DF-2542

mfc entries kmalloc without M_ZERO only [0 numvifs) of mfc_ttls/mfc_flags initialized later add_vif exposes uninitialized heap bytes via forwarding decisions

Summary

Both mfc allocation sites use kmalloc no M_ZERO then init_mfc_params/update_mfc_params only initialize mfc_ttls[i]/mfc_flags[i] for i in [0 numvifs). If numvifs later grows (add_vif :853) new vif slots in every existing mfc remain uninitialized heap data. ip_mdq reads them (:1569) to decide whether to forward. Non-zero garbage with ip_ttl>garbage causes multicast replicated onto unauthorized vif. Weak byte-is-nonzero oracle into prior kernel heap. Root-only multicast routing daemon. [VERIFIED pass2: not reproduced in budget - mechanism code-certain kmalloc no M_ZERO with growth and ip_mdq read, full MRTsockopt control-plane succeeded but forwarding oracle stayed 0; M_ZERO fix in combined kernel]

Discussion (0)

No comments yet.

Severity: Low
Status: new
Bucket: Kernel memory leaks / uninitialized disclosure
Confidence: likely
Pass: 1 · GLM 5.2 first pass
CVSS 3.1: CVSS:3.1/AV:L/AC:H/PR:H/UI:N/S:U:C:L/I:N/A:L
CWE: CWE-908 Use of Uninitialized Resource
File: sys/net/ip_mroute/ip_mroute.c
Lines: 1055-1569
Area: net
Discovered: 2026-07-28
Updated: 2026-07-28

Team triage

untriaged

priority: — · owner: —

PoC verification

not_reproduced

Status: not_reproduced
Reproduced: no
Impact: none
Confidence: likely
Tested: 2026-10-24 14:11:27
Attempts: 4
Runtime: 600s
Code hash: —
Guest: DragonFly dfbsd 6.5-DEVELOPMENT DragonFly 6.5-DEVELOPMENT #0: Thu Jul 2 06:02:54 UTC 2026 root@dfbsd:/usr/obj/usr/src/sys/X86_64_GENERIC x86_64
Build cmd: cc -O -o mroute2542 mroute2542.c
Run cmd: ./mroute2542

jump to evidence pack ↓

Fix verification

not_testable

Fix status: not_testable
Baseline reproduced: —
Patched reproduced: —
Confidence: likely
Tested: 2026-10-24 14:11:27
Runtime: —
Kernel: per-fix-DF-2542
fix.diff: findings/poc/DF-2542/fix.diff
Runner: df-bsd-reviewer

jump to fix verification ↓

PoC verification

Evidence pack

findings/poc/DF-2542 · 2 files

File	Type	Description	Size
mroute2542.c	—		6.8 KB	view raw
run.log	—		861 B	view raw

Fix verification

not_testable

↓ fix.diffper-fix-DF-2542

Confirmed kernel references

Detail

Evidence (decisive lines)

mroute2542_run.log: full control-plane sequence (all sockopts 0); forwarding oracle delta +0

PoC changes

real-API rig (MRT_* sockopts + SIOCGETVIFCNT); data-plane injection added; grooming via freed MFC slot

Verified recommended fix

allocate mfcs with M_ZERO (included in the combined fix kernel)

Verdict

The allocation and partial-initialization mechanics are confirmed in source: both mfc allocations use kmalloc without M_ZERO (ip_mroute.c:1056/1061), update_mfc_params initializes mfc_ttls/mfc_flags only for i<numvifs (:934), add_vif grows numvifs without touching existing mfcs (:853), and ip_mdq reads the stale slots (:1569). The control-plane rig (MRT_INIT/MRT_ADD_VIF/groom-MFC/del-vif/re-add-vif on the real ip_mroute.ko via the real socket API) executed every step cleanly, but the data-plane forwarding oracle (SIOCGETVIFCNT delta after injecting a multicast packet for the crafted (S,G)) stayed 0 in the timeboxed runs: kernel-originated loopback multicast did not reach ip_mdq with a cache hit in this setup (cache-miss handling expects a listening multicast daemon per design). The mechanism remains code-certain; the unprivileged/reachable impact ceiling is as filed (root-configured multicast router).