Skip to contentSkip to conversation

ceph: only allow punch hole mode in fallocate

PATCH Applied Luis Henriques 3 participants 5 messages Oct 9, 2018 – Oct 10, 2018 Ceph client

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ceph: only allow punch hole mode in fallocate

Current implementation of cephfs fallocate isn't correct as it doesn't really reserve the space in the cluster, which means that a subsequent call to a write may actually fail due to lack of space. In fact, it is currently possible to fallocate an amount space that is larger than the free space in the cluster.

Since there's no easy solution to fix this at the moment, this patch simply removes support for all fallocate operations but FALLOC_FL_PUNCH_HOLE (which implies FALLOC_FL_KEEP_SIZE).

Fixes ad7a60de882a ("ceph: punch hole support")
---
 fs/ceph/file.c | 45 +++++++++------------------------------------
 1 file changed, 9 insertions(+), 36 deletions(-)
fs/ceph/file.c +9 -36
index 92ab20433682..91a7ad259bcf 100644
@@ -1735,7 +1735,6 @@ static long ceph_fallocate(struct file *file, int mode,
 	struct ceph_file_info *fi = file->private_data;
 	struct inode *inode = file_inode(file);
 	struct ceph_inode_info *ci = ceph_inode(inode);
-	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
 	struct ceph_cap_flush *prealloc_cf;
 	int want, got = 0;
 	int dirty;
@@ -1743,10 +1742,7 @@ static long ceph_fallocate(struct file *file, int mode,
 	loff_t endoff = 0;
 	loff_t size;
 
-	if ((offset + length) > max(i_size_read(inode), fsc->max_file_size))
-		return -EFBIG;
-
-	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
+	if (mode != (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
 		return -EOPNOTSUPP;
 
 	if (!S_ISREG(inode->i_mode))
@@ -1763,18 +1759,6 @@ static long ceph_fallocate(struct file *file, int mode,
 		goto unlock;
 	}
 
-	if (!(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE)) &&
-	    ceph_quota_is_max_bytes_exceeded(inode, offset + length)) {
-		ret = -EDQUOT;
-		goto unlock;
-	}
-
-	if (ceph_osdmap_flag(&fsc->client->osdc, CEPH_OSDMAP_FULL) &&
-	    !(mode & FALLOC_FL_PUNCH_HOLE)) {
-		ret = -ENOSPC;
-		goto unlock;
-	}
-
 	if (ci->i_inline_version != CEPH_INLINE_NONE) {
 		ret = ceph_uninline_data(file, NULL);
 		if (ret < 0)
@@ -1782,12 +1766,12 @@ static long ceph_fallocate(struct file *file, int mode,
 	}
 
 	size = i_size_read(inode);
-	if (!(mode & FALLOC_FL_KEEP_SIZE)) {
-		endoff = offset + length;
-		ret = inode_newsize_ok(inode, endoff);
-		if (ret)
-			goto unlock;
-	}
+
+	/* Are we punching a hole beyond EOF? */
+	if (offset >= size)
+		goto unlock;
+	if ((offset + length) > size)
+		length = size - offset;
 
 	if (fi->fmode & CEPH_FILE_MODE_LAZY)
 		want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO;
@@ -1798,16 +1782,8 @@ static long ceph_fallocate(struct file *file, int mode,
 	if (ret < 0)
 		goto unlock;
 
-	if (mode & FALLOC_FL_PUNCH_HOLE) {
-		if (offset < size)
-			ceph_zero_pagecache_range(inode, offset, length);
-		ret = ceph_zero_objects(inode, offset, length);
-	} else if (endoff > size) {
-		truncate_pagecache_range(inode, size, -1);
-		if (ceph_inode_set_size(inode, endoff))
-			ceph_check_caps(ceph_inode(inode),
-				CHECK_CAPS_AUTHONLY, NULL);
-	}
+	ceph_zero_pagecache_range(inode, offset, length);
+	ret = ceph_zero_objects(inode, offset, length);
 
 	if (!ret) {
 		spin_lock(&ci->i_ceph_lock);
@@ -1817,9 +1793,6 @@ static long ceph_fallocate(struct file *file, int mode,
 		spin_unlock(&ci->i_ceph_lock);
 		if (dirty)
 			__mark_inode_dirty(inode, dirty);
-		if ((endoff > size) &&
-		    ceph_quota_is_max_bytes_approaching(inode, endoff))
-			ceph_check_caps(ci, CHECK_CAPS_NODELAY, NULL);
 	}
 
 	ceph_put_cap_refs(ci, got);
Applied
#1 lore raw
Quoted 101 lines · Luis Henriques

Current implementation of cephfs fallocate isn't correct as it doesn't really reserve the space in the cluster, which means that a subsequent call to a write may actually fail due to lack of space. In fact, it is currently possible to fallocate an amount space that is larger than the free space in the cluster.

Since there's no easy solution to fix this at the moment, this patch simply removes support for all fallocate operations but FALLOC_FL_PUNCH_HOLE (which implies FALLOC_FL_KEEP_SIZE).

Fixes ad7a60de882a ("ceph: punch hole support")
---
 fs/ceph/file.c | 45 +++++++++------------------------------------
 1 file changed, 9 insertions(+), 36 deletions(-)
fs/ceph/file.c +9 -33
index 92ab20433682..91a7ad259bcf 100644
@@ -1735,7 +1735,6 @@ static long ceph_fallocate(struct file *file, int mode,
        struct ceph_file_info *fi = file->private_data;
        struct inode *inode = file_inode(file);
        struct ceph_inode_info *ci = ceph_inode(inode);
-       struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
        struct ceph_cap_flush *prealloc_cf;
        int want, got = 0;
        int dirty;
@@ -1743,10 +1742,7 @@ static long ceph_fallocate(struct file *file, int mode,
        loff_t endoff = 0;
        loff_t size;
 
-       if ((offset + length) > max(i_size_read(inode), fsc->max_file_size))
-               return -EFBIG;
-
-       if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
+       if (mode != (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
                return -EOPNOTSUPP;
 
        if (!S_ISREG(inode->i_mode))
@@ -1763,18 +1759,6 @@ static long ceph_fallocate(struct file *file, int mode,
                goto unlock;
        }
 
-       if (!(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE)) &&
-           ceph_quota_is_max_bytes_exceeded(inode, offset + length)) {
-               ret = -EDQUOT;
-               goto unlock;
-       }
-
-       if (ceph_osdmap_flag(&fsc->client->osdc, CEPH_OSDMAP_FULL) &&
-           !(mode & FALLOC_FL_PUNCH_HOLE)) {
-               ret = -ENOSPC;
-               goto unlock;
-       }
-
        if (ci->i_inline_version != CEPH_INLINE_NONE) {
                ret = ceph_uninline_data(file, NULL);
                if (ret < 0)
@@ -1782,12 +1766,12 @@ static long ceph_fallocate(struct file *file, int mode,
        }
 
        size = i_size_read(inode);
-       if (!(mode & FALLOC_FL_KEEP_SIZE)) {
-               endoff = offset + length;
-               ret = inode_newsize_ok(inode, endoff);
-               if (ret)
-                       goto unlock;
-       }
+
+       /* Are we punching a hole beyond EOF? */
+       if (offset >= size)
+               goto unlock;
+       if ((offset + length) > size)
+               length = size - offset;
 
        if (fi->fmode & CEPH_FILE_MODE_LAZY)
                want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO;
@@ -1798,16 +1782,8 @@ static long ceph_fallocate(struct file *file, int mode,
        if (ret < 0)
                goto unlock;
 
-       if (mode & FALLOC_FL_PUNCH_HOLE) {
-               if (offset < size)
-                       ceph_zero_pagecache_range(inode, offset, length);
-               ret = ceph_zero_objects(inode, offset, length);
-       } else if (endoff > size) {
-               truncate_pagecache_range(inode, size, -1);
-               if (ceph_inode_set_size(inode, endoff))
-                       ceph_check_caps(ceph_inode(inode),
-                               CHECK_CAPS_AUTHONLY, NULL);
-       }
+       ceph_zero_pagecache_range(inode, offset, length);
+       ret = ceph_zero_objects(inode, offset, length);
 
        if (!ret) {
                spin_lock(&ci->i_ceph_lock);

fs/ceph/file.c static long ceph_fallocate(struct file *file, int mode, · line 1798

Quoted 9 lines
fs/ceph/file.c +0 -3
@@ -1817,9 +1793,6 @@ static long ceph_fallocate(struct file *file, int mode,
                spin_unlock(&ci->i_ceph_lock);
                if (dirty)
                        __mark_inode_dirty(inode, dirty);
-               if ((endoff > size) &&
-                   ceph_quota_is_max_bytes_approaching(inode, endoff))
-                       ceph_check_caps(ci, CHECK_CAPS_NODELAY, NULL);
        }
 
        ceph_put_cap_refs(ci, got);

Applied, thanks

Yan, Zheng

maintainer
#2 lore raw
Quoted 111 lines · Zheng
Quoted 109 lines · Luis Henriques

Current implementation of cephfs fallocate isn't correct as it doesn't really reserve the space in the cluster, which means that a subsequent call to a write may actually fail due to lack of space. In fact, it is currently possible to fallocate an amount space that is larger than the free space in the cluster.

Since there's no easy solution to fix this at the moment, this patch simply removes support for all fallocate operations but FALLOC_FL_PUNCH_HOLE (which implies FALLOC_FL_KEEP_SIZE).

Fixes ad7a60de882a ("ceph: punch hole support")
---
 fs/ceph/file.c | 45 +++++++++------------------------------------
 1 file changed, 9 insertions(+), 36 deletions(-)
fs/ceph/file.c +9 -36
index 92ab20433682..91a7ad259bcf 100644
@@ -1735,7 +1735,6 @@ static long ceph_fallocate(struct file *file, int mode,
        struct ceph_file_info *fi = file->private_data;
        struct inode *inode = file_inode(file);
        struct ceph_inode_info *ci = ceph_inode(inode);
-       struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
        struct ceph_cap_flush *prealloc_cf;
        int want, got = 0;
        int dirty;
@@ -1743,10 +1742,7 @@ static long ceph_fallocate(struct file *file, int mode,
        loff_t endoff = 0;
        loff_t size;
 
-       if ((offset + length) > max(i_size_read(inode), fsc->max_file_size))
-               return -EFBIG;
-
-       if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
+       if (mode != (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
                return -EOPNOTSUPP;
 
        if (!S_ISREG(inode->i_mode))
@@ -1763,18 +1759,6 @@ static long ceph_fallocate(struct file *file, int mode,
                goto unlock;
        }
 
-       if (!(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE)) &&
-           ceph_quota_is_max_bytes_exceeded(inode, offset + length)) {
-               ret = -EDQUOT;
-               goto unlock;
-       }
-
-       if (ceph_osdmap_flag(&fsc->client->osdc, CEPH_OSDMAP_FULL) &&
-           !(mode & FALLOC_FL_PUNCH_HOLE)) {
-               ret = -ENOSPC;
-               goto unlock;
-       }
-
        if (ci->i_inline_version != CEPH_INLINE_NONE) {
                ret = ceph_uninline_data(file, NULL);
                if (ret < 0)
@@ -1782,12 +1766,12 @@ static long ceph_fallocate(struct file *file, int mode,
        }
 
        size = i_size_read(inode);
-       if (!(mode & FALLOC_FL_KEEP_SIZE)) {
-               endoff = offset + length;
-               ret = inode_newsize_ok(inode, endoff);
-               if (ret)
-                       goto unlock;
-       }
+
+       /* Are we punching a hole beyond EOF? */
+       if (offset >= size)
+               goto unlock;
+       if ((offset + length) > size)
+               length = size - offset;
 
        if (fi->fmode & CEPH_FILE_MODE_LAZY)
                want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO;
@@ -1798,16 +1782,8 @@ static long ceph_fallocate(struct file *file, int mode,
        if (ret < 0)
                goto unlock;
 
-       if (mode & FALLOC_FL_PUNCH_HOLE) {
-               if (offset < size)
-                       ceph_zero_pagecache_range(inode, offset, length);
-               ret = ceph_zero_objects(inode, offset, length);
-       } else if (endoff > size) {
-               truncate_pagecache_range(inode, size, -1);
-               if (ceph_inode_set_size(inode, endoff))
-                       ceph_check_caps(ceph_inode(inode),
-                               CHECK_CAPS_AUTHONLY, NULL);
-       }
+       ceph_zero_pagecache_range(inode, offset, length);
+       ret = ceph_zero_objects(inode, offset, length);
 
        if (!ret) {
                spin_lock(&ci->i_ceph_lock);
@@ -1817,9 +1793,6 @@ static long ceph_fallocate(struct file *file, int mode,
                spin_unlock(&ci->i_ceph_lock);
                if (dirty)
                        __mark_inode_dirty(inode, dirty);
-               if ((endoff > size) &&
-                   ceph_quota_is_max_bytes_approaching(inode, endoff))
-                       ceph_check_caps(ci, CHECK_CAPS_NODELAY, NULL);
        }
 
Quoted 3 lines

fs/ceph/file.c static long ceph_fallocate(struct file *file, int mode, · line 1798

Quoted 1 line
fs/ceph/file.c +0 -0
        ceph_put_cap_refs(ci, got);

Applied, thanks

I don't think it should go to stable kernels. Strictly speaking it's a behaviour change -- it's been this way for many years and, unless you are close to ENOSPC, it's sort of appears to work. I'll take off the stable tag unless I hear objections.

Thanks,

Ilya

#3 lore raw
Quoted 3 lines · Ilya Dryomov
Quoted 2 lines · Zheng

On Wed, Oct 10, 2018 at 1:54 AM Luis Henriques <[email protected]> wrote:

<snip>

Quoted 6 lines
Quoted 1 line

Applied, thanks

I don't think it should go to stable kernels. Strictly speaking it's a behaviour change -- it's been this way for many years and, unless you are close to ENOSPC, it's sort of appears to work. I'll take off the stable tag unless I hear objections.

Right, it can in fact break applications that rely on the previous (bogus) behaviour. But it can also be claimed that it *will* break applications anyway with an updated kernel, so backporting it to older kernels will just allow a consistent behaviour.

Anyway, I'm OK either way. But if you drop the stable tag make sure you also remove the 'Fixes:' tag as I believe the stable folks will still pick this patch if it includes a valid SHA1 in it.

Cheers,

Signature · 1 line
Luis
maintainer
#4 lore raw
Quoted 16 lines · Luis Henriques
Quoted 3 lines · Ilya Dryomov
Quoted 2 lines · Zheng

On Wed, Oct 10, 2018 at 1:54 AM Luis Henriques <[email protected]> wrote:

<snip>

Quoted 6 lines
Quoted 1 line

Applied, thanks

I don't think it should go to stable kernels. Strictly speaking it's a behaviour change -- it's been this way for many years and, unless you are close to ENOSPC, it's sort of appears to work. I'll take off the stable tag unless I hear objections.

Quoted 8 lines

Right, it can in fact break applications that rely on the previous (bogus) behaviour. But it can also be claimed that it *will* break applications anyway with an updated kernel, so backporting it to older kernels will just allow a consistent behaviour.

Anyway, I'm OK either way. But if you drop the stable tag make sure you also remove the 'Fixes:' tag as I believe the stable folks will still pick this patch if it includes a valid SHA1 in it.

Yeah, we've run into this in the past.

Thanks,

Ilya