[SRU Focal 3/6] x86/mm: Randomize per-cpu entry area

[Cengiz Can]

From: Peter Zijlstra <peterz at infradead.org>

Seth found that the CPU-entry-area; the piece of per-cpu data that is
mapped into the userspace page-tables for kPTI is not subject to any
randomization -- irrespective of kASLR settings.

On x86_64 a whole P4D (512 GB) of virtual address space is reserved for
this structure, which is plenty large enough to randomize things a
little.

As such, use a straight forward randomization scheme that avoids
duplicates to spread the existing CPUs over the available space.

  [ bp: Fix le build. ]

Reported-by: Seth Jenkins <sethjenkins at google.com>
Reviewed-by: Kees Cook <keescook at chromium.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz at infradead.org>
Signed-off-by: Dave Hansen <dave.hansen at linux.intel.com>
Signed-off-by: Borislav Petkov <bp at suse.de>
(backported from commit 97e3d26b5e5f371b3ee223d94dd123e6c442ba80)
[cascardo: macros in a different file]
[cascardo: no checks in breakpoint]
[cascardo: get_cpu_entry_area is not noinstr]
[cascardo: include <linux/prandom.h>]
CVE-2023-0597
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo at canonical.com>
---
 arch/x86/include/asm/cpu_entry_area.h | 13 ++++----
 arch/x86/mm/cpu_entry_area.c          | 47 +++++++++++++++++++++++++--
 2 files changed, 52 insertions(+), 8 deletions(-)

diff --git a/arch/x86/include/asm/cpu_entry_area.h b/arch/x86/include/asm/cpu_entry_area.h
index 0d1d37d8b279..1a569407312f 100644
--- a/arch/x86/include/asm/cpu_entry_area.h
+++ b/arch/x86/include/asm/cpu_entry_area.h
@@ -111,10 +111,6 @@ struct cpu_entry_area {
 };
 
 #define CPU_ENTRY_AREA_SIZE		(sizeof(struct cpu_entry_area))
-#define CPU_ENTRY_AREA_ARRAY_SIZE	(CPU_ENTRY_AREA_SIZE * NR_CPUS)
-
-/* Total size includes the readonly IDT mapping page as well: */
-#define CPU_ENTRY_AREA_TOTAL_SIZE	(CPU_ENTRY_AREA_ARRAY_SIZE + PAGE_SIZE)
 
 DECLARE_PER_CPU(struct cpu_entry_area *, cpu_entry_area);
 DECLARE_PER_CPU(struct cea_exception_stacks *, cea_exception_stacks);
@@ -128,8 +124,13 @@ extern void cea_set_pte(void *cea_vaddr, phys_addr_t pa, pgprot_t flags);
 
 #define CPU_ENTRY_AREA_RO_IDT_VADDR	((void *)CPU_ENTRY_AREA_RO_IDT)
 
-#define CPU_ENTRY_AREA_MAP_SIZE			\
-	(CPU_ENTRY_AREA_PER_CPU + CPU_ENTRY_AREA_ARRAY_SIZE - CPU_ENTRY_AREA_BASE)
+#ifdef CONFIG_X86_32
+#define CPU_ENTRY_AREA_MAP_SIZE		(CPU_ENTRY_AREA_PER_CPU +		\
+					 (CPU_ENTRY_AREA_SIZE * NR_CPUS) -	\
+					 CPU_ENTRY_AREA_BASE)
+#else
+#define CPU_ENTRY_AREA_MAP_SIZE		P4D_SIZE
+#endif
 
 extern struct cpu_entry_area *get_cpu_entry_area(int cpu);
 
diff --git a/arch/x86/mm/cpu_entry_area.c b/arch/x86/mm/cpu_entry_area.c
index 102e1d63f99a..e7a201f3fa65 100644
--- a/arch/x86/mm/cpu_entry_area.c
+++ b/arch/x86/mm/cpu_entry_area.c
@@ -4,6 +4,7 @@
 #include <linux/percpu.h>
 #include <linux/kallsyms.h>
 #include <linux/kcore.h>
+#include <linux/prandom.h>
 
 #include <asm/cpu_entry_area.h>
 #include <asm/pgtable.h>
@@ -16,11 +17,52 @@ static DEFINE_PER_CPU_PAGE_ALIGNED(struct entry_stack_page, entry_stack_storage)
 #ifdef CONFIG_X86_64
 static DEFINE_PER_CPU_PAGE_ALIGNED(struct exception_stacks, exception_stacks);
 DEFINE_PER_CPU(struct cea_exception_stacks*, cea_exception_stacks);
+
+static DEFINE_PER_CPU_READ_MOSTLY(unsigned long, _cea_offset);
+
+static __always_inline unsigned int cea_offset(unsigned int cpu)
+{
+	return per_cpu(_cea_offset, cpu);
+}
+
+static __init void init_cea_offsets(void)
+{
+	unsigned int max_cea;
+	unsigned int i, j;
+
+	max_cea = (CPU_ENTRY_AREA_MAP_SIZE - PAGE_SIZE) / CPU_ENTRY_AREA_SIZE;
+
+	/* O(sodding terrible) */
+	for_each_possible_cpu(i) {
+		unsigned int cea;
+
+again:
+		cea = prandom_u32_max(max_cea);
+
+		for_each_possible_cpu(j) {
+			if (cea_offset(j) == cea)
+				goto again;
+
+			if (i == j)
+				break;
+		}
+
+		per_cpu(_cea_offset, i) = cea;
+	}
+}
+#else /* !X86_64 */
+DECLARE_PER_CPU_PAGE_ALIGNED(struct doublefault_stack, doublefault_stack);
+
+static __always_inline unsigned int cea_offset(unsigned int cpu)
+{
+	return cpu;
+}
+static inline void init_cea_offsets(void) { }
 #endif
 
 struct cpu_entry_area *get_cpu_entry_area(int cpu)
 {
-	unsigned long va = CPU_ENTRY_AREA_PER_CPU + cpu * CPU_ENTRY_AREA_SIZE;
+	unsigned long va = CPU_ENTRY_AREA_PER_CPU + cea_offset(cpu) * CPU_ENTRY_AREA_SIZE;
 	BUILD_BUG_ON(sizeof(struct cpu_entry_area) % PAGE_SIZE != 0);
 
 	return (struct cpu_entry_area *) va;
@@ -186,7 +228,6 @@ static __init void setup_cpu_entry_area_ptes(void)
 
 	/* The +1 is for the readonly IDT: */
 	BUILD_BUG_ON((CPU_ENTRY_AREA_PAGES+1)*PAGE_SIZE != CPU_ENTRY_AREA_MAP_SIZE);
-	BUILD_BUG_ON(CPU_ENTRY_AREA_TOTAL_SIZE != CPU_ENTRY_AREA_MAP_SIZE);
 	BUG_ON(CPU_ENTRY_AREA_BASE & ~PMD_MASK);
 
 	start = CPU_ENTRY_AREA_BASE;
@@ -202,6 +243,8 @@ void __init setup_cpu_entry_areas(void)
 {
 	unsigned int cpu;
 
+	init_cea_offsets();
+
 	setup_cpu_entry_area_ptes();
 
 	for_each_possible_cpu(cpu)
-- 
2.39.2