cdc acm drops packets silently

[Explorer001]

cdc acm drops packets silently

[nordicjm]

Have you tried #65112 ?

[jfischer-no]

I was testing on a nrf52840dk with the cdc_acm_echo sample.

[00:02:35.662,628] cdc_acm_echo: Received byte does not match expected byte! C0 != 80
[00:02:35.662,689] cdc_acm_echo: Sent byte does not match expected byte! C0 != 80

Which sample is it?

[Explorer001]

Its the cdc_acm sample: samples/subsys/usb/cdc_acm

[jfischer-no]

Its the cdc_acm sample: samples/subsys/usb/cdc_acm

The log output clearly shows that it is not the sample in the tree, please add your modifications to the bug description.

[Explorer001]

Added the modified interrupt handler

[jfischer-no]

No idea why git diff does not work for you, but assuming your changes are:

diff --git a/samples/subsys/usb/cdc_acm/src/main.c b/samples/subsys/usb/cdc_acm/src/main.c
index 91f7a8db7f7..012f1a6b893 100644
--- a/samples/subsys/usb/cdc_acm/src/main.c
+++ b/samples/subsys/usb/cdc_acm/src/main.c
@@ -108,6 +108,11 @@ static int enable_usb_device_next(void)
 }
 #endif /* IS_ENABLED(CONFIG_USB_DEVICE_STACK_NEXT) */
 
+uint8_t tx_byte = 0x00;
+uint8_t rx_byte = 0x00;
+bool rx_err = false;
+bool tx_err = false;
+
 static void interrupt_handler(const struct device *dev, void *user_data)
 {
        ARG_UNUSED(user_data);
@@ -132,6 +137,14 @@ static void interrupt_handler(const struct device *dev, void *user_data)
                                recv_len = 0;
                        };
 
+                       for (int i = 0; i < recv_len; i++) {
+                               if (!rx_err && (buffer[i] != rx_byte)) {
+                                       LOG_ERR("Received byte does not match expected byte! %02X != %02X", buffer[i], rx_byte);
+                                       rx_err = true;
+                               }
+                               rx_byte++;
+                       }
+
                        rb_len = ring_buf_put(&ringbuf, buffer, recv_len);
                        if (rb_len < recv_len) {
                                LOG_ERR("Drop %u bytes", recv_len - rb_len);
@@ -154,6 +167,14 @@ static void interrupt_handler(const struct device *dev, void *user_data)
                                continue;
                        }
 
+                       for (int i = 0; i < rb_len; i++) {
+                               if (!tx_err && (buffer[i] != tx_byte)) {
+                                       LOG_ERR("Sent byte does not match expected byte! %02X != %02X", buffer[i], tx_byte);
+                                       tx_err = true;
+                               }
+                               tx_byte++;
+                       }
+
                        if (rx_throttled) {
                                uart_irq_rx_enable(dev);
                                rx_throttled = false;

I cannot reproduce it on nRF52840, my steps are something like:

Python 3.11.9 (main, Apr 10 2024, 13:16:36) [GCC 13.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import serial
>>> dev = serial.Serial('/dev/ttyACM1', 115200, timeout=0.1, writeTimeout=0.1)
>>> data_out = serial.to_bytes(range(256))
>>> dev.writable()
True
>>> dev.write(data_out)
256
>>> dev.readable()
True
>>> data_in = dev.read(len(data_out))
>>> if data_out != data_in:
...     print("Error");
... 
>>> 

[Explorer001]

The changes look correct. It does not happen on the first packet. Try sending and receiving in a loop. On my system the byte drop happens after several minutes.

[jfischer-no]

The changes look correct. It does not happen on the first packet. Try sending and receiving in a loop. On my system the byte drop happens after several minutes.

Well, with example above I do not see any issue running in a loop for about 10 minutes.

It seems like the first 64 bytes are dropped on the usb receive side of the nrf52.

What is the exact name of the SoC or development board?

[Explorer001]

What is the exact name of the SoC or development board?

nRF21540-DK