mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Dmitry Safonov via B4 Relay <devnull+0x7f454c46.gmail.com@kernel.org>
To: "David S. Miller" <davem@davemloft.net>,
	 Eric Dumazet <edumazet@google.com>,
	Jakub Kicinski <kuba@kernel.org>,
	 Paolo Abeni <pabeni@redhat.com>, Simon Horman <horms@kernel.org>,
	 David Ahern <dsahern@kernel.org>,
	Ivan Delalande <colona@arista.com>,
	 Matthieu Baerts <matttbe@kernel.org>,
	Mat Martineau <martineau@kernel.org>,
	 Geliang Tang <geliang@kernel.org>,
	Boris Pismenny <borisp@nvidia.com>,
	 John Fastabend <john.fastabend@gmail.com>
Cc: netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
	 mptcp@lists.linux.dev, Dmitry Safonov <0x7f454c46@gmail.com>
Subject: [PATCH net 5/6] net/diag: Limit TCP-MD5-diag array by max attribute length
Date: Wed, 06 Nov 2024 18:10:18 +0000	[thread overview]
Message-ID: <20241106-tcp-md5-diag-prep-v1-5-d62debf3dded@gmail.com> (raw)
In-Reply-To: <20241106-tcp-md5-diag-prep-v1-0-d62debf3dded@gmail.com>

From: Dmitry Safonov <0x7f454c46@gmail.com>

Currently TCP-MD5 keys are dumped as an array of
(struct tcp_diag_md5sig). All the keys from a socket go
into the same netlink attribute. The maximum amount of TCP-MD5 keys on
any socket is limited by /proc/sys/net/core/optmem_max, which post
commit 4944566706b2 ("net: increase optmem_max default value") is now by
default 128 KB. With the help of selftest I've figured out that equals
to 963 keys, without user having to increase optmem_max:
> test_set_md5() [963/1024]: Cannot allocate memory

The maximum length of nlattr is limited by typeof(nlattr::nla_len),
which is (U16_MAX - 1). When there are too many keys the array written
overflows the netlink attribute. Here is what one can see on a test,
with no adjustments to optmem_max defaults:

> recv() = 65180
> socket: 10.0.254.1:7013->0.0.0.0:0 (intf 3)
>      family: 2 state: 10 timer: 0 retrans: 0
>      expires: 0 rqueu: 0 wqueue: 1 uid: 0 inode: 456
>              attr type: 8 (5)
>              attr type: 15 (8)
>              attr type: 21 (12)
>              attr type: 22 (6)
>              attr type: 2 (252)
>              attr type: 18 (64804)
> recv() = 130680
> socket: 10.0.254.1:7013->0.0.0.0:0 (intf 3)
>      family: 2 state: 10 timer: 0 retrans: 0
>      expires: 0 rqueu: 0 wqueue: 1 uid: 0 inode: 456
>              attr type: 8 (5)
>              attr type: 15 (8)
>              attr type: 21 (12)
>              attr type: 22 (6)
>              attr type: 2 (252)
>              attr type: 18 (64768)
>              attr type: 29555 (25966)
> recv() = 130680
> socket: 10.0.254.1:7013->0.0.0.0:0 (intf 3)
>      family: 2 state: 10 timer: 0 retrans: 0
>      expires: 0 rqueu: 0 wqueue: 1 uid: 0 inode: 456
>              attr type: 8 (5)
>              attr type: 15 (8)
>              attr type: 21 (12)
>              attr type: 22 (6)
>              attr type: 2 (252)
>              attr type: 18 (64768)
>              attr type: 29555 (25966)
>              attr type: 8265 (8236)

Here attribute type 18 is INET_DIAG_MD5SIG, the following nlattr types
are junk made of tcp_diag_md5sig's content.

Here is the overflow of the nlattr size:
>>> hex(64768)
'0xfd00'
>>> hex(130300)
'0x1fcfc'

Limit the size of (struct tcp_diag_md5sig) array in the netlink reply by
maximum attribute length. Not perfect as NLM_F_DUMP_INTR will be set on
the netlink header flags, but the userspace can differ if it's due to
inconsistency or due to maximum size of the netlink attribute.

In a following patch set, I'm planning to address this and re-introduce
TCP-MD5-diag that actually works.

Signed-off-by: Dmitry Safonov <0x7f454c46@gmail.com>
---
 net/ipv4/tcp_diag.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/net/ipv4/tcp_diag.c b/net/ipv4/tcp_diag.c
index 722dbfd54d247b4def1e77b1674c5b207c5a939d..d55a0ac39fa0853806efd4a6b38591255e0f4930 100644
--- a/net/ipv4/tcp_diag.c
+++ b/net/ipv4/tcp_diag.c
@@ -72,6 +72,7 @@ static int tcp_diag_put_md5sig(struct sk_buff *skb,
 		return 0;
 
 	attrlen = skb_availroom(skb) - NLA_HDRLEN;
+	attrlen = min(attrlen, U16_MAX - 1); /* attr->nla_len */
 	md5sig_count = min(md5sig_count, attrlen / key_size);
 	attr = nla_reserve(skb, INET_DIAG_MD5SIG, md5sig_count * key_size);
 	if (!attr)

-- 
2.42.2



  parent reply	other threads:[~2024-11-06 18:49 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-11-06 18:10 [PATCH net 0/6] Make TCP-MD5-diag slightly less broken Dmitry Safonov via B4 Relay
2024-11-06 18:10 ` [PATCH net 1/6] net/diag: Do not race on dumping MD5 keys with adding new MD5 keys Dmitry Safonov via B4 Relay
2024-11-07  4:21   ` kernel test robot
2024-11-06 18:10 ` [PATCH net 2/6] net/diag: Warn only once on EMSGSIZE Dmitry Safonov via B4 Relay
2024-11-06 18:10 ` [PATCH net 3/6] net/diag: Pre-allocate optional info only if requested Dmitry Safonov via B4 Relay
2024-11-06 18:10 ` [PATCH net 4/6] net/diag: Always pre-allocate tcp_ulp info Dmitry Safonov via B4 Relay
2024-11-07  0:21   ` Kuniyuki Iwashima
2024-11-07 17:35     ` Dmitry Safonov
2024-11-06 18:10 ` Dmitry Safonov via B4 Relay [this message]
2024-11-07  0:25   ` [PATCH net 5/6] net/diag: Limit TCP-MD5-diag array by max attribute length Kuniyuki Iwashima
2024-11-07 17:46     ` Dmitry Safonov
2024-11-06 18:10 ` [PATCH net 6/6] net/netlink: Correct the comment on netlink message max cap Dmitry Safonov via B4 Relay

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=20241106-tcp-md5-diag-prep-v1-5-d62debf3dded@gmail.com \
    --to=devnull+0x7f454c46.gmail.com@kernel.org \
    --cc=0x7f454c46@gmail.com \
    --cc=borisp@nvidia.com \
    --cc=colona@arista.com \
    --cc=davem@davemloft.net \
    --cc=dsahern@kernel.org \
    --cc=edumazet@google.com \
    --cc=geliang@kernel.org \
    --cc=horms@kernel.org \
    --cc=john.fastabend@gmail.com \
    --cc=kuba@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=martineau@kernel.org \
    --cc=matttbe@kernel.org \
    --cc=mptcp@lists.linux.dev \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    /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®