On 2/9/21 2:20 AM, Andrea Bolognani wrote:
On Mon, 2021-02-08 at 17:30 -0700, Jim Fehlig wrote:Hi All, I received a private bug report that starting a VM with swtpm device fails with qemu-system-aarch64. The VM config has <tpm model='tpm-tis'> <backend type='emulator' version='2.0'/> </tpm> QEMU reports the following error error: internal error: process exited while connecting to monitor: 2021-02-07T05:15:35.378927Z qemu-system-aarch64: -device tpm-tis,tpmdev=tpm-tpm0,id=tpm0: 'tpm-tis' is not a valid device model name Indeed it appears the device name is 'tpm-tis-device' [1][2] for ARM virt. The similar yet different device naming between x86 and ARM is unfortunate. IIUC, the devices are the same between the architectures, i.e. they both emulate a TPM device conforming to the TIS spec. If so, adding a 'tpm-tis-device' to virDomainTPMModel seems redundant. I assume the naming difference should be handled internally, e.g. something like the below patch to qemuBuildTPMDevStr? I can send a proper patch if this is an acceptable approach. Regards, Jim [1] https://qemu.readthedocs.io/en/latest/specs/tpm.html [2] https://github.com/qemu/qemu/commit/c294ac327ca99342b90bd3a83d2cef9b447afaa7 diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index 92036d26c0..5154611ccd 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -9198,6 +9198,9 @@ qemuBuildTPMDevStr(const virDomainDef *def, g_auto(virBuffer) buf = VIR_BUFFER_INITIALIZER; const char *model = virDomainTPMModelTypeToString(tpm->model); + if (tpm->model == VIR_DOMAIN_TPM_MODEL_TIS && def->os.arch == VIR_ARCH_AARCH64) + model = "tpm-tis-device"; + virBufferAsprintf(&buf, "%s,tpmdev=tpm-%s,id=%s", model, tpm->info.alias, tpm->info.alias);This looks fairly reasonable. Please make sure you also include a test case for this usage.
Thanks for taking a look! I've included a test case.
If you're feeling fancy, you might consider augmenting qemuValidateDomainDeviceDefTPM() to ensure VIR_DOMAIN_TPM_MODEL_TIS is only used on x86 or with the aarch64 virt machine type.
I've done this as well and provided packages to the reporter to test in his environment. I'll send the patches to the list after hearing the test results.
Regards, Jim
