We need to be able to exclude page_mkwrite from happening concurrently
with certain operations. To facilitate this, add a i_mmap_lock to our
inode, down_read() it in our mkwrite, and add a new ILOCK flag to
indicate that we want to take the i_mmap_lock as well. I used pahole to
check the size of the btrfs_inode, the sizes are as follows
no lockdep:
before: 1120 (3 per 4k page)
after: 1160 (3 per 4k page)
lockdep:
before: 2072 (1 per 4k page)
after: 2224 (1 per 4k page)
We're slightly larger but it doesn't change how many objects we can fit
per page.
Reviewed-by: Filipe Manana <fdman...@suse.com>
Signed-off-by: Josef Bacik <jo...@toxicpanda.com>
---
fs/btrfs/btrfs_inode.h | 1 +
fs/btrfs/ctree.h | 1 +
fs/btrfs/inode.c | 10 ++++++++++
3 files changed, 12 insertions(+)
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index 28e202e89660..26837c3ca7f6 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -220,6 +220,7 @@ struct btrfs_inode {
/* Hook into fs_info->delayed_iputs */
struct list_head delayed_iput;
+ struct rw_semaphore i_mmap_lock;
struct inode vfs_inode;
};
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 3bc00aed13b2..5a410c812978 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -3174,6 +3174,7 @@ extern const struct iomap_dio_ops btrfs_dio_ops;
/* Inode locking type flags, by default the exclusive lock is taken */
#define BTRFS_ILOCK_SHARED (1U << 0)
#define BTRFS_ILOCK_TRY (1U << 1)
+#define BTRFS_ILOCK_MMAP (1U << 2)
int btrfs_inode_lock(struct inode *inode, unsigned int ilock_flags);
void btrfs_inode_unlock(struct inode *inode, unsigned int ilock_flags);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 535abf898225..4c3ba0a3e0e6 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -102,6 +102,7 @@ static void __endio_write_update_ordered(struct btrfs_inode
*inode,
* BTRFS_ILOCK_SHARED - acquire a shared lock on the inode
* BTRFS_ILOCK_TRY - try to acquire the lock, if fails on first attempt
* return -EAGAIN
+ * BTRFS_ILOCK_MMAP - acquire a write lock on the i_mmap_lock
*/
int btrfs_inode_lock(struct inode *inode, unsigned int ilock_flags)
{
@@ -122,6 +123,8 @@ int btrfs_inode_lock(struct inode *inode, unsigned int
ilock_flags)
}
inode_lock(inode);
}
+ if (ilock_flags & BTRFS_ILOCK_MMAP)
+ down_write(&BTRFS_I(inode)->i_mmap_lock);
return 0;
}
@@ -133,6 +136,8 @@ int btrfs_inode_lock(struct inode *inode, unsigned int
ilock_flags)
*/
void btrfs_inode_unlock(struct inode *inode, unsigned int ilock_flags)
{
+ if (ilock_flags & BTRFS_ILOCK_MMAP)
+ up_write(&BTRFS_I(inode)->i_mmap_lock);
if (ilock_flags & BTRFS_ILOCK_SHARED)
inode_unlock_shared(inode);
else
@@ -8538,6 +8543,7 @@ vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf)
ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
again:
+ down_read(&BTRFS_I(inode)->i_mmap_lock);
lock_page(page);
size = i_size_read(inode);
@@ -8566,6 +8572,7 @@ vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf)
unlock_extent_cached(io_tree, page_start, page_end,
&cached_state);
unlock_page(page);
+ up_read(&BTRFS_I(inode)->i_mmap_lock);
btrfs_start_ordered_extent(ordered, 1);
btrfs_put_ordered_extent(ordered);
goto again;
@@ -8623,6 +8630,7 @@ vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf)
BTRFS_I(inode)->last_log_commit = BTRFS_I(inode)->root->last_log_commit;
unlock_extent_cached(io_tree, page_start, page_end, &cached_state);
+ up_read(&BTRFS_I(inode)->i_mmap_lock);
btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
sb_end_pagefault(inode->i_sb);
@@ -8631,6 +8639,7 @@ vm_fault_t btrfs_page_mkwrite(struct vm_fault *vmf)
out_unlock:
unlock_page(page);
+ up_read(&BTRFS_I(inode)->i_mmap_lock);
out:
btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved, page_start,
@@ -8882,6 +8891,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
INIT_LIST_HEAD(&ei->delalloc_inodes);
INIT_LIST_HEAD(&ei->delayed_iput);
RB_CLEAR_NODE(&ei->rb_node);
+ init_rwsem(&ei->i_mmap_lock);
return inode;
}
--
2.26.2
