[PATCH] Add tmpfs options for memory placement policies

Anything that writes into a tmpfs filesystem is liable to disproportionately
decrease the available memory on a particular node. Since there's no telling
what sort of application (e.g. dd/cp/cat) might be dropping large files
there, this lets the admin choose the appropriate default behavior for their
site's situation.

Introduce a tmpfs mount option which allows specifying a memory policy and
a second option to specify the nodelist for that policy. With the default
policy, tmpfs will behave as it does today. This patch adds support for
preferred, bind, and interleave policies.

The default policy will cause pages to be added to tmpfs files on the node
which is doing the writing. Some jobs expect a single process to create
and manage the tmpfs files. This results in a node which has a
significantly reduced number of free pages.

With this patch, the administrator can specify the policy and nodes for
that policy where they would prefer allocations.

This patch was originally written by Brent Casavant and Hugh Dickins. I
added support for the bind and preferred policies and the mpol_nodelist
mount option.

Signed-off-by: Brent Casavant <[e-mail]>
Signed-off-by: Hugh Dickins <[e-mail]>
Signed-off-by: Robin Holt <[e-mail]>
Signed-off-by: Andrew Morton <[e-mail]>
Signed-off-by: Linus Torvalds <[e-mail]>

Documentation/filesystems/tmpfs.txt

···
 that instance in a system with many cpus making intensive use of it.
 
 
+tmpfs has a mount option to set the NUMA memory allocation policy for
+all files in that instance:
+mpol=interleave		prefers to allocate memory from each node in turn
+mpol=default		prefers to allocate memory from the local node
+mpol=bind		prefers to allocate from mpol_nodelist
+mpol=preferred		prefers to allocate from first node in mpol_nodelist
+
+The following mount option is used in conjunction with mpol=interleave,
+mpol=bind or mpol=preferred:
+mpol_nodelist:	nodelist suitable for parsing with nodelist_parse.
+
+
 To specify the initial root directory you can use the following mount
 options:
 

fs/hugetlbfs/inode.c

···
 		inode->i_mapping->backing_dev_info =&hugetlbfs_backing_dev_info;
 		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
 		info = HUGETLBFS_I(inode);
-		mpol_shared_policy_init(&info->policy);
+		mpol_shared_policy_init(&info->policy, MPOL_DEFAULT, NULL);
 		switch (mode & S_IFMT) {
 		default:
 			init_special_inode(inode, mode, dev);

include/linux/mempolicy.h

···
 	spinlock_t lock;
 };
 
-static inline void mpol_shared_policy_init(struct shared_policy *info)
-{
-	info->root = RB_ROOT;
-	spin_lock_init(&info->lock);
-}
-
+void mpol_shared_policy_init(struct shared_policy *info, int policy,
+				nodemask_t *nodes);
 int mpol_set_shared_policy(struct shared_policy *info,
 				struct vm_area_struct *vma,
 				struct mempolicy *new);
···
 	return -EINVAL;
 }
 
-static inline void mpol_shared_policy_init(struct shared_policy *info)
+static inline void mpol_shared_policy_init(struct shared_policy *info,
+					int policy, nodemask_t *nodes)
 {
 }
 

include/linux/shmem_fs.h

···
 	unsigned long free_blocks;  /* How many are left for allocation */
 	unsigned long max_inodes;   /* How many inodes are allowed */
 	unsigned long free_inodes;  /* How many are left for allocation */
+	int policy;		    /* Default NUMA memory alloc policy */
+	nodemask_t policy_nodes;    /* nodemask for preferred and bind */
 	spinlock_t    stat_lock;
 };
 

mm/mempolicy.c

···
 	return 0;
 }
 
+void mpol_shared_policy_init(struct shared_policy *info, int policy,
+				nodemask_t *policy_nodes)
+{
+	info->root = RB_ROOT;
+	spin_lock_init(&info->lock);
+
+	if (policy != MPOL_DEFAULT) {
+		struct mempolicy *newpol;
+
+		/* Falls back to MPOL_DEFAULT on any error */
+		newpol = mpol_new(policy, policy_nodes);
+		if (!IS_ERR(newpol)) {
+			/* Create pseudo-vma that contains just the policy */
+			struct vm_area_struct pvma;
+
+			memset(&pvma, 0, sizeof(struct vm_area_struct));
+			/* Policy covers entire file */
+			pvma.vm_end = TASK_SIZE;
+			mpol_set_shared_policy(info, &pvma, newpol);
+			mpol_free(newpol);
+		}
+	}
+}
+
 int mpol_set_shared_policy(struct shared_policy *info,
 			struct vm_area_struct *vma, struct mempolicy *npol)
 {

mm/shmem.c

···
 		case S_IFREG:
 			inode->i_op = &shmem_inode_operations;
 			inode->i_fop = &shmem_file_operations;
-			mpol_shared_policy_init(&info->policy);
+			mpol_shared_policy_init(&info->policy, sbinfo->policy,
+							&sbinfo->policy_nodes);
 			break;
 		case S_IFDIR:
 			inode->i_nlink++;
···
 			 * Must not load anything in the rbtree,
 			 * mpol_free_shared_policy will not be called.
 			 */
-			mpol_shared_policy_init(&info->policy);
+			mpol_shared_policy_init(&info->policy, MPOL_DEFAULT,
+						NULL);
 			break;
 		}
 	} else if (sbinfo->max_inodes) {
···
 	.put_link	= shmem_put_link,
 };
 
-static int shmem_parse_options(char *options, int *mode, uid_t *uid, gid_t *gid, unsigned long *blocks, unsigned long *inodes)
+static int shmem_parse_options(char *options, int *mode, uid_t *uid,
+	gid_t *gid, unsigned long *blocks, unsigned long *inodes,
+	int *policy, nodemask_t *policy_nodes)
 {
 	char *this_char, *value, *rest;
 
···
 			*gid = simple_strtoul(value,&rest,0);
 			if (*rest)
 				goto bad_val;
+		} else if (!strcmp(this_char,"mpol")) {
+			if (!strcmp(value,"default"))
+				*policy = MPOL_DEFAULT;
+			else if (!strcmp(value,"preferred"))
+				*policy = MPOL_PREFERRED;
+			else if (!strcmp(value,"bind"))
+				*policy = MPOL_BIND;
+			else if (!strcmp(value,"interleave"))
+				*policy = MPOL_INTERLEAVE;
+			else
+				goto bad_val;
+		} else if (!strcmp(this_char,"mpol_nodelist")) {
+			nodelist_parse(value, *policy_nodes);
 		} else {
 			printk(KERN_ERR "tmpfs: Bad mount option %s\n",
 			       this_char);
···
 	struct shmem_sb_info *sbinfo = SHMEM_SB(sb);
 	unsigned long max_blocks = sbinfo->max_blocks;
 	unsigned long max_inodes = sbinfo->max_inodes;
+	int policy = sbinfo->policy;
+	nodemask_t policy_nodes = sbinfo->policy_nodes;
 	unsigned long blocks;
 	unsigned long inodes;
 	int error = -EINVAL;
 
-	if (shmem_parse_options(data, NULL, NULL, NULL,
-				&max_blocks, &max_inodes))
+	if (shmem_parse_options(data, NULL, NULL, NULL, &max_blocks,
+				&max_inodes, &policy, &policy_nodes))
 		return error;
 
 	spin_lock(&sbinfo->stat_lock);
···
 	sbinfo->free_blocks = max_blocks - blocks;
 	sbinfo->max_inodes  = max_inodes;
 	sbinfo->free_inodes = max_inodes - inodes;
+	sbinfo->policy = policy;
+	sbinfo->policy_nodes = policy_nodes;
 out:
 	spin_unlock(&sbinfo->stat_lock);
 	return error;
···
 	struct shmem_sb_info *sbinfo;
 	unsigned long blocks = 0;
 	unsigned long inodes = 0;
+	int policy = MPOL_DEFAULT;
+	nodemask_t policy_nodes = node_online_map;
 
 #ifdef CONFIG_TMPFS
 	/*
···
 		inodes = totalram_pages - totalhigh_pages;
 		if (inodes > blocks)
 			inodes = blocks;
-		if (shmem_parse_options(data, &mode, &uid, &gid,
-					&blocks, &inodes))
+		if (shmem_parse_options(data, &mode, &uid, &gid, &blocks,
+					&inodes, &policy, &policy_nodes))
 			return -EINVAL;
 	}
 #else
···
 	sbinfo->free_blocks = blocks;
 	sbinfo->max_inodes = inodes;
 	sbinfo->free_inodes = inodes;
+	sbinfo->policy = policy;
+	sbinfo->policy_nodes = policy_nodes;
 
 	sb->s_fs_info = sbinfo;
 	sb->s_maxbytes = SHMEM_MAX_BYTES;