mm/nds32: use general page fault accounting
authorPeter Xu <peterx@redhat.com>
Wed, 12 Aug 2020 01:38:19 +0000 (18:38 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 12 Aug 2020 17:58:03 +0000 (10:58 -0700)
Use the general page fault accounting by passing regs into
handle_mm_fault().  It naturally solve the issue of multiple page fault
accounting when page fault retry happened.

Fix PERF_COUNT_SW_PAGE_FAULTS perf event manually for page fault retries,
by moving it before taking mmap_sem.

Signed-off-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Greentime Hu <green.hu@gmail.com>
Cc: Nick Hu <nickhu@andestech.com>
Cc: Vincent Chen <deanbo422@gmail.com>
Link: http://lkml.kernel.org/r/20200707225021.200906-13-peterx@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
arch/nds32/mm/fault.c

index d0ecc8f..f02524e 100644 (file)
@@ -121,6 +121,8 @@ void do_page_fault(unsigned long entry, unsigned long addr,
        if (unlikely(faulthandler_disabled() || !mm))
                goto no_context;
 
+       perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
+
        /*
         * As per x86, we may deadlock here. However, since the kernel only
         * validly references user space from well defined areas of the code,
@@ -206,7 +208,7 @@ good_area:
         * the fault.
         */
 
-       fault = handle_mm_fault(vma, addr, flags, NULL);
+       fault = handle_mm_fault(vma, addr, flags, regs);
 
        /*
         * If we need to retry but a fatal signal is pending, handle the
@@ -228,22 +230,7 @@ good_area:
                        goto bad_area;
        }
 
-       /*
-        * 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.
-        */
-       perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
        if (flags & FAULT_FLAG_ALLOW_RETRY) {
-               if (fault & VM_FAULT_MAJOR) {
-                       tsk->maj_flt++;
-                       perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ,
-                                     1, regs, addr);
-               } else {
-                       tsk->min_flt++;
-                       perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN,
-                                     1, regs, addr);
-               }
                if (fault & VM_FAULT_RETRY) {
                        flags |= FAULT_FLAG_TRIED;