Skip to contentSkip to conversation

cdc_ncm kernel log spam with trendnet 2.5G USB adapter

PATCH Applied Roland Dreier 4 participants 18 messages Nov 16, 2020 – Jan 6, 2021 2 messages not in the archive

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

Hi, I recently got a 2.5G USB adapter, and although it works fine, the driver continually spams the kernel log with messages like

[127662.025647] cdc_ncm 4-1:2.0 enx3c8cf8f942e0: network connection: connected
[127662.057680] cdc_ncm 4-1:2.0 enx3c8cf8f942e0: 2500 mbit/s downlink
2500 mbit/s uplink
[127662.089794] cdc_ncm 4-1:2.0 enx3c8cf8f942e0: network connection: connected
[127662.121831] cdc_ncm 4-1:2.0 enx3c8cf8f942e0: 2500 mbit/s downlink
2500 mbit/s uplink
[127662.153858] cdc_ncm 4-1:2.0 enx3c8cf8f942e0: network connection: connected

...

Looking at the code in cdc_ncm.c it seems the device is just sending USB_CDC_NOTIFY_NETWORK_CONNECTION and USB_CDC_NOTIFY_SPEED_CHANGE urbs over and over, and the driver logs every single one.

Should we add code to the driver to keep track of what the last event was and only log if the state has really change?

Thanks!
Roland

Full details on the adapter, in case it matters:

Bus 004 Device 005: ID 20f4:e02b TRENDnet USB 10/100/1G/2.5G LAN Device Descriptor:
bLength 18
bDescriptorType 1
bcdUSB 3.20

  bDeviceClass            0
  bDeviceSubClass         0
  bDeviceProtocol         0
  bMaxPacketSize0         9
  idVendor           0x20f4 TRENDnet
  idProduct          0xe02b
  bcdDevice           30.04
  iManufacturer           1
  iProduct                2
  iSerial                 6
  bNumConfigurations      3

Configuration Descriptor:
bLength 9
bDescriptorType 2
wTotalLength 0x0039
bNumInterfaces 1
bConfigurationValue 1
iConfiguration 0
bmAttributes 0xa0
(Bus Powered)
Remote Wakeup
MaxPower 512mA
Interface Descriptor:
bLength 9
bDescriptorType 4
bInterfaceNumber 0
bAlternateSetting 0
bNumEndpoints 3
bInterfaceClass 255 Vendor Specific Class
bInterfaceSubClass 255 Vendor Specific Subclass
bInterfaceProtocol 0
iInterface 0
Endpoint Descriptor:
bLength 7
bDescriptorType 5
bEndpointAddress 0x81 EP 1 IN
bmAttributes 2
Transfer Type Bulk
Synch Type None
Usage Type Data
wMaxPacketSize 0x0400 1x 1024 bytes
bInterval 0
bMaxBurst 3
Endpoint Descriptor:
bLength 7
bDescriptorType 5
bEndpointAddress 0x02 EP 2 OUT
bmAttributes 2
Transfer Type Bulk
Synch Type None
Usage Type Data
wMaxPacketSize 0x0400 1x 1024 bytes
bInterval 0
bMaxBurst 3
Endpoint Descriptor:
bLength 7
bDescriptorType 5
bEndpointAddress 0x83 EP 3 IN
bmAttributes 3
Transfer Type Interrupt
Synch Type None
Usage Type Data
wMaxPacketSize 0x0002 1x 2 bytes
bInterval 11
bMaxBurst 0
Configuration Descriptor:
bLength 9
bDescriptorType 2
wTotalLength 0x0068
bNumInterfaces 2
bConfigurationValue 2
iConfiguration 0
bmAttributes 0xa0
(Bus Powered)
Remote Wakeup
MaxPower 512mA
Interface Descriptor:
bLength 9
bDescriptorType 4
bInterfaceNumber 0
bAlternateSetting 0
bNumEndpoints 1
bInterfaceClass 2 Communications
bInterfaceSubClass 13
bInterfaceProtocol 0
iInterface 5
CDC Header:
bcdCDC 1.10
CDC Union:
bMasterInterface 0
bSlaveInterface 1
CDC Ethernet:
iMacAddress 3 (??)
bmEthernetStatistics 0x0031500f
wMaxSegmentSize 1518
wNumberMCFilters 0x8000
bNumberPowerFilters 0
CDC NCM:
bcdNcmVersion 1.00
bmNetworkCapabilities 0x2b
8-byte ntb input size
max datagram size
net address
packet filter
Endpoint Descriptor:
bLength 7
bDescriptorType 5
bEndpointAddress 0x83 EP 3 IN
bmAttributes 3
Transfer Type Interrupt
Synch Type None
Usage Type Data
wMaxPacketSize 0x0010 1x 16 bytes
bInterval 8
bMaxBurst 0
Interface Descriptor:
bLength 9
bDescriptorType 4
bInterfaceNumber 1
bAlternateSetting 0
bNumEndpoints 0
bInterfaceClass 10 CDC Data
bInterfaceSubClass 0
bInterfaceProtocol 1
iInterface 0
Interface Descriptor:
bLength 9
bDescriptorType 4
bInterfaceNumber 1
bAlternateSetting 1
bNumEndpoints 2
bInterfaceClass 10 CDC Data
bInterfaceSubClass 0
bInterfaceProtocol 1
iInterface 4
Endpoint Descriptor:
bLength 7
bDescriptorType 5
bEndpointAddress 0x81 EP 1 IN
bmAttributes 2
Transfer Type Bulk
Synch Type None
Usage Type Data
wMaxPacketSize 0x0400 1x 1024 bytes
bInterval 0
bMaxBurst 3
Endpoint Descriptor:
bLength 7
bDescriptorType 5
bEndpointAddress 0x02 EP 2 OUT
bmAttributes 2
Transfer Type Bulk
Synch Type None
Usage Type Data
wMaxPacketSize 0x0400 1x 1024 bytes
bInterval 0
bMaxBurst 3
Configuration Descriptor:
bLength 9
bDescriptorType 2
wTotalLength 0x0062
bNumInterfaces 2
bConfigurationValue 3
iConfiguration 0
bmAttributes 0xa0
(Bus Powered)
Remote Wakeup
MaxPower 512mA
Interface Descriptor:
bLength 9
bDescriptorType 4
bInterfaceNumber 0
bAlternateSetting 0
bNumEndpoints 1
bInterfaceClass 2 Communications
bInterfaceSubClass 6 Ethernet Networking
bInterfaceProtocol 0
iInterface 5
CDC Header:
bcdCDC 1.10
CDC Union:
bMasterInterface 0
bSlaveInterface 1
CDC Ethernet:
iMacAddress 3 (??)
bmEthernetStatistics 0x0031500f
wMaxSegmentSize 1518
wNumberMCFilters 0x8000
bNumberPowerFilters 0
Endpoint Descriptor:
bLength 7
bDescriptorType 5
bEndpointAddress 0x83 EP 3 IN
bmAttributes 3
Transfer Type Interrupt
Synch Type None
Usage Type Data
wMaxPacketSize 0x0010 1x 16 bytes
bInterval 8
bMaxBurst 0
Interface Descriptor:
bLength 9
bDescriptorType 4
bInterfaceNumber 1
bAlternateSetting 0
bNumEndpoints 0
bInterfaceClass 10 CDC Data
bInterfaceSubClass 0
bInterfaceProtocol 0
iInterface 0
Interface Descriptor:
bLength 9
bDescriptorType 4
bInterfaceNumber 1
bAlternateSetting 1
bNumEndpoints 2
bInterfaceClass 10 CDC Data
bInterfaceSubClass 0
bInterfaceProtocol 0
iInterface 4
Endpoint Descriptor:
bLength 7
bDescriptorType 5
bEndpointAddress 0x81 EP 1 IN
bmAttributes 2
Transfer Type Bulk
Synch Type None
Usage Type Data
wMaxPacketSize 0x0400 1x 1024 bytes
bInterval 0
bMaxBurst 3
Endpoint Descriptor:
bLength 7
bDescriptorType 5
bEndpointAddress 0x02 EP 2 OUT
bmAttributes 2
Transfer Type Bulk
Synch Type None
Usage Type Data
wMaxPacketSize 0x0400 1x 1024 bytes
bInterval 0
bMaxBurst 3

maintainer
#1 lore raw

Am Montag, den 16.11.2020, 09:23 -0800 schrieb Roland Dreier:

Quoted 10 lines

Hi, I recently got a 2.5G USB adapter, and although it works fine, the driver continually spams the kernel log with messages like

[127662.025647] cdc_ncm 4-1:2.0 enx3c8cf8f942e0: network connection: connected
[127662.057680] cdc_ncm 4-1:2.0 enx3c8cf8f942e0: 2500 mbit/s downlink
2500 mbit/s uplink
[127662.089794] cdc_ncm 4-1:2.0 enx3c8cf8f942e0: network connection: connected
[127662.121831] cdc_ncm 4-1:2.0 enx3c8cf8f942e0: 2500 mbit/s downlink
2500 mbit/s uplink
[127662.153858] cdc_ncm 4-1:2.0 enx3c8cf8f942e0: network connection: connected
Quoted 8 lines

...

Looking at the code in cdc_ncm.c it seems the device is just sending USB_CDC_NOTIFY_NETWORK_CONNECTION and USB_CDC_NOTIFY_SPEED_CHANGE urbs over and over, and the driver logs every single one.

Should we add code to the driver to keep track of what the last event was and only log if the state has really change?

Hi,

in theory I suppose we could do that, but it would be kind of stupid. Events like these should be reported to higher layers and not logged at all.

Regards
Oliver

A message that is not in the archive (sent off-list or not yet received): [email protected]

#3 lore raw
Quoted 1 line

your criticism was valid. Could you test the attached patches?

Thanks - I will build a kernel and report back.

- R.

author Tested-by
#4 lore raw
In reply to a message not in the archive

(Apologies, trying one more time with a better mailer)

Sorry it took so long, but I finally got a chance to test the patches. They seem to work well, but they only get rid of the downlink / uplink speed spam - I still get the following filling my kernel log with a patched kernel:

  [   29.830383] cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected
  [   29.894359] cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected
  [   29.958601] cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected
  [   30.022473] cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected
  [   30.086548] cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected

with the below patch on top of your 3, then my kernel log is clean.

Please apply your patches plus my patch, and feel free to add

to the other three.

--- 8< ---------- 8< ---

Subject: [PATCH] CDC-NCM: remove "connected" log message

The cdc_ncm driver passes network connection notifications up to usbnet_link_change(), which is the right place for any logging. Remove the netdev_info() duplicating this from the driver itself.

This stops devices such as my "TRENDnet USB 10/100/1G/2.5G LAN" (ID 20f4:e02b) adapter from spamming the kernel log with

cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected

messages every 60 msec or so.

---
 drivers/net/usb/cdc_ncm.c | 3 ---
 1 file changed, 3 deletions(-)
drivers/net/usb/cdc_ncm.c +0 -3
index a45fcc44facf..50d3a4e6d445 100644
@@ -1850,9 +1850,6 @@ static void cdc_ncm_status(struct usbnet *dev, struct urb *urb)
 		 * USB_CDC_NOTIFY_NETWORK_CONNECTION notification shall be
 		 * sent by device after USB_CDC_NOTIFY_SPEED_CHANGE.
 		 */
-		netif_info(dev, link, dev->net,
-			   "network connection: %sconnected\n",
-			   !!event->wValue ? "" : "dis");
 		usbnet_link_change(dev, !!event->wValue, 0);
 		break;
 
Signature · 1 line
2.29.2
maintainer
#5 lore raw
Quoted 12 lines · On Sat, 19 Dec 2020 14:21:40 -0800 Roland Dreier wrote:

(Apologies, trying one more time with a better mailer)

Sorry it took so long, but I finally got a chance to test the patches. They seem to work well, but they only get rid of the downlink / uplink speed spam - I still get the following filling my kernel log with a patched kernel:

  [   29.830383] cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected
  [   29.894359] cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected
  [   29.958601] cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected
  [   30.022473] cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected
  [   30.086548] cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected
Quoted 7 lines

with the below patch on top of your 3, then my kernel log is clean.

Please apply your patches plus my patch, and feel free to add

to the other three.

Hi Ronald, thanks for the patch.

I'm not sure what the story here is but if this change is expected to get into the networking tree we'll need a fresh posting. This sort of scissored reply does not get into patchwork.

Quoted 23 lines

Subject: [PATCH] CDC-NCM: remove "connected" log message

The cdc_ncm driver passes network connection notifications up to usbnet_link_change(), which is the right place for any logging. Remove the netdev_info() duplicating this from the driver itself.

This stops devices such as my "TRENDnet USB 10/100/1G/2.5G LAN" (ID 20f4:e02b) adapter from spamming the kernel log with

cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected

messages every 60 msec or so.

---
 drivers/net/usb/cdc_ncm.c | 3 ---
 1 file changed, 3 deletions(-)
drivers/net/usb/cdc_ncm.c +0 -0
index a45fcc44facf..50d3a4e6d445 100644

drivers/net/usb/cdc_ncm.c static void cdc_ncm_status(struct usbnet *dev, struct urb *urb) · line 1855

Quoted 9 lines
drivers/net/usb/cdc_ncm.c +0 -3
@@ -1850,9 +1850,6 @@ static void cdc_ncm_status(struct usbnet *dev, struct urb *urb)
 		 * USB_CDC_NOTIFY_NETWORK_CONNECTION notification shall be
 		 * sent by device after USB_CDC_NOTIFY_SPEED_CHANGE.
 		 */
-		netif_info(dev, link, dev->net,
-			   "network connection: %sconnected\n",
-			   !!event->wValue ? "" : "dis");
 		usbnet_link_change(dev, !!event->wValue, 0);
 		break;
 

It sounds like you're getting tens of those messages a second, we can remove the message but the device is still generating spurious events, wasting CPU cycles. Was blocking those events deemed unfeasible?

#6 lore raw
Quoted 3 lines · Jakub Kicinski

I'm not sure what the story here is but if this change is expected to get into the networking tree we'll need a fresh posting. This sort of scissored reply does not get into patchwork.

OK, will resend. Too bad about patchwork, "git am" drops everything before scissors lines by default.

Quoted 3 lines

It sounds like you're getting tens of those messages a second, we can remove the message but the device is still generating spurious events, wasting CPU cycles. Was blocking those events deemed unfeasible?

I certainly don't know enough about the USB CDC class to know why the spurious messages are showing up or whether they could be suppressed without a fix in the adapter firmware. But even ~30 spurious messages per second doesn't seem so bad for a multi-gig adapter that might be handling 100,000 or more packets per second.

- R.

#7 lore raw

PATCH 1/1 CDC-NCM: remove "connected" log message

In reply to Jakub Kicinski

The cdc_ncm driver passes network connection notifications up to usbnet_link_change(), which is the right place for any logging. Remove the netdev_info() duplicating this from the driver itself.

This stops devices such as my "TRENDnet USB 10/100/1G/2.5G LAN" (ID 20f4:e02b) adapter from spamming the kernel log with

cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected

messages every 60 msec or so.

---
 drivers/net/usb/cdc_ncm.c | 3 ---
 1 file changed, 3 deletions(-)
drivers/net/usb/cdc_ncm.c +0 -3
index a45fcc44facf..50d3a4e6d445 100644
@@ -1850,9 +1850,6 @@ static void cdc_ncm_status(struct usbnet *dev, struct urb *urb)
 		 * USB_CDC_NOTIFY_NETWORK_CONNECTION notification shall be
 		 * sent by device after USB_CDC_NOTIFY_SPEED_CHANGE.
 		 */
-		netif_info(dev, link, dev->net,
-			   "network connection: %sconnected\n",
-			   !!event->wValue ? "" : "dis");
 		usbnet_link_change(dev, !!event->wValue, 0);
 		break;
 
Signature · 1 line
2.29.2
maintainer Reviewed-by on v1 1/1
#8 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

Quoted 33 lines · Roland Dreier

The cdc_ncm driver passes network connection notifications up to usbnet_link_change(), which is the right place for any logging. Remove the netdev_info() duplicating this from the driver itself.

This stops devices such as my "TRENDnet USB 10/100/1G/2.5G LAN" (ID 20f4:e02b) adapter from spamming the kernel log with

cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected

messages every 60 msec or so.

---
 drivers/net/usb/cdc_ncm.c | 3 ---
 1 file changed, 3 deletions(-)
drivers/net/usb/cdc_ncm.c +0 -3
index a45fcc44facf..50d3a4e6d445 100644
@@ -1850,9 +1850,6 @@ static void cdc_ncm_status(struct usbnet *dev, struct urb *urb)
 		 * USB_CDC_NOTIFY_NETWORK_CONNECTION notification shall be
 		 * sent by device after USB_CDC_NOTIFY_SPEED_CHANGE.
 		 */
-		netif_info(dev, link, dev->net,
-			   "network connection: %sconnected\n",
-			   !!event->wValue ? "" : "dis");
 		usbnet_link_change(dev, !!event->wValue, 0);
 		break;
 
Signature · 1 line
2.29.2
maintainer Applied
#9 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

Quoted 8 lines · On Thu, 24 Dec 2020 08:53:52 +0100 Greg KH wrote:
Quoted 7 lines · Roland Dreier

The cdc_ncm driver passes network connection notifications up to usbnet_link_change(), which is the right place for any logging. Remove the netdev_info() duplicating this from the driver itself.

This stops devices such as my "TRENDnet USB 10/100/1G/2.5G LAN" (ID 20f4:e02b) adapter from spamming the kernel log with

Quoted 7 lines
Quoted 5 lines

cdc_ncm 2-2:2.0 enp0s2u2c2: network connection: connected

messages every 60 msec or so.

Applied to net and queued for LTS, thanks!

#10 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

Quoted 1 line

Applied to net and queued for LTS, thanks!

Thanks - is Oliver's series of 3 patches that get rid of the other half of the log spam also on the way upstream?

- R.

maintainer
#11 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

Am Montag, den 28.12.2020, 23:56 -0800 schrieb Roland Dreier:

Quoted 4 lines
Quoted 1 line

Applied to net and queued for LTS, thanks!

Thanks - is Oliver's series of 3 patches that get rid of the other half of the log spam also on the way upstream?

Hi,

I looked at them again and found that there is a way to get the same effect that will make maintenance easier in the long run. Could I send them to you later this week for testing?

Regards
Oliver

A message that is not in the archive (sent off-list or not yet received): [email protected]

#13 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

Quoted 3 lines

I looked at them again and found that there is a way to get the same effect that will make maintenance easier in the long run. Could I send them to you later this week for testing?

Yes, please. I have a good test setup now so I can easily try out patches.

Thanks,
Roland

maintainer
#14 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

Am Dienstag, den 29.12.2020, 11:50 -0800 schrieb Roland Dreier:

Quoted 5 lines
Quoted 3 lines

I looked at them again and found that there is a way to get the same effect that will make maintenance easier in the long run. Could I send them to you later this week for testing?

Yes, please. I have a good test setup now so I can easily try out patches.

Thank you,

here we go.

Regards
Oliver

From 7dfb5f35933ebbe12076e41606b178fa7d8d2e7b Mon Sep 17 00:00:00 2001
From: Oliver Neukum <[email protected]>
Date: Tue, 1 Dec 2020 11:31:15 +0100
Subject: [PATCH 1/2] usbnet: add method for reporting speed without MDIO

The old method for reporting network speed upwards assumed that a device uses MDIO and uses the generic phy functions based on that.
Add a a primitive internal version not making the assumption reporting back directly what the status operations record.

---
 drivers/net/usb/usbnet.c   | 30 +++++++++++++++++++++++++++++-
 include/linux/usb/usbnet.h |  7 ++++++-
 2 files changed, 35 insertions(+), 2 deletions(-)
drivers/net/usb/usbnet.c +29 -1
index 1447da1d5729..bcd17f6d6de6 100644
@@ -944,7 +944,7 @@ EXPORT_SYMBOL_GPL(usbnet_open);
  * they'll probably want to use this base set.
  */
 
-int usbnet_get_link_ksettings(struct net_device *net,
+int usbnet_get_link_ksettings_mdio(struct net_device *net,
 			      struct ethtool_link_ksettings *cmd)
 {
 	struct usbnet *dev = netdev_priv(net);
@@ -956,6 +956,32 @@ int usbnet_get_link_ksettings(struct net_device *net,
 
 	return 0;
 }
+EXPORT_SYMBOL_GPL(usbnet_get_link_ksettings_mdio);
+
+int usbnet_get_link_ksettings(struct net_device *net,
+					struct ethtool_link_ksettings *cmd)
+{
+	struct usbnet *dev = netdev_priv(net);
+
+	/* the assumption that speed is equal on tx and rx
+	 * is deeply engrained into the networking layer.
+	 * For wireless stuff it is not true.
+	 * We assume that rxspeed matters more.
+	 */
+	if (dev->rxspeed != SPEED_UNKNOWN)
+		cmd->base.speed = dev->rxspeed / 1000000;
+	else if (dev->txspeed != SPEED_UNKNOWN)
+		cmd->base.speed = dev->txspeed / 1000000;
+	/* if a minidriver does not record speed we try to
+	 * fall back on MDIO
+	 */
+	else if (!dev->mii.mdio_read)
+		cmd->base.speed = SPEED_UNKNOWN;
+	else
+		mii_ethtool_get_link_ksettings(&dev->mii, cmd);
+
+	return 0;
+}
 EXPORT_SYMBOL_GPL(usbnet_get_link_ksettings);
 
 int usbnet_set_link_ksettings(struct net_device *net,
@@ -1661,6 +1687,8 @@ usbnet_probe (struct usb_interface *udev, const struct usb_device_id *prod)
 	dev->intf = udev;
 	dev->driver_info = info;
 	dev->driver_name = name;
+	dev->rxspeed = -1; /* unknown or handled by MII */
+	dev->txspeed = -1;
 
 	net->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 	if (!net->tstats)
include/linux/usb/usbnet.h +6 -1
index 88a7673894d5..f748c758f82a 100644
@@ -53,6 +53,8 @@ struct usbnet {
 	u32			hard_mtu;	/* count any extra framing */
 	size_t			rx_urb_size;	/* size for rx urbs */
 	struct mii_if_info	mii;
+	int			rxspeed;	/* if MII is not used */
+	int			txspeed;	/* if MII is not used */
 
 	/* various kinds of pending driver work */
 	struct sk_buff_head	rxq;
@@ -267,8 +269,11 @@ extern void usbnet_purge_paused_rxq(struct usbnet *);
 
 extern int usbnet_get_link_ksettings(struct net_device *net,
 				     struct ethtool_link_ksettings *cmd);
-extern int usbnet_set_link_ksettings(struct net_device *net,
+extern int usbnet_set_link_ksettings_mdio(struct net_device *net,
 				     const struct ethtool_link_ksettings *cmd);
+/* Legacy - to be used if you really need an error to be returned */
+extern int usbnet_set_link_ksettings(struct net_device *net,
+					const struct ethtool_link_ksettings *cmd);
 extern u32 usbnet_get_link(struct net_device *net);
 extern u32 usbnet_get_msglevel(struct net_device *);
 extern void usbnet_set_msglevel(struct net_device *, u32);
Signature · 1 line
2.26.2

From 447850c0e90aef5a8bb569d3888c76032a82c7c7 Mon Sep 17 00:00:00 2001
From: Oliver Neukum <[email protected]>
Date: Tue, 1 Dec 2020 11:33:38 +0100
Subject: [PATCH 2/2] CDC-NCM: record speed in status method

The driver has a status method for receiving speed updates. The framework, however, had support functions only for devices that reported their speed upon an explicit query over a MDIO interface.
CDC_NCM however gets direct notifications from the device. As new support functions have become available, we shall now record such notifications and tell the usbnet framework to make direct use of them without going through the PHY layer.

---
 drivers/net/usb/cdc_ncm.c | 17 ++---------------
 1 file changed, 2 insertions(+), 15 deletions(-)
drivers/net/usb/cdc_ncm.c +2 -15
index 2bac57d5e8d5..78cb3da8de0b 100644
@@ -1823,21 +1823,8 @@ cdc_ncm_speed_change(struct usbnet *dev,
 	uint32_t rx_speed = le32_to_cpu(data->DLBitRRate);
 	uint32_t tx_speed = le32_to_cpu(data->ULBitRate);
 
-	/*
-	 * Currently the USB-NET API does not support reporting the actual
-	 * device speed. Do print it instead.
-	 */
-	if ((tx_speed > 1000000) && (rx_speed > 1000000)) {
-		netif_info(dev, link, dev->net,
-			   "%u mbit/s downlink %u mbit/s uplink\n",
-			   (unsigned int)(rx_speed / 1000000U),
-			   (unsigned int)(tx_speed / 1000000U));
-	} else {
-		netif_info(dev, link, dev->net,
-			   "%u kbit/s downlink %u kbit/s uplink\n",
-			   (unsigned int)(rx_speed / 1000U),
-			   (unsigned int)(tx_speed / 1000U));
-	}
+	dev->rxspeed = rx_speed;
+	dev->txspeed = tx_speed;
 }
 
 static void cdc_ncm_status(struct usbnet *dev, struct urb *urb)
Signature · 1 line
2.26.2
#15 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

I haven't tried these patches yet but they don't look quite right to me. inlining the first 0001 patch:

Quoted 14 lines
drivers/net/usb/usbnet.c +1 -1
index 1447da1d5729..bcd17f6d6de6 100644
@@ -944,7 +944,7 @@ EXPORT_SYMBOL_GPL(usbnet_open);
  * they'll probably want to use this base set.
  */
 
-int usbnet_get_link_ksettings(struct net_device *net,
+int usbnet_get_link_ksettings_mdio(struct net_device *net,
                   struct ethtool_link_ksettings *cmd)
 {
     struct usbnet *dev = netdev_priv(net);

drivers/net/usb/usbnet.c int usbnet_get_link_ksettings(struct net_device *net, · line 959

Quoted 4 lines
drivers/net/usb/usbnet.c +1 -0
@@ -956,6 +956,32 @@ int usbnet_get_link_ksettings(struct net_device *net,
 
     return 0;
 }
+EXPORT_SYMBOL_GPL(usbnet_get_link_ksettings_mdio);

why keep and export the old function when it will have no callers?

Quoted 13 lines
drivers/net/usb/usbnet.c +13 -0
+int usbnet_get_link_ksettings(struct net_device *net,
+                    struct ethtool_link_ksettings *cmd)
+{
+    struct usbnet *dev = netdev_priv(net);
+
+    /* the assumption that speed is equal on tx and rx
+     * is deeply engrained into the networking layer.
+     * For wireless stuff it is not true.
+     * We assume that rxspeed matters more.
+     */
+    if (dev->rxspeed != SPEED_UNKNOWN)
+        cmd->base.speed = dev->rxspeed / 1000000;
+    else if (dev->txspeed != SPEED_UNKNOWN)

drivers/net/usb/usbnet.c int usbnet_get_link_ksettings(struct net_device *net, · line 983

Quoted 10 lines
drivers/net/usb/usbnet.c +10 -0
+        cmd->base.speed = dev->txspeed / 1000000;
+    /* if a minidriver does not record speed we try to
+     * fall back on MDIO
+     */
+    else if (!dev->mii.mdio_read)
+        cmd->base.speed = SPEED_UNKNOWN;
+    else
+        mii_ethtool_get_link_ksettings(&dev->mii, cmd);
+
+    return 0;

This is a change in behavior for every driver that doesn't set rxspeed / txspeed - the old get_link function would return EOPNOTSUPP if mdio_read isn't implemented, now we give SPEED_UNKNOWN with a successful return code.

drivers/net/usb/usbnet.c usbnet_probe (struct usb_interface *udev,

Quoted 1 line
drivers/net/usb/usbnet.c +0 -0
@@ -1661,6 +1687,8 @@ usbnet_probe (struct usb_interface *udev,

const struct usb_device_id *prod)

drivers/net/usb/usbnet.c usbnet_probe (struct usb_interface *udev, const struct usb_device_id *prod) · line 1691

Quoted 5 lines
drivers/net/usb/usbnet.c +2 -0
     dev->intf = udev;
     dev->driver_info = info;
     dev->driver_name = name;
+    dev->rxspeed = -1; /* unknown or handled by MII */
+    dev->txspeed = -1;

Minor nit: if we're going to test these against SPEED_UNKNOWN above, then I think it's clearer to initialize them to that value via the same constant.

Quoted 7 lines
include/linux/usb/usbnet.h +0 -0
index 88a7673894d5..f748c758f82a 100644
@@ -267,8 +269,11 @@ extern void usbnet_purge_paused_rxq(struct usbnet *);
 
 extern int usbnet_get_link_ksettings(struct net_device *net,

include/linux/usb/usbnet.h extern void usbnet_purge_paused_rxq(struct usbnet *); · line 279

Quoted 10 lines
include/linux/usb/usbnet.h +4 -1
                      struct ethtool_link_ksettings *cmd);
-extern int usbnet_set_link_ksettings(struct net_device *net,
+extern int usbnet_set_link_ksettings_mdio(struct net_device *net,
                      const struct ethtool_link_ksettings *cmd);
+/* Legacy - to be used if you really need an error to be returned */
+extern int usbnet_set_link_ksettings(struct net_device *net,
+                    const struct ethtool_link_ksettings *cmd);
 extern u32 usbnet_get_link(struct net_device *net);
 extern u32 usbnet_get_msglevel(struct net_device *);
 extern void usbnet_set_msglevel(struct net_device *, u32);

I think this was meant to be changing get_link, not set_link.

Also I don't understand the "Legacy" comment. Is that referring to the EOPNOTSUPP change I mentioned above? If so, wouldn't it be better to preserve the legacy behavior rather than changing the behavior of every usbnet driver all at once? Like make a new usbnet_get_link_ksettings_nonmdio and update only cdc_ncm to use it?

- R.

maintainer
#16 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

Am Donnerstag, den 31.12.2020, 10:51 -0800 schrieb Roland Dreier:

Quoted 2 lines

I haven't tried these patches yet but they don't look quite right to me. inlining the first 0001 patch:

OK, let's see.

Quoted 4 lines
Quoted 4 lines
drivers/net/usb/usbnet.c +0 -0
index 1447da1d5729..bcd17f6d6de6 100644

[...]

Quoted 3 lines
Quoted 1 line

+EXPORT_SYMBOL_GPL(usbnet_get_link_ksettings_mdio);

why keep and export the old function when it will have no callers?

But they will callers. Have a dozen drivers use them. The goal of this patch set is to not touch them.

Quoted 21 lines
Quoted 21 lines
drivers/net/usb/usbnet.c +20 -0
+int usbnet_get_link_ksettings(struct net_device *net,
+                    struct ethtool_link_ksettings *cmd)
+{
+    struct usbnet *dev = netdev_priv(net);
+
+    /* the assumption that speed is equal on tx and rx
+     * is deeply engrained into the networking layer.
+     * For wireless stuff it is not true.
+     * We assume that rxspeed matters more.
+     */
+    if (dev->rxspeed != SPEED_UNKNOWN)
+        cmd->base.speed = dev->rxspeed / 1000000;
+    else if (dev->txspeed != SPEED_UNKNOWN)
+        cmd->base.speed = dev->txspeed / 1000000;
+    /* if a minidriver does not record speed we try to
+     * fall back on MDIO
+     */
+    else if (!dev->mii.mdio_read)
+        cmd->base.speed = SPEED_UNKNOWN;
+    else
Quoted 8 lines

drivers/net/usb/usbnet.c int usbnet_get_link_ksettings(struct net_device *net, · line 983

Quoted 3 lines
drivers/net/usb/usbnet.c +3 -0
+        mii_ethtool_get_link_ksettings(&dev->mii, cmd);
+
+    return 0;

This is a change in behavior for every driver that doesn't set rxspeed / txspeed - the old get_link function would return EOPNOTSUPP if mdio_read isn't implemented, now we give SPEED_UNKNOWN with a successful return code.

Yes. This is a drawback. Yet the speed is unknown is it not?

Quoted 11 lines

drivers/net/usb/usbnet.c usbnet_probe (struct usb_interface *udev,

Quoted 1 line
drivers/net/usb/usbnet.c +0 -0
@@ -1661,6 +1687,8 @@ usbnet_probe (struct usb_interface *udev,

const struct usb_device_id *prod)

drivers/net/usb/usbnet.c usbnet_probe (struct usb_interface *udev, const struct usb_device_id *prod) · line 1691

Quoted 5 lines
drivers/net/usb/usbnet.c +2 -0
     dev->intf = udev;
     dev->driver_info = info;
     dev->driver_name = name;
+    dev->rxspeed = -1; /* unknown or handled by MII */
+    dev->txspeed = -1;

Minor nit: if we're going to test these against SPEED_UNKNOWN above, then I think it's clearer to initialize them to that value via the same constant.

Correct. The next iteration will do that.

Quoted 20 lines
Quoted 7 lines
include/linux/usb/usbnet.h +0 -0
index 88a7673894d5..f748c758f82a 100644
@@ -267,8 +269,11 @@ extern void usbnet_purge_paused_rxq(struct usbnet *);
 
 extern int usbnet_get_link_ksettings(struct net_device *net,

include/linux/usb/usbnet.h extern void usbnet_purge_paused_rxq(struct usbnet *); · line 279

Quoted 10 lines
include/linux/usb/usbnet.h +4 -1
                      struct ethtool_link_ksettings *cmd);
-extern int usbnet_set_link_ksettings(struct net_device *net,
+extern int usbnet_set_link_ksettings_mdio(struct net_device *net,
                      const struct ethtool_link_ksettings *cmd);
+/* Legacy - to be used if you really need an error to be returned */
+extern int usbnet_set_link_ksettings(struct net_device *net,
+                    const struct ethtool_link_ksettings *cmd);
 extern u32 usbnet_get_link(struct net_device *net);
 extern u32 usbnet_get_msglevel(struct net_device *);
 extern void usbnet_set_msglevel(struct net_device *, u32);

I think this was meant to be changing get_link, not set_link.

Quoted 2 lines

Also I don't understand the "Legacy" comment. Is that referring to the EOPNOTSUPP change I mentioned above? If so, wouldn't it be better

Yes.

Quoted 3 lines

to preserve the legacy behavior rather than changing the behavior of every usbnet driver all at once? Like make a new usbnet_get_link_ksettings_nonmdio and update only cdc_ncm to use it?

Then I would have to touch them all. The problem is that the MDIO stuff really is pretty much a layering violation. It should never have been default. But now it is.

Regards
Oliver

#17 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

Quoted 7 lines
Quoted 3 lines

to preserve the legacy behavior rather than changing the behavior of every usbnet driver all at once? Like make a new usbnet_get_link_ksettings_nonmdio and update only cdc_ncm to use it?

Then I would have to touch them all. The problem is that the MDIO stuff really is pretty much a layering violation. It should never have been default. But now it is.

I don't understand this. Your 0001 patch changes the behavior of usbnet_get_link_ksettings() and you have to touch all of the 8 drivers that use it if you don't want to change their behavior. If you keep the old usbnet_get_link_ksettings() and add usbnet_get_link_ksettings_nonmdio() then you can just update cdc_ncm to start with, and then gradually migrate other drivers. And eventually fix the layering violation and get rid of the legacy function when the whole transition is done.

- R.

maintainer
#18 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

Am Montag, den 04.01.2021, 11:13 -0800 schrieb Roland Dreier:

Quoted 8 lines
Quoted 7 lines
Quoted 3 lines

to preserve the legacy behavior rather than changing the behavior of every usbnet driver all at once? Like make a new usbnet_get_link_ksettings_nonmdio and update only cdc_ncm to use it?

Then I would have to touch them all. The problem is that the MDIO stuff really is pretty much a layering violation. It should never have been default. But now it is.

Quoted 8 lines

I don't understand this. Your 0001 patch changes the behavior of usbnet_get_link_ksettings() and you have to touch all of the 8 drivers that use it if you don't want to change their behavior. If you keep the old usbnet_get_link_ksettings() and add usbnet_get_link_ksettings_nonmdio() then you can just update cdc_ncm to start with, and then gradually migrate other drivers. And eventually fix the layering violation and get rid of the legacy function when the whole transition is done.

Hi,

now that you put it that way, I get the merit of what you are saying. Very well. I will submit the first set of patches.

May I add your "Tested-by"?

Regards
Oliver

author Tested-by
#19 lore raw

Re: [PATCH] CDC-NCM: remove "connected" log message

Quoted 4 lines

now that you put it that way, I get the merit of what you are saying. Very well. I will submit the first set of patches.

May I add your "Tested-by"?

Yes, absolutely: