Re: [PATCH] Replace completions with semaphores

[Peter Zijlstra <peterz@infradead.org>]

On Sat, 2008-04-12 at 13:04 -0600, Matthew Wilcox wrote:
> On Sat, Apr 12, 2008 at 08:05:49PM +0200, Peter Zijlstra wrote:
> > On Sat, 2008-04-12 at 11:26 -0600, Matthew Wilcox wrote:
> > >  We can continue the
> > > discussion about eliminating one or the other API, but my opinion is that
> > > eliminating either is a mistake.  The choice of API indicates how the
> > > author thinks about the code, which is crucial for those reading the code.
> > 
> > Which is exactly the point, semaphores are rarely the right tool for the
> > job (I'm currently only aware of XFS that actually uses them as actual
> > semaphores, and the driver model that uses it to get around lockdep -
> > still need to come up with a good solution for that).
> 
> I guess you haven't grepped the tree recently then.  We have 11 places
> which call sema_init with a number other than 0 or 1 as the second
> argument:
> 
> ../fs/xfs/linux-2.6/sema.h:#define initnsema(sp, val, name)      sema_init(sp, val)
> ../drivers/net/mlx4/cmd.c:       sema_init(&priv->cmd.event_sem, priv->cmd.max_cmds);
> ../drivers/char/viotape.c:       sema_init(&reqSem, VIOTAPE_MAXREQ);
> ../drivers/infiniband/hw/mthca/mthca_cmd.c:      sema_init(&dev->cmd.event_sem, dev->cmd.max_cmds);
> ../drivers/acpi/osl.c:   sema_init(sem, initial_units);
> ../drivers/scsi/megaraid/megaraid_sas.c: sema_init(&instance->ioctl_sem, MEGASAS_INT_CMDS);
> ../drivers/scsi/megaraid/megaraid_mm.c:  sema_init(&adapter->kioc_semaphore, lld_adp->max_kioc);
> ../drivers/video/omap/hwa742.c:  sema_init(&hwa742.req_sema, i - IRQ_REQ_POOL_SIZE);
> ../drivers/video/omap/blizzard.c:        sema_init(&blizzard.req_sema, i - IRQ_REQ_POOL_SIZE);
> ../drivers/usb/misc/usblcd.c:    sema_init(&dev->limit_sem, USB_LCD_CONCURRENT_WRITES);
> ../drivers/usb/usb-skeleton.c:   sema_init(&dev->limit_sem, WRITES_IN_FLIGHT);
> 
> It's nice to know that nobody is using __DECLARE_SEMAPHORE_GENERIC or
> __SEMAPHORE_INITIALIZER at this time.  I'll delete those from my semaphore
> tree soon.

Luckily those are mostly drivers, but yeah - I haven't actually grepped
the tree I just know the few sem users I've ran into myself.

Still doesn't make it a good idea to have semaphores.