Comment 6 for bug 540940

Thanks for looking at this! I definitely want to take something like this for Lucid, and this looks fairly close to being usable. A few comments:

Could you use 'quilt refresh -pab' so that you don't change the format of all the patches, please?

Try to pay attention to the style of the surrounding code. Your error-check-BLKPG.patch is formatted completely differently: you have spaces inside parentheses, different brace indentation, tab/space differences, and so on. I know it seems picky but it really makes the patch easier to read when it's in line with what's around it, and you'll need to sync this up for upstream acceptance anyway.

Your removal of the _has_partitions call in linux_disk_commit seems risky; I recall this being added to fix a bug. I would be more comfortable if this were left in, even if it turns out that it isn't strictly necessary now for some reason; it seems to me that it's at least needed for the _kernel_reread_part_table fallback, so it seems that it would be best to check it at the top level.

Is the length the only change that might necessitate telling the kernel to re-read a partition? Are there any circumstances in which a type change might matter too? I'm hoping not but was wondering if you'd thought about this.