On Wed, Jun 25, 2014 at 05:50:03PM +0100, Mark Rutland wrote:

Hi Steve,

On Wed, Jun 25, 2014 at 04:40:24PM +0100, Steve Capper wrote:

quoted

Activate the RCU fast_gup for ARM64. We also need to force THP splits
to broadcast an IPI s.t. we block in the fast_gup page walker. As THP
splits are comparatively rare, this should not lead to a noticeable
performance degradation.

Some pre-requisite functions pud_write and pud_page are also added.

Signed-off-by: Steve Capper <redacted>
---
 arch/arm64/Kconfig               |  3 +++
 arch/arm64/include/asm/pgtable.h | 11 ++++++++++-
 arch/arm64/mm/flush.c            | 19 +++++++++++++++++++
 3 files changed, 32 insertions(+), 1 deletion(-)

[...]

quoted

diff --git a/arch/arm64/mm/flush.c b/arch/arm64/mm/flush.c
index e4193e3..ddf96c1 100644
--- a/arch/arm64/mm/flush.c
+++ b/arch/arm64/mm/flush.c

@@ -103,3 +103,22 @@ EXPORT_SYMBOL(flush_dcache_page);
  */
 EXPORT_SYMBOL(flush_cache_all);
 EXPORT_SYMBOL(flush_icache_range);
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+static void thp_splitting_flush_sync(void *arg)
+{
+}
+
+void pmdp_splitting_flush(struct vm_area_struct *vma, unsigned long address,
+			  pmd_t *pmdp)
+{
+	pmd_t pmd = pmd_mksplitting(*pmdp);
+	VM_BUG_ON(address & ~PMD_MASK);
+	set_pmd_at(vma->vm_mm, address, pmdp, pmd);
+
+	/* dummy IPI to serialise against fast_gup */
+	smp_call_function(thp_splitting_flush_sync, NULL, 1);

Is there some reason we can't use kick_all_cpus_sync()?

Yes that would be equivalent. But looking at that, I worry about the
smp_mb(); archs are supposed to make sure IPIs are serializing.