x86/mm: Don't reenter flush_tlb_func_common()
authorAndy Lutomirski <luto@kernel.org>
Thu, 29 Jun 2017 15:53:13 +0000 (08:53 -0700)
committerIngo Molnar <mingo@kernel.org>
Fri, 30 Jun 2017 08:12:35 +0000 (10:12 +0200)
It was historically possible to have two concurrent TLB flushes
targetting the same CPU: one initiated locally and one initiated
remotely.  This can now cause an OOPS in leave_mm() at
arch/x86/mm/tlb.c:47:

        if (this_cpu_read(cpu_tlbstate.state) == TLBSTATE_OK)
                BUG();

with this call trace:
 flush_tlb_func_local arch/x86/mm/tlb.c:239 [inline]
 flush_tlb_mm_range+0x26d/0x370 arch/x86/mm/tlb.c:317

Without reentrancy, this OOPS is impossible: leave_mm() is only
called if we're not in TLBSTATE_OK, but then we're unexpectedly
in TLBSTATE_OK in leave_mm().

This can be caused by flush_tlb_func_remote() happening between
the two checks and calling leave_mm(), resulting in two consecutive
leave_mm() calls on the same CPU with no intervening switch_mm()
calls.

We never saw this OOPS before because the old leave_mm()
implementation didn't put us back in TLBSTATE_OK, so the assertion
didn't fire.

Nadav noticed the reentrancy issue in a different context, but
neither of us realized that it caused a problem yet.

Reported-by: Levin, Alexander (Sasha Levin) <alexander.levin@verizon.com>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Reviewed-by: Nadav Amit <nadav.amit@gmail.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: linux-mm@kvack.org
Fixes: 3d28ebceaffa ("x86/mm: Rework lazy TLB to track the actual loaded mm")
Link: http://lkml.kernel.org/r/855acf733268d521c9f2e191faee2dcc23a29729.1498751203.git.luto@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/mm/tlb.c

index b2485d69f7c2c593d921bc2267c2b898ae0a4da6..1cc47838d1e817b3579f7b78c64ee5ead263ffb4 100644 (file)
@@ -192,6 +192,9 @@ void switch_mm_irqs_off(struct mm_struct *prev, struct mm_struct *next,
 static void flush_tlb_func_common(const struct flush_tlb_info *f,
                                  bool local, enum tlb_flush_reason reason)
 {
+       /* This code cannot presently handle being reentered. */
+       VM_WARN_ON(!irqs_disabled());
+
        if (this_cpu_read(cpu_tlbstate.state) != TLBSTATE_OK) {
                leave_mm(smp_processor_id());
                return;
@@ -297,8 +300,13 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start,
                info.end = TLB_FLUSH_ALL;
        }
 
-       if (mm == this_cpu_read(cpu_tlbstate.loaded_mm))
+       if (mm == this_cpu_read(cpu_tlbstate.loaded_mm)) {
+               VM_WARN_ON(irqs_disabled());
+               local_irq_disable();
                flush_tlb_func_local(&info, TLB_LOCAL_MM_SHOOTDOWN);
+               local_irq_enable();
+       }
+
        if (cpumask_any_but(mm_cpumask(mm), cpu) < nr_cpu_ids)
                flush_tlb_others(mm_cpumask(mm), &info);
        put_cpu();
@@ -354,8 +362,13 @@ void arch_tlbbatch_flush(struct arch_tlbflush_unmap_batch *batch)
 
        int cpu = get_cpu();
 
-       if (cpumask_test_cpu(cpu, &batch->cpumask))
+       if (cpumask_test_cpu(cpu, &batch->cpumask)) {
+               VM_WARN_ON(irqs_disabled());
+               local_irq_disable();
                flush_tlb_func_local(&info, TLB_LOCAL_SHOOTDOWN);
+               local_irq_enable();
+       }
+
        if (cpumask_any_but(&batch->cpumask, cpu) < nr_cpu_ids)
                flush_tlb_others(&batch->cpumask, &info);
        cpumask_clear(&batch->cpumask);