On Tue, Aug 24, 2021 at 02:32:32PM +0200, Willy Tarreau wrote:

On Tue, Aug 24, 2021 at 02:19:26PM +0200, Johan Hovold wrote:

quoted

This reverts commit 3c18e9baee0ef97510dcda78c82285f52626764b.

These devices do not appear to send a zero-length packet when the
transfer size is a multiple of the bulk-endpoint max-packet size. This
means that incoming data may not be processed by the driver until a
short packet is received or the receive buffer is full.

Revert back to using endpoint-sized receive buffers to avoid stalled
reads.

Sorry for this, I didn't notice any issue here (aside for the chip
working where it used not to). I have no idea what these zero-length
packets correspond to, nor why they're affected by the transfer size.
Do you have any idea what I should look for ? Because without that
patch, the device is unusable for me :-/

Zero-length packets are used to indicate completion of bulk transfers
that are multiples of the endpoint max-packet size (as per the USB
spec). Without those the host controller driver doesn't now that the
transfer is complete and that it should call the driver completion
callback (and instead waits for the other completion conditions).

It may be possible to configure the device to send ZLPs somehow but
since there's no public documentation for the protocol that may require
some reverse engineering.

Johan