device: 0.888
instruction: 0.722
graphic: 0.679
other: 0.660
semantic: 0.622
mistranslation: 0.589
KVM: 0.440
boot: 0.337
network: 0.290
socket: 0.279
vnc: 0.248
assembly: 0.131

Clock drift issue with -soundhw hda

Here's the scenario: I'm working on code for loopback audio recording (i.e. recording what you're hearing) using WASAPI on Windows. As I usually develop on Linux, I'm using qemu to test this on a Windows 10 VM. The heart of WASAPI audio recording is the IAudioCaptureClient::GetBuffer function (https://docs.microsoft.com/en-us/windows/win32/api/audioclient/nf-audioclient-iaudiocaptureclient-getbuffer). Among other things, this function produces a timestamp for when the audio buffer it returned is supposed to be played.

When the audio device in question is the qemu hda device, this timestamp is wrong.

There is a clock drift error (I measured it to be about 0.1%, i.e. 1ms drift every second = a full second after 16 minutes) that causes the audio clock to advance faster than the system clock. Paradoxically, this does not affect audio playback through qemu at all, no delay there. Only the timestamps returned to recording applications are completely bogus.

Unfortunately I'm not intimately familiar with the inner workings of Intel HD Audio. All I can tell you is that this timestamp is supposedly obtained directly from the hardware (which would be qemu in this case), which is also why e.g. chromium implements a workaround for buggy hardware that returns incorrect timestamps.

Here are the relevant parts of my command line (version 4.2.0):
-enable-kvm -machine pc-q35-3.1,kernel-irqchip=on -cpu host,kvm=off,hv_time,hv_relaxed,hv_vapic,hv_spinlocks=0x1fff,hv_vendor_id=NvidiaFuckU -rtc base=localtime -nodefaults -soundhw hda

Just wanted to let you know about this because it took me three days of utter confusion and frustration to figure this out.

The QEMU project is currently considering to move its bug tracking to
another system. For this we need to know which bugs are still valid
and which could be closed already. Thus we are setting older bugs to
"Incomplete" now.

If you still think this bug report here is valid, then please switch
the state back to "New" within the next 60 days, otherwise this report
will be marked as "Expired". Or please mark it as "Fix Released" if
the problem has been solved with a newer version of QEMU already.

Thank you and sorry for the inconvenience.


[Expired for QEMU because there has been no activity for 60 days.]