Re: [PATCH v2] wwan: core: Support slicing in port TX flow of WWAN subsystem

[Loic Poulain]

On Wed, 9 Nov 2022 at 12:23, Haozhe Chang (常浩哲)
<Haozhe.Chang@xxxxxxxxxxxx> wrote:
>
> Hi Loic
>
> On Tue, 2022-11-08 at 13:14 +0100, Loic Poulain wrote:
> > Hi Haozhe,
> >
> > On Tue, 8 Nov 2022 at 11:54, <haozhe.chang@xxxxxxxxxxxx> wrote:
> > >
> > > From: haozhe chang <haozhe.chang@xxxxxxxxxxxx>
> > >
> > > wwan_port_fops_write inputs the SKB parameter to the TX callback of
> > > the WWAN device driver. However, the WWAN device (e.g., t7xx) may
> > > have an MTU less than the size of SKB, causing the TX buffer to be
> > > sliced and copied once more in the WWAN device driver.
> > >
> > > This patch implements the slicing in the WWAN subsystem and gives
> > > the WWAN devices driver the option to slice(by chunk) or not. By
> > > doing so, the additional memory copy is reduced.
> > >
> > > Meanwhile, this patch gives WWAN devices driver the option to
> > > reserve
> > > headroom in SKB for the device-specific metadata.
> > >
> > > Signed-off-by: haozhe chang <haozhe.chang@xxxxxxxxxxxx>
> > >
> > > ---
> > > Changes in v2
> > >   -send fragments to device driver by skb frag_list.
> > > ---
> > >  drivers/net/wwan/t7xx/t7xx_port_wwan.c | 42 ++++++++++-------
> > >  drivers/net/wwan/wwan_core.c           | 65 ++++++++++++++++++++
> > > ------
> > >  include/linux/wwan.h                   |  5 +-
> > >  3 files changed, 80 insertions(+), 32 deletions(-)
> > >
> > > diff --git a/drivers/net/wwan/t7xx/t7xx_port_wwan.c
> > > b/drivers/net/wwan/t7xx/t7xx_port_wwan.c
> > > index 33931bfd78fd..74fa58575d5a 100644
> > > --- a/drivers/net/wwan/t7xx/t7xx_port_wwan.c
> > > +++ b/drivers/net/wwan/t7xx/t7xx_port_wwan.c
> > > @@ -54,13 +54,13 @@ static void t7xx_port_ctrl_stop(struct
> > > wwan_port *port)
> >
> > [...]
> > >  static const struct wwan_port_ops wwan_ops = {
> > >         .start = t7xx_port_ctrl_start,
> > >         .stop = t7xx_port_ctrl_stop,
> > >         .tx = t7xx_port_ctrl_tx,
> > > +       .needed_headroom = t7xx_port_tx_headroom,
> > > +       .tx_chunk_len = t7xx_port_tx_chunk_len,
> >
> > Can you replace 'chunk' with 'frag' everywhere?
> >
> OK
> > >  };
> > >
> > >  static int t7xx_port_wwan_init(struct t7xx_port *port)
> > > diff --git a/drivers/net/wwan/wwan_core.c
> > > b/drivers/net/wwan/wwan_core.c
> > > index 62e9f7d6c9fe..ed78471f9e38 100644
> > > --- a/drivers/net/wwan/wwan_core.c
> > > +++ b/drivers/net/wwan/wwan_core.c
> > > @@ -20,7 +20,7 @@
> > >  #include <uapi/linux/wwan.h>
> > >
> > >  /* Maximum number of minors in use */
> > > -#define WWAN_MAX_MINORS                (1 << MINORBITS)
> > > +#define WWAN_MAX_MINORS                BIT(MINORBITS)
> > >
> > >  static DEFINE_MUTEX(wwan_register_lock); /* WWAN device
> > > create|remove lock */
> > >  static DEFINE_IDA(minors); /* minors for WWAN port chardevs */
> > > @@ -67,6 +67,8 @@ struct wwan_device {
> > >   * @rxq: Buffer inbound queue
> > >   * @waitqueue: The waitqueue for port fops (read/write/poll)
> > >   * @data_lock: Port specific data access serialization
> > > + * @headroom_len: SKB reserved headroom size
> > > + * @chunk_len: Chunk len to split packet
> > >   * @at_data: AT port specific data
> > >   */
> > >  struct wwan_port {
> > > @@ -79,6 +81,8 @@ struct wwan_port {
> > >         struct sk_buff_head rxq;
> > >         wait_queue_head_t waitqueue;
> > >         struct mutex data_lock; /* Port specific data access
> > > serialization */
> > > +       size_t headroom_len;
> > > +       size_t chunk_len;
> > >         union {
> > >                 struct {
> > >                         struct ktermios termios;
> > > @@ -550,8 +554,13 @@ static int wwan_port_op_start(struct wwan_port
> > > *port)
> > >         }
> > >
> > >         /* If port is already started, don't start again */
> > > -       if (!port->start_count)
> > > +       if (!port->start_count) {
> > >                 ret = port->ops->start(port);
> > > +               if (port->ops->tx_chunk_len)
> > > +                       port->chunk_len = port->ops-
> > > >tx_chunk_len(port);
> >
> > So, maybe frag len and headroom should be parameters of
> > wwan_create_port() instead of port ops, as we really need this info
> > only once.
> >
> If frag_len and headroom are added, wwan_create_port will have 6
> parameters, is it too much? And for similar requirements in the future,
> it may be difficult to add more parameters.

I think 6 is still fine, if we need more fields in the future we can
always have a port_config struct as a parameter instead.

>
> Is it a good solution to provide wwan_port_set_frag_len and
> wwan_port_set_headroom_len to the device driver? if so, the device
> driver has a chance to modify the wwan port's field after calling
> wwan_create_port.

Would be preferable to not have these values changeable at runtime.

> > > +               if (port->ops->needed_headroom)
> > > +                       port->headroom_len = port->ops-
> > > >needed_headroom(port);
> > > +       }
> > >
> > >         if (!ret)
> > >                 port->start_count++;
> > > @@ -698,30 +707,56 @@ static ssize_t wwan_port_fops_read(struct
> > > file *filp, char __user *buf,
> > >  static ssize_t wwan_port_fops_write(struct file *filp, const char
> > > __user *buf,
> > >                                     size_t count, loff_t *offp)
> > >  {
> > > +       size_t len, chunk_len, offset, allowed_chunk_len;
> > > +       struct sk_buff *skb, *head = NULL, *tail = NULL;
> > >         struct wwan_port *port = filp->private_data;
> > > -       struct sk_buff *skb;
> > >         int ret;
> > >
> > >         ret = wwan_wait_tx(port, !!(filp->f_flags & O_NONBLOCK));
> > >         if (ret)
> > >                 return ret;
> > >
> > > -       skb = alloc_skb(count, GFP_KERNEL);
> > > -       if (!skb)
> > > -               return -ENOMEM;
> > > +       allowed_chunk_len = port->chunk_len ? port->chunk_len :
> > > count;
> >
> > I would suggest making port->chunk_len (frag_len) always valid, by
> > setting it to -1 (MAX size_t) when creating a port without frag_len
> > requirement.
> >
> Ok, it will help to reduce some code.
> > > +       for (offset = 0; offset < count; offset += chunk_len) {
> > > +               chunk_len = min(count - offset, allowed_chunk_len);
> > > +               len = chunk_len + port->headroom_len;
> > > +               skb = alloc_skb(len, GFP_KERNEL);
> >
> > That works but would prefer a simpler solution like:
> > do {
> >     len = min(port->frag_len, remain);
> >     skb = alloc_skb(len + port->needed_headroom; GFP_KERNEL);
> >     [...]
> >     copy_from_user(skb_put(skb, len), buf + count - remain)
> > } while ((remain -= len));
> >
> May I know if the suggestion is because "while" is more efficient
> than  "for", or is it more readable?

It's more readable to me, but it's a subjective opinion here.

> > > +               if (!skb) {
> > > +                       ret = -ENOMEM;
> > > +                       goto freeskb;
> > > +               }
> > > +               skb_reserve(skb, port->headroom_len);
> > > +
> > > +               if (!head) {
> > > +                       head = skb;
> > > +               } else if (!tail) {
> > > +                       skb_shinfo(head)->frag_list = skb;
> > > +                       tail = skb;
> > > +               } else {
> > > +                       tail->next = skb;
> > > +                       tail = skb;
> > > +               }
> > >
> > > -       if (copy_from_user(skb_put(skb, count), buf, count)) {
> > > -               kfree_skb(skb);
> > > -               return -EFAULT;
> > > -       }
> > > +               if (copy_from_user(skb_put(skb, chunk_len), buf +
> > > offset, chunk_len)) {
> > > +                       ret = -EFAULT;
> > > +                       goto freeskb;
> > > +               }
> > >
> > > -       ret = wwan_port_op_tx(port, skb, !!(filp->f_flags &
> > > O_NONBLOCK));
> > > -       if (ret) {
> > > -               kfree_skb(skb);
> > > -               return ret;
> > > +               if (skb != head) {
> > > +                       head->data_len += skb->len;
> > > +                       head->len += skb->len;
> > > +                       head->truesize += skb->truesize;
> > > +               }
> > >         }
> > >
> > > -       return count;
> > > +       if (head) {
> >
> > How head can be null here?
> >
> if the parameter "count" is 0, the for loop will not be executed.

Ok, right (with do/while version it should not happen).
But maybe the !count case should be caught earlier, if even possible in fops.

Regards,
Loic