mm: show total hugetlb memory consumption in /proc/meminfo
authorRoman Gushchin <guro@fb.com>
Thu, 1 Feb 2018 00:16:22 +0000 (16:16 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 1 Feb 2018 01:18:36 +0000 (17:18 -0800)
Currently we display some hugepage statistics (total, free, etc) in
/proc/meminfo, but only for default hugepage size (e.g.  2Mb).

If hugepages of different sizes are used (like 2Mb and 1Gb on x86-64),
/proc/meminfo output can be confusing, as non-default sized hugepages
are not reflected at all, and there are no signs that they are existing
and consuming system memory.

To solve this problem, let's display the total amount of memory,
consumed by hugetlb pages of all sized (both free and used).  Let's call
it "Hugetlb", and display size in kB to match generic /proc/meminfo
style.

For example, (1024 2Mb pages and 2 1Gb pages are pre-allocated):
  $ cat /proc/meminfo
  MemTotal:        8168984 kB
  MemFree:         3789276 kB
  <...>
  CmaFree:               0 kB
  HugePages_Total:    1024
  HugePages_Free:     1024
  HugePages_Rsvd:        0
  HugePages_Surp:        0
  Hugepagesize:       2048 kB
  Hugetlb:         4194304 kB
  DirectMap4k:       32632 kB
  DirectMap2M:     4161536 kB
  DirectMap1G:     6291456 kB

Also, this patch updates corresponding docs to reflect Hugetlb entry
meaning and difference between Hugetlb and HugePages_Total * Hugepagesize.

Link: http://lkml.kernel.org/r/20171115231409.12131-1-guro@fb.com
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Documentation/vm/hugetlbpage.txt
mm/hugetlb.c

index 59cbc80..faf077d 100644 (file)
@@ -20,19 +20,20 @@ options.
 
 The /proc/meminfo file provides information about the total number of
 persistent hugetlb pages in the kernel's huge page pool.  It also displays
-information about the number of free, reserved and surplus huge pages and the
-default huge page size.  The huge page size is needed for generating the
-proper alignment and size of the arguments to system calls that map huge page
-regions.
+default huge page size and information about the number of free, reserved
+and surplus huge pages in the pool of huge pages of default size.
+The huge page size is needed for generating the proper alignment and
+size of the arguments to system calls that map huge page regions.
 
 The output of "cat /proc/meminfo" will include lines like:
 
 .....
-HugePages_Total: vvv
-HugePages_Free:  www
-HugePages_Rsvd:  xxx
-HugePages_Surp:  yyy
-Hugepagesize:    zzz kB
+HugePages_Total: uuu
+HugePages_Free:  vvv
+HugePages_Rsvd:  www
+HugePages_Surp:  xxx
+Hugepagesize:    yyy kB
+Hugetlb:         zzz kB
 
 where:
 HugePages_Total is the size of the pool of huge pages.
@@ -47,6 +48,14 @@ HugePages_Surp  is short for "surplus," and is the number of huge pages in
                 the pool above the value in /proc/sys/vm/nr_hugepages. The
                 maximum number of surplus huge pages is controlled by
                 /proc/sys/vm/nr_overcommit_hugepages.
+Hugepagesize    is the default hugepage size (in Kb).
+Hugetlb         is the total amount of memory (in kB), consumed by huge
+                pages of all sizes.
+                If huge pages of different sizes are in use, this number
+                will exceed HugePages_Total * Hugepagesize. To get more
+                detailed information, please, refer to
+                /sys/kernel/mm/hugepages (described below).
+
 
 /proc/filesystems should also show a filesystem of type "hugetlbfs" configured
 in the kernel.
index 9a334f5..1e6a5ad 100644 (file)
@@ -2975,20 +2975,32 @@ out:
 
 void hugetlb_report_meminfo(struct seq_file *m)
 {
-       struct hstate *h = &default_hstate;
+       struct hstate *h;
+       unsigned long total = 0;
+
        if (!hugepages_supported())
                return;
-       seq_printf(m,
-                       "HugePages_Total:   %5lu\n"
-                       "HugePages_Free:    %5lu\n"
-                       "HugePages_Rsvd:    %5lu\n"
-                       "HugePages_Surp:    %5lu\n"
-                       "Hugepagesize:   %8lu kB\n",
-                       h->nr_huge_pages,
-                       h->free_huge_pages,
-                       h->resv_huge_pages,
-                       h->surplus_huge_pages,
-                       1UL << (huge_page_order(h) + PAGE_SHIFT - 10));
+
+       for_each_hstate(h) {
+               unsigned long count = h->nr_huge_pages;
+
+               total += (PAGE_SIZE << huge_page_order(h)) * count;
+
+               if (h == &default_hstate)
+                       seq_printf(m,
+                                  "HugePages_Total:   %5lu\n"
+                                  "HugePages_Free:    %5lu\n"
+                                  "HugePages_Rsvd:    %5lu\n"
+                                  "HugePages_Surp:    %5lu\n"
+                                  "Hugepagesize:   %8lu kB\n",
+                                  count,
+                                  h->free_huge_pages,
+                                  h->resv_huge_pages,
+                                  h->surplus_huge_pages,
+                                  (PAGE_SIZE << huge_page_order(h)) / 1024);
+       }
+
+       seq_printf(m, "Hugetlb:        %8lu kB\n", total / 1024);
 }
 
 int hugetlb_report_node_meminfo(int nid, char *buf)