Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit f66fd9f0 authored by Yan, Zheng's avatar Yan, Zheng Committed by Ilya Dryomov
Browse files

ceph: pre-allocate data structure that tracks caps flushing 



Signed-off-by: default avatarYan, Zheng <zyan@redhat.com>
parent e548e9b9

fs/ceph/addr.c

+15 −4
Original line number Diff line number Diff line
@@ -1308,12 +1308,17 @@ static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) 
	struct inode *inode = file_inode(vma->vm_file);

	struct ceph_inode_info *ci = ceph_inode(inode);

	struct ceph_file_info *fi = vma->vm_file->private_data;

	struct ceph_cap_flush *prealloc_cf;

	struct page *page = vmf->page;

	loff_t off = page_offset(page);

	loff_t size = i_size_read(inode);

	size_t len;

	int want, got, ret;



	prealloc_cf = ceph_alloc_cap_flush();

	if (!prealloc_cf)

		return VM_FAULT_SIGBUS;



	if (ci->i_inline_version != CEPH_INLINE_NONE) {

		struct page *locked_page = NULL;

		if (off == 0) {
@@ -1323,8 +1328,10 @@ static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) 
		ret = ceph_uninline_data(vma->vm_file, locked_page);

		if (locked_page)

			unlock_page(locked_page);

		if (ret < 0)

			return VM_FAULT_SIGBUS;

		if (ret < 0) {

			ret = VM_FAULT_SIGBUS;

			goto out_free;

		}

	}



	if (off + PAGE_CACHE_SIZE <= size)
@@ -1346,7 +1353,8 @@ static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) 
			break;

		if (ret != -ERESTARTSYS) {

			WARN_ON(1);

			return VM_FAULT_SIGBUS;

			ret = VM_FAULT_SIGBUS;

			goto out_free;

		}

	}

	dout("page_mkwrite %p %llu~%zd got cap refs on %s\n",
@@ -1381,7 +1389,8 @@ static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) 
		int dirty;

		spin_lock(&ci->i_ceph_lock);

		ci->i_inline_version = CEPH_INLINE_NONE;

		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR);

		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR,

					       &prealloc_cf);

		spin_unlock(&ci->i_ceph_lock);

		if (dirty)

			__mark_inode_dirty(inode, dirty);
@@ -1390,6 +1399,8 @@ static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) 
	dout("page_mkwrite %p %llu~%zd dropping cap refs on %s ret %d\n",

	     inode, off, len, ceph_cap_string(got), ret);

	ceph_put_cap_refs(ci, got);

out_free:

	ceph_free_cap_flush(prealloc_cf);



	return ret;

}

fs/ceph/caps.c

+22 −4
Original line number Diff line number Diff line
@@ -1356,7 +1356,8 @@ static void ceph_flush_snaps(struct ceph_inode_info *ci) 
 * Caller is then responsible for calling __mark_inode_dirty with the

 * returned flags value.

 */

int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask)

int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,

			   struct ceph_cap_flush **pcf)

{

	struct ceph_mds_client *mdsc =

		ceph_sb_to_client(ci->vfs_inode.i_sb)->mdsc;
@@ -1376,6 +1377,9 @@ int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask) 
	     ceph_cap_string(was | mask));

	ci->i_dirty_caps |= mask;

	if (was == 0) {

		WARN_ON_ONCE(ci->i_prealloc_cap_flush);

		swap(ci->i_prealloc_cap_flush, *pcf);



		if (!ci->i_head_snapc) {

			WARN_ON_ONCE(!rwsem_is_locked(&mdsc->snap_rwsem));

			ci->i_head_snapc = ceph_get_snap_context(
@@ -1391,6 +1395,8 @@ int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask) 
			ihold(inode);

			dirty |= I_DIRTY_SYNC;

		}

	} else {

		WARN_ON_ONCE(!ci->i_prealloc_cap_flush);

	}

	BUG_ON(list_empty(&ci->i_dirty_item));

	if (((was | ci->i_flushing_caps) & CEPH_CAP_FILE_BUFFER) &&
@@ -1446,6 +1452,17 @@ static void __add_cap_flushing_to_mdsc(struct ceph_mds_client *mdsc, 
	rb_insert_color(&cf->g_node, &mdsc->cap_flush_tree);

}



struct ceph_cap_flush *ceph_alloc_cap_flush(void)

{

	return kmem_cache_alloc(ceph_cap_flush_cachep, GFP_KERNEL);

}



void ceph_free_cap_flush(struct ceph_cap_flush *cf)

{

	if (cf)

		kmem_cache_free(ceph_cap_flush_cachep, cf);

}



static u64 __get_oldest_flush_tid(struct ceph_mds_client *mdsc)

{

	struct rb_node *n = rb_first(&mdsc->cap_flush_tree);
@@ -1469,11 +1486,12 @@ static int __mark_caps_flushing(struct inode *inode, 
{

	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;

	struct ceph_inode_info *ci = ceph_inode(inode);

	struct ceph_cap_flush *cf;

	struct ceph_cap_flush *cf = NULL;

	int flushing;



	BUG_ON(ci->i_dirty_caps == 0);

	BUG_ON(list_empty(&ci->i_dirty_item));

	BUG_ON(!ci->i_prealloc_cap_flush);



	flushing = ci->i_dirty_caps;

	dout("__mark_caps_flushing flushing %s, flushing_caps %s -> %s\n",
@@ -1484,7 +1502,7 @@ static int __mark_caps_flushing(struct inode *inode, 
	ci->i_dirty_caps = 0;

	dout(" inode %p now !dirty\n", inode);



	cf = kmalloc(sizeof(*cf), GFP_ATOMIC);

	swap(cf, ci->i_prealloc_cap_flush);

	cf->caps = flushing;

	cf->kick = false;
@@ -3075,7 +3093,7 @@ static void handle_cap_flush_ack(struct inode *inode, u64 flush_tid, 
		cf = list_first_entry(&to_remove,

				      struct ceph_cap_flush, list);

		list_del(&cf->list);

		kfree(cf);

		ceph_free_cap_flush(cf);

	}

	if (drop)

		iput(inode);

fs/ceph/file.c

+16 −2
Original line number Diff line number Diff line
@@ -939,6 +939,7 @@ static ssize_t ceph_write_iter(struct kiocb *iocb, struct iov_iter *from) 
	struct ceph_inode_info *ci = ceph_inode(inode);

	struct ceph_osd_client *osdc =

		&ceph_sb_to_client(inode->i_sb)->client->osdc;

	struct ceph_cap_flush *prealloc_cf;

	ssize_t count, written = 0;

	int err, want, got;

	loff_t pos;
@@ -946,6 +947,10 @@ static ssize_t ceph_write_iter(struct kiocb *iocb, struct iov_iter *from) 
	if (ceph_snap(inode) != CEPH_NOSNAP)

		return -EROFS;



	prealloc_cf = ceph_alloc_cap_flush();

	if (!prealloc_cf)

		return -ENOMEM;



	mutex_lock(&inode->i_mutex);



	/* We can write back this queue in page reclaim */
@@ -1050,7 +1055,8 @@ static ssize_t ceph_write_iter(struct kiocb *iocb, struct iov_iter *from) 
		int dirty;

		spin_lock(&ci->i_ceph_lock);

		ci->i_inline_version = CEPH_INLINE_NONE;

		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR);

		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR,

					       &prealloc_cf);

		spin_unlock(&ci->i_ceph_lock);

		if (dirty)

			__mark_inode_dirty(inode, dirty);
@@ -1074,6 +1080,7 @@ static ssize_t ceph_write_iter(struct kiocb *iocb, struct iov_iter *from) 
out:

	mutex_unlock(&inode->i_mutex);

out_unlocked:

	ceph_free_cap_flush(prealloc_cf);

	current->backing_dev_info = NULL;

	return written ? written : err;

}
@@ -1270,6 +1277,7 @@ static long ceph_fallocate(struct file *file, int mode, 
	struct ceph_inode_info *ci = ceph_inode(inode);

	struct ceph_osd_client *osdc =

		&ceph_inode_to_client(inode)->client->osdc;

	struct ceph_cap_flush *prealloc_cf;

	int want, got = 0;

	int dirty;

	int ret = 0;
@@ -1282,6 +1290,10 @@ static long ceph_fallocate(struct file *file, int mode, 
	if (!S_ISREG(inode->i_mode))

		return -EOPNOTSUPP;



	prealloc_cf = ceph_alloc_cap_flush();

	if (!prealloc_cf)

		return -ENOMEM;



	mutex_lock(&inode->i_mutex);



	if (ceph_snap(inode) != CEPH_NOSNAP) {
@@ -1328,7 +1340,8 @@ static long ceph_fallocate(struct file *file, int mode, 
	if (!ret) {

		spin_lock(&ci->i_ceph_lock);

		ci->i_inline_version = CEPH_INLINE_NONE;

		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR);

		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR,

					       &prealloc_cf);

		spin_unlock(&ci->i_ceph_lock);

		if (dirty)

			__mark_inode_dirty(inode, dirty);
@@ -1337,6 +1350,7 @@ static long ceph_fallocate(struct file *file, int mode, 
	ceph_put_cap_refs(ci, got);

unlock:

	mutex_unlock(&inode->i_mutex);

	ceph_free_cap_flush(prealloc_cf);

	return ret;

}

fs/ceph/inode.c

+13 −2
Original line number Diff line number Diff line
@@ -416,6 +416,7 @@ struct inode *ceph_alloc_inode(struct super_block *sb) 
	ci->i_flushing_caps = 0;

	INIT_LIST_HEAD(&ci->i_dirty_item);

	INIT_LIST_HEAD(&ci->i_flushing_item);

	ci->i_prealloc_cap_flush = NULL;

	ci->i_cap_flush_tree = RB_ROOT;

	init_waitqueue_head(&ci->i_cap_wq);

	ci->i_hold_caps_min = 0;
@@ -1720,6 +1721,7 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) 
	const unsigned int ia_valid = attr->ia_valid;

	struct ceph_mds_request *req;

	struct ceph_mds_client *mdsc = ceph_sb_to_client(dentry->d_sb)->mdsc;

	struct ceph_cap_flush *prealloc_cf;

	int issued;

	int release = 0, dirtied = 0;

	int mask = 0;
@@ -1734,10 +1736,16 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) 
	if (err != 0)

		return err;



	prealloc_cf = ceph_alloc_cap_flush();

	if (!prealloc_cf)

		return -ENOMEM;



	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SETATTR,

				       USE_AUTH_MDS);

	if (IS_ERR(req))

	if (IS_ERR(req)) {

		ceph_free_cap_flush(prealloc_cf);

		return PTR_ERR(req);

	}



	spin_lock(&ci->i_ceph_lock);

	issued = __ceph_caps_issued(ci, NULL);
@@ -1895,7 +1903,8 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) 
		dout("setattr %p ATTR_FILE ... hrm!\n", inode);



	if (dirtied) {

		inode_dirty_flags = __ceph_mark_dirty_caps(ci, dirtied);

		inode_dirty_flags = __ceph_mark_dirty_caps(ci, dirtied,

							   &prealloc_cf);

		inode->i_ctime = CURRENT_TIME;

	}
@@ -1927,9 +1936,11 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) 
	ceph_mdsc_put_request(req);

	if (mask & CEPH_SETATTR_SIZE)

		__ceph_do_pending_vmtruncate(inode);

	ceph_free_cap_flush(prealloc_cf);

	return err;

out_put:

	ceph_mdsc_put_request(req);

	ceph_free_cap_flush(prealloc_cf);

	return err;

}

fs/ceph/mds_client.c

+5 −1
Original line number Diff line number Diff line
@@ -1189,6 +1189,10 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap, 
		}

		spin_unlock(&mdsc->cap_dirty_lock);



		if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {

			list_add(&ci->i_prealloc_cap_flush->list, &to_remove);

			ci->i_prealloc_cap_flush = NULL;

		}

	}

	spin_unlock(&ci->i_ceph_lock);

	while (!list_empty(&to_remove)) {
@@ -1196,7 +1200,7 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap, 
		cf = list_first_entry(&to_remove,

				      struct ceph_cap_flush, list);

		list_del(&cf->list);

		kfree(cf);

		ceph_free_cap_flush(cf);

	}

	while (drop--)

		iput(inode);