is this a problem of numactl in RedHat7.0 ?

I use numactl(--localalloc) tool run a test case, but it shows that the numa policy is prefer, I don't know why.

[root@localhost test]# numactl --hard
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 4 5 12 13 14 15 16 17
node 0 size: 32739 MB
node 0 free: 29134 MB
node 1 cpus: 6 7 8 9 10 11 18 19 20 21 22 23
node 1 size: 32768 MB
node 1 free: 31682 MB
node distances:
node 0 1

  0:  10  20
  1:  20  10

[root@localhost test]# numactl --localalloc --physcpubind=22 ./test.exe

[root@localhost linux-3.10-redhat]# ps -ef| grep test

root      18532  13994  1 05:24 pts/0    00:00:00 ./test.exe
root      18534  14037  0 05:24 pts/1    00:00:00 grep --color=auto test

[root@localhost linux-3.10-redhat]# cat /proc/18532/numa_maps 00400000 prefer:0 file=/home/qiuxishi/test/test.exe mapped=1 N0=1 00600000 prefer:0 file=/home/qiuxishi/test/test.exe anon=1 dirty=1 N1=1 00601000 prefer:0 file=/home/qiuxishi/test/test.exe anon=1 dirty=1 N1=1 7f1a15b9b000 prefer:0 file=/usr/lib64/libc-2.17.so mapped=51 mapmax=44 N0=49 N1=2 7f1a15d51000 prefer:0 file=/usr/lib64/libc-2.17.so
7f1a15f51000 prefer:0 file=/usr/lib64/libc-2.17.so anon=4 dirty=4 N1=4 7f1a15f55000 prefer:0 file=/usr/lib64/libc-2.17.so anon=2 dirty=2 N1=2 7f1a15f57000 prefer:0 anon=3 dirty=3 N1=3
7f1a15f5c000 prefer:0 file=/usr/lib64/ld-2.17.so mapped=28 mapmax=15 N0=23 N1=5 7f1a16166000 prefer:0 anon=3 dirty=3 N1=3
7f1a1617c000 prefer:0 anon=1 dirty=1 N1=1
7f1a1617d000 prefer:0 file=/usr/lib64/ld-2.17.so anon=1 dirty=1 N1=1 7f1a1617e000 prefer:0 file=/usr/lib64/ld-2.17.so anon=1 dirty=1 N1=1 7f1a1617f000 prefer:0 anon=1 dirty=1 N1=1
7ffd0d2ed000 prefer:0 stack anon=3 dirty=3 N1=3
7ffd0d3fc000 prefer:0

The kernel implements MPOL_PREFERRED and MPOL_LOCAL in the same way. Look at this code in mpol_new(), in mm/mempolicy.c:

/*
* MPOL_PREFERRED cannot be used with MPOL_F_STATIC_NODES or
* MPOL_F_RELATIVE_NODES if the nodemask is empty (local
allocation).

         * All other modes require a valid pointer to a non-empty nodemask.
         */
        if (mode == MPOL_PREFERRED) {
                if (nodes_empty(*nodes)) {
                        if (((flags & MPOL_F_STATIC_NODES) ||
                             (flags & MPOL_F_RELATIVE_NODES)))
                                return ERR_PTR(-EINVAL);
                }
        } else if (mode == MPOL_LOCAL) {
                if (!nodes_empty(*nodes))
                        return ERR_PTR(-EINVAL);
                mode = MPOL_PREFERRED;
        } else if (nodes_empty(*nodes))
                return ERR_PTR(-EINVAL);

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Hi Rik,

Thank you for your reply. I find the reason is this patch, and it is not backport to RedHat 7.0

8790c71a18e5d2d93532ae250bcf5eddbba729cd

index 873de7e..ae3c8f3 100644
@@ -2930,7 +2930,7 @@ void mpol_to_str(char *buffer, int maxlen, struct mempolicy *pol)
        unsigned short mode = MPOL_DEFAULT;
        unsigned short flags = 0;
 
-       if (pol && pol != &default_policy) {
+       if (pol && pol != &default_policy && !(pol->flags & MPOL_F_MORON)) {
                mode = pol->mode;
                flags = pol->flags;
        }

Thanks,
Xishi Qiu

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

user:
"numactl --localalloc" wil call

	main()
	  numa_set_localalloc()
	    setpol(MPOL_DEFAULT, numa_no_nodes_ptr);
	      set_mempolicy()
	        syscall(__NR_set_mempolicy,mode,nmask,maxnode);

kernel:

	do_set_mempolicy()
	  mpol_new()
		if (mode == MPOL_DEFAULT) {
			if (nodes && !nodes_empty(*nodes))
				return ERR_PTR(-EINVAL);
			return NULL;  // return from here
		}

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>