mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Yeonju Bae <iwasbaeyz@gmail.com>
To: netdev@vger.kernel.org
Cc: gregkh@linuxfoundation.org, security@kernel.org,
	john.fastabend@gmail.com, kuba@kernel.org, sd@queasysnail.net,
	davem@davemloft.net, edumazet@google.com, pabeni@redhat.com,
	horms@kernel.org, linux-kernel@vger.kernel.org,
	Yeonju Bae <iwasbaeyz@gmail.com>
Subject: [PATCH net] tls: avoid zc receive for file-backed pages
Date: Fri, 22 May 2026 01:53:28 +0900	[thread overview]
Message-ID: <20260521165328.16112-1-iwasbaeyz@gmail.com> (raw)
In-Reply-To: <2026052150-stylus-germicide-780e@gregkh>

kTLS RX zc decrypt writes unauthenticated AEAD output directly into
pages pinned from the recvmsg iterator via tls_setup_from_iter().
For MAP_SHARED, PROT_WRITE file-backed destinations, those pages are
live page-cache pages rather than anonymous copies: MAP_SHARED does not
trigger copy-on-write, so FOLL_WRITE returns the actual page-cache page.

crypto_aead_decrypt() writes CTR-mode decryption output into the
scatter-gather list before the authentication tag is verified.  If the
tag check fails (-EBADMSG), the plaintext-like output is already
resident in the page-cache page.  exit_free_pages() calls put_page()
without any content cleanup, so the modification persists through the
backing file.  An independent open(O_RDONLY)/read() of the same file
returns different content and its SHA-256 changes.  MAP_PRIVATE is safe
via COW; PROT_READ-only destinations fail at iov_iter_get_pages2()
before any decryption occurs.

Avoid zc receive for file-backed destination pages.  In
tls_setup_from_iter(), after iov_iter_get_pages2() pins pages, check
each page with folio_mapping(page_folio(page)).  If any pinned page is
file-backed (mapping != NULL), release the pinned pages and return
-EOPNOTSUPP.  Handle -EOPNOTSUPP in tls_decrypt_sw() by clearing
darg->zc and retrying, which causes tls_decrypt_sg() to allocate a
kernel bounce buffer instead.  Decryption output never reaches the
file-backed page; on tag failure the bounce buffer is discarded.

This follows the existing opportunistic zc retry pattern already used
for TLS 1.3 record type mismatches in tls_decrypt_sw().

Verified on linux-7.0-rc3 QEMU (x86-64), four destination types:
  MAP_SHARED+RW:   file_changed=0  (was 4077/4096 bytes before patch)
  MAP_PRIVATE+RW:  file_changed=0  (COW isolation; unchanged)
  anonymous heap:  no file backing  (unchanged)
  PROT_READ only:  file_changed=0  (EFAULT before decrypt; unchanged)

Signed-off-by: Yeonju Bae <iwasbaeyz@gmail.com>
---
 net/tls/tls_sw.c | 37 +++++++++++++++++++++++++++++++++++++
 1 file changed, 37 insertions(+)

diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c
index a977b0434..c312a83b4 100644
--- a/net/tls/tls_sw.c
+++ b/net/tls/tls_sw.c
@@ -36,6 +36,7 @@
  */
 
 #include <linux/bug.h>
+#include <linux/pagemap.h>
 #include <linux/sched/signal.h>
 #include <linux/module.h>
 #include <linux/kernel.h>
@@ -1443,6 +1444,34 @@ static int tls_setup_from_iter(struct iov_iter *from,
 
 		length -= copied;
 		size += copied;
+		/* Reject file-backed destination pages.  Writing unauthenticated
+		 * AEAD output into a page-cache page before tag verification
+		 * leaves the backing file modified even when recvmsg() returns
+		 * -EBADMSG.  Return -EOPNOTSUPP so the caller retries via the
+		 * non-ZC bounce-buffer path.
+		 */
+		{
+			ssize_t remain = copied;
+			size_t  off    = offset;
+			int     np = 0, j;
+
+			while (remain > 0) {
+				remain -= min_t(ssize_t, remain,
+						(ssize_t)(PAGE_SIZE - off));
+				off = 0;
+				np++;
+			}
+			for (j = 0; j < np; j++) {
+				if (folio_mapping(page_folio(pages[j]))) {
+					int k;
+
+					for (k = 0; k < np; k++)
+						put_page(pages[k]);
+					rc = -EOPNOTSUPP;
+					goto out;
+				}
+			}
+		}
 		while (copied) {
 			use = min_t(int, copied, PAGE_SIZE - offset);
 
@@ -1699,6 +1728,14 @@ tls_decrypt_sw(struct sock *sk, struct tls_context *tls_ctx,
 	if (err < 0) {
 		if (err == -EBADMSG)
 			TLS_INC_STATS(sock_net(sk), LINUX_MIB_TLSDECRYPTERROR);
+		if (err == -EOPNOTSUPP && darg->zc) {
+			/* tls_setup_from_iter detected file-backed destination
+			 * pages; retry without ZC via the bounce-buffer path.
+			 */
+			darg->zc = false;
+			TLS_INC_STATS(sock_net(sk), LINUX_MIB_TLSDECRYPTRETRY);
+			return tls_decrypt_sw(sk, tls_ctx, msg, darg);
+		}
 		return err;
 	}
 	/* keep going even for ->async, the code below is TLS 1.3 */
-- 
2.43.0


       reply	other threads:[~2026-05-21 16:54 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <2026052150-stylus-germicide-780e@gregkh>
2026-05-21 16:53 ` Yeonju Bae [this message]
2026-05-25 17:54   ` Jakub Kicinski
2026-05-26  7:05     ` Christoph Hellwig

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260521165328.16112-1-iwasbaeyz@gmail.com \
    --to=iwasbaeyz@gmail.com \
    --cc=davem@davemloft.net \
    --cc=edumazet@google.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=horms@kernel.org \
    --cc=john.fastabend@gmail.com \
    --cc=kuba@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=sd@queasysnail.net \
    --cc=security@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®