ocfs2: Support xfs style space reservation ioctls

We re-use the RESVSP/UNRESVSP ioctls from xfs which allow the user to
allocate and deallocate regions to a file without zeroing data or changing
i_size.

Though renamed, the structure passed in from user is identical to struct
xfs_flock64. The three fields that are actually used right now are l_whence,
l_start and l_len.

This should get ocfs2 immediate compatibility with userspace software using
the pre-existing xfs ioctls.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>

6 files changed, 216 insertions, 16 deletions

diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 11f7cf9f2511..f04c7aa834cb 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -1111,17 +1111,16 @@ out:
         return ret;
 }
 
-static int ocfs2_write_remove_suid(struct inode *inode)
+static int __ocfs2_write_remove_suid(struct inode *inode,
+                                     struct buffer_head *bh)
 {
         int ret;
-        struct buffer_head *bh = NULL;
-        struct ocfs2_inode_info *oi = OCFS2_I(inode);
         handle_t *handle;
         struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
         struct ocfs2_dinode *di;
 
         mlog_entry("(Inode %llu, mode 0%o)\n",
-                   (unsigned long long)oi->ip_blkno, inode->i_mode);
+                   (unsigned long long)OCFS2_I(inode)->ip_blkno, inode->i_mode);
 
         handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
         if (handle == NULL) {
@@ -1130,17 +1129,11 @@ static int ocfs2_write_remove_suid(struct inode *inode)
                 goto out;
         }
 
-        ret = ocfs2_read_block(osb, oi->ip_blkno, &bh, OCFS2_BH_CACHED, inode);
-        if (ret < 0) {
-                mlog_errno(ret);
-                goto out_trans;
-        }
-
         ret = ocfs2_journal_access(handle, inode, bh,
                                    OCFS2_JOURNAL_ACCESS_WRITE);
         if (ret < 0) {
                 mlog_errno(ret);
-                goto out_bh;
+                goto out_trans;
         }
 
         inode->i_mode &= ~S_ISUID;
@@ -1153,8 +1146,7 @@ static int ocfs2_write_remove_suid(struct inode *inode)
         ret = ocfs2_journal_dirty(handle, bh);
         if (ret < 0)
                 mlog_errno(ret);
-out_bh:
-        brelse(bh);
+
 out_trans:
         ocfs2_commit_trans(osb, handle);
 out:
@@ -1200,6 +1192,25 @@ out:
         return ret;
 }
 
+static int ocfs2_write_remove_suid(struct inode *inode)
+{
+        int ret;
+        struct buffer_head *bh = NULL;
+        struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+        ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
+                               oi->ip_blkno, &bh, OCFS2_BH_CACHED, inode);
+        if (ret < 0) {
+                mlog_errno(ret);
+                goto out;
+        }
+
+        ret =  __ocfs2_write_remove_suid(inode, bh);
+out:
+        brelse(bh);
+        return ret;
+}
+
 /*
  * Allocate enough extents to cover the region starting at byte offset
  * start for len bytes. Existing extents are skipped, any extents
@@ -1490,6 +1501,151 @@ out:
         return ret;
 }
 
+/*
+ * Parts of this function taken from xfs_change_file_space()
+ */
+int ocfs2_change_file_space(struct file *file, unsigned int cmd,
+                            struct ocfs2_space_resv *sr)
+{
+        int ret;
+        s64 llen;
+        struct inode *inode = file->f_path.dentry->d_inode;
+        struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+        struct buffer_head *di_bh = NULL;
+        handle_t *handle;
+        unsigned long long max_off = ocfs2_max_file_offset(inode->i_sb->s_blocksize_bits);
+
+        if ((cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) &&
+            !ocfs2_writes_unwritten_extents(osb))
+                return -ENOTTY;
+        else if ((cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) &&
+                 !ocfs2_sparse_alloc(osb))
+                return -ENOTTY;
+
+        if (!S_ISREG(inode->i_mode))
+                return -EINVAL;
+
+        if (!(file->f_mode & FMODE_WRITE))
+                return -EBADF;
+
+        if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
+                return -EROFS;
+
+        mutex_lock(&inode->i_mutex);
+
+        /*
+         * This prevents concurrent writes on other nodes
+         */
+        ret = ocfs2_rw_lock(inode, 1);
+        if (ret) {
+                mlog_errno(ret);
+                goto out;
+        }
+
+        ret = ocfs2_meta_lock(inode, &di_bh, 1);
+        if (ret) {
+                mlog_errno(ret);
+                goto out_rw_unlock;
+        }
+
+        if (inode->i_flags & (S_IMMUTABLE|S_APPEND)) {
+                ret = -EPERM;
+                goto out_meta_unlock;
+        }
+
+        switch (sr->l_whence) {
+        case 0: /*SEEK_SET*/
+                break;
+        case 1: /*SEEK_CUR*/
+                sr->l_start += file->f_pos;
+                break;
+        case 2: /*SEEK_END*/
+                sr->l_start += i_size_read(inode);
+                break;
+        default:
+                ret = -EINVAL;
+                goto out_meta_unlock;
+        }
+        sr->l_whence = 0;
+
+        llen = sr->l_len > 0 ? sr->l_len - 1 : sr->l_len;
+
+        if (sr->l_start < 0
+            || sr->l_start > max_off
+            || (sr->l_start + llen) < 0
+            || (sr->l_start + llen) > max_off) {
+                ret = -EINVAL;
+                goto out_meta_unlock;
+        }
+
+        if (cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) {
+                if (sr->l_len <= 0) {
+                        ret = -EINVAL;
+                        goto out_meta_unlock;
+                }
+        }
+
+        if (should_remove_suid(file->f_path.dentry)) {
+                ret = __ocfs2_write_remove_suid(inode, di_bh);
+                if (ret) {
+                        mlog_errno(ret);
+                        goto out_meta_unlock;
+                }
+        }
+
+        down_write(&OCFS2_I(inode)->ip_alloc_sem);
+        switch (cmd) {
+        case OCFS2_IOC_RESVSP:
+        case OCFS2_IOC_RESVSP64:
+                /*
+                 * This takes unsigned offsets, but the signed ones we
+                 * pass have been checked against overflow above.
+                 */
+                ret = ocfs2_allocate_unwritten_extents(inode, sr->l_start,
+                                                       sr->l_len);
+                break;
+        case OCFS2_IOC_UNRESVSP:
+        case OCFS2_IOC_UNRESVSP64:
+                ret = ocfs2_remove_inode_range(inode, di_bh, sr->l_start,
+                                               sr->l_len);
+                break;
+        default:
+                ret = -EINVAL;
+        }
+        up_write(&OCFS2_I(inode)->ip_alloc_sem);
+        if (ret) {
+                mlog_errno(ret);
+                goto out_meta_unlock;
+        }
+
+        /*
+         * We update c/mtime for these changes
+         */
+        handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
+        if (IS_ERR(handle)) {
+                ret = PTR_ERR(handle);
+                mlog_errno(ret);
+                goto out_meta_unlock;
+        }
+
+        inode->i_ctime = inode->i_mtime = CURRENT_TIME;
+        ret = ocfs2_mark_inode_dirty(handle, inode, di_bh);
+        if (ret < 0)
+                mlog_errno(ret);
+
+        ocfs2_commit_trans(osb, handle);
+
+out_meta_unlock:
+        brelse(di_bh);
+        ocfs2_meta_unlock(inode, 1);
+out_rw_unlock:
+        ocfs2_rw_unlock(inode, 1);
+
+        mutex_unlock(&inode->i_mutex);
+out:
+        return ret;
+}
+
 static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
                                          loff_t *ppos,
                                          size_t count,
diff --git a/fs/ocfs2/file.h b/fs/ocfs2/file.h
index 79115c92dc30..36fe27f268ee 100644
--- a/fs/ocfs2/file.h
+++ b/fs/ocfs2/file.h
@@ -62,4 +62,7 @@ int ocfs2_should_update_atime(struct inode *inode,
 int ocfs2_update_inode_atime(struct inode *inode,
                              struct buffer_head *bh);
 
+int ocfs2_change_file_space(struct file *file, unsigned int cmd,
+                            struct ocfs2_space_resv *sr);
+
 #endif /* OCFS2_FILE_H */
diff --git a/fs/ocfs2/ioctl.c b/fs/ocfs2/ioctl.c
index f3ad21ad9aed..bd68c3f2afbe 100644
--- a/fs/ocfs2/ioctl.c
+++ b/fs/ocfs2/ioctl.c
@@ -14,6 +14,7 @@
 #include "ocfs2.h"
 #include "alloc.h"
 #include "dlmglue.h"
+#include "file.h"
 #include "inode.h"
 #include "journal.h"
 
@@ -115,6 +116,7 @@ int ocfs2_ioctl(struct inode * inode, struct file * filp,
 {
         unsigned int flags;
         int status;
+        struct ocfs2_space_resv sr;
 
         switch (cmd) {
         case OCFS2_IOC_GETFLAGS:
@@ -130,6 +132,14 @@ int ocfs2_ioctl(struct inode * inode, struct file * filp,
 
                 return ocfs2_set_inode_attr(inode, flags,
                         OCFS2_FL_MODIFIABLE);
+        case OCFS2_IOC_RESVSP:
+        case OCFS2_IOC_RESVSP64:
+        case OCFS2_IOC_UNRESVSP:
+        case OCFS2_IOC_UNRESVSP64:
+                if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
+                        return -EFAULT;
+
+                return ocfs2_change_file_space(filp, cmd, &sr);
         default:
                 return -ENOTTY;
         }
@@ -148,6 +158,11 @@ long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
         case OCFS2_IOC32_SETFLAGS:
                 cmd = OCFS2_IOC_SETFLAGS;
                 break;
+        case OCFS2_IOC_RESVSP:
+        case OCFS2_IOC_RESVSP64:
+        case OCFS2_IOC_UNRESVSP:
+        case OCFS2_IOC_UNRESVSP64:
+                break;
         default:
                 return -ENOIOCTLCMD;
         }
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index c20a74b99d87..82f8a75b207e 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -175,6 +175,32 @@
 #define OCFS2_IOC32_SETFLAGS    _IOW('f', 2, int)
 
 /*
+ * Space reservation / allocation / free ioctls and argument structure
+ * are designed to be compatible with XFS.
+ *
+ * ALLOCSP* and FREESP* are not and will never be supported, but are
+ * included here for completeness.
+ */
+struct ocfs2_space_resv {
+        __s16           l_type;
+        __s16           l_whence;
+        __s64           l_start;
+        __s64           l_len;          /* len == 0 means until end of file */
+        __s32           l_sysid;
+        __u32           l_pid;
+        __s32           l_pad[4];       /* reserve area                     */
+};
+
+#define OCFS2_IOC_ALLOCSP               _IOW ('X', 10, struct ocfs2_space_resv)
+#define OCFS2_IOC_FREESP                _IOW ('X', 11, struct ocfs2_space_resv)
+#define OCFS2_IOC_RESVSP                _IOW ('X', 40, struct ocfs2_space_resv)
+#define OCFS2_IOC_UNRESVSP      _IOW ('X', 41, struct ocfs2_space_resv)
+#define OCFS2_IOC_ALLOCSP64     _IOW ('X', 36, struct ocfs2_space_resv)
+#define OCFS2_IOC_FREESP64      _IOW ('X', 37, struct ocfs2_space_resv)
+#define OCFS2_IOC_RESVSP64      _IOW ('X', 42, struct ocfs2_space_resv)
+#define OCFS2_IOC_UNRESVSP64    _IOW ('X', 43, struct ocfs2_space_resv)
+
+/*
  * Journal Flags (ocfs2_dinode.id1.journal1.i_flags)
  */
 #define OCFS2_JOURNAL_DIRTY_FL  (0x00000001)    /* Journal needs recovery */
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index f07718a7552b..3a5a1ed09ac9 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -115,8 +115,6 @@ static void ocfs2_write_super(struct super_block *sb);
 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 static void ocfs2_destroy_inode(struct inode *inode);
 
-static unsigned long long ocfs2_max_file_offset(unsigned int blockshift);
-
 static const struct super_operations ocfs2_sops = {
         .statfs         = ocfs2_statfs,
         .alloc_inode    = ocfs2_alloc_inode,
@@ -321,7 +319,7 @@ static void ocfs2_destroy_inode(struct inode *inode)
 /* From xfs_super.c:xfs_max_file_offset
  * Copyright (c) 2000-2004 Silicon Graphics, Inc.
  */
-static unsigned long long ocfs2_max_file_offset(unsigned int blockshift)
+unsigned long long ocfs2_max_file_offset(unsigned int blockshift)
 {
         unsigned int pagefactor = 1;
         unsigned int bitshift = BITS_PER_LONG - 1;
diff --git a/fs/ocfs2/super.h b/fs/ocfs2/super.h
index 783f5270f2a1..3b9cb3d0b008 100644
--- a/fs/ocfs2/super.h
+++ b/fs/ocfs2/super.h
@@ -45,4 +45,6 @@ void __ocfs2_abort(struct super_block *sb,
 
 #define ocfs2_abort(sb, fmt, args...) __ocfs2_abort(sb, __PRETTY_FUNCTION__, fmt, ##args)
 
+unsigned long long ocfs2_max_file_offset(unsigned int blockshift);
+
 #endif /* OCFS2_SUPER_H */