All of lore.kernel.org
 help / color / mirror / Atom feed
From: Lai Jiangshan <jiangshanlai@gmail.com>
To: linux-kernel@vger.kernel.org
Cc: Lai Jiangshan <jiangshan.ljs@antgroup.com>,
	Hou Wenlong <houwenlong.hwl@antgroup.com>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Sean Christopherson <seanjc@google.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Borislav Petkov <bp@alien8.de>, Ingo Molnar <mingo@redhat.com>,
	kvm@vger.kernel.org, Paolo Bonzini <pbonzini@redhat.com>,
	x86@kernel.org, Kees Cook <keescook@chromium.org>,
	Juergen Gross <jgross@suse.com>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	"H. Peter Anvin" <hpa@zytor.com>
Subject: [RFC PATCH 32/73] KVM: x86/PVM: Enable guest debugging functions
Date: Mon, 26 Feb 2024 22:35:49 +0800	[thread overview]
Message-ID: <20240226143630.33643-33-jiangshanlai@gmail.com> (raw)
In-Reply-To: <20240226143630.33643-1-jiangshanlai@gmail.com>

From: Lai Jiangshan <jiangshan.ljs@antgroup.com>

The guest DR7 is loaded before VM enter to enable debugging functions
for the guest. If guest debugging is not enabled, the #DB and #BP
exceptions are reinjected into the guest directly; otherwise, they are
handled by the hypervisor.

However, DR7_GD is cleared since debug register read/write is a
privileged instruction, which always leads to a VM exit for #GP. The
address of breakpoints is limited to the allowed address range, similar
to the check in the #PF path.  Guest DR7 is loaded before VM enter to
enable debug function for guest.  If guest debug is not enabled, the #DB
and #BP are reinjected into guest directly, otherwise, they are handled
by hypervisor similar to VMX.

Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
Signed-off-by: Hou Wenlong <houwenlong.hwl@antgroup.com>
---
 arch/x86/kvm/pvm/pvm.c | 96 ++++++++++++++++++++++++++++++++++++++++++
 arch/x86/kvm/pvm/pvm.h |  3 ++
 2 files changed, 99 insertions(+)

diff --git a/arch/x86/kvm/pvm/pvm.c b/arch/x86/kvm/pvm/pvm.c
index 4ec8c2c514ca..299305903005 100644
--- a/arch/x86/kvm/pvm/pvm.c
+++ b/arch/x86/kvm/pvm/pvm.c
@@ -383,6 +383,8 @@ static void pvm_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 {
 	struct vcpu_pvm *pvm = to_pvm(vcpu);
 
+	pvm->host_debugctlmsr = get_debugctlmsr();
+
 	if (__this_cpu_read(active_pvm_vcpu) == pvm && vcpu->cpu == cpu)
 		return;
 
@@ -533,6 +535,9 @@ static int pvm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 	case MSR_IA32_SYSENTER_ESP:
 		msr_info->data = pvm->unused_MSR_IA32_SYSENTER_ESP;
 		break;
+	case MSR_IA32_DEBUGCTLMSR:
+		msr_info->data = 0;
+		break;
 	case MSR_PVM_VCPU_STRUCT:
 		msr_info->data = pvm->msr_vcpu_struct;
 		break;
@@ -619,6 +624,9 @@ static int pvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 	case MSR_IA32_SYSENTER_ESP:
 		pvm->unused_MSR_IA32_SYSENTER_ESP = data;
 		break;
+	case MSR_IA32_DEBUGCTLMSR:
+		/* It is ignored now. */
+		break;
 	case MSR_PVM_VCPU_STRUCT:
 		if (!PAGE_ALIGNED(data))
 			return 1;
@@ -810,6 +818,10 @@ static bool pvm_apic_init_signal_blocked(struct kvm_vcpu *vcpu)
 	return false;
 }
 
+static void update_exception_bitmap(struct kvm_vcpu *vcpu)
+{
+}
+
 static struct pvm_vcpu_struct *pvm_get_vcpu_struct(struct vcpu_pvm *pvm)
 {
 	struct gfn_to_pfn_cache *gpc = &pvm->pvcs_gpc;
@@ -1235,6 +1247,72 @@ static int pvm_vcpu_pre_run(struct kvm_vcpu *vcpu)
 	return 1;
 }
 
+static void pvm_sync_dirty_debug_regs(struct kvm_vcpu *vcpu)
+{
+	WARN_ONCE(1, "pvm never sets KVM_DEBUGREG_WONT_EXIT\n");
+}
+
+static void pvm_set_dr7(struct kvm_vcpu *vcpu, unsigned long val)
+{
+	to_pvm(vcpu)->guest_dr7 = val;
+}
+
+static __always_inline unsigned long __dr7_enable_mask(int drnum)
+{
+	unsigned long bp_mask = 0;
+
+	bp_mask |= (DR_LOCAL_ENABLE << (drnum * DR_ENABLE_SIZE));
+	bp_mask |= (DR_GLOBAL_ENABLE << (drnum * DR_ENABLE_SIZE));
+
+	return bp_mask;
+}
+
+static __always_inline unsigned long __dr7_mask(int drnum)
+{
+	unsigned long bp_mask = 0xf;
+
+	bp_mask <<= (DR_CONTROL_SHIFT + drnum * DR_CONTROL_SIZE);
+	bp_mask |= __dr7_enable_mask(drnum);
+
+	return bp_mask;
+}
+
+/*
+ * Calculate the correct dr7 for the hardware to avoid the host
+ * being watched.
+ *
+ * It only needs to be calculated each time when vcpu->arch.eff_db or
+ * pvm->guest_dr7 is changed.  But now it is calculated each time on
+ * VM-enter since there is no proper callback for vcpu->arch.eff_db and
+ * it is slow path.
+ */
+static __always_inline unsigned long pvm_eff_dr7(struct kvm_vcpu *vcpu)
+{
+	unsigned long eff_dr7 = to_pvm(vcpu)->guest_dr7;
+	int i;
+
+	/*
+	 * DR7_GD should not be set to hardware. And it doesn't need to be
+	 * set to hardware since PVM guest is running on hardware ring3.
+	 * All access to debug registers will be trapped and the emulation
+	 * code can handle DR7_GD correctly for PVM.
+	 */
+	eff_dr7 &= ~DR7_GD;
+
+	/*
+	 * Disallow addresses that are not for the guest, especially addresses
+	 * on the host entry code.
+	 */
+	for (i = 0; i < KVM_NR_DB_REGS; i++) {
+		if (!pvm_guest_allowed_va(vcpu, vcpu->arch.eff_db[i]))
+			eff_dr7 &= ~__dr7_mask(i);
+		if (!pvm_guest_allowed_va(vcpu, vcpu->arch.eff_db[i] + 7))
+			eff_dr7 &= ~__dr7_mask(i);
+	}
+
+	return eff_dr7;
+}
+
 // Save guest registers from host sp0 or IST stack.
 static __always_inline void save_regs(struct kvm_vcpu *vcpu, struct pt_regs *guest)
 {
@@ -1301,6 +1379,9 @@ static noinstr void pvm_vcpu_run_noinstr(struct kvm_vcpu *vcpu)
 	// Load guest registers into the host sp0 stack for switcher.
 	load_regs(vcpu, sp0_regs);
 
+	if (unlikely(pvm->guest_dr7 & DR7_BP_EN_MASK))
+		set_debugreg(pvm_eff_dr7(vcpu), 7);
+
 	// Call into switcher and enter guest.
 	ret_regs = switcher_enter_guest();
 
@@ -1309,6 +1390,11 @@ static noinstr void pvm_vcpu_run_noinstr(struct kvm_vcpu *vcpu)
 	pvm->exit_vector = (ret_regs->orig_ax >> 32);
 	pvm->exit_error_code = (u32)ret_regs->orig_ax;
 
+	// dr7 requires to be zero when the controling of debug registers
+	// passes back to the host.
+	if (unlikely(pvm->guest_dr7 & DR7_BP_EN_MASK))
+		set_debugreg(0, 7);
+
 	// handle noinstr vmexits reasons.
 	switch (pvm->exit_vector) {
 	case PF_VECTOR:
@@ -1387,8 +1473,15 @@ static fastpath_t pvm_vcpu_run(struct kvm_vcpu *vcpu)
 
 	pvm_set_host_cr3(pvm);
 
+	if (pvm->host_debugctlmsr)
+		update_debugctlmsr(0);
+
 	pvm_vcpu_run_noinstr(vcpu);
 
+	/* MSR_IA32_DEBUGCTLMSR is zeroed before vmenter. Restore it if needed */
+	if (pvm->host_debugctlmsr)
+		update_debugctlmsr(pvm->host_debugctlmsr);
+
 	if (is_smod(pvm)) {
 		struct pvm_vcpu_struct *pvcs = pvm->pvcs_gpc.khva;
 
@@ -1696,6 +1789,7 @@ static struct kvm_x86_ops pvm_x86_ops __initdata = {
 	.vcpu_load = pvm_vcpu_load,
 	.vcpu_put = pvm_vcpu_put,
 
+	.update_exception_bitmap = update_exception_bitmap,
 	.get_msr_feature = pvm_get_msr_feature,
 	.get_msr = pvm_get_msr,
 	.set_msr = pvm_set_msr,
@@ -1709,6 +1803,8 @@ static struct kvm_x86_ops pvm_x86_ops __initdata = {
 	.set_gdt = pvm_set_gdt,
 	.get_idt = pvm_get_idt,
 	.set_idt = pvm_set_idt,
+	.set_dr7 = pvm_set_dr7,
+	.sync_dirty_debug_regs = pvm_sync_dirty_debug_regs,
 	.get_rflags = pvm_get_rflags,
 	.set_rflags = pvm_set_rflags,
 	.get_if_flag = pvm_get_if_flag,
diff --git a/arch/x86/kvm/pvm/pvm.h b/arch/x86/kvm/pvm/pvm.h
index bf3a6a1837c0..4cdcbed1c813 100644
--- a/arch/x86/kvm/pvm/pvm.h
+++ b/arch/x86/kvm/pvm/pvm.h
@@ -37,6 +37,7 @@ struct vcpu_pvm {
 	unsigned long switch_flags;
 
 	u16 host_ds_sel, host_es_sel;
+	u64 host_debugctlmsr;
 
 	union {
 		unsigned long exit_extra;
@@ -52,6 +53,8 @@ struct vcpu_pvm {
 	int int_shadow;
 	bool nmi_mask;
 
+	unsigned long guest_dr7;
+
 	struct gfn_to_pfn_cache pvcs_gpc;
 
 	// emulated x86 msrs
-- 
2.19.1.6.gb485710b


  parent reply	other threads:[~2024-02-26 14:36 UTC|newest]

Thread overview: 82+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-02-26 14:35 [RFC PATCH 00/73] KVM: x86/PVM: Introduce a new hypervisor Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 01/73] KVM: Documentation: Add the specification for PVM Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 02/73] x86/ABI/PVM: Add PVM-specific ABI header file Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 03/73] x86/entry: Implement switcher for PVM VM enter/exit Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 04/73] x86/entry: Implement direct switching for the switcher Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 05/73] KVM: x86: Set 'vcpu->arch.exception.injected' as true before vendor callback Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 06/73] KVM: x86: Move VMX interrupt/nmi handling into kvm.ko Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 07/73] KVM: x86/mmu: Adapt shadow MMU for PVM Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 08/73] KVM: x86: Allow hypercall handling to not skip the instruction Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 09/73] KVM: x86: Add PVM virtual MSRs into emulated_msrs_all[] Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 10/73] KVM: x86: Introduce vendor feature to expose vendor-specific CPUID Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 11/73] KVM: x86: Implement gpc refresh for guest usage Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 12/73] KVM: x86: Add NR_VCPU_SREG in SREG enum Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 13/73] KVM: x86/emulator: Reinject #GP if instruction emulation failed for PVM Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 14/73] KVM: x86: Create stubs for PVM module as a new vendor Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 15/73] mm/vmalloc: Add a helper to reserve a contiguous and aligned kernel virtual area Lai Jiangshan
2024-02-27 14:56   ` Christoph Hellwig
2024-02-27 17:07     ` Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 16/73] KVM: x86/PVM: Implement host mmu initialization Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 17/73] KVM: x86/PVM: Implement module initialization related callbacks Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 18/73] KVM: x86/PVM: Implement VM/VCPU " Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 19/73] x86/entry: Export 32-bit ignore syscall entry and __ia32_enabled variable Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 20/73] KVM: x86/PVM: Implement vcpu_load()/vcpu_put() related callbacks Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 21/73] KVM: x86/PVM: Implement vcpu_run() callbacks Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 22/73] KVM: x86/PVM: Handle some VM exits before enable interrupts Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 23/73] KVM: x86/PVM: Handle event handling related MSR read/write operation Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 24/73] KVM: x86/PVM: Introduce PVM mode switching Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 25/73] KVM: x86/PVM: Implement APIC emulation related callbacks Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 26/73] KVM: x86/PVM: Implement event delivery flags " Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 27/73] KVM: x86/PVM: Implement event injection " Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 28/73] KVM: x86/PVM: Handle syscall from user mode Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 29/73] KVM: x86/PVM: Implement allowed range checking for #PF Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 30/73] KVM: x86/PVM: Implement segment related callbacks Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 31/73] KVM: x86/PVM: Implement instruction emulation for #UD and #GP Lai Jiangshan
2024-02-26 14:35 ` Lai Jiangshan [this message]
2024-02-26 14:35 ` [RFC PATCH 33/73] KVM: x86/PVM: Handle VM-exit due to hardware exceptions Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 34/73] KVM: x86/PVM: Handle ERETU/ERETS synthetic instruction Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 35/73] KVM: x86/PVM: Handle PVM_SYNTHETIC_CPUID " Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 36/73] KVM: x86/PVM: Handle KVM hypercall Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 37/73] KVM: x86/PVM: Use host PCID to reduce guest TLB flushing Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 38/73] KVM: x86/PVM: Handle hypercalls for privilege instruction emulation Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 39/73] KVM: x86/PVM: Handle hypercall for CR3 switching Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 40/73] KVM: x86/PVM: Handle hypercall for loading GS selector Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 41/73] KVM: x86/PVM: Allow to load guest TLS in host GDT Lai Jiangshan
2024-02-26 14:35 ` [RFC PATCH 42/73] KVM: x86/PVM: Support for kvm_exit() tracepoint Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 43/73] KVM: x86/PVM: Enable direct switching Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 44/73] KVM: x86/PVM: Implement TSC related callbacks Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 45/73] KVM: x86/PVM: Add dummy PMU " Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 46/73] KVM: x86/PVM: Support for CPUID faulting Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 47/73] KVM: x86/PVM: Handle the left supported MSRs in msrs_to_save_base[] Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 48/73] KVM: x86/PVM: Implement system registers setting callbacks Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 49/73] KVM: x86/PVM: Implement emulation for non-PVM mode Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 50/73] x86/tools/relocs: Cleanup cmdline options Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 51/73] x86/tools/relocs: Append relocations into input file Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 52/73] x86/boot: Allow to do relocation for uncompressed kernel Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 53/73] x86/pvm: Add Kconfig option and the CPU feature bit for PVM guest Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 54/73] x86/pvm: Detect PVM hypervisor support Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 55/73] x86/pvm: Relocate kernel image to specific virtual address range Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 56/73] x86/pvm: Relocate kernel image early in PVH entry Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 57/73] x86/pvm: Make cpu entry area and vmalloc area variable Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 58/73] x86/pvm: Relocate kernel address space layout Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 59/73] x86/pti: Force enabling KPTI for PVM guest Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 60/73] x86/pvm: Add event entry/exit and dispatch code Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 61/73] x86/pvm: Allow to install a system interrupt handler Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 62/73] x86/pvm: Add early kernel event entry and dispatch code Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 63/73] x86/pvm: Add hypercall support Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 64/73] x86/pvm: Enable PVM event delivery Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 65/73] x86/kvm: Patch KVM hypercall as PVM hypercall Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 66/73] x86/pvm: Use new cpu feature to describe XENPV and PVM Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 67/73] x86/pvm: Implement cpu related PVOPS Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 68/73] x86/pvm: Implement irq " Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 69/73] x86/pvm: Implement mmu " Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 70/73] x86/pvm: Don't use SWAPGS for gsbase read/write Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 71/73] x86/pvm: Adapt pushf/popf in this_cpu_cmpxchg16b_emu() Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 72/73] x86/pvm: Use RDTSCP as default in vdso_read_cpunode() Lai Jiangshan
2024-02-26 14:36 ` [RFC PATCH 73/73] x86/pvm: Disable some unsupported syscalls and features Lai Jiangshan
2024-02-26 14:49 ` [RFC PATCH 00/73] KVM: x86/PVM: Introduce a new hypervisor Paolo Bonzini
2024-02-27 17:27   ` Sean Christopherson
2024-02-29  9:33     ` David Woodhouse
2024-03-01 14:00     ` Lai Jiangshan
2024-02-29 14:55   ` Lai Jiangshan
2024-03-06 11:05 ` Like Xu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20240226143630.33643-33-jiangshanlai@gmail.com \
    --to=jiangshanlai@gmail.com \
    --cc=bp@alien8.de \
    --cc=dave.hansen@linux.intel.com \
    --cc=houwenlong.hwl@antgroup.com \
    --cc=hpa@zytor.com \
    --cc=jgross@suse.com \
    --cc=jiangshan.ljs@antgroup.com \
    --cc=keescook@chromium.org \
    --cc=kvm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=peterz@infradead.org \
    --cc=seanjc@google.com \
    --cc=tglx@linutronix.de \
    --cc=torvalds@linux-foundation.org \
    --cc=x86@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.