Re: [PATCH 01/12] Use mutex instead of semaphore in driver core

["Dave Young" <hidave.darkstar@gmail.com>]

On Jan 2, 2008 1:18 PM, Greg KH <gregkh@suse.de> wrote:
> On Wed, Jan 02, 2008 at 08:54:53AM +0800, Dave Young wrote:
> > On Dec 30, 2007 1:07 AM, Greg KH <greg@kroah.com> wrote:
> > >
> > > On Sat, Dec 29, 2007 at 03:07:30PM +0800, Dave Young wrote:
> > > > On Dec 29, 2007 1:06 PM, Dave Young <hidave.darkstar@gmail.com> wrote:
> > > > >
> > > > > On Dec 29, 2007 12:42 PM, Greg KH <gregkh@suse.de> wrote:
> > > > > > On Sat, Dec 29, 2007 at 10:36:49AM +0800, Dave Young wrote:
> > > > > > > >
> > > > > > > The full boot dmesg with lockdep output is out, there's one warnings in it :
> > > > > >
> > > > > > Please fix that warning before the next repost of these patches (along
> > > > > > with fixing the problem of them not being able to be applied and
> > > > > > successfully built at every point in the series...)
> > > > > >
> > > > >
> > > > > Ok, thanks, I will fix them and repost.
> > > > >
> > > >
> > > > Hi,
> > > > After digging the code, I feel hard to fix the lockdep warning due to
> > > > some misterious relationship with usb.
> > > >
> > > > Could someone help on this? thanks.
> > >
> > > My question back to you is why are you doing this conversion?  Have you
> > > found that it is needed for something?  Are there problems in the -rt
> > > kernels that this conversion helps with?  Or is it just a janitorial
> > > "convert semaphore to mutex" type thing.
> >
> > Sorry for delay-reply because recently I have little free time.
> >
> > Mutex is lightweighter than semaphore, the device/class is used
> > heavily in kernel, so I think the convert would be worth.
>
> But is the usage of this semaphore in the class code really a problem?
> Has it been seen to cause issues anywhere?  Does it show up on any
> benchmarks as being something that really needs to be replaced?
>
> All of the places this is used should be on a "slow" code-path, and the
> semaphores themselves should very rarely ever have to block anyone under
> normal usages.
>
> Without any real problems being reported for this, I wouldn't worry
> about this, the effort involved is non-trivial as you are quickly
> finding out :)
>

Maybe not a big problem, but AFAIK convertion is better.
For Benchmarks, not yet.

OTOH, as you said the effort to do this is indeed a problem.

Regards
dave