| 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
| ||
|
Message-ID: <CAGRyCJEekOwNwdtzMoW7LYGzDDcaoDdc-n5L+rJ9LgfbckFzXQ@mail.gmail.com>
Date: Thu, 12 Aug 2021 14:02:27 +0200
From: Daniele Palmas <dnlplm@...il.com>
To: Subash Abhinov Kasiviswanathan <subashab@...eaurora.org>
Cc: Bjørn Mork <bjorn@...k.no>,
Aleksander Morgado <aleksander@...ksander.es>,
Network Development <netdev@...r.kernel.org>,
Sean Tranchetti <stranche@...eaurora.org>
Subject: Re: RMNET QMAP data aggregation with size greater than 16384
Hi Subash,
Il giorno lun 9 ago 2021 alle ore 23:40 <subashab@...eaurora.org> ha scritto:
>
> > No need for () around a constant, is there?
> >
> > Either I'm blind, or you don't actuelly change the rx_urb_size for the
> > mux and pass through modes?
> >
>
> I seem to have missed this and the other stuff you have pointed out.
> Can you please review this update-
>
> >
> > I'd also prefer this to reset back to syncing with hard_mtu if/when
> > muxing or passthrough is disabled. Calling usbnet_change_mtu() won't
> > do
> > that. It doesn't touch rx_urb_size if it is different from hard_mtu.
> >
> > I also think that it might be useful to keep the mtu/hard_mtu control,
> > wouldn't it?
> >
> >
> > Something like
> >
> > old_rx_urb_size = dev->rx_urb_size;
> > if (mux|passthrough)
> > dev->rx_urb_size = MAP_DL_URB_SIZE;
> > else
> > dev->rx_urb_size = dev->hard_mtu;
> > if (dev->rx_urb_size > old_rx_urb_size)
> > unlink_urbs etc;
> > return usbnet_change_mtu(net, new_mtu);
> >
> > should do that, I think. Completely untested....
> >
> > And add it to the (!qmimux_has_slaves(dev)) cas in del_mux_store() too.
> >
>
> Assuming this patch doesn't have too many other issues, can I request
> Aleksander / Daniele to try this out.
>
I'm currently in vacation for a few weeks, so I can't test the change
at the moment, will try to do when I come back.
> diff --git a/drivers/net/usb/qmi_wwan.c b/drivers/net/usb/qmi_wwan.c
> index 6a2e4f8..4676544 100644
> --- a/drivers/net/usb/qmi_wwan.c
> +++ b/drivers/net/usb/qmi_wwan.c
> @@ -75,6 +75,8 @@ struct qmimux_priv {
> u8 mux_id;
> };
>
> +#define MAP_DL_URB_SIZE 32768
Just an heads-up that when I proposed that urb size there were doubts
about the value (see
https://patchwork.ozlabs.org/project/netdev/patch/20200909091302.20992-1-dnlplm@gmail.com/#2523774
and https://patchwork.ozlabs.org/project/netdev/patch/20200909091302.20992-1-dnlplm@gmail.com/#2523958):
it is true that this time you are setting the size just when qmap is
enabled, but I think that Carl's comment about low-cat chipsets could
still apply.
Thanks,
Daniele
> +
> static int qmimux_open(struct net_device *dev)
> {
> struct qmimux_priv *priv = netdev_priv(dev);
> @@ -303,6 +305,39 @@ static void qmimux_unregister_device(struct
> net_device *dev,
> dev_put(real_dev);
> }
>
> +static int qmi_wwan_change_mtu(struct net_device *net, int new_mtu)
> +{
> + struct usbnet *dev = netdev_priv(net);
> + struct qmi_wwan_state *info = (void *)&dev->data;
> + int old_rx_urb_size = dev->rx_urb_size;
> +
> + /* mux and pass through modes use a fixed rx_urb_size and the
> value
> + * is independent of mtu
> + */
> + if (info->flags & (QMI_WWAN_FLAG_MUX |
> QMI_WWAN_FLAG_PASS_THROUGH)) {
> + if (old_rx_urb_size == MAP_DL_URB_SIZE)
> + return 0;
> +
> + if (old_rx_urb_size < MAP_DL_URB_SIZE) {
> + dev->rx_urb_size = MAP_DL_URB_SIZE;
> +
> + usbnet_pause_rx(dev);
> + usbnet_unlink_rx_urbs(dev);
> + usbnet_resume_rx(dev);
> + usbnet_update_max_qlen(dev);
> + }
> +
> + return 0;
> + }
> +
> + /* rawip mode uses existing logic of setting rx_urb_size based
> on mtu.
> + * rx_urb_size will be updated within usbnet_change_mtu only if
> it is
> + * equal to existing hard_mtu
> + */
> + dev->rx_urb_size = dev->hard_mtu;
> + return usbnet_change_mtu(net, new_mtu);
> +}
> +
> static void qmi_wwan_netdev_setup(struct net_device *net)
> {
> struct usbnet *dev = netdev_priv(net);
> @@ -326,7 +361,7 @@ static void qmi_wwan_netdev_setup(struct net_device
> *net)
> }
>
> /* recalculate buffers after changing hard_header_len */
> - usbnet_change_mtu(net, net->mtu);
> + qmi_wwan_change_mtu(net, net->mtu);
> }
>
> static ssize_t raw_ip_show(struct device *d, struct device_attribute
> *attr, char *buf)
> @@ -433,6 +468,7 @@ static ssize_t add_mux_store(struct device *d,
> struct device_attribute *attr, c
> if (!ret) {
> info->flags |= QMI_WWAN_FLAG_MUX;
> ret = len;
> + qmi_wwan_change_mtu(dev->net, dev->net->mtu);
> }
> err:
> rtnl_unlock();
> @@ -466,8 +502,11 @@ static ssize_t del_mux_store(struct device *d,
> struct device_attribute *attr, c
> }
> qmimux_unregister_device(del_dev, NULL);
>
> - if (!qmimux_has_slaves(dev))
> + if (!qmimux_has_slaves(dev)) {
> info->flags &= ~QMI_WWAN_FLAG_MUX;
> + qmi_wwan_change_mtu(dev->net, dev->net->mtu);
> + }
> +
> ret = len;
> err:
> rtnl_unlock();
> @@ -514,6 +553,8 @@ static ssize_t pass_through_store(struct device *d,
> else
> info->flags &= ~QMI_WWAN_FLAG_PASS_THROUGH;
>
> + qmi_wwan_change_mtu(dev->net, dev->net->mtu);
> +
> return len;
> }
>
> @@ -643,7 +684,7 @@ static const struct net_device_ops
> qmi_wwan_netdev_ops = {
> .ndo_stop = usbnet_stop,
> .ndo_start_xmit = usbnet_start_xmit,
> .ndo_tx_timeout = usbnet_tx_timeout,
> - .ndo_change_mtu = usbnet_change_mtu,
> + .ndo_change_mtu = qmi_wwan_change_mtu,
> .ndo_get_stats64 = dev_get_tstats64,
> .ndo_set_mac_address = qmi_wwan_mac_addr,
> .ndo_validate_addr = eth_validate_addr,
Powered by blists - more mailing lists