Re: [PATCH v3 28/46] kmsan: entry: handle register passing from uninstrumented code

[Thomas Gleixner <tglx@xxxxxxxxxxxxx>]

On Thu, May 12 2022 at 18:17, Thomas Gleixner wrote:
> On Thu, May 12 2022 at 14:24, Alexander Potapenko wrote:
>> We could try to figure out the places in idtentry code where normal
>> kmsan_unpoison_memory() can be called in IRQ context, but as far as I
>> can see it will depend on the type of the entry point.
>
> NMI is covered as it increments before it invokes the unpoison().
>
> Let me figure out why we increment the preempt count late for
> interrupts. IIRC it's for symmetry reasons related to softirq processing
> on return, but let me double check.

It's even documented:

 https://www.kernel.org/doc/html/latest/core-api/entry.html#interrupts-and-regular-exceptions

But who reads documentation? :)

So, I think the simplest and least intrusive solution is to have special
purpose unpoison functions. See the patch below for illustration.

The reasons why I used specific ones:

  1) User entry

     Whether that's a syscall or interrupt/exception does not
     matter. It's always on the task stack and your machinery cannot be
     running at that point because it came from user space.

  2) Interrupt/exception/NMI entry kernel
  
     Those can nest into an already active context, so you really want
     to unpoison @regs.

     Also while regular interrupts cannot nest because of interrupts
     staying disabled, exceptions triggered in the interrupt handler and
     NMIs can nest.

     -> device interrupt()
           irqentry_enter(regs)

        -> NMI()
           irqentry_nmi_enter(regs)

           -> fault()
              irqentry_enter(regs)
          
              --> debug_exception()
                  irqentry_nmi_enter(regs)

     Soft interrupt processing on return from interrupt makes it more
     interesting:

     interrupt()
       handler()
       do_softirq()
         local_irq_enable()
            interrupt()
              NMI
                ....

     And everytime you get a new @regs pointer to deal with.

Wonderful, isn't it?

Thanks,

        tglx

---
--- a/kernel/entry/common.c
+++ b/kernel/entry/common.c
@@ -24,6 +24,7 @@ static __always_inline void __enter_from
 	user_exit_irqoff();
 
 	instrumentation_begin();
+	unpoison_user(regs);
 	trace_hardirqs_off_finish();
 	instrumentation_end();
 }
@@ -352,6 +353,7 @@ noinstr irqentry_state_t irqentry_enter(
 		lockdep_hardirqs_off(CALLER_ADDR0);
 		rcu_irq_enter();
 		instrumentation_begin();
+		unpoison_irq(regs);
 		trace_hardirqs_off_finish();
 		instrumentation_end();
 
@@ -367,6 +369,7 @@ noinstr irqentry_state_t irqentry_enter(
 	 */
 	lockdep_hardirqs_off(CALLER_ADDR0);
 	instrumentation_begin();
+	unpoison_irq(regs);
 	rcu_irq_enter_check_tick();
 	trace_hardirqs_off_finish();
 	instrumentation_end();
@@ -452,6 +455,7 @@ irqentry_state_t noinstr irqentry_nmi_en
 	rcu_nmi_enter();
 
 	instrumentation_begin();
+	unpoison_irq(regs);
 	trace_hardirqs_off_finish();
 	ftrace_nmi_enter();
 	instrumentation_end();