Changes between Version 5 and Version 6 of hba_driver

Timestamp:

Mar 25, 2015, 7:18:41 PM (9 years ago)

Author:

alain

Comment:

--

hba_driver

@@ -11 +11 @@
 This driver implements two operating mode: 
  * In '''synchronous mode''', the calling task uses a polling strategy on the HBA_PXCI register to detect the command completion (busy waiting). It does not use interrupt. This mode is used by the boot code to load the map.bin file into memory (before MMU activation), or to load the .elf files (after MMU activation). 
- * In '''descheduling mode''', the calling task is descheduled, and the calling task global index is saved in the ''_hba_gtid[32] array, indexed by the command index. The task must be reactivated  when the command is completed. This mode is used, to handle the user system calls. 
+ * In '''descheduling mode''', the calling task is descheduled, and the calling task global index is saved in the ''_hba_gtid[32]'' array, indexed by the command index. The task must be reactivated  when the command is completed. This mode is used, to handle the user system calls. 
 
-As several user tasks can concurrently register commands in the command list, and there is only one HBA interrupt per channel, this interrupt is not linked to a specific calling task: in descheduling mode, the HBA IRQ is a "global" IRQ that is statically routed to processor P[x_io,y_io,0] in cluster_io.  The associated global HBA_ISR send a WAKUP WTI to all tasks that have a completed command. This HBA_ISR uses a read pointer on the Command List to identify the expected command completion. The incrementation of this read pointer does not require atomic_increment as there is no concurrent access on this pointer. 
+As several user tasks can concurrently register commands in the command list, and there is only one HBA interrupt per channel, this interrupt is not linked to a specific calling task: in descheduling mode, the HBA IRQ is a "global" IRQ that is statically routed to processor P[x_io,y_io,0] in cluster_io.  The associated global HBA_ISR send a WAKUP WTI to all tasks that have a completed command. This HBA_ISR uses a read pointer on the Command List to identify the expected command completion. The incrementation of this read pointer does not require atomic_increment() as there is no concurrent access on this pointer. 
 
 The SEG_IOC_BASE address must be defined in the hard_config.h file.