powerpc/book3s64/radix: Remove WARN_ON in destroy_context()

On failed task initialization due to memory allocation failures, we can
call into destroy_context() with process_tb entry already populated.
This patch forces the process_tb entry to zero in destroy_context().
With this patch, we lose the ability to track if we are destroying a
context without flushing the process table entry.

  WARNING: CPU: 4 PID: 6368 at arch/powerpc/mm/mmu_context_book3s64.c:246 destroy_context+0x58/0x340
  NIP [c0000000000875f8] destroy_context+0x58/0x340
  LR [c00000000013da18] __mmdrop+0x78/0x270
  Call Trace:
  [c000000f7db77c80] [c00000000013da18] __mmdrop+0x78/0x270
  [c000000f7db77cf0] [c0000000004d6a34] __do_execve_file.isra.13+0xbd4/0x1000
  [c000000f7db77e00] [c0000000004d7428] sys_execve+0x58/0x70
  [c000000f7db77e30] [c00000000000b388] system_call+0x5c/0x70

Reported-by: Priya M.A <priyama2@in.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
[mpe: Reformat/tweak comment wording]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190918140103.24395-1-aneesh.kumar@linux.ibm.com

diff --git a/arch/powerpc/mm/book3s64/mmu_context.c b/arch/powerpc/mm/book3s64/mmu_context.c
index 2d0cb5ba9a47e89317a1ebc9a46c3ff7d1a1a423..0ba30b8b935bcf2a0251185b2655555542c053c0 100644 (file)
--- a/arch/powerpc/mm/book3s64/mmu_context.c
+++ b/arch/powerpc/mm/book3s64/mmu_context.c
@@ -256,8 +256,21 @@ void destroy_context(struct mm_struct *mm)
 #ifdef CONFIG_SPAPR_TCE_IOMMU
        WARN_ON_ONCE(!list_empty(&mm->context.iommu_group_mem_list));
 #endif
+       /*
+        * For tasks which were successfully initialized we end up calling
+        * arch_exit_mmap() which clears the process table entry. And
+        * arch_exit_mmap() is called before the required fullmm TLB flush
+        * which does a RIC=2 flush. Hence for an initialized task, we do clear
+        * any cached process table entries.
+        *
+        * The condition below handles the error case during task init. We have
+        * set the process table entry early and if we fail a task
+        * initialization, we need to ensure the process table entry is zeroed.
+        * We need not worry about process table entry caches because the task
+        * never ran with the PID value.
+        */
        if (radix_enabled())
-               WARN_ON(process_tb[mm->context.id].prtb0 != 0);
+               process_tb[mm->context.id].prtb0 = 0;
        else
                subpage_prot_free(mm);
        destroy_contexts(&mm->context);