Initial recvmsg(3) implementation (fixes traceroute)

kenardi · Dec 20, 2018 · 5a4af30 · 5a4af30
1 parent 0a51d8f
commit 5a4af30
Show file tree

Hide file tree

Showing 2 changed files with 164 additions and 2 deletions.
diff --git a/fs/sock.c b/fs/sock.c
@@ -369,6 +369,128 @@ dword_t sys_getsockopt(fd_t sock_fd, dword_t level, dword_t option, addr_t value
     return 0;
 }
 
+dword_t sys_recvmsg(fd_t sock_fd, addr_t msghdr_addr, dword_t flags) {
+    STRACE("recvmsg(%d, 0x%x, %d)", sock_fd, msghdr_addr, flags);
+    struct fd *sock = sock_getfd(sock_fd);
+    if (sock == NULL)
+        return _EBADF;
+
+    struct msghdr msg;
+    memset(&msg, 0, sizeof(msg));
+
+    struct msghdr_ msg_fake;
+    if (user_get(msghdr_addr, msg_fake)) {
+        return _EFAULT;
+    }
+
+    // msg_name
+    struct sockaddr msg_name;
+    memset(&msg_name, 0, sizeof(msg_name));
+
+    struct sockaddr_ msg_name_fake;
+    if (msg_fake.msg_name != 0 && msg_fake.msg_namelen != 0) {
+        if (user_get(msg_fake.msg_name, msg_name_fake))
+            return _EFAULT;
+#ifdef __APPLE__
+        msg_name.sa_len = sizeof(msg_name);
+#endif
+        msg_name.sa_family = msg_name_fake.family;
+        memcpy(&msg_name.sa_data, msg_name_fake.data, sizeof(msg_name_fake.data));
+
+        msg.msg_name = (void *)&msg_name;
+        msg.msg_namelen = sizeof(msg_name);
+    }
+
+    // msg_control (no initial content)
+    char msg_control[msg_fake.msg_controllen];
+    memset(&msg_control, 0, sizeof(msg_control));
+
+    msg.msg_control = (void *)&msg_control;
+    msg.msg_controllen = sizeof(msg_control);
+
+    // msg_iovec (no initial content)
+    struct iovec msg_iov[msg_fake.msg_iovlen];
+    memset(&msg_iov, 0, sizeof(msg_iov));
+
+    struct iovec_ msg_iov_fake[msg_fake.msg_iovlen];
+    if (user_get(msg_fake.msg_iov, msg_iov_fake))
+        return _EFAULT;
+
+    for (int i = 0; i < msg_fake.msg_iovlen; i++) {
+        char iov_base[msg_iov_fake[i].iov_len];
+        memset(&iov_base, 0, sizeof(iov_base));
+
+        msg_iov[i].iov_base = (void *)&iov_base;
+        msg_iov[i].iov_len = sizeof(iov_base);
+    }
+
+    msg.msg_iov = (void *)&msg_iov;
+    msg.msg_iovlen = msg_fake.msg_iovlen;
+
+    // msg_flags (no initial content)
+    msg.msg_flags = sock_flags_to_real(msg_fake.msg_flags);
+
+    int real_flags = sock_flags_to_real(flags);
+    if (real_flags < 0)
+        return _EINVAL;
+
+    ssize_t res = recvmsg(sock->real_fd, &msg, real_flags);
+
+    // msg_name (changed)
+    memset(&msg_name_fake, 0, sizeof(msg_name_fake));
+
+    if (msg.msg_name != 0 && msg.msg_namelen != 0) {
+        struct sockaddr *msg_name = msg.msg_name;
+
+        msg_name_fake.family = msg_name->sa_family;
+        memcpy(&msg_name_fake.data, msg_name->sa_data, sizeof(msg_name->sa_data));
+
+        if (user_put(msg_fake.msg_name, msg_name_fake)) {
+            return _EFAULT;
+        }
+        msg_fake.msg_namelen = sizeof(msg_name_fake);
+    }
+
+    // msg_control (changed)
+    uint_t message_offset = 0;
+    for (struct cmsghdr *cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
+        void *cmsg_data = CMSG_DATA(cmsg);
+        uint_t cmsg_data_size = cmsg->cmsg_len - sizeof(cmsg);
+
+        if (user_write(msg_fake.msg_control + message_offset, cmsg, sizeof(cmsg)))
+            return _EFAULT;
+
+        if (user_write(msg_fake.msg_control + message_offset + sizeof(cmsg), cmsg_data, cmsg_data_size))
+            return _EFAULT;
+
+        message_offset += cmsg->cmsg_len;
+    }
+
+    msg_fake.msg_controllen = message_offset;
+
+    // msg_iovec (changed)
+    for (int i = 0; i < msg.msg_iovlen; i++) {
+        if (user_write(msg_iov_fake[i].iov_base, msg_iov[i].iov_base, msg_iov[i].iov_len))
+            return _EFAULT;
+
+        msg_iov_fake[i].iov_len = msg_iov[i].iov_len;
+    }
+
+    if (user_put(msg_fake.msg_iov, msg_iov_fake))
+        return _EFAULT;
+
+    // msg_flags (changed)
+    msg_fake.msg_flags = sock_flags_from_real(msg.msg_flags);
+
+    if (user_put(msghdr_addr, msg_fake)) {
+        return _EFAULT;
+    }
+
+    if (res < 0)
+        return errno_map();
+    return res;
+}
+
 static struct fd_ops socket_fdops = {
     .read = realfs_read,
     .write = realfs_write,
@@ -398,7 +520,7 @@ static struct socket_call {
     {(syscall_t) sys_setsockopt, 5},
     {(syscall_t) sys_getsockopt, 5},
     {NULL}, // sendmsg
-    {NULL}, // recvmsg
+    {(syscall_t) sys_recvmsg, 3},
     {NULL}, // accept4
     {NULL}, // recvmmsg
     {NULL}, // sendmmsg

diff --git a/fs/sock.h b/fs/sock.h
@@ -19,6 +19,21 @@ size_t sockaddr_size(void *p);
 // result comes from malloc
 struct sockaddr *sockaddr_to_real(void *p);
 
+struct msghdr_ {
+  addr_t msg_name;
+  int_t msg_namelen;
+  addr_t msg_iov;
+  uint_t msg_iovlen;
+  addr_t msg_control;
+  uint_t msg_controllen;
+  int_t msg_flags;
+};
+
+struct iovec_ {
+  addr_t iov_base;
+  uint_t iov_len;
+};
+
 #define PF_LOCAL_ 1
 #define PF_INET_ 2
 #define PF_INET6_ 10
@@ -69,17 +84,38 @@ static inline int sock_type_to_real(int type, int protocol) {
 
 #define MSG_OOB_ 0x1
 #define MSG_PEEK_ 0x2
+#define MSG_CTRUNC_  0x8
+#define MSG_TRUNC_  0x20
+#define MSG_DONTWAIT_ 0x40
+#define MSG_EOR_    0x80
 #define MSG_WAITALL_ 0x100
 
 static inline int sock_flags_to_real(int fake) {
     int real = 0;
     if (fake & MSG_OOB_) real |= MSG_OOB;
     if (fake & MSG_PEEK_) real |= MSG_PEEK;
+    if (fake & MSG_CTRUNC_) real |= MSG_CTRUNC;
+    if (fake & MSG_TRUNC_) real |= MSG_TRUNC;
+    if (fake & MSG_DONTWAIT_) real |= MSG_DONTWAIT;
+    if (fake & MSG_EOR_) real |= MSG_EOR;
     if (fake & MSG_WAITALL_) real |= MSG_WAITALL;
-    if (fake & ~(MSG_OOB_|MSG_PEEK_|MSG_WAITALL_))
+    if (fake & ~(MSG_OOB_|MSG_PEEK_|MSG_CTRUNC_|MSG_TRUNC_|MSG_DONTWAIT_|MSG_EOR_|MSG_WAITALL_))
         TRACE("unimplemented socket flags %d\n", fake);
     return real;
 }
+static inline int sock_flags_from_real(int real) {
+  int fake = 0;
+  if (real & MSG_OOB) fake |= MSG_OOB_;
+  if (real & MSG_PEEK) fake |= MSG_PEEK_;
+  if (real & MSG_CTRUNC) fake |= MSG_CTRUNC_;
+  if (real & MSG_TRUNC) fake |= MSG_TRUNC_;
+  if (real & MSG_DONTWAIT) fake |= MSG_DONTWAIT_;
+  if (real & MSG_EOR) fake |= MSG_EOR_;
+  if (real & MSG_WAITALL) fake |= MSG_WAITALL_;
+  if (real & ~(MSG_OOB|MSG_PEEK|MSG_CTRUNC|MSG_TRUNC|MSG_DONTWAIT|MSG_EOR|MSG_WAITALL))
+      TRACE("unimplemented socket flags %d\n", real);
+  return fake;
+}
 
 #define SOL_SOCKET_ 1
 #define IPPROTO_TCP_ 6
@@ -93,7 +129,9 @@ static inline int sock_flags_to_real(int fake) {
 #define SO_ERROR_ 4
 #define SO_BROADCAST_ 6
 #define SO_KEEPALIVE_ 9
+#define SO_SNDBUF_ 7
 #define IP_TOS_ 1
+#define IP_TTL_ 2
 #define TCP_NODELAY_ 1
 #define IPV6_TCLASS_ 67
 #define ICMP6_FILTER_ 1
@@ -106,12 +144,14 @@ static inline int sock_opt_to_real(int fake, int level) {
             case SO_ERROR_: return SO_ERROR;
             case SO_BROADCAST_: return SO_BROADCAST;
             case SO_KEEPALIVE_: return SO_KEEPALIVE;
+            case SO_SNDBUF_: return SO_SNDBUF;
         } break;
         case IPPROTO_TCP_: switch (fake) {
             case TCP_NODELAY_: return TCP_NODELAY;
         } break;
         case IPPROTO_IP_: switch (fake) {
             case IP_TOS_: return IP_TOS;
+            case IP_TTL_: return IP_TTL;
         } break;
         case IPPROTO_IPV6_: switch (fake) {
             case IPV6_TCLASS_: return IPV6_TCLASS;