snd_mixer_selem_ask_playback_vol_dB calculates wrong volume with SND_CTL_TLVT_DB_MINMAX_MUTE

[pv]

snd_mixer_selem_ask_playback_dB_vol() and snd_mixer_selem_ask_playback_vol_dB() are inconsistent between each other when SND_CTL_TLVT_DB_MINMAX_MUTE is enabled. Similarly for snd_mixer_selem_get_playback_dB after setting volume dB_vol returned.

Version: alsa-lib-1.2.15.3-3.fc44.x86_64

In Pipewire, we have

if ((r = snd_mixer_selem_ask_playback_dB_vol(me, value_dB, rounding, &val)) < 0)
    return r;
if ((r = snd_mixer_selem_ask_playback_vol_dB(me, val, &dB)) < 0)
    return r;

In the problem case value_dB = -3259 and rounding = +1 but I'm seeing

(gdb) p	val
$55 = 1
(gdb) p	dB
$56 = -9999999

The val = 1 is correct (the device is muted only at HW volume 0), but dB from snd_mixer_selem_ask_playback_vol_dB is wrong, the device is not muted at volume 1. (The inconsistency results to Pipewire volume control being broken at low volume.)

The device in question is "1532:055e Razer USA, Ltd Razer Nommo V2 X" -- it is muted at lowest ALSA hw volume 0, and anything > 0 is not muted (so it needs the mixer_playback_min_mute quirk).

In snd_tlv_convert_to_dB:

snd_tlv_convert_to_dB (tlv=0x7b460cd21610, rangemin=0, rangemax=7248, volume=1, db_gain=0x7b260b455e40) at control/tlv.c:212
254                     if (*db_gain < mindb)
(gdb) p	type
$80 = 5
(gdb) p	mindb
$72 = -2837
(gdb) p	maxdb
$73 = -6

The *db_gain = (maxdb - mindb) * (volume - rangemin) / (rangemax - rangemin) + mindb results to *db_gain == mindb and the function returns with SND_CTL_TLV_DB_GAIN_MUTE which is wrong.

Reverting 41039b4 appears to fix it for me (get dB == -2837). Alternatively one could change snd_tlv_convert_from_dB to match so the rounding is right.

I don't understand the new logic, doesn't SND_CTL_TLVT_DB_MINMAX_MUTE mean that the lowest hardware volume value is muted? (Also, the ALSA 1/100 dB values round to zero earlier than the USB 1/256 dB ones.)

Link: https://gitlab.freedesktop.org/pipewire/pipewire/-/work_items/5235 (I also have access to similar device locally)

[perexg]

Thanks. I reverted 41039b4 and put examples to commit description. It seems like a wrong fix.

[qaqland]

Sorry for the noise, and thanks for catching this.