while others dislike them being extra commits, let's clean them up once and for all for the existing code. If it's ugly, let it only be ugly once :-) Signed-off-by: Stefan Reinauer <stepan@coresystems.de> Acked-by: Stefan Reinauer <stepan@coresystems.de> git-svn-id: svn://svn.coreboot.org/coreboot/trunk@5507 2b7e53f0-3cfb-0310-b3e9-8179ed1497e1
		
			
				
	
	
		
			23 lines
		
	
	
		
			1.1 KiB
		
	
	
	
		
			Plaintext
		
	
	
	
	
	
			
		
		
	
	
			23 lines
		
	
	
		
			1.1 KiB
		
	
	
	
		
			Plaintext
		
	
	
	
	
	
| - handle error conditions
 | |
| - handle disconnect more gracefully (ie. make calling layer aware that the device doesn't exist somehow)
 | |
| - usbhub:
 | |
|   - proper client enumeration (esp. detach)
 | |
|   - change detection
 | |
|   - power management
 | |
| - handle interrupts more cleverly:
 | |
|   create a new queue for the interrupt with a couple of TD sequences,
 | |
|     - each ending with "breadth first" flag
 | |
|     - linked as a chain
 | |
|   add that queue at the appropriate times in front of the default structure so the max latency is honored
 | |
|     - only one intr chain per framelist item, so it must be arranged appropriately
 | |
|   reads from usb device just look at "invalidated" tds and the results they got
 | |
|   handled tds get reactivated as a ring structure
 | |
|     - added as child of the oldest td
 | |
|     - queue header already dropped the td, so no issue there
 | |
| 
 | |
|   this setup ensures that:
 | |
|   - the max latency of the device is honored
 | |
|   - the client knows the right order of the data
 | |
|   - there is no need for an interrupt handler
 | |
|     - but must be polled at least max latency * num tds times -> more tds = less time pressure
 |