]> pilppa.com Git - linux-2.6-omap-h63xx.git/commitdiff
ocfs2: Add splice support
authorTiger Yang <tiger.yang@oracle.com>
Wed, 18 Oct 2006 01:29:52 +0000 (18:29 -0700)
committerMark Fasheh <mark.fasheh@oracle.com>
Sat, 2 Dec 2006 02:28:46 +0000 (18:28 -0800)
Add splice read/write support in ocfs2.

ocfs2_file_splice_read/write are very similar to ocfs2_file_aio_read/write.

Signed-off-by: Tiger Yang <tiger.yang@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
fs/ocfs2/file.c

index 39b2f1653e2554387bece37f0f6bbc43ccaa81ed..b32cdb3bf7c5abe7eeb5f9f9ac0f93049a52a258 100644 (file)
@@ -31,6 +31,7 @@
 #include <linux/pagemap.h>
 #include <linux/uio.h>
 #include <linux/sched.h>
+#include <linux/pipe_fs_i.h>
 
 #define MLOG_MASK_PREFIX ML_INODE
 #include <cluster/masklog.h>
@@ -943,53 +944,21 @@ out:
        return ret;
 }
 
-static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
-                                   const struct iovec *iov,
-                                   unsigned long nr_segs,
-                                   loff_t pos)
+static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
+                                        loff_t *ppos,
+                                        size_t count,
+                                        int appending)
 {
-       int ret, rw_level = -1, meta_level = -1, have_alloc_sem = 0;
+       int ret = 0, meta_level = appending;
+       struct inode *inode = dentry->d_inode;
        u32 clusters;
-       struct file *filp = iocb->ki_filp;
-       struct inode *inode = filp->f_dentry->d_inode;
        loff_t newsize, saved_pos;
 
-       mlog_entry("(0x%p, %u, '%.*s')\n", filp,
-                  (unsigned int)nr_segs,
-                  filp->f_dentry->d_name.len,
-                  filp->f_dentry->d_name.name);
-
-       /* happy write of zero bytes */
-       if (iocb->ki_left == 0)
-               return 0;
-
-       if (!inode) {
-               mlog(0, "bad inode\n");
-               return -EIO;
-       }
-
-       mutex_lock(&inode->i_mutex);
-       /* to match setattr's i_mutex -> i_alloc_sem -> rw_lock ordering */
-       if (filp->f_flags & O_DIRECT) {
-               have_alloc_sem = 1;
-               down_read(&inode->i_alloc_sem);
-       }
-
-       /* concurrent O_DIRECT writes are allowed */
-       rw_level = (filp->f_flags & O_DIRECT) ? 0 : 1;
-       ret = ocfs2_rw_lock(inode, rw_level);
-       if (ret < 0) {
-               rw_level = -1;
-               mlog_errno(ret);
-               goto out;
-       }
-
        /* 
         * We sample i_size under a read level meta lock to see if our write
         * is extending the file, if it is we back off and get a write level
         * meta lock.
         */
-       meta_level = (filp->f_flags & O_APPEND) ? 1 : 0;
        for(;;) {
                ret = ocfs2_meta_lock(inode, NULL, meta_level);
                if (ret < 0) {
@@ -1007,7 +976,7 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
                 * inode. There's also the dinode i_size state which
                 * can be lost via setattr during extending writes (we
                 * set inode->i_size at the end of a write. */
-               if (should_remove_suid(filp->f_dentry)) {
+               if (should_remove_suid(dentry)) {
                        if (meta_level == 0) {
                                ocfs2_meta_unlock(inode, meta_level);
                                meta_level = 1;
@@ -1017,19 +986,19 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
                        ret = ocfs2_write_remove_suid(inode);
                        if (ret < 0) {
                                mlog_errno(ret);
-                               goto out;
+                               goto out_unlock;
                        }
                }
 
                /* work on a copy of ppos until we're sure that we won't have
                 * to recalculate it due to relocking. */
-               if (filp->f_flags & O_APPEND) {
+               if (appending) {
                        saved_pos = i_size_read(inode);
                        mlog(0, "O_APPEND: inode->i_size=%llu\n", saved_pos);
                } else {
-                       saved_pos = iocb->ki_pos;
+                       saved_pos = *ppos;
                }
-               newsize = iocb->ki_left + saved_pos;
+               newsize = count + saved_pos;
 
                mlog(0, "pos=%lld newsize=%lld cursize=%lld\n",
                     (long long) saved_pos, (long long) newsize,
@@ -1062,19 +1031,66 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
                if (!clusters)
                        break;
 
-               ret = ocfs2_extend_file(inode, NULL, newsize, iocb->ki_left);
+               ret = ocfs2_extend_file(inode, NULL, newsize, count);
                if (ret < 0) {
                        if (ret != -ENOSPC)
                                mlog_errno(ret);
-                       goto out;
+                       goto out_unlock;
                }
                break;
        }
 
-       /* ok, we're done with i_size and alloc work */
-       iocb->ki_pos = saved_pos;
+       if (appending)
+               *ppos = saved_pos;
+
+out_unlock:
        ocfs2_meta_unlock(inode, meta_level);
-       meta_level = -1;
+
+out:
+       return ret;
+}
+
+static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
+                                   const struct iovec *iov,
+                                   unsigned long nr_segs,
+                                   loff_t pos)
+{
+       int ret, rw_level, have_alloc_sem = 0;
+       struct file *filp = iocb->ki_filp;
+       struct inode *inode = filp->f_dentry->d_inode;
+       int appending = filp->f_flags & O_APPEND ? 1 : 0;
+
+       mlog_entry("(0x%p, %u, '%.*s')\n", filp,
+                  (unsigned int)nr_segs,
+                  filp->f_dentry->d_name.len,
+                  filp->f_dentry->d_name.name);
+
+       /* happy write of zero bytes */
+       if (iocb->ki_left == 0)
+               return 0;
+
+       mutex_lock(&inode->i_mutex);
+       /* to match setattr's i_mutex -> i_alloc_sem -> rw_lock ordering */
+       if (filp->f_flags & O_DIRECT) {
+               have_alloc_sem = 1;
+               down_read(&inode->i_alloc_sem);
+       }
+
+       /* concurrent O_DIRECT writes are allowed */
+       rw_level = (filp->f_flags & O_DIRECT) ? 0 : 1;
+       ret = ocfs2_rw_lock(inode, rw_level);
+       if (ret < 0) {
+               rw_level = -1;
+               mlog_errno(ret);
+               goto out;
+       }
+
+       ret = ocfs2_prepare_inode_for_write(filp->f_dentry, &iocb->ki_pos,
+                                           iocb->ki_left, appending);
+       if (ret < 0) {
+               mlog_errno(ret);
+               goto out;
+       }
 
        /* communicate with ocfs2_dio_end_io */
        ocfs2_iocb_set_rw_locked(iocb);
@@ -1100,8 +1116,6 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
        }
 
 out:
-       if (meta_level != -1)
-               ocfs2_meta_unlock(inode, meta_level);
        if (have_alloc_sem)
                up_read(&inode->i_alloc_sem);
        if (rw_level != -1) 
@@ -1112,6 +1126,77 @@ out:
        return ret;
 }
 
+static ssize_t ocfs2_file_splice_write(struct pipe_inode_info *pipe,
+                                      struct file *out,
+                                      loff_t *ppos,
+                                      size_t len,
+                                      unsigned int flags)
+{
+       int ret;
+       struct inode *inode = out->f_dentry->d_inode;
+
+       mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", out, pipe,
+                  (unsigned int)len,
+                  out->f_dentry->d_name.len,
+                  out->f_dentry->d_name.name);
+
+       inode_double_lock(inode, pipe->inode);
+
+       ret = ocfs2_rw_lock(inode, 1);
+       if (ret < 0) {
+               mlog_errno(ret);
+               goto out;
+       }
+
+       ret = ocfs2_prepare_inode_for_write(out->f_dentry, ppos, len, 0);
+       if (ret < 0) {
+               mlog_errno(ret);
+               goto out_unlock;
+       }
+
+       /* ok, we're done with i_size and alloc work */
+       ret = generic_file_splice_write_nolock(pipe, out, ppos, len, flags);
+
+out_unlock:
+       ocfs2_rw_unlock(inode, 1);
+out:
+       inode_double_unlock(inode, pipe->inode);
+
+       mlog_exit(ret);
+       return ret;
+}
+
+static ssize_t ocfs2_file_splice_read(struct file *in,
+                                     loff_t *ppos,
+                                     struct pipe_inode_info *pipe,
+                                     size_t len,
+                                     unsigned int flags)
+{
+       int ret = 0;
+       struct inode *inode = in->f_dentry->d_inode;
+
+       mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", in, pipe,
+                  (unsigned int)len,
+                  in->f_dentry->d_name.len,
+                  in->f_dentry->d_name.name);
+
+       /*
+        * See the comment in ocfs2_file_aio_read()
+        */
+       ret = ocfs2_meta_lock(inode, NULL, 0);
+       if (ret < 0) {
+               mlog_errno(ret);
+               goto bail;
+       }
+       ocfs2_meta_unlock(inode, 0);
+
+       ret = generic_file_splice_read(in, ppos, pipe, len, flags);
+
+bail:
+       mlog_exit(ret);
+       return ret;
+}
+
 static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
                                   const struct iovec *iov,
                                   unsigned long nr_segs,
@@ -1210,6 +1295,8 @@ const struct file_operations ocfs2_fops = {
        .aio_read       = ocfs2_file_aio_read,
        .aio_write      = ocfs2_file_aio_write,
        .ioctl          = ocfs2_ioctl,
+       .splice_read    = ocfs2_file_splice_read,
+       .splice_write   = ocfs2_file_splice_write,
 };
 
 const struct file_operations ocfs2_dops = {