[PATCH v2 2/3] block: don't hide inode from block_device users

From: Pavel Begunkov
Date: Wed Oct 13 2021 - 04:58:15 EST


Instead of tricks with struct bdev_inode, just openly place the inode
inside struct block_device. First, it allows us to inline I_BDEV, which
is simple but non-inline nature of it impacts performance. Also, make it
possible to get rid of ->bd_inode pointer and hooping with extra
indirection, the amount of which became a noticeable problem for the
block layer.

Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
---
block/bdev.c | 44 ++++++++++-----------------------------
block/fops.c | 20 +++++++-----------
include/linux/blk_types.h | 1 +
include/linux/blkdev.h | 8 +++++--
4 files changed, 26 insertions(+), 47 deletions(-)

diff --git a/block/bdev.c b/block/bdev.c
index 567534c63f3d..541e24d240d1 100644
--- a/block/bdev.c
+++ b/block/bdev.c
@@ -30,22 +30,6 @@
#include "../fs/internal.h"
#include "blk.h"

-struct bdev_inode {
- struct block_device bdev;
- struct inode vfs_inode;
-};
-
-static inline struct bdev_inode *BDEV_I(struct inode *inode)
-{
- return container_of(inode, struct bdev_inode, vfs_inode);
-}
-
-struct block_device *I_BDEV(struct inode *inode)
-{
- return &BDEV_I(inode)->bdev;
-}
-EXPORT_SYMBOL(I_BDEV);
-
static void bdev_write_inode(struct block_device *bdev)
{
struct inode *inode = bdev->bd_inode;
@@ -389,12 +373,13 @@ static struct kmem_cache * bdev_cachep __read_mostly;

static struct inode *bdev_alloc_inode(struct super_block *sb)
{
- struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL);
+ struct block_device *bdev = kmem_cache_alloc(bdev_cachep, GFP_KERNEL);

- if (!ei)
+ if (!bdev)
return NULL;
- memset(&ei->bdev, 0, sizeof(ei->bdev));
- return &ei->vfs_inode;
+ memset(bdev, 0, sizeof(*bdev));
+ inode_init_once(&bdev->inode);
+ return &bdev->inode;
}

static void bdev_free_inode(struct inode *inode)
@@ -413,14 +398,7 @@ static void bdev_free_inode(struct inode *inode)
if (MAJOR(bdev->bd_dev) == BLOCK_EXT_MAJOR)
blk_free_ext_minor(MINOR(bdev->bd_dev));

- kmem_cache_free(bdev_cachep, BDEV_I(inode));
-}
-
-static void init_once(void *data)
-{
- struct bdev_inode *ei = data;
-
- inode_init_once(&ei->vfs_inode);
+ kmem_cache_free(bdev_cachep, bdev);
}

static void bdev_evict_inode(struct inode *inode)
@@ -462,10 +440,10 @@ void __init bdev_cache_init(void)
int err;
static struct vfsmount *bd_mnt;

- bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode),
- 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
- SLAB_MEM_SPREAD|SLAB_ACCOUNT|SLAB_PANIC),
- init_once);
+ bdev_cachep = kmem_cache_create("bdev_cache",
+ sizeof(struct block_device), 0,
+ (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
+ SLAB_MEM_SPREAD|SLAB_ACCOUNT|SLAB_PANIC), NULL);
err = register_filesystem(&bd_type);
if (err)
panic("Cannot register bdev pseudo-fs");
@@ -744,7 +722,7 @@ struct block_device *blkdev_get_no_open(dev_t dev)
}

/* switch from the inode reference to a device mode one: */
- bdev = &BDEV_I(inode)->bdev;
+ bdev = I_BDEV(inode);
if (!kobject_get_unless_zero(&bdev->bd_device.kobj))
bdev = NULL;
iput(inode);
diff --git a/block/fops.c b/block/fops.c
index c71e91cd6bcb..9ae795f27f74 100644
--- a/block/fops.c
+++ b/block/fops.c
@@ -17,11 +17,6 @@
#include <linux/fs.h>
#include "blk.h"

-static inline struct inode *bdev_file_inode(struct file *file)
-{
- return file->f_mapping->host;
-}
-
static int blkdev_get_block(struct inode *inode, sector_t iblock,
struct buffer_head *bh, int create)
{
@@ -390,7 +385,8 @@ const struct address_space_operations def_blk_aops = {
*/
static loff_t blkdev_llseek(struct file *file, loff_t offset, int whence)
{
- struct inode *bd_inode = bdev_file_inode(file);
+ struct block_device *bdev = file->private_data;
+ struct inode *bd_inode = &bdev->inode;
loff_t retval;

inode_lock(bd_inode);
@@ -446,7 +442,7 @@ static int blkdev_open(struct inode *inode, struct file *filp)
return PTR_ERR(bdev);

filp->private_data = bdev;
- filp->f_mapping = bdev->bd_inode->i_mapping;
+ filp->f_mapping = bdev->inode.i_mapping;
filp->f_wb_err = filemap_sample_wb_err(filp->f_mapping);
return 0;
}
@@ -469,7 +465,7 @@ static int blkdev_close(struct inode *inode, struct file *filp)
static ssize_t blkdev_write_iter(struct kiocb *iocb, struct iov_iter *from)
{
struct block_device *bdev = iocb->ki_filp->private_data;
- struct inode *bd_inode = bdev->bd_inode;
+ struct inode *bd_inode = &bdev->inode;
loff_t size = i_size_read(bd_inode);
struct blk_plug plug;
size_t shorted = 0;
@@ -508,7 +504,7 @@ static ssize_t blkdev_write_iter(struct kiocb *iocb, struct iov_iter *from)
static ssize_t blkdev_read_iter(struct kiocb *iocb, struct iov_iter *to)
{
struct block_device *bdev = iocb->ki_filp->private_data;
- loff_t size = i_size_read(bdev->bd_inode);
+ loff_t size = i_size_read(&bdev->inode);
loff_t pos = iocb->ki_pos;
size_t shorted = 0;
ssize_t ret;
@@ -534,8 +530,8 @@ static ssize_t blkdev_read_iter(struct kiocb *iocb, struct iov_iter *to)
static long blkdev_fallocate(struct file *file, int mode, loff_t start,
loff_t len)
{
- struct inode *inode = bdev_file_inode(file);
- struct block_device *bdev = I_BDEV(inode);
+ struct block_device *bdev = file->private_data;
+ struct inode *inode = &bdev->inode;
loff_t end = start + len - 1;
loff_t isize;
int error;
@@ -545,7 +541,7 @@ static long blkdev_fallocate(struct file *file, int mode, loff_t start,
return -EOPNOTSUPP;

/* Don't go off the end of the device. */
- isize = i_size_read(bdev->bd_inode);
+ isize = i_size_read(inode);
if (start >= isize)
return -EINVAL;
if (end >= isize) {
diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index 3b967053e9f5..e94d08eb4fa9 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -49,6 +49,7 @@ struct block_device {
#ifdef CONFIG_FAIL_MAKE_REQUEST
bool bd_make_it_fail;
#endif
+ struct inode inode;
} __randomize_layout;

#define bdev_whole(_bdev) \
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 17705c970d7e..75d2682a8a55 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -1141,7 +1141,7 @@ static inline unsigned int blksize_bits(unsigned int size)

static inline unsigned int block_size(struct block_device *bdev)
{
- return 1 << bdev->bd_inode->i_blkbits;
+ return 1 << bdev->inode.i_blkbits;
}

int kblockd_schedule_work(struct work_struct *work);
@@ -1271,10 +1271,14 @@ void blkdev_put_no_open(struct block_device *bdev);

struct block_device *bdev_alloc(struct gendisk *disk, u8 partno);
void bdev_add(struct block_device *bdev, dev_t dev);
-struct block_device *I_BDEV(struct inode *inode);
int truncate_bdev_range(struct block_device *bdev, fmode_t mode, loff_t lstart,
loff_t lend);

+static inline struct block_device *I_BDEV(struct inode *inode)
+{
+ return container_of(inode, struct block_device, inode);
+}
+
#ifdef CONFIG_BLOCK
void invalidate_bdev(struct block_device *bdev);
int sync_blockdev(struct block_device *bdev);
--
2.33.0