Messages in this thread | | | Date | Tue, 5 Sep 2023 21:57:52 +0530 | Subject | Re: [PATCH v4 2/3] greybus: Add BeaglePlay Linux Driver | From | Ayush Singh <> |
| |
>> +static void hdlc_handle_rx_frame(struct gb_beagleplay *bg) >> +{ >> + u8 address = bg->rx_buffer[0]; >> + char *buffer = &bg->rx_buffer[2]; >> + size_t buffer_len = bg->rx_buffer_len - 4; >> + >> + switch (address) { >> + case ADDRESS_DBG: >> + hdlc_handle_dbg_frame(bg, buffer, buffer_len); >> + break; >> + case ADDRESS_GREYBUS: >> + hdlc_handle_greybus_frame(bg, buffer, buffer_len); >> + break; >> + default: >> + dev_warn(&bg->serdev->dev, "Got Unknown Frame %u", address); > ratelimit > Probably as well in several places with possible flooding.
I don't think `hdlc_handle_rx_frame` is the correct place since it only processes a single completed HDLC frame. The more appropriate place would be `hdlc_rx` if we want to limit based on the number of HDLC frames or `gb_beagleplay_tty_receive` to limit based on the number of bytes.
I would like to ask, though, why is rate limiting required here? Won't `serdev_device_ops->receive_buf` already rate limit the number of bytes somewhat? Or is it related to blocking in the `serdev_device_ops->receive_buf` callback? In the case of latter, it would probably make sense to ratelimit based on number of frames, I think.
Ayush Singh
| |