On Wed, Jan 22, 2014 at 03:51:56PM -0700, Ross Zwisler wrote:
> > + if (hole) {
> > addr = NULL;
> > - hole = true;
> > size = bh->b_size;
> > + } else {
> > + unsigned first;
> > + retval = xip_get_addr(inode, bh, &addr);
> > + if (retval < 0)
> > + break;
> > + size = retval;
> > + first = offset - (block << inode->i_blkbits);
> > + if (buffer_unwritten(bh))
> > + memset(addr, 0, first);
> > + addr += first;
>
> + size -= first;
>
> This is needed so that we don't overrun the XIP buffer we are given in the
> event that our user buffer >= our XIP buffer and the start of our I/O isn't
> block aligned.
You're right! Thank you! However, we also need it for the hole ==
true case, don't we? So maybe something like this, incrementally on top of
patch 22/22:
P.S. Can someone come up with a better name for this variable than 'first'?
I'd usually use 'offset', but that's already taken. 'annoying_bit' seems a
bit judgemental. 'misaligned', maybe? 'skip' or 'seek' like dd uses?
diff --git a/fs/xip.c b/fs/xip.c
index 92157ff..1ae00db 100644
--- a/fs/xip.c
+++ b/fs/xip.c
@@ -103,6 +103,7 @@ static ssize_t xip_io(int rw, struct inode *inode, const struct iovec *iov,
if (max == offset) {
sector_t block = offset >> inode->i_blkbits;
+ unsigned first = offset - (block << inode->i_blkbits);
long size;
memset(bh, 0, sizeof(*bh));
bh->b_size = ALIGN(end - offset, PAGE_SIZE);
@@ -121,14 +122,12 @@ static ssize_t xip_io(int rw, struct inode *inode, const struct iovec *iov,
if (hole) {
addr = NULL;
- size = bh->b_size;
+ size = bh->b_size - first;
} else {
- unsigned first;
retval = xip_get_addr(inode, bh, &addr);
if (retval < 0)
break;
- size = retval;
- first = offset - (block << inode->i_blkbits);
+ size = retval - first;
if (buffer_unwritten(bh))
memset(addr, 0, first);
addr += first;
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
