mm: retry page fault when blocking on disk transfer

This change reduces mmap_sem hold times that are caused by waiting for
disk transfers when accessing file mapped VMAs.

It introduces the VM_FAULT_ALLOW_RETRY flag, which indicates that the call
site wants mmap_sem to be released if blocking on a pending disk transfer.
In that case, filemap_fault() returns the VM_FAULT_RETRY status bit and
do_page_fault() will then re-acquire mmap_sem and retry the page fault.

It is expected that the retry will hit the same page which will now be
cached, and thus it will complete with a low mmap_sem hold time.

Tests:

- microbenchmark: thread A mmaps a large file and does random read accesses
  to the mmaped area - achieves about 55 iterations/s. Thread B does
  mmap/munmap in a loop at a separate location - achieves 55 iterations/s
  before, 15000 iterations/s after.

- We are seeing related effects in some applications in house, which show
  significant performance regressions when running without this change.

[akpm@linux-foundation.org: fix warning & crash]
Signed-off-by: Michel Lespinasse <walken@google.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Reviewed-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: Ying Han <yinghan@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 26 insertions, 12 deletions

diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index 852b319edbdc..9b2345c9e0c3 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -956,8 +956,10 @@ do_page_fault(struct pt_regs *regs, unsigned long error_code)
         struct task_struct *tsk;
         unsigned long address;
         struct mm_struct *mm;
-        int write;
         int fault;
+        int write = error_code & PF_WRITE;
+        unsigned int flags = FAULT_FLAG_ALLOW_RETRY |
+                                        (write ? FAULT_FLAG_WRITE : 0);
 
         tsk = current;
         mm = tsk->mm;
@@ -1068,6 +1070,7 @@ do_page_fault(struct pt_regs *regs, unsigned long error_code)
                         bad_area_nosemaphore(regs, error_code, address);
                         return;
                 }
+retry:
                 down_read(&mm->mmap_sem);
         } else {
                 /*
@@ -1111,8 +1114,6 @@ do_page_fault(struct pt_regs *regs, unsigned long error_code)
          * we can handle it..
          */
 good_area:
-        write = error_code & PF_WRITE;
-
         if (unlikely(access_error(error_code, write, vma))) {
                 bad_area_access_error(regs, error_code, address);
                 return;
@@ -1123,21 +1124,34 @@ good_area:
          * make sure we exit gracefully rather than endlessly redo
          * the fault:
          */
-        fault = handle_mm_fault(mm, vma, address, write ? FAULT_FLAG_WRITE : 0);
+        fault = handle_mm_fault(mm, vma, address, flags);
 
         if (unlikely(fault & VM_FAULT_ERROR)) {
                 mm_fault_error(regs, error_code, address, fault);
                 return;
         }
 
-        if (fault & VM_FAULT_MAJOR) {
-                tsk->maj_flt++;
-                perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0,
-                                     regs, address);
-        } else {
-                tsk->min_flt++;
-                perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0,
-                                     regs, address);
+        /*
+         * Major/minor page fault accounting is only done on the
+         * initial attempt. If we go through a retry, it is extremely
+         * likely that the page will be found in page cache at that point.
+         */
+        if (flags & FAULT_FLAG_ALLOW_RETRY) {
+                if (fault & VM_FAULT_MAJOR) {
+                        tsk->maj_flt++;
+                        perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0,
+                                      regs, address);
+                } else {
+                        tsk->min_flt++;
+                        perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0,
+                                      regs, address);
+                }
+                if (fault & VM_FAULT_RETRY) {
+                        /* Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk
+                         * of starvation. */
+                        flags &= ~FAULT_FLAG_ALLOW_RETRY;
+                        goto retry;
+                }
         }
 
         check_v8086_mode(regs, address, tsk);