Re: deprecate and convert some sleep_on variants.

[Dave Jones <davej@redhat.com>]

On Tue, Aug 01, 2006 at 01:20:28PM -0500, Nish Aravamudan wrote:

 > >+  wait_queue_t __wait;
 > >+
 > >+  init_waitqueue_entry(&__wait, current);
 > >
 > >   spin_lock_irqsave(&Controller->queue_lock, flags);
 > >   while ((Command = DAC960_AllocateCommand(Controller)) == NULL)
 > >@@ -6314,11 +6317,18 @@ static boolean DAC960_V2_ExecuteUserComm
 > >                                        .SegmentByteCount =
 > >            CommandMailbox->ControllerInfo.DataTransferSize;
 > >          DAC960_ExecuteCommand(Command);
 > >+         add_wait_queue(&Controller->CommandWaitQueue, &__wait);
 > >+         set_current_state(TASK_UNINTERRUPTIBLE);
 > 
 > Could this use prepare_to_wait()

Maybe, though I'd rather not do that conversion with the hardware to test it.
sidenote: prepare_to_wait() and friends could really use some kerneldoc explaining
their purpose rather than their internal workings.

 > >          while 
 > >          (Controller->V2.NewControllerInformation->PhysicalScanActive)
 > >            {
 > >              DAC960_ExecuteCommand(Command);
 > >-             sleep_on_timeout(&Controller->CommandWaitQueue, HZ);
 > >+             schedule_timeout(HZ);
 > >+             set_current_state(TASK_UNINTERRUPTIBLE);
 > 
 > and schedule_timeout_uninterruptible() (which is redundant for the
 > first invocation, I suppose)

Makes sense.

 > >+         current->state = TASK_RUNNING;
 > >+         remove_wait_queue(&Controller->CommandWaitQueue, &__wait);
 > 
 > and finish_wait()?
 > 
 > Same for ibmtr.c ?

Same comments as above.

 > Also, would these changes:
 > 
 > >diff -urNp --exclude-from=/home/davej/.exclude 
 > >linux-1060/include/linux/wait.h linux-1070/include/linux/wait.h
 > >--- linux-1060/include/linux/wait.h
 > >+++ linux-1070/include/linux/wait.h
 > 
 > Be better in a separate patch?

A split-up patchset would for sure make sense for committing upstream.
Though, at least each file touched here is a separate cset.

		Dave

-- 
http://www.codemonkey.org.uk