CVE-2024-26923: af_unix: Fix garbage collector racing against connect()

In the Linux kernel, the following vulnerability has been resolved:

afunix: Fix garbage collector racing against connect()

Garbage collector does not take into account the risk of embryo getting enqueued during the garbage collection. If such embryo has a peer that carries SCMRIGHTS, two consecutive passes of scanchildren() may see a different set of children. Leading to an incorrectly elevated inflight count, and then a dangling pointer within the gcinflightlist.

sockets are AFUNIX/SOCKSTREAM S is an unconnected socket L is a listening in-flight socket bound to addr, not in fdtable V's fd will be passed via sendmsg(), gets inflight count bumped

connect(S, addr) sendmsg(S, [V]); close(V) unixgc() ---------------- ------------------------- -----------

NS = unixcreate1() skb1 = sockwmalloc(NS) L = unixfindother(addr) unixstatelock(L) unixpeer(S) = NS // V count=1 inflight=0

NS = unixpeer(S) skb2 = sockalloc() skbqueuetail(NS, skb2[V])

// V became in-flight // V count=2 inflight=1

close(V)

// V count=1 inflight=1 // GC candidate condition met

for u in gcinflightlist: if (totalrefs == inflightrefs) add u to gccandidates

// gccandidates={L, V}

for u in gccandidates: scanchildren(u, decinflight)

// embryo (skb1) was not // reachable from L yet, so V's // inflight remains unchanged skbqueuetail(L, skb1) unixstateunlock(L) for u in gccandidates: if (u.inflight) scanchildren(u, incinflightmovetail)

// V count=1 inflight=2 (!)

If there is a GC-candidate listening socket, lock/unlock its state. This makes GC wait until the end of any ongoing connect() to that socket. After flipping the lock, a possibly SCM-laden embryo is already enqueued. And if there is another embryo coming, it can not possibly carry SCMRIGHTS. At this point, unixinflight() can not happen because unixgclock is already taken. Inflight graph remains unaffected.