Re: [PATCH v19 038/130] KVM: TDX: create/destroy VM structure

[Huang, Kai]

On 4/04/2024 6:24 am, Yamahata, Isaku wrote:
> On Thu, Mar 28, 2024 at 12:33:35PM +1300,
> "Huang, Kai" <kai.huang@xxxxxxxxx> wrote:
>
> > > +	kvm_tdx->tdr_pa = tdr_pa;
> > > +
> > > +	for_each_online_cpu(i) {
> > > +		int pkg = topology_physical_package_id(i);
> > > +
> > > +		if (cpumask_test_and_set_cpu(pkg, packages))
> > > +			continue;
> > > +
> > > +		/*
> > > +		 * Program the memory controller in the package with an
> > > +		 * encryption key associated to a TDX private host key id
> > > +		 * assigned to this TDR.  Concurrent operations on same memory
> > > +		 * controller results in TDX_OPERAND_BUSY.  Avoid this race by
> > > +		 * mutex.
> > > +		 */
> >
> > IIUC the race can only happen when you are creating multiple TDX guests
> > simulatenously?  Please clarify this in the comment.
> >
> > And I even don't think you need all these TDX module details:
> >
> > 		/*
> > 		 * Concurrent run of TDH.MNG.KEY.CONFIG on the same
> > 		 * package resluts in TDX_OPERAND_BUSY.  When creating
> > 		 * multiple TDX guests simultaneously this can run
> > 		 * concurrently.  Take the per-package lock to
> > 		 * serialize.
> > 		 */
>
> As pointed by Chao, those mutex will be dropped.
> https://lore.kernel.org/kvm/ZfpwIespKy8qxWWE@chao-email/
> Also we would simplify cpu masks to track which package is online/offline,
> which cpu to use for each package somehow.

Please see my reply there.  I might be missing something, though.

> > > +		mutex_lock(&tdx_mng_key_config_lock[pkg]);
> > > +		ret = smp_call_on_cpu(i, tdx_do_tdh_mng_key_config,
> > > +				      &kvm_tdx->tdr_pa, true);
> > > +		mutex_unlock(&tdx_mng_key_config_lock[pkg]);
> > > +		if (ret)
> > > +			break;
> > > +	}
> > > +	cpus_read_unlock();
> > > +	free_cpumask_var(packages);
> > > +	if (ret) {
> > > +		i = 0;
> > > +		goto teardown;
> > > +	}
> > > +
> > > +	kvm_tdx->tdcs_pa = tdcs_pa;
> > > +	for (i = 0; i < tdx_info->nr_tdcs_pages; i++) {
> > > +		err = tdh_mng_addcx(kvm_tdx->tdr_pa, tdcs_pa[i]);
> > > +		if (err == TDX_RND_NO_ENTROPY) {
> > > +			/* Here it's hard to allow userspace to retry. */
> > > +			ret = -EBUSY;
> > > +			goto teardown;
> > > +		}
> > > +		if (WARN_ON_ONCE(err)) {
> > > +			pr_tdx_error(TDH_MNG_ADDCX, err, NULL);
> > > +			ret = -EIO;
> > > +			goto teardown;
> > > +		}
> > > +	}
> > > +
> > > +	/*
> > > +	 * Note, TDH_MNG_INIT cannot be invoked here.  TDH_MNG_INIT requires a dedicated
> > > +	 * ioctl() to define the configure CPUID values for the TD.
> > > +	 */
> >
> > Then, how about renaming this function to __tdx_td_create()?
>
> So do we want to rename also ioctl name for consistency?
> i.e. KVM_TDX_INIT_VM => KVM_TDX_CREATE_VM.

Hmm.. but this __tdx_td_create() (the __tdx_td_init() in this patch) is 
called via kvm_x86_ops->vm_init(), but not IOCTL()?

If I read correctly, only TDH.MNG.INIT is called via IOCTL(), in that 
sense it makes more sense to name the IOCTL() as KVM_TDX_INIT_VM.

> I don't have strong opinion those names. Maybe
> KVM_TDX_{INIT, CREATE, or CONFIG}_VM?
> And we can rename the function name to match it.