Received-SPF: pass (google.com: domain of linux-kernel+bounces-81721-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
From: Lai Jiangshan <jiangshanlai@gmail.com>
To: linux-kernel@vger.kernel.org
Cc: Hou Wenlong <houwenlong.hwl@antgroup.com>,
	Lai Jiangshan <jiangshan.ljs@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>,
	Sami Tolvanen <samitolvanen@google.com>,
	Fangrui Song <maskray@google.com>,
	Willy Tarreau <w@1wt.eu>,
	Thomas Garnier <thgarnie@chromium.org>,
	Josh Poimboeuf <jpoimboe@kernel.org>,
	Xin Li <xin3.li@intel.com>
Subject: [RFC PATCH 72/73] x86/pvm: Use RDTSCP as default in vdso_read_cpunode()
Date: Mon, 26 Feb 2024 22:36:29 +0800
Message-Id: <20240226143630.33643-73-jiangshanlai@gmail.com>
In-Reply-To: <20240226143630.33643-1-jiangshanlai@gmail.com>
References: <20240226143630.33643-1-jiangshanlai@gmail.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

From: Hou Wenlong <houwenlong.hwl@antgroup.com>

The CPUNODE description of the guest cannot be installed into the host's
GDT, as this index is also used for the host to retrieve the current CPU
in paranoid entry. As a result, LSL in vdso_read_cpunode() does not work
correctly for the PVM guest. To address this issue, use RDTSCP as the
default in vdso_read_cpunode(), as it is supported by the hypervisor.

Suggested-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
Signed-off-by: Hou Wenlong <houwenlong.hwl@antgroup.com>
Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
---
 arch/x86/include/asm/alternative.h | 14 ++++++++++++++
 arch/x86/include/asm/segment.h     | 14 ++++++++++----
 2 files changed, 24 insertions(+), 4 deletions(-)

diff --git a/arch/x86/include/asm/alternative.h b/arch/x86/include/asm/alternative.h
index cf4b236b47a3..caebb49c5d61 100644
--- a/arch/x86/include/asm/alternative.h
+++ b/arch/x86/include/asm/alternative.h
@@ -299,6 +299,20 @@ static inline int alternatives_text_reserved(void *start, void *end)
 	asm_inline volatile (ALTERNATIVE(oldinstr, newinstr, ft_flags)	\
 		: output : "i" (0), ## input)
 
+/*
+ * This is similar to alternative_io. But it has two features and
+ * respective instructions.
+ *
+ * If CPU has feature2, newinstr2 is used.
+ * Otherwise, if CPU has feature1, newinstr1 is used.
+ * Otherwise, oldinstr is used.
+ */
+#define alternative_io_2(oldinstr, newinstr1, ft_flags1, newinstr2,	     \
+			 ft_flags2, output, input...)			     \
+	asm_inline volatile (ALTERNATIVE_2(oldinstr, newinstr1, ft_flags1,   \
+		newinstr2, ft_flags2)					     \
+		: output : "i" (0), ## input)
+
 /* Like alternative_io, but for replacing a direct call with another one. */
 #define alternative_call(oldfunc, newfunc, ft_flags, output, input...)	\
 	asm_inline volatile (ALTERNATIVE("call %P[old]", "call %P[new]", ft_flags) \
diff --git a/arch/x86/include/asm/segment.h b/arch/x86/include/asm/segment.h
index 9d6411c65920..555966922e8f 100644
--- a/arch/x86/include/asm/segment.h
+++ b/arch/x86/include/asm/segment.h
@@ -253,11 +253,17 @@ static inline void vdso_read_cpunode(unsigned *cpu, unsigned *node)
 	 * hoisting it out of the calling function.
 	 *
 	 * If RDPID is available, use it.
+	 *
+	 * If it is PVM guest and RDPID is not available, use RDTSCP.
 	 */
-	alternative_io ("lsl %[seg],%[p]",
-			".byte 0xf3,0x0f,0xc7,0xf8", /* RDPID %eax/rax */
-			X86_FEATURE_RDPID,
-			[p] "=a" (p), [seg] "r" (__CPUNODE_SEG));
+	alternative_io_2("lsl %[seg],%[p]",
+			 ".byte 0x0f,0x01,0xf9\n\t" /* RDTSCP %eax:%edx, %ecx */
+			 "mov %%ecx,%%eax\n\t",
+			 X86_FEATURE_KVM_PVM_GUEST,
+			 ".byte 0xf3,0x0f,0xc7,0xf8", /* RDPID %eax/rax */
+			 X86_FEATURE_RDPID,
+			 [p] "=a" (p), [seg] "r" (__CPUNODE_SEG)
+			 : "cx", "dx");
 
 	if (cpu)
 		*cpu = (p & VDSO_CPUNODE_MASK);
-- 
2.19.1.6.gb485710b