RE: [PATCH 3/3] net: cdc_ncm: Add ACPI MAC address pass through functionality

[<Mario.Limonciello@dell.com>]

> -----Original Message-----
> From: Hyde, Charles - Dell Team
> Sent: Tuesday, September 3, 2019 12:50 PM
> To: Oliver Neukum; lenb@kernel.org; rjw@rjwysocki.net
> Cc: Limonciello, Mario; nic_swsd@realtek.com; linux-acpi@vger.kernel.org; linux-
> usb@vger.kernel.org
> Subject: RE: [PATCH 3/3] net: cdc_ncm: Add ACPI MAC address pass through
> functionality
> 
> > > This change adds support to cdc_ncm for ACPI MAC address pass through
> > > functionality that also exists in the Realtek r8152 driver.  This is
> > > in support of Dell's Universal Dock D6000, to give it the same feature
> > > capability as is currently available in Windows and advertized on
> > > Dell's product web site.
> > >
> > > Signed-off-by: Charles Hyde <charles.hyde@dellteam.com>
> > > Cc: Mario Limonciello <mario.limonciello@dell.com>
> > > Cc: Oliver Neukum <oliver@neukum.org>
> > > Cc: Rafael J. Wysocki <rjw@rjwysocki.net>
> > > Cc: Len Brown <lenb@kernel.org>
> > > Cc: linux-usb@vger.kernel.org
> > > Cc: linux-acpi@vger.kernel.org
> > > ---
> > >  drivers/net/usb/cdc_ncm.c | 67
> > > +++++++++++++++++++++++++++++++++++++--
> > >  1 file changed, 64 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/drivers/net/usb/cdc_ncm.c b/drivers/net/usb/cdc_ncm.c
> > > index 85093579612f..11a04dc2298d 100644
> > > --- a/drivers/net/usb/cdc_ncm.c
> > > +++ b/drivers/net/usb/cdc_ncm.c
> > > @@ -52,6 +52,7 @@
> > >  #include <linux/usb/usbnet.h>
> > >  #include <linux/usb/cdc.h>
> > >  #include <linux/usb/cdc_ncm.h>
> > > +#include <acpi/acpi_mac_passthru.h>
> > >
> > >  #if IS_ENABLED(CONFIG_USB_NET_CDC_MBIM)
> > >  static bool prefer_mbim = true;
> > > @@ -984,11 +985,30 @@ int cdc_ncm_bind_common(struct usbnet *dev,
> > struct usb_interface *intf, u8 data_
> > >  	usb_set_intfdata(ctx->control, dev);
> > >
> > >  	if (ctx->ether_desc) {
> > > -		temp = usbnet_get_ethernet_addr(dev, ctx->ether_desc-
> > >iMACAddress);
> > > +		struct sockaddr sa;
> > > +
> > > +		temp = cdc_ncm_get_ethernet_address(dev, ctx);
> > >  		if (temp) {
> > >  			dev_dbg(&intf->dev, "failed to get mac address\n");
> > >  			goto error2;
> > >  		}
> > > +
> > > +		/* Check for a Dell Universal Dock D6000 before checking if
> > > +		 * ACPI supports MAC address pass through.
> > > +		 */
> > > +		if (!strstr(dev->udev->product, "D6000"))
> > > +			goto skip_acpi_mapt_in_bind;
> > > +
> > > +		if (get_acpi_mac_passthru(sa.sa_data) != 0)
> > > +			goto skip_acpi_mapt_in_bind;
> > > +
> > > +		if (memcmp(dev->net->dev_addr, sa.sa_data, ETH_ALEN) == 0)
> > > +			goto skip_acpi_mapt_in_bind;
> > > +
> > > +		if (cdc_ncm_set_ethernet_address(dev, &sa) == 0)
> > > +			memcpy(dev->net->dev_addr, sa.sa_data, ETH_ALEN);
> > > +
> > > +skip_acpi_mapt_in_bind:
> > >  		dev_info(&intf->dev, "MAC-Address: %pM\n", dev->net-
> > >dev_addr);
> > >  	}
> > >
> > > @@ -1716,6 +1736,47 @@ static void cdc_ncm_status(struct usbnet *dev,
> > struct urb *urb)
> > >  	}
> > >  }
> > >
> > > +static int cdc_ncm_resume(struct usb_interface *intf) {
> > > +	struct usbnet *dev = usb_get_intfdata(intf);
> > > +	struct cdc_ncm_ctx *ctx = (struct cdc_ncm_ctx *)dev->data[0];
> > > +	int ret;
> > > +
> > > +	ret = usbnet_resume(intf);
> > > +	if (ret != 0)
> > > +		goto error2;
> > > +
> > > +	if (ctx->ether_desc) {
> > > +		struct sockaddr sa;
> > > +
> > > +		if (cdc_ncm_get_ethernet_address(dev, ctx)) {
> > > +			dev_dbg(&intf->dev, "failed to get mac address\n");
> > > +			goto error2;
> > > +		}
> > > +
> > > +		/* Check for a Dell Universal Dock D6000 before checking if
> > > +		 * ACPI supports MAC address pass through.
> > > +		 */
> > > +		if (!strstr(dev->udev->product, "D6000"))
> > > +			goto skip_acpi_mapt_in_resume;
> >
> > This is too ugly. Use a flag for the need to restore the MAC.
> 
> It is ugly to me, also, because of restrictions that Linux source cannot be longer
> than 80 columns.  I much prefer readability and maintainability over 80 column
> line length limits.   Besides, I have not used an 80 column keypunch machine in
> more than 35 years.

This is kernel style though.
https://www.kernel.org/doc/html/v5.2/process/coding-style.html

> 
> > And have you consider the case somebody triggers a reset through usbfs? It
> > looks to me like you should encapsulate the restoration of the MAC and also
> > use it in post_reset()
> >
> > 	Regards
> > 		Oliver
> 
> Will there need to be a pre_reset() function to go with the post_reset() you are
> suggesting?

It should match implementing what 25766271e42f6b15b72ba156cb42a3fea91b5b21
did for r8152.