Hi Tom,
On 15 February 2016 at 15:37, Tom Rini trini@konsulko.com wrote:
On Sun, Feb 14, 2016 at 07:16:29PM -0700, Simon Glass wrote:
Recent additions of the MMC and DISK uclasses have indicated that it is time to look at adding a uclass for block devices. This series does this and includes a few clean-ups to the partition code also.
A block device is typically a child device of its storage parent. For example an MMC device will have a block-device child. A USB storage device may have multiple block-device children, one for each LUN.
With this series only USB storage and 'host' are converted over to use the new support. Several more remain, including SCSI, IDE and MMC. Each of these should get its own uclass.
The uclass implements only a few basic features. A few tests are added to check that things work as expected.
The code size impact of switching to driver model for block devices is small. One benefit is that it becomes possible to enumerate all block devices, regardless of their type.
Things generally make sense. I've tossed out a few Reviewed-by's for some things but I looked over everything. My only question is have you run buildman -S and compared the numbers, especially for things that aren't opting in yet? A few oddballs have been spotted by that already :)
Yes I found quite a few problems that way. There was SPL code bloat and also general bloat with more code being linked in than before. But I believe all of that was fixed. Let me know if you see anything though.
Regards, Simon