Merge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[pandora-kernel.git] / fs / ocfs2 / file.c
index b75b2e1..57e0d30 100644 (file)
@@ -51,6 +51,7 @@
 #include "inode.h"
 #include "ioctl.h"
 #include "journal.h"
+#include "locks.h"
 #include "mmap.h"
 #include "suballoc.h"
 #include "super.h"
@@ -63,6 +64,35 @@ static int ocfs2_sync_inode(struct inode *inode)
        return sync_mapping_buffers(inode->i_mapping);
 }
 
+static int ocfs2_init_file_private(struct inode *inode, struct file *file)
+{
+       struct ocfs2_file_private *fp;
+
+       fp = kzalloc(sizeof(struct ocfs2_file_private), GFP_KERNEL);
+       if (!fp)
+               return -ENOMEM;
+
+       fp->fp_file = file;
+       mutex_init(&fp->fp_mutex);
+       ocfs2_file_lock_res_init(&fp->fp_flock, fp);
+       file->private_data = fp;
+
+       return 0;
+}
+
+static void ocfs2_free_file_private(struct inode *inode, struct file *file)
+{
+       struct ocfs2_file_private *fp = file->private_data;
+       struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+       if (fp) {
+               ocfs2_simple_drop_lockres(osb, &fp->fp_flock);
+               ocfs2_lock_res_free(&fp->fp_flock);
+               kfree(fp);
+               file->private_data = NULL;
+       }
+}
+
 static int ocfs2_file_open(struct inode *inode, struct file *file)
 {
        int status;
@@ -89,7 +119,18 @@ static int ocfs2_file_open(struct inode *inode, struct file *file)
 
        oi->ip_open_count++;
        spin_unlock(&oi->ip_lock);
-       status = 0;
+
+       status = ocfs2_init_file_private(inode, file);
+       if (status) {
+               /*
+                * We want to set open count back if we're failing the
+                * open.
+                */
+               spin_lock(&oi->ip_lock);
+               oi->ip_open_count--;
+               spin_unlock(&oi->ip_lock);
+       }
+
 leave:
        mlog_exit(status);
        return status;
@@ -108,11 +149,24 @@ static int ocfs2_file_release(struct inode *inode, struct file *file)
                oi->ip_flags &= ~OCFS2_INODE_OPEN_DIRECT;
        spin_unlock(&oi->ip_lock);
 
+       ocfs2_free_file_private(inode, file);
+
        mlog_exit(0);
 
        return 0;
 }
 
+static int ocfs2_dir_open(struct inode *inode, struct file *file)
+{
+       return ocfs2_init_file_private(inode, file);
+}
+
+static int ocfs2_dir_release(struct inode *inode, struct file *file)
+{
+       ocfs2_free_file_private(inode, file);
+       return 0;
+}
+
 static int ocfs2_sync_file(struct file *file,
                           struct dentry *dentry,
                           int datasync)
@@ -382,18 +436,13 @@ static int ocfs2_truncate_file(struct inode *inode,
 
        down_write(&OCFS2_I(inode)->ip_alloc_sem);
 
-       /* This forces other nodes to sync and drop their pages. Do
-        * this even if we have a truncate without allocation change -
-        * ocfs2 cluster sizes can be much greater than page size, so
-        * we have to truncate them anyway.  */
-       status = ocfs2_data_lock(inode, 1);
-       if (status < 0) {
-               up_write(&OCFS2_I(inode)->ip_alloc_sem);
-
-               mlog_errno(status);
-               goto bail;
-       }
-
+       /*
+        * The inode lock forced other nodes to sync and drop their
+        * pages, which (correctly) happens even if we have a truncate
+        * without allocation change - ocfs2 cluster sizes can be much
+        * greater than page size, so we have to truncate them
+        * anyway.
+        */
        unmap_mapping_range(inode->i_mapping, new_i_size + PAGE_SIZE - 1, 0, 1);
        truncate_inode_pages(inode->i_mapping, new_i_size);
 
@@ -403,7 +452,7 @@ static int ocfs2_truncate_file(struct inode *inode,
                if (status)
                        mlog_errno(status);
 
-               goto bail_unlock_data;
+               goto bail_unlock_sem;
        }
 
        /* alright, we're going to need to do a full blown alloc size
@@ -413,25 +462,23 @@ static int ocfs2_truncate_file(struct inode *inode,
        status = ocfs2_orphan_for_truncate(osb, inode, di_bh, new_i_size);
        if (status < 0) {
                mlog_errno(status);
-               goto bail_unlock_data;
+               goto bail_unlock_sem;
        }
 
        status = ocfs2_prepare_truncate(osb, inode, di_bh, &tc);
        if (status < 0) {
                mlog_errno(status);
-               goto bail_unlock_data;
+               goto bail_unlock_sem;
        }
 
        status = ocfs2_commit_truncate(osb, inode, di_bh, tc);
        if (status < 0) {
                mlog_errno(status);
-               goto bail_unlock_data;
+               goto bail_unlock_sem;
        }
 
        /* TODO: orphan dir cleanup here. */
-bail_unlock_data:
-       ocfs2_data_unlock(inode, 1);
-
+bail_unlock_sem:
        up_write(&OCFS2_I(inode)->ip_alloc_sem);
 
 bail:
@@ -579,7 +626,7 @@ int ocfs2_lock_allocators(struct inode *inode, struct ocfs2_dinode *di,
 
        mlog(0, "extend inode %llu, i_size = %lld, di->i_clusters = %u, "
             "clusters_to_add = %u, extents_to_split = %u\n",
-            (unsigned long long)OCFS2_I(inode)->ip_blkno, i_size_read(inode),
+            (unsigned long long)OCFS2_I(inode)->ip_blkno, (long long)i_size_read(inode),
             le32_to_cpu(di->i_clusters), clusters_to_add, extents_to_split);
 
        num_free_extents = ocfs2_num_free_extents(osb, inode, di);
@@ -760,7 +807,7 @@ restarted_transaction:
             le32_to_cpu(fe->i_clusters),
             (unsigned long long)le64_to_cpu(fe->i_size));
        mlog(0, "inode: ip_clusters=%u, i_size=%lld\n",
-            OCFS2_I(inode)->ip_clusters, i_size_read(inode));
+            OCFS2_I(inode)->ip_clusters, (long long)i_size_read(inode));
 
 leave:
        if (handle) {
@@ -917,7 +964,7 @@ static int ocfs2_extend_file(struct inode *inode,
                             struct buffer_head *di_bh,
                             u64 new_i_size)
 {
-       int ret = 0, data_locked = 0;
+       int ret = 0;
        struct ocfs2_inode_info *oi = OCFS2_I(inode);
 
        BUG_ON(!di_bh);
@@ -943,20 +990,6 @@ static int ocfs2_extend_file(struct inode *inode,
            && ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
                goto out_update_size;
 
-       /* 
-        * protect the pages that ocfs2_zero_extend is going to be
-        * pulling into the page cache.. we do this before the
-        * metadata extend so that we don't get into the situation
-        * where we've extended the metadata but can't get the data
-        * lock to zero.
-        */
-       ret = ocfs2_data_lock(inode, 1);
-       if (ret < 0) {
-               mlog_errno(ret);
-               goto out;
-       }
-       data_locked = 1;
-
        /*
         * The alloc sem blocks people in read/write from reading our
         * allocation until we're done changing it. We depend on
@@ -980,7 +1013,7 @@ static int ocfs2_extend_file(struct inode *inode,
                        up_write(&oi->ip_alloc_sem);
 
                        mlog_errno(ret);
-                       goto out_unlock;
+                       goto out;
                }
        }
 
@@ -991,7 +1024,7 @@ static int ocfs2_extend_file(struct inode *inode,
 
        if (ret < 0) {
                mlog_errno(ret);
-               goto out_unlock;
+               goto out;
        }
 
 out_update_size:
@@ -999,10 +1032,6 @@ out_update_size:
        if (ret < 0)
                mlog_errno(ret);
 
-out_unlock:
-       if (data_locked)
-               ocfs2_data_unlock(inode, 1);
-
 out:
        return ret;
 }
@@ -1019,6 +1048,10 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
        mlog_entry("(0x%p, '%.*s')\n", dentry,
                   dentry->d_name.len, dentry->d_name.name);
 
+       /* ensuring we don't even attempt to truncate a symlink */
+       if (S_ISLNK(inode->i_mode))
+               attr->ia_valid &= ~ATTR_SIZE;
+
        if (attr->ia_valid & ATTR_MODE)
                mlog(0, "mode change: %d\n", attr->ia_mode);
        if (attr->ia_valid & ATTR_UID)
@@ -1050,7 +1083,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
                }
        }
 
-       status = ocfs2_meta_lock(inode, &bh, 1);
+       status = ocfs2_inode_lock(inode, &bh, 1);
        if (status < 0) {
                if (status != -ENOENT)
                        mlog_errno(status);
@@ -1102,7 +1135,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
 bail_commit:
        ocfs2_commit_trans(osb, handle);
 bail_unlock:
-       ocfs2_meta_unlock(inode, 1);
+       ocfs2_inode_unlock(inode, 1);
 bail_unlock_rw:
        if (size_change)
                ocfs2_rw_unlock(inode, 1);
@@ -1149,7 +1182,7 @@ int ocfs2_permission(struct inode *inode, int mask, struct nameidata *nd)
 
        mlog_entry_void();
 
-       ret = ocfs2_meta_lock(inode, NULL, 0);
+       ret = ocfs2_inode_lock(inode, NULL, 0);
        if (ret) {
                if (ret != -ENOENT)
                        mlog_errno(ret);
@@ -1158,7 +1191,7 @@ int ocfs2_permission(struct inode *inode, int mask, struct nameidata *nd)
 
        ret = generic_permission(inode, mask, NULL);
 
-       ocfs2_meta_unlock(inode, 0);
+       ocfs2_inode_unlock(inode, 0);
 out:
        mlog_exit(ret);
        return ret;
@@ -1630,7 +1663,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
                goto out;
        }
 
-       ret = ocfs2_meta_lock(inode, &di_bh, 1);
+       ret = ocfs2_inode_lock(inode, &di_bh, 1);
        if (ret) {
                mlog_errno(ret);
                goto out_rw_unlock;
@@ -1638,7 +1671,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
 
        if (inode->i_flags & (S_IMMUTABLE|S_APPEND)) {
                ret = -EPERM;
-               goto out_meta_unlock;
+               goto out_inode_unlock;
        }
 
        switch (sr->l_whence) {
@@ -1652,7 +1685,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
                break;
        default:
                ret = -EINVAL;
-               goto out_meta_unlock;
+               goto out_inode_unlock;
        }
        sr->l_whence = 0;
 
@@ -1663,14 +1696,14 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
            || (sr->l_start + llen) < 0
            || (sr->l_start + llen) > max_off) {
                ret = -EINVAL;
-               goto out_meta_unlock;
+               goto out_inode_unlock;
        }
        size = sr->l_start + sr->l_len;
 
        if (cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) {
                if (sr->l_len <= 0) {
                        ret = -EINVAL;
-                       goto out_meta_unlock;
+                       goto out_inode_unlock;
                }
        }
 
@@ -1678,7 +1711,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
                ret = __ocfs2_write_remove_suid(inode, di_bh);
                if (ret) {
                        mlog_errno(ret);
-                       goto out_meta_unlock;
+                       goto out_inode_unlock;
                }
        }
 
@@ -1704,7 +1737,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
        up_write(&OCFS2_I(inode)->ip_alloc_sem);
        if (ret) {
                mlog_errno(ret);
-               goto out_meta_unlock;
+               goto out_inode_unlock;
        }
 
        /*
@@ -1714,7 +1747,7 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
        if (IS_ERR(handle)) {
                ret = PTR_ERR(handle);
                mlog_errno(ret);
-               goto out_meta_unlock;
+               goto out_inode_unlock;
        }
 
        if (change_size && i_size_read(inode) < size)
@@ -1727,9 +1760,9 @@ static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
 
        ocfs2_commit_trans(osb, handle);
 
-out_meta_unlock:
+out_inode_unlock:
        brelse(di_bh);
-       ocfs2_meta_unlock(inode, 1);
+       ocfs2_inode_unlock(inode, 1);
 out_rw_unlock:
        ocfs2_rw_unlock(inode, 1);
 
@@ -1799,7 +1832,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
         * if we need to make modifications here.
         */
        for(;;) {
-               ret = ocfs2_meta_lock(inode, NULL, meta_level);
+               ret = ocfs2_inode_lock(inode, NULL, meta_level);
                if (ret < 0) {
                        meta_level = -1;
                        mlog_errno(ret);
@@ -1817,7 +1850,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
                 * set inode->i_size at the end of a write. */
                if (should_remove_suid(dentry)) {
                        if (meta_level == 0) {
-                               ocfs2_meta_unlock(inode, meta_level);
+                               ocfs2_inode_unlock(inode, meta_level);
                                meta_level = 1;
                                continue;
                        }
@@ -1886,7 +1919,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
                *ppos = saved_pos;
 
 out_unlock:
-       ocfs2_meta_unlock(inode, meta_level);
+       ocfs2_inode_unlock(inode, meta_level);
 
 out:
        return ret;
@@ -2099,12 +2132,12 @@ static ssize_t ocfs2_file_splice_read(struct file *in,
        /*
         * See the comment in ocfs2_file_aio_read()
         */
-       ret = ocfs2_meta_lock(inode, NULL, 0);
+       ret = ocfs2_inode_lock(inode, NULL, 0);
        if (ret < 0) {
                mlog_errno(ret);
                goto bail;
        }
-       ocfs2_meta_unlock(inode, 0);
+       ocfs2_inode_unlock(inode, 0);
 
        ret = generic_file_splice_read(in, ppos, pipe, len, flags);
 
@@ -2160,12 +2193,12 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
         * like i_size. This allows the checks down below
         * generic_file_aio_read() a chance of actually working. 
         */
-       ret = ocfs2_meta_lock_atime(inode, filp->f_vfsmnt, &lock_level);
+       ret = ocfs2_inode_lock_atime(inode, filp->f_vfsmnt, &lock_level);
        if (ret < 0) {
                mlog_errno(ret);
                goto bail;
        }
-       ocfs2_meta_unlock(inode, lock_level);
+       ocfs2_inode_unlock(inode, lock_level);
 
        ret = generic_file_aio_read(iocb, iov, nr_segs, iocb->ki_pos);
        if (ret == -EINVAL)
@@ -2204,6 +2237,7 @@ const struct inode_operations ocfs2_special_file_iops = {
 };
 
 const struct file_operations ocfs2_fops = {
+       .llseek         = generic_file_llseek,
        .read           = do_sync_read,
        .write          = do_sync_write,
        .mmap           = ocfs2_mmap,
@@ -2212,20 +2246,25 @@ const struct file_operations ocfs2_fops = {
        .open           = ocfs2_file_open,
        .aio_read       = ocfs2_file_aio_read,
        .aio_write      = ocfs2_file_aio_write,
-       .ioctl          = ocfs2_ioctl,
+       .unlocked_ioctl = ocfs2_ioctl,
 #ifdef CONFIG_COMPAT
        .compat_ioctl   = ocfs2_compat_ioctl,
 #endif
+       .flock          = ocfs2_flock,
        .splice_read    = ocfs2_file_splice_read,
        .splice_write   = ocfs2_file_splice_write,
 };
 
 const struct file_operations ocfs2_dops = {
+       .llseek         = generic_file_llseek,
        .read           = generic_read_dir,
        .readdir        = ocfs2_readdir,
        .fsync          = ocfs2_sync_file,
-       .ioctl          = ocfs2_ioctl,
+       .release        = ocfs2_dir_release,
+       .open           = ocfs2_dir_open,
+       .unlocked_ioctl = ocfs2_ioctl,
 #ifdef CONFIG_COMPAT
        .compat_ioctl   = ocfs2_compat_ioctl,
 #endif
+       .flock          = ocfs2_flock,
 };