dma-buf: give each buffer a full-fledged inode

By traversing /proc/*/fd and /proc/*/map_files, processes with CAP_ADMIN
can get a lot of fine-grained data about how shmem buffers are shared
among processes.  stat(2) on each entry gives the caller a unique
ID (st_ino), the buffer's size (st_size), and even the number of pages
currently charged to the buffer (st_blocks / 512).

In contrast, all dma-bufs share the same anonymous inode.  So while we
can count how many dma-buf fds or mappings a process has, we can't get
the size of the backing buffers or tell if two entries point to the same
dma-buf.  On systems with debugfs, we can get a per-buffer breakdown of
size and reference count, but can't tell which processes are actually
holding the references to each buffer.

Replace the singleton inode with full-fledged inodes allocated by
alloc_anon_inode().  This involves creating and mounting a
mini-pseudo-filesystem for dma-buf, following the example in fs/aio.c.

Signed-off-by: Greg Hackmann <[email protected]>
Signed-off-by: Chenbo Feng <[email protected]>
Signed-off-by: Sumit Semwal <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

Author: greghackmann

drivers/dma-buf/dma-buf.c

@@ -34,8 +34,10 @@
 #include <linux/poll.h>
 #include <linux/reservation.h>
 #include <linux/mm.h>
+#include <linux/mount.h>
 
 #include <uapi/linux/dma-buf.h>
+#include <uapi/linux/magic.h>
 
 static inline int is_dma_buf_file(struct file *);
 
@@ -46,6 +48,25 @@ struct dma_buf_list {
 
 static struct dma_buf_list db_list;
 
+static const struct dentry_operations dma_buf_dentry_ops = {
+	.d_dname = simple_dname,
+};
+
+static struct vfsmount *dma_buf_mnt;
+
+static struct dentry *dma_buf_fs_mount(struct file_system_type *fs_type,
+		int flags, const char *name, void *data)
+{
+	return mount_pseudo(fs_type, "dmabuf:", NULL, &dma_buf_dentry_ops,
+			DMA_BUF_MAGIC);
+}
+
+static struct file_system_type dma_buf_fs_type = {
+	.name = "dmabuf",
+	.mount = dma_buf_fs_mount,
+	.kill_sb = kill_anon_super,
+};
+
 static int dma_buf_release(struct inode *inode, struct file *file)
 {
 	struct dma_buf *dmabuf;
@@ -342,6 +363,31 @@ static inline int is_dma_buf_file(struct file *file)
 	return file->f_op == &dma_buf_fops;
 }
 
+static struct file *dma_buf_getfile(struct dma_buf *dmabuf, int flags)
+{
+	struct file *file;
+	struct inode *inode = alloc_anon_inode(dma_buf_mnt->mnt_sb);
+
+	if (IS_ERR(inode))
+		return ERR_CAST(inode);
+
+	inode->i_size = dmabuf->size;
+	inode_set_bytes(inode, dmabuf->size);
+
+	file = alloc_file_pseudo(inode, dma_buf_mnt, "dmabuf",
+				 flags, &dma_buf_fops);
+	if (IS_ERR(file))
+		goto err_alloc_file;
+	file->f_flags = flags & (O_ACCMODE | O_NONBLOCK);
+	file->private_data = dmabuf;
+
+	return file;
+
+err_alloc_file:
+	iput(inode);
+	return file;
+}
+
 /**
  * DOC: dma buf device access
  *
@@ -436,8 +482,7 @@ struct dma_buf *dma_buf_export(const struct dma_buf_export_info *exp_info)
 	}
 	dmabuf->resv = resv;
 
-	file = anon_inode_getfile("dmabuf", &dma_buf_fops, dmabuf,
-					exp_info->flags);
+	file = dma_buf_getfile(dmabuf, exp_info->flags);
 	if (IS_ERR(file)) {
 		ret = PTR_ERR(file);
 		goto err_dmabuf;
@@ -1055,8 +1100,8 @@ static int dma_buf_debug_show(struct seq_file *s, void *unused)
 		return ret;
 
 	seq_puts(s, "\nDma-buf Objects:\n");
-	seq_printf(s, "%-8s\t%-8s\t%-8s\t%-8s\texp_name\n",
-		   "size", "flags", "mode", "count");
+	seq_printf(s, "%-8s\t%-8s\t%-8s\t%-8s\texp_name\t%-8s\n",
+		   "size", "flags", "mode", "count", "ino");
 
 	list_for_each_entry(buf_obj, &db_list.head, list_node) {
 		ret = mutex_lock_interruptible(&buf_obj->lock);
@@ -1067,11 +1112,12 @@ static int dma_buf_debug_show(struct seq_file *s, void *unused)
 			continue;
 		}
 
-		seq_printf(s, "%08zu\t%08x\t%08x\t%08ld\t%s\n",
+		seq_printf(s, "%08zu\t%08x\t%08x\t%08ld\t%s\t%08lu\n",
 				buf_obj->size,
 				buf_obj->file->f_flags, buf_obj->file->f_mode,
 				file_count(buf_obj->file),
-				buf_obj->exp_name);
+				buf_obj->exp_name,
+				file_inode(buf_obj->file)->i_ino);
 
 		robj = buf_obj->resv;
 		while (true) {
@@ -1167,6 +1213,10 @@ static inline void dma_buf_uninit_debugfs(void)
 
 static int __init dma_buf_init(void)
 {
+	dma_buf_mnt = kern_mount(&dma_buf_fs_type);
+	if (IS_ERR(dma_buf_mnt))
+		return PTR_ERR(dma_buf_mnt);
+
 	mutex_init(&db_list.lock);
 	INIT_LIST_HEAD(&db_list.head);
 	dma_buf_init_debugfs();
@@ -1177,5 +1227,6 @@ subsys_initcall(dma_buf_init);
 static void __exit dma_buf_deinit(void)
 {
 	dma_buf_uninit_debugfs();
+	kern_unmount(dma_buf_mnt);
 }
 __exitcall(dma_buf_deinit);

include/uapi/linux/magic.h

@@ -91,5 +91,6 @@
 #define UDF_SUPER_MAGIC		0x15013346
 #define BALLOON_KVM_MAGIC	0x13661366
 #define ZSMALLOC_MAGIC		0x58295829
+#define DMA_BUF_MAGIC		0x444d4142	/* "DMAB" */
 
 #endif /* __LINUX_MAGIC_H__ */