netdev - Re: [PATCH v5 09/11] net: ieee802154: atusb: Call _xmit

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Mon, 25 Apr 2022 15:16:44 +0200
From:   Miquel Raynal <miquel.raynal@...tlin.com>
To:     Alexander Aring <alex.aring@...il.com>
Cc:     Stefan Schmidt <stefan@...enfreihafen.org>,
        linux-wpan - ML <linux-wpan@...r.kernel.org>,
        "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        "open list:NETWORKING [GENERAL]" <netdev@...r.kernel.org>,
        David Girault <david.girault@...vo.com>,
        Romuald Despres <romuald.despres@...vo.com>,
        Frederic Blain <frederic.blain@...vo.com>,
        Nicolas Schodet <nico@...fr.eu.org>,
        Thomas Petazzoni <thomas.petazzoni@...tlin.com>
Subject: Re: [PATCH v5 09/11] net: ieee802154: atusb: Call _xmit_error()
 when a transmission fails

Hi Alexander,

alex.aring@...il.com wrote on Mon, 25 Apr 2022 09:05:49 -0400:

> Hi,
> 
> On Mon, Apr 25, 2022 at 8:35 AM Alexander Aring <alex.aring@...il.com> wrote:
> >
> > Hi,
> >
> > On Thu, Apr 7, 2022 at 4:06 AM Miquel Raynal <miquel.raynal@...tlin.com> wrote:  
> > >
> > > Hi Alexander,
> > >
> > > alex.aring@...il.com wrote on Wed, 6 Apr 2022 17:58:59 -0400:
> > >  
> > > > Hi,
> > > >
> > > > On Wed, Apr 6, 2022 at 11:34 AM Miquel Raynal <miquel.raynal@...tlin.com> wrote:  
> > > > >
> > > > > ieee802154_xmit_error() is the right helper to call when a transmission
> > > > > has failed. Let's use it instead of open-coding it.
> > > > >
> > > > > Signed-off-by: Miquel Raynal <miquel.raynal@...tlin.com>
> > > > > ---
> > > > >  drivers/net/ieee802154/atusb.c | 5 ++---
> > > > >  1 file changed, 2 insertions(+), 3 deletions(-)
> > > > >
> > > > > diff --git a/drivers/net/ieee802154/atusb.c b/drivers/net/ieee802154/atusb.c
> > > > > index f27a5f535808..d04db4d07a64 100644
> > > > > --- a/drivers/net/ieee802154/atusb.c
> > > > > +++ b/drivers/net/ieee802154/atusb.c
> > > > > @@ -271,9 +271,8 @@ static void atusb_tx_done(struct atusb *atusb, u8 seq)
> > > > >                  * unlikely case now that seq == expect is then true, but can
> > > > >                  * happen and fail with a tx_skb = NULL;
> > > > >                  */
> > > > > -               ieee802154_wake_queue(atusb->hw);
> > > > > -               if (atusb->tx_skb)
> > > > > -                       dev_kfree_skb_irq(atusb->tx_skb);
> > > > > +               ieee802154_xmit_error(atusb->hw, atusb->tx_skb,
> > > > > +                                     IEEE802154_SYSTEM_ERROR);  
> > > >
> > > > That should then call the xmit_error for ANY other reason which is not
> > > > 802.15.4 specific which is the bus_error() function?  
> > >
> > > I'll drop the bus error function so we can stick to a regular
> > > _xmit_error() call.
> > >  
> >
> > Okay, this error is only hardware related.
> >  
> > > Besides, we do not have any trac information nor any easy access to
> > > what failed exactly, so it's probably best anyway.  
> >
> > This is correct, Somebody needs to write support for it for atusb firmware. [0]
> > However some transceivers can't yet or will never (because lack of
> > functionality?) report such errors back... they will act a little bit
> > weird.
> >
> > However, this return value is a BIG step moving into the right
> > direction that other drivers can follow.
> >
> > I think for MLME ops we can definitely handle some transmit errors now
> > and retry so that we don't wait for anything when we know the
> > transceiver was never submitting.
> >  
> 
> s/submitting/transmitted/
> 
> I could more deeper into that topic:
> 
> 1.
> 
> In my opinion this result value was especially necessary for MLME-ops,
> for others which do not directly work with MAC... they provide an
> upper layer protocol if they want something reliable.
> 
> 2.
> 
> Later on we can do statistics like what was already going around in
> the linux-wpan community to have something like whatever dump to see
> all neighbors and see how many ack failures there, etc. Some people
> want to make some predictions about link quality here. The kernel
> should therefore only capture some stats and the $WHATEVER userspace
> capable netlink monitor daemon should make some heuristic by dumping
> those stats.

I like the idea of having a per-device dump of the stats. It would be
really straightforward to implement with the current scan
implementation that I am about to share. We already have a per PAN
structure (with information like ID, channel, last time it was seen,
strength, etc). We might improve this structure with counters for all
the common mac errors. Maybe an option to the "pans dump" command
(again, in the pipe) might be a good start to get all the stats, like
"pans dump [stats]". I'll keep this in mind.

> 3.
> 
> Sometimes even IP capable protocols (using 6LoWPAN) want to know if
> ack was received or not, as mentioned. But this required additional
> handling in the socket layers... I didn't look into that topic yet but
> I know wireless solved it because they have some similar problems.

I did not look at the upper layers yet, but that would indeed be a nice
use case of these MAC statuses.

Thanks,
Miquèl