[for-6.13,01/19] nfs/localio: must clear res.replen in nfs_local_read_done

Commit Message

Mike Snitzer Nov. 8, 2024, 11:39 p.m. UTC

From: NeilBrown <[e-mail]>

Otherwise memory corruption can occur due to NFSv3 LOCALIO reads
leaving garbage in res.replen:
- nfs3_read_done() copies that into server->read_hdrsize; from there
  nfs3_proc_read_setup() copies it to args.replen in new requests.
- nfs3_xdr_enc_read3args() passes that to rpc_prepare_reply_pages()
  which includes it in hdrsize for xdr_init_pages, so that rq_rcv_buf
  contains a ridiculous len.
- This is copied to rq_private_buf and xs_read_stream_request()
  eventually passes the kvec to sock_recvmsg() which receives incoming
  data into entirely the wrong place.

This is easily reproduced with NFSv3 LOCALIO that is servicing reads
when it is made to pivot back to using normal RPC.  This switch back
to using normal NFSv3 with RPC can occur for a few reasons but this
issue was exposed with a test that stops and then restarts the NFSv3
server while LOCALIO is performing heavy read IO.

Fixes: 70ba381e1a43 ("nfs: add LOCALIO support")
Reported-by: Mike Snitzer <[e-mail]>
Signed-off-by: NeilBrown <[e-mail]>
Co-developed-by: Mike Snitzer <[e-mail]>
Signed-off-by: Mike Snitzer <[e-mail]>
---
 fs/nfs/localio.c | 6 ++++++
 1 file changed, 6 insertions(+)

Comments

NeilBrown Nov. 11, 2024, 12:36 a.m. UTC | #1

On Sat, 09 Nov 2024, Mike Snitzer wrote:
> From: NeilBrown <[e-mail]>
> 
> Otherwise memory corruption can occur due to NFSv3 LOCALIO reads
> leaving garbage in res.replen:

I'm not comfortable with this patch.  It doesn't tell us *why* there is
garbage in res.replen.
This is part of nfs_pgio_header and whenever that is allocated it
initialised to all zeros.  So where does the garbage come from?

Answer: it comes from
	hdr->res.verf    = &hdr->verf;
in nfs_pgio_rpcsetup().


struct nfs_pgio_res contains a union.  'replen' is present for read.
'verf' is present for write (and there is other stuff).

so I think that init of res.verf should only happen for write.

I cannot see an easy way to do that.  The best I can come up with is
to add a new pg_ioflags flag which says "this is a write", and only
initialise res.verf if that is set.

If we do stick with the current patch, I'd like a comment where we set
res.replen saying that it was corrupted when res.verf was initialised in
nfs_gpio_rpcsetup().

Or maybe move res.replen out of the union.  There is a 4byte hole before
the union (on x86_64).  It would be cleaner to move verf out, but that
is bigger....

Thanks,
NeilBrown

Patch

diff --git a/fs/nfs/localio.c b/fs/nfs/localio.c
index 8f0ce82a677e..637528e6368e 100644
--- a/fs/nfs/localio.c
+++ b/fs/nfs/localio.c
@@ -354,6 +354,12 @@  nfs_local_read_done(struct nfs_local_kiocb *iocb, long status)
 
 	nfs_local_pgio_done(hdr, status);
 
+	/*
+	 * Must clear replen otherwise NFSv3 data corruption will occur
+	 * if/when switching from LOCALIO back to using normal RPC.
+	 */
+	hdr->res.replen = 0;
+
 	if (hdr->res.count != hdr->args.count ||
 	    hdr->args.offset + hdr->res.count >= i_size_read(file_inode(filp)))
 		hdr->res.eof = true;