On 1/13/21 3:35 PM, Vincent MAILHOL wrote:

quoted

My understanding is that having a queue size being a power of two is
required in order not to use spinlocks (else, modulo operations would
break when the index wraparound back to zero). I tried to minimize the
number of spinlock: only one per bulk send or bulk receive.

Or do you mean to round up the skb_echo array length to the next power
of two in the driver despite the actual size of the device FIFO? Did
not think about that in the past but that should work.

I am going to think a bit more of how to improve that.

For the echo_skb rounding up might work, but you should round down, so that your
can directly use the masked value.

Marc
-- 
Pengutronix e.K.                 | Marc Kleine-Budde           |
Embedded Linux                   | https://www.pengutronix.de  |
Vertretung West/Dortmund         | Phone: +49-231-2826-924     |
Amtsgericht Hildesheim, HRA 2686 | Fax:   +49-5121-206917-5555 |