tree 61d2636d83fc8bd408e739bfc27fca47b30b6348
parent 6e7f6b82c60afb46ff71c2127421c66207966d6d
author Nicholas Piggin <npiggin@gmail.com> 1536471555 +1000
committer Greg Kroah-Hartman <gregkh@linuxfoundation.org> 1536595471 +0200

tty: hvc: hvc_poll() fix read loop batching

Commit ec97eaad1383 ("tty: hvc: hvc_poll() break hv read loop")
removes get_chars batching entirely, which slows down large console
operations like paste -- virtio console "feels worse than a 9600 baud
serial line," reports Matteo.

This adds back batching in a more latency friendly way. If the caller
can sleep then we try to fill the entire flip buffer, releasing the
lock and scheduling between each iteration. If it can not sleep, then
batches are limited to 128 bytes. Matteo confirms this fixes the
performance problem.

Latency testing the powerpc OPAL console with OpenBMC UART with a
large paste shows about 0.25ms latency, which seems reasonable. 10ms
latencies were typical for this case before the latency breaking work,
so we still see most of the benefit.

  kopald-1204    0d.h.    5us : hvc_poll <-hvc_handle_interrupt
  kopald-1204    0d.h.    5us : __hvc_poll <-hvc_handle_interrupt
  kopald-1204    0d.h.    5us : _raw_spin_lock_irqsave <-__hvc_poll
  kopald-1204    0d.h.    5us : tty_port_tty_get <-__hvc_poll
  kopald-1204    0d.h.    6us : _raw_spin_lock_irqsave <-tty_port_tty_get
  kopald-1204    0d.h.    6us : _raw_spin_unlock_irqrestore <-tty_port_tty_get
  kopald-1204    0d.h.    6us : tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.    7us : __tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.    7us+: opal_get_chars <-__hvc_poll
  kopald-1204    0d.h.   36us : tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.   36us : __tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.   36us+: opal_get_chars <-__hvc_poll
  kopald-1204    0d.h.   65us : tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.   65us : __tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.   66us+: opal_get_chars <-__hvc_poll
  kopald-1204    0d.h.   94us : tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.   95us : __tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.   95us+: opal_get_chars <-__hvc_poll
  kopald-1204    0d.h.  124us : tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.  124us : __tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.  125us+: opal_get_chars <-__hvc_poll
  kopald-1204    0d.h.  154us : tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.  154us : __tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.  154us+: opal_get_chars <-__hvc_poll
  kopald-1204    0d.h.  183us : tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.  184us : __tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.  184us+: opal_get_chars <-__hvc_poll
  kopald-1204    0d.h.  213us : tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.  213us : __tty_buffer_request_room <-__hvc_poll
  kopald-1204    0d.h.  213us+: opal_get_chars <-__hvc_poll
  kopald-1204    0d.h.  242us : _raw_spin_unlock_irqrestore <-__hvc_poll
  kopald-1204    0d.h.  242us : tty_flip_buffer_push <-__hvc_poll
  kopald-1204    0d.h.  243us : queue_work_on <-tty_flip_buffer_push
  kopald-1204    0d.h.  243us : tty_kref_put <-__hvc_poll
  kopald-1204    0d.h.  243us : hvc_kick <-hvc_handle_interrupt
  kopald-1204    0d.h.  243us : wake_up_process <-hvc_kick
  kopald-1204    0d.h.  244us : try_to_wake_up <-hvc_kick
  kopald-1204    0d.h.  244us : _raw_spin_lock_irqsave <-try_to_wake_up
  kopald-1204    0d.h.  244us : _raw_spin_unlock_irqrestore <-try_to_wake_up

Reported-by: Matteo Croce <mcroce@redhat.com>
Tested-by: Matteo Croce <mcroce@redhat.com>
Tested-by: Jason Gunthorpe <jgg@mellanox.com>
Tested-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>