Skip to contentSkip to conversation

fat: Editions to support fat_fallocate()

PATCH Grant Likely 1 participant 1 message Dec 14, 2007 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): 20071214193017.7545.88989.stgit@jpe-laptop

#1 lore raw

Thanks Steve; I've cc'd LKML and Hirofumi in my reply.

Cheers,
g.

fs/fat/cache.c

Quoted 18 lines · Steven Cavanagh

From: Steven Cavanagh <[email protected]>

Added support for fallocate for a msdos fat driver. This allows preallocation of clusters to an inode before writes to reduce file fragmentation

---

 fs/fat/cache.c |    9 +++++++++
 fs/fat/file.c  |   47 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/fat/inode.c |   15 +++++++++++++++
 3 files changed, 71 insertions(+), 0 deletions(-)
fs/fat/cache.c +0 -0
index 639b3b4..1a69ce4 100644

Drop your changes to this file; they are just pr_debug statements that don't need to be in mainline.

<snip>

fs/fat/file.c line 10

Quoted 10 lines
fs/fat/file.c +2 -0
index 69a83b5..de3f9ee 100644
@@ -6,6 +6,8 @@
  *  regular file handling primitives for fat-based filesystems
  */
 
+#undef DEBUG
+

Drop these 2 lines

Quoted 27 lines
fs/fat/file.c +13 -0
 #include <linux/capability.h>
 #include <linux/module.h>
 #include <linux/time.h>
@@ -15,6 +17,7 @@ #include <linux/buffer_head.h>
 #include <linux/writeback.h>
 #include <linux/backing-dev.h>
 #include <linux/blkdev.h>
+#include <linux/falloc.h>
 
 int fat_generic_ioctl(struct inode *inode, struct file *filp,
                      unsigned int cmd, unsigned long arg)
@@ -312,8 +315,52 @@ int fat_getattr(struct vfsmount *mnt, st
 }
 EXPORT_SYMBOL_GPL(fat_getattr);
 
+/*
+ * preallocate space for a file. This implements fat fallocate inode
+ * operation, which gets called from sys_fallocate system call. User
+ * space requests len bytes at offset.
+ */
+long fat_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len)
+{
+       int ret = 0;
+       loff_t filesize = inode->i_size;
+
+       /* preallocation to directories is currently not supported */
+       if (S_ISDIR(inode->i_mode)) {

fs/fat/file.c int fat_getattr(struct vfsmount *mnt, st · line 339

Quoted 10 lines
fs/fat/file.c +10 -0
+               printk(KERN_ERR
+               "fat_fallocate(): Directory prealloc not supported\n");
+               return -ENODEV;
+       }
+
+       if ((offset + len) <= filesize) {
+               printk(KERN_ERR
+                       "fat_fallocate():Blocks already allocated\n");
+                       return -EIO;
+       }

Drop the printk... in fact, dorp the error code too and just return 0.
It's not an IO error if the space has already been allocated.

In fact, this test is probably irrelevant. Since we're allocating clusters and not necessarily increasing the file size, you should instead test to see if the requested region already has clusters allocated.

fs/fat/file.c int fat_getattr(struct vfsmount *mnt, st · line 344

Quoted 5 lines
fs/fat/file.c +5 -0
+       if (offset > filesize) {
+                       printk(KERN_ERR
+                       "fat_fallocate():Offset error\n");
+                       return -EIO;
+       }

I though we agreed that we *want* to support this case. ie. if the caller specifies an offset beyond the length of the file, then allocate clusters to cover both the requested region and the 'gap'.

fs/fat/file.c int fat_getattr(struct vfsmount *mnt, st · line 346

Quoted 2 lines
fs/fat/file.c +2 -0
+
+       if ((offset + len) > filesize) {

Again, you don't want to test against the filesize; you want to test against the number of allocated sectors.

fs/fat/file.c int fat_getattr(struct vfsmount *mnt, st · line 350

Quoted 4 lines
fs/fat/file.c +4 -0
+               pr_debug("fat_fallocate():fat_cont_expand(): size: %llu\n",
+                               (offset+len));
+               ret = fat_cont_expand(inode, (offset + len));
+       }

What if fat_cont_expand fails? You'll end up increasing the filesize regardless.

fs/fat/file.c int fat_getattr(struct vfsmount *mnt, st · line 352

Quoted 2 lines
fs/fat/file.c +2 -0
+       if (mode & FALLOC_FL_KEEP_SIZE) {
+               mutex_lock(&inode->i_mutex);

The lock/unlock needs to also protect fat_cont_expand.

fs/fat/inode.c

Quoted 19 lines
fs/fat/file.c +9 -0
+               i_size_write(inode, filesize);
+               mutex_unlock(&inode->i_mutex);
+               pr_debug(
+               "fat_fallocate():FALLOC_FL_KEEP_SIZE: %llu\n", filesize);
+       }
+       return ret;
+}
+
 const struct inode_operations fat_file_inode_operations = {
        .truncate       = fat_truncate,
        .setattr        = fat_notify_change,
        .getattr        = fat_getattr,
+       .fallocate      = fat_fallocate,
 };
fs/fat/inode.c +0 -0
index 920a576..ad6f069 100644

Same for this file; this is just the addition of pr_debugs; drop from this patch

<snip>

Signature · 4 lines
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.
[email protected]
(403) 399-0195