FreeBSD Bugzilla – Attachment 209951 Details for
Bug 242637
emulators/virtualbox-ose: Fix bridging when ng_ether node name does not match interface name
Home
|
New
|
Browse
|
Search
|
[?]
|
Reports
|
Help
|
New Account
|
Log In
Remember
[x]
|
Forgot Password
Login:
[x]
[patch]
Patch for current (5.2.34) emulators/virtualbox-ose
virtualbox-ose-NGM_ETHER_GET_IFNAME-5.2.34.diff (text/plain), 32.90 KB, created by
Landon J. Fuller
on 2019-12-14 19:25:57 UTC
(
hide
)
Description:
Patch for current (5.2.34) emulators/virtualbox-ose
Filename:
MIME Type:
Creator:
Landon J. Fuller
Created:
2019-12-14 19:25:57 UTC
Size:
32.90 KB
patch
obsolete
>diff --git a/emulators/virtualbox-ose/files/patch-src-VBox-HostDrivers-VBoxNetFlt-freebsd-VBoxNetFlt-freebsd.c b/emulators/virtualbox-ose/files/patch-src-VBox-HostDrivers-VBoxNetFlt-freebsd-VBoxNetFlt-freebsd.c >index 4c0df373227c..555073a3b971 100644 >--- a/emulators/virtualbox-ose/files/patch-src-VBox-HostDrivers-VBoxNetFlt-freebsd-VBoxNetFlt-freebsd.c >+++ b/emulators/virtualbox-ose/files/patch-src-VBox-HostDrivers-VBoxNetFlt-freebsd-VBoxNetFlt-freebsd.c >@@ -1,75 +1,345 @@ >---- src/VBox/HostDrivers/VBoxNetFlt/freebsd/VBoxNetFlt-freebsd.c.orig 2018-10-15 14:30:58 UTC >-+++ src/VBox/HostDrivers/VBoxNetFlt/freebsd/VBoxNetFlt-freebsd.c >-@@ -52,6 +52,7 @@ >+--- src/VBox/HostDrivers/VBoxNetFlt/freebsd/VBoxNetFlt-freebsd.c 2019-12-14 12:05:06.742420000 -0700 >++++ src/VBox/HostDrivers/VBoxNetFlt/freebsd/VBoxNetFlt-freebsd.c 2019-12-14 12:05:06.742420000 -0700 >+@@ -52,10 +52,12 @@ > #include <net/if_dl.h> > #include <net/if_types.h> > #include <net/ethernet.h> > +#include <net/if_vlan_var.h> > > #include <netgraph/ng_message.h> > #include <netgraph/netgraph.h> >-@@ -73,6 +74,7 @@ >+ #include <netgraph/ng_parse.h> >++#include <netgraph/ng_ether.h> >+ >+ #define LOG_GROUP LOG_GROUP_NET_FLT_DRV >+ #include <VBox/version.h> >+@@ -73,6 +75,9 @@ > > #define VBOXNETFLT_OS_SPECFIC 1 > #include "../VBoxNetFltInternal.h" > +#include "freebsd/the-freebsd-kernel.h" >++ >++MALLOC_DEFINE(M_VBOXNETFLT, "vboxnetflt", "VBoxNetFlt netgraph node"); > > static int vboxnetflt_modevent(struct module *, int, void *); > static ng_constructor_t ng_vboxnetflt_constructor; >-@@ -370,7 +372,11 @@ static int ng_vboxnetflt_rcvdata(hook_p hook, item_p i >+@@ -93,6 +98,9 @@ >+ /** Output netgraph hook name */ >+ #define NG_VBOXNETFLT_HOOK_OUT "output" >+ >++/** Maximum time (in seconds) to wait for ng_ether replies */ >++#define NG_VBOXNETFLT_RPLY_TIMEOUT (10*hz) >++ >+ /** mbuf tag identifier */ >+ #define MTAG_VBOX 0x56424f58 >+ /** mbuf packet tag */ >+@@ -115,6 +123,18 @@ >+ #endif /* !defined(__FreeBSD_version) || __FreeBSD_version < 800500 */ >+ >+ /* >++ * Context used to track synchronous netgraph request state. >++ */ >++struct vboxnetflt_sreq { >++ uint32_t typecookie; /**< Matched type cookie. */ >++ uint32_t cmd; /**< Matched command. */ >++ uint32_t token; /**< Matched token. */ >++ struct ng_mesg *resp; /**< On completion, a copy of the reply */ >++ >++ STAILQ_ENTRY(vboxnetflt_sreq) sr_link; >++}; >++ >++/* >+ * Netgraph command list, we don't support any >+ * additional commands. >+ */ >+@@ -303,24 +323,225 @@ >+ } >+ >+ /** >+- * Netgraph message processing for node specific messages. >+- * We don't accept any special messages so this is not used. >++ * Enqueue and return a reference to a new synchronous response handler, or >++ * NULL if allocation fails. >+ */ >++static struct vboxnetflt_sreq * >++vboxnetflt_sreq_enqueue(PVBOXNETFLTINS pThis, uint32_t cookie, uint32_t cmd, >++ uint32_t token) >++{ >++ struct vboxnetflt_sreq *sreq; >++ >++ mtx_assert(&pThis->u.s.sreq_lck, MA_OWNED); >++ >++ sreq = malloc(sizeof(*sreq), M_VBOXNETFLT, M_NOWAIT|M_ZERO); >++ if (sreq == NULL) >++ return (NULL); >++ >++ sreq->typecookie = cookie; >++ sreq->cmd = cmd; >++ sreq->token = token; >++ >++ STAILQ_INSERT_TAIL(&pThis->u.s.sreq_list, sreq, sr_link); >++ >++ return (sreq); >++} >++ >++/** >++ * Dequeue and return the first synchronous responder handler capable of >++ * handling the given message, or NULL if no handler is found. >++ * >++ * The caller assumes ownership of any returned handler. >++ */ >++static struct vboxnetflt_sreq * >++vboxnetflt_sreq_dequeue(PVBOXNETFLTINS pThis, struct ng_mesg *msg) >++{ >++ struct vboxnetflt_sreq *sreq, *sr_next; >++ int error; >++ >++ mtx_assert(&pThis->u.s.sreq_lck, MA_OWNED); >++ >++ if (!(msg->header.flags & NGF_RESP)) >++ return (EINVAL); >++ >++ STAILQ_FOREACH_SAFE(sreq, &pThis->u.s.sreq_list, sr_link, sr_next) >++ { >++ if (sreq->typecookie != msg->header.typecookie) >++ continue; >++ >++ if (sreq->cmd != msg->header.cmd) >++ continue; >++ >++ if (sreq->token != msg->header.token) >++ continue; >++ >++ STAILQ_REMOVE(&pThis->u.s.sreq_list, sreq, vboxnetflt_sreq, >++ sr_link); >++ >++ return (sreq); >++ } >++ >++ return (NULL); >++} >++ >++/** >++ * Wait for completion of a previously enqueued response handler. On >++ * return, the handler will no longer be enqueued, and the caller is >++ * responsible for deallocating it via vboxnetflt_sreq_free(). >++ */ >++static int >++vboxnetflt_sreq_wait_enqueued(PVBOXNETFLTINS pThis, >++ struct vboxnetflt_sreq *sreq, int timo) >++{ >++ int error; >++ >++ mtx_assert(&pThis->u.s.sreq_lck, MA_OWNED); >++ >++ while (sreq->resp == NULL) { >++ error = mtx_sleep(sreq, &pThis->u.s.sreq_lck, 0, "vboxnetflt_sreq", >++ timo); >++ >++ if (error) >++ break; >++ } >++ >++ /* If our request was completed, we don't care if mtx_sleep() >++ * timed out or otherwise failed */ >++ if (sreq->resp != NULL) >++ return (0); >++ >++ /* >++ * Otherwise, we stopped waiting before a response has been >++ * received, and we're responsible for removing this handler from >++ * the queue. >++ * >++ * If the expected message is delivered later, it will be dropped as >++ * spurious if it fails to match against another registered handler. >++ */ >++ STAILQ_REMOVE(&pThis->u.s.sreq_list, sreq, vboxnetflt_sreq, >++ sr_link); >++ >++ return (ETIMEDOUT); >++} >++ >++/** >++ * Free the given synchronous request handler and all associated >++ * state. >++ * >++ * The handler must be dequeued and owned by the caller. >++ */ >++static void >++vboxnetflt_sreq_free(struct vboxnetflt_sreq *sreq) >++{ >++ if (sreq->resp != NULL) >++ NG_FREE_MSG(sreq->resp); >++ >++ free(sreq, M_VBOXNETFLT); >++} >++ >++ >++ >++/** >++ * Send a simple control message with no additional bytes, blocking until >++ * a response is received or a timeout is reached. >++ * >++ * On success, the caller assumes ownership of the response message, and >++ * is responsible for deallocating it via NG_FREE_MSG(). >++ */ >++int >++vboxnetflt_sreq_simple_rpc(PVBOXNETFLTINS pThis, node_p node, ng_ID_t peer, >++ uint32_t cookie, uint32_t cmd, struct ng_mesg **resp) >++{ >++ struct ng_mesg *msg; >++ struct vboxnetflt_sreq *sr; >++ int error; >++ >++ NG_MKMESSAGE(msg, cookie, cmd, 0, M_NOWAIT); >++ if (msg == NULL) >++ return (ENOMEM); >++ >++ NG_SEND_MSG_ID(error, node, msg, peer, 0); >++ if (error) { >++ LogRel(("VBoxNetFlt: error sending netgraph message (cookie=%u, " >++ "cmd=%u) to [%x]\n", cookie, cmd, peer)); >++ return (error); >++ } >++ >++ mtx_lock(&pThis->u.s.sreq_lck); >++ >++ /* Enqueue our response handler and wait for completion */ >++ if ((sr = vboxnetflt_sreq_enqueue(pThis, cookie, cmd, 0)) == NULL) { >++ mtx_unlock(&pThis->u.s.sreq_lck); >++ return (ENOMEM); >++ } >++ >++ error = vboxnetflt_sreq_wait_enqueued(pThis, sr, >++ NG_VBOXNETFLT_RPLY_TIMEOUT); >++ >++ mtx_unlock(&pThis->u.s.sreq_lck); >++ >++ if (error) >++ goto cleanup; >++ >++ /* Transfer ownership of the message to the caller */ >++ *resp = sr->resp; >++ sr->resp = NULL; >++ >++ /* fallthrough to common cleanup */ >++ >++cleanup: >++ vboxnetflt_sreq_free(sr); >++ return (error); >++} >++ >++/** >++ * Netgraph control message processing >++ */ >+ static int ng_vboxnetflt_rcvmsg(node_p node, item_p item, hook_p lasthook) >+ { >+ PVBOXNETFLTINS pThis = NG_NODE_PRIVATE(node); >+ struct ng_mesg *msg; >+- int error = 0; >++ struct vboxnetflt_sreq *sreq; >++ int error; >+ >+ NGI_GET_MSG(item, msg); >+- if (msg->header.typecookie != NGM_VBOXNETFLT_COOKIE) >+- return (EINVAL); >+ >+- switch (msg->header.cmd) >++ /* We don't accept any control messages of our own; drop anything that is >++ * not a message response. */ >++ if (!(msg->header.flags & NGF_RESP)) >+ { >+- default: >+- error = EINVAL; >++ error = EINVAL; >++ goto failed; >+ } >++ >++ /* Deqeueue the first matching response handler, if any */ >++ mtx_lock(&pThis->u.s.sreq_lck); >++ if ((sreq = vboxnetflt_sreq_dequeue(pThis, msg)) == NULL) >++ { >++ mtx_unlock(&pThis->u.s.sreq_lck); >++ >++ LogRel(("VBoxNetFlt: dropping spurious netgraph reply (typecookie=%u, " >++ "cmd=%u, token=%u)\n", msg->header.typecookie, msg->header.cmd, >++ msg->header.token)); >++ >++ error = EINVAL; >++ goto failed; >++ } >++ >++ /* Populate response data, transfering ownership of the message >++ * to the response handler. */ >++ KASSERT(sreq->resp == NULL, ("request already completed")); >++ sreq->resp = msg; >++ msg = NULL; >++ >++ /* Wake up any waiting threads */ >++ mtx_unlock(&pThis->u.s.sreq_lck); >++ wakeup(sreq); >++ >++ return (0); >++ >++failed: >++ NG_FREE_MSG(msg); >+ return (error); >+ } >+ >+@@ -370,7 +591,11 @@ > mtx_lock_spin(&pThis->u.s.inq.ifq_mtx); > _IF_ENQUEUE(&pThis->u.s.inq, m); > mtx_unlock_spin(&pThis->u.s.inq.ifq_mtx); > +#if __FreeBSD_version >= 1100100 > + taskqueue_enqueue(taskqueue_fast, &pThis->u.s.tskin); > +#else > taskqueue_enqueue_fast(taskqueue_fast, &pThis->u.s.tskin); > +#endif > } > /* > * Handle mbufs on the outgoing hook, frames going to the interface >-@@ -388,7 +394,11 @@ static int ng_vboxnetflt_rcvdata(hook_p hook, item_p i >+@@ -388,7 +613,11 @@ > mtx_lock_spin(&pThis->u.s.outq.ifq_mtx); > _IF_ENQUEUE(&pThis->u.s.outq, m); > mtx_unlock_spin(&pThis->u.s.outq.ifq_mtx); > +#if __FreeBSD_version >= 1100100 > + taskqueue_enqueue(taskqueue_fast, &pThis->u.s.tskout); > +#else > taskqueue_enqueue_fast(taskqueue_fast, &pThis->u.s.tskout); > +#endif > } > else > { >-@@ -428,6 +438,8 @@ static void vboxNetFltFreeBSDinput(void *arg, int pend >+@@ -428,6 +657,8 @@ > struct ifnet *ifp = pThis->u.s.ifp; > unsigned int cSegs = 0; > bool fDropIt = false, fActive; > + bool is_vl_tagged = false; > + uint16_t vl_tag; > PINTNETSG pSG; > > VBOXCURVNET_SET(ifp->if_vnet); >-@@ -440,6 +452,19 @@ static void vboxNetFltFreeBSDinput(void *arg, int pend >+@@ -440,6 +671,19 @@ > if (m == NULL) > break; > > + /* Prepend a VLAN header for consumption by the virtual switch */ > + if (m->m_flags & M_VLANTAG) { > + vl_tag = m->m_pkthdr.ether_vtag; > + is_vl_tagged = true; > + > + m = ether_vlanencap(m, m->m_pkthdr.ether_vtag); > + if (m == NULL) { > + printf("vboxflt: unable to prepend VLAN header\n"); > + break; > + } > + m->m_flags &= ~M_VLANTAG; > + } > + > for (m0 = m; m0 != NULL; m0 = m0->m_next) > if (m0->m_len > 0) > cSegs++; >-@@ -454,6 +479,27 @@ static void vboxNetFltFreeBSDinput(void *arg, int pend >+@@ -454,6 +698,27 @@ > vboxNetFltFreeBSDMBufToSG(pThis, m, pSG, cSegs, 0); > fDropIt = pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, INTNETTRUNKDIR_WIRE); > RTMemTmpFree(pSG); >@@ -97,175 +367,602 @@ > if (fDropIt) > m_freem(m); > else >-@@ -513,6 +559,7 @@ static void vboxNetFltFreeBSDoutput(void *arg, int pen >+@@ -513,6 +778,7 @@ > */ > int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst) > { > + IPRT_FREEBSD_SAVE_EFL_AC(); > NOREF(pvIfData); > > void (*input_f)(struct ifnet *, struct mbuf *); >-@@ -529,10 +576,16 @@ int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *p >+@@ -529,10 +795,16 @@ > { > m = vboxNetFltFreeBSDSGMBufFromSG(pThis, pSG); > if (m == NULL) > + { > + IPRT_FREEBSD_RESTORE_EFL_AC(); > return VERR_NO_MEMORY; > + } > m = m_pullup(m, ETHER_HDR_LEN); > if (m == NULL) > + { > + IPRT_FREEBSD_RESTORE_EFL_AC(); > return VERR_NO_MEMORY; > + } > > m->m_flags |= M_PKTHDR; > ether_output_frame(ifp, m); >-@@ -542,10 +595,16 @@ int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *p >+@@ -542,10 +814,16 @@ > { > m = vboxNetFltFreeBSDSGMBufFromSG(pThis, pSG); > if (m == NULL) > + { > + IPRT_FREEBSD_RESTORE_EFL_AC(); > return VERR_NO_MEMORY; > + } > m = m_pullup(m, ETHER_HDR_LEN); > if (m == NULL) > + { > + IPRT_FREEBSD_RESTORE_EFL_AC(); > return VERR_NO_MEMORY; > + } > /* > * Delivering packets to the host will be captured by the > * input hook. Tag the packet with a mbuf tag so that we >-@@ -556,6 +615,7 @@ int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *p >+@@ -556,6 +834,7 @@ > if (mtag == NULL) > { > m_freem(m); > + IPRT_FREEBSD_RESTORE_EFL_AC(); > return VERR_NO_MEMORY; > } > >-@@ -566,6 +626,7 @@ int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *p >+@@ -566,6 +845,7 @@ > ifp->if_input(ifp, m); > } > VBOXCURVNET_RESTORE(); > + IPRT_FREEBSD_RESTORE_EFL_AC(); > return VINF_SUCCESS; > } > >-@@ -578,6 +639,7 @@ static bool vboxNetFltFreeBsdIsPromiscuous(PVBOXNETFLT >+@@ -576,9 +856,43 @@ >+ return (pThis->u.s.flags & IFF_PROMISC) ? true : false; >+ } > >++static bool vboxNetFltFreeBSDNodeName(const char *ifname, char *output, size_t len) >++{ >++ size_t i; >++ >++ /* rewrite all characters in netgraph's reserved set '[]:.' to '_' */ >++ for (i = 0; i < len; i++) >++ { >++ char c = ifname[i]; >++ >++ switch (c) >++ { >++ case '[': >++ case ']': >++ case ':': >++ case '.': >++ output[i] = '_'; >++ break; >++ default: >++ output[i] = c; >++ break; >++ } >++ >++ if (c == '\0') >++ break; >++ } >++ >++ if (ifname[i] != '\0') >++ return false; >++ >++ return true; >++} >++ > int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext) > { > + IPRT_FREEBSD_SAVE_EFL_AC(); > char nam[NG_NODESIZ]; >++ char nam_if[IFNAMSIZ]; > struct ifnet *ifp; > node_p node; >-@@ -586,7 +648,10 @@ int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, voi >+ >+@@ -586,11 +900,27 @@ > NOREF(pvContext); > ifp = ifunit(pThis->szName); > if (ifp == NULL) > + { >++ VBOXCURVNET_RESTORE(); > + IPRT_FREEBSD_RESTORE_EFL_AC(); > return VERR_INTNET_FLT_IF_NOT_FOUND; > + } > >++ /* Rewrite netgraph-reserved characters in the interface name to produce >++ * our node's name suffix */ >++ if (!vboxNetFltFreeBSDNodeName(pThis->szName, nam_if, sizeof(nam_if))) >++ { >++ VBOXCURVNET_RESTORE(); >++ IPRT_FREEBSD_RESTORE_EFL_AC(); >++ return VERR_INTNET_FLT_IF_NOT_FOUND; >++ } >++ > /* Create a new netgraph node for this instance */ >- if (ng_make_node_common(&ng_vboxnetflt_typestruct, &node) != 0) >-@@ -630,12 +695,14 @@ int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, voi >+- if (ng_make_node_common(&ng_vboxnetflt_typestruct, &node) != 0) >++ if (ng_make_node_common(&ng_vboxnetflt_typestruct, &node) != 0) { >++ VBOXCURVNET_RESTORE(); >++ IPRT_FREEBSD_RESTORE_EFL_AC(); >+ return VERR_INTERNAL_ERROR; >++ } >+ >+ RTSpinlockAcquire(pThis->hSpinlock); >+ >+@@ -599,6 +929,10 @@ >+ bcopy(IF_LLADDR(ifp), &pThis->u.s.MacAddr, ETHER_ADDR_LEN); >+ ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false); >+ >++ /* Initialize synchronous request state */ >++ mtx_init(&pThis->u.s.sreq_lck, "vboxnetflt ng_recv", NULL, MTX_DEF|MTX_NEW); >++ STAILQ_INIT(&pThis->u.s.sreq_list); >++ >+ /* Initialize deferred input queue */ >+ bzero(&pThis->u.s.inq, sizeof(struct ifqueue)); >+ mtx_init(&pThis->u.s.inq.ifq_mtx, "vboxnetflt inq", NULL, MTX_SPIN); >+@@ -614,7 +948,7 @@ >+ NG_NODE_SET_PRIVATE(node, pThis); >+ >+ /* Attempt to name it vboxnetflt_<ifname> */ >+- snprintf(nam, NG_NODESIZ, "vboxnetflt_%s", pThis->szName); >++ snprintf(nam, NG_NODESIZ, "vboxnetflt_%s", nam_if); >+ ng_name_node(node, nam); >+ >+ /* Report MAC address, promiscuous mode and GSO capabilities. */ >+@@ -630,12 +964,14 @@ > vboxNetFltRelease(pThis, true /*fBusy*/); > } > VBOXCURVNET_RESTORE(); > + IPRT_FREEBSD_RESTORE_EFL_AC(); > > return VINF_SUCCESS; > } > > bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis) > { > + IPRT_FREEBSD_SAVE_EFL_AC(); > struct ifnet *ifp, *ifp0; > > ifp = ASMAtomicUoReadPtrT(&pThis->u.s.ifp, struct ifnet *); >-@@ -652,6 +719,7 @@ bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThi >+@@ -652,6 +988,7 @@ > pThis->u.s.node = NULL; > } > VBOXCURVNET_RESTORE(); > + IPRT_FREEBSD_RESTORE_EFL_AC(); > > if (ifp0 != NULL) > { >-@@ -664,6 +732,7 @@ bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThi >+@@ -664,6 +1001,7 @@ > > void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis) > { > + IPRT_FREEBSD_SAVE_EFL_AC(); > > taskqueue_drain(taskqueue_fast, &pThis->u.s.tskin); > taskqueue_drain(taskqueue_fast, &pThis->u.s.tskout); >-@@ -676,6 +745,7 @@ void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis) >+@@ -676,106 +1014,345 @@ > ng_rmnode_self(pThis->u.s.node); > VBOXCURVNET_RESTORE(); > pThis->u.s.node = NULL; >++ >++ /* Must be done *after* the node has been destroyed, as any incoming >++ * netgraph replies will need to acquire this lock. */ >++ mtx_destroy(&pThis->u.s.sreq_lck); >++ >++ /* Not possible unless a state management bug in VBoxNetFlt triggers >++ * instance deletion while also blocked in another thread on an active >++ * call into our instance */ >++ KASSERT(STAILQ_EMPTY(&pThis->u.s.sreq_list), >++ ("incomplete synchronous netgraph requests")); >++ > + IPRT_FREEBSD_RESTORE_EFL_AC(); > } > > int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis) >-@@ -689,6 +759,7 @@ int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis) >+ { >+- >+ pThis->u.s.ifp = NULL; >+ pThis->u.s.flags = 0; >+ pThis->u.s.node = NULL; >+ return VINF_SUCCESS; >+ } > >- void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive) >+-void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive) >++/** >++ * Fetch the list of netgraph nodes. >++ * >++ * On success, the caller assumes ownership of the list of node names, and >++ * is responsible for deallocating it via vboxNetFltFreeBSDFreeNodeList(). >++ */ >++int vboxNetFltFreeBSDListNodes(PVBOXNETFLTINS pThis, node_p node, >++ struct namelist **names) > { >-+ IPRT_FREEBSD_SAVE_EFL_AC(); >- struct ifnet *ifp; >- struct ifreq ifreq; >+- struct ifnet *ifp; >+- struct ifreq ifreq; >++ struct ng_mesg *msg; >++ struct namelist *nl; >++ size_t len, ni_len, ni_count; > int error; >-@@ -722,7 +793,10 @@ void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, b >- NG_MKMESSAGE(msg, NGM_GENERIC_COOKIE, NGM_CONNECT, >- sizeof(struct ngm_connect), M_NOWAIT); >- if (msg == NULL) >-+ { >-+ IPRT_FREEBSD_RESTORE_EFL_AC(); >- return; >+- node_p node; >++ >++ /* Fetch the node list */ >++ error = vboxnetflt_sreq_simple_rpc(pThis, node, NG_NODE_ID(node), >++ NGM_GENERIC_COOKIE, NGM_LISTNODES, &msg); >++ if (error) >++ { >++ LogRel(("VBoxNetFlt: NGM_LISTNODES failed (%d)\n", error)); >++ return error; >++ } >++ >++ /* Must be at least large enough for the fixed entry count. */ >++ nl = (struct namelist *)msg->data; >++ if (msg->header.arglen < sizeof(*nl)) { >++ error = ENXIO; >++ goto cleanup; >++ } >++ >++ /* Must be large enough for the header + (numnames * sizeof(nodeinfo)) */ >++ ni_count = nl->numnames; >++ ni_len = msg->header.arglen - sizeof(*nl); >++ >++ if (SIZE_MAX / sizeof(nl->nodeinfo[0]) < ni_count) { >++ /* Would overflow */ >++ error = ENXIO; >++ goto cleanup; >++ } >++ >++ if (ni_len < (sizeof(nl->nodeinfo[0]) * ni_count)) { >++ error = ENXIO; >++ goto cleanup; >++ } >++ >++ /* Copy out the result */ >++ nl = malloc(msg->header.arglen, M_VBOXNETFLT, M_NOWAIT); >++ if (nl == NULL) { >++ >++ error = ENOMEM; >++ goto cleanup; >++ } >++ memcpy(nl, msg->data, msg->header.arglen); >++ >++ *names = nl; >++ error = 0; >++ >++ /* fallthrough */ >++ >++cleanup: >++ NG_FREE_MSG(msg); >++ return (error); >++} >++ >++int vboxNetFltFreeBSDFreeNodeList(struct namelist *names) >++{ >++ free(names, M_VBOXNETFLT); >++} >++ >++/** >++ * Fetch the interface name from an ng_ether node. >++ */ >++int vboxNetFltFreeBSDEtherGetIfName(PVBOXNETFLTINS pThis, node_p node, >++ ng_ID_t ether_node, char *buf, size_t buflen) >++{ >+ struct ng_mesg *msg; >++ char *ifname; >++ size_t namelen; >++ int error; >++ >++ /* Fetch the interface name */ >++ error = vboxnetflt_sreq_simple_rpc(pThis, node, ether_node, >++ NGM_ETHER_COOKIE, NGM_ETHER_GET_IFNAME, &msg); >++ if (error) >++ { >++ /* May fail for innocuous reasons; e.g. if the node disappeared while >++ * iterating the node list, etc. */ >++ Log(("VBoxNetFlt: NGM_ETHER_GET_IFNAME failed (%d)\n", error)); >++ return (error); >++ } >++ >++ ifname = (const char *)msg->data; >++ namelen = strnlen(ifname, msg->header.arglen); >++ >++ /* Must be NUL-terminated */ >++ if (namelen == msg->header.arglen) { >++ NG_FREE_MSG(msg); >++ return (ENXIO); >++ } >++ >++ /* Copy out the result */ >++ if (buflen < namelen+1) { >++ NG_FREE_MSG(msg); >++ return (ENOMEM); >++ } >++ >++ strlcpy(buf, ifname, buflen); >++ NG_FREE_MSG(msg); >++ >++ return (0); >++} >++ >++/** >++ * Find the ng_ether node correspnding to our target interface. >++ */ >++static int vboxNetFltFreeBSDFindEtherNode(PVBOXNETFLTINS pThis, node_p node, >++ ng_ID_t *id) >++{ >++ struct namelist *nl; >++ int error; >++ >++ /* Fetch the full node list */ >++ if ((error = vboxNetFltFreeBSDListNodes(pThis, node, &nl))) { >++ LogRel(("VBoxNetFlt: failed to fetch node list (%d)\n", error)); >++ return (error); >++ } >++ >++ /* Look for a matching NG_ETHER node */ >++ for (uint32_t i = 0; i < nl->numnames; i++) { >++ struct nodeinfo *ni; >++ char ifname[IF_NAMESIZE]; >++ >++ ni = &nl->nodeinfo[i]; >++ >++ /* We only care about ng_ether nodes */ >++ if (strcmp(ni->type, NG_ETHER_NODE_TYPE) != 0) >++ continue; >++ >++ /* Fetch the interface name */ >++ error = vboxNetFltFreeBSDEtherGetIfName(pThis, node, ni->id, ifname, >++ sizeof(ifname)); >++ if (error) { >++ Log(("VBoxNetFlt: failed to fetch interface name (%d)\n", >++ error)); >++ continue; > + } >- con = (struct ngm_connect *)msg->data; >- snprintf(con->path, NG_PATHSIZ, "vboxnetflt_%s:", ifp->if_xname); >- strlcpy(con->ourhook, "lower", NG_HOOKSIZ); >-@@ -736,7 +810,10 @@ void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, b >- NG_MKMESSAGE(msg, NGM_GENERIC_COOKIE, NGM_CONNECT, >- sizeof(struct ngm_connect), M_NOWAIT); >- if (msg == NULL) >-+ { >-+ IPRT_FREEBSD_RESTORE_EFL_AC(); >- return; >-+ } >- con = (struct ngm_connect *)msg->data; >- snprintf(con->path, NG_PATHSIZ, "vboxnetflt_%s:", >- ifp->if_xname); >-@@ -759,7 +836,10 @@ void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, b >- NG_MKMESSAGE(msg, NGM_GENERIC_COOKIE, NGM_RMHOOK, >- sizeof(struct ngm_rmhook), M_NOWAIT); >- if (msg == NULL) >-+ { >-+ IPRT_FREEBSD_RESTORE_EFL_AC(); >- return; >-+ } >- rm = (struct ngm_rmhook *)msg->data; >- strlcpy(rm->ourhook, "input", NG_HOOKSIZ); >- NG_SEND_MSG_PATH(error, node, msg, path, 0); >-@@ -770,12 +850,16 @@ void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, b >- NG_MKMESSAGE(msg, NGM_GENERIC_COOKIE, NGM_RMHOOK, >- sizeof(struct ngm_rmhook), M_NOWAIT); >- if (msg == NULL) >-+ { >-+ IPRT_FREEBSD_RESTORE_EFL_AC(); >- return; >-+ } >- rm = (struct ngm_rmhook *)msg->data; >- strlcpy(rm->ourhook, "output", NG_HOOKSIZ); >- NG_SEND_MSG_PATH(error, node, msg, path, 0); >++ >++ /* Skip non-matching interfaces */ >++ if (strcmp(ifname, pThis->szName) != 0) >++ continue; >++ >++ *id = ni->id; >++ vboxNetFltFreeBSDFreeNodeList(nl); >++ return (0); >++ } >++ >++ vboxNetFltFreeBSDFreeNodeList(nl); >++ return (ENOENT); >++} >++ >++/** >++ * Send an NGM_CONNECT message. >++ */ >++static int vboxNetFltFreeBSDConnectHook(PVBOXNETFLTINS pThis, node_p node, >++ ng_ID_t peer, const char *ourhook, const char *peerhook) >++{ >++ struct ng_mesg *msg; >+ struct ngm_connect *con; >++ int error; >++ >++ if (strlen(ourhook) > NG_HOOKSIZ || strlen(peerhook) > NG_HOOKSIZ) >++ return (EINVAL); >++ >++ NG_MKMESSAGE(msg, NGM_GENERIC_COOKIE, NGM_CONNECT, >++ sizeof(struct ngm_connect), M_NOWAIT); >++ if (msg == NULL) >++ return (ENOMEM); >++ >++ con = (struct ngm_connect *)msg->data; >++ snprintf(con->path, NG_PATHSIZ, "[%x]:", NG_NODE_ID(node)); >++ strlcpy(con->ourhook, ourhook, NG_HOOKSIZ); >++ strlcpy(con->peerhook, peerhook, NG_HOOKSIZ); >++ >++ NG_SEND_MSG_ID(error, node, msg, peer, 0); >++ return (error); >++} >++ >++/** >++ * Send an NGM_RMHOOK message. >++ */ >++static int vboxNetFltFreeBSDRmHook(PVBOXNETFLTINS pThis, node_p node, >++ const char *ourhook) >++{ >++ struct ng_mesg *msg; >+ struct ngm_rmhook *rm; >+- char path[NG_PATHSIZ]; >++ int error; >+ >+- Log(("%s: fActive:%d\n", __func__, fActive)); >++ if (strlen(ourhook) > NG_HOOKSIZ) >++ return (EINVAL); >+ >+- ifp = ASMAtomicUoReadPtrT(&pThis->u.s.ifp, struct ifnet *); >+- VBOXCURVNET_SET(ifp->if_vnet); >+- node = ASMAtomicUoReadPtrT(&pThis->u.s.node, node_p); >++ NG_MKMESSAGE(msg, NGM_GENERIC_COOKIE, NGM_RMHOOK, >++ sizeof(struct ngm_rmhook), M_NOWAIT); >++ if (msg == NULL) >++ return (ENOMEM); >+ >+- memset(&ifreq, 0, sizeof(struct ifreq)); >+- /* Activate interface */ >+- if (fActive) >+- { >+- pThis->u.s.flags = ifp->if_flags; >+- ifpromisc(ifp, 1); >++ rm = (struct ngm_rmhook *)msg->data; >++ strlcpy(rm->ourhook, ourhook, NG_HOOKSIZ); >+ >+- /* ng_ether nodes are named after the interface name */ >+- snprintf(path, sizeof(path), "%s:", ifp->if_xname); >++ NG_SEND_MSG_PATH(error, node, msg, ".:", 0); >++ return (error); >++} >+ >+- /* >+- * Send a netgraph connect message to the ng_ether node >+- * assigned to the bridged interface. Connecting >+- * the hooks 'lower' (ng_ether) to out 'input'. >+- */ >+- NG_MKMESSAGE(msg, NGM_GENERIC_COOKIE, NGM_CONNECT, >+- sizeof(struct ngm_connect), M_NOWAIT); >+- if (msg == NULL) >+- return; >+- con = (struct ngm_connect *)msg->data; >+- snprintf(con->path, NG_PATHSIZ, "vboxnetflt_%s:", ifp->if_xname); >+- strlcpy(con->ourhook, "lower", NG_HOOKSIZ); >+- strlcpy(con->peerhook, "input", NG_HOOKSIZ); >+- NG_SEND_MSG_PATH(error, node, msg, path, 0); >++/** >++ * Called by vboxNetFltPortOsSetActive() to activate the filter driver instance. >++ */ >++static int vboxNetFltFreeBSDActivate(PVBOXNETFLTINS pThis, struct ifnet *ifp, >++ node_p node) >++{ >++ ng_ID_t ether_node; >++ int error; >+ >+- /* >+- * Do the same for the hooks 'upper' (ng_ether) and our >+- * 'output' hook. >+- */ >+- NG_MKMESSAGE(msg, NGM_GENERIC_COOKIE, NGM_CONNECT, >+- sizeof(struct ngm_connect), M_NOWAIT); >+- if (msg == NULL) >+- return; >+- con = (struct ngm_connect *)msg->data; >+- snprintf(con->path, NG_PATHSIZ, "vboxnetflt_%s:", >+- ifp->if_xname); >+- strlcpy(con->ourhook, "upper", sizeof(con->ourhook)); >+- strlcpy(con->peerhook, "output", sizeof(con->peerhook)); >+- NG_SEND_MSG_PATH(error, node, msg, path, 0); >++ pThis->u.s.flags = ifp->if_flags; >++ ifpromisc(ifp, 1); >++ >++ /* Locate the interface's corresponding NG_ETHER netgraph node */ >++ if ((error = vboxNetFltFreeBSDFindEtherNode(pThis, node, ðer_node))) { >++ LogRel(("VBoxNetFlt: Failed to locate netgraph node for %s: %d\n", >++ ifp->if_xname, error)); >++ return (error); >+ } >+- else >+- { >+- /* De-activate interface */ >+- pThis->u.s.flags = 0; >+- ifpromisc(ifp, 0); >+ >+- /* Disconnect msgs are addressed to ourself */ >+- snprintf(path, sizeof(path), "vboxnetflt_%s:", ifp->if_xname); >++ /* >++ * Send a netgraph connect message to the ng_ether node >++ * assigned to the bridged interface. Connecting >++ * the hooks 'lower' (ng_ether) to out 'input'. >++ */ >++ error = vboxNetFltFreeBSDConnectHook(pThis, node, ether_node, >++ NG_ETHER_HOOK_LOWER, NG_VBOXNETFLT_HOOK_IN); >++ if (error) { >++ LogRel(("VBoxNetFlt: Failed to connect netgraph input hook for %s:" >++ " %d\n", ifp->if_xname, error)); >++ return (error); >++ } >+ >+- /* >+- * Send a netgraph message to disconnect our 'input' hook >+- */ >+- NG_MKMESSAGE(msg, NGM_GENERIC_COOKIE, NGM_RMHOOK, >+- sizeof(struct ngm_rmhook), M_NOWAIT); >+- if (msg == NULL) >+- return; >+- rm = (struct ngm_rmhook *)msg->data; >+- strlcpy(rm->ourhook, "input", NG_HOOKSIZ); >+- NG_SEND_MSG_PATH(error, node, msg, path, 0); >++ /* >++ * Do the same for the hooks 'upper' (ng_ether) and our >++ * 'output' hook. >++ */ >++ error = vboxNetFltFreeBSDConnectHook(pThis, node, ether_node, >++ NG_ETHER_HOOK_UPPER, NG_VBOXNETFLT_HOOK_OUT); >++ if (error) { >++ LogRel(("VBoxNetFlt: Failed to connect netgraph output hook for %s:" >++ " %d\n", ifp->if_xname, error)); >++ return (error); >++ } >+ >+- /* >+- * Send a netgraph message to disconnect our 'output' hook >+- */ >+- NG_MKMESSAGE(msg, NGM_GENERIC_COOKIE, NGM_RMHOOK, >+- sizeof(struct ngm_rmhook), M_NOWAIT); >+- if (msg == NULL) >+- return; >+- rm = (struct ngm_rmhook *)msg->data; >+- strlcpy(rm->ourhook, "output", NG_HOOKSIZ); >+- NG_SEND_MSG_PATH(error, node, msg, path, 0); >++ return (0); >++} >++ >++/** >++ * Called by vboxNetFltPortOsSetActive() to deactivate the filter driver >++ * instance. >++ */ >++static int vboxNetFltFreeBSDDeactivate(PVBOXNETFLTINS pThis, struct ifnet *ifp, >++ node_p node) >++{ >++ int ng_error; >++ int error = 0; >++ >++ pThis->u.s.flags = 0; >++ ifpromisc(ifp, 0); >++ >++ /* >++ * Send a netgraph message to disconnect our input and output hooks; on >++ * failure, we still attempt to disconnect any remaining hooks. >++ */ >++ ng_error = vboxNetFltFreeBSDRmHook(pThis, node, NG_VBOXNETFLT_HOOK_IN); >++ if (ng_error) { >++ LogRel(("VBoxNetFlt: Failed to disconnect netgraph input hook for %s:" >++ " %d\n", ifp->if_xname, error)); >++ error = ng_error; > } >++ >++ ng_error = vboxNetFltFreeBSDRmHook(pThis, node, NG_VBOXNETFLT_HOOK_OUT); >++ if (ng_error) { >++ LogRel(("VBoxNetFlt: Failed to disconnect netgraph output hook for %s:" >++ " %d\n", ifp->if_xname, error)); >++ error = ng_error; >++ } >++ >++ return (error); >++} >++ >++void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive) >++{ >++ IPRT_FREEBSD_SAVE_EFL_AC(); >++ struct ifnet *ifp; >++ int error; >++ node_p node; >++ >++ Log(("%s: fActive:%d\n", __func__, fActive)); >++ >++ ifp = ASMAtomicUoReadPtrT(&pThis->u.s.ifp, struct ifnet *); >++ VBOXCURVNET_SET(ifp->if_vnet); >++ node = ASMAtomicUoReadPtrT(&pThis->u.s.node, node_p); >++ >++ if (fActive) >++ error = vboxNetFltFreeBSDActivate(pThis, ifp, node); >++ else >++ error = vboxNetFltFreeBSDDeactivate(pThis, ifp, node); >++ >++ if (error) >++ LogRel(("VBoxNetFlt: Failed to %s %s filter driver: %d\n", >++ (fActive ? "activate" : "deactivate"), ifp->if_xname, error)); >++ > VBOXCURVNET_RESTORE(); > + IPRT_FREEBSD_RESTORE_EFL_AC(); > } >diff --git a/emulators/virtualbox-ose/files/patch-src-VBox-HostDrivers-VBoxNetFlt-VBoxNetFltInternal.h b/emulators/virtualbox-ose/files/patch-src-VBox-HostDrivers-VBoxNetFlt-VBoxNetFltInternal.h >new file mode 100644 >index 000000000000..81dd868519c5 >--- /dev/null >+++ b/emulators/virtualbox-ose/files/patch-src-VBox-HostDrivers-VBoxNetFlt-VBoxNetFltInternal.h >@@ -0,0 +1,13 @@ >+--- src/VBox/HostDrivers/VBoxNetFlt/VBoxNetFltInternal.h.orig 2019-06-15 19:42:35 UTC >++++ src/VBox/HostDrivers/VBoxNetFlt/VBoxNetFltInternal.h >+@@ -240,6 +240,10 @@ typedef struct VBOXNETFLTINS >+ hook_p input; >+ /** Output hook */ >+ hook_p output; >++ /** Pending synchronous netgraph requests */ >++ STAILQ_HEAD(,vboxnetflt_sreq) sreq_list; >++ /** Synchronous request lock */ >++ struct mtx sreq_lck; >+ /** Original interface flags */ >+ unsigned int flags; >+ /** Input queue */
<b>You cannot view the attachment while viewing its details because your browser does not support IFRAMEs. <a href="attachment.cgi?id=209951">View the attachment on a separate page</a>.</b>
View Attachment As Diff
View Attachment As Raw
Flags:
grahamperrin
:
maintainer-approval?
(
vbox
)
Actions:
View
|
Diff
Attachments on
bug 242637
: 209951 |
209952