Re: [PATCH v19 070/130] KVM: TDX: TDP MMU TDX support

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, May 28, 2024 at 05:55:19PM -0700,
Isaku Yamahata <isaku.yamahata@xxxxxxxxx> wrote:

> On Sun, May 26, 2024 at 04:45:15PM +0800,
> Chen Yu <yu.c.chen@xxxxxxxxx> wrote:
> 
> > On 2024-03-28 at 11:12:57 +0800, Chao Gao wrote:
> > > >+#if IS_ENABLED(CONFIG_HYPERV)
> > > >+static int vt_flush_remote_tlbs(struct kvm *kvm);
> > > >+#endif
> > > >+
> > > > static __init int vt_hardware_setup(void)
> > > > {
> > > > 	int ret;
> > > >@@ -49,11 +53,29 @@ static __init int vt_hardware_setup(void)
> > > > 		pr_warn_ratelimited("TDX requires mmio caching.  Please enable mmio caching for TDX.\n");
> > > > 	}
> > > > 
> > > >+#if IS_ENABLED(CONFIG_HYPERV)
> > > >+	/*
> > > >+	 * TDX KVM overrides flush_remote_tlbs method and assumes
> > > >+	 * flush_remote_tlbs_range = NULL that falls back to
> > > >+	 * flush_remote_tlbs.  Disable TDX if there are conflicts.
> > > >+	 */
> > > >+	if (vt_x86_ops.flush_remote_tlbs ||
> > > >+	    vt_x86_ops.flush_remote_tlbs_range) {
> > > >+		enable_tdx = false;
> > > >+		pr_warn_ratelimited("TDX requires baremetal. Not Supported on VMM guest.\n");
> > > >+	}
> > > >+#endif
> > > >+
> > > > 	enable_tdx = enable_tdx && !tdx_hardware_setup(&vt_x86_ops);
> > > > 	if (enable_tdx)
> > > > 		vt_x86_ops.vm_size = max_t(unsigned int, vt_x86_ops.vm_size,
> > > > 					   sizeof(struct kvm_tdx));
> > > > 
> > > >+#if IS_ENABLED(CONFIG_HYPERV)
> > > >+	if (enable_tdx)
> > > >+		vt_x86_ops.flush_remote_tlbs = vt_flush_remote_tlbs;
> > > 
> > > Is this hook necessary/beneficial to TDX?
> > >
> > 
> > I think so.
> > 
> > We happended to encounter the following error and breaks the boot up:
> > "SEAMCALL (0x000000000000000f) failed: 0xc0000b0800000001"
> > 0xc0000b0800000001 indicates the TDX_TLB_TRACKING_NOT_DONE, and it is caused
> > by page demotion but not yet doing a tlb shotdown by tlb track.
> > 
> > 
> > It was found on my system the CONFIG_HYPERV is not set, and it makes
> > kvm_arch_flush_remote_tlbs() not invoking tdx_track() before the
> > tdh_mem_page_demote(), which caused the problem.
> > 
> > > if no, we can leave .flush_remote_tlbs as NULL. if yes, we should do:
> > > 
> > > struct kvm_x86_ops {
> > > ...
> > > #if IS_ENABLED(CONFIG_HYPERV) || IS_ENABLED(TDX...)
> > > 	int  (*flush_remote_tlbs)(struct kvm *kvm);
> > > 	int  (*flush_remote_tlbs_range)(struct kvm *kvm, gfn_t gfn,
> > > 					gfn_t nr_pages);
> > > #endif
> > 
> > If the flush_remote_tlbs implementation are both available in HYPERV and TDX,
> > does it make sense to remove the config checks? I thought when commit 0277022a77a5
> > was introduced, the only user of flush_remote_tlbs() is hyperv, and now
> > there is TDX.
> 
> You don't like IS_ENABLED(CONFIG_HYPERV) || IS_ENABLED(CONFIG_TDX_HOST) in many
> places?  Then, we can do something like the followings.  Although It would be
> a bit ugly than the commit of 0277022a77a5, it's better to keep the intention
> of it.
> 

Ah, we have already __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE. We can use it.
-- 
Isaku Yamahata <isaku.yamahata@xxxxxxxxx>




[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux