On Tue, 13 Jan 2004, Doug Holland wrote:
On Tue 13 Jan 2004 1:31 pm, Doug Holland wrote:
On Tue 13 Jan 2004 11:35 am, Peter Osterlund wrote:
I don't know yet, but I have the same problem on one of my IDE drives (Sony CRX175A). I only have a problem when using native ide support in a 2.6 kernel. In 2.4 the drive works with both native ide and scsi emulation, and it also works with scsi emulation in 2.6.
I have another IDE drive (HP 8100) on the same ide channel which works perfectly with native ide in 2.6, so it can't be that packet writing using native ide is completely broken in 2.6.
It looks like I/O requests sent from the packet driver to the ide driver get shifted 4 sectors. In the log file below, the requested sectors (the numbers within parenthesis on the pkt_gather_data lines) exactly match the reported read sectors (in pkt_end_io_read), except for the 4 sector offset.
This was probably a false alarm. I get the same shift also for cases that work.
I also get an "access beyond end of device" if I try to access the last sector on the disc, so it seems like the offset is added before the actual I/O operation starts.
Or maybe not. The Sony drive reports: pktcdvd: 65534kB available on disc and the HP drive reports (for the same disc): pktcdvd: 65536kB available on disc They report the same values whether I use scsi emulation or native ide, but only in the ide case do I get the "access beyond end of device" errors. However, this is not the only problem, because if I try to create a smaller filesystem that doesn't go anywhere near the last sector, I still get the reported IDE errors.
I would really like to avoid turning on SCSI emulation - it's always been a pain for me to set up, and my drives otherwise work fine without it.
Agreed, it should work without scsi emulation. I'm working on it...
Another strange behavior... After I copy some files to the disk, after I umount and remount it, and do an ls -al, I get the following:
$ ls -l /cdrw ls: /cdrw/climbatize.mp3: Permission denied ls: /cdrw/smack_my_bitch_up.mp3: Permission denied ...
Not that strange. If the block device driver doesn't work correctly, any filesystem you try to put on top of it can develop all sorts of funny filesystem errors. -- Peter Osterlund - petero2@telia.com http://w1.894.telia.com/~u89404340