RE: [PATCH] hv: hv_fcopy: drop the obsolete message on transfer failure

[Dexuan Cui <decui@microsoft.com>]

> -----Original Message-----
> From: KY Srinivasan
> Sent: Friday, November 21, 2014 1:58 AM
> To: Dexuan Cui; gregkh@linuxfoundation.org; linux-kernel@vger.kernel.org;
> driverdev-devel@linuxdriverproject.org; olaf@aepfle.de;
> apw@canonical.com; jasowang@redhat.com
> Cc: Haiyang Zhang
> Subject: RE: [PATCH] hv: hv_fcopy: drop the obsolete message on transfer
> failure
> > -----Original Message-----
> > From: Dexuan Cui
> > Sent: Wednesday, November 19, 2014 11:48 PM
> > To: KY Srinivasan; gregkh@linuxfoundation.org; linux-
> > kernel@vger.kernel.org; driverdev-devel@linuxdriverproject.org;
> > olaf@aepfle.de; apw@canonical.com; jasowang@redhat.com
> > Cc: Haiyang Zhang
> > Subject: RE: [PATCH] hv: hv_fcopy: drop the obsolete message on transfer
> > failure
> >
> > > -----Original Message-----
> > > From: KY Srinivasan
> > > Sent: Thursday, November 20, 2014 6:59 AM
> > > > diff --git a/drivers/hv/hv_fcopy.c b/drivers/hv/hv_fcopy.c index
> > > > 23b2ce2..177122a 100644
> > > > --- a/drivers/hv/hv_fcopy.c
> > > > +++ b/drivers/hv/hv_fcopy.c
> > > > @@ -86,6 +86,15 @@ static void fcopy_work_func(struct
> work_struct
> > > > *dummy)
> > > >   * process the pending transaction.
> > > >   */
> > > >  fcopy_respond_to_host(HV_E_FAIL);
> > > > +
> > > > +/* In the case the user-space daemon crashes, hangs or is killed,
> > > > +we
> > > > + * need to down the semaphore, otherwise, after the daemon starts
> > > > next
> > > > + * time, the obsolete data in fcopy_transaction.message or
> > > > + * fcopy_transaction.fcopy_msg will be used immediately.
> > > > + */
> > > > +if (down_trylock(&fcopy_transaction.read_sema))
> > > > +pr_debug("FCP: failed to acquire the semaphore\n");
> > > > +
> > > >  }
> > >
> > > When the daemon is killed, we currently reset the state in the release
> > > function. Why can't we cleanup the semaphore state (initialize) here as
> > well.
> > >
> > > K. Y
> >
> > Hi KY,
> > 1) The down_trylock() here is necessary: the daemon can fail to respond
> in 5
> > seconds due to many reasons, e.g., the VM's CPU and I/O are too busy. In
> > this case, the daemon may become running later(NOTE: in this example,
> the
> > daemon is not killed), but from the host user's point of view, the
> PowerShell
> > copy-vmfile command has failed, so here we have to 'down' the
> semaphore
> > anyway, otherwise, the daemon can get obsolete data.
> >
> > 2) If we add a line
> > sema_init(&fcopy_transaction.read_sema, 0); in fcopy_release(), it seems
> > OK at a glance, but we have to handle the race
> > condition: the above down_trylock() and the sema_init() can, in theory,
> run
> > simultaneously on different virtual CPUs.  It's tricky to address this.
> >
> > 3) So I think we can reuse the same semaphore without an actually
> > unnecessary re-initialization. :-)
> 
> Agreed; you may want to get rid of the pr_debug() call though.
> 
> Thanks,
> 
> K. Y

The pr_debug() is added intentionally according to suggestion of 
Redhat's Vitaly Kuznetsov in the bugzilla:
https://bugzilla.redhat.com/show_bug.cgi?id=1162100#c5

The function is declared with__must_check in include/linux/semaphore.h:
extern int __must_check down_trylock(struct semaphore *sem);

Without checking the return value, we'll get these warning if the
"Kernel hacking" options are enabled:

drivers/hv/hv_fcopy.c: In function 'fcopy_work_func':
drivers/hv/hv_fcopy.c:95:2: warning: ignoring return value of 'down_trylock', declared with attribute warn_unused_result [-Wunused-result]
  (void)down_trylock(&fcopy_transaction.read_sema);
  ^

In practice, the message I add should be very rare since it's very unlikely to fail
to get the semaphore in this timeout case -- and in case this happens, it's actually
OK, because the driver has told the host user the PowerShell command should fail.

Thanks,
-- Dexuan