Re: [PATCH 07/17] new helpers: skb_copy_datagram_from_iter() and zerocopy_sg_from_iter()

On 11/24/2014 08:02 AM, Ben Hutchings wrote:

On Sat, 2014-11-22 at 04:33 +0000, Al Viro wrote:
[...]

quoted

--- a/net/core/datagram.c
+++ b/net/core/datagram.c

@@ -572,6 +572,77 @@ fault:
 }
 EXPORT_SYMBOL(skb_copy_datagram_from_iovec);
 

Missing kernel-doc.

quoted

+int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
+				 struct iov_iter *from,
+				 int len)
+{
+	int start = skb_headlen(skb);
+	int i, copy = start - offset;
+	struct sk_buff *frag_iter;
+
+	/* Copy header. */
+	if (copy > 0) {
+		if (copy > len)
+			copy = len;
+		if (copy_from_iter(skb->data + offset, copy, from) != copy)
+			goto fault;
+		if ((len -= copy) == 0)
+			return 0;
+		offset += copy;
+	}
+
+	/* Copy paged appendix. Hmm... why does this look so complicated? */
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+		int end;
+		const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
+
+		WARN_ON(start > offset + len);
+
+		end = start + skb_frag_size(frag);
+		if ((copy = end - offset) > 0) {
+			size_t copied;

Blank line needed after a declaration.

quoted

+			if (copy > len)
+				copy = len;
+			copied = copy_page_from_iter(skb_frag_page(frag),
+					  frag->page_offset + offset - start,
+					  copy, from);
+			if (copied != copy)
+				goto fault;
+
+			if (!(len -= copy))
+				return 0;

The other two instances of this condition are written as:

			if ((len -= copy) == 0)

Similarly in skb_copy_bits().

quoted

+			offset += copy;
+		}
+		start = end;
+	}
+
+	skb_walk_frags(skb, frag_iter) {
+		int end;
+
+		WARN_ON(start > offset + len);
+
+		end = start + frag_iter->len;
+		if ((copy = end - offset) > 0) {
+			if (copy > len)
+				copy = len;
+			if (skb_copy_datagram_from_iter(frag_iter,
+							offset - start,
+							from, copy))
+				goto fault;
+			if ((len -= copy) == 0)
+				return 0;
+			offset += copy;
+		}
+		start = end;
+	}
+	if (!len)
+		return 0;
+
+fault:
+	return -EFAULT;
+}
+EXPORT_SYMBOL(skb_copy_datagram_from_iter);
+
 /**
  *	zerocopy_sg_from_iovec - Build a zerocopy datagram from an iovec
  *	@skb: buffer to copy

@@ -643,6 +714,50 @@ int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *from,
 }
 EXPORT_SYMBOL(zerocopy_sg_from_iovec);
 

Missing kernel-doc.

quoted

+int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *from)
+{
+	int len = iov_iter_count(from);
+	int copy = min_t(int, skb_headlen(skb), len);
+	int i = 0;
+
+	/* copy up to skb headlen */
+	if (skb_copy_datagram_from_iter(skb, 0, from, copy))
+		return -EFAULT;
+
+	while (iov_iter_count(from)) {
+		struct page *pages[MAX_SKB_FRAGS];
+		size_t start;
+		ssize_t copied;
+		unsigned long truesize;
+		int n = 0;
+
+		copied = iov_iter_get_pages(from, pages, ~0U, MAX_SKB_FRAGS, &start);
+		if (copied < 0)
+			return -EFAULT;
+
+		truesize = DIV_ROUND_UP(copied + start, PAGE_SIZE) * PAGE_SIZE;

PAGE_ALIGN(copied + start) ?

quoted

+		skb->data_len += copied;
+		skb->len += copied;
+		skb->truesize += truesize;
+		atomic_add(truesize, &skb->sk->sk_wmem_alloc);
+		while (copied) {
+			int off = start;

This variable seems redundant.  Can't we use start directly and move the
'start = 0' to the bottom of the loop?

quoted

+			int size = min_t(int, copied, PAGE_SIZE - off);
+			start = 0;
+			if (i < MAX_SKB_FRAGS)
+				skb_fill_page_desc(skb, i, pages[n], off, size);
+			else
+				put_page(pages[n]);

Why is this condition needed, given we told iov_iter_get_pages() to
limit to MAX_SKB_FRAGS pages?

We don't want to send truncated packets and there's no other way to put
those pages since it was not in the frag array.