rx: Create AFS_ADAPT_PMTU and AFS_RXERRQ_ENV

Currently we have the ADAPT_PMTU define, which turns on functionality
in Linux to detect PMTU-related ICMP errors for Rx. However, this is
really turning on two separate pieces of functionality: the PMTU
processing, and the processing for ICMP errors in general.

So split this out into two defines: AFS_ADAPT_PMTU, and
AFS_RXERRQ_ENV. The former is for processing PMTU discovery, and the
latter is for processing ICMP errors. Both of these are left disabled
due to issues in the error processing. Although PMTU discovery is the
only functionality which makes use of ICMP errors, this will change in
the future.

Change-Id: Ia334c68ce5eb3fa01c01a8a1c52a0e0a2e41b2c0
Reviewed-on: http://gerrit.openafs.org/7925
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

diff --git a/acinclude.m4 b/acinclude.m4
index e67d220..7f69dad 100644 (file)
--- a/acinclude.m4
+++ b/acinclude.m4
@@ -1141,8 +1141,8 @@ setsockopt(0, SOL_IP, IP_RECVERR, &on, sizeof(on));],
        [ac_cv_setsockopt_iprecverr=no])])
 
 AS_IF([test "$ac_cv_setsockopt_iprecverr" = "yes"],
-      [AC_DEFINE(ADAPT_PMTU_RECVERR, 1,
-                [define if asynchronous socket errors can be received])])
+      [AC_DEFINE([HAVE_SETSOCKOPT_IP_RECVERR], [1],
+                [define if we can receive socket errors via IP_RECVERR])])
 
 PTHREAD_LIBS=error
 if test "x$MKAFS_OSTYPE" = OBSD; then
@@ -1213,10 +1213,6 @@ else
 fi
 AC_SUBST(USE_UNIX_SOCKETS)
 
-dnl if test "$ac_cv_setsockopt_iprecverr" = "yes"; then
-dnl    AC_DEFINE(ADAPT_PMTU, 1, [define if you want to decode icmp unreachable packets to discover path mtu])
-dnl fi
-
 if test "$enable_namei_fileserver" = "yes"; then
        AC_DEFINE(AFS_NAMEI_ENV, 1, [define if you want to want namei fileserver])
        VFSCK=""

diff --git a/src/config/param.linux26.h b/src/config/param.linux26.h
index 4e192b8..4a7b942 100644 (file)
--- a/src/config/param.linux26.h
+++ b/src/config/param.linux26.h
@@ -99,6 +99,15 @@
 #include <features.h>
 #endif
 
+#ifdef notyet
+#if defined(HAVE_LINUX_ERRQUEUE_H) && defined(HAVE_SETSOCKOPT_IP_RECVERR)
+# define AFS_RXERRQ_ENV
+#endif
+#ifdef AFS_RXERRQ_ENV
+# define AFS_ADAPT_PMTU
+#endif
+#endif
+
 #ifdef __GLIBC__
 #if (__GLIBC__ > 2) || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 3)
 #define USE_UCONTEXT

diff --git a/src/rx/LINUX/rx_knet.c b/src/rx/LINUX/rx_knet.c
index 942a072..fccdbc4 100644 (file)
--- a/src/rx/LINUX/rx_knet.c
+++ b/src/rx/LINUX/rx_knet.c
@@ -18,8 +18,15 @@
 
 #include <linux/version.h>
 #include "rx/rx_kcommon.h"
+#include "rx.h"
+#include "rx_atomic.h"
+#include "rx_globals.h"
+#include "rx_stats.h"
+#include "rx_peer.h"
+#include "rx_packet.h"
+#include "rx_internal.h"
 #include <asm/uaccess.h>
-#ifdef ADAPT_PMTU
+#ifdef AFS_RXERRQ_ENV
 #include <linux/errqueue.h>
 #include <linux/icmp.h>
 #endif
@@ -35,9 +42,8 @@ rxk_NewSocketHost(afs_uint32 ahost, short aport)
     struct socket *sockp;
     struct sockaddr_in myaddr;
     int code;
-#ifdef ADAPT_PMTU
+#ifdef AFS_ADAPT_PMTU
     int pmtu = IP_PMTUDISC_WANT;
-    int do_recverr = 1;
 #else
     int pmtu = IP_PMTUDISC_DONT;
 #endif
@@ -67,9 +73,12 @@ rxk_NewSocketHost(afs_uint32 ahost, short aport)
 
     kernel_setsockopt(sockp, SOL_IP, IP_MTU_DISCOVER, (char *)&pmtu,
                      sizeof(pmtu));
-#ifdef ADAPT_PMTU
-    kernel_setsockopt(sockp, SOL_IP, IP_RECVERR, (char *)&do_recverr,
-                      sizeof(do_recverr));
+#ifdef AFS_RXERRQ_ENV
+    {
+       int recverr = 1;
+       kernel_setsockopt(sockp, SOL_IP, IP_RECVERR, (char *)&recverr,
+                         sizeof(recverr));
+    }
 #endif
     return (osi_socket *)sockp;
 }
@@ -88,7 +97,7 @@ rxk_FreeSocket(struct socket *asocket)
     return 0;
 }
 
-#ifdef ADAPT_PMTU
+#ifdef AFS_RXERRQ_ENV
 void
 handle_socket_error(osi_socket so)
 {
@@ -130,12 +139,14 @@ handle_socket_error(osi_socket so)
 
     memcpy(&addr, offender, sizeof(addr));
 
+# ifdef AFS_ADAPT_PMTU
     if (err->ee_origin == SO_EE_ORIGIN_ICMP &&
        err->ee_type == ICMP_DEST_UNREACH &&
        err->ee_code == ICMP_FRAG_NEEDED) {
        rxi_SetPeerMtu(NULL, ntohl(addr.sin_addr.s_addr), ntohs(addr.sin_port),
                       err->ee_info);
     }
+# endif
     /* other DEST_UNREACH's and TIME_EXCEEDED should be dealt with too */
 
 out:
@@ -156,7 +167,7 @@ osi_NetSend(osi_socket sop, struct sockaddr_in *to, struct iovec *iovec,
 {
     struct msghdr msg;
     int code;
-#ifdef ADAPT_PMTU
+#ifdef AFS_RXERRQ_ENV
     int sockerr;
     int esize;
 
@@ -208,7 +219,7 @@ osi_NetReceive(osi_socket so, struct sockaddr_in *from, struct iovec *iov,
 {
     struct msghdr msg;
     int code;
-#ifdef ADAPT_PMTU
+#ifdef AFS_RXERRQ_ENV
     int sockerr;
     int esize;
 #endif
@@ -218,7 +229,7 @@ osi_NetReceive(osi_socket so, struct sockaddr_in *from, struct iovec *iov,
     if (iovcnt > RX_MAXWVECS + 2) {
        osi_Panic("Too many (%d) iovecs passed to osi_NetReceive\n", iovcnt);
     }
-#ifdef ADAPT_PMTU
+#ifdef AFS_RXERRQ_ENV
     while (1) {
        sockerr=0;
        esize = sizeof(sockerr);

diff --git a/src/rx/rx.c b/src/rx/rx.c
index 53db0a3..17ab869 100644 (file)
--- a/src/rx/rx.c
+++ b/src/rx/rx.c
@@ -6221,7 +6221,7 @@ rxi_CheckCall(struct rx_call *call)
      * number of seconds. */
     if (now > (call->lastReceiveTime + deadTime)) {
        if (call->state == RX_STATE_ACTIVE) {
-#ifdef ADAPT_PMTU
+#ifdef AFS_ADAPT_PMTU
 # if defined(KERNEL) && defined(AFS_SUN5_ENV)
            ire_t *ire;
 #  if defined(AFS_SUN510_ENV) && defined(GLOBAL_NETSTACKID)
@@ -6247,7 +6247,7 @@ rxi_CheckCall(struct rx_call *call)
            netstack_rele(ns);
 #  endif
 # endif
-#endif /* ADAPT_PMTU */
+#endif /* AFS_ADAPT_PMTU */
            cerror = RX_CALL_DEAD;
            goto mtuout;
        } else {

diff --git a/src/rx/rx_lwp.c b/src/rx/rx_lwp.c
index 229b1ec..0923642 100644 (file)
--- a/src/rx/rx_lwp.c
+++ b/src/rx/rx_lwp.c
@@ -410,7 +410,7 @@ rxi_Listen(osi_socket sock)
 int
 rxi_Recvmsg(osi_socket socket, struct msghdr *msg_p, int flags)
 {
-#if defined(HAVE_LINUX_ERRQUEUE_H) && defined(ADAPT_PMTU)
+#ifdef AFS_RXERRQ_ENV
     while((rxi_HandleSocketError(socket)) > 0)
        ;
 #endif
@@ -438,7 +438,7 @@ rxi_Sendmsg(osi_socket socket, struct msghdr *msg_p, int flags)
            }
            FD_SET(socket, sfds);
        }
-#if defined(HAVE_LINUX_ERRQUEUE_H) && defined(ADAPT_PMTU)
+#ifdef AFS_RXERRQ_ENV
        while((rxi_HandleSocketError(socket)) > 0)
          ;
 #endif

diff --git a/src/rx/rx_pthread.c b/src/rx/rx_pthread.c
index a8a8ef0..d76ed6d 100644 (file)
--- a/src/rx/rx_pthread.c
+++ b/src/rx/rx_pthread.c
@@ -391,7 +391,7 @@ int
 rxi_Recvmsg(osi_socket socket, struct msghdr *msg_p, int flags)
 {
     int ret;
-#if defined(HAVE_LINUX_ERRQUEUE_H) && defined(ADAPT_PMTU)
+#ifdef AFS_RXERRQ_ENV
     while((rxi_HandleSocketError(socket)) > 0)
       ;
 #endif

diff --git a/src/rx/rx_user.c b/src/rx/rx_user.c
index c55949e..a0427e4 100644 (file)
--- a/src/rx/rx_user.c
+++ b/src/rx/rx_user.c
@@ -36,10 +36,10 @@
 #define IPPORT_USERRESERVED 5000
 # endif
 
-#if defined(HAVE_LINUX_ERRQUEUE_H) && defined(ADAPT_PMTU)
+#if defined(AFS_LINUX22_ENV) && defined(AFS_RXERRQ_ENV)
 # include <linux/types.h>
 # include <linux/errqueue.h>
-# ifndef IP_MTU
+# if defined(AFS_ADAPT_PMTU) && !defined(IP_MTU)
 #  define IP_MTU 14
 # endif
 #endif
@@ -50,6 +50,7 @@
 #include "rx_stats.h"
 #include "rx_peer.h"
 #include "rx_packet.h"
+#include "rx_internal.h"
 
 #ifdef AFS_PTHREAD_ENV
 
@@ -94,9 +95,8 @@ rxi_GetHostUDPSocket(u_int ahost, u_short port)
     struct sockaddr_in taddr;
     char *name = "rxi_GetUDPSocket: ";
 #ifdef AFS_LINUX22_ENV
-# if defined(ADAPT_PMTU)
+# if defined(AFS_ADAPT_PMTU)
     int pmtu = IP_PMTUDISC_WANT;
-    int recverr = 1;
 # else
     int pmtu = IP_PMTUDISC_DONT;
 # endif
@@ -198,9 +198,12 @@ rxi_GetHostUDPSocket(u_int ahost, u_short port)
 
 #ifdef AFS_LINUX22_ENV
     setsockopt(socketFd, SOL_IP, IP_MTU_DISCOVER, &pmtu, sizeof(pmtu));
-#if defined(ADAPT_PMTU)
-    setsockopt(socketFd, SOL_IP, IP_RECVERR, &recverr, sizeof(recverr));
 #endif
+#ifdef AFS_RXERRQ_ENV
+    {
+       int recverr = 1;
+       setsockopt(socketFd, SOL_IP, IP_RECVERR, &recverr, sizeof(recverr));
+    }
 #endif
     if (rxi_Listen(socketFd) < 0) {
        goto error;
@@ -668,7 +671,7 @@ rxi_InitPeerParams(struct rx_peer *pp)
     afs_uint32 ppaddr;
     u_short rxmtu;
     int ix;
-#if defined(ADAPT_PMTU) && defined(IP_MTU)
+#ifdef AFS_ADAPT_PMTU
     int sock;
     struct sockaddr_in addr;
 #endif
@@ -715,7 +718,7 @@ rxi_InitPeerParams(struct rx_peer *pp)
        rx_rto_setPeerTimeoutSecs(pp, 3);
        pp->ifMTU = MIN(rx_MyMaxSendSize, RX_REMOTE_PACKET_SIZE);
     }
-#if defined(ADAPT_PMTU) && defined(IP_MTU)
+#ifdef AFS_ADAPT_PMTU
     sock=socket(PF_INET, SOCK_DGRAM, IPPROTO_UDP);
     if (sock != OSI_NULLSOCKET) {
         addr.sin_family = AF_INET;
@@ -773,12 +776,11 @@ rx_SetMaxMTU(int mtu)
     return 0;
 }
 
-#if defined(ADAPT_PMTU)
+#ifdef AFS_RXERRQ_ENV
 int
 rxi_HandleSocketError(int socket)
 {
-    int ret=0;
-#if defined(HAVE_LINUX_ERRQUEUE_H)
+    int ret = 0;
     struct msghdr msg;
     struct cmsghdr *cmsg;
     struct sock_extended_err *err;
@@ -813,14 +815,15 @@ rxi_HandleSocketError(int socket)
     ret = 1;
     err = (struct sock_extended_err *) CMSG_DATA(cmsg);
 
+# ifdef AFS_ADAPT_PMTU
     if (err->ee_errno == EMSGSIZE && err->ee_info >= 68) {
        rxi_SetPeerMtu(NULL, addr.sin_addr.s_addr, addr.sin_port,
                        err->ee_info - RX_IPUDP_SIZE);
     }
+# endif
     /* other DEST_UNREACH's and TIME_EXCEEDED should be dealt with too */
 
 out:
-#endif
     return ret;
 }
 #endif