[Bluez-devel] Re: [UPDATE PATCH 18/39] net/core: use wait_event_timeout()

[Marcel Holtmann <marcel@holtmann.org>]

Hi Nish,

> > > Description: Use wait_event_timeout() instead of custom wait-queue code.
> > > The current code uses TASK_INTERRUPTIBLE but only cares about timing out and the
> > > wq event taking place (does not actively do anything in response to signals), so
> > > wait_event_timeout() should be ok.
> > >
> > > Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
> > >
> > >
> > > --- 2.6.11-rc1-kj-v/net/bluetooth/hidp/core.c 2005-01-15 16:55:44.000000000 -0800
> > > +++ 2.6.11-rc1-kj/net/bluetooth/hidp/core.c   2005-01-21 11:07:11.000000000 -0800
> > > @@ -36,6 +36,7 @@
> > >  #include <linux/ioctl.h>
> > >  #include <linux/file.h>
> > >  #include <linux/init.h>
> > > +#include <linux/wait.h>
> > >  #include <net/sock.h>
> > >
> > >  #include <linux/input.h>
> > > @@ -325,7 +326,6 @@ static int hidp_session(void *arg)
> > >       struct sk_buff *skb;
> > >       int vendor = 0x0000, product = 0x0000;
> > >       wait_queue_t ctrl_wait, intr_wait;
> > > -     unsigned long timeo = HZ;
> > >
> > >       BT_DBG("session %p", session);
> > >
> > > @@ -370,28 +370,14 @@ static int hidp_session(void *arg)
> > >
> > >       hidp_del_timer(session);
> > >
> > > -     if (intr_sk->sk_state != BT_CONNECTED) {
> > > -             init_waitqueue_entry(&ctrl_wait, current);
> > > -             add_wait_queue(ctrl_sk->sk_sleep, &ctrl_wait);
> > > -             while (timeo && ctrl_sk->sk_state != BT_CLOSED) {
> > > -                     set_current_state(TASK_INTERRUPTIBLE);
> > > -                     timeo = schedule_timeout(timeo);
> > > -             }
> > > -             set_current_state(TASK_RUNNING);
> > > -             remove_wait_queue(ctrl_sk->sk_sleep, &ctrl_wait);
> > > -             timeo = HZ;
> > > -     }
> > > +     if (intr_sk->sk_state != BT_CONNECTED)
> > > +             wait_event_timeout(*(ctrl_sk->sk_sleep),
> > > +                             (ctrl_sk->sk_state == BT_CLOSED), HZ);
> > 
> > and this does not make it look better. For me this is too much pointer
> > stuff here. I think the interface of wait_event_timeout() is wrong. It
> > should take a pointer to wait_queue_head_t.
> 
> Maybe; I'm not the designer of the interface, though, so I was just
> trying to use what exists. To be honest, this particular driver is the
> *only* one of a good number I've tried to convert that has this issue.
> Everywhere else, {add,remove}_wait_queue() is called like so:
> 
> {add,remove}_wait_queue(&wq, &wait);
> 
> just as it is called via prepare_to_wait() in wait_event(). core.c is
> the first driver I've come across that called the functions via:
> 
> {add,remove}_wait_queue(wq, &wait);
> 
> wait_event() is just a large macro which expands to the first form,
> given wq as a parameter, so I think it should be ok, as that form is
> the far more common one in the kernel currently.

we are using the wait queue from struct sock and sk_sleep is already a
pointer. Converting from and to a pointer and later to the struct again
to only match an API looks wrong to me and thus this API is wrong, but
this is only my thought.

Dave, any comments from you about this?

Regards

Marcel




-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel