mm: split 'tlb_flush_mmu()' into tlb flushing and memory freeing parts

The mmu-gather operation 'tlb_flush_mmu()' has done two things: the
actual tlb flush operation, and the batched freeing of the pages that
the TLB entries pointed at.

This splits the operation into separate phases, so that the forced
batched flushing done by zap_pte_range() can now do the actual TLB flush
while still holding the page table lock, but delay the batched freeing
of all the pages to after the lock has been dropped.

This in turn allows us to avoid a race condition between
set_page_dirty() (as called by zap_pte_range() when it finds a dirty
shared memory pte) and page_mkclean(): because we now flush all the
dirty page data from the TLB's while holding the pte lock,
page_mkclean() will be held up walking the (recently cleaned) page
tables until after the TLB entries have been flushed from all CPU's.

Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Tested-by: Dave Hansen <dave.hansen@intel.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Russell King - ARM Linux <linux@arm.linux.org.uk>
Cc: Tony Luck <tony.luck@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 34 insertions, 19 deletions

diff --git a/mm/memory.c b/mm/memory.c
index 93e332d5ed77..037b812a9531 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -232,17 +232,18 @@ void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned long
 #endif
 }
 
-void tlb_flush_mmu(struct mmu_gather *tlb)
+static void tlb_flush_mmu_tlbonly(struct mmu_gather *tlb)
 {
-        struct mmu_gather_batch *batch;
-
-        if (!tlb->need_flush)
-                return;
         tlb->need_flush = 0;
         tlb_flush(tlb);
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
         tlb_table_flush(tlb);
 #endif
+}
+
+static void tlb_flush_mmu_free(struct mmu_gather *tlb)
+{
+        struct mmu_gather_batch *batch;
 
         for (batch = &tlb->local; batch; batch = batch->next) {
                 free_pages_and_swap_cache(batch->pages, batch->nr);
@@ -251,6 +252,14 @@ void tlb_flush_mmu(struct mmu_gather *tlb)
         tlb->active = &tlb->local;
 }
 
+void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+        if (!tlb->need_flush)
+                return;
+        tlb_flush_mmu_tlbonly(tlb);
+        tlb_flush_mmu_free(tlb);
+}
+
 /* tlb_finish_mmu
  *      Called at the end of the shootdown operation to free up any resources
  *      that were required.
@@ -1127,8 +1136,10 @@ again:
                         if (PageAnon(page))
                                 rss[MM_ANONPAGES]--;
                         else {
-                                if (pte_dirty(ptent))
+                                if (pte_dirty(ptent)) {
+                                        force_flush = 1;
                                         set_page_dirty(page);
+                                }
                                 if (pte_young(ptent) &&
                                     likely(!(vma->vm_flags & VM_SEQ_READ)))
                                         mark_page_accessed(page);
@@ -1137,9 +1148,10 @@ again:
                         page_remove_rmap(page);
                         if (unlikely(page_mapcount(page) < 0))
                                 print_bad_pte(vma, addr, ptent, page);
-                        force_flush = !__tlb_remove_page(tlb, page);
-                        if (force_flush)
+                        if (unlikely(!__tlb_remove_page(tlb, page))) {
+                                force_flush = 1;
                                 break;
+                        }
                         continue;
                 }
                 /*
@@ -1174,18 +1186,11 @@ again:
 
         add_mm_rss_vec(mm, rss);
         arch_leave_lazy_mmu_mode();
-        pte_unmap_unlock(start_pte, ptl);
 
-        /*
-         * mmu_gather ran out of room to batch pages, we break out of
-         * the PTE lock to avoid doing the potential expensive TLB invalidate
-         * and page-free while holding it.
-         */
+        /* Do the actual TLB flush before dropping ptl */
         if (force_flush) {
                 unsigned long old_end;
 
-                force_flush = 0;
-
                 /*
                  * Flush the TLB just for the previous segment,
                  * then update the range to be the remaining
@@ -1193,11 +1198,21 @@ again:
                  */
                 old_end = tlb->end;
                 tlb->end = addr;
-
-                tlb_flush_mmu(tlb);
-
+                tlb_flush_mmu_tlbonly(tlb);
                 tlb->start = addr;
                 tlb->end = old_end;
+        }
+        pte_unmap_unlock(start_pte, ptl);
+
+        /*
+         * If we forced a TLB flush (either due to running out of
+         * batch buffers or because we needed to flush dirty TLB
+         * entries before releasing the ptl), free the batched
+         * memory too. Restart if we didn't do everything.
+         */
+        if (force_flush) {
+                force_flush = 0;
+                tlb_flush_mmu_free(tlb);
 
                 if (addr != end)
                         goto again;