From: Fang Xieyan <fangxy@xiaopeng.com>
To: Loic Poulain <loic.poulain@oss.qualcomm.com>,
Sergey Ryazanov <ryazanov.s.a@gmail.com>,
netdev@vger.kernel.org
Cc: "David S . Miller" <davem@davemloft.net>,
Eric Dumazet <edumazet@kernel.org>,
Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
Andrew Lunn <andrew+netdev@lunn.ch>,
Simon Horman <horms@kernel.org>,
Johannes Berg <johannes@sipsolutions.net>,
Maoyi Xie <maoyixie.tju@gmail.com>,
Guanglei Zhu <zhugl3@xiaopeng.com>,
linux-kernel@vger.kernel.org, stable@vger.kernel.org
Subject: [PATCH net v3 0/5] net: wwan: iosm: validate MUX DL lengths and offsets
Date: Fri, 20 Nov 2026 18:55:54 +0800 [thread overview]
Message-ID: <20261120105559.54519-1-fangxy@xiaopeng.com> (raw)
The MUX downlink decoders trust several modem supplied lengths and
offsets without bounding them against what actually arrived: the block
size from the transfer descriptor, the ADGH datagram length, the ACBH
block length, the command chain offsets and the declared parameter
length. A truncated or enlarged DL block makes the driver read past the
received data, hand the network stack an undeclared payload, or walk a
command chain that loops or redecodes the same bytes. The aggregated
datagram block path is already covered by commit 526b8ef54668 ("net:
wwan: iosm: bound device offsets in the MUX downlink decoder") and is
not touched here.
This series adds those bounds. No command type, response or protocol
semantics change, so a conforming modem behaves exactly as before. A
parameter too short for the member its command type reads is dropped
with -EMSGSIZE, which also ends an aggregated block at that command.
Patch 4 is the prep change for patch 5; both carry Cc: stable, as patch
5 cannot be backported without patch 4.
cmd_len is not a parameter bound in either path. In an aggregated block
the boundary is next_cmd_index, or the end of the received block for the
last command. In MUX Lite there is one command per skb, so the received
length is already everything that command owns; the flow control decoder
treats a mismatching cmd_len as a legacy device, so the field is a
compatibility flag, not a length bound.
Patch 3 bounds the ACBH command chain by the received length rather
than block_length. Although the structure describes block_length as
the length of the aggregated command block, the uplink encoder
initializes it to the ACBH header size and does not update it as
commands are appended, so it cannot delimit the received command
chain. The field is sanity checked, while first_cmd_index and
next_cmd_index are bounded against skb->len. The walk's
array_index_nospec() call is removed: it bounded a byte offset by
sizeof(struct mux_cmdh), zeroing every legal offset past that size.
The offset indexes no array; array_index_nospec() is kept only for
if_id, an actual array index.
Patch 2 relies on the ADGH header check from patch 1, patch 5 on the
ACBH validation from patch 3 and the parameter pointer from patch 4.
Every commit applies and builds individually, W=1 drivers/net/wwan/iosm/
is clean. Found by reviewing the DL decode path against the MUX protocol
structures, not reproduced on a live modem: reaching it needs a device
that truncates or enlarges its own downlink blocks. The series is based
on net-next at 9068e05d35f3.
Changes in v3:
- Rebase onto net-next (9068e05d35f3) so the series applies on top of
66974df65a0b ("net: wwan: iosm: drop redundant skb->data check in MUX
DL decode"). Patch 1 no longer anchors its signature-length check on
the removed skb->data test; the check now sits directly before the
signature read. This clears the v2 patchwork context failure at
iosm_ipc_mux_codec.c:696 and the dependent context-hw gate failure.
- Patch 5 wraps the over-long continuation line in
ipc_mux_dl_cmd_decode(): param_len now goes on its own line, clearing
the checkpatch "line length of 82 exceeds 80 columns" warning.
- Cc Guanglei Zhu, whom get_maintainer.pl now lists for this file after
his review of 66974df65a0b (the v2 cc_maintainers warning).
Changes in v2, all three points from the Sashiko AI review of v1:
- Patch 5 bounds the MUX Lite parameter by the received length alone,
not by cmd_len: v1 dropped legacy FLOW_CTL (cmd_len 16..19) and
OPEN_SESSION_RESP (16..27) even when the parameter bytes had arrived,
defeating the backward compatibility branch in
ipc_mux_dl_cmds_decode_process() that exists for exactly these (the
protocol minimum for cmd_len is 0x10).
- Patch 2 rejects adgh_len == packet_offset too: the zero length clone
reached ipc_wwan_receive(), which reads skb->data[0] without checking
skb->len. mux_dl_process_dg() already rejects the same case for the
aggregated path with dl_head_pad_len >= dg_len.
- Patch 4 gains Fixes: 9413491e20e1 and Cc: stable@vger.kernel.org: its
by-value signature copied all twelve union bytes even for a four byte
response, and patch 5 bounds the session responses, so the two are
needed together. A prerequisite itself marked for stable and present
in the series need not be listed, per stable-kernel-rules.rst.
Patch 1 is unchanged. Patch 3 changes only its changelog and comments.
v2:
https://lore.kernel.org/netdev/20261117173940.34970-1-fangxy@xiaopeng.com/
v1:
https://lore.kernel.org/netdev/20261112125201.1019-1-fangxy@xiaopeng.com/
Fang Xieyan (5):
net: wwan: iosm: validate the MUX DL block length
net: wwan: iosm: validate the ADGH datagram length
net: wwan: iosm: bound the ACBH command chain walk
net: wwan: iosm: pass the DL command parameter by pointer
net: wwan: iosm: validate the command parameter length
drivers/net/wwan/iosm/iosm_ipc_mux_codec.c | 186 ++++++++++++++++-----
1 file changed, 141 insertions(+), 45 deletions(-)
--
2.50.1
next reply other threads:[~2026-11-20 10:56 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-20 10:55 Fang Xieyan [this message]
2026-11-20 10:55 ` [PATCH net v3 1/5] net: wwan: iosm: validate the MUX DL block length Fang Xieyan
2026-11-20 10:55 ` [PATCH net v3 2/5] net: wwan: iosm: validate the ADGH datagram length Fang Xieyan
2026-11-20 10:55 ` [PATCH net v3 3/5] net: wwan: iosm: bound the ACBH command chain walk Fang Xieyan
2026-11-20 10:55 ` [PATCH net v3 4/5] net: wwan: iosm: pass the DL command parameter by pointer Fang Xieyan
2026-11-20 10:55 ` [PATCH net v3 5/5] net: wwan: iosm: validate the command parameter length Fang Xieyan
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=20261120105559.54519-1-fangxy@xiaopeng.com \
--to=fangxy@xiaopeng.com \
--cc=andrew+netdev@lunn.ch \
--cc=davem@davemloft.net \
--cc=edumazet@kernel.org \
--cc=horms@kernel.org \
--cc=johannes@sipsolutions.net \
--cc=kuba@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=loic.poulain@oss.qualcomm.com \
--cc=maoyixie.tju@gmail.com \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=ryazanov.s.a@gmail.com \
--cc=stable@vger.kernel.org \
--cc=zhugl3@xiaopeng.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®