The chip supports high transfer rates, but with the small default buffers (64 bytes read), some entire blocks are regularly lost. This typically happens at 1.5 Mbps (which is the default speed on Rockchip devices) when used as a console to access U-Boot where the output of the "help" command misses many lines and where "printenv" mangles the environment.
The FTDI driver doesn't suffer at all from this. One difference is that it uses 512 bytes rx buffers and 256 bytes tx buffers. Adopting these values completely resolved the issue, even the output of "dmesg" is reliable. I preferred to leave the Tx value unchanged as it is not involved in this issue, while a change could increase the risk of triggering the same issue with other devices having too small buffers.
I verified that it backports well (and works) at least to 5.4. It's of low importance enough to be dropped where it doesn't trivially apply anymore.
Cc: stable@vger.kernel.org Signed-off-by: Willy Tarreau w@1wt.eu --- drivers/usb/serial/ch341.c | 1 + 1 file changed, 1 insertion(+)
diff --git a/drivers/usb/serial/ch341.c b/drivers/usb/serial/ch341.c index 2db917eab799..8a521b5ea769 100644 --- a/drivers/usb/serial/ch341.c +++ b/drivers/usb/serial/ch341.c @@ -851,6 +851,7 @@ static struct usb_serial_driver ch341_device = { .owner = THIS_MODULE, .name = "ch341-uart", }, + .bulk_in_size = 512, .id_table = id_table, .num_ports = 1, .open = ch341_open,
On Sat, Jul 24, 2021 at 05:27:39PM +0200, Willy Tarreau wrote:
The chip supports high transfer rates, but with the small default buffers (64 bytes read), some entire blocks are regularly lost. This typically happens at 1.5 Mbps (which is the default speed on Rockchip devices) when used as a console to access U-Boot where the output of the "help" command misses many lines and where "printenv" mangles the environment.
The FTDI driver doesn't suffer at all from this. One difference is that it uses 512 bytes rx buffers and 256 bytes tx buffers. Adopting these values completely resolved the issue, even the output of "dmesg" is reliable. I preferred to leave the Tx value unchanged as it is not involved in this issue, while a change could increase the risk of triggering the same issue with other devices having too small buffers.
Since these device do not support automatic flow control this is indeed the best we can to do here (otherwise I'd probably prefer framing it more as an optimisation than a fix).
I verified that it backports well (and works) at least to 5.4. It's of low importance enough to be dropped where it doesn't trivially apply anymore.
This should be fine to backport to all stable trees.
Cc: stable@vger.kernel.org Signed-off-by: Willy Tarreau w@1wt.eu
Now applied for 5.14, thanks.
Johan
linux-stable-mirror@lists.linaro.org
-
Johan Hovold -
Willy Tarreau