On Mon, Jun 8, 2015 at 4:10 PM, Peter Hurley <peter@xxxxxxxxxxxxxxxxxx> wrote:
> On 06/08/2015 03:12 PM, Steven Walter wrote:
>> Use kthread_worker instead of workqueues. For now there is only a
>> single workqueue, but the intention is to bring back the "low_latency"
>> tty option, along with a second high-priority kthread worker.
>> ---
>> drivers/tty/tty_buffer.c | 17 +++++++++++------
>> drivers/tty/tty_io.c | 3 ++-
>> drivers/tty/tty_port.c | 2 +-
>> include/linux/tty.h | 5 +++--
>> 4 files changed, 17 insertions(+), 10 deletions(-)
>>
>> diff --git a/drivers/tty/tty_buffer.c b/drivers/tty/tty_buffer.c
>> index ed7b5c8..4564202 100644
>> --- a/drivers/tty/tty_buffer.c
>> +++ b/drivers/tty/tty_buffer.c
>> @@ -3,6 +3,7 @@
>> */
>>
>> #include <linux/types.h>
>> +#include <linux/kthread.h>
>> #include <linux/errno.h>
>> #include <linux/tty.h>
>> #include <linux/tty_driver.h>
>> @@ -431,7 +432,7 @@ receive_buf(struct tty_struct *tty, struct tty_buffer *head, int count)
>> * 'consumer'
>> */
>>
>> -static void flush_to_ldisc(struct work_struct *work)
>> +static void flush_to_ldisc(struct kthread_work *work)
>> {
>> struct tty_port *port = container_of(work, struct tty_port, buf.work);
>> struct tty_bufhead *buf = &port->buf;
>> @@ -520,17 +521,21 @@ EXPORT_SYMBOL(tty_flip_buffer_push);
>> * Must be called before the other tty buffer functions are used.
>> */
>>
>> +static DEFINE_KTHREAD_WORKER(tty_buffer_worker);
>
> One i/o thread for a server running 5000 containers?
> Nope.
Are you really concerned that you would need more than one full CPU's
worth of horsepower just to drive the line disciplines? I find
that... surprising.
What might you propose instead? I could envision creating one kthread
per CPU, but that seems heavy-weight. Altenately, at the cost of
making the code somewhat trickier, I could continue to use workqueues
for "normal" priority ports, and only use the kthread worker for
low-latency ports.
Thoughts?
--
-Steven Walter <stevenrwalter@xxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
