Merge tag 'x86_cpu_for_v6.1_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 cpu updates from Borislav Petkov:

 - Print the CPU number at segfault time.

   The number printed is not always accurate (preemption is enabled at
   that time) but the print string contains "likely" and after a lot of
   back'n'forth on this, this was the consensus that was reached. See
   thread at [1].

 - After a *lot* of testing and polishing, finally the clear_user()
   improvements to inline REP; STOSB by default

Link: https://lore.kernel.org/r/5d62c1d0-7425-d5bb-ecb5-1dc3b4d7d245@intel.com [1]

* tag 'x86_cpu_for_v6.1_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/mm: Print likely CPU at segfault time
  x86/clear_user: Make it faster

5 files changed, 195 insertions, 43 deletions

diff --git a/arch/x86/include/asm/uaccess.h b/arch/x86/include/asm/uaccess.h
index 913e593a3b45..c46207946e05 100644
--- a/arch/x86/include/asm/uaccess.h
+++ b/arch/x86/include/asm/uaccess.h
@@ -502,9 +502,6 @@ strncpy_from_user(char *dst, const char __user *src, long count);
 
 extern __must_check long strnlen_user(const char __user *str, long n);
 
-unsigned long __must_check clear_user(void __user *mem, unsigned long len);
-unsigned long __must_check __clear_user(void __user *mem, unsigned long len);
-
 #ifdef CONFIG_ARCH_HAS_COPY_MC
 unsigned long __must_check
 copy_mc_to_kernel(void *to, const void *from, unsigned len);
@@ -526,6 +523,8 @@ extern struct movsl_mask {
 #define ARCH_HAS_NOCACHE_UACCESS 1
 
 #ifdef CONFIG_X86_32
+unsigned long __must_check clear_user(void __user *mem, unsigned long len);
+unsigned long __must_check __clear_user(void __user *mem, unsigned long len);
 # include <asm/uaccess_32.h>
 #else
 # include <asm/uaccess_64.h>
diff --git a/arch/x86/include/asm/uaccess_64.h b/arch/x86/include/asm/uaccess_64.h
index 45697e04d771..d13d71af5cf6 100644
--- a/arch/x86/include/asm/uaccess_64.h
+++ b/arch/x86/include/asm/uaccess_64.h
@@ -79,4 +79,49 @@ __copy_from_user_flushcache(void *dst, const void __user *src, unsigned size)
 	kasan_check_write(dst, size);
 	return __copy_user_flushcache(dst, src, size);
 }
+
+/*
+ * Zero Userspace.
+ */
+
+__must_check unsigned long
+clear_user_original(void __user *addr, unsigned long len);
+__must_check unsigned long
+clear_user_rep_good(void __user *addr, unsigned long len);
+__must_check unsigned long
+clear_user_erms(void __user *addr, unsigned long len);
+
+static __always_inline __must_check unsigned long __clear_user(void __user *addr, unsigned long size)
+{
+	might_fault();
+	stac();
+
+	/*
+	 * No memory constraint because it doesn't change any memory gcc
+	 * knows about.
+	 */
+	asm volatile(
+		"1:\n\t"
+		ALTERNATIVE_3("rep stosb",
+			      "call clear_user_erms",	  ALT_NOT(X86_FEATURE_FSRM),
+			      "call clear_user_rep_good", ALT_NOT(X86_FEATURE_ERMS),
+			      "call clear_user_original", ALT_NOT(X86_FEATURE_REP_GOOD))
+		"2:\n"
+	       _ASM_EXTABLE_UA(1b, 2b)
+	       : "+c" (size), "+D" (addr), ASM_CALL_CONSTRAINT
+	       : "a" (0)
+		/* rep_good clobbers %rdx */
+	       : "rdx");
+
+	clac();
+
+	return size;
+}
+
+static __always_inline unsigned long clear_user(void __user *to, unsigned long n)
+{
+	if (access_ok(to, n))
+		return __clear_user(to, n);
+	return n;
+}
 #endif /* _ASM_X86_UACCESS_64_H */
diff --git a/arch/x86/lib/clear_page_64.S b/arch/x86/lib/clear_page_64.S
index fe59b8ac4fcc..ecbfb4dd3b01 100644
--- a/arch/x86/lib/clear_page_64.S
+++ b/arch/x86/lib/clear_page_64.S
@@ -1,5 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0-only */
 #include <linux/linkage.h>
+#include <asm/asm.h>
 #include <asm/export.h>
 
 /*
@@ -50,3 +51,140 @@ SYM_FUNC_START(clear_page_erms)
 	RET
 SYM_FUNC_END(clear_page_erms)
 EXPORT_SYMBOL_GPL(clear_page_erms)
+
+/*
+ * Default clear user-space.
+ * Input:
+ * rdi destination
+ * rcx count
+ *
+ * Output:
+ * rcx: uncleared bytes or 0 if successful.
+ */
+SYM_FUNC_START(clear_user_original)
+	/*
+	 * Copy only the lower 32 bits of size as that is enough to handle the rest bytes,
+	 * i.e., no need for a 'q' suffix and thus a REX prefix.
+	 */
+	mov %ecx,%eax
+	shr $3,%rcx
+	jz .Lrest_bytes
+
+	# do the qwords first
+	.p2align 4
+.Lqwords:
+	movq $0,(%rdi)
+	lea 8(%rdi),%rdi
+	dec %rcx
+	jnz .Lqwords
+
+.Lrest_bytes:
+	and $7,  %eax
+	jz .Lexit
+
+	# now do the rest bytes
+.Lbytes:
+	movb $0,(%rdi)
+	inc %rdi
+	dec %eax
+	jnz .Lbytes
+
+.Lexit:
+	/*
+	 * %rax still needs to be cleared in the exception case because this function is called
+	 * from inline asm and the compiler expects %rax to be zero when exiting the inline asm,
+	 * in case it might reuse it somewhere.
+	 */
+        xor %eax,%eax
+        RET
+
+.Lqwords_exception:
+        # convert remaining qwords back into bytes to return to caller
+        shl $3, %rcx
+        and $7, %eax
+        add %rax,%rcx
+        jmp .Lexit
+
+.Lbytes_exception:
+        mov %eax,%ecx
+        jmp .Lexit
+
+        _ASM_EXTABLE_UA(.Lqwords, .Lqwords_exception)
+        _ASM_EXTABLE_UA(.Lbytes, .Lbytes_exception)
+SYM_FUNC_END(clear_user_original)
+EXPORT_SYMBOL(clear_user_original)
+
+/*
+ * Alternative clear user-space when CPU feature X86_FEATURE_REP_GOOD is
+ * present.
+ * Input:
+ * rdi destination
+ * rcx count
+ *
+ * Output:
+ * rcx: uncleared bytes or 0 if successful.
+ */
+SYM_FUNC_START(clear_user_rep_good)
+	# call the original thing for less than a cacheline
+	cmp $64, %rcx
+	jb clear_user_original
+
+.Lprep:
+	# copy lower 32-bits for rest bytes
+	mov %ecx, %edx
+	shr $3, %rcx
+	jz .Lrep_good_rest_bytes
+
+.Lrep_good_qwords:
+	rep stosq
+
+.Lrep_good_rest_bytes:
+	and $7, %edx
+	jz .Lrep_good_exit
+
+.Lrep_good_bytes:
+	mov %edx, %ecx
+	rep stosb
+
+.Lrep_good_exit:
+	# see .Lexit comment above
+	xor %eax, %eax
+	RET
+
+.Lrep_good_qwords_exception:
+	# convert remaining qwords back into bytes to return to caller
+	shl $3, %rcx
+	and $7, %edx
+	add %rdx, %rcx
+	jmp .Lrep_good_exit
+
+	_ASM_EXTABLE_UA(.Lrep_good_qwords, .Lrep_good_qwords_exception)
+	_ASM_EXTABLE_UA(.Lrep_good_bytes, .Lrep_good_exit)
+SYM_FUNC_END(clear_user_rep_good)
+EXPORT_SYMBOL(clear_user_rep_good)
+
+/*
+ * Alternative clear user-space when CPU feature X86_FEATURE_ERMS is present.
+ * Input:
+ * rdi destination
+ * rcx count
+ *
+ * Output:
+ * rcx: uncleared bytes or 0 if successful.
+ *
+ */
+SYM_FUNC_START(clear_user_erms)
+	# call the original thing for less than a cacheline
+	cmp $64, %rcx
+	jb clear_user_original
+
+.Lerms_bytes:
+	rep stosb
+
+.Lerms_exit:
+	xorl %eax,%eax
+	RET
+
+	_ASM_EXTABLE_UA(.Lerms_bytes, .Lerms_exit)
+SYM_FUNC_END(clear_user_erms)
+EXPORT_SYMBOL(clear_user_erms)
diff --git a/arch/x86/lib/usercopy_64.c b/arch/x86/lib/usercopy_64.c
index 0ae6cf804197..6c1f8ac5e721 100644
--- a/arch/x86/lib/usercopy_64.c
+++ b/arch/x86/lib/usercopy_64.c
@@ -14,46 +14,6 @@
  * Zero Userspace
  */
 
-unsigned long __clear_user(void __user *addr, unsigned long size)
-{
-	long __d0;
-	might_fault();
-	/* no memory constraint because it doesn't change any memory gcc knows
-	   about */
-	stac();
-	asm volatile(
-		"	testq  %[size8],%[size8]\n"
-		"	jz     4f\n"
-		"	.align 16\n"
-		"0:	movq $0,(%[dst])\n"
-		"	addq   $8,%[dst]\n"
-		"	decl %%ecx ; jnz   0b\n"
-		"4:	movq  %[size1],%%rcx\n"
-		"	testl %%ecx,%%ecx\n"
-		"	jz     2f\n"
-		"1:	movb   $0,(%[dst])\n"
-		"	incq   %[dst]\n"
-		"	decl %%ecx ; jnz  1b\n"
-		"2:\n"
-
-		_ASM_EXTABLE_TYPE_REG(0b, 2b, EX_TYPE_UCOPY_LEN8, %[size1])
-		_ASM_EXTABLE_UA(1b, 2b)
-
-		: [size8] "=&c"(size), [dst] "=&D" (__d0)
-		: [size1] "r"(size & 7), "[size8]" (size / 8), "[dst]"(addr));
-	clac();
-	return size;
-}
-EXPORT_SYMBOL(__clear_user);
-
-unsigned long clear_user(void __user *to, unsigned long n)
-{
-	if (access_ok(to, n))
-		return __clear_user(to, n);
-	return n;
-}
-EXPORT_SYMBOL(clear_user);
-
 #ifdef CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE
 /**
  * clean_cache_range - write back a cache range with CLWB
diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index fa71a5d12e87..a498ae1fbe66 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -769,6 +769,8 @@ show_signal_msg(struct pt_regs *regs, unsigned long error_code,
 		unsigned long address, struct task_struct *tsk)
 {
 	const char *loglvl = task_pid_nr(tsk) > 1 ? KERN_INFO : KERN_EMERG;
+	/* This is a racy snapshot, but it's better than nothing. */
+	int cpu = raw_smp_processor_id();
 
 	if (!unhandled_signal(tsk, SIGSEGV))
 		return;
@@ -782,6 +784,14 @@ show_signal_msg(struct pt_regs *regs, unsigned long error_code,
 
 	print_vma_addr(KERN_CONT " in ", regs->ip);
 
+	/*
+	 * Dump the likely CPU where the fatal segfault happened.
+	 * This can help identify faulty hardware.
+	 */
+	printk(KERN_CONT " likely on CPU %d (core %d, socket %d)", cpu,
+	       topology_core_id(cpu), topology_physical_package_id(cpu));
+
+
 	printk(KERN_CONT "\n");
 
 	show_opcodes(regs, loglvl);