On Monday 15 of October 2012 10:40:25 Stephen Warren wrote:
On 10/13/2012 01:38 PM, Pavel Herrmann wrote:
Hi
On Wednesday 10 October 2012 12:14:00 Stephen Warren wrote:
From: Stephen Warren swarren@nvidia.com
This removes the standalone cur_part_nr variable, opening the way to replacing fat_register_device() with fat_set_blk_dev().
Note that when get_partition_info() fails and we use the entire disk, the correct partition number is 0 (whole disk) not 1 (first partition), so that change is also made here.
An alternative to this (and the previous) patch might be to simply remove the partition number from the printf(). I don't know how attached people are to it.
Signed-off-by: Stephen Warren swarren@nvidia.com
Just as a heads up, in the DM any difference between a partition and a whole block device (also between different interfaces for disks) is hidden from any user code (code other than the one keeping track of partitions/disks, that only uses such information to determine whether to scan for partitions), you only get some object that can read/write blocks if you ask it nicely, and you have to make do with that (if you need more then you're probably doing something wrong). As a result, there is no notion of partition number, and the string you are patching up here (along with many others, due to unification of disk interfaces) is changed.
OK, so do you think it'd be better right now to drop patches 1 and 2 in this series, and just remove the partition number from fat.c's printf() call? That'd certainly be simple to do.
Well, in my case I have done a broader abstraction, that could be used for non-continuous partitions as well (think LVM) with minimal effort (think extend identifier structure used for searching to more than interface:number:partnumber, no changes in FS code), and partition number loses any meaning in that context. Whether dropping the number now is an acceptable change would be up to Tom Rini, I would vote for it though, if that meant anything around here.
Pavel Herrmann