On Fri, Nov 30, 2012 at 11:36:01AM -0500, Christoph Hellwig wrote:
> On Fri, Nov 30, 2012 at 01:49:10PM +1100, Dave Chinner wrote:
> > > Ugh. That's a big violation of how buffer-heads are supposed to work:
> > > the block number is very much defined to be in multiples of b_size
> > > (see for example "submit_bh()" that turns it into a sector number).
> > >
> > > But you're right. The direct-IO code really *is* violating that, and
> > > knows that get_block() ends up being defined in i_blkbits regardless
> > > of b_size.
> >
> > Same with mpage_readpages(), so it's not just direct IO that has
> > this problem....
>
> The mpage code may actually fall back to BHs.
>
> I have a version of the direct I/O code that uses the iomap_ops from the
> multi-page write code that you originally started. It uses the new op
> as primary interface for direct I/O and provides a helper for
> filesystems that still use buffer heads internally. I'll try to dust it
> off and send out a version for the current kernel.
So it was based on this interface?
(I went looking for this code on google a couple of days ago so I
could point at it and say "we should be using an iomap structure,
not buffer heads", but it looks like I never posted it to fsdevel or
the xfs lists...)
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 090f0ea..e247d62 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -522,6 +522,7 @@ enum positive_aop_returns {
struct page;
struct address_space;
struct writeback_control;
+struct iomap;
struct iov_iter {
const struct iovec *iov;
@@ -614,6 +615,9 @@ struct address_space_operations {
int (*is_partially_uptodate) (struct page *, read_descriptor_t *,
unsigned long);
int (*error_remove_page)(struct address_space *, struct page *);
+
+ int (*iomap)(struct address_space *mapping, loff_t pos,
+ ssize_t length, struct iomap *iomap, int cmd);
};
/*
diff --git a/include/linux/iomap.h b/include/linux/iomap.h
new file mode 100644
index 0000000..7708614
--- /dev/null
+++ b/include/linux/iomap.h
@@ -0,0 +1,45 @@
+#ifndef _IOMAP_H
+#define _IOMAP_H
+
+/* ->iomap a_op command types */
+#define IOMAP_READ 0x01 /* read the current mapping starting at the
+ given position, trimmed to a maximum length.
+ FS's should use this to obtain and lock
+ resources within this range */
+#define IOMAP_RESERVE 0x02 /* reserve space for an allocation that spans
+ the given iomap */
+#define IOMAP_ALLOCATE 0x03 /* allocate space in a given iomap - must have
+ first been reserved */
+#define IOMAP_UNRESERVE 0x04 /* return unused reserved space for the given
+ iomap and used space. This will always be
+ called after a IOMAP_READ so as to allow the
+ FS to release held resources. */
+
+/* types of block ranges for multipage write mappings. */
+#define IOMAP_HOLE 0x01 /* no blocks allocated, need allocation */
+#define IOMAP_DELALLOC 0x02 /* delayed allocation blocks */
+#define IOMAP_MAPPED 0x03 /* blocks allocated @blkno */
+#define IOMAP_UNWRITTEN 0x04 /* blocks allocated @blkno in unwritten state */
+
+struct iomap {
+ sector_t blkno; /* first sector of mapping */
+ loff_t offset; /* file offset of mapping, bytes */
+ ssize_t length; /* length of mapping, bytes */
+ int type; /* type of mapping */
+ void *priv; /* fs private data associated with map */
+};
+
+static inline bool
+iomap_needs_allocation(struct iomap *iomap)
+{
+ return iomap->type == IOMAP_HOLE;
+}
+
+/* multipage write interfaces use iomaps */
+typedef int (*mpw_actor_t)(struct address_space *mapping, void *src,
+ loff_t pos, ssize_t len, struct iomap *iomap);
+
+ssize_t multipage_write_segment(struct address_space *mapping, void *src,
+ loff_t pos, ssize_t length, mpw_actor_t actor);
+
+#endif /* _IOMAP_H */
Cheers,
Dave.
>
>
> --
> This message has been scanned for viruses and
> dangerous content by MailScanner, and is
> believed to be clean.
>
>
--
Dave Chinner
david@xxxxxxxxxxxxx
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
