On Tue, Jun 28, 2022 at 09:05:09PM +0300, Vitaliy Makkoveev wrote:
> The updated diff. It was triggered by Hrvoje Popovski, we could do
> direct (*if_qstart)() call in pipex(4) PPPOE input path, and this
> provides deadlock. The updated diff uses ip{,6}_send() instead of
> ipv{4,6}_input().
>

I think, I found better solution. The diff below still uses
ipv{4,6}_input(), but introduces custom pipex_if_enqueue()
(*if_enqueue)() handler. Now we can be sure, our (*if_qstart)() handlers
will never be called directly. Also we can be sure about netlock state
within our (*if_qstart)() handlers and remove pipexintr().

We do direct (*if_qstart)() call only if we overflow `if_snd' queue, and
of course we loose packets in this case, so the behavior mostly the
same.

Index: sys/net/if_pppx.c
===================================================================
RCS file: /cvs/src/sys/net/if_pppx.c,v
retrieving revision 1.117
diff -u -p -r1.117 if_pppx.c
--- sys/net/if_pppx.c   26 Jun 2022 22:51:58 -0000      1.117
+++ sys/net/if_pppx.c   28 Jun 2022 19:56:47 -0000
@@ -651,6 +651,7 @@ pppx_add_session(struct pppx_dev *pxd, s
        ifp->if_mtu = req->pr_peer_mru; /* XXX */
        ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST | IFF_UP;
        ifp->if_xflags = IFXF_CLONED | IFXF_MPSAFE;
+       ifp->if_enqueue = pipex_if_enqueue;
        ifp->if_qstart = pppx_if_qstart;
        ifp->if_output = pppx_if_output;
        ifp->if_ioctl = pppx_if_ioctl;
@@ -659,8 +660,6 @@ pppx_add_session(struct pppx_dev *pxd, s
        ifp->if_softc = pxi;
        /* ifp->if_rdomain = req->pr_rdomain; */
        if_counters_alloc(ifp);
-       /* XXXSMP: be sure pppx_if_qstart() called with NET_LOCK held */
-       ifq_set_maxlen(&ifp->if_snd, 1);
 
        /* XXXSMP breaks atomicity */
        NET_UNLOCK();
@@ -801,13 +800,16 @@ pppx_if_qstart(struct ifqueue *ifq)
        struct mbuf *m;
        int proto;
 
-       NET_ASSERT_LOCKED();
+       mtx_enter(&pxi->pxi_session->pxs_mtx);
+
        while ((m = ifq_dequeue(ifq)) != NULL) {
                proto = *mtod(m, int *);
                m_adj(m, sizeof(proto));
 
                pipex_ppp_output(m, pxi->pxi_session, proto);
        }
+
+       mtx_leave(&pxi->pxi_session->pxs_mtx);
 }
 
 int
@@ -1041,11 +1043,10 @@ pppacopen(dev_t dev, int flags, int mode
        ifp->if_flags = IFF_SIMPLEX | IFF_BROADCAST;
        ifp->if_xflags = IFXF_CLONED | IFXF_MPSAFE;
        ifp->if_rtrequest = p2p_rtrequest; /* XXX */
+       ifp->if_enqueue = pipex_if_enqueue;
        ifp->if_output = pppac_output;
        ifp->if_qstart = pppac_qstart;
        ifp->if_ioctl = pppac_ioctl;
-       /* XXXSMP: be sure pppac_qstart() called with NET_LOCK held */
-       ifq_set_maxlen(&ifp->if_snd, 1);
 
        if_counters_alloc(ifp);
        if_attach(ifp);
@@ -1441,7 +1442,6 @@ pppac_qstart(struct ifqueue *ifq)
        struct ip ip;
        int rv;
 
-       NET_ASSERT_LOCKED();
        while ((m = ifq_dequeue(ifq)) != NULL) {
 #if NBPFILTER > 0
                if (ifp->if_bpf) {
Index: sys/net/pipex.c
===================================================================
RCS file: /cvs/src/sys/net/pipex.c,v
retrieving revision 1.142
diff -u -p -r1.142 pipex.c
--- sys/net/pipex.c     28 Jun 2022 08:01:40 -0000      1.142
+++ sys/net/pipex.c     28 Jun 2022 19:56:47 -0000
@@ -91,7 +91,6 @@ struct pool mppe_key_pool;
  * Locks used to protect global data
  *       A       atomic operation
  *       I       immutable after creation
- *       N       net lock
  *       L       pipex_list_mtx
  */
 
@@ -172,7 +171,6 @@ pipex_ioctl(void *ownersc, u_long cmd, c
 {
        int ret = 0;
 
-       NET_ASSERT_LOCKED();
        switch (cmd) {
        case PIPEXCSESSION:
                ret = pipex_config_session(
@@ -197,6 +195,19 @@ pipex_ioctl(void *ownersc, u_long cmd, c
        return (ret);
 }
 
+int
+pipex_if_enqueue(struct ifnet *ifp, struct mbuf *m)
+{
+       struct ifqueue *ifq = &ifp->if_snd;
+       int error;
+
+       if ((error = ifq_enqueue(ifq, m)) != 0)
+               return (error);
+       task_add(ifq->ifq_softnet, &ifq->ifq_bundle);
+
+       return (0);
+}
+
 /************************************************************************
  * Software Interrupt Handler
  ************************************************************************/
@@ -575,18 +586,20 @@ pipex_config_session(struct pipex_sessio
        struct pipex_session *session;
        int error = 0;
 
-       NET_ASSERT_LOCKED();
-
        session = pipex_lookup_by_session_id(req->pcr_protocol,
            req->pcr_session_id);
        if (session == NULL)
                return (EINVAL);
 
        if (session->ownersc == ownersc) {
+               mtx_enter(&session->pxs_mtx);
+
                if (req->pcr_ip_forward != 0)
                        session->flags |= PIPEX_SFLAGS_IP_FORWARD;
                else
                        session->flags &= ~PIPEX_SFLAGS_IP_FORWARD;
+
+               mtx_leave(&session->pxs_mtx);
        } else
                error = EINVAL;
 
@@ -601,8 +614,6 @@ pipex_get_stat(struct pipex_session_stat
        struct pipex_session *session;
        int error = 0;
 
-       NET_ASSERT_LOCKED();
-
        session = pipex_lookup_by_session_id(req->psr_protocol,
            req->psr_session_id);
        if (session == NULL)
@@ -811,6 +822,9 @@ pipex_ip_output(struct mbuf *m0, struct 
                /*
                 * Multicast packet is a idle packet and it's not TCP.
                 */
+
+               mtx_enter(&session->pxs_mtx);
+
                if ((session->flags & (PIPEX_SFLAGS_IP_FORWARD |
                    PIPEX_SFLAGS_IP6_FORWARD)) == 0)
                        goto drop;
@@ -837,6 +851,8 @@ pipex_ip_output(struct mbuf *m0, struct 
                }
 
                pipex_ppp_output(m0, session, PPP_IP);
+
+               mtx_leave(&session->pxs_mtx);
        } else {
                struct pipex_session *session_tmp;
                struct mbuf *m;
@@ -846,16 +862,21 @@ pipex_ip_output(struct mbuf *m0, struct 
                LIST_FOREACH(session_tmp, &pipex_session_list, session_list) {
                        if (session_tmp->ownersc != session->ownersc)
                                continue;
+
+                       mtx_enter(&session->pxs_mtx);
+
                        if ((session->flags & (PIPEX_SFLAGS_IP_FORWARD |
                            PIPEX_SFLAGS_IP6_FORWARD)) == 0)
-                               continue;
+                               goto next;
                        m = m_copym(m0, 0, M_COPYALL, M_NOWAIT);
                        if (m == NULL) {
                                counters_inc(session->stat_counters,
                                    pxc_oerrors);
-                               continue;
+                               goto next;
                        }
                        pipex_ppp_output(m, session_tmp, PPP_IP);
+next:
+                       mtx_leave(&session->pxs_mtx);
                }
                m_freem(m0);
        }
@@ -864,6 +885,7 @@ pipex_ip_output(struct mbuf *m0, struct 
 drop:
        m_freem(m0);
 dropped:
+       mtx_leave(&session->pxs_mtx);
        counters_inc(session->stat_counters, pxc_oerrors);
 }
 
@@ -872,7 +894,7 @@ pipex_ppp_output(struct mbuf *m0, struct
 {
        u_char *cp, hdr[16];
 
-       NET_ASSERT_LOCKED();
+       MUTEX_ASSERT_LOCKED(&session->pxs_mtx);
 
 #ifdef PIPEX_MPPE
        if (pipex_session_is_mppe_enabled(session)) {
@@ -927,6 +949,8 @@ pipex_ppp_input(struct mbuf *m0, struct 
        int proto, hlen = 0;
        struct mbuf *n;
 
+       MUTEX_ASSERT_LOCKED(&session->pxs_mtx);
+
        KASSERT(m0->m_pkthdr.len >= PIPEX_PPPMINLEN);
        proto = pipex_ppp_proto(m0, session, 0, &hlen);
 #ifdef PIPEX_MPPE
@@ -1288,11 +1312,16 @@ pipex_pppoe_input(struct mbuf *m0, struc
            sizeof(struct pipex_pppoe_header), &pppoe);
 
        hlen = sizeof(struct ether_header) + sizeof(struct pipex_pppoe_header);
-       if ((m0 = pipex_common_input(session, m0, hlen, ntohs(pppoe.length)))
-           == NULL)
-               return (NULL);
-       m_freem(m0);
-       counters_inc(session->stat_counters, pxc_ierrors);
+
+       mtx_enter(&session->pxs_mtx);
+       m0 = pipex_common_input(session, m0, hlen, ntohs(pppoe.length));
+       mtx_leave(&session->pxs_mtx);
+
+       if (m0 != NULL) {
+               m_freem(m0);
+               counters_inc(session->stat_counters, pxc_ierrors);
+       }
+
        return (NULL);
 }
 
@@ -1354,6 +1383,8 @@ pipex_pptp_output(struct mbuf *m0, struc
        struct ip *ip;
        u_char *cp;
 
+       MUTEX_ASSERT_LOCKED(&session->pxs_mtx);
+
        reqlen = PIPEX_IPGRE_HDRLEN + (has_seq + has_ack) * 4;
 
        len = 0;
@@ -1507,7 +1538,6 @@ pipex_pptp_input(struct mbuf *m0, struct
        struct pipex_pptp_session *pptp_session;
        int rewind = 0;
 
-       NET_ASSERT_LOCKED();
        KASSERT(m0->m_pkthdr.len >= PIPEX_IPGRE_HDRLEN);
        pptp_session = &session->proto.pptp;
 
@@ -1537,6 +1567,9 @@ pipex_pptp_input(struct mbuf *m0, struct
                seqp = cp;
                GETLONG(seq, cp);
        }
+
+       mtx_enter(&session->pxs_mtx);
+
        if (has_ack) {
                ackp = cp;
                GETLONG(ack, cp);
@@ -1585,6 +1618,7 @@ pipex_pptp_input(struct mbuf *m0, struct
                /* ok,  The packet is for PIPEX */
                if (!rewind)
                        session->proto.pptp.rcv_gap += nseq;
+               mtx_leave(&session->pxs_mtx);
                return (NULL);
        }
 
@@ -1618,6 +1652,8 @@ not_ours:
                PUTLONG(ack, ackp);
        }
 
+       mtx_leave(&session->pxs_mtx);
+
        return (m0);
 out_seq:
        pipex_session_log(session, LOG_DEBUG,
@@ -1626,6 +1662,7 @@ out_seq:
            pptp_session->rcv_nxt + pptp_session->maxwinsz,
            ack, pptp_session->snd_una,
            pptp_session->snd_nxt);
+       mtx_leave(&session->pxs_mtx);
 
        /* FALLTHROUGH */
 drop:
@@ -1755,6 +1792,8 @@ pipex_pptp_userland_output(struct mbuf *
        gre = mtod(m0, struct pipex_gre_header *);
        cp = PIPEX_SEEK_NEXTHDR(gre, sizeof(struct pipex_gre_header), u_char *);
 
+       mtx_enter(&session->pxs_mtx);
+
        /*
         * overwrite sequence numbers to adjust a gap between pipex and
         * userland.
@@ -1775,6 +1814,8 @@ pipex_pptp_userland_output(struct mbuf *
                        session->proto.pptp.rcv_acked = val32;
        }
 
+       mtx_leave(&session->pxs_mtx);
+
        return (m0);
 }
 #endif /* PIPEX_PPTP */
@@ -1796,6 +1837,8 @@ pipex_l2tp_output(struct mbuf *m0, struc
 #endif
        struct m_tag *mtag;
 
+       MUTEX_ASSERT_LOCKED(&session->pxs_mtx);
+
        hlen = sizeof(struct pipex_l2tp_header) +
            ((pipex_session_is_l2tp_data_sequencing_on(session))
                    ? sizeof(struct pipex_l2tp_seq_header) : 0) +
@@ -1967,17 +2010,15 @@ pipex_l2tp_input(struct mbuf *m0, int of
     uint32_t ipsecflowinfo)
 {
        struct pipex_l2tp_session *l2tp_session;
-       int length, offset, hlen, nseq;
-       u_char *cp, *nsp, *nrp;
+       int length = 0, offset = 0, hlen, nseq;
+       u_char *cp, *nsp = NULL, *nrp = NULL;
        uint16_t flags, ns = 0, nr = 0;
        int rewind = 0;
 
-       NET_ASSERT_LOCKED();
+       mtx_enter(&session->pxs_mtx);
 
-       length = offset = ns = nr = 0;
        l2tp_session = &session->proto.l2tp;
        l2tp_session->ipsecflowinfo = ipsecflowinfo;
-       nsp = nrp = NULL;
 
        m_copydata(m0, off0, sizeof(flags), &flags);
 
@@ -2039,6 +2080,8 @@ pipex_l2tp_input(struct mbuf *m0, int of
                /* ok,  The packet is for PIPEX */
                if (!rewind)
                        session->proto.l2tp.nr_gap += nseq;
+               mtx_leave(&session->pxs_mtx);
+
                return (NULL);
        }
 
@@ -2069,15 +2112,18 @@ pipex_l2tp_input(struct mbuf *m0, int of
                PUTSHORT(nr, nrp);
        }
 
+       mtx_leave(&session->pxs_mtx);
+
        return (m0);
 out_seq:
        pipex_session_log(session, LOG_DEBUG,
            "Received bad data packet: out of sequence: seq=%u(%u-) "
            "ack=%u(%u-%u)", ns, l2tp_session->nr_nxt, nr, l2tp_session->ns_una,
            l2tp_session->ns_nxt);
-
        /* FALLTHROUGH */
 drop:
+       mtx_leave(&session->pxs_mtx);
+
        m_freem(m0);
        counters_inc(session->stat_counters, pxc_ierrors);
 
@@ -2202,6 +2248,8 @@ pipex_l2tp_userland_output(struct mbuf *
        ns = ntohs(seq->ns);
        nr = ntohs(seq->nr);
 
+       mtx_enter(&session->pxs_mtx);
+
        ns += session->proto.l2tp.ns_gap;
        seq->ns = htons(ns);
        session->proto.l2tp.ns_nxt++;
@@ -2211,6 +2259,8 @@ pipex_l2tp_userland_output(struct mbuf *
        if (SEQ16_GT(nr, session->proto.l2tp.nr_acked))
                session->proto.l2tp.nr_acked = nr;
 
+       mtx_leave(&session->pxs_mtx);
+
        return (m0);
 }
 #endif /* PIPEX_L2TP */
@@ -2379,6 +2429,8 @@ pipex_mppe_input(struct mbuf *m0, struct
        u_char *cp;
        int rewind = 0;
 
+       MUTEX_ASSERT_LOCKED(&session->pxs_mtx);
+
        /* pullup */
        PIPEX_PULLUP(m0, sizeof(coher_cnt));
        if (m0 == NULL)
@@ -2471,10 +2523,8 @@ pipex_mppe_input(struct mbuf *m0, struct
                        /* Send CCP ResetReq */
                        PIPEX_DBG((session, LOG_DEBUG, "CCP SendResetReq"));
 
-                       mtx_enter(&session->pxs_mtx);
                        ccp_id = session->ccp_id;
                        session->ccp_id++;
-                       mtx_leave(&session->pxs_mtx);
 
                        pipex_ccp_output(session, CCP_RESETREQ, ccp_id);
                        goto drop;
Index: sys/net/pipex_local.h
===================================================================
RCS file: /cvs/src/sys/net/pipex_local.h,v
retrieving revision 1.47
diff -u -p -r1.47 pipex_local.h
--- sys/net/pipex_local.h       26 Jun 2022 15:50:21 -0000      1.47
+++ sys/net/pipex_local.h       28 Jun 2022 19:56:47 -0000
@@ -56,8 +56,8 @@ extern struct mutex pipex_list_mtx;
 
 /*
  * Locks used to protect struct members:
+ *      A       atomic operation
  *      I       immutable after creation
- *      N       net lock
  *      L       pipex_list_mtx
  *      s       this pipex_session' `pxs_mtx'
  *      m       this pipex_mppe' `pxm_mtx'
@@ -91,14 +91,14 @@ struct pipex_pppoe_session {
 #ifdef PIPEX_PPTP
 struct pipex_pptp_session {
        /* sequence number gap between pipex and userland */
-       int32_t snd_gap;                /* [N] gap of our sequence */
-       int32_t rcv_gap;                /* [N] gap of peer's sequence */
-       int32_t ul_snd_una;             /* [N] userland send acked seq */
-
-       uint32_t snd_nxt;               /* [N] send next */
-       uint32_t rcv_nxt;               /* [N] receive next */
-       uint32_t snd_una;               /* [N] send acked sequence */
-       uint32_t rcv_acked;             /* [N] recv acked sequence */
+       int32_t snd_gap;                /* [s] gap of our sequence */
+       int32_t rcv_gap;                /* [s] gap of peer's sequence */
+       int32_t ul_snd_una;             /* [s] userland send acked seq */
+
+       uint32_t snd_nxt;               /* [s] send next */
+       uint32_t rcv_nxt;               /* [s] receive next */
+       uint32_t snd_una;               /* [s] send acked sequence */
+       uint32_t rcv_acked;             /* [s] recv acked sequence */
 
        int winsz;                      /* [I] windows size */
        int maxwinsz;                   /* [I] max windows size */
@@ -141,16 +141,16 @@ struct pipex_l2tp_session {
 
        uint32_t option_flags;          /* [I] protocol options */
 
-       int16_t ns_gap;         /* [N] gap between userland and pipex */
-       int16_t nr_gap;         /* [N] gap between userland and pipex */
-       uint16_t ul_ns_una;     /* [N] unacked sequence number (userland) */
-
-       uint16_t ns_nxt;        /* [N] next sequence number to send */
-       uint16_t ns_una;        /* [N] unacked sequence number to send */
-
-       uint16_t nr_nxt;        /* [N] next sequence number to recv */
-       uint16_t nr_acked;      /* [N] acked sequence number to recv */
-       uint32_t ipsecflowinfo; /* [N] IPsec SA flow id for NAT-T */
+       int16_t ns_gap;         /* [s] gap between userland and pipex */
+       int16_t nr_gap;         /* [s] gap between userland and pipex */
+       uint16_t ul_ns_una;     /* [s] unacked sequence number (userland) */
+
+       uint16_t ns_nxt;        /* [s] next sequence number to send */
+       uint16_t ns_una;        /* [s] unacked sequence number to send */
+
+       uint16_t nr_nxt;        /* [s] next sequence number to recv */
+       uint16_t nr_acked;      /* [s] acked sequence number to recv */
+       uint32_t ipsecflowinfo; /* [s] IPsec SA flow id for NAT-T */
 };
 #endif /* PIPEX_L2TP */
 
@@ -180,9 +180,9 @@ struct pipex_session {
 
        uint32_t        idle_time;      /* [L] idle time in seconds */
 
-       u_int           flags;          /* [N] flags, see below */
-#define PIPEX_SFLAGS_IP_FORWARD                0x01 /* [N] enable IP 
forwarding */
-#define PIPEX_SFLAGS_IP6_FORWARD       0x02 /* [N] enable IPv6 forwarding */
+       u_int           flags;          /* [s] flags, see below */
+#define PIPEX_SFLAGS_IP_FORWARD                0x01 /* [s] enable IP 
forwarding */
+#define PIPEX_SFLAGS_IP6_FORWARD       0x02 /* [s] enable IPv6 forwarding */
 #define PIPEX_SFLAGS_MULTICAST         0x04 /* [I] virtual entry for
                                                multicast */
 #define PIPEX_SFLAGS_PPPX              0x08 /* [I] interface is
@@ -200,7 +200,7 @@ struct pipex_session {
        struct sockaddr_in6 ip6_address; /* [I] remote IPv6 address */
        int             ip6_prefixlen;   /* [I] remote IPv6 prefixlen */
 
-       u_int           ifindex;                /* [N] interface index */
+       u_int           ifindex;                /* [A] interface index */
        void            *ownersc;               /* [I] owner context */
 
        uint32_t        ppp_flags;              /* [I] configure flags */
@@ -469,3 +469,4 @@ int                   pipex_ppp_enqueue 
 void                  pipex_timer_start (void);
 void                  pipex_timer_stop (void);
 void                  pipex_timer (void *);
+int                   pipex_if_enqueue(struct ifnet *, struct mbuf *);

Reply via email to