bdev: open block device as files

Add two new helpers to allow opening block devices as files.
This is not the final infrastructure. This still opens the block device
before opening a struct a file. Until we have removed all references to
struct bdev_handle we can't switch the order:

* Introduce blk_to_file_flags() to translate from block specific to
  flags usable to pen a new file.
* Introduce bdev_file_open_by_{dev,path}().
* Introduce temporary sb_bdev_handle() helper to retrieve a struct
  bdev_handle from a block device file and update places that directly
  reference struct bdev_handle to rely on it.
* Don't count block device openes against the number of open files. A
  bdev_file_open_by_{dev,path}() file is never installed into any
  file descriptor table.

One idea that came to mind was to use kernel_tmpfile_open() which
would require us to pass a path and it would then call do_dentry_open()
going through the regular fops->open::blkdev_open() path. But then we're
back to the problem of routing block specific flags such as
BLK_OPEN_RESTRICT_WRITES through the open path and would have to waste
FMODE_* flags every time we add a new one. With this we can avoid using
a flag bit and we have more leeway in how we open block devices from
bdev_open_by_{dev,path}().

Link: https://lore.kernel.org/r/20240123-vfs-bdev-file-v2-1-adbd023e19cc@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>

block/bdev.c

@@ -49,6 +49,13 @@ struct block_device *I_BDEV(struct inode *inode)
 }
 EXPORT_SYMBOL(I_BDEV);
 
+struct block_device *file_bdev(struct file *bdev_file)
+{
+	struct bdev_handle *handle = bdev_file->private_data;
+	return handle->bdev;
+}
+EXPORT_SYMBOL(file_bdev);
+
 static void bdev_write_inode(struct block_device *bdev)
 {
 	struct inode *inode = bdev->bd_inode;
@@ -368,12 +375,12 @@ static struct file_system_type bd_type = {
 };
 
 struct super_block *blockdev_superblock __ro_after_init;
+struct vfsmount *blockdev_mnt __ro_after_init;
 EXPORT_SYMBOL_GPL(blockdev_superblock);
 
 void __init bdev_cache_init(void)
 {
 	int err;
-	static struct vfsmount *bd_mnt __ro_after_init;
 
 	bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode),
 			0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
@@ -382,10 +389,10 @@ void __init bdev_cache_init(void)
 	err = register_filesystem(&bd_type);
 	if (err)
 		panic("Cannot register bdev pseudo-fs");
-	bd_mnt = kern_mount(&bd_type);
-	if (IS_ERR(bd_mnt))
+	blockdev_mnt = kern_mount(&bd_type);
+	if (IS_ERR(blockdev_mnt))
 		panic("Cannot create bdev pseudo-fs");
-	blockdev_superblock = bd_mnt->mnt_sb;   /* For writeback */
+	blockdev_superblock = blockdev_mnt->mnt_sb;   /* For writeback */
 }
 
 struct block_device *bdev_alloc(struct gendisk *disk, u8 partno)
@@ -911,6 +918,92 @@ free_handle:
 }
 EXPORT_SYMBOL(bdev_open_by_dev);
 
+/*
+ * If BLK_OPEN_WRITE_IOCTL is set then this is a historical quirk
+ * associated with the floppy driver where it has allowed ioctls if the
+ * file was opened for writing, but does not allow reads or writes.
+ * Make sure that this quirk is reflected in @f_flags.
+ *
+ * It can also happen if a block device is opened as O_RDWR | O_WRONLY.
+ */
+static unsigned blk_to_file_flags(blk_mode_t mode)
+{
+	unsigned int flags = 0;
+
+	if ((mode & (BLK_OPEN_READ | BLK_OPEN_WRITE)) ==
+	    (BLK_OPEN_READ | BLK_OPEN_WRITE))
+		flags |= O_RDWR;
+	else if (mode & BLK_OPEN_WRITE_IOCTL)
+		flags |= O_RDWR | O_WRONLY;
+	else if (mode & BLK_OPEN_WRITE)
+		flags |= O_WRONLY;
+	else if (mode & BLK_OPEN_READ)
+		flags |= O_RDONLY; /* homeopathic, because O_RDONLY is 0 */
+	else
+		WARN_ON_ONCE(true);
+
+	if (mode & BLK_OPEN_NDELAY)
+		flags |= O_NDELAY;
+
+	return flags;
+}
+
+struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder,
+				   const struct blk_holder_ops *hops)
+{
+	struct file *bdev_file;
+	struct bdev_handle *handle;
+	unsigned int flags;
+
+	handle = bdev_open_by_dev(dev, mode, holder, hops);
+	if (IS_ERR(handle))
+		return ERR_CAST(handle);
+
+	flags = blk_to_file_flags(mode);
+	bdev_file = alloc_file_pseudo_noaccount(handle->bdev->bd_inode,
+			blockdev_mnt, "", flags | O_LARGEFILE, &def_blk_fops);
+	if (IS_ERR(bdev_file)) {
+		bdev_release(handle);
+		return bdev_file;
+	}
+	ihold(handle->bdev->bd_inode);
+
+	bdev_file->f_mode |= FMODE_BUF_RASYNC | FMODE_CAN_ODIRECT;
+	if (bdev_nowait(handle->bdev))
+		bdev_file->f_mode |= FMODE_NOWAIT;
+
+	bdev_file->f_mapping = handle->bdev->bd_inode->i_mapping;
+	bdev_file->f_wb_err = filemap_sample_wb_err(bdev_file->f_mapping);
+	bdev_file->private_data = handle;
+	return bdev_file;
+}
+EXPORT_SYMBOL(bdev_file_open_by_dev);
+
+struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode,
+				    void *holder,
+				    const struct blk_holder_ops *hops)
+{
+	struct file *bdev_file;
+	dev_t dev;
+	int error;
+
+	error = lookup_bdev(path, &dev);
+	if (error)
+		return ERR_PTR(error);
+
+	bdev_file = bdev_file_open_by_dev(dev, mode, holder, hops);
+	if (!IS_ERR(bdev_file) && (mode & BLK_OPEN_WRITE)) {
+		struct bdev_handle *handle = bdev_file->private_data;
+		if (bdev_read_only(handle->bdev)) {
+			fput(bdev_file);
+			bdev_file = ERR_PTR(-EACCES);
+		}
+	}
+
+	return bdev_file;
+}
+EXPORT_SYMBOL(bdev_file_open_by_path);
+
 /**
  * bdev_open_by_path - open a block device by name
  * @path: path to the block device to open

fs/cramfs/inode.c

@@ -495,7 +495,7 @@ static void cramfs_kill_sb(struct super_block *sb)
 		sb->s_mtd = NULL;
 	} else if (IS_ENABLED(CONFIG_CRAMFS_BLOCKDEV) && sb->s_bdev) {
 		sync_blockdev(sb->s_bdev);
-		bdev_release(sb->s_bdev_handle);
+		fput(sb->s_bdev_file);
 	}
 	kfree(sbi);
 }

fs/f2fs/super.c

@@ -4247,7 +4247,7 @@ static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
 
 	for (i = 0; i < max_devices; i++) {
 		if (i == 0)
-			FDEV(0).bdev_handle = sbi->sb->s_bdev_handle;
+			FDEV(0).bdev_handle = sb_bdev_handle(sbi->sb);
 		else if (!RDEV(i).path[0])
 			break;
 

fs/jfs/jfs_logmgr.c

@@ -1162,7 +1162,7 @@ static int open_inline_log(struct super_block *sb)
 	init_waitqueue_head(&log->syncwait);
 
 	set_bit(log_INLINELOG, &log->flag);
-	log->bdev_handle = sb->s_bdev_handle;
+	log->bdev_handle = sb_bdev_handle(sb);
 	log->base = addressPXD(&JFS_SBI(sb)->logpxd);
 	log->size = lengthPXD(&JFS_SBI(sb)->logpxd) >>
 	    (L2LOGPSIZE - sb->s_blocksize_bits);

fs/romfs/super.c

@@ -594,7 +594,7 @@ static void romfs_kill_sb(struct super_block *sb)
 #ifdef CONFIG_ROMFS_ON_BLOCK
 	if (sb->s_bdev) {
 		sync_blockdev(sb->s_bdev);
-		bdev_release(sb->s_bdev_handle);
+		fput(sb->s_bdev_file);
 	}
 #endif
 }

fs/super.c

@@ -1532,16 +1532,16 @@ int setup_bdev_super(struct super_block *sb, int sb_flags,
 		struct fs_context *fc)
 {
 	blk_mode_t mode = sb_open_mode(sb_flags);
-	struct bdev_handle *bdev_handle;
+	struct file *bdev_file;
 	struct block_device *bdev;
 
-	bdev_handle = bdev_open_by_dev(sb->s_dev, mode, sb, &fs_holder_ops);
-	if (IS_ERR(bdev_handle)) {
+	bdev_file = bdev_file_open_by_dev(sb->s_dev, mode, sb, &fs_holder_ops);
+	if (IS_ERR(bdev_file)) {
 		if (fc)
 			errorf(fc, "%s: Can't open blockdev", fc->source);
-		return PTR_ERR(bdev_handle);
+		return PTR_ERR(bdev_file);
 	}
-	bdev = bdev_handle->bdev;
+	bdev = file_bdev(bdev_file);
 
 	/*
 	 * This really should be in blkdev_get_by_dev, but right now can't due
@@ -1549,7 +1549,7 @@ int setup_bdev_super(struct super_block *sb, int sb_flags,
 	 * writable from userspace even for a read-only block device.
 	 */
 	if ((mode & BLK_OPEN_WRITE) && bdev_read_only(bdev)) {
-		bdev_release(bdev_handle);
+		fput(bdev_file);
 		return -EACCES;
 	}
 
@@ -1560,11 +1560,11 @@ int setup_bdev_super(struct super_block *sb, int sb_flags,
 	if (atomic_read(&bdev->bd_fsfreeze_count) > 0) {
 		if (fc)
 			warnf(fc, "%pg: Can't mount, blockdev is frozen", bdev);
-		bdev_release(bdev_handle);
+		fput(bdev_file);
 		return -EBUSY;
 	}
 	spin_lock(&sb_lock);
-	sb->s_bdev_handle = bdev_handle;
+	sb->s_bdev_file = bdev_file;
 	sb->s_bdev = bdev;
 	sb->s_bdi = bdi_get(bdev->bd_disk->bdi);
 	if (bdev_stable_writes(bdev))
@@ -1680,7 +1680,7 @@ void kill_block_super(struct super_block *sb)
 	generic_shutdown_super(sb);
 	if (bdev) {
 		sync_blockdev(bdev);
-		bdev_release(sb->s_bdev_handle);
+		fput(sb->s_bdev_file);
 	}
 }
 

fs/xfs/xfs_super.c

@@ -467,7 +467,7 @@ xfs_open_devices(
 	 * Setup xfs_mount buffer target pointers
 	 */
 	error = -ENOMEM;
-	mp->m_ddev_targp = xfs_alloc_buftarg(mp, sb->s_bdev_handle);
+	mp->m_ddev_targp = xfs_alloc_buftarg(mp, sb_bdev_handle(sb));
 	if (!mp->m_ddev_targp)
 		goto out_close_rtdev;
 

include/linux/blkdev.h

@@ -24,6 +24,7 @@
 #include <linux/sbitmap.h>
 #include <linux/uuid.h>
 #include <linux/xarray.h>
+#include <linux/file.h>
 
 struct module;
 struct request_queue;
@@ -1474,6 +1475,7 @@ extern const struct blk_holder_ops fs_holder_ops;
 	(BLK_OPEN_READ | BLK_OPEN_RESTRICT_WRITES | \
 	 (((flags) & SB_RDONLY) ? 0 : BLK_OPEN_WRITE))
 
+/* @bdev_handle will be removed soon. */
 struct bdev_handle {
 	struct block_device *bdev;
 	void *holder;
@@ -1484,6 +1486,10 @@ struct bdev_handle *bdev_open_by_dev(dev_t dev, blk_mode_t mode, void *holder,
 		const struct blk_holder_ops *hops);
 struct bdev_handle *bdev_open_by_path(const char *path, blk_mode_t mode,
 		void *holder, const struct blk_holder_ops *hops);
+struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder,
+		const struct blk_holder_ops *hops);
+struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode,
+		void *holder, const struct blk_holder_ops *hops);
 int bd_prepare_to_claim(struct block_device *bdev, void *holder,
 		const struct blk_holder_ops *hops);
 void bd_abort_claiming(struct block_device *bdev, void *holder);
@@ -1494,6 +1500,7 @@ struct block_device *blkdev_get_no_open(dev_t dev);
 void blkdev_put_no_open(struct block_device *bdev);
 
 struct block_device *I_BDEV(struct inode *inode);
+struct block_device *file_bdev(struct file *bdev_file);
 
 #ifdef CONFIG_BLOCK
 void invalidate_bdev(struct block_device *bdev);

include/linux/fs.h

@@ -1228,8 +1228,8 @@ struct super_block {
 #endif
 	struct hlist_bl_head	s_roots;	/* alternate root dentries for NFS */
 	struct list_head	s_mounts;	/* list of mounts; _not_ for fs use */
-	struct block_device	*s_bdev;
-	struct bdev_handle	*s_bdev_handle;
+	struct block_device	*s_bdev;	/* can go away once we use an accessor for @s_bdev_file */
+	struct file		*s_bdev_file;
 	struct backing_dev_info *s_bdi;
 	struct mtd_info		*s_mtd;
 	struct hlist_node	s_instances;
@@ -1327,6 +1327,12 @@ struct super_block {
 	struct list_head	s_inodes_wb;	/* writeback inodes */
 } __randomize_layout;
 
+/* Temporary helper that will go away. */
+static inline struct bdev_handle *sb_bdev_handle(struct super_block *sb)
+{
+	return sb->s_bdev_file->private_data;
+}
+
 static inline struct user_namespace *i_user_ns(const struct inode *inode)
 {
 	return inode->i_sb->s_user_ns;