I am not sure this went out, so, resending. If you look at the
original question, you will see the error comes from SCSI, and
not from a Linux kernel LBA check.
In article [EMAIL PROTECTED] you wrote:
decided that the partition table is written in hardware sectorsize
units, not in units of 512-byte sectors. (No doubt because MSDOS
Ok, assuming the problem comes from a wrong partition table. Why doesn't
the block device filter the erroneous LBA, but instead sends it
as is to the SCSI device which answers with an error ?
I was assuming that block devices did check for last_lba. Even if the
geometry is wrong, the number of blocks is still known.
PS: it would be interesting to see if the 1024-bytes problem also happens
without a partition table (ie mke2fs -b 4096 /dev/sda instead of
/dev/sda1). If it does, well, something else is borken somewhere.