[net-next] net: dsa: mv88e6xxx: assert SMI lock

It's easy to forget to lock the smi_mutex before calling the low-level
_mv88e6xxx_reg_{read,write}, so add a assert_smi_lock function in them.

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
---
 drivers/net/dsa/mv88e6xxx.c | 31 ++++++++++++++++++-------------
 1 file changed, 18 insertions(+), 13 deletions(-)

Hi Andrew,

On Oct. Friday 30 (44) 09:41 PM, Andrew Lunn wrote:
> On Fri, Oct 30, 2015 at 04:35:42PM -0400, Vivien Didelot wrote:
> > It's easy to forget to lock the smi_mutex before calling the low-level
> > _mv88e6xxx_reg_{read,write}, so add a assert_smi_lock function in them.
> > 
> > Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
> > ---
> >  drivers/net/dsa/mv88e6xxx.c | 31 ++++++++++++++++++-------------
> >  1 file changed, 18 insertions(+), 13 deletions(-)
> > 
> > diff --git a/drivers/net/dsa/mv88e6xxx.c b/drivers/net/dsa/mv88e6xxx.c
> > index b1b14f5..70a0106 100644
> > --- a/drivers/net/dsa/mv88e6xxx.c
> > +++ b/drivers/net/dsa/mv88e6xxx.c
> > @@ -24,6 +24,16 @@
> >  #include <net/switchdev.h>
> >  #include "mv88e6xxx.h"
> >  
> > +static inline void assert_smi_lock(struct dsa_switch *ds)
> 
> No need for inline. Gcc will automatically inline it, if it thinks it
> is small enough.

OK, I will respin this without the inline keyword then.

> 
> > +{
> > +	struct mv88e6xxx_priv_state *ps = ds_to_priv(ds);
> > +
> > +	if (unlikely(!mutex_is_locked(&ps->smi_mutex))) {
> > +		dev_err(ds->master_dev, "SMI lock not held!\n");
> > +		dump_stack();
> > +	}
> > +}
> > +
> 
> > -/* Must be called with SMI mutex held */
> >  static int _mv88e6xxx_reg_read(struct dsa_switch *ds, int addr, int reg)
> >  {
> > -	struct mii_bus *bus = dsa_host_dev_to_mii_bus(ds->master_dev);
> > +	struct mii_bus *bus;
> >  	int ret;
> >  
> > +	assert_smi_lock(ds);
> > +
> > +	bus = dsa_host_dev_to_mii_bus(ds->master_dev);
> 
> Is this change of when bus is assigned actually required?

No, but I found not necessary to issue this "mdio_bus" lookup if the
lock is not held (see net/dsa/dsa.c:555). Do you prefer not to do that?

Also are you OK with removing all the "Must be called with..." comments,
which I found not necessary too (some function have this comment, some
others don't).

Thanks,
-v
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

> > >  static int _mv88e6xxx_reg_read(struct dsa_switch *ds, int addr, int reg)
> > >  {
> > > -	struct mii_bus *bus = dsa_host_dev_to_mii_bus(ds->master_dev);
> > > +	struct mii_bus *bus;
> > >  	int ret;
> > >  
> > > +	assert_smi_lock(ds);
> > > +
> > > +	bus = dsa_host_dev_to_mii_bus(ds->master_dev);
> > 
> > Is this change of when bus is assigned actually required?
> 
> No, but I found not necessary to issue this "mdio_bus" lookup if the
> lock is not held (see net/dsa/dsa.c:555). Do you prefer not to do that?

You are optimising for an error condition. If this optimisation saves
anything, it means we have a locking bug!

As a separate patch, i would do this lookup once in a setup function
and save it away in ps. We just need to watch out for the probe
register accesses.
 
> Also are you OK with removing all the "Must be called with..." comments,

Yes, it will become a lot more clear when the kernel outputs a stack dump!

     Andrew
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On Oct. Friday 30 (44) 10:01 PM, Andrew Lunn wrote:
> > > >  static int _mv88e6xxx_reg_read(struct dsa_switch *ds, int addr, int reg)
> > > >  {
> > > > -	struct mii_bus *bus = dsa_host_dev_to_mii_bus(ds->master_dev);
> > > > +	struct mii_bus *bus;
> > > >  	int ret;
> > > >  
> > > > +	assert_smi_lock(ds);
> > > > +
> > > > +	bus = dsa_host_dev_to_mii_bus(ds->master_dev);
> > > 
> > > Is this change of when bus is assigned actually required?
> > 
> > No, but I found not necessary to issue this "mdio_bus" lookup if the
> > lock is not held (see net/dsa/dsa.c:555). Do you prefer not to do that?
> 
> You are optimising for an error condition. If this optimisation saves
> anything, it means we have a locking bug!

Very good point, I'll get rid of that then.

> As a separate patch, i would do this lookup once in a setup function
> and save it away in ps. We just need to watch out for the probe
> register accesses.

Yes, I already put that on a todo list ;-)

> > Also are you OK with removing all the "Must be called with..." comments,
> 
> Yes, it will become a lot more clear when the kernel outputs a stack dump!

Good, sending a v2 right away then.

Thanks,
-v
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html