Hiya,
On two different systems (arm and x86) I've seen an issue with module-
null-sink. When samples are streamed to a null sink using aplay, the
cpu gets pinned. From the arm system:
top - 04:01:05 up 1:14, 4 users, load average: 1.42, 0.83, 0.60
Tasks: 82 total, 2 running, 80 sleeping, 0 stopped, 0 zombie
Cpu(s): 81.9%us, 18.1%sy, 0.0%ni, 0.0%id, 0.0%wa, 0.0%hi,
0.0%si, 0.0%st
Mem: 252572k total, 245456k used, 7116k free, 10136k buffers
Swap: 0k total, 0k used, 0k free, 121888k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
1921 root 20 0 77468 1712 1480 R 90.1 0.7 0:57.44 aplay
1895 root 9 -11 166m 4556 2976 S 1.0 1.8 0:02.67 pulseaudio
1917 root 20 0 2440 1148 896 R 1.0 0.5 0:00.94 top
110 root 15 -5 0 0 0 S 0.7 0.0 0:08.56
omap2_mcspi
1150 root -97 0 126m 2880 2224 S 0.7 1.1 0:28.08 hidd
1426 root 20 0 69216 21m 3916 S 0.7 8.7 0:52.38 java
5 root 15 -5 0 0 0 S 0.3 0.0 0:05.82 events/0
1209 root 20 0 14080 2148 1840 S 0.3 0.9 0:14.20 powerd
1296 root 20 0 47612 1672 1256 S 0.3 0.7 0:01.64 extractfs
1914 root 20 0 2588 1032 812 S 0.3 0.4 0:00.35 dropbear
1 root 20 0 1892 964 688 S 0.0 0.4 0:00.42 upstart
2 root 15 -5 0 0 0 S 0.0 0.0 0:00.00 kthreadd
3 root 15 -5 0 0 0 S 0.0 0.0 0:00.01
ksoftirqd/0
4 root RT -5 0 0 0 S 0.0 0.0 0:00.00 watchdog/0
I see similar on x86.
paplay barely shows up in top playing the same file, so it might seem
something is hawked with the alsa interface. Output from oprofile
would also indicate something is up here:
CPU: ARM V7 PMNC, speed 0 MHz (estimated)
Counted CPU_CYCLES events (Number of CPU cycles) with a unit mask of
0x00 (No unit mask) count 10000
CPU_CYCLES:10000|
samples| %|
------------------
444 81.3187 aplay
CPU_CYCLES:10000|
samples| %|
------------------
134 30.1802 libpulse.so.0.7.0
103 23.1982 libasound.so.2.0.0
84 18.9189 System.map-2.6.24-palm-joplin-3430
61 13.7387 libpthread-2.5.so
43 9.6847 libasound_module_pcm_pulse.so
19 4.2793 libc-2.5.so
...
12 2.1978 pulseaudio
CPU_CYCLES:10000|
samples| %|
------------------
3 25.0000 System.map-2.6.24-palm-joplin-3430
3 25.0000 libpulsecore.so.8.0.0
2 16.6667 libc-2.5.so
1 8.3333 libm-2.5.so
1 8.3333 libpthread-2.5.so
1 8.3333 librt-2.5.so
1 8.3333 libpstream.so
detail for aplay looks like this:
CPU: ARM V7 PMNC, speed 0 MHz (estimated)
Counted CPU_CYCLES events (Number of CPU cycles) with a unit mask of
0x00 (No unit mask) count 10000
samples % image name symbol name
61 13.7387 libpthread-2.5.so (no symbols)
22 4.9550 libasound.so.2.0.05045 libpulse.so.0.7.0
pa_stream_writable_size
19 4.2793 libc-2.5.so (no symbols)
18 4.0541 libpulse.so.0.7.0 pa_thread_self
16 3.6036 libasound_module_pcm_pulse.so pulse_pointer
16 3.6036 libpulse.so.0.7.0 pa_thread_is_running
16 3.6036 libpulse.so.0.7.0 pa_threaded_mainloop_lock
15 3.3784 libasound.so.2.0.0 snd_pcm_ioplug_avail_update
13 2.9279 System.map-2.6.24-palm-joplin-3430 vector_swi
12 2.7027 libpulse.so.0.7.0 pa_mutex_lock
12 2.7027 libpulse.so.0.7.0 pa_threaded_mainloop_unlock
11 2.4775 libasound.so.2.0.0 snd_pcm_hwsync
11 2.4775 libpulse.so.0.7.0 pa_run_once
10 2.2523 libpulse.so.0.7.0 pa_context_get_state
9 2.0270 System.map-2.6.24-palm-joplin-3430 pipe_poll
9 2.0270 libasound.so.2.0.0 snd_pcm_ioplug_hw_ptr_update
8 1.8018 System.map-2.6.24-palm-joplin-3430 fget_light
8 1.8018 libasound_module_pcm_pulse.so pulse_pcm_poll_revents
8 1.8018 libasound_module_pcm_pulse.so update_ptr
7 1.5766 libasound.so.2.0.0 snd_pcm_bytes_to_frames
7 1.5766 libpulse.so.0.7.0 pa_once_begin
6 1.3514 System.map-2.6.24-palm-joplin-3430 fput
6 1.3514 libasound.so.2.0.0 .plt
6 1.3514 libasound.so.2.0.0 snd_pcm_wait
6 1.3514 libpulse.so.0.7.0 .plt
5 1.1261 System.map-2.6.24-palm-joplin-3430 poll_freewait
5 1.1261 System.map-2.6.24-palm-joplin-3430 sys_poll
5 1.1261 libasound_module_pcm_pulse.so pulse_check_connection
4 0.9009 libasound.so.2.0.0 snd_pcm_ioplug_hwsync
4 0.9009 libasound.so.2.0.0
snd_pcm_poll_descriptors_count
4 0.9009 libasound.so.2.0.0
snd_pcm_poll_descriptors_revents
4 0.9009 libasound_module_pcm_pulse.so check_active
3 0.6757 System.map-2.6.24-palm-joplin-3430 add_wait_queue
3 0.6757 System.map-2.6.24-palm-joplin-3430 remove_wait_queue
3 0.6757 System.map-2.6.24-palm-joplin-3430 ret_fast_syscall
3 0.6757 libasound.so.2.0.0 __udivsi3
3 0.6757 libasound.so.2.0.0 snd_pcm_avail_update
3 0.6757 libpulse.so.0.7.0 pa_tls_get
2 0.4505 System.map-2.6.24-palm-joplin-3430 fput_light
2 0.4505 libasound.so.2.0.0 snd1_pcm_write_areas
2 0.4505 libasound.so.2.0.0
snd_pcm_ioplug_poll_descriptors_count
2 0.4505 libasound.so.2.0.0 snd_pcm_ioplug_poll_revents
2 0.4505 libasound.so.2.0.0 snd_pcm_poll_descriptors
2 0.4505 libasound_module_pcm_pulse.so .plt
2 0.4505 libpulse.so.0.7.0 pa_mutex_unlock
1 0.2252 System.map-2.6.24-palm-joplin-3430 __copy_to_user
1 0.2252 System.map-2.6.24-palm-joplin-3430 __do_softirq
1 0.2252 System.map-2.6.24-palm-joplin-3430 cpu_v7_switch_mm
1 0.2252 System.map-2.6.24-palm-joplin-3430 omap_32k_read
1 0.2252 System.map-2.6.24-palm-joplin-3430 schedule
1 0.2252 System.map-2.6.24-palm-joplin-3430 sys_sendto
1 0.2252 System.map-2.6.24-palm-joplin-3430 verify_iovec
1 0.2252 libasound.so.2.0.0
snd_pcm_ioplug_poll_descriptors
1 0.2252 libpulse.so.0.7.0 __udivsi3
Let me know if more info would be useful.
Has anyone else seen issues with module-null-sink in 0.9.13?
Thanks!!
Nick
