Help understanding USB CDC ACM read flow

RE: Help understanding USB CDC ACM read flow

kv1980 — Sun, 14 Mar 2021 20:37:56 GMT

Ahhhh! Once I see it I realize I'm dumb :-P It's all so clear now. Thank you so much Dmitry!

RE: Help understanding USB CDC ACM read flow

Dmitry — Sun, 14 Mar 2021 19:38:42 GMT

[quote userid="95738" url="~/f/nordic-q-a/72752/help-understanding-usb-cdc-acm-read-flow/299618#299618"]but the end of the do loop in the handler is while (ret == NRF_SUCCESS) so how can I call app_usbd_cdc_acm_write if I am stuck in the loop?[/quote]

The code should look something like this:

            do
            {
                /* here the next 64 bytes are in the buffer */
                app_usbd_cdc_acm_write(p_cdc_acm, m_rx_buffer[m_cdc_buffer_idx].buffer, READ_SIZE);
                if (++m_cdc_buffer_idx > ARRAY_LIST_SIZE) m_cdc_buffer_idx = 0;
                
                /* Fetch data until internal buffer is empty */
                ret = app_usbd_cdc_acm_read(p_cdc_acm,
                                            m_rx_buffer[m_cdc_buffer_idx].buffer,
                                            READ_SIZE);
            } while (ret == NRF_SUCCESS);

            /* last app_usbd_cdc_acm_read() returned IO_PENDING */

[quote userid="95738" url="~/f/nordic-q-a/72752/help-understanding-usb-cdc-acm-read-flow/299618#299618"]Sorry, but does this mean I just make one 64 byte buffer and just keep pointing it back into the app_usbd_cdc_acm_read? What if I want to do some processing on it?[/quote]

What I mean - the buffer is not touched by USBD until you call app_usbd_cdc_acm_read() again - you can process it directly in APP_USBD_CDC_ACM_USER_EVT_RX_DONE handler, while the next data are received into the internal buffer. Of course, if circular buffer is more convenient in your case, it's a good decision.

RE: Help understanding USB CDC ACM read flow

kv1980 — Sun, 14 Mar 2021 19:18:23 GMT

Thanks Dmitry, I think I'm getting there. Just two little things I can't quite wrap my brain around. You say

[quote userid="5418" url="~/f/nordic-q-a/72752/help-understanding-usb-cdc-acm-read-flow/299617#299617"]If the result of next app_usbd_cdc_acm_read() is NRF_SUCCESS, you can call app_usbd_cdc_acm_write() again.[/quote]

but the end of the do loop in the handler is while (ret == NRF_SUCCESS) so how can I call app_usbd_cdc_acm_write if I am stuck in the loop?

Also,

[quote userid="5418" url="~/f/nordic-q-a/72752/help-understanding-usb-cdc-acm-read-flow/299617#299617"]CDC ACM has two rx buffers internally, you don't need to switch buffers in your code.[/quote]

Sorry, but does this mean I just make one 64 byte buffer and just keep pointing it back into the app_usbd_cdc_acm_read? What if I want to do some processing on it?

RE: Help understanding USB CDC ACM read flow

Dmitry — Sun, 14 Mar 2021 19:04:21 GMT

Hi,

the documentation for CDC ACM module explains these things quite clear.

In APP_USBD_CDC_ACM_USER_EVT_PORT_OPEN handler, you're preparing to receive 64 bytes into the buffer. There is still no data, so you shouldn't call app_usbd_cdc_acm_write() here - it will send some garbage.

When APP_USBD_CDC_ACM_USER_EVT_RX_DONE comes, you have 64 bytes ready in your buffer - here you should call app_usbd_cdc_acm_write(). Then in the loop, you're reading next chunks that may already have been received by USB peripheral. If the result of next app_usbd_cdc_acm_read() is NRF_SUCCESS, you can call app_usbd_cdc_acm_write() again. Otherwise it will return NRF_ERROR_IO_PENDING - data is not yet ready, but app_usbd_cdc_acm_read() prepares USBD to receive next 64 bytes - the following APP_USBD_CDC_ACM_USER_EVT_RX_DONE will have the buffer filled in.

CDC ACM has two rx buffers internally, you don't need to switch buffers in your code.

[quote userid="95738" url="~/f/nordic-q-a/72752/help-understanding-usb-cdc-acm-read-flow/299610#299610"]How do other people process variable length data from USB CDC ACM?[/quote]

Eother by reading one byte at a time, or with app_usbd_cdc_acm_read_any function.

RE: Help understanding USB CDC ACM read flow

kv1980 — Sun, 14 Mar 2021 14:18:41 GMT

After a little more testing, I think I am narrowing down the my question a little. Here's the sequence:

On PORT_OPEN, the first read is executed, the buffer is empty (since no data has been sent), and the index of the buffer is incremented by one, so m_cdc_buffer_idx = 1.
When I send the 662 bytes of data (this is the header for a standard WAV file) and I set a break, I see the beginning of the data get read into m_rx_buffer[0], and subsequent data read accordingly into m_rx_buffer[1], etc.

How does this work? When I send the data I do the read with the buffer located at m_rx_buffer[m_cdc_buffer_idx] and m_cdc_idx = 1 - so shouldn't the beginning of my data appear in m_rx_buffer[1] instead of m_rx_buffer[0]? I'm guessing that since the index gets increased but the data read into previous buffers, that's why it is not echoed to the terminal.

Here is the test data I'm using:

RIFFüakWAVEfmt      D¬  Ì   bext[                                                                                                                                                                                                                                                                  iZotope RX 7 Audio Editor       USIZT0037408408461551490069551812021-03-1115:51:49                                                                                                                                                                                                                                                                          datas_kÿÿÿ

I was planning on keeping track of which buffer I was on with the m_cdc_buffer_idx variable, but if it isn't on the current buffer, that kind of defeats the purpose. How do other people process variable length data from USB CDC ACM?