Re: aconnect occasionally causes kernel oopses

From: folkert <folkert@vanheusden.com>
To: Takashi Iwai <tiwai@suse.de>
Cc: alsa-devel@alsa-project.org
Subject: Re: aconnect occasionally causes kernel oopses
Date: Mon, 2 Aug 2021 11:10:12 +0200	[thread overview]
Message-ID: <20210802091012.GR890690@belle.intranet.vanheusden.com> (raw)
In-Reply-To: <s5him0oz5zq.wl-tiwai@suse.de>

> > > In which situation?
> > 
> > I was testing something that listens for alsa events and for that I ran
> > a continuous loop that did:
> > 
> > while true
> > do
> > 	aconnect 128:1 14:0
> > 	aconnect -d 128:1 14:0
> > 	aconnect -d 128:2 128:1
> > 	aconnect 128:2 128:1
> > done
> > 
> > I ran 5 instances in parallel.
> > 
> > 14 is midi through
> > 128 is rtpmidi
> 
> So rtpmidi process keeps running during the loop, that is, it's only
> about connection and disconnection, right?
> Also, you're listening to an event during that -- but how?

I tried it again but with a simpler setup:

I've got these devices:

root@lappiemctopface:~# aplaymidi -l
 Port    Client name                      Port name
 14:0    Midi Through                     Midi Through Port-0
130:0    FLUID Synth (17032)              Synth input port (17032:0)
131:0    VMPK Input                       in
root@lappiemctopface:~# arecordmidi -l
 Port    Client name                      Port name
 14:0    Midi Through                     Midi Through Port-0
132:0    VMPK Output                      out

I run this in 3x parallel:

while true
do
        aconnect 132:0 130:0
        aconnect -d 132:0 130:0
done

and then in less than a minute I get a backtrace.

[ma aug  2 11:05:13 2021] Call Trace:
[ma aug  2 11:05:13 2021]  ? snd_seq_deliver_event+0x38/0x90 [snd_seq]
[ma aug  2 11:05:13 2021]  ? snd_seq_kernel_client_dispatch+0x72/0x90 [snd_seq]
[ma aug  2 11:05:13 2021]  kfree+0x3bc/0x3e0
[ma aug  2 11:05:13 2021]  ? snd_seq_port_disconnect+0x10c/0x140 [snd_seq]
[ma aug  2 11:05:13 2021]  snd_seq_port_disconnect+0x10c/0x140 [snd_seq]
[ma aug  2 11:05:13 2021]  snd_seq_ioctl_unsubscribe_port+0xb9/0x180 [snd_seq]
[ma aug  2 11:05:13 2021]  ? snd_seq_port_get_subscription+0xbb/0xd0 [snd_seq]
[ma aug  2 11:05:13 2021]  ? __check_object_size.part.0+0x3a/0x150
[ma aug  2 11:05:13 2021]  snd_seq_ioctl+0xe8/0x1b0 [snd_seq]
[ma aug  2 11:05:13 2021]  __x64_sys_ioctl+0x91/0xc0
[ma aug  2 11:05:13 2021]  do_syscall_64+0x38/0x90
[ma aug  2 11:05:13 2021]  entry_SYSCALL_64_after_hwframe+0x44/0xa9

Using:
fluidsynth                                 2.1.7-1
vmpk                                       0.7.2-1build1

On: 5.11.0-25-generic (ubuntu) on 2 cores, 2 threads/core cpu (64b).