On Thu, Jul 26, 2007 at 05:25:50AM -0400, Corey Osgood wrote:
You could also try using my FILO IDE patch with the debugging suggested in my message, it may provide some more debugging hints:
Still no dice, doesn't seem to give any extra info (with your patch and debugging). New log file attached, FILO can now find an SATA drive (formatted in reiserfs, which I didn't include support for in FILO).
Add the reiserfs support to FILO and it should load the kernel.
IDE is still showing up as a floating bus.
No?
[..]
file_open: dev=hda1, path=/vmlinuz find_ide_controller: found PCI IDE controller 1106:3149 prog_if=0x8f find_ide_controller: primary channel: native PCI mode find_ide_controller: cmd_base=0x2020 ctrl_base=0x2040 ide_software_reset: Waiting for ide0 to become ready for reset... ok init_drive: Testing for hda init_drive: Probing for hda init_drive: LBA mode, sectors=160836480 init_drive: LBA48 mode, sectors=160836480 init_drive: Init device params... ok hda: LBA48 82GB: HDS728080PLA380
Bingo, here's an LBA48 82GB disk on hda.
init_drive: Testing for hdb init_drive: Testing for hdb ide_readmany: sector 0 to 001105c0 devopen: Partition 1 start 63 length 156456972 ide_readmany: sector 63 to 001183c0 ide_readmany: sector 65 to 001107c0 ide_readmany: sector 66 to 001109c0 Unknown filesystem type
It's reading the partition table and tries to read a magic superblock, but FILO doesn't recognize what it is and gives up.
[..]
file_open: dev=hda, path=<NULL> devopen: already open malloc_diag: alloc: 296 bytes (4 blocks), free: 65232 bytes (1 blocks) ide_readmany: sector 1 to 001107c0 load_linux_header: This looks like a bootdisk image but not like Linux... Unsupported image format
Then FILO tries to boot a kernel written directly to the start of the block device.
.. and then hdb..hdj are tried manually.
Add reiserfs to FILO and try again. :)
Note that my patch will not give as much improvement with reiser as with ext2. I've only patched ext2 to read multiple blocks,it should be done in the other fs drivers too.
(But using 32-bit IO should shorten the time by half at least.)
//Peter