Merge tag 'audio-pull-request' of https://gitlab.com/marcandre.lureau/qemu into staging

Audio patches for QEMU 8.0

Cleanups and improvements from Volker Rümelin.

# -----BEGIN PGP SIGNATURE-----
#
# iQJQBAABCAA6FiEEh6m9kz+HxgbSdvYt2ujhCXWWnOUFAmQFjUccHG1hcmNhbmRy
# ZS5sdXJlYXVAcmVkaGF0LmNvbQAKCRDa6OEJdZac5ZLMD/sFWmVYccRqvOfOJvuC
# dJ7EtABmC6b4MxLkMyP5WLMoCezSha0vgcYwAL3oZX1dIBoGfaUXXdxIM8IK8rAv
# J2Fe2p9D+6tqzqc8jkbx0DmY040A3Gb18P60jL3NeQfmqCt7KmzuQoCMn6htPyyy
# NMgSRG3wy+esglmeU2majhxQZh2Vmwsi6i37A5Fg8b27oWLYHxeS6Lgp3PNAwJsA
# k/RKJeZDMjDh0f0wSHZN+8w0VVFQa3BjqoNV4UAgqSjsPaRlnQxppG2FEXNNrtLb
# bJ4rbXgSO68wrcweHgC2L7qWhLnYwdcdQFAPotmt2V1jtH4xx6f+ovwCdf0vZ5Fo
# YhHixpP41gFvF7sJWSZO0wUFKXTyLSdWIPaUKP9iMTZfXpBxo5H6rkPsZ6v5yyG2
# tLFlL0Rv/h3liEkngmRbw81gB9te92ZGXOjkB2Q69tGQl/tXRkhrn5VT2mufb7BV
# WUsdtSFl6P231lDuzQkFp85PkoRtr8eDPkkZW8/xgRx5h8j5I2bc6LRMd63w2tE6
# TGdUfiaqH396r1T87ynBdBWSc4H6YuImTouy27rAv/50x8w4cgk66wY+UGB7D8FX
# Eg4rU5Dzco2D+1RY50zSqZgPxIWjnC4xtYQXLaJURe0Way1HQqt/Pvp/aVFKUOef
# snYV1FLDvHg2Cf9LCGqvpW8b/w==
# =d4LY
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 06 Mar 2023 06:50:47 GMT
# gpg:                using RSA key 87A9BD933F87C606D276F62DDAE8E10975969CE5
# gpg:                issuer "marcandre.lureau@redhat.com"
# gpg: Good signature from "Marc-André Lureau <marcandre.lureau@redhat.com>" [full]
# gpg:                 aka "Marc-André Lureau <marcandre.lureau@gmail.com>" [full]
# Primary key fingerprint: 87A9 BD93 3F87 C606 D276  F62D DAE8 E109 7596 9CE5

* tag 'audio-pull-request' of https://gitlab.com/marcandre.lureau/qemu: (27 commits)
  audio: remove sw->ratio
  audio/audio_template: substitute sw->hw with hw
  audio: handle leftover audio frame from upsampling
  audio: make recording packet length calculation exact
  audio: rename variables in audio_pcm_sw_read()
  audio: replace the resampling loop in audio_pcm_sw_read()
  audio: make playback packet length calculation exact
  audio: remove unused noop_conv() function
  audio: don't misuse audio_pcm_sw_write()
  audio: rename variables in audio_pcm_sw_write()
  audio: remove sw == NULL check
  audio: replace the resampling loop in audio_pcm_sw_write()
  audio: make the resampling code greedy
  audio: change type and name of the resample buffer
  audio: change type of mix_buf and conv_buf
  alsaaudio: reintroduce default recording settings
  alsaaudio: change default playback settings
  audio: remove audio_calloc() function
  audio/audio_template: use g_new0() to replace audio_calloc()
  audio/audio_template: use g_malloc0() to replace audio_calloc()
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

MAINTAINERS

@@ -2490,6 +2490,7 @@ Subsystems
 ----------
 Overall Audio backends
 M: Gerd Hoffmann <kraxel@redhat.com>
+M: Marc-André Lureau <marcandre.lureau@redhat.com>
 S: Odd Fixes
 F: audio/
 X: audio/alsaaudio.c
@@ -2785,6 +2786,7 @@ F: docs/spice-port-fqdn.txt
 
 Graphics
 M: Gerd Hoffmann <kraxel@redhat.com>
+M: Marc-André Lureau <marcandre.lureau@redhat.com>
 S: Odd Fixes
 F: ui/
 F: include/ui/

audio/alsaaudio.c

@@ -222,11 +222,7 @@ static int alsa_poll_helper (snd_pcm_t *handle, struct pollhlp *hlp, int mask)
         return -1;
     }
 
-    pfds = audio_calloc ("alsa_poll_helper", count, sizeof (*pfds));
-    if (!pfds) {
-        dolog ("Could not initialize poll mode\n");
-        return -1;
-    }
+    pfds = g_new0(struct pollfd, count);
 
     err = snd_pcm_poll_descriptors (handle, pfds, count);
     if (err < 0) {
@@ -917,28 +913,23 @@ static void *alsa_audio_init(Audiodev *dev)
     alsa_init_per_direction(aopts->in);
     alsa_init_per_direction(aopts->out);
 
-    /*
-     * need to define them, as otherwise alsa produces no sound
-     * doesn't set has_* so alsa_open can identify it wasn't set by the user
-     */
+    /* don't set has_* so alsa_open can identify it wasn't set by the user */
     if (!dev->u.alsa.out->has_period_length) {
-        /* 1024 frames assuming 44100Hz */
-        dev->u.alsa.out->period_length = 1024 * 1000000 / 44100;
+        /* 256 frames assuming 44100Hz */
+        dev->u.alsa.out->period_length = 5805;
     }
     if (!dev->u.alsa.out->has_buffer_length) {
         /* 4096 frames assuming 44100Hz */
-        dev->u.alsa.out->buffer_length = 4096ll * 1000000 / 44100;
+        dev->u.alsa.out->buffer_length = 92880;
     }
 
-    /*
-     * OptsVisitor sets unspecified optional fields to zero, but do not depend
-     * on it...
-     */
     if (!dev->u.alsa.in->has_period_length) {
-        dev->u.alsa.in->period_length = 0;
+        /* 256 frames assuming 44100Hz */
+        dev->u.alsa.in->period_length = 5805;
     }
     if (!dev->u.alsa.in->has_buffer_length) {
-        dev->u.alsa.in->buffer_length = 0;
+        /* 4096 frames assuming 44100Hz */
+        dev->u.alsa.in->buffer_length = 92880;
     }
 
     return dev;

audio/audio_int.h

@@ -58,7 +58,7 @@ typedef struct SWVoiceCap SWVoiceCap;
 
 typedef struct STSampleBuffer {
     size_t pos, size;
-    st_sample samples[];
+    st_sample *buffer;
 } STSampleBuffer;
 
 typedef struct HWVoiceOut {
@@ -71,7 +71,7 @@ typedef struct HWVoiceOut {
     f_sample *clip;
     uint64_t ts_helper;
 
-    STSampleBuffer *mix_buf;
+    STSampleBuffer mix_buf;
     void *buf_emul;
     size_t pos_emul, pending_emul, size_emul;
 
@@ -93,7 +93,7 @@ typedef struct HWVoiceIn {
     size_t total_samples_captured;
     uint64_t ts_helper;
 
-    STSampleBuffer *conv_buf;
+    STSampleBuffer conv_buf;
     void *buf_emul;
     size_t pos_emul, pending_emul, size_emul;
 
@@ -108,8 +108,7 @@ struct SWVoiceOut {
     AudioState *s;
     struct audio_pcm_info info;
     t_sample *conv;
-    int64_t ratio;
-    struct st_sample *buf;
+    STSampleBuffer resample_buf;
     void *rate;
     size_t total_hw_samples_mixed;
     int active;
@@ -126,10 +125,9 @@ struct SWVoiceIn {
     AudioState *s;
     int active;
     struct audio_pcm_info info;
-    int64_t ratio;
     void *rate;
     size_t total_hw_samples_acquired;
-    struct st_sample *buf;
+    STSampleBuffer resample_buf;
     f_sample *clip;
     HWVoiceIn *hw;
     char *name;
@@ -151,8 +149,8 @@ struct audio_driver {
     int can_be_default;
     int max_voices_out;
     int max_voices_in;
-    int voice_size_out;
-    int voice_size_in;
+    size_t voice_size_out;
+    size_t voice_size_in;
     QLIST_ENTRY(audio_driver) next;
 };
 
@@ -251,7 +249,6 @@ void audio_pcm_init_info (struct audio_pcm_info *info, struct audsettings *as);
 void audio_pcm_info_clear_buf (struct audio_pcm_info *info, void *buf, int len);
 
 int audio_bug (const char *funcname, int cond);
-void *audio_calloc (const char *funcname, int nmemb, size_t size);
 
 void audio_run(AudioState *s, const char *msg);
 
@@ -294,9 +291,6 @@ static inline size_t audio_ring_posb(size_t pos, size_t dist, size_t len)
 #define ldebug(fmt, ...) (void)0
 #endif
 
-#define AUDIO_STRINGIFY_(n) #n
-#define AUDIO_STRINGIFY(n) AUDIO_STRINGIFY_(n)
-
 typedef struct AudiodevListEntry {
     Audiodev *dev;
     QSIMPLEQ_ENTRY(AudiodevListEntry) next;

audio/audio_template.h

@@ -40,7 +40,7 @@ static void glue(audio_init_nb_voices_, TYPE)(AudioState *s,
                                               struct audio_driver *drv)
 {
     int max_voices = glue (drv->max_voices_, TYPE);
-    int voice_size = glue (drv->voice_size_, TYPE);
+    size_t voice_size = glue(drv->voice_size_, TYPE);
 
     if (glue (s->nb_hw_voices_, TYPE) > max_voices) {
         if (!max_voices) {
@@ -63,16 +63,17 @@ static void glue(audio_init_nb_voices_, TYPE)(AudioState *s,
     }
 
     if (audio_bug(__func__, voice_size && !max_voices)) {
-        dolog ("drv=`%s' voice_size=%d max_voices=0\n",
-               drv->name, voice_size);
+        dolog("drv=`%s' voice_size=%zu max_voices=0\n",
+              drv->name, voice_size);
     }
 }
 
 static void glue (audio_pcm_hw_free_resources_, TYPE) (HW *hw)
 {
     g_free(hw->buf_emul);
-    g_free (HWBUF);
-    HWBUF = NULL;
+    g_free(HWBUF.buffer);
+    HWBUF.buffer = NULL;
+    HWBUF.size = 0;
 }
 
 static void glue(audio_pcm_hw_alloc_resources_, TYPE)(HW *hw)
@@ -83,56 +84,67 @@ static void glue(audio_pcm_hw_alloc_resources_, TYPE)(HW *hw)
             dolog("Attempted to allocate empty buffer\n");
         }
 
-        HWBUF = g_malloc0(sizeof(STSampleBuffer) + sizeof(st_sample) * samples);
-        HWBUF->size = samples;
+        HWBUF.buffer = g_new0(st_sample, samples);
+        HWBUF.size = samples;
+        HWBUF.pos = 0;
     } else {
-        HWBUF = NULL;
+        HWBUF.buffer = NULL;
+        HWBUF.size = 0;
     }
 }
 
 static void glue (audio_pcm_sw_free_resources_, TYPE) (SW *sw)
 {
-    g_free (sw->buf);
+    g_free(sw->resample_buf.buffer);
+    sw->resample_buf.buffer = NULL;
+    sw->resample_buf.size = 0;
 
     if (sw->rate) {
         st_rate_stop (sw->rate);
     }
-
-    sw->buf = NULL;
     sw->rate = NULL;
 }
 
 static int glue (audio_pcm_sw_alloc_resources_, TYPE) (SW *sw)
 {
-    int samples;
+    HW *hw = sw->hw;
+    uint64_t samples;
 
     if (!glue(audio_get_pdo_, TYPE)(sw->s->dev)->mixing_engine) {
         return 0;
     }
 
-#ifdef DAC
-    samples = ((int64_t) sw->HWBUF->size << 32) / sw->ratio;
-#else
-    samples = (int64_t)sw->HWBUF->size * sw->ratio >> 32;
-#endif
+    samples = muldiv64(HWBUF.size, sw->info.freq, hw->info.freq);
+    if (samples == 0) {
+        uint64_t f_fe_min;
+        uint64_t f_be = (uint32_t)hw->info.freq;
 
-    sw->buf = audio_calloc(__func__, samples, sizeof(struct st_sample));
-    if (!sw->buf) {
-        dolog ("Could not allocate buffer for `%s' (%d samples)\n",
-               SW_NAME (sw), samples);
+        /* f_fe_min = ceil(1 [frames] * f_be [Hz] / size_be [frames]) */
+        f_fe_min = (f_be + HWBUF.size - 1) / HWBUF.size;
+        qemu_log_mask(LOG_UNIMP,
+                      AUDIO_CAP ": The guest selected a " NAME " sample rate"
+                      " of %d Hz for %s. Only sample rates >= %" PRIu64 " Hz"
+                      " are supported.\n",
+                      sw->info.freq, sw->name, f_fe_min);
         return -1;
     }
 
+    /*
+     * Allocate one additional audio frame that is needed for upsampling
+     * if the resample buffer size is small. For large buffer sizes take
+     * care of overflows and truncation.
+     */
+    samples = samples < SIZE_MAX ? samples + 1 : SIZE_MAX;
+    sw->resample_buf.buffer = g_new0(st_sample, samples);
+    sw->resample_buf.size = samples;
+    sw->resample_buf.pos = 0;
+
 #ifdef DAC
-    sw->rate = st_rate_start (sw->info.freq, sw->hw->info.freq);
+    sw->rate = st_rate_start(sw->info.freq, hw->info.freq);
 #else
-    sw->rate = st_rate_start (sw->hw->info.freq, sw->info.freq);
+    sw->rate = st_rate_start(hw->info.freq, sw->info.freq);
 #endif
-    if (!sw->rate) {
-        g_free (sw->buf);
-        sw->buf = NULL;
-        return -1;
-    }
+
     return 0;
 }
 
@@ -149,11 +161,8 @@ static int glue (audio_pcm_sw_init_, TYPE) (
     sw->hw = hw;
     sw->active = 0;
 #ifdef DAC
-    sw->ratio = ((int64_t) sw->hw->info.freq << 32) / sw->info.freq;
     sw->total_hw_samples_mixed = 0;
     sw->empty = 1;
-#else
-    sw->ratio = ((int64_t) sw->info.freq << 32) / sw->hw->info.freq;
 #endif
 
     if (sw->info.is_float) {
@@ -264,13 +273,11 @@ static HW *glue(audio_pcm_hw_add_new_, TYPE)(AudioState *s,
         return NULL;
     }
 
-    hw = audio_calloc(__func__, 1, glue(drv->voice_size_, TYPE));
-    if (!hw) {
-        dolog ("Can not allocate voice `%s' size %d\n",
-               drv->name, glue (drv->voice_size_, TYPE));
-        return NULL;
-    }
-
+    /*
+     * Since glue(s->nb_hw_voices_, TYPE) is != 0, glue(drv->voice_size_, TYPE)
+     * is guaranteed to be != 0. See the audio_init_nb_voices_* functions.
+     */
+    hw = g_malloc0(glue(drv->voice_size_, TYPE));
     hw->s = s;
     hw->pcm_ops = drv->pcm_ops;
 
@@ -418,33 +425,28 @@ static SW *glue(audio_pcm_create_voice_pair_, TYPE)(
         hw_as = *as;
     }
 
-    sw = audio_calloc(__func__, 1, sizeof(*sw));
-    if (!sw) {
-        dolog ("Could not allocate soft voice `%s' (%zu bytes)\n",
-               sw_name ? sw_name : "unknown", sizeof (*sw));
-        goto err1;
-    }
+    sw = g_new0(SW, 1);
     sw->s = s;
 
     hw = glue(audio_pcm_hw_add_, TYPE)(s, &hw_as);
     if (!hw) {
-        goto err2;
+        dolog("Could not create a backend for voice `%s'\n", sw_name);
+        goto err1;
     }
 
     glue (audio_pcm_hw_add_sw_, TYPE) (hw, sw);
 
     if (glue (audio_pcm_sw_init_, TYPE) (sw, hw, sw_name, as)) {
-        goto err3;
+        goto err2;
     }
 
     return sw;
 
-err3:
+err2:
     glue (audio_pcm_hw_del_sw_, TYPE) (sw);
     glue (audio_pcm_hw_gc_, TYPE) (&hw);
-err2:
-    g_free (sw);
 err1:
+    g_free(sw);
     return NULL;
 }
 
@@ -515,8 +517,8 @@ SW *glue (AUD_open_, TYPE) (
         HW *hw = sw->hw;
 
         if (!hw) {
-            dolog ("Internal logic error voice `%s' has no hardware store\n",
-                   SW_NAME (sw));
+            dolog("Internal logic error: voice `%s' has no backend\n",
+                  SW_NAME(sw));
             goto fail;
         }
 
@@ -527,7 +529,6 @@ SW *glue (AUD_open_, TYPE) (
     } else {
         sw = glue(audio_pcm_create_voice_pair_, TYPE)(s, name, as);
         if (!sw) {
-            dolog ("Failed to create voice `%s'\n", name);
             return NULL;
         }
     }

audio/mixeng.c

@@ -414,12 +414,7 @@ struct rate {
  */
 void *st_rate_start (int inrate, int outrate)
 {
-    struct rate *rate = audio_calloc(__func__, 1, sizeof(*rate));
-
-    if (!rate) {
-        dolog ("Could not allocate resampler (%zu bytes)\n", sizeof (*rate));
-        return NULL;
-    }
+    struct rate *rate = g_new0(struct rate, 1);
 
     rate->opos = 0;
 
@@ -445,6 +440,86 @@ void st_rate_stop (void *opaque)
     g_free (opaque);
 }
 
+/**
+ * st_rate_frames_out() - returns the number of frames the resampling code
+ * generates from frames_in frames
+ *
+ * @opaque: pointer to struct rate
+ * @frames_in: number of frames
+ *
+ * When upsampling, there may be more than one correct result. In this case,
+ * the function returns the maximum number of output frames the resampling
+ * code can generate.
+ */
+uint32_t st_rate_frames_out(void *opaque, uint32_t frames_in)
+{
+    struct rate *rate = opaque;
+    uint64_t opos_end, opos_delta;
+    uint32_t ipos_end;
+    uint32_t frames_out;
+
+    if (rate->opos_inc == 1ULL << 32) {
+        return frames_in;
+    }
+
+    /* no output frame without at least one input frame */
+    if (!frames_in) {
+        return 0;
+    }
+
+    /* last frame read was at rate->ipos - 1 */
+    ipos_end = rate->ipos - 1 + frames_in;
+    opos_end = (uint64_t)ipos_end << 32;
+
+    /* last frame written was at rate->opos - rate->opos_inc */
+    if (opos_end + rate->opos_inc <= rate->opos) {
+        return 0;
+    }
+    opos_delta = opos_end - rate->opos + rate->opos_inc;
+    frames_out = opos_delta / rate->opos_inc;
+
+    return opos_delta % rate->opos_inc ? frames_out : frames_out - 1;
+}
+
+/**
+ * st_rate_frames_in() - returns the number of frames needed to
+ * get frames_out frames after resampling
+ *
+ * @opaque: pointer to struct rate
+ * @frames_out: number of frames
+ *
+ * When downsampling, there may be more than one correct result. In this
+ * case, the function returns the maximum number of input frames needed.
+ */
+uint32_t st_rate_frames_in(void *opaque, uint32_t frames_out)
+{
+    struct rate *rate = opaque;
+    uint64_t opos_start, opos_end;
+    uint32_t ipos_start, ipos_end;
+
+    if (rate->opos_inc == 1ULL << 32) {
+        return frames_out;
+    }
+
+    if (frames_out) {
+        opos_start = rate->opos;
+        ipos_start = rate->ipos;
+    } else {
+        uint64_t offset;
+
+        /* add offset = ceil(opos_inc) to opos and ipos to avoid an underflow */
+        offset = (rate->opos_inc + (1ULL << 32) - 1) & ~((1ULL << 32) - 1);
+        opos_start = rate->opos + offset;
+        ipos_start = rate->ipos + (offset >> 32);
+    }
+    /* last frame written was at opos_start - rate->opos_inc */
+    opos_end = opos_start - rate->opos_inc + rate->opos_inc * frames_out;
+    ipos_end = (opos_end >> 32) + 1;
+
+    /* last frame read was at ipos_start - 1 */
+    return ipos_end + 1 > ipos_start ? ipos_end + 1 - ipos_start : 0;
+}
+
 void mixeng_clear (struct st_sample *buf, int len)
 {
     memset (buf, 0, len * sizeof (struct st_sample));

audio/mixeng.h

@@ -52,6 +52,8 @@ void st_rate_flow(void *opaque, st_sample *ibuf, st_sample *obuf,
 void st_rate_flow_mix(void *opaque, st_sample *ibuf, st_sample *obuf,
                       size_t *isamp, size_t *osamp);
 void st_rate_stop (void *opaque);
+uint32_t st_rate_frames_out(void *opaque, uint32_t frames_in);
+uint32_t st_rate_frames_in(void *opaque, uint32_t frames_out);
 void mixeng_clear (struct st_sample *buf, int len);
 void mixeng_volume (struct st_sample *buf, int len, struct mixeng_volume *vol);
 

audio/rate_template.h

@@ -40,8 +40,6 @@ void NAME (void *opaque, struct st_sample *ibuf, struct st_sample *obuf,
     int64_t t;
 #endif
 
-    ilast = rate->ilast;
-
     istart = ibuf;
     iend = ibuf + *isamp;
 
@@ -59,15 +57,17 @@ void NAME (void *opaque, struct st_sample *ibuf, struct st_sample *obuf,
         return;
     }
 
-    while (obuf < oend) {
+    /* without input samples, there's nothing to do */
+    if (ibuf >= iend) {
+        *osamp = 0;
+        return;
+    }
 
-        /* Safety catch to make sure we have input samples.  */
-        if (ibuf >= iend) {
-            break;
-        }
+    ilast = rate->ilast;
+
+    while (true) {
 
         /* read as many input samples so that ipos > opos */
-
         while (rate->ipos <= (rate->opos >> 32)) {
             ilast = *ibuf++;
             rate->ipos++;
@@ -78,6 +78,11 @@ void NAME (void *opaque, struct st_sample *ibuf, struct st_sample *obuf,
             }
         }
 
+        /* make sure that the next output sample can be written */
+        if (obuf >= oend) {
+            break;
+        }
+
         icur = *ibuf;
 
         /* wrap ipos and opos around long before they overflow */