On 11/23/2013 07:25 PM, One Thousand Gnomes wrote:

On Fri, 22 Nov 2013 10:59:25 -0500
Peter Hurley [off-list ref] wrote:

quoted

Instrument .receive_buf() and read() paths with trace_printk's
to aid in debugging flow control changes.

tty devices have a device, we have a dev_dbg() layer. The old tty trace
predates this but there isn't really any excuse for not using it now that
I can see ?

I was using the ftrace facility because it has significantly less performance
impact than printk (which was an important factor while debugging flow control
problems).

That said, I could further macro-ize n_tty_trace() with selectable facility
(which would be useful when debugging problems that end in cpu death).

Regards,
Peter Hurley