USB Host Dual and Quad FTDI support with possible bugs?

[fertinator]

Hi all,

I am working on a LPC55S69 with USB host on USB0 with the OHCI driver.
I need to connect a FT4232H to the USB and interface with all 4 ports.

I got it partially working now but I did need to add 2 lines of code to the library to make it work better.

First the important configuration settings that I used:

#define CFG_TUH_CDC              4 
#define CFG_TUH_CDC_FTDI         1
#define CFG_TUH_INTERFACE_MAX    4

When connecting the FT4232 the device is recognized and proper endpoints are opened. It went wrong with the configuration thereafter. With any value of CFG_TUH_CDC higher than 1, the configuration got stuck into an endless loop.

I fixed this by changing the ift_offset in cdc_host.c in function set_config_complete from:
const uint8_t itf_offset = (p_cdc->serial_drid == SERIAL_DRIVER_ACM) ? 1 : 0;
to
const uint8_t itf_offset = p_cdc->ftdi.channel - 1;
Also the ftdi.channel is not properly updated in the driver so that needed also fixing in cdc_host.c in function ftdi_proccess_set_config

if (p_cdc->bInterfaceNumber == 0) {
TU_ASSERT(ftdi_determine_type(p_cdc));
p_cdc->ftdi.channel = 1 + xfer->setup->wIndex;
}

This solution made it at least possible to send data from LPC to the TX ports. However I am still only able to receive from 1 port, still debugging that.

It's not yet fully clear to me what the proper configuration is. For example 1 FT4232 vs 4 FT232.
In case of the FT4232 should the bInterfaceNumber property be always 0 while the ftdi_private_t ftdi.channel should be 1-4?
And for the 4 FT232 there should be 4 different bInterfaceNumber values?
What I am missing is the distinction of 1 device with 4 uarts and 4 devices with each 1 uart. I feel that is also missing from the implementation?

[ingmichaelknecht]

Hi,
have you found a solution?
I'm searching for a way to connect an embedded system (most likekly an esp 32 because it's easy to fetch one)
to a FT4232.

[HiFiPhile]

You can try #3501

[hathach]

3501 is done by openclaw + cdoex. It does not seems to be well written enough. replaced by #3513

[fertinator]

Hi,
I managed to get it working with 0.19.0, not with 0.20.0 (that just breaks during enumeration). But I am using only 2 of the 4 uarts
I had to change the following:
In cdc_host.c:
case CONFIG_FTDI_DETERMINE_TYPE: if (p_cdc->bInterfaceNumber == 0) { TU_ASSERT(ftdi_determine_type(p_cdc)); p_cdc->ftdi.channel = 1 + xfer->setup->wIndex; } else {
The "p_cdc->ftdi.channel = 1 + xfer->setup->wIndex;" line is needed as the device stays the same, but index changes

Also from the 2nd device the p_cdc doesn't line up with the interface, only the endpoints are stored properly. Transfer is therefore also not initiated properly. Running tuh_cdc_read_clear(idx) after all enumeration is done will start all transactions.

uint8_t count = 0; void tuh_cdc_mount_cb(uint8_t idx) { (void) idx; /* Little workaround to have the 2nd serial port starting to read data. Just flushing the buffers */ /* Should be fixed in library */ count++; if(count == 2) { tuh_cdc_read_clear(0); tuh_cdc_read_clear(1); } }

I added following 2 defines in the tusb_config.h:
#define CFG_TUH_CDC 2
#define CFG_TUH_CDC_FTDI 1

I haven't tried above commit, it could work without my mods, but it's up to you to test it.

Ugh. Code doesn't format well. Sorry about that

[hathach]

should be fixed by #3513