it works by > using the values from "fdisk -l" (which are: 253 heads, 62 > sectors/track)Your partition table has bizarre geometry, and the BIOS is probably obtaining the geometry from the partition table. That's all fine and good; the problem is that it then doesn't report then correct geometry when Syslinux asks "what's your geometry"? In fact, the fact that --sectors and --heads worked when used indicates that the geometry query BIOS call in fact just returns an error. I generally recommend using zipdrive geometry (64 heads, 32 sectors) for USB sticks -- it works on a few systems which are otherwise broken. This should be used for fdisk as well as extlinux. -hpa -- H. Peter Anvin, Intel Open Source Technology Center I work for Intel. I don't speak on their behalf.
Hi Peter, Thank you for your fast reply and I'm sorry my previous message was a complaint. I didn't intended to do so. I'm very happy with Syslinux. I couldn't change the partition table of the USB stick, I've tried to do so by using the eXpert function of fdisk but when I Wrote the new partition table using a different geometry and re-read the partition table by restarting fdisk the geometry was restored to H253/S62. It would be great if extlinux could report the sectors/heads settings which are asked by Syslinux from the BIOS to stdout. greetings, Dick
Dick Marinus wrote:> Hi Peter, > > Thank you for your fast reply and I'm sorry my previous message was a > complaint. I didn't intended to do so. I'm very happy with Syslinux. > > I couldn't change the partition table of the USB stick, I've tried to do > so by using the eXpert function of fdisk but when I Wrote the new > partition table using a different geometry and re-read the partition > table by restarting fdisk the geometry was restored to H253/S62. > > It would be great if extlinux could report the sectors/heads settings > which are asked by Syslinux from the BIOS to stdout. >It would be, but it's not possible; at the point where you're seeing the boot failed message you're seeing that message because *that's the length of message that fits*. Adding a binary-to-ascii conversion routine in the few bytes there are is functionally impossible. Keep in mind that the geometry fdisk reports has nothing to do with reality, either on-disk or in-BIOS. So relying on it is probably not the right thing to do. On the other hand, some flash media has been known to "help" with the geometry (to try to get better alignment) by clobbering the MBR; that's also rather suboptimal behavior and such flash media should be considered faulty. -hpa -- H. Peter Anvin, Intel Open Source Technology Center I work for Intel. I don't speak on their behalf.