[net,v2] drivers/net/wan/hdlc_fr: Add needed_headroom for PVC devices

PVC devices are virtual devices in this driver stacked on top of the
actual HDLC device. They are the devices normal users would use.
PVC devices have two types: normal PVC devices and Ethernet-emulating
PVC devices.

When transmitting data with PVC devices, the ndo_start_xmit function
will prepend a header of 4 or 10 bytes. Currently this driver requests
this headroom to be reserved for normal PVC devices by setting their
hard_header_len to 10. However, this does not work when these devices
are used with AF_PACKET/RAW sockets. Also, this driver does not request
this headroom for Ethernet-emulating PVC devices (but deals with this
problem by reallocating the skb when needed, which is not optimal).

This patch replaces hard_header_len with needed_headroom, and set
needed_headroom for Ethernet-emulating PVC devices, too. This makes
the driver to request headroom for all PVC devices in all cases.

Cc: Krzysztof Halasa <khc@pm.waw.pl>
Cc: Martin Schiller <ms@dev.tdt.de>
Signed-off-by: Xie He <xie.he.0141@gmail.com>
---

Change from v1:

English language fix for the commit message.

Changed "Ethernet-emulated" to "Ethernet-emulating" because the device
is emulating an Ethernet device, rather than being emulated by an
Ethernet device.

I'm sorry for my poor English.

---
 drivers/net/wan/hdlc_fr.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

On Wed,  2 Sep 2020 17:06:58 -0700 Xie He wrote:
> PVC devices are virtual devices in this driver stacked on top of the
> actual HDLC device. They are the devices normal users would use.
> PVC devices have two types: normal PVC devices and Ethernet-emulating
> PVC devices.
> 
> When transmitting data with PVC devices, the ndo_start_xmit function
> will prepend a header of 4 or 10 bytes. Currently this driver requests
> this headroom to be reserved for normal PVC devices by setting their
> hard_header_len to 10. However, this does not work when these devices
> are used with AF_PACKET/RAW sockets. Also, this driver does not request
> this headroom for Ethernet-emulating PVC devices (but deals with this
> problem by reallocating the skb when needed, which is not optimal).
> 
> This patch replaces hard_header_len with needed_headroom, and set
> needed_headroom for Ethernet-emulating PVC devices, too. This makes
> the driver to request headroom for all PVC devices in all cases.

Since this is a tunnel protocol on top of HDLC interfaces, and
hdlc_setup_dev() sets dev->hard_header_len = 16; should we actually 
set the needed_headroom to 10 + 16 = 26? I'm not clear on where/if 
hdlc devices actually prepend 16 bytes of header, though.

CC: Willem as he was reviewing your similar patch recently.

> diff --git a/drivers/net/wan/hdlc_fr.c b/drivers/net/wan/hdlc_fr.c
> index 9acad651ea1f..12b35404cd8e 100644
> --- a/drivers/net/wan/hdlc_fr.c
> +++ b/drivers/net/wan/hdlc_fr.c
> @@ -1041,7 +1041,7 @@ static void pvc_setup(struct net_device *dev)
>  {
>  	dev->type = ARPHRD_DLCI;
>  	dev->flags = IFF_POINTOPOINT;
> -	dev->hard_header_len = 10;
> +	dev->hard_header_len = 0;

Is there a need to set this to 0? Will it not be zero after allocation?

>  	dev->addr_len = 2;
>  	netif_keep_dst(dev);
>  }
> @@ -1093,6 +1093,7 @@ static int fr_add_pvc(struct net_device *frad, unsigned int dlci, int type)
>  	dev->mtu = HDLC_MAX_MTU;
>  	dev->min_mtu = 68;
>  	dev->max_mtu = HDLC_MAX_MTU;
> +	dev->needed_headroom = 10;
>  	dev->priv_flags |= IFF_NO_QUEUE;
>  	dev->ml_priv = pvc;
>

Thank you for your email, Jakub!

On Fri, Sep 4, 2020 at 3:14 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> Since this is a tunnel protocol on top of HDLC interfaces, and
> hdlc_setup_dev() sets dev->hard_header_len = 16; should we actually
> set the needed_headroom to 10 + 16 = 26? I'm not clear on where/if
> hdlc devices actually prepend 16 bytes of header, though.

The HDLC device is not actually prepending any header when it is used
with this driver. When the PVC device has prepended its header and
handed over the skb to the HDLC device, the HDLC device just hands it
over to the hardware driver for transmission without prepending any
header.

If we grep "header_ops" and "skb_push" in "hdlc.c" and "hdlc_fr.c", we
can see there is no "header_ops" implemented in these two files and
all "skb_push" happen in the PVC device in hdlc_fr.c.

For this reason, I have previously submitted a patch to change the
value of hard_header_len of the HDLC device from 16 to 0, because it
is not actually used.

See:
2b7bcd967a0f (drivers/net/wan/hdlc: Change the default of hard_header_len to 0)

> > diff --git a/drivers/net/wan/hdlc_fr.c b/drivers/net/wan/hdlc_fr.c
> > index 9acad651ea1f..12b35404cd8e 100644
> > --- a/drivers/net/wan/hdlc_fr.c
> > +++ b/drivers/net/wan/hdlc_fr.c
> > @@ -1041,7 +1041,7 @@ static void pvc_setup(struct net_device *dev)
> >  {
> >       dev->type = ARPHRD_DLCI;
> >       dev->flags = IFF_POINTOPOINT;
> > -     dev->hard_header_len = 10;
> > +     dev->hard_header_len = 0;
>
> Is there a need to set this to 0? Will it not be zero after allocation?

Oh. I understand your point. Theoretically we don't need to set it to
0 because it already has the default value of 0. I'm setting it to 0
only because I want to tell future developers that this value is
intentionally set to 0, and it is not carelessly missed out.

On Fri, Sep 4, 2020 at 9:36 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> Applied to net, thank you!

Thank you, Jakub!

On Sun, Sep 13, 2020 at 10:26 PM Krzysztof Hałasa <khalasa@piap.pl> wrote:
>
> Xie He <xie.he.0141@gmail.com> writes:
>
> > The HDLC device is not actually prepending any header when it is used
> > with this driver. When the PVC device has prepended its header and
> > handed over the skb to the HDLC device, the HDLC device just hands it
> > over to the hardware driver for transmission without prepending any
> > header.
>
> That's correct. IIRC:
> - Cisco and PPP modes add 4 bytes
> - Frame Relay adds 4 (specific protocols - mostly IPv4) or 10 (general
>   case) bytes. There is that pvcX->hdlcX transition which adds nothing
>   (the header is already in place when the packet leaves pvcX device).
> - Raw mode adds nothing (IPv4 only, though it could be modified for
>   both IPv4/v6 easily)
> - Ethernet (hdlc_raw_eth.c) adds normal Ethernet header.

Thank you for the nice summary!

I just realized that we can indeed support both IPv4/v6 in hdlc_raw.
Both IPv4 and v6 packets have a version field in the header, so we can
use it to distinguish v4 and v6 packets on receiving. We can introduce
it as a new feature some time.