arm64: entry: fix NMI {user, kernel}->kernel transitions

Exceptions which can be taken at (almost) any time are consdiered to be
NMIs. On arm64 that includes:

* SDEI events
* GICv3 Pseudo-NMIs
* Kernel stack overflows
* Unexpected/unhandled exceptions

... but currently debug exceptions (BRKs, breakpoints, watchpoints,
single-step) are not considered NMIs.

As these can be taken at any time, kernel features (lockdep, RCU,
ftrace) may not be in a consistent kernel state. For example, we may
take an NMI from the idle code or partway through an entry/exit path.

While nmi_enter() and nmi_exit() handle most of this state, notably they
don't save/restore the lockdep state across an NMI being taken and
handled. When interrupts are enabled and an NMI is taken, lockdep may
see interrupts become disabled within the NMI code, but not see
interrupts become enabled when returning from the NMI, leaving lockdep
believing interrupts are disabled when they are actually disabled.

The x86 code handles this in idtentry_{enter,exit}_nmi(), which will
shortly be moved to the generic entry code. As we can't use either yet,
we copy the x86 approach in arm64-specific helpers. All the NMI
entrypoints are marked as noinstr to prevent any instrumentation
handling code being invoked before the state has been corrected.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Will Deacon <will@kernel.org>
Link: https://lore.kernel.org/r/20201130115950.22492-11-mark.rutland@arm.comSigned-off-by: Will Deacon <will@kernel.org>

arch/arm64/include/asm/exception.h

@@ -35,6 +35,8 @@ asmlinkage void noinstr enter_el1_irq_or_nmi(struct pt_regs *regs);
 asmlinkage void noinstr exit_el1_irq_or_nmi(struct pt_regs *regs);
 asmlinkage void enter_from_user_mode(void);
 asmlinkage void exit_to_user_mode(void);
+void arm64_enter_nmi(struct pt_regs *regs);
+void arm64_exit_nmi(struct pt_regs *regs);
 void do_mem_abort(unsigned long addr, unsigned int esr, struct pt_regs *regs);
 void do_undefinstr(struct pt_regs *regs);
 void do_bti(struct pt_regs *regs);

arch/arm64/kernel/entry-common.c

@@ -63,10 +63,40 @@ static void noinstr exit_to_kernel_mode(struct pt_regs *regs)
 	}
 }
 
+void noinstr arm64_enter_nmi(struct pt_regs *regs)
+{
+	regs->lockdep_hardirqs = lockdep_hardirqs_enabled();
+
+	__nmi_enter();
+	lockdep_hardirqs_off(CALLER_ADDR0);
+	lockdep_hardirq_enter();
+	rcu_nmi_enter();
+
+	trace_hardirqs_off_finish();
+	ftrace_nmi_enter();
+}
+
+void noinstr arm64_exit_nmi(struct pt_regs *regs)
+{
+	bool restore = regs->lockdep_hardirqs;
+
+	ftrace_nmi_exit();
+	if (restore) {
+		trace_hardirqs_on_prepare();
+		lockdep_hardirqs_on_prepare(CALLER_ADDR0);
+	}
+
+	rcu_nmi_exit();
+	lockdep_hardirq_exit();
+	if (restore)
+		lockdep_hardirqs_on(CALLER_ADDR0);
+	__nmi_exit();
+}
+
 asmlinkage void noinstr enter_el1_irq_or_nmi(struct pt_regs *regs)
 {
 	if (IS_ENABLED(CONFIG_ARM64_PSEUDO_NMI) && !interrupts_enabled(regs))
-		nmi_enter();
+		arm64_enter_nmi(regs);
 	else
 		enter_from_kernel_mode(regs);
 }
@@ -74,7 +104,7 @@ asmlinkage void noinstr enter_el1_irq_or_nmi(struct pt_regs *regs)
 asmlinkage void noinstr exit_el1_irq_or_nmi(struct pt_regs *regs)
 {
 	if (IS_ENABLED(CONFIG_ARM64_PSEUDO_NMI) && !interrupts_enabled(regs))
-		nmi_exit();
+		arm64_exit_nmi(regs);
 	else
 		exit_to_kernel_mode(regs);
 }

arch/arm64/kernel/sdei.c

@@ -10,6 +10,7 @@
 #include <linux/uaccess.h>
 
 #include <asm/alternative.h>
+#include <asm/exception.h>
 #include <asm/kprobes.h>
 #include <asm/mmu.h>
 #include <asm/ptrace.h>
@@ -223,16 +224,16 @@ static __kprobes unsigned long _sdei_handler(struct pt_regs *regs,
 }
 
 
-asmlinkage __kprobes notrace unsigned long
+asmlinkage noinstr unsigned long
 __sdei_handler(struct pt_regs *regs, struct sdei_registered_event *arg)
 {
 	unsigned long ret;
 
-	nmi_enter();
+	arm64_enter_nmi(regs);
 
 	ret = _sdei_handler(regs, arg);
 
-	nmi_exit();
+	arm64_exit_nmi(regs);
 
 	return ret;
 }

arch/arm64/kernel/traps.c

@@ -34,6 +34,7 @@
 #include <asm/daifflags.h>
 #include <asm/debug-monitors.h>
 #include <asm/esr.h>
+#include <asm/exception.h>
 #include <asm/extable.h>
 #include <asm/insn.h>
 #include <asm/kprobes.h>
@@ -753,8 +754,10 @@ const char *esr_get_class_string(u32 esr)
  * bad_mode handles the impossible case in the exception vector. This is always
  * fatal.
  */
-asmlinkage void bad_mode(struct pt_regs *regs, int reason, unsigned int esr)
+asmlinkage void notrace bad_mode(struct pt_regs *regs, int reason, unsigned int esr)
 {
+	arm64_enter_nmi(regs);
+
 	console_verbose();
 
 	pr_crit("Bad mode in %s handler detected on CPU%d, code 0x%08x -- %s\n",
@@ -786,7 +789,7 @@ void bad_el0_sync(struct pt_regs *regs, int reason, unsigned int esr)
 DEFINE_PER_CPU(unsigned long [OVERFLOW_STACK_SIZE/sizeof(long)], overflow_stack)
 	__aligned(16);
 
-asmlinkage void handle_bad_stack(struct pt_regs *regs)
+asmlinkage void noinstr handle_bad_stack(struct pt_regs *regs)
 {
 	unsigned long tsk_stk = (unsigned long)current->stack;
 	unsigned long irq_stk = (unsigned long)this_cpu_read(irq_stack_ptr);
@@ -794,6 +797,8 @@ asmlinkage void handle_bad_stack(struct pt_regs *regs)
 	unsigned int esr = read_sysreg(esr_el1);
 	unsigned long far = read_sysreg(far_el1);
 
+	arm64_enter_nmi(regs);
+
 	console_verbose();
 	pr_emerg("Insufficient stack space to handle exception!");
 
@@ -865,15 +870,15 @@ bool arm64_is_fatal_ras_serror(struct pt_regs *regs, unsigned int esr)
 	}
 }
 
-asmlinkage void do_serror(struct pt_regs *regs, unsigned int esr)
+asmlinkage void noinstr do_serror(struct pt_regs *regs, unsigned int esr)
 {
-	nmi_enter();
+	arm64_enter_nmi(regs);
 
 	/* non-RAS errors are not containable */
 	if (!arm64_is_ras_serror(esr) || arm64_is_fatal_ras_serror(regs, esr))
 		arm64_serror_panic(regs, esr);
 
-	nmi_exit();
+	arm64_exit_nmi(regs);
 }
 
 /* GENERIC_BUG traps */