Skip to contentSkip to conversation

audio/jack: fixes to overall jack behaviour

PATCH In review Geoffrey McRae 3 participants 14 messages Jun 14, 2020 – Jun 22, 2020

lore
Share & tools
lore
#0 lore raw

cover audio/jack: fixes to overall jack behaviour

This patch set addresses several issues that cause inconsistent behaviour in the guest when the sound device is stopped and started or the JACK server stops responding on the host.

Geoffrey McRae (6):
audio/jack: fix invalid minimum buffer size check audio/jack: remove unused stopped state audio/jack: remove invalid set of input support bool audio/jack: do not remove ports when finishing audio/jack: honour the enable state of the audio device audio/jack: simplify the re-init code path

 audio/jackaudio.c | 73 ++++++++++++++++++++++++-----------------------
 1 file changed, 38 insertions(+), 35 deletions(-)
Signature · 1 line
2.20.1
#1 lore raw

PATCH 1/6 audio/jack: fix invalid minimum buffer size check

JACK does not provide us with the configured buffer size until after activiation which was overriding this minimum value. JACK itself doesn't have this minimum limitation, but the QEMU virtual hardware and as such it must be enforced, failure to do so results in audio discontinuities.

---
 audio/jackaudio.c | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)
audio/jackaudio.c +11 -11
index 722ddb1dfe..d0b6f748f2 100644
@@ -434,17 +434,6 @@ static int qjack_client_init(QJackClient *c)
     jack_set_xrun_callback(c->client, qjack_xrun, c);
     jack_on_shutdown(c->client, qjack_shutdown, c);
 
-    /*
-     * ensure the buffersize is no smaller then 512 samples, some (all?) qemu
-     * virtual devices do not work correctly otherwise
-     */
-    if (c->buffersize < 512) {
-        c->buffersize = 512;
-    }
-
-    /* create a 2 period buffer */
-    qjack_buffer_create(&c->fifo, c->nchannels, c->buffersize * 2);
-
     /* allocate and register the ports */
     c->port = g_malloc(sizeof(jack_port_t *) * c->nchannels);
     for (int i = 0; i < c->nchannels; ++i) {
@@ -468,6 +457,17 @@ static int qjack_client_init(QJackClient *c)
     jack_activate(c->client);
     c->buffersize = jack_get_buffer_size(c->client);
 
+    /*
+     * ensure the buffersize is no smaller then 512 samples, some (all?) qemu
+     * virtual devices do not work correctly otherwise
+     */
+    if (c->buffersize < 512) {
+        c->buffersize = 512;
+    }
+
+    /* create a 2 period buffer */
+    qjack_buffer_create(&c->fifo, c->nchannels, c->buffersize * 2);
+
     qjack_client_connect_ports(c);
     c->state = QJACK_STATE_RUNNING;
     return 0;
Signature · 1 line
2.20.1
#2 lore raw

PATCH 2/6 audio/jack: remove unused stopped state

In reply to Geoffrey McRae
---
 audio/jackaudio.c | 4 ----
 1 file changed, 4 deletions(-)
audio/jackaudio.c +0 -4
index d0b6f748f2..fb8efd7af7 100644
@@ -38,7 +38,6 @@ struct QJack;
 
 typedef enum QJackState {
     QJACK_STATE_DISCONNECTED,
-    QJACK_STATE_STOPPED,
     QJACK_STATE_RUNNING,
     QJACK_STATE_SHUTDOWN
 }
@@ -549,9 +548,6 @@ static void qjack_client_fini(QJackClient *c)
 {
     switch (c->state) {
     case QJACK_STATE_RUNNING:
-        /* fallthrough */
-
-    case QJACK_STATE_STOPPED:
         for (int i = 0; i < c->nchannels; ++i) {
             jack_port_unregister(c->client, c->port[i]);
         }
Signature · 1 line
2.20.1
#3 lore raw

PATCH 3/6 audio/jack: remove invalid set of input support bool

In reply to Geoffrey McRae

Initial code for JACK did not support audio input and as such this boolean was set to let QEMU know, however JACK ended up including input support making this invalid. Further investigation shows it was invalid to set it in the first instance anyway due to a failure on my part understand properly what this was for when the audodev was initially developed.

---
 audio/jackaudio.c | 3 ---
 1 file changed, 3 deletions(-)
audio/jackaudio.c +0 -3
index fb8efd7af7..58c7344497 100644
@@ -607,9 +607,6 @@ static int qjack_thread_creator(jack_native_thread_t *thread,
 static void *qjack_init(Audiodev *dev)
 {
     assert(dev->driver == AUDIODEV_DRIVER_JACK);
-
-    dev->u.jack.has_in = false;
-
     return dev;
 }
 
Signature · 1 line
2.20.1
#4 lore raw

PATCH 4/6 audio/jack: do not remove ports when finishing

In reply to Geoffrey McRae

This fixes a hang when there is a communications issue with the JACK server. Simply closing the connection is enough to completely clean up and as such we do not need to remove the ports first. As JACK uses a socket based protocol that relies on the select call, if there is a communication breakdown with the server the client library waits forever for a response to the unregister request.

---
 audio/jackaudio.c | 3 ---
 1 file changed, 3 deletions(-)
audio/jackaudio.c +0 -3
index 58c7344497..249cbd3265 100644
@@ -548,9 +548,6 @@ static void qjack_client_fini(QJackClient *c)
 {
     switch (c->state) {
     case QJACK_STATE_RUNNING:
-        for (int i = 0; i < c->nchannels; ++i) {
-            jack_port_unregister(c->client, c->port[i]);
-        }
         jack_deactivate(c->client);
         /* fallthrough */
 
Signature · 1 line
2.20.1
#5 lore raw

PATCH 5/6 audio/jack: honour the enable state of the audio device

In reply to Geoffrey McRae

When the guest closes the audio device we must start dropping input samples from JACK and zeroing the output buffer samples. Failure to do so causes sound artifacts during operations such as guest OS reboot, and causes a hang of the input pipeline breaking it until QEMU is restated.

Closing and reconnecting to JACK was tested during these enable/disable calls which works well for Linux guests, however Windows re-opens the audio hardware repeatedly even when doing simple tasks like playing a system sounds. As such it was decided it is better to feed silence to JACK while the device is disabled.

---
 audio/jackaudio.c | 29 +++++++++++++++++++++--------
 1 file changed, 21 insertions(+), 8 deletions(-)
audio/jackaudio.c +21 -8
index 249cbd3265..b2b53985ae 100644
@@ -56,7 +56,7 @@ typedef struct QJackClient {
     AudiodevJackPerDirectionOptions *opt;
 
     bool out;
-    bool finished;
+    bool enabled;
     bool connect_ports;
     int  packets;
 
@@ -271,9 +271,17 @@ static int qjack_process(jack_nframes_t nframes, void *arg)
     }
 
     if (c->out) {
-        qjack_buffer_read_l(&c->fifo, buffers, nframes);
+        if (likely(c->enabled)) {
+            qjack_buffer_read_l(&c->fifo, buffers, nframes);
+        } else {
+            for(int i = 0; i < c->nchannels; ++i) {
+                memset(buffers[i], 0, nframes * sizeof(float));
+            }
+        }
     } else {
-        qjack_buffer_write_l(&c->fifo, buffers, nframes);
+        if (likely(c->enabled)) {
+            qjack_buffer_write_l(&c->fifo, buffers, nframes);
+        }
     }
 
     return 0;
@@ -314,8 +322,8 @@ static void qjack_client_recover(QJackClient *c)
     if (c->state == QJACK_STATE_DISCONNECTED &&
         c->packets % 100 == 0) {
 
-        /* if not finished then attempt to recover */
-        if (!c->finished) {
+        /* if enabled then attempt to recover */
+        if (c->enabled) {
             dolog("attempting to reconnect to server\n");
             qjack_client_init(c);
         }
@@ -387,7 +395,6 @@ static int qjack_client_init(QJackClient *c)
     char client_name[jack_client_name_size()];
     jack_options_t options = JackNullOption;
 
-    c->finished      = false;
     c->connect_ports = true;
 
     snprintf(client_name, sizeof(client_name), "%s-%s",
@@ -483,8 +490,10 @@ static int qjack_init_out(HWVoiceOut *hw, struct audsettings *as,
     }
 
     jo->c.out       = true;
+    jo->c.enabled   = false;
     jo->c.nchannels = as->nchannels;
     jo->c.opt       = dev->u.jack.out;
+
     int ret = qjack_client_init(&jo->c);
     if (ret != 0) {
         return ret;
@@ -519,8 +528,10 @@ static int qjack_init_in(HWVoiceIn *hw, struct audsettings *as,
     }
 
     ji->c.out       = false;
+    ji->c.enabled   = false;
     ji->c.nchannels = as->nchannels;
     ji->c.opt       = dev->u.jack.in;
+
     int ret = qjack_client_init(&ji->c);
     if (ret != 0) {
         return ret;
@@ -568,23 +579,25 @@ static void qjack_client_fini(QJackClient *c)
 static void qjack_fini_out(HWVoiceOut *hw)
 {
     QJackOut *jo = (QJackOut *)hw;
-    jo->c.finished = true;
     qjack_client_fini(&jo->c);
 }
 
 static void qjack_fini_in(HWVoiceIn *hw)
 {
     QJackIn *ji = (QJackIn *)hw;
-    ji->c.finished = true;
     qjack_client_fini(&ji->c);
 }
 
 static void qjack_enable_out(HWVoiceOut *hw, bool enable)
 {
+    QJackOut *jo = (QJackOut *)hw;
+    jo->c.enabled = enable;
 }
 
 static void qjack_enable_in(HWVoiceIn *hw, bool enable)
 {
+    QJackIn *ji = (QJackIn *)hw;
+    ji->c.enabled = enable;
 }
 
 static int qjack_thread_creator(jack_native_thread_t *thread,
Signature · 1 line
2.20.1
#6 lore raw

Re: [PATCH 5/6] audio/jack: honour the enable state of the audio device

Quoted 10 lines · Geoffrey McRae

When the guest closes the audio device we must start dropping input samples from JACK and zeroing the output buffer samples. Failure to do so causes sound artifacts during operations such as guest OS reboot, and causes a hang of the input pipeline breaking it until QEMU is restated.

Closing and reconnecting to JACK was tested during these enable/disable calls which works well for Linux guests, however Windows re-opens the audio hardware repeatedly even when doing simple tasks like playing a system sounds. As such it was decided it is better to feed silence to JACK while the device is disabled.

Hmm, I guess feeding silence into jack needs some cpu cycles? Maybe add a timer to close the jack server connection? Keep the connection open for re-use for a while, but in case the guest stops playing sound altogether close the jack connection after being unused for a few minutes?

[ Doesn't render the patch invalid, consider it a suggestion for future
improvements ]

take care,
Gerd

#7 lore raw

Re: [PATCH 5/6] audio/jack: honour the enable state of the audio device

Quoted 12 lines · Gerd Hoffmann
Quoted 10 lines · Geoffrey McRae

When the guest closes the audio device we must start dropping input samples from JACK and zeroing the output buffer samples. Failure to do so causes sound artifacts during operations such as guest OS reboot, and causes a hang of the input pipeline breaking it until QEMU is restated.

Closing and reconnecting to JACK was tested during these enable/disable calls which works well for Linux guests, however Windows re-opens the audio hardware repeatedly even when doing simple tasks like playing a system sounds. As such it was decided it is better to feed silence to JACK while the device is disabled.

Quoted 8 lines

Hmm, I guess feeding silence into jack needs some cpu cycles? Maybe add a timer to close the jack server connection? Keep the connection open for re-use for a while, but in case the guest stops playing sound altogether close the jack connection after being unused for a few minutes?

[ Doesn't render the patch invalid, consider it a suggestion for future
improvements ]

Thanks, I had considered that however there is a usability issue to consider. Each time a jack client registers, it removes the client entirely and disconnects any custom plumbing that may have been done by the user. JACK does not remember this custom routing and as such it's lost until the user re-establishes it, or they have some automation set up to do it. While this could likely be worked around, it would likely cause more headaches then the few CPU cycles lost in a memset.

Further to this, while I have added some automation to connect ports via regex matching, this is likely not going to be used by many as it's not the normal method of connecting things up. This was added to open up this for general use for those that don't care for custom filters but just want the reliable audio interface.

Quoted 3 lines

take care,
Gerd

#8 lore raw

Re: [PATCH 5/6] audio/jack: honour the enable state of the audio device

Hi,

Quoted 6 lines
Quoted 6 lines

Hmm, I guess feeding silence into jack needs some cpu cycles? Maybe add a timer to close the jack server connection? Keep the connection open for re-use for a while, but in case the guest stops playing sound altogether close the jack connection after being unused for a few minutes?

Quoted 6 lines
Quoted 2 lines

[ Doesn't render the patch invalid, consider it a suggestion for future
improvements ]

Thanks, I had considered that however there is a usability issue to consider. Each time a jack client registers, it removes the client entirely and disconnects any custom plumbing that may have been done by the user.

Hmm, ok, that is bad indeed.

Can you stop the stream without closing the connection?

Quoted 4 lines

JACK does not remember this custom routing and as such it's lost until the user re-establishes it, or they have some automation set up to do it. While this could likely be worked around, it would likely cause more headaches then the few CPU cycles lost in a memset.

I'm more concerned about the frequent wakeups to feed the next chunk of (silence) data to jack.

take care,
Gerd

#9 lore raw

Re: [PATCH 5/6] audio/jack: honour the enable state of the audio device

Quoted 11 lines · Gerd Hoffmann

Hi,

Quoted 9 lines
Quoted 8 lines

Hmm, I guess feeding silence into jack needs some cpu cycles? Maybe add a timer to close the jack server connection? Keep the connection open for re-use for a while, but in case the guest stops playing sound altogether close the jack connection after being unused for a few minutes?

[ Doesn't render the patch invalid, consider it a suggestion for future
improvements ]

Quoted 7 lines
Quoted 3 lines

Thanks, I had considered that however there is a usability issue to consider. Each time a jack client registers, it removes the client entirely and disconnects any custom plumbing that may have been done by the user.

Hmm, ok, that is bad indeed.

Can you stop the stream without closing the connection?

Not as far as I can tell, it seems the JACK API doesn't allow for this in a way that is useful to us.

Quoted 11 lines
Quoted 5 lines

JACK does not remember this custom routing and as such it's lost until the user re-establishes it, or they have some automation set up to do it. While this could likely be worked around, it would likely cause more headaches then the few CPU cycles lost in a memset.

I'm more concerned about the frequent wakeups to feed the next chunk of (silence) data to jack.

take care,
Gerd

#10 lore raw

Re: [PATCH 5/6] audio/jack: honour the enable state of the audio device

Quoted 5 lines · Geoffrey McRae
Quoted 2 lines

Can you stop the stream without closing the connection?

Not as far as I can tell, it seems the JACK API doesn't allow for this in a way that is useful to us.

What happens if you don't feed data to jack? The cracking you hear on reboots etc. sounds like jack might reuses the buffers in that case.

So, maybe you can stop sending data to jack when all buffers are already filled with silence?

take care,
Gerd

#11 lore raw

Re: [PATCH 5/6] audio/jack: honour the enable state of the audio device

Quoted 4 lines · Gerd Hoffmann
Quoted 3 lines · Geoffrey McRae
Quoted 2 lines

Can you stop the stream without closing the connection?

Quoted 8 lines
Quoted 2 lines

Not as far as I can tell, it seems the JACK API doesn't allow for this in a way that is useful to us.

What happens if you don't feed data to jack? The cracking you hear on reboots etc. sounds like jack might reuses the buffers in that case.

So, maybe you can stop sending data to jack when all buffers are already filled with silence?

Jack hands us a buffer we have to set, it's uninitialized as it's a ring buffer, if we do not zero it then we get repeating samples like your sound device has hung.

Quoted 3 lines

take care,
Gerd

#12 lore raw

PATCH 6/6 audio/jack: simplify the re-init code path

In reply to Geoffrey McRae

Instead of checking for the audodev state in each code path, centralize the check into the initialize function itself to make it safe to call it at any time.

---
 audio/jackaudio.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
audio/jackaudio.c +6 -6
index b2b53985ae..72ed7c4929 100644
@@ -395,6 +395,10 @@ static int qjack_client_init(QJackClient *c)
     char client_name[jack_client_name_size()];
     jack_options_t options = JackNullOption;
 
+    if (c->state == QJACK_STATE_RUNNING) {
+        return 0;
+    }
+
     c->connect_ports = true;
 
     snprintf(client_name, sizeof(client_name), "%s-%s",
@@ -485,9 +489,7 @@ static int qjack_init_out(HWVoiceOut *hw, struct audsettings *as,
     QJackOut *jo  = (QJackOut *)hw;
     Audiodev *dev = (Audiodev *)drv_opaque;
 
-    if (jo->c.state != QJACK_STATE_DISCONNECTED) {
-        return 0;
-    }
+    qjack_client_fini(&jo->c);
 
     jo->c.out       = true;
     jo->c.enabled   = false;
@@ -523,9 +525,7 @@ static int qjack_init_in(HWVoiceIn *hw, struct audsettings *as,
     QJackIn  *ji  = (QJackIn *)hw;
     Audiodev *dev = (Audiodev *)drv_opaque;
 
-    if (ji->c.state != QJACK_STATE_DISCONNECTED) {
-        return 0;
-    }
+    qjack_client_fini(&ji->c);
 
     ji->c.out       = false;
     ji->c.enabled   = false;
Signature · 1 line
2.20.1
#13 lore raw
In reply to Geoffrey McRae
Patchew URL: https://patchew.org/QEMU/[email protected]/

Patchew URL: https://patchew.org/QEMU/[email protected]/

Hi,

This series seems to have some coding style problems. See output below for more information:

Message-id: [email protected]
Subject: [PATCH 0/6] audio/jack: fixes to overall jack behaviour
Type: series

=== TEST SCRIPT BEGIN ===
#!/bin/bash
git rev-parse base > /dev/null || exit 0
git config --local diff.renamelimit 0
git config --local diff.renames True
git config --local diff.algorithm histogram
./scripts/checkpatch.pl --mailback base..
=== TEST SCRIPT END ===

Updating 3c8cf5a9c21ff8782164d1def7f44bd888713384 Switched to a new branch 'test'
c962352 audio/jack: simplify the re-init code path 1b80c6e audio/jack: honour the enable state of the audio device d549d5e0 audio/jack: do not remove ports when finishing d9f3c84 audio/jack: remove invalid set of input support bool c612550 audio/jack: remove unused stopped state 62d579e audio/jack: fix invalid minimum buffer size check

=== OUTPUT BEGIN ===
1/6 Checking commit 62d579e00e55 (audio/jack: fix invalid minimum buffer size check) 2/6 Checking commit c612550cb58e (audio/jack: remove unused stopped state) 3/6 Checking commit d9f3c846ec2f (audio/jack: remove invalid set of input support bool) 4/6 Checking commit d549d5e05a30 (audio/jack: do not remove ports when finishing) 5/6 Checking commit 1b80c6eb42f5 (audio/jack: honour the enable state of the audio device)
ERROR: space required before the open parenthesis '('
#42: FILE: audio/jackaudio.c:277:
+ for(int i = 0; i < c->nchannels; ++i) {

total: 1 errors, 0 warnings, 91 lines checked

Patch 5/6 has style problems, please review. If any of these errors are false positives report them to the maintainer, see CHECKPATCH in MAINTAINERS.

6/6 Checking commit c9623522cfe0 (audio/jack: simplify the re-init code path) === OUTPUT END ===

Test command exited with code: 1

The full log is available at
http://patchew.org/logs/[email protected]/testing.checkpatch/?type=message.
---
Email generated automatically by Patchew [https://patchew.org/].
Please send your feedback to [email protected]