Re: [PATCH 4/4] acpi: semaphore removal

[Daniel Walker <dwalker@mvista.com>]

On Tue, 2008-08-26 at 13:50 -0600, Matthew Wilcox wrote:
> On Tue, Aug 26, 2008 at 12:30:46PM -0700, Daniel Walker wrote:
> > On Tue, 2008-08-26 at 13:13 -0600, Matthew Wilcox wrote:
> > > On Tue, Aug 26, 2008 at 11:59:49AM -0700, Daniel Walker wrote:
> > > > The semaphore usage in ACPI is more like completions. The ASL
> > > 
> > > Huh?  They are semaphores.  They're not 'more like completions' at all.
> > 
> > You can clearly make a completion out of a semaphore, but we have a
> > completion API .. ACPI is using locked semaphores, and essentially
> > re-making completions with the semaphore API..
> 
> What makes you think that?
> 
> executer/excreate.c:    status = acpi_os_create_semaphore(ACPI_NO_UNIT_LIMIT, 0,
> executer/exsystem.c:        acpi_os_create_semaphore(ACPI_NO_UNIT_LIMIT, 0, &temp_semaphore);
> namespace/nsaccess.c:                                       acpi_os_create_semaphore(1, 0,
> osl.c:acpi_os_create_semaphore(u32 max_units, u32 initial_units, acpi_handle * handle)
> 
> All users set 'initial_units' to 0.
> 

We have from ACPI,

acpi_status
acpi_os_create_semaphore(u32 max_units, u32 initial_units, acpi_handle * handle)
{
        struct semaphore *sem = NULL;
...
        sema_init(sem, initial_units);



Then from semaphore.h,

#define init_MUTEX(sem)         sema_init(sem, 1)
#define init_MUTEX_LOCKED(sem)  sema_init(sem, 0)

So initial units of 0 means make it locked initially .. Initialize to 1
would be a regular unlocked mutex ..

Daniel