Re: 4.9.0 regression in pipe-backed iov_iter with systemd-nspawn

[Al Viro]

On Fri, Jan 13, 2017 at 09:54:02AM +0000, Alan J. Wylie wrote:
> root      1669  0.0  0.0  76144  5412 ?        S    09:51   0:00          \_ /usr/sbin/postdrop -r
> 
> Another hang.
> 
> # dmesg  | tail
> [   22.352442] r8169 0000:03:00.0: loading /lib/firmware/4.9.3-dirty/rtl_nic/rtl8168e-3.fw failed with error -2
> [   22.408814] r8169 0000:03:00.0: direct-loading rtl_nic/rtl8168e-3.fw
> [   22.408821] fw_set_page_data: fw-rtl_nic/rtl8168e-3.fw buf=ffff92b7b1cb8c80 data=ffffad1641179000 size=3872
> [   22.536043] r8169 0000:03:00.0 enp3s0: link down
> [   22.536079] r8169 0000:03:00.0 enp3s0: link down
> [   24.873801] r8169 0000:03:00.0 enp3s0: link up
> [   24.874766] br0: port 1(enp3s0) entered blocking state
> [   24.876622] br0: port 1(enp3s0) entered forwarding state
> [   24.878560] IPv6: ADDRCONF(NETDEV_CHANGE): br0: link becomes ready
> [  219.683974] nr: 0->16, cur: 5->5, buffers: 16->16

OK, so it is iov_iter_advance() failing to free the shit allocated, either
due to some breakage in pipe_advance() or buggered 'copied'...  Let's
see which one; could you apply the following and run your reproducer?  The
only difference from the previous is that it collects and prints a bit more,
so it should be just as reproducible...

diff --git a/fs/splice.c b/fs/splice.c
index 873d83104e79..11477609e7f7 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -393,6 +393,10 @@ static ssize_t default_file_splice_read(struct file *in, loff_t *ppos,
 	size_t offset, dummy, copied = 0;
 	ssize_t res;
 	int i;
+	unsigned nrbufs = pipe->nrbufs,
+		 curbuf = pipe->curbuf,
+		 buffers = pipe->buffers;
+	int idx, count, offs;
 
 	if (pipe->nrbufs == pipe->buffers)
 		return -EAGAIN;
@@ -444,7 +448,22 @@ static ssize_t default_file_splice_read(struct file *in, loff_t *ppos,
 	for (i = 0; i < nr_pages; i++)
 		put_page(pages[i]);
 	kvfree(pages);
+	count = to.count;
+	idx = to.idx;
+	offs = to.iov_offset;
 	iov_iter_advance(&to, copied);	/* truncates and discards */
+	if (res == -EAGAIN && (
+		pipe->nrbufs != nrbufs ||
+		pipe->curbuf != curbuf ||
+		pipe->buffers != buffers)
+	    ) {
+		printk(KERN_ERR "nr: %d->%d, cur: %d->%d, buffers: %d->%d\n",
+			nrbufs, pipe->nrbufs,
+			curbuf, pipe->curbuf,
+			buffers, pipe->buffers);
+		printk(KERN_ERR "copied: %zd, count:%d, idx:%d, offs:%d\n",
+			copied, count, idx, offs);
+	}
 	return res;
 }