On Mon, May 31, 2021 at 1:08 PM HAGIO KAZUHITO(萩尾 一仁) <k-hagio-ab@nec.com> wrote:
The nid member of struct memory_block is a 4-byte integer, but read
and printed as a 8-byte integer on 64-bit machines.  Without the patch,
the option displays wrong NIDs.

Good findings, Kazu. 
Acked-by: Lianbo Jiang <lijiang@redhat.com>


  crash> kmem -n
  ...
     MEM_BLOCK        NAME          PHYSICAL RANGE       NODE  STATE   START_SECTION_NO
   ffff9edeff2b9400   memory0             0 -   7fffffff 14195095130662240256  ONLINE  0
   ffff9edeff2bb400   memory2     100000000 -  17fffffff 14195094718345379840  ONLINE  32

The issue seems to appear on Linux 5.12 and later kernels that contain
commit e9a2e48e8704c ("drivers/base/memory: don't store phys_device in
memory blocks"), which changed the arrangement of the members of struct
memory_block.

Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
---
 memory.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/memory.c b/memory.c
index 8c6bbe409922..600f2de336d9 100644
--- a/memory.c
+++ b/memory.c
@@ -17564,13 +17564,13 @@ print_memory_block(ulong memory_block)

        if (MEMBER_EXISTS("memory_block", "nid")) {
                readmem(memory_block + OFFSET(memory_block_nid), KVADDR, &nid,
-                       sizeof(void *), "memory_block nid", FAULT_ON_ERROR);
+                       sizeof(int), "memory_block nid", FAULT_ON_ERROR);
                fprintf(fp, " %s %s %s %s  %s %s\n",
                        mkstring(buf1, VADDR_PRLEN, LJUST|LONG_HEX,
                        MKSTR(memory_block)),
                        mkstring(buf2, 12, CENTER, name),
                        parangebuf,
-                       mkstring(buf5, strlen("NODE"), CENTER|LONG_DEC,
+                       mkstring(buf5, strlen("NODE"), CENTER|INT_DEC,
                        MKSTR(nid)),
                        mkstring(buf6, strlen("OFFLINE"), LJUST,
                        statebuf),
--
2.27.0