[PATCH v7 2/4] x86/syscalls: Architecture specific pre-usermode check

[Thomas Garnier]

Disable the generic pre-usermode check in favor of an optimized
implementation. This patch adds specific checks on user-mode return path
to make it faster and smaller.

The user-mode state check is added to the prepare_exit_to_usermode
function. This function is called before any user-mode return on 32-bit
and on the 64-bit syscall slowpath. For the 64-bit syscall fast path, an
assembly address limit check redirects to the slow path if the address
limit is different.

The TASK_SIZE_MAX define is moved to the pgtable_64_types header so it
can be used in assembly code.

Signed-off-by: Thomas Garnier <thgarnie at google.com>
---
Based on next-20170410
---
 arch/x86/Kconfig                        |  1 +
 arch/x86/entry/common.c                 |  3 +++
 arch/x86/entry/entry_64.S               |  8 ++++++++
 arch/x86/include/asm/pgtable_64_types.h | 11 +++++++++++
 arch/x86/include/asm/processor.h        | 11 -----------
 5 files changed, 23 insertions(+), 11 deletions(-)

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index c43f47622440..74185795eb24 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -63,6 +63,7 @@ config X86
 	select ARCH_MIGHT_HAVE_ACPI_PDC		if ACPI
 	select ARCH_MIGHT_HAVE_PC_PARPORT
 	select ARCH_MIGHT_HAVE_PC_SERIO
+	select ARCH_NO_SYSCALL_VERIFY_PRE_USERMODE_STATE
 	select ARCH_SUPPORTS_ATOMIC_RMW
 	select ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT
 	select ARCH_SUPPORTS_NUMA_BALANCING	if X86_64
diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
index cdefcfdd9e63..76ef050255c9 100644
--- a/arch/x86/entry/common.c
+++ b/arch/x86/entry/common.c
@@ -23,6 +23,7 @@
 #include <linux/user-return-notifier.h>
 #include <linux/uprobes.h>
 #include <linux/livepatch.h>
+#include <linux/syscalls.h>
 
 #include <asm/desc.h>
 #include <asm/traps.h>
@@ -183,6 +184,8 @@ __visible inline void prepare_exit_to_usermode(struct pt_regs *regs)
 	struct thread_info *ti = current_thread_info();
 	u32 cached_flags;
 
+	verify_pre_usermode_state();
+
 	if (IS_ENABLED(CONFIG_PROVE_LOCKING) && WARN_ON(!irqs_disabled()))
 		local_irq_disable();
 
diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
index 607d72c4a485..62aca6dcdaf4 100644
--- a/arch/x86/entry/entry_64.S
+++ b/arch/x86/entry/entry_64.S
@@ -218,6 +218,14 @@ entry_SYSCALL_64_fastpath:
 	testl	$_TIF_ALLWORK_MASK, TASK_TI_flags(%r11)
 	jnz	1f
 
+	/*
+	 * If address limit is not based on user-mode, jump to slow path for
+	 * additional security checks.
+	 */
+	movq	$TASK_SIZE_MAX, %rcx
+	cmp	%rcx, TASK_addr_limit(%r11)
+	jne	1f
+
 	LOCKDEP_SYS_EXIT
 	TRACE_IRQS_ON		/* user mode is traced as IRQs on */
 	movq	RIP(%rsp), %rcx
diff --git a/arch/x86/include/asm/pgtable_64_types.h b/arch/x86/include/asm/pgtable_64_types.h
index 06470da156ba..78af4d43a7ce 100644
--- a/arch/x86/include/asm/pgtable_64_types.h
+++ b/arch/x86/include/asm/pgtable_64_types.h
@@ -104,4 +104,15 @@ typedef struct { pteval_t pte; } pte_t;
 
 #define EARLY_DYNAMIC_PAGE_TABLES	64
 
+/*
+ * User space process size. 47bits minus one guard page.  The guard
+ * page is necessary on Intel CPUs: if a SYSCALL instruction is at
+ * the highest possible canonical userspace address, then that
+ * syscall will enter the kernel with a non-canonical return
+ * address, and SYSRET will explode dangerously.  We avoid this
+ * particular problem by preventing anything from being mapped
+ * at the maximum canonical address.
+ */
+#define TASK_SIZE_MAX	((_AC(1, UL) << 47) - PAGE_SIZE)
+
 #endif /* _ASM_X86_PGTABLE_64_DEFS_H */
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 3cada998a402..e80822582d3e 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -825,17 +825,6 @@ static inline void spin_lock_prefetch(const void *x)
 #define KSTK_ESP(task)		(task_pt_regs(task)->sp)
 
 #else
-/*
- * User space process size. 47bits minus one guard page.  The guard
- * page is necessary on Intel CPUs: if a SYSCALL instruction is at
- * the highest possible canonical userspace address, then that
- * syscall will enter the kernel with a non-canonical return
- * address, and SYSRET will explode dangerously.  We avoid this
- * particular problem by preventing anything from being mapped
- * at the maximum canonical address.
- */
-#define TASK_SIZE_MAX	((1UL << 47) - PAGE_SIZE)
-
 /* This decides where the kernel will search for a free chunk of vm
  * space during mmap's.
  */
-- 
2.12.2.715.g7642488e1d-goog