[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Minios-devel] [UNIKRAFT/LIBMICROPYTHON PATCH 3/7] Add Unikraft-adapted external modules.



For now we add just modusocket. modlwip we introduce by patching the
original version in the sources; we dcan't do this for modusocket
since this file does not exist in the unix sub-directory of
Micropython.

Signed-off-by: Felipe Huici <felipe.huici@xxxxxxxxx>
---
 extmod/modusocket.c | 624 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 624 insertions(+)
 create mode 100644 extmod/modusocket.c

diff --git a/extmod/modusocket.c b/extmod/modusocket.c
new file mode 100644
index 0000000..61402e0
--- /dev/null
+++ b/extmod/modusocket.c
@@ -0,0 +1,624 @@
+/*
+ * This file is part of the MicroPython project, http://micropython.org/
+ *
+ * The MIT License (MIT)
+ *
+ * Copyright (c) 2013, 2014 Damien P. George
+ * Copyright (c) 2014 Paul Sokolovsky
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to 
deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 
FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include <stdio.h>
+#include <assert.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
+#include <netdb.h>
+#include <errno.h>
+#include <math.h>
+
+#include "py/objtuple.h"
+#include "py/objstr.h"
+#include "py/runtime.h"
+#include "py/stream.h"
+#include "py/builtin.h"
+#include "py/mphal.h"
+
+/*
+  The idea of this module is to implement reasonable minimum of
+  socket-related functions to write typical clients and servers.
+  The module named "usocket" on purpose, to allow to make
+  Python-level module more (or fully) compatible with CPython
+  "socket", e.g.:
+  ---- socket.py ----
+  from usocket import *
+  from socket_more_funcs import *
+  from socket_more_funcs2 import *
+  -------------------
+  I.e. this module should stay lean, and more functions (if needed)
+  should be add to separate modules (C or Python level).
+ */
+
+// This type must "inherit" from mp_obj_fdfile_t, i.e. matching subset of
+// fields should have the same layout.
+typedef struct _mp_obj_socket_t {
+    mp_obj_base_t base;
+    int fd;
+    bool blocking;
+} mp_obj_socket_t;
+
+const mp_obj_type_t mp_type_socket;
+
+// Helper functions
+static inline mp_obj_t mp_obj_from_sockaddr(const struct sockaddr *addr, 
socklen_t len) {
+    return mp_obj_new_bytes((const byte *)addr, len);
+}
+
+STATIC mp_obj_socket_t *socket_new(int fd) {
+    mp_obj_socket_t *o = m_new_obj(mp_obj_socket_t);
+    o->base.type = &mp_type_socket;
+    o->fd = fd;
+    o->blocking = true;
+    return o;
+}
+
+
+STATIC void socket_print(const mp_print_t *print, mp_obj_t self_in, 
mp_print_kind_t kind) {
+    (void)kind;
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(self_in);
+    mp_printf(print, "<_socket %d>", self->fd);
+}
+
+STATIC mp_uint_t socket_read(mp_obj_t o_in, void *buf, mp_uint_t size, int 
*errcode) {
+    mp_obj_socket_t *o = MP_OBJ_TO_PTR(o_in);
+    mp_int_t r = read(o->fd, buf, size);
+    if (r == -1) {
+        int err = errno;
+        // On blocking socket, we get EAGAIN in case SO_RCVTIMEO/SO_SNDTIMEO
+        // timed out, and need to convert that to ETIMEDOUT.
+        if (err == EAGAIN && o->blocking) {
+            err = MP_ETIMEDOUT;
+        }
+
+        *errcode = err;
+        return MP_STREAM_ERROR;
+    }
+    return r;
+}
+
+STATIC mp_uint_t socket_write(mp_obj_t o_in, const void *buf, mp_uint_t size, 
int *errcode) {
+    mp_obj_socket_t *o = MP_OBJ_TO_PTR(o_in);
+    mp_int_t r = write(o->fd, buf, size);
+    if (r == -1) {
+        int err = errno;
+        // On blocking socket, we get EAGAIN in case SO_RCVTIMEO/SO_SNDTIMEO
+        // timed out, and need to convert that to ETIMEDOUT.
+        if (err == EAGAIN && o->blocking) {
+            err = MP_ETIMEDOUT;
+        }
+
+        *errcode = err;
+        return MP_STREAM_ERROR;
+    }
+    return r;
+}
+
+STATIC mp_uint_t socket_ioctl(mp_obj_t o_in, mp_uint_t request, uintptr_t arg, 
int *errcode) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(o_in);
+    (void)arg;
+    switch (request) {
+        case MP_STREAM_CLOSE:
+            // There's a POSIX drama regarding return value of close in 
general,
+            // and EINTR error in particular. See e.g.
+            // http://lwn.net/Articles/576478/
+            // http://austingroupbugs.net/view.php?id=529
+            // The rationale MicroPython follows is that close() just releases
+            // file descriptor. If you're interested to catch I/O errors before
+            // closing fd, fsync() it.
+            close(self->fd);
+            return 0;
+
+        case MP_STREAM_GET_FILENO:
+            return self->fd;
+
+        default:
+            *errcode = MP_EINVAL;
+            return MP_STREAM_ERROR;
+    }
+}
+
+STATIC mp_obj_t socket_fileno(mp_obj_t self_in) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(self_in);
+    return MP_OBJ_NEW_SMALL_INT(self->fd);
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_1(socket_fileno_obj, socket_fileno);
+
+STATIC mp_obj_t socket_connect(mp_obj_t self_in, mp_obj_t addr_in) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(self_in);
+    mp_buffer_info_t bufinfo;
+    mp_get_buffer_raise(addr_in, &bufinfo, MP_BUFFER_READ);
+    int r = connect(self->fd, (const struct sockaddr *)bufinfo.buf, 
bufinfo.len);
+    RAISE_ERRNO(r, errno);
+    return mp_const_none;
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_2(socket_connect_obj, socket_connect);
+
+STATIC mp_obj_t socket_bind(mp_obj_t self_in, mp_obj_t addr_in) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(self_in);
+    mp_buffer_info_t bufinfo;
+    mp_get_buffer_raise(addr_in, &bufinfo, MP_BUFFER_READ);
+    int r = bind(self->fd, (const struct sockaddr *)bufinfo.buf, bufinfo.len);
+    RAISE_ERRNO(r, errno);
+    return mp_const_none;
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_2(socket_bind_obj, socket_bind);
+
+STATIC mp_obj_t socket_listen(mp_obj_t self_in, mp_obj_t backlog_in) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(self_in);
+    int r = listen(self->fd, MP_OBJ_SMALL_INT_VALUE(backlog_in));
+    RAISE_ERRNO(r, errno);
+    return mp_const_none;
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_2(socket_listen_obj, socket_listen);
+
+STATIC mp_obj_t socket_accept(mp_obj_t self_in) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(self_in);
+    // sockaddr_storage isn't stack-friendly (129 bytes or so)
+    //struct sockaddr_storage addr;
+    byte addr[32];
+    socklen_t addr_len = sizeof(addr);
+    int fd = accept(self->fd, (struct sockaddr*)&addr, &addr_len);
+    RAISE_ERRNO(fd, errno);
+
+    mp_obj_tuple_t *t = MP_OBJ_TO_PTR(mp_obj_new_tuple(2, NULL));
+    t->items[0] = MP_OBJ_FROM_PTR(socket_new(fd));
+    t->items[1] = mp_obj_new_bytearray(addr_len, &addr);
+
+    return MP_OBJ_FROM_PTR(t);
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_1(socket_accept_obj, socket_accept);
+
+// Note: besides flag param, this differs from read() in that
+// this does not swallow blocking errors (EAGAIN, EWOULDBLOCK) -
+// these would be thrown as exceptions.
+STATIC mp_obj_t socket_recv(size_t n_args, const mp_obj_t *args) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(args[0]);
+    int sz = MP_OBJ_SMALL_INT_VALUE(args[1]);
+    int flags = 0;
+
+    if (n_args > 2) {
+        flags = MP_OBJ_SMALL_INT_VALUE(args[2]);
+    }
+
+    byte *buf = m_new(byte, sz);
+    int out_sz = recv(self->fd, buf, sz, flags);
+    RAISE_ERRNO(out_sz, errno);
+
+    mp_obj_t ret = mp_obj_new_str_of_type(&mp_type_bytes, buf, out_sz);
+    m_del(char, buf, sz);
+    return ret;
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(socket_recv_obj, 2, 3, socket_recv);
+
+STATIC mp_obj_t socket_recvfrom(size_t n_args, const mp_obj_t *args) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(args[0]);
+    int sz = MP_OBJ_SMALL_INT_VALUE(args[1]);
+    int flags = 0;
+
+    if (n_args > 2) {
+        flags = MP_OBJ_SMALL_INT_VALUE(args[2]);
+    }
+
+    struct sockaddr_storage addr;
+    socklen_t addr_len = sizeof(addr);
+
+    byte *buf = m_new(byte, sz);
+    int out_sz = recvfrom(self->fd, buf, sz, flags, (struct sockaddr*)&addr, 
&addr_len);
+    RAISE_ERRNO(out_sz, errno);
+
+    mp_obj_t buf_o = mp_obj_new_str_of_type(&mp_type_bytes, buf, out_sz);
+    m_del(char, buf, sz);
+
+    mp_obj_tuple_t *t = MP_OBJ_TO_PTR(mp_obj_new_tuple(2, NULL));
+    t->items[0] = buf_o;
+    t->items[1] = mp_obj_from_sockaddr((struct sockaddr*)&addr, addr_len);
+
+    return MP_OBJ_FROM_PTR(t);
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(socket_recvfrom_obj, 2, 3, 
socket_recvfrom);
+
+// Note: besides flag param, this differs from write() in that
+// this does not swallow blocking errors (EAGAIN, EWOULDBLOCK) -
+// these would be thrown as exceptions.
+STATIC mp_obj_t socket_send(size_t n_args, const mp_obj_t *args) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(args[0]);
+    int flags = 0;
+
+    if (n_args > 2) {
+        flags = MP_OBJ_SMALL_INT_VALUE(args[2]);
+    }
+
+    mp_buffer_info_t bufinfo;
+    mp_get_buffer_raise(args[1], &bufinfo, MP_BUFFER_READ);
+    int out_sz = send(self->fd, bufinfo.buf, bufinfo.len, flags);
+    RAISE_ERRNO(out_sz, errno);
+
+    return MP_OBJ_NEW_SMALL_INT(out_sz);
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(socket_send_obj, 2, 3, socket_send);
+
+STATIC mp_obj_t socket_sendto(size_t n_args, const mp_obj_t *args) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(args[0]);
+    int flags = 0;
+
+    mp_obj_t dst_addr = args[2];
+    if (n_args > 3) {
+        flags = MP_OBJ_SMALL_INT_VALUE(args[2]);
+        dst_addr = args[3];
+    }
+
+    mp_buffer_info_t bufinfo, addr_bi;
+    mp_get_buffer_raise(args[1], &bufinfo, MP_BUFFER_READ);
+    mp_get_buffer_raise(dst_addr, &addr_bi, MP_BUFFER_READ);
+    int out_sz = sendto(self->fd, bufinfo.buf, bufinfo.len, flags,
+                        (struct sockaddr *)addr_bi.buf, addr_bi.len);
+    RAISE_ERRNO(out_sz, errno);
+
+    return MP_OBJ_NEW_SMALL_INT(out_sz);
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(socket_sendto_obj, 3, 4, 
socket_sendto);
+
+STATIC mp_obj_t socket_setsockopt(size_t n_args, const mp_obj_t *args) {
+    (void)n_args; // always 4
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(args[0]);
+    int level = MP_OBJ_SMALL_INT_VALUE(args[1]);
+    int option = mp_obj_get_int(args[2]);
+
+    const void *optval;
+    socklen_t optlen;
+    int val;
+    if (MP_OBJ_IS_INT(args[3])) {
+        val = mp_obj_int_get_truncated(args[3]);
+        optval = &val;
+        optlen = sizeof(val);
+    } else {
+        mp_buffer_info_t bufinfo;
+        mp_get_buffer_raise(args[3], &bufinfo, MP_BUFFER_READ);
+        optval = bufinfo.buf;
+        optlen = bufinfo.len;
+    }
+    int r = setsockopt(self->fd, level, option, optval, optlen);
+    RAISE_ERRNO(r, errno);
+    return mp_const_none;
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(socket_setsockopt_obj, 4, 4, 
socket_setsockopt);
+
+STATIC mp_obj_t socket_setblocking(mp_obj_t self_in, mp_obj_t flag_in) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(self_in);
+    int val = mp_obj_is_true(flag_in);
+    int flags = fcntl(self->fd, F_GETFL, 0);
+    RAISE_ERRNO(flags, errno);
+    if (val) {
+        flags &= ~O_NONBLOCK;
+    } else {
+        flags |= O_NONBLOCK;
+    }
+    flags = fcntl(self->fd, F_SETFL, flags);
+    RAISE_ERRNO(flags, errno);
+    self->blocking = val;
+    return mp_const_none;
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_2(socket_setblocking_obj, socket_setblocking);
+
+STATIC mp_obj_t socket_settimeout(mp_obj_t self_in, mp_obj_t timeout_in) {
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(self_in);
+    struct timeval tv = {0,};
+    bool new_blocking = true;
+
+    if (timeout_in == mp_const_none) {
+        setsockopt(self->fd, SOL_SOCKET, SO_RCVTIMEO, NULL, 0);
+        setsockopt(self->fd, SOL_SOCKET, SO_SNDTIMEO, NULL, 0);
+    } else {
+        #if MICROPY_PY_BUILTINS_FLOAT
+        mp_float_t val = mp_obj_get_float(timeout_in);
+        double ipart;
+        tv.tv_usec = round(modf(val, &ipart) * 1000000);
+        tv.tv_sec = ipart;
+        #else
+        tv.tv_sec = mp_obj_get_int(timeout_in);
+        #endif
+
+        // For SO_RCVTIMEO/SO_SNDTIMEO, zero timeout means infinity, but
+        // for Python API it means non-blocking.
+        if (tv.tv_sec == 0 && tv.tv_usec == 0) {
+            new_blocking = false;
+        } else {
+            setsockopt(self->fd, SOL_SOCKET, SO_RCVTIMEO,
+                       &tv, sizeof(struct timeval));
+            setsockopt(self->fd, SOL_SOCKET, SO_SNDTIMEO,
+                       &tv, sizeof(struct timeval));
+        }
+    }
+
+    if (self->blocking != new_blocking) {
+        socket_setblocking(self_in, mp_obj_new_bool(new_blocking));
+    }
+
+    return mp_const_none;
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_2(socket_settimeout_obj, socket_settimeout);
+
+STATIC mp_obj_t socket_makefile(size_t n_args, const mp_obj_t *args) {
+    // TODO: CPython explicitly says that closing returned object doesn't close
+    // the original socket (Python2 at all says that fd is dup()ed). But we
+    // save on the bloat.
+    mp_obj_socket_t *self = MP_OBJ_TO_PTR(args[0]);
+    mp_obj_t *new_args = alloca(n_args * sizeof(mp_obj_t));
+    memcpy(new_args + 1, args + 1, (n_args - 1) * sizeof(mp_obj_t));
+    new_args[0] = MP_OBJ_NEW_SMALL_INT(self->fd);
+    return mp_builtin_open(n_args, new_args, (mp_map_t*)&mp_const_empty_map);
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(socket_makefile_obj, 1, 3, 
socket_makefile);
+
+STATIC mp_obj_t socket_make_new(const mp_obj_type_t *type_in, size_t n_args, 
size_t n_kw, const mp_obj_t *args) {
+    (void)type_in;
+    (void)n_kw;
+
+    int family = AF_INET;
+    int type = SOCK_STREAM;
+    int proto = 0;
+
+    if (n_args > 0) {
+        assert(MP_OBJ_IS_SMALL_INT(args[0]));
+        family = MP_OBJ_SMALL_INT_VALUE(args[0]);
+        if (n_args > 1) {
+            assert(MP_OBJ_IS_SMALL_INT(args[1]));
+            type = MP_OBJ_SMALL_INT_VALUE(args[1]);
+            if (n_args > 2) {
+                assert(MP_OBJ_IS_SMALL_INT(args[2]));
+                proto = MP_OBJ_SMALL_INT_VALUE(args[2]);
+            }
+        }
+    }
+
+    int fd = socket(family, type, proto);
+    RAISE_ERRNO(fd, errno);
+    return MP_OBJ_FROM_PTR(socket_new(fd));
+}
+
+STATIC const mp_rom_map_elem_t usocket_locals_dict_table[] = {
+    { MP_ROM_QSTR(MP_QSTR_fileno), MP_ROM_PTR(&socket_fileno_obj) },
+    { MP_ROM_QSTR(MP_QSTR_makefile), MP_ROM_PTR(&socket_makefile_obj) },
+    { MP_ROM_QSTR(MP_QSTR_read), MP_ROM_PTR(&mp_stream_read_obj) },
+    { MP_ROM_QSTR(MP_QSTR_readinto), MP_ROM_PTR(&mp_stream_readinto_obj) },
+    { MP_ROM_QSTR(MP_QSTR_readline), 
MP_ROM_PTR(&mp_stream_unbuffered_readline_obj) },
+    { MP_ROM_QSTR(MP_QSTR_write), MP_ROM_PTR(&mp_stream_write_obj) },
+    { MP_ROM_QSTR(MP_QSTR_connect), MP_ROM_PTR(&socket_connect_obj) },
+    { MP_ROM_QSTR(MP_QSTR_bind), MP_ROM_PTR(&socket_bind_obj) },
+    { MP_ROM_QSTR(MP_QSTR_listen), MP_ROM_PTR(&socket_listen_obj) },
+    { MP_ROM_QSTR(MP_QSTR_accept), MP_ROM_PTR(&socket_accept_obj) },
+    { MP_ROM_QSTR(MP_QSTR_recv), MP_ROM_PTR(&socket_recv_obj) },
+    { MP_ROM_QSTR(MP_QSTR_recvfrom), MP_ROM_PTR(&socket_recvfrom_obj) },
+    { MP_ROM_QSTR(MP_QSTR_send), MP_ROM_PTR(&socket_send_obj) },
+    { MP_ROM_QSTR(MP_QSTR_sendto), MP_ROM_PTR(&socket_sendto_obj) },
+    { MP_ROM_QSTR(MP_QSTR_setsockopt), MP_ROM_PTR(&socket_setsockopt_obj) },
+    { MP_ROM_QSTR(MP_QSTR_setblocking), MP_ROM_PTR(&socket_setblocking_obj) },
+    { MP_ROM_QSTR(MP_QSTR_settimeout), MP_ROM_PTR(&socket_settimeout_obj) },
+    { MP_ROM_QSTR(MP_QSTR_close), MP_ROM_PTR(&mp_stream_close_obj) },
+};
+
+STATIC MP_DEFINE_CONST_DICT(usocket_locals_dict, usocket_locals_dict_table);
+
+STATIC const mp_stream_p_t usocket_stream_p = {
+    .read = socket_read,
+    .write = socket_write,
+    .ioctl = socket_ioctl,
+};
+
+const mp_obj_type_t mp_type_socket = {
+    { &mp_type_type },
+    .name = MP_QSTR_socket,
+    .print = socket_print,
+    .make_new = socket_make_new,
+    .getiter = NULL,
+    .iternext = NULL,
+    .protocol = &usocket_stream_p,
+    .locals_dict = (mp_obj_dict_t*)&usocket_locals_dict,
+};
+
+#define BINADDR_MAX_LEN sizeof(struct in6_addr)
+STATIC mp_obj_t mod_socket_inet_pton(mp_obj_t family_in, mp_obj_t addr_in) {
+    int family = mp_obj_get_int(family_in);
+    byte binaddr[BINADDR_MAX_LEN];
+    int r = inet_pton(family, mp_obj_str_get_str(addr_in), binaddr);
+    RAISE_ERRNO(r, errno);
+    if (r == 0) {
+        mp_raise_OSError(MP_EINVAL);
+    }
+    int binaddr_len = 0;
+    switch (family) {
+        case AF_INET:
+            binaddr_len = sizeof(struct in_addr);
+            break;
+        case AF_INET6:
+            binaddr_len = sizeof(struct in6_addr);
+            break;
+    }
+    return mp_obj_new_bytes(binaddr, binaddr_len);
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_2(mod_socket_inet_pton_obj, 
mod_socket_inet_pton);
+
+STATIC mp_obj_t mod_socket_inet_ntop(mp_obj_t family_in, mp_obj_t binaddr_in) {
+    int family = mp_obj_get_int(family_in);
+    mp_buffer_info_t bufinfo;
+    mp_get_buffer_raise(binaddr_in, &bufinfo, MP_BUFFER_READ);
+    vstr_t vstr;
+    vstr_init_len(&vstr, family == AF_INET ? INET_ADDRSTRLEN : 
INET6_ADDRSTRLEN);
+    if (inet_ntop(family, bufinfo.buf, vstr.buf, vstr.len) == NULL) {
+        mp_raise_OSError(errno);
+    }
+    vstr.len = strlen(vstr.buf);
+    return mp_obj_new_str_from_vstr(&mp_type_str, &vstr);
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_2(mod_socket_inet_ntop_obj, 
mod_socket_inet_ntop);
+
+STATIC mp_obj_t mod_socket_getaddrinfo(size_t n_args, const mp_obj_t *args) {
+    // TODO: Implement 5th and 6th args
+
+    const char *host = mp_obj_str_get_str(args[0]);
+    const char *serv = NULL;
+    struct addrinfo hints;
+    char buf[6];
+    memset(&hints, 0, sizeof(hints));
+    // getaddrinfo accepts port in string notation, so however
+    // it may seem stupid, we need to convert int to str
+    if (MP_OBJ_IS_SMALL_INT(args[1])) {
+        unsigned port = (unsigned short)MP_OBJ_SMALL_INT_VALUE(args[1]);
+        snprintf(buf, sizeof(buf), "%u", port);
+        serv = buf;
+        hints.ai_flags = AI_NUMERICSERV;
+#ifdef __UCLIBC_MAJOR__
+#if __UCLIBC_MAJOR__ == 0 && (__UCLIBC_MINOR__ < 9 || (__UCLIBC_MINOR__ == 9 
&& __UCLIBC_SUBLEVEL__ <= 32))
+// "warning" requires -Wno-cpp which is a relatively new gcc option, so we 
choose not to use it.
+//#warning Working around uClibc bug with numeric service name
+        // Older versions og uClibc have bugs when numeric ports in service
+        // arg require also hints.ai_socktype (or hints.ai_protocol) != 0
+        // This actually was fixed in 0.9.32.1, but uClibc doesn't allow to
+        // test for that.
+        // 
http://git.uclibc.org/uClibc/commit/libc/inet/getaddrinfo.c?id=bc3be18145e4d5
+        // Note that this is crude workaround, precluding UDP socket addresses
+        // to be returned. TODO: set only if not set by Python args.
+        hints.ai_socktype = SOCK_STREAM;
+#endif
+#endif
+    } else {
+        serv = mp_obj_str_get_str(args[1]);
+    }
+
+    if (n_args > 2) {
+        hints.ai_family = MP_OBJ_SMALL_INT_VALUE(args[2]);
+        if (n_args > 3) {
+            hints.ai_socktype = MP_OBJ_SMALL_INT_VALUE(args[3]);
+        }
+    }
+
+    struct addrinfo *addr_list;
+    int res = getaddrinfo(host, serv, &hints, &addr_list);
+
+    if (res != 0) {
+        // CPython: socket.gaierror
+        nlr_raise(mp_obj_new_exception_msg_varg(&mp_type_OSError, "[addrinfo 
error %d]", res));
+    }
+    assert(addr_list);
+
+    mp_obj_t list = mp_obj_new_list(0, NULL);
+    for (struct addrinfo *addr = addr_list; addr; addr = addr->ai_next) {
+        mp_obj_tuple_t *t = MP_OBJ_TO_PTR(mp_obj_new_tuple(5, NULL));
+        t->items[0] = MP_OBJ_NEW_SMALL_INT(addr->ai_family);
+        t->items[1] = MP_OBJ_NEW_SMALL_INT(addr->ai_socktype);
+        t->items[2] = MP_OBJ_NEW_SMALL_INT(addr->ai_protocol);
+        // "canonname will be a string representing the canonical name of the 
host
+        // if AI_CANONNAME is part of the flags argument; else canonname will 
be empty." ??
+        if (addr->ai_canonname) {
+            t->items[3] = MP_OBJ_NEW_QSTR(qstr_from_str(addr->ai_canonname));
+        } else {
+            t->items[3] = mp_const_none;
+        }
+        t->items[4] = mp_obj_new_bytearray(addr->ai_addrlen, addr->ai_addr);
+        mp_obj_list_append(list, MP_OBJ_FROM_PTR(t));
+    }
+    freeaddrinfo(addr_list);
+    return list;
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(mod_socket_getaddrinfo_obj, 2, 4, 
mod_socket_getaddrinfo);
+
+STATIC mp_obj_t mod_socket_sockaddr(mp_obj_t sockaddr_in) {
+    mp_buffer_info_t bufinfo;
+    mp_get_buffer_raise(sockaddr_in, &bufinfo, MP_BUFFER_READ);
+    switch (((struct sockaddr*)bufinfo.buf)->sa_family) {
+        case AF_INET: {
+            struct sockaddr_in *sa = (struct sockaddr_in*)bufinfo.buf;
+            mp_obj_tuple_t *t = MP_OBJ_TO_PTR(mp_obj_new_tuple(3, NULL));
+            t->items[0] = MP_OBJ_NEW_SMALL_INT(AF_INET);
+            t->items[1] = mp_obj_new_bytes((byte*)&sa->sin_addr, 
sizeof(sa->sin_addr));
+            t->items[2] = MP_OBJ_NEW_SMALL_INT(ntohs(sa->sin_port));
+            return MP_OBJ_FROM_PTR(t);
+        }
+        case AF_INET6: {
+            struct sockaddr_in6 *sa = (struct sockaddr_in6*)bufinfo.buf;
+            mp_obj_tuple_t *t = MP_OBJ_TO_PTR(mp_obj_new_tuple(5, NULL));
+            t->items[0] = MP_OBJ_NEW_SMALL_INT(AF_INET6);
+            t->items[1] = mp_obj_new_bytes((byte*)&sa->sin6_addr, 
sizeof(sa->sin6_addr));
+            t->items[2] = MP_OBJ_NEW_SMALL_INT(ntohs(sa->sin6_port));
+            t->items[3] = MP_OBJ_NEW_SMALL_INT(ntohl(sa->sin6_flowinfo));
+            t->items[4] = MP_OBJ_NEW_SMALL_INT(ntohl(sa->sin6_scope_id));
+            return MP_OBJ_FROM_PTR(t);
+        }
+        default: {
+            struct sockaddr *sa = (struct sockaddr*)bufinfo.buf;
+            mp_obj_tuple_t *t = MP_OBJ_TO_PTR(mp_obj_new_tuple(2, NULL));
+            t->items[0] = MP_OBJ_NEW_SMALL_INT(sa->sa_family);
+            t->items[1] = mp_obj_new_bytes((byte*)sa->sa_data, bufinfo.len - 
offsetof(struct sockaddr, sa_data));
+            return MP_OBJ_FROM_PTR(t);
+        }
+    }
+    return mp_const_none;
+}
+STATIC MP_DEFINE_CONST_FUN_OBJ_1(mod_socket_sockaddr_obj, mod_socket_sockaddr);
+
+STATIC const mp_rom_map_elem_t mp_module_socket_globals_table[] = {
+    { MP_ROM_QSTR(MP_QSTR___name__), MP_ROM_QSTR(MP_QSTR_usocket) },
+    { MP_ROM_QSTR(MP_QSTR_socket), MP_ROM_PTR(&mp_type_socket) },
+    { MP_ROM_QSTR(MP_QSTR_getaddrinfo), 
MP_ROM_PTR(&mod_socket_getaddrinfo_obj) },
+    { MP_ROM_QSTR(MP_QSTR_inet_pton), MP_ROM_PTR(&mod_socket_inet_pton_obj) },
+    { MP_ROM_QSTR(MP_QSTR_inet_ntop), MP_ROM_PTR(&mod_socket_inet_ntop_obj) },
+    { MP_ROM_QSTR(MP_QSTR_sockaddr), MP_ROM_PTR(&mod_socket_sockaddr_obj) },
+
+#define C(name) { MP_ROM_QSTR(MP_QSTR_ ## name), MP_ROM_INT(name) }
+    C(AF_UNIX),
+    C(AF_INET),
+    C(AF_INET6),
+    C(SOCK_STREAM),
+    C(SOCK_DGRAM),
+    C(SOCK_RAW),
+
+    C(MSG_DONTROUTE),
+    C(MSG_DONTWAIT),
+
+    C(SOL_SOCKET),
+    C(SO_BROADCAST),
+    C(SO_ERROR),
+    C(SO_KEEPALIVE),
+    C(SO_LINGER),
+    C(SO_REUSEADDR),
+#undef C
+};
+
+STATIC MP_DEFINE_CONST_DICT(mp_module_socket_globals, 
mp_module_socket_globals_table);
+
+const mp_obj_module_t mp_module_socket = {
+    .base = { &mp_type_module },
+    .globals = (mp_obj_dict_t*)&mp_module_socket_globals,
+};
-- 
2.20.1


_______________________________________________
Minios-devel mailing list
Minios-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/minios-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.