-----Original Message----- From: Pedro Falcato pfalcato@suse.de Sent: Wednesday, 23 July 2025 12:41 To: Jason Gunthorpe jgg@ziepe.ca; Bernard Metzler BMT@zurich.ibm.com; Leon Romanovsky leon@kernel.org; Vlastimil Babka vbabka@suse.cz Cc: Jakub Kicinski kuba@kernel.org; David Howells dhowells@redhat.com; Tom Talpey tom@talpey.com; linux-rdma@vger.kernel.org; linux- kernel@vger.kernel.org; linux-mm@kvack.org; Pedro Falcato pfalcato@suse.de; stable@vger.kernel.org; kernel test robot oliver.sang@intel.com Subject: [EXTERNAL] [PATCH] RDMA/siw: Fix the sendmsg byte count in siw_tcp_sendpages
Ever since commit c2ff29e99a76 ("siw: Inline do_tcp_sendpages()"), we have been doing this:
static int siw_tcp_sendpages(struct socket *s, struct page **page, int offset, size_t size) [...] /* Calculate the number of bytes we need to push, for this page * specifically */ size_t bytes = min_t(size_t, PAGE_SIZE - offset, size); /* If we can't splice it, then copy it in, as normal */ if (!sendpage_ok(page[i])) msg.msg_flags &= ~MSG_SPLICE_PAGES; /* Set the bvec pointing to the page, with len $bytes */ bvec_set_page(&bvec, page[i], bytes, offset); /* Set the iter to $size, aka the size of the whole sendpages (!!!) */ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size); try_page_again: lock_sock(sk); /* Sendmsg with $size size (!!!) */ rv = tcp_sendmsg_locked(sk, &msg, size);
This means we've been sending oversized iov_iters and tcp_sendmsg calls for a while. This has a been a benign bug because sendpage_ok() always returned true. With the recent slab allocator changes being slowly introduced into next (that disallow sendpage on large kmalloc allocations), we have recently hit out-of-bounds crashes, due to slight differences in iov_iter behavior between the MSG_SPLICE_PAGES and "regular" copy paths:
(MSG_SPLICE_PAGES) skb_splice_from_iter iov_iter_extract_pages iov_iter_extract_bvec_pages uses i->nr_segs to correctly stop in its tracks before OoB'ing everywhere skb_splice_from_iter gets a "short" read
(!MSG_SPLICE_PAGES) skb_copy_to_page_nocache copy=iov_iter_count [...] copy_from_iter /* this doesn't help */ if (unlikely(iter->count < len)) len = iter->count; iterate_bvec ... and we run off the bvecs
Fix this by properly setting the iov_iter's byte count, plus sending the correct byte count to tcp_sendmsg_locked.
Cc: stable@vger.kernel.org Fixes: c2ff29e99a76 ("siw: Inline do_tcp_sendpages()") Reported-by: kernel test robot oliver.sang@intel.com Closes: https% 3A__lore.kernel.org_oe-2Dlkp_202507220801.50a7210-2Dlkp- 40intel.com&d=DwIDAg&c=BSDicqBQBDjDI9RkVyTcHQ&r=4ynb4Sj_4MUcZXbhvovE4tYSbqx yOwdSiLedP4yO55g&m=8FIDji_gvHEZEUL08IM8h6Slg9f_hp4n8OxkdR_OWLnZ9CkPknHCzHVC BYkCKt_q&s=IP0c71OgDL-cEe5hFduy0qNhy5WICEkTBTQLrVicotc&e= Signed-off-by: Pedro Falcato pfalcato@suse.de
drivers/infiniband/sw/siw/siw_qp_tx.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/infiniband/sw/siw/siw_qp_tx.c b/drivers/infiniband/sw/siw/siw_qp_tx.c index 3a08f57d2211..9576a2b766c4 100644 --- a/drivers/infiniband/sw/siw/siw_qp_tx.c +++ b/drivers/infiniband/sw/siw/siw_qp_tx.c @@ -340,11 +340,11 @@ static int siw_tcp_sendpages(struct socket *s, struct page **page, int offset, if (!sendpage_ok(page[i])) msg.msg_flags &= ~MSG_SPLICE_PAGES; bvec_set_page(&bvec, page[i], bytes, offset);
iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, bytes);try_page_again: lock_sock(sk);
rv = tcp_sendmsg_locked(sk, &msg, size);
rv = tcp_sendmsg_locked(sk, &msg, bytes);
Pedro, many thanks for catching this! I completely missed it during my too sloppy review of that patch. It's a serious bug which must be fixed asap. BUT, looking closer, I do not see the offset being taken into account when retrying a current segment. So, resend attempts seem to send old data which are already out. Shouldn't the try_page_again: label be above bvec_set_page()??
Thanks! Bernard.
if (rv > 0) {-- 2.50.1