Re: [PATCH wpan-next] mac802154: Avoid superfluous endianness handling

[Miquel Raynal <miquel.raynal@bootlin.com>]

Hi Alexander,

aahringo@redhat.com wrote on Sun, 5 Feb 2023 20:37:45 -0500:

> Hi,
> 
> On Tue, Jan 31, 2023 at 6:03 AM Miquel Raynal <miquel.raynal@bootlin.com> wrote:
> >
> > Hi Alexander,
> >
> > aahringo@redhat.com wrote on Mon, 30 Jan 2023 11:41:20 -0500:
> >  
> > > Hi,
> > >
> > > On Mon, Jan 30, 2023 at 11:34 AM Stefan Schmidt
> > > <stefan@datenfreihafen.org> wrote:  
> > > >
> > > > Hello.
> > > >
> > > > On 30.01.23 16:43, Miquel Raynal wrote:  
> > > > > When compiling scan.c with C=1, Sparse complains with:
> > > > >
> > > > >     sparse:     expected unsigned short [usertype] val
> > > > >     sparse:     got restricted __le16 [usertype] pan_id
> > > > >     sparse: sparse: cast from restricted __le16
> > > > >
> > > > >     sparse:     expected unsigned long long [usertype] val
> > > > >     sparse:     got restricted __le64 [usertype] extended_addr
> > > > >     sparse: sparse: cast from restricted __le64
> > > > >
> > > > > The tool is right, both pan_id and extended_addr already are rightfully
> > > > > defined as being __le16 and __le64 on both sides of the operations and
> > > > > do not require extra endianness handling.
> > > > >
> > > > > Fixes: 3accf4762734 ("mac802154: Handle basic beaconing")
> > > > > Reported-by: kernel test robot <lkp@intel.com>
> > > > > Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
> > > > > ---
> > > > >   net/mac802154/scan.c | 4 ++--
> > > > >   1 file changed, 2 insertions(+), 2 deletions(-)
> > > > >
> > > > > diff --git a/net/mac802154/scan.c b/net/mac802154/scan.c
> > > > > index cfbe20b1ec5e..8f98efec7753 100644
> > > > > --- a/net/mac802154/scan.c
> > > > > +++ b/net/mac802154/scan.c
> > > > > @@ -419,8 +419,8 @@ int mac802154_send_beacons_locked(struct ieee802154_sub_if_data *sdata,
> > > > >       local->beacon.mhr.fc.source_addr_mode = IEEE802154_EXTENDED_ADDRESSING;
> > > > >       atomic_set(&request->wpan_dev->bsn, -1);
> > > > >       local->beacon.mhr.source.mode = IEEE802154_ADDR_LONG;
> > > > > -     local->beacon.mhr.source.pan_id = cpu_to_le16(request->wpan_dev->pan_id);
> > > > > -     local->beacon.mhr.source.extended_addr = cpu_to_le64(request->wpan_dev->extended_addr);
> > > > > +     local->beacon.mhr.source.pan_id = request->wpan_dev->pan_id;
> > > > > +     local->beacon.mhr.source.extended_addr = request->wpan_dev->extended_addr;
> > > > >       local->beacon.mac_pl.beacon_order = request->interval;
> > > > >       local->beacon.mac_pl.superframe_order = request->interval;
> > > > >       local->beacon.mac_pl.final_cap_slot = 0xf;  
> > > >
> > > > This patch has been applied to the wpan-next tree and will be
> > > > part of the next pull request to net-next. Thanks!  
> > >
> > > fyi: in my opinion, depending on system endianness this is actually a bug.  
> >
> > Actually there are many uses of __le16 and __le64 for PAN IDs, short
> > and extended addresses. I did follow the existing patterns, I think
> > they are legitimate. Can you clarify what you think is a bug in the
> > current state? I always feel a bit flaky when it comes to properly
> > handling endianness, so all feedback welcome, if you have any hints
> > of what should be fixed after this patch, I'll do it.
> >  
> 
> net/ policy is so far I understood to always use endianness how it's
> stored on wire. There is no bug that addresses are stored as little
> endian, but there is a bug doing a conversion when it's not necessary.
> In this example, so far I see, big endian does a byteswap here.
> And as you figured there will be less byteswaps when it's stored as
> little endian. Consider this in netlink, that we store things as it is
> on wire (or in our case on air/wireless).

Ah ok I get it: "keep the endianness as on the medium". All right so
the __le16/__le64 variables are the right thing to do. Thanks for the
validation.

Thanks,
Miquèl