Thread (7 messages) 7 messages, 2 authors, 19h ago
HOTtoday

Revision v5 of 3 in this series.

Revisions (3)
  1. v3 [diff vs current]
  2. v4 [diff vs current]
  3. v5 current

[PATCH nf-next v5 0/4] netfilter: conntrack: shared port parser for helpers

From: Rahul Chandelkar <hidden>
Date: 2026-09-28 16:29:12
Also in: netfilter-devel

The IRC and Amanda conntrack helpers parse port numbers from packet
payload with simple_strtoul(), which returns unsigned long without a
range check, so a port above 65535 is truncated when it is stored in a
u16 (65537 becomes 1) and an expectation is created for a port that
never appeared in the payload.  Amanda also cuts a port field of six or
more digits down to its first five.  Both problems are present in
current nf-next.

Instead of open-coding a range check in each helper, this series moves
the sip_strtouint() parser of nf_conntrack_sip into the helper core as
nf_ct_helper_parse_uint(), adds nf_ct_helper_parse_port() on top of it,
and converts the three helpers:

  1/4 adds the two parsers
  2/4 and 3/4 fix the IRC and Amanda helpers
  4/4 removes the now duplicate parsing code from the SIP helper
      (no functional change)

Testing: allmodconfig and allyesconfig builds with W=1, sparse, an
allnoconfig build, i386 and big-endian powerpc cross-builds, and a
build of each patch in turn.  The old and new sip_parse_port() were
compared in a userspace harness built from the kernel source on three
million random inputs with no difference.  The helpers have not been
run-time tested with IRC, Amanda or SIP traffic.

Tool use: v4 and v5 were prepared with Claude Code, an AI coding
assistant, as recorded in the Assisted-by tags.  It wrote most of the
code changes and changelogs from v3 and the review feedback, and ran
the builds and tests listed above.  v4 should have carried the tags as
well; it did not, sorry about that.

Changes in v5:
  - nf_ct_helper_parse_port(): reject digit runs longer than five
    characters, so that sip_parse_port() still rejects zero-padded
    ports and 4/4 has no functional change; return -EINVAL rather
    than -1
  - add kernel-doc for both parsers, including their digit limits
  - amanda: make pbuf one byte larger so that a port field of six or
    more digits is rejected instead of being cut to its first five
    digits
  - irc: remove the dcc_port == 0 test in help(), which can no longer
    be true
  - sip: rewrap call sites that no longer fit in 80 columns
  - add Assisted-by tags
  (the first four items address the Sashiko review of v4)
v4: https://lore.kernel.org/all/20260923091608.2617604-1-rc@rexion.ai/ (local)
  - rebase onto nf-next; move sip_strtouint() into the helper core
    instead of adding a second parser, and convert nf_conntrack_sip
    (Pablo Neira Ayuso, Phil Sutter, Florian Westphal)
  - use a real name and target nf-next (Pablo Neira Ayuso)
v3: https://lore.kernel.org/all/20260503083220.630655-1-rc@rexion.ai/ (local)
  - add nf_ct_helper_parse_uint() and convert nf_conntrack_sip
    (Phil Sutter)
v2: https://lore.kernel.org/all/20260501063156.2520780-1-rc@rexion.ai/ (local)
  - add a shared nf_ct_helper_parse_port() in the helper core instead
    of open-coding range checks (Florian Westphal, Pablo Neira Ayuso)
v1: https://lore.kernel.org/all/20260430161230.3438973-1-rc@rexion.ai/ (local)

Rahul Chandelkar (4):
  netfilter: conntrack: add shared uint and port parsers for helpers
  netfilter: nf_conntrack_irc: use nf_ct_helper_parse_port()
  netfilter: nf_conntrack_amanda: use nf_ct_helper_parse_port()
  netfilter: nf_conntrack_sip: use shared helper parsers

 include/net/netfilter/nf_conntrack_helper.h |  5 ++
 net/netfilter/nf_conntrack_amanda.c         | 13 ++-
 net/netfilter/nf_conntrack_helper.c         | 79 +++++++++++++++++
 net/netfilter/nf_conntrack_irc.c            |  9 +-
 net/netfilter/nf_conntrack_sip.c            | 93 ++++++---------------
 5 files changed, 123 insertions(+), 76 deletions(-)


base-commit: 87b80c2f6b05cad9f0ff9136709c62a0f59923e3
-- 
2.43.0
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help