allow custom MTU

From https://github.com/rootless-containers/slirp4netns/blob/4889f5299f407d7d7566c76a3b8b5f71c99b6db5/qemu_patches/0002-slirp-allow-custom-MTU.patch

Original commits:
* https://github.com/rootless-containers/slirp4netns/commit/ea630a7e945cf538184ff1b1b4bd7b8ddc01993e
* https://github.com/rootless-containers/slirp4netns/commit/1508a66c93c223555f08651592dde3d2d708b166
* https://github.com/rootless-containers/slirp4netns/commit/19f3f41df4066d6103e6f882500e24db7ea7d9e1
* https://github.com/rootless-containers/slirp4netns/commit/a11abedafcc627ef0657999e63b211b0f26d4c02
* https://github.com/rootless-containers/slirp4netns/commit/2adbd7c449944d3b837164c86eedd3dcabbba1a6

Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>

12 files changed, 46 insertions, 33 deletions

diff --git a/src/dhcpv6.c b/src/dhcpv6.c
index bc041d5..ce5f203 100644
--- a/src/dhcpv6.c
+++ b/src/dhcpv6.c
@@ -178,7 +178,7 @@ static void dhcpv6_info_request(Slirp *slirp, struct sockaddr_in6 *srcsas,
 
         *resp++ = OPTION_BOOTFILE_URL >> 8; /* option-code high byte */
         *resp++ = OPTION_BOOTFILE_URL; /* option-code low byte */
-        smaxlen = (uint8_t *)m->m_data + IF_MTU - (resp + 2);
+        smaxlen = (uint8_t *)m->m_data + slirp->if_mtu - (resp + 2);
         slen = snprintf((char *)resp + 2, smaxlen,
                         "tftp://[%02x%02x:%02x%02x:%02x%02x:%02x%02x:"
                         "%02x%02x:%02x%02x:%02x%02x:%02x%02x]/%s",
diff --git a/src/if.h b/src/if.h
index 3288298..7cf9d27 100644
--- a/src/if.h
+++ b/src/if.h
@@ -11,8 +11,12 @@
 #define IF_AUTOCOMP 0x04 /* Autodetect (default) */
 #define IF_NOCIDCOMP 0x08 /* CID compression */
 
-#define IF_MTU 1500
-#define IF_MRU 1500
+#define IF_MTU_DEFAULT 1500
+#define IF_MTU_MIN 68
+#define IF_MTU_MAX 65521
+#define IF_MRU_DEFAULT 1500
+#define IF_MRU_MIN 68
+#define IF_MRU_MAX 65521
 #define IF_COMP IF_AUTOCOMP /* Flags for compression */
 
 /* 2 for alignment, 14 for ethernet */
diff --git a/src/ip6_icmp.c b/src/ip6_icmp.c
index 8512686..28ec2be 100644
--- a/src/ip6_icmp.c
+++ b/src/ip6_icmp.c
@@ -94,8 +94,8 @@ void icmp6_send_error(struct mbuf *m, uint8_t type, uint8_t code)
     DEBUG_ARG("target = %s", addrstr);
 
     rip->ip_nh = IPPROTO_ICMPV6;
-    const int error_data_len =
-        MIN(m->m_len, IF_MTU - (sizeof(struct ip6) + ICMP6_ERROR_MINLEN));
+    const int error_data_len = MIN(
+        m->m_len, slirp->if_mtu - (sizeof(struct ip6) + ICMP6_ERROR_MINLEN));
     rip->ip_pl = htons(ICMP6_ERROR_MINLEN + error_data_len);
     t->m_len = sizeof(struct ip6) + ntohs(rip->ip_pl);
 
@@ -112,7 +112,7 @@ void icmp6_send_error(struct mbuf *m, uint8_t type, uint8_t code)
         ricmp->icmp6_err.unused = 0;
         break;
     case ICMP6_TOOBIG:
-        ricmp->icmp6_err.mtu = htonl(IF_MTU);
+        ricmp->icmp6_err.mtu = htonl(slirp->if_mtu);
         break;
     case ICMP6_PARAMPROB:
         /* TODO: Handle this case */
diff --git a/src/ip6_input.c b/src/ip6_input.c
index d9d2b7e..dfcbfd6 100644
--- a/src/ip6_input.c
+++ b/src/ip6_input.c
@@ -44,7 +44,7 @@ void ip6_input(struct mbuf *m)
         goto bad;
     }
 
-    if (ntohs(ip6->ip_pl) > IF_MTU) {
+    if (ntohs(ip6->ip_pl) > slirp->if_mtu) {
         icmp6_send_error(m, ICMP6_TOOBIG, 0);
         goto bad;
     }
diff --git a/src/ip_output.c b/src/ip_output.c
index f820359..22916a3 100644
--- a/src/ip_output.c
+++ b/src/ip_output.c
@@ -72,7 +72,7 @@ int ip_output(struct socket *so, struct mbuf *m0)
     /*
      * If small enough for interface, can just send directly.
      */
-    if ((uint16_t)ip->ip_len <= IF_MTU) {
+    if ((uint16_t)ip->ip_len <= slirp->if_mtu) {
         ip->ip_len = htons((uint16_t)ip->ip_len);
         ip->ip_off = htons((uint16_t)ip->ip_off);
         ip->ip_sum = 0;
@@ -91,7 +91,7 @@ int ip_output(struct socket *so, struct mbuf *m0)
         goto bad;
     }
 
-    len = (IF_MTU - hlen) & ~7; /* ip databytes per packet */
+    len = (slirp->if_mtu - hlen) & ~7; /* ip databytes per packet */
     if (len < 8) {
         error = -1;
         goto bad;
diff --git a/src/libslirp.h b/src/libslirp.h
index ba6a142..2ba4388 100644
--- a/src/libslirp.h
+++ b/src/libslirp.h
@@ -85,6 +85,10 @@ typedef struct SlirpConfig {
     struct in6_addr vnameserver6;
     const char **vdnssearch;
     const char *vdomainname;
+    /* Default: IF_MTU_DEFAULT */
+    size_t if_mtu;
+    /* Default: IF_MRU_DEFAULT */
+    size_t if_mru;
 } SlirpConfig;
 
 Slirp *slirp_new(const SlirpConfig *cfg, const SlirpCb *callbacks,
diff --git a/src/mbuf.c b/src/mbuf.c
index 4fd6228..54ec721 100644
--- a/src/mbuf.c
+++ b/src/mbuf.c
@@ -20,8 +20,8 @@
 /*
  * Find a nice value for msize
  */
-#define SLIRP_MSIZE \
-    (offsetof(struct mbuf, m_dat) + IF_MAXLINKHDR + TCPIPHDR_DELTA + IF_MTU)
+#define SLIRP_MSIZE(mtu) \
+    (offsetof(struct mbuf, m_dat) + IF_MAXLINKHDR + TCPIPHDR_DELTA + (mtu))
 
 void m_init(Slirp *slirp)
 {
@@ -66,7 +66,7 @@ struct mbuf *m_get(Slirp *slirp)
     DEBUG_CALL("m_get");
 
     if (slirp->m_freelist.qh_link == &slirp->m_freelist) {
-        m = g_malloc(SLIRP_MSIZE);
+        m = g_malloc(SLIRP_MSIZE(slirp->if_mtu));
         slirp->mbuf_alloced++;
         if (slirp->mbuf_alloced > MBUF_THRESH)
             flags = M_DOFREE;
@@ -81,7 +81,7 @@ struct mbuf *m_get(Slirp *slirp)
     m->m_flags = (flags | M_USEDLIST);
 
     /* Initialise it */
-    m->m_size = SLIRP_MSIZE - offsetof(struct mbuf, m_dat);
+    m->m_size = SLIRP_MSIZE(slirp->if_mtu) - offsetof(struct mbuf, m_dat);
     m->m_data = m->m_dat;
     m->m_len = 0;
     m->m_nextpkt = NULL;
diff --git a/src/slirp.c b/src/slirp.c
index c82f675..3be982c 100644
--- a/src/slirp.c
+++ b/src/slirp.c
@@ -271,6 +271,10 @@ Slirp *slirp_new(const SlirpConfig *cfg, const SlirpCb *callbacks, void *opaque)
 {
     Slirp *slirp;
     g_return_val_if_fail(cfg != NULL, NULL);
+    g_return_val_if_fail(cfg->if_mtu >= IF_MTU_MIN || cfg->if_mtu == 0, NULL);
+    g_return_val_if_fail(cfg->if_mtu <= IF_MTU_MAX, NULL);
+    g_return_val_if_fail(cfg->if_mru >= IF_MRU_MIN || cfg->if_mru == 0, NULL);
+    g_return_val_if_fail(cfg->if_mru <= IF_MRU_MAX, NULL);
 
     slirp = g_malloc0(sizeof(Slirp));
 
@@ -311,6 +315,8 @@ Slirp *slirp_new(const SlirpConfig *cfg, const SlirpCb *callbacks, void *opaque)
     if (cfg->vdnssearch) {
         translate_dnssearch(slirp, cfg->vdnssearch);
     }
+    slirp->if_mtu = cfg->if_mtu == 0 ? IF_MTU_DEFAULT : cfg->if_mtu;
+    slirp->if_mru = cfg->if_mru == 0 ? IF_MRU_DEFAULT : cfg->if_mru;
 
     return slirp;
 }
@@ -917,7 +923,7 @@ static int if_encap6(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
  */
 int if_encap(Slirp *slirp, struct mbuf *ifm)
 {
-    uint8_t buf[1600];
+    uint8_t buf[IF_MTU_MAX + 100];
     struct ethhdr *eh = (struct ethhdr *)buf;
     uint8_t ethaddr[ETH_ALEN];
     const struct ip *iph = (const struct ip *)ifm->m_data;
diff --git a/src/slirp.h b/src/slirp.h
index 37ba6ed..5fbee5c 100644
--- a/src/slirp.h
+++ b/src/slirp.h
@@ -144,6 +144,9 @@ struct Slirp {
     int restricted;
     struct gfwd_list *guestfwd_list;
 
+    int if_mtu;
+    int if_mru;
+
     /* mbuf states */
     struct quehead m_freelist;
     struct quehead m_usedlist;
diff --git a/src/tcp.h b/src/tcp.h
index 22625e6..70a9760 100644
--- a/src/tcp.h
+++ b/src/tcp.h
@@ -41,8 +41,9 @@ typedef uint32_t tcp_seq;
 #define PR_SLOWHZ 2 /* 2 slow timeouts per second (approx) */
 #define PR_FASTHZ 5 /* 5 fast timeouts per second (not important) */
 
-#define TCP_SNDSPACE 8192
-#define TCP_RCVSPACE 8192
+#define TCP_SNDSPACE 1024 * 128
+#define TCP_RCVSPACE 1024 * 128
+#define TCP_MAXSEG_MAX 32768
 
 /*
  * TCP header.
@@ -100,19 +101,6 @@ struct tcphdr {
 #define TCPOLEN_TSTAMP_APPA (TCPOLEN_TIMESTAMP + 2) /* appendix A */
 #endif
 
-/*
- * Default maximum segment size for TCP.
- * With an IP MSS of 576, this is 536,
- * but 512 is probably more convenient.
- * This should be defined as MIN(512, IP_MSS - sizeof (struct tcpiphdr)).
- *
- * We make this 1460 because we only care about Ethernet in the qemu context.
- */
-#undef TCP_MSS
-#define TCP_MSS 1460
-#undef TCP6_MSS
-#define TCP6_MSS 1440
-
 #undef TCP_MAXWIN
 #define TCP_MAXWIN 65535 /* largest value for (unscaled) window */
 
diff --git a/src/tcp_input.c b/src/tcp_input.c
index 228c2aa..30265d6 100644
--- a/src/tcp_input.c
+++ b/src/tcp_input.c
@@ -1510,10 +1510,12 @@ int tcp_mss(struct tcpcb *tp, unsigned offer)
 
     switch (so->so_ffamily) {
     case AF_INET:
-        mss = MIN(IF_MTU, IF_MRU) - sizeof(struct tcphdr) - sizeof(struct ip);
+        mss = MIN(so->slirp->if_mtu, so->slirp->if_mru) -
+              sizeof(struct tcphdr) - sizeof(struct ip);
         break;
     case AF_INET6:
-        mss = MIN(IF_MTU, IF_MRU) - sizeof(struct tcphdr) - sizeof(struct ip6);
+        mss = MIN(so->slirp->if_mtu, so->slirp->if_mru) -
+              sizeof(struct tcphdr) - sizeof(struct ip6);
         break;
     default:
         g_assert_not_reached();
@@ -1523,7 +1525,7 @@ int tcp_mss(struct tcpcb *tp, unsigned offer)
         mss = MIN(mss, offer);
     mss = MAX(mss, 32);
     if (mss < tp->t_maxseg || offer != 0)
-        tp->t_maxseg = mss;
+        tp->t_maxseg = MIN(mss, TCP_MAXSEG_MAX);
 
     tp->snd_cwnd = mss;
 
diff --git a/src/tcp_subr.c b/src/tcp_subr.c
index d6dd133..83c7520 100644
--- a/src/tcp_subr.c
+++ b/src/tcp_subr.c
@@ -261,7 +261,13 @@ struct tcpcb *tcp_newtcpcb(struct socket *so)
 
     memset((char *)tp, 0, sizeof(struct tcpcb));
     tp->seg_next = tp->seg_prev = (struct tcpiphdr *)tp;
-    tp->t_maxseg = (so->so_ffamily == AF_INET) ? TCP_MSS : TCP6_MSS;
+    /*
+     * 40: length of IPv4 header (20) + TCP header (20)
+     * 60: length of IPv6 header (40) + TCP header (20)
+     */
+    tp->t_maxseg =
+        MIN(so->slirp->if_mtu - ((so->so_ffamily == AF_INET) ? 40 : 60),
+            TCP_MAXSEG_MAX);
 
     tp->t_flags = TCP_DO_RFC1323 ? (TF_REQ_SCALE | TF_REQ_TSTMP) : 0;
     tp->t_socket = so;