[PATCH] partial io completion problem

There's an issue with bio_uncopy_user() in that it expects ->bv_len to
be untouched from io submission to io completion, which it wont be for
some transfer lengths on some hardware. So we end up not copying all the
data back to user space, not so good. This patch adds a bio_map_data
local structure to track the io_vec.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

fs/bio.c

@@ -372,6 +372,38 @@ int bio_add_page(struct bio *bio, struct page *page, unsigned int len,
 			      len, offset);
 }
 
+struct bio_map_data {
+	struct bio_vec *iovecs;
+	void __user *userptr;
+};
+
+static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio)
+{
+	memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt);
+	bio->bi_private = bmd;
+}
+
+static void bio_free_map_data(struct bio_map_data *bmd)
+{
+	kfree(bmd->iovecs);
+	kfree(bmd);
+}
+
+static struct bio_map_data *bio_alloc_map_data(int nr_segs)
+{
+	struct bio_map_data *bmd = kmalloc(sizeof(*bmd), GFP_KERNEL);
+
+	if (!bmd)
+		return NULL;
+
+	bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, GFP_KERNEL);
+	if (bmd)
+		return bmd;
+
+	kfree(bmd);
+	return NULL;
+}
+
 /**
  *	bio_uncopy_user	-	finish previously mapped bio
  *	@bio: bio being terminated
@@ -381,20 +413,22 @@ int bio_add_page(struct bio *bio, struct page *page, unsigned int len,
  */
 int bio_uncopy_user(struct bio *bio)
 {
+	struct bio_map_data *bmd = bio->bi_private;
+	const int read = bio_data_dir(bio) == READ;
 	struct bio_vec *bvec;
 	int i, ret = 0;
 
-	char *uaddr = bio->bi_private;
-
 	__bio_for_each_segment(bvec, bio, i, 0) {
 		char *addr = page_address(bvec->bv_page);
-		if (bio_data_dir(bio) == READ && !ret &&
-		    copy_to_user(uaddr, addr, bvec->bv_len))
+		unsigned int len = bmd->iovecs[i].bv_len;
+
+		if (read && !ret && copy_to_user(bmd->userptr, addr, len))
 			ret = -EFAULT;
 
 		__free_page(bvec->bv_page);
-		uaddr += bvec->bv_len;
+		bmd->userptr += len;
 	}
+	bio_free_map_data(bmd);
 	bio_put(bio);
 	return ret;
 }
@@ -415,14 +449,25 @@ struct bio *bio_copy_user(request_queue_t *q, unsigned long uaddr,
 {
 	unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
 	unsigned long start = uaddr >> PAGE_SHIFT;
+	struct bio_map_data *bmd;
 	struct bio_vec *bvec;
 	struct page *page;
 	struct bio *bio;
 	int i, ret;
 
+	bmd = bio_alloc_map_data(end - start);
+	if (!bmd)
+		return ERR_PTR(-ENOMEM);
+
+	bmd->userptr = (void __user *) uaddr;
+
 	bio = bio_alloc(GFP_KERNEL, end - start);
-	if (!bio)
+	if (!bio) {
+		bio_free_map_data(bmd);
 		return ERR_PTR(-ENOMEM);
+	}
+
+	bio->bi_rw |= (!write_to_vm << BIO_RW);
 
 	ret = 0;
 	while (len) {
@@ -445,13 +490,15 @@ struct bio *bio_copy_user(request_queue_t *q, unsigned long uaddr,
 		len -= bytes;
 	}
 
+	if (ret)
+		goto cleanup;
+
 	/*
 	 * success
 	 */
-	if (!ret) {
 	if (!write_to_vm) {
 		unsigned long p = uaddr;
-			bio->bi_rw |= (1 << BIO_RW);
+
 		/*
 		 * for a write, copy in data to kernel pages
 		 */
@@ -465,13 +512,8 @@ struct bio *bio_copy_user(request_queue_t *q, unsigned long uaddr,
 		}
 	}
 
-		bio->bi_private = (void *) uaddr;
+	bio_set_map_data(bmd, bio);
 	return bio;
-	}
-
-	/*
-	 * cleanup
-	 */
 cleanup:
 	bio_for_each_segment(bvec, bio, i)
 		__free_page(bvec->bv_page);