We are sending data to controller at wrong average rate not equal to
1 packet / SDU interval, if Transport_Latency is not an integer multiple
of SDU_Interval. The calculation currently may also give zero, so no
data gets sent.
We are sending data in bursts of num ~= Transport_Latency/SDU_Interval
packets, in hopes that possibly larger timer interval makes things more
efficient.
Fix the data rate by sending num packets every num*SDU_Interval, so that
the average data rate is correct.
Also fix use of itimerspect.it_value with TFD_TIMER_ABSTIME. The value
set previously is going to always be in the past in CLOCK_MONOTONIC so
just set it to 1, and leave sending the initial packets to the main
loop.
---
Notes:
This assumes kernel shall set qos.interval to SDU_Interval and not
ISO_Interval.
client/player.c | 20 ++++++++++++++------
1 file changed, 14 insertions(+), 6 deletions(-)
diff --git a/client/player.c b/client/player.c
index 7f67425aa..8d17022de 100644
--- a/client/player.c
+++ b/client/player.c
@@ -5066,22 +5066,30 @@ static int transport_send(struct transport *transport, int fd,
if (timer_fd < 0)
return -errno;
+ /* Send data in bursts of
+ * num = ROUND_CLOSEST(Transport_Latency (ms) / SDU_Interval (us))
+ * with average data rate = 1 packet / SDU_Interval
+ */
+ transport->num = ROUND_CLOSEST(qos->latency * 1000, qos->interval);
+ if (!transport->num)
+ transport->num = 1;
+
memset(&ts, 0, sizeof(ts));
- ts.it_value.tv_nsec = qos->latency * 1000000;
- ts.it_interval.tv_nsec = qos->latency * 1000000;
+ ts.it_value.tv_nsec = 1;
+ ts.it_interval.tv_nsec = transport->num * qos->interval * 1000;
- if (timerfd_settime(timer_fd, TFD_TIMER_ABSTIME, &ts, NULL) < 0)
+ if (timerfd_settime(timer_fd, TFD_TIMER_ABSTIME, &ts, NULL) < 0) {
+ close(timer_fd);
return -errno;
+ }
transport->fd = fd;
- /* num of packets = ROUND_CLOSEST(latency (ms) / interval (us)) */
- transport->num = ROUND_CLOSEST(qos->latency * 1000, qos->interval);
transport->timer_io = io_new(timer_fd);
io_set_read_handler(transport->timer_io, transport_timer_read,
transport, NULL);
- return transport_send_seq(transport, fd, 1);
+ return 0;
}
static void cmd_send_transport(int argc, char *argv[])
--
2.45.0
This is automated email and please do not reply to this email!
Dear submitter,
Thank you for submitting the patches to the linux bluetooth mailing list.
This is a CI test results with your patch series:
PW Link:https://patchwork.kernel.org/project/bluetooth/list/?series=852520
---Test result---
Test Summary:
CheckPatch PASS 0.41 seconds
GitLint PASS 0.29 seconds
BuildEll PASS 24.81 seconds
BluezMake PASS 1755.50 seconds
MakeCheck PASS 13.61 seconds
MakeDistcheck PASS 181.36 seconds
CheckValgrind PASS 252.17 seconds
CheckSmatch PASS 359.45 seconds
bluezmakeextell PASS 122.46 seconds
IncrementalBuild PASS 1490.77 seconds
ScanBuild PASS 1058.36 seconds
---
Regards,
Linux Bluetooth
Hi Pauli,
On Sat, May 11, 2024 at 7:34 AM Pauli Virtanen <[email protected]> wrote:
>
> We are sending data to controller at wrong average rate not equal to
> 1 packet / SDU interval, if Transport_Latency is not an integer multiple
> of SDU_Interval. The calculation currently may also give zero, so no
> data gets sent.
>
> We are sending data in bursts of num ~= Transport_Latency/SDU_Interval
> packets, in hopes that possibly larger timer interval makes things more
> efficient.
>
> Fix the data rate by sending num packets every num*SDU_Interval, so that
> the average data rate is correct.
>
> Also fix use of itimerspect.it_value with TFD_TIMER_ABSTIME. The value
> set previously is going to always be in the past in CLOCK_MONOTONIC so
> just set it to 1, and leave sending the initial packets to the main
> loop.
> ---
>
> Notes:
> This assumes kernel shall set qos.interval to SDU_Interval and not
> ISO_Interval.
>
> client/player.c | 20 ++++++++++++++------
> 1 file changed, 14 insertions(+), 6 deletions(-)
>
> diff --git a/client/player.c b/client/player.c
> index 7f67425aa..8d17022de 100644
> --- a/client/player.c
> +++ b/client/player.c
> @@ -5066,22 +5066,30 @@ static int transport_send(struct transport *transport, int fd,
> if (timer_fd < 0)
> return -errno;
>
> + /* Send data in bursts of
> + * num = ROUND_CLOSEST(Transport_Latency (ms) / SDU_Interval (us))
> + * with average data rate = 1 packet / SDU_Interval
> + */
> + transport->num = ROUND_CLOSEST(qos->latency * 1000, qos->interval);
> + if (!transport->num)
> + transport->num = 1;
> +
> memset(&ts, 0, sizeof(ts));
> - ts.it_value.tv_nsec = qos->latency * 1000000;
> - ts.it_interval.tv_nsec = qos->latency * 1000000;
> + ts.it_value.tv_nsec = 1;
> + ts.it_interval.tv_nsec = transport->num * qos->interval * 1000;
>
> - if (timerfd_settime(timer_fd, TFD_TIMER_ABSTIME, &ts, NULL) < 0)
> + if (timerfd_settime(timer_fd, TFD_TIMER_ABSTIME, &ts, NULL) < 0) {
> + close(timer_fd);
> return -errno;
> + }
>
> transport->fd = fd;
> - /* num of packets = ROUND_CLOSEST(latency (ms) / interval (us)) */
> - transport->num = ROUND_CLOSEST(qos->latency * 1000, qos->interval);
> transport->timer_io = io_new(timer_fd);
>
> io_set_read_handler(transport->timer_io, transport_timer_read,
> transport, NULL);
>
> - return transport_send_seq(transport, fd, 1);
The above was actually done on purpose so we are always one interval
ahead to keep controller buffer full as much as possible.
> + return 0;
> }
>
> static void cmd_send_transport(int argc, char *argv[])
> --
> 2.45.0
>
>
--
Luiz Augusto von Dentz