speaker-test: inconsistent behaviour of --buffer option.

[AndrewAmmerlaan]

Without specifying the --buffer option speaker-test returns this:

andrew@andrew-gentoo-pc ~ % speaker-test -c2 -twav -Dplughw:CARD=PCH,DEV=3

speaker-test 1.2.9

Playback device is plughw:CARD=PCH,DEV=3
Stream parameters are 48000Hz, S16_LE, 2 channels
WAV file(s)
Rate set to 48000Hz (requested 48000Hz)
Buffer size range from 64 to 1048576
Period size range from 32 to 524288
Using max buffer size 1048576
Periods = 4
was set period_size = 262144
was set buffer_size = 1048576
0 - Front Left
1 - Front Right
Write error: -77,File descriptor in bad state
xrun_recovery failed: -77,File descriptor in bad state
Transfer failed: File descriptor in bad state

Note that it says Using max buffer size 1048576. But now if we try to explicitly specify this maximum buffer size:

andrew@andrew-gentoo-pc ~ % speaker-test -c2 -twav -Dplughw:CARD=PCH,DEV=3 --buffer 1048576

speaker-test 1.2.9

Playback device is plughw:CARD=PCH,DEV=3
Stream parameters are 48000Hz, S16_LE, 2 channels
WAV file(s)
Rate set to 48000Hz (requested 48000Hz)
Buffer size range from 64 to 1048576
Period size range from 32 to 524288
Requested buffer time 1000000 us
Periods = 4
was set period_size = 12000
was set buffer_size = 48000
0 - Front Left
1 - Front Right
Time per period = 2.251287

Note that now we get Requested buffer time 1000000 us which is less then what we actually specified. In fact this will happen with any --buffer parameter larger then 1000000:

andrew@andrew-gentoo-pc ~ % speaker-test -c2 -twav -Dplughw:CARD=PCH,DEV=3 --buffer 1000001

speaker-test 1.2.9

Playback device is plughw:CARD=PCH,DEV=3
Stream parameters are 48000Hz, S16_LE, 2 channels
WAV file(s)
Rate set to 48000Hz (requested 48000Hz)
Buffer size range from 64 to 1048576
Period size range from 32 to 524288
Requested buffer time 1000000 us
Periods = 4
was set period_size = 12000
was set buffer_size = 48000
0 - Front Left
1 - Front Right
Time per period = 2.251042

Leading me to believe that despite what speaker-test claims, 1000000 is actually the maximum, not 1048576.

Discovered while debugging https://gitlab.freedesktop.org/drm/intel/-/issues/8462
This behaviour of speaker-test is confusing though and feels to me like a separate issue.

[perexg]

There was 1 second wrap for buffer and period time in the code. I increased this in the above commit by 100 to allow tests for the large buffers.

Otherwise, you mix two things - sizes (in frames) and time (in microseconds).

If you have rate 48000Hz, the buffer size 48000 it is equal to 1000000us. There is no issue.

BTW: Buffer size (limited by the driver) 1048576 means 1048576 / 48000 = ~21.81 seconds not 1048576us.

[AndrewAmmerlaan]

Oops, I indeed got the units mixed up.
Thanks for the fix, this should help me further debug my i915 hdmi audio problem.