USB: Remote NDIS USB Ethernet support #4999

finikorg · 2017-11-15T14:53:27Z

Add RNDIS protocol, WIP.

finikorg · 2017-11-15T14:55:13Z

@loicpoulain

loicpoulain · 2017-11-15T16:17:37Z

General notes since I have EEM in the pipe as well:

We systematically use an extra buffer to send data to usb, usually linerarization in a ep buffer. Maybe we could avoid this by making usb subsystem handling net_buf directly.
In my case I need to send multi-packet in a same transfer which does work today with stm32 driver only, we should align both driver. e.g. sending a packet greater than ep packet size is split in multiple packets but sent during the same usb transfer, this improves throughput and reduces latency. Don't know if RNDIS support/request this btw.

tbursztyka

Tiny comments. Overall looks fine to me

tbursztyka · 2017-11-16T15:23:58Z

subsys/usb/class/netusb/function_rndis.c

+	}
+
+	len = sys_le32_to_cpu(hdr->len);
+


you can remove that empty line, your if condition is testing the previous assignment.

tbursztyka · 2017-11-16T15:25:15Z

subsys/usb/class/netusb/function_rndis.c

+	    /*
+	     * Calculate additional offset since payload_offset is calculated
+	     * from the start of itself ;)
+	     */


Ouch this comment in the middle of the if condition makes it hard to read. move it before the if

This comment is related to part of the "if" condition, if I move it before we would not know where it relates to.

You could explain the whole if then. It's quite ugly to cut the condition in the middle

tbursztyka · 2017-11-16T15:38:34Z

subsys/usb/class/netusb/function_rndis.c

+	}
+}
+
+struct tlv {


declare that structure at the beginning of the file.

tbursztyka · 2017-11-16T15:42:27Z

subsys/usb/class/netusb/function_rndis.c

+}
+
+static int rndis_class_handler(struct usb_setup_packet *setup, s32_t *len,
+			       u8_t **data)


wrong indent here it seems.

Looks OK in text editor

ok so it's github messing up again.

tbursztyka · 2017-11-16T15:42:33Z

subsys/usb/class/netusb/function_rndis.c

+		 */
+		queue_encapsulated_cmd(*data, *len);
+	} else if (setup->bRequest == CDC_GET_ENC_RSP &&
+		 REQTYPE_GET_DIR(setup->bmRequestType) == REQTYPE_DIR_TO_HOST) {


tbursztyka · 2017-11-16T15:44:51Z

subsys/usb/class/netusb/function_rndis.c

+				try_write(CONFIG_RNDIS_IN_EP_ADDR, zero,
+					  sizeof(zero));
+			}
+		}


you could move this code block to its own function.

tbursztyka · 2017-11-16T15:48:37Z

include/usb/class/usb_cdc.h

@@ -65,6 +65,8 @@
 * for ACM devices
 * @note PSTN120.pdf, 6.3, Table 13
 */
+#define CDC_SEND_ENC_CMD		0x00
+#define CDC_GET_ENC_RSP			0x01


Is github messing up the order of the commits again? I guess this commit comes before adding RNDIS

Yes, it comes before.

tbursztyka · 2017-11-16T15:50:26Z

subsys/usb/class/netusb/Kconfig

@@ -78,4 +78,41 @@ config USB_DEVICE_NETWORK_ECM_MAC

 endif # USB_DEVICE_NETWORK_ECM

+if USB_DEVICE_NETWORK_RNDIS


Merge that patch with "usb: rndis: Add Remote NDIS protocol handling"

lpereira · 2017-11-17T00:39:44Z

subsys/usb/class/netusb/function_rndis.c

+	u32_t rx_no_buf;
+
+	atomic_t notify_count;
+} rndis = {


Is there any possibility this struct can be accessed from different contexts, such as different threads or an interrupt context? notify_count being an atomic_t makes me think that this is the case. If so, this must be protected.

lpereira · 2017-11-17T00:40:53Z

subsys/usb/class/netusb/function_rndis.c

+	.skip_bytes = 0,
+};
+
+static u8_t manufacturer[] = { 't', 'e', 's', 't' };


Is "test" an acceptable hardcoded value? Can this be marked as const as well, since I don't see anything modifying it?

This is my left-overs from the old code base, now we use special CONFIG_USB_DEVICE_MANUFACTURER, I will fix it.

lpereira · 2017-11-17T00:45:22Z

subsys/usb/class/netusb/function_rndis.c

+	     * Calculate additional offset since payload_offset is calculated
+	     * from the start of itself ;)
+	     */
+	    (u8_t *)&hdr->payload_offset - (u8_t *)hdr) {


If I'm understanding correctly, payload_offset came directly from the wire and it has not been properly validated. This could lead to problems if the header is malformed, maliciously or not.

Also, the type of a pointer difference is ptrdiff_t, a signed integral type; len is an unsigned integral type; will this work in all occasions?

NO! This is offset of the payload_offset field from the start of the structure hdr.

The header is created in rndis_hdr_add() when sending packets, which is fine (it's correctly sent over the wire for the peer). However, this function is called right after receiving a command from USB -- and could be malformed. This value must be validated since it's treated like a pointer.

(Consider also that's undefined behavior in C to create a pointer outside of a valid memory region; one exception is to create a pointer value one element past the last on arrays.)

lpereira · 2017-11-17T00:48:11Z

subsys/usb/class/netusb/function_rndis.c

+{
+
+	if (!k_fifo_is_empty(&rndis_tx_queue)) {
+#if 0


Please remove commented-out code

lpereira · 2017-11-17T00:50:06Z

subsys/usb/class/netusb/function_rndis.c

+struct tlv {
+	u32_t type;
+	u32_t len;
+	u8_t data[0];


Do we need to support very old GCC compilers? Might be better to just declare data as u8_t data[], that is standard C99.

lpereira · 2017-11-17T01:04:38Z

subsys/usb/class/netusb/function_rndis.c

+	enum {
+		UNINITIALIZED,
+		INITIALIZED,
+	} state;


Will there be another state? Can this be a bool initialized instead?

at the moment I have 2 states, maybe there will be more in the future

lpereira · 2017-11-17T01:05:13Z

subsys/usb/class/netusb/function_rndis.c

+static K_THREAD_STACK_DEFINE(cmd_stack, 1024);
+static struct k_thread cmd_thread_data;
+
+static struct __rndis {


This struct can be better packed by reorganizing its members so that there are no holes due to alignment.

I am not sure we need to organize members of the structures by alignment rather then by logical order.

I agree with @lpereira here, we should align structs for minimizing memory consumption. The ordering of items might look funny but that is not very relevant. You could add more comments into this struct to explain more what the various integers like net_filter, ep_in, ep_notify are in this struct.

Actually I will remove most of fields since we have now new interface for netusb.

lpereira · 2017-11-17T01:08:07Z

subsys/usb/class/netusb/function_rndis.c

+
+	SYS_LOG_DBG("EP 0x%x status %d len %u", ep, ep_status, len);
+
+	usb_read(ep, buffer, len, NULL);


Is there any possibility of len being larger than CONFIG_RNDIS_BULK_EP_MPS?

It shouldn't but I have added check and limit read to MPS.

lpereira · 2017-11-17T01:16:49Z

subsys/usb/class/netusb/function_rndis.c

+		hdr_offset = sizeof(struct rndis_payload_packet);
+	}
+
+	ret = net_pkt_append_all(rndis.in_pkt, len - hdr_offset,


hdr_offset should be declared as an u32_t to avoid converting it from u8_t here

attributing to it the result of the sizeof() operator will truncate size_t to u8_t (as the code is written now); is this OK?

Can len be smaller than hdr_offset in a way that the unsigned subtraction will lead to overflow?

I have changed hdr_offset to u32_t though it cannot be so large since this is offset inside 64 bytes array.
hdr_offset is not zero if we get a new packet which starts with header of this offset size. In parse_rndis_header() we already have a check
if (buf_len < sizeof(*hdr)) {
SYS_LOG_ERR("Too small packet len %u", buf_len);
return -EINVAL;
}

net_pkt_append_all() can still be called if rndis.in_pkt is 0, so len must be checked here as well.

lpereira · 2017-11-17T01:17:25Z

subsys/usb/class/netusb/function_rndis.c

+	}
+
+	ret = net_pkt_append_all(rndis.in_pkt, len - hdr_offset,
+				 buffer + hdr_offset, K_FOREVER);


Is buffer + hdr_offset guaranteed to be within the boundaries of buffer?

Yes, hdr_offset is not zero only for new packet with header

lpereira · 2017-11-17T01:18:41Z

subsys/usb/class/netusb/function_rndis.c

+	rsp->max_packets = sys_cpu_to_le32(1);
+	rsp->max_transfer_size = sys_cpu_to_le32(rndis.mtu +
+						 sizeof(struct net_eth_hdr) +
+						 44 + 22);


What are these magic numbers?

lpereira · 2017-11-17T01:20:12Z

subsys/usb/class/netusb/function_rndis.c

+		break;
+	case RNDIS_OBJECT_ID_GEN_MAX_TOTAL_SIZE:
+		SYS_LOG_DBG("RNDIS_OBJECT_ID_GEN_MAX_TOTAL_SIZE");
+		net_buf_add_le32(buf, 1558);


What is this magic value?

This is maximum total packet length
https://docs.microsoft.com/en-us/windows-hardware/drivers/network/oid-gen-maximum-total-size

lpereira · 2017-11-17T01:22:01Z

subsys/usb/class/netusb/function_rndis.c

+static u8_t *rndis_hdr_add(u8_t *buf, u32_t len)
+{
+	struct rndis_payload_packet *hdr = (void *)buf;
+	u8_t offset = (u8_t *)&hdr->payload_offset - (u8_t *)hdr;


Beware of casting ptrdiff_t to u8_t.

Please use either ptrdiff_t here or at the very least u32_t. There's no reason to truncate a size to u8_t: even if the payload offset will never be larger than 255 bytes, this will generate code to convert from 8-bit to 32-bit wide registers. Due to stack alignment it's even possible that using u32_t won't even use more memory.

lpereira · 2017-11-17T01:22:30Z

subsys/usb/class/netusb/function_rndis.c

+
+static int rndis_send(struct net_pkt *pkt)
+{
+	u8_t buf[1500], *p; /* FIXME */


Shouldn't 1500 be a named constant that's used also by the rndis struct?

Fixed, I am planning also to refactor send() function similar to ECM one.

lpereira · 2017-11-17T01:24:01Z

subsys/usb/class/netusb/function_rndis.c

+		       net_pkt_ll_reserve(pkt) + pkt->frags->len);
+		p += net_pkt_ll_reserve(pkt) + pkt->frags->len;
+
+		for (frag = pkt->frags->frags; frag; frag = frag->frags) {


A chain of fragments can easily overflow buf here.

This looks like something that still needs to be addressed.

lpereira · 2017-11-17T01:26:21Z

I haven't finished reviewing this, but I'm sure I'll find more things if I keep poking around. Please review validation, pointer arithmetic, and type truncation (specially when pointer arithmetic is used; you may want to build with -Wconversion to catch these errors).

tbursztyka · 2017-11-17T08:22:35Z

subsys/usb/class/netusb/function_rndis.c

+
+	usb_read(ep, buffer, len, NULL);
+
+	/* We already keep frame keeping with len, warn here about


this comment would need to be rephrased.

finikorg · 2017-11-20T13:34:56Z

I addresses most of the comments, still need to address couple of issues, @lpereira how can I pass now with cmake -W parameters? I tried "cd outdir/ && make -Wconversion", no warnings ;)

lpereira · 2017-11-20T20:52:59Z

subsys/usb/class/netusb/function_rndis.c

+	     * Calculate additional offset since payload_offset is calculated
+	     * from the start of itself ;)
+	     */
+	    (u8_t *)&hdr->payload_offset - (u8_t *)hdr) {


The header is created in rndis_hdr_add() when sending packets, which is fine (it's correctly sent over the wire for the peer). However, this function is called right after receiving a command from USB -- and could be malformed. This value must be validated since it's treated like a pointer.

(Consider also that's undefined behavior in C to create a pointer outside of a valid memory region; one exception is to create a pointer value one element past the last on arrays.)

lpereira · 2017-11-20T20:56:32Z

subsys/usb/class/netusb/function_rndis.c

+		hdr_offset = sizeof(struct rndis_payload_packet);
+	}
+
+	ret = net_pkt_append_all(rndis.in_pkt, len - hdr_offset,


net_pkt_append_all() can still be called if rndis.in_pkt is 0, so len must be checked here as well.

lpereira · 2017-11-20T20:59:53Z

subsys/usb/class/netusb/function_rndis.c

+static u8_t *rndis_hdr_add(u8_t *buf, u32_t len)
+{
+	struct rndis_payload_packet *hdr = (void *)buf;
+	u8_t offset = (u8_t *)&hdr->payload_offset - (u8_t *)hdr;


Please use either ptrdiff_t here or at the very least u32_t. There's no reason to truncate a size to u8_t: even if the payload offset will never be larger than 255 bytes, this will generate code to convert from 8-bit to 32-bit wide registers. Due to stack alignment it's even possible that using u32_t won't even use more memory.

lpereira · 2017-11-20T21:01:55Z

subsys/usb/class/netusb/function_rndis.c

+		       net_pkt_ll_reserve(pkt) + pkt->frags->len);
+		p += net_pkt_ll_reserve(pkt) + pkt->frags->len;
+
+		for (frag = pkt->frags->frags; frag; frag = frag->frags) {


This looks like something that still needs to be addressed.

lpereira · 2017-11-20T21:04:18Z

subsys/usb/class/netusb/function_rndis.c

+	return buf + sizeof(*hdr);
+}
+
+static u8_t buf[RNDIS_GEN_MAX_TOTAL_SIZE];  /* FIXME */


buf is not a good name for a global variable; it's in fact very easy to shadow it locally. Please be more specific: send_buffer is preferable.

Also, is there a possibility that two threads will call rndis_send()?

In addition: what's that FIXME comment is referring to?

About "(u8_t *)&hdr->payload_offset - (u8_t *)hdr)" construction: those are pointers to struct hdr, they should not be validated!
Now for some reason I cannot reply to all github comments separately. So how net_pkt_append_all() can be called if in_pkt == NULL?
For other comments I am rewriting send() function so those issues would be fixed.

I also have a check:
if (buf_len < sizeof(*hdr)) {
SYS_LOG_ERR("Too small packet len %u", buf_len);
return -EINVAL;
}

lpereira · 2017-11-20T21:32:57Z

BTW, you should be able to pass CFLAGS as environment variables to CMake: CFLAGS="-Wconversion -Wall -Wextra -Werror" cmake ...

finikorg · 2017-11-21T15:30:58Z

Updated rndis_send() function and other fixes

finikorg · 2018-01-08T09:39:12Z

@lpereira have you had time to review the changes?

lpereira · 2018-01-13T01:14:41Z

subsys/usb/class/netusb/function_rndis.c

+	 */
+	if (len < sys_le32_to_cpu(hdr->payload_offset) +
+	    sys_le32_to_cpu(hdr->payload_len) +
+	    (u8_t *)&hdr->payload_offset - (u8_t *)hdr) {


To calculate this expression ((u8_t *)&hdr->payload_offset - (u8_t *)hdr), it's better to use offsetof(), as it's performed in compile time.

I will use this macro

lpereira · 2018-01-13T01:16:23Z

subsys/usb/class/netusb/function_rndis.c

+	 * Calculate additional offset since payload_offset is calculated
+	 * from the start of itself ;)
+	 */
+	if (len < sys_le32_to_cpu(hdr->payload_offset) +


Would there be any issue if payload_offset + payload_len - offsetof(header, payload_offset) overflowed? Unsigned overflow would mean that this would wrap around, creating a value that is smaller than len.

We get from this function only len, of course USB device could put anything to any field and we cannot always verify it, so in a case packet is broken it would be dropped by network interface.

@lpereira what do you think?

codecov-io · 2018-01-15T08:47:45Z

Codecov Report

Merging #4999 into master will not change coverage.
The diff coverage is n/a.

@@          Coverage Diff           @@
##           master   #4999   +/-   ##
======================================
  Coverage    51.3%   51.3%           
======================================
  Files         441     441           
  Lines       42259   42259           
  Branches     8061    8061           
======================================
  Hits        21680   21680           
  Misses      20060   20060           
  Partials      519     519

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5f5e163...77ebc2b. Read the comment docs.

Remove old unneeded debug statements. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Making configuration files useful. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

At the moment CONFIG_SYS_LOG_USB_LEVEL name does not specify that this is log level for the Device Stack. Make it clear renaming to the proper name. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Change hardcoded value to configuration defined. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Add missing CDC definitions for RNDIS. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Add init() callback needed for RNDIS and maybe others functions. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Add needed descriptors for USB Remote NDIS Ethernet Network. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Add implementation of Microsoft Remote NDIS USB Ethernet protocol. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Set default value for RNDIS. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Windows works better with this option. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Add configuration for zperf sample allowing to build USB dongle with zperf for testing. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Adding Networking USB Device Debug level allows to fine-tune logging. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Add check for rndis_set_handle() packet. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

Use macro offsetof() for calculating offset inside structure. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

lpereira · 2018-01-17T22:17:35Z

I'm approving the changes to move this forward; there might be other issues lingering around, though.

jli157 · 2020-05-25T17:30:27Z

@finikorg How have you done the tests on Windows? I found the latest Windows 10 has problem to support the device with this driver enabled. I can't manually set the IP address on the host side(Windows 10 always reports "it can't save the IP configs"). Should I report an issue for it given that the RNDIS driver is supposed to be well supported by Windows?

finikorg · 2020-05-26T05:45:32Z

@jli157 was working in Windows when I used it last time, depending on Windows version there might be issues of selecting right driver, IIRC "Compatible RNDIS adapter, etc". With MS OS Descriptors driver should be selected automatically.

jli157 · 2020-05-26T07:08:25Z

@jli157 was working in Windows when I used it last time, depending on Windows version there might be issues of selecting right driver, IIRC "Compatible RNDIS adapter, etc". With MS OS Descriptors driver should be selected automatically.

Windows 10 seems not to be able to find the driver automatically. I have to manually update it to either "compatible RNDIS adapter" or "USB RNDIS Adapter". However, the IP address on the host can't be changed on either one.

Did you mean to enable this one USB_DEVICE_OS_DESC so that Windows can choose the driver automatically?

jli157 · 2020-06-08T15:50:36Z

While using the driver with Ubuntu 18.04, I often found lots of errors like below:

[00:00:21.063,568] <err> usb_rndis: Not enough pkt buffers, len 118, skip 54
[00:00:21.073,822] <wrn> usb_rndis: Skip 54 bytes out of remaining 54 bytes
[00:00:21.084,014] <err> usb_rndis: Not enough pkt buffers, len 615, skip 551
[00:00:21.094,360] <wrn> usb_rndis: Skip 64 bytes out of remaining 551 bytes
[00:00:21.104,614] <wrn> usb_rndis: Skip 64 bytes out of remaining 487 bytes

And these:

[00:04:01.442,443] <err> usb_rndis: Error parsing RNDIS header
[00:04:01.451,477] <err> usb_rndis: Wrong data packet type 0x785bc0ef
[00:04:01.460,937] <err> usb_rndis: Error parsing RNDIS header
[00:04:01.469,909] <err> usb_rndis: Too small packet len 21
[00:04:01.478,515] <err> usb_rndis: Error parsing RNDIS header
[00:04:15.355,102] <err> usb_rndis: Error writing data to pkt: 0x2001f30c
[00:04:15.365,081] <err> usb_rndis: Wrong data packet type 0x11ff0040
[00:04:15.374,572] <err> usb_rndis: Error parsing RNDIS header

Any suggestions to fix these issues, increasing the RX buffer size?

finikorg requested review from andyross, jukkar and tbursztyka as code owners November 15, 2017 14:53

finikorg mentioned this pull request Nov 15, 2017

Usb: USB Device Stack fixes and improvements #4942

Closed

finikorg mentioned this pull request Nov 16, 2017

Implement Virtual USB ethernet Adapter support #2760

Closed

tbursztyka requested changes Nov 16, 2017

View reviewed changes

finikorg changed the title ~~USB: WIP: Remote NDIS USB Ethernet support~~ USB: Remote NDIS USB Ethernet support Nov 16, 2017

lpereira suggested changes Nov 17, 2017

View reviewed changes

lpereira reviewed Nov 17, 2017

View reviewed changes

tbursztyka reviewed Nov 17, 2017

View reviewed changes

finikorg force-pushed the rndis branch from 5023166 to c576382 Compare November 20, 2017 13:30

finikorg requested a review from youvedeep as a code owner November 20, 2017 13:30

lpereira suggested changes Nov 20, 2017

View reviewed changes

finikorg force-pushed the rndis branch from eca5445 to bd1125c Compare November 21, 2017 13:27

finikorg force-pushed the rndis branch from 854e711 to 3416f7c Compare November 22, 2017 14:29

galak added the area: USB Universal Serial Bus label Dec 6, 2017

finikorg force-pushed the rndis branch 2 times, most recently from 89d0622 to cc6e72a Compare December 18, 2017 10:39

finikorg requested a review from laperie December 18, 2017 10:40

lpereira reviewed Jan 13, 2018

View reviewed changes

finikorg added 14 commits January 17, 2018 16:48

usb: netusb: Remove unneeded debug

808a57c

Remove old unneeded debug statements. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

samples: Correct default SYS_LOG_USB and SYS_LOG_USB_DRIVER levels

149b13b

Making configuration files useful. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: Rename SYS_LOG_USB to SYS_LOG_USB_DEVICE

320ceeb

At the moment CONFIG_SYS_LOG_USB_LEVEL name does not specify that this is log level for the Device Stack. Make it clear renaming to the proper name. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: Use standard SYS_LOG_USB_DEVICE_LEVEL for log level

1cf9d10

Change hardcoded value to configuration defined. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: Add missing definitions needed for RNDIS

de6ee8b

Add missing CDC definitions for RNDIS. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: Add init() callback to network device functions

e168a5a

Add init() callback needed for RNDIS and maybe others functions. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: rndis: Add RNDIS device descriptors to the table

99d254c

Add needed descriptors for USB Remote NDIS Ethernet Network. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: rndis: Add Remote NDIS protocol handling

958d1aa

Add implementation of Microsoft Remote NDIS USB Ethernet protocol. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: Set Class Handler buffer size to 256 for RNDIS

926cc52

Set default value for RNDIS. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: rndis: Select USB_COMPOSITE_DEVICE for RNDIS

28dcc8d

Windows works better with this option. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

samples: zperf: Add USB RNDIS configuration check

fb7a187

Add configuration for zperf sample allowing to build USB dongle with zperf for testing. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: netusb: Add Networking logging debug level

0d097d6

Adding Networking USB Device Debug level allows to fine-tune logging. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: Add RNDIS packet boundary check

28ea075

Add check for rndis_set_handle() packet. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

usb: rndis: Use offsetof macro to ease offset calculations

77ebc2b

Use macro offsetof() for calculating offset inside structure. Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>

finikorg force-pushed the rndis branch from ad32929 to 77ebc2b Compare January 17, 2018 14:54

lpereira approved these changes Jan 17, 2018

View reviewed changes

nashif merged commit ca9c3ad into zephyrproject-rtos:master Jan 17, 2018

finikorg deleted the rndis branch February 15, 2018 08:22

		@@ -78,4 +78,41 @@ config USB_DEVICE_NETWORK_ECM_MAC

		endif # USB_DEVICE_NETWORK_ECM

		if USB_DEVICE_NETWORK_RNDIS


		SYS_LOG_DBG("EP 0x%x status %d len %u", ep, ep_status, len);

		usb_read(ep, buffer, len, NULL);


		usb_read(ep, buffer, len, NULL);

		/* We already keep frame keeping with len, warn here about

USB: Remote NDIS USB Ethernet support #4999

USB: Remote NDIS USB Ethernet support #4999

Conversation

finikorg commented Nov 15, 2017

finikorg commented Nov 15, 2017

loicpoulain commented Nov 15, 2017

tbursztyka left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lpereira Nov 17, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lpereira commented Nov 17, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

finikorg commented Nov 20, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lpereira commented Nov 20, 2017

finikorg commented Nov 21, 2017 • edited

finikorg commented Jan 8, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

codecov-io commented Jan 15, 2018 • edited

Codecov Report

lpereira commented Jan 17, 2018

jli157 commented May 25, 2020 • edited

finikorg commented May 26, 2020

lpereira Nov 17, 2017 •

edited

finikorg commented Nov 21, 2017 •

edited

codecov-io commented Jan 15, 2018 •

edited

jli157 commented May 25, 2020 •

edited

jli157 commented May 26, 2020 •

edited

jli157 commented Jun 8, 2020 •

edited