svn commit: r192535 - head/sys/kern

[Attilio Rao]

2009/5/21 Kostik Belousov <kostikbel at gmail.com>:
> On Thu, May 21, 2009 at 07:05:17PM +0200, Attilio Rao wrote:
>> 2009/5/21 Kostik Belousov <kostikbel at gmail.com>:
>> > On Thu, May 21, 2009 at 09:23:15AM -0700, Scott Long wrote:
>> >> Kostik Belousov wrote:
>> >> >We do have the KPI for the callers that cannot drop the locks and need
>> >> >to do destroy_dev, destroy_dev_sched(9).
>> >>
>> >> Good to know, I'll look at destroy_dev_sched().  I'd rather not have to
>> >> roll my own decoupled version.  And I understand the argument about
>> >> destroy_dev being a drain point for the API.  However, what about
>> >> create_dev()?  Making that non-blocking would help a lot.
>> >
>> > create_dev() can be made non-blocking, and this is the first argument pro
>> > Attilio patch.
>> >
>> > From the quick look, all that is needed is to replace M_WAITOK with
>> > M_NOWAIT inside prep_cdevsw() and devfs_alloc(). Untested patch below.
>> >
>> > diff --git a/sys/fs/devfs/devfs_devs.c b/sys/fs/devfs/devfs_devs.c
>> > index 4041911..f470ee8 100644
>> > --- a/sys/fs/devfs/devfs_devs.c
>> > +++ b/sys/fs/devfs/devfs_devs.c
>> > @@ -120,7 +120,7 @@ devfs_alloc(void)
>> >        struct cdev *cdev;
>> >        struct timespec ts;
>> >
>> > -       cdp = malloc(sizeof *cdp, M_CDEVP, M_USE_RESERVE | M_ZERO | M_WAITOK);
>> > +       cdp = malloc(sizeof *cdp, M_CDEVP, M_USE_RESERVE | M_ZERO | M_NOWAIT);
>> >
>> >        cdp->cdp_dirents = &cdp->cdp_dirent0;
>> >        cdp->cdp_dirent0 = NULL;
>> > diff --git a/sys/kern/kern_conf.c b/sys/kern/kern_conf.c
>> > index 284f482..acdd44a 100644
>> > --- a/sys/kern/kern_conf.c
>> > +++ b/sys/kern/kern_conf.c
>> > @@ -559,7 +559,7 @@ prep_cdevsw(struct cdevsw *devsw)
>> >                return;
>> >        if (devsw->d_flags & D_NEEDGIANT) {
>> >                dev_unlock();
>> > -               dsw2 = malloc(sizeof *dsw2, M_DEVT, M_WAITOK);
>> > +               dsw2 = malloc(sizeof *dsw2, M_DEVT, M_NOWAIT);
>> >                dev_lock();
>> >        } else
>> >                dsw2 = NULL;
>>
>> You need to check return values here if it returns NULL.
>>
>> IMHO, having a non-sleepable version of destroy_dev(), create_dev()
>> and such would be ideal.
>> Ideally, we should resolve all the sleeping point and do the conversion.
>> I'm unable to check the code right now.
>
> Sure. Something like this.
>

At this point I wonder what's the purpose of maintaining the sleeping
version for such functions?

Attilio


-- 
Peace can only be achieved by understanding - A. Einstein