libwebsockets/lib/roles/http/server/server.c
2024-11-03 07:59:12 +00:00

3512 lines
88 KiB
C

/*
* libwebsockets - small server side websockets and web server implementation
*
* Copyright (C) 2010 - 2019 Andy Green <andy@warmcat.com>
*
* 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 "private-lib-core.h"
#if !defined(SOL_TCP) && defined(IPPROTO_TCP)
#define SOL_TCP IPPROTO_TCP
#endif
const char * const method_names[] = {
"GET", "POST",
#if defined(LWS_WITH_HTTP_UNCOMMON_HEADERS)
"OPTIONS", "PUT", "PATCH", "DELETE",
#endif
"CONNECT", "HEAD",
#ifdef LWS_WITH_HTTP2
":path",
#endif
};
#if defined(LWS_WITH_FILE_OPS)
static const char * const intermediates[] = { "private", "public" };
#endif
/*
* return 0: all done
* 1: nonfatal error
* <0: fatal error
*
* REQUIRES CONTEXT LOCK HELD
*/
#if defined(LWS_WITH_SERVER)
struct vh_sock_args {
const struct lws_context_creation_info *info;
struct lws_vhost *vhost;
int af;
};
static int
check_extant(struct lws_dll2 *d, void *user)
{
struct lws *wsi = lws_container_of(d, struct lws, listen_list);
struct vh_sock_args *a = (struct vh_sock_args *)user;
if (!lws_vhost_compare_listen(wsi->a.vhost, a->vhost))
return 0;
if (wsi->af != a ->af)
return 0;
if (a->info && a->info->vh_listen_sockfd &&
wsi->desc.sockfd != a->info->vh_listen_sockfd)
return 0;
lwsl_notice(" using listen skt from vhost %s\n", wsi->a.vhost->name);
return 1;
}
/*
* Creates a single listen socket of a specific AF
*/
int
_lws_vhost_init_server_af(struct vh_sock_args *a)
{
struct lws_context *cx = a->vhost->context;
struct lws_context_per_thread *pt;
int n, opt = 1, limit = 1, san = 2;
lws_sockfd_type sockfd;
struct lws *wsi;
int m = 0, is = 0;
#if defined(LWS_WITH_IPV6)
int value = 1;
#endif
(void)method_names;
(void)opt;
lwsl_info("%s: af %d\n", __func__, (int)a->af);
if (lws_vhost_foreach_listen_wsi(a->vhost->context, a, check_extant))
return 0;
deal:
if (!san--)
return -1;
if (a->vhost->iface && (!a->info || !a->info->vh_listen_sockfd)) {
/*
* let's check before we do anything else about the disposition
* of the interface he wants to bind to...
*/
is = lws_socket_bind(a->vhost, NULL, LWS_SOCK_INVALID,
a->vhost->listen_port, a->vhost->iface,
a->af);
lwsl_debug("initial if check says %d\n", is);
if (is == LWS_ITOSA_BUSY)
/* treat as fatal */
return -1;
lws_start_foreach_llp(struct lws_vhost **, pv,
cx->no_listener_vhost_list) {
if (is >= LWS_ITOSA_USABLE && *pv == a->vhost) {
/* on the list and shouldn't be: remove it */
lwsl_debug("deferred iface: removing vh %s\n",
(*pv)->name);
*pv = a->vhost->no_listener_vhost_list;
a->vhost->no_listener_vhost_list = NULL;
goto done_list;
}
if (is < LWS_ITOSA_USABLE && *pv == a->vhost)
goto done_list;
} lws_end_foreach_llp(pv, no_listener_vhost_list);
/* not on the list... */
if (is < LWS_ITOSA_USABLE) {
/* ... but needs to be: so add it */
lwsl_debug("deferred iface: adding vh %s\n",
a->vhost->name);
a->vhost->no_listener_vhost_list =
cx->no_listener_vhost_list;
cx->no_listener_vhost_list = a->vhost;
}
done_list:
switch (is) {
default:
break;
case LWS_ITOSA_NOT_EXIST:
/* can't add it */
if (!a->info)
return -1;
/* first time */
lwsl_err("%s: VH %s: iface %s port %d DOESN'T EXIST\n",
__func__, a->vhost->name, a->vhost->iface,
a->vhost->listen_port);
return (a->info->options &
LWS_SERVER_OPTION_FAIL_UPON_UNABLE_TO_BIND) ==
LWS_SERVER_OPTION_FAIL_UPON_UNABLE_TO_BIND ?
-1 : 1;
case LWS_ITOSA_NOT_USABLE:
/* can't add it */
if (!a->info) /* first time */
return -1;
lwsl_err("%s: VH %s: iface %s port %d NOT USABLE\n",
__func__, a->vhost->name, a->vhost->iface,
a->vhost->listen_port);
return (a->info->options &
LWS_SERVER_OPTION_FAIL_UPON_UNABLE_TO_BIND) ==
LWS_SERVER_OPTION_FAIL_UPON_UNABLE_TO_BIND ?
-1 : 1;
}
} else {
if (a->info && a->info->vh_listen_sockfd) {
a->vhost->iface = "inherited";
a->vhost->listen_port = a->info->port;
}
}
(void)n;
#if defined(__linux__)
/*
* A Unix domain sockets cannot be bound multiple times, even if we
* set the SO_REUSE* options on.
*
* However on recent linux, each thread is able to independently listen.
*
* So we can assume creating just one listening socket for a multi-
* threaded environment will typically work.
*/
if (a->af != AF_UNIX)
limit = cx->count_threads;
#endif
for (m = 0; m < limit; m++) {
if (a->info && a->info->vh_listen_sockfd)
{
#if defined(_WIN32)
if (!DuplicateHandle(GetCurrentProcess(),
(HANDLE)a->info->vh_listen_sockfd,
GetCurrentProcess(), (HANDLE*)&sockfd, 0,
FALSE, DUPLICATE_SAME_ACCESS))
sockfd = LWS_SOCK_INVALID;
#else
#if defined(LWS_PLAT_FREERTOS)
sockfd = a->info->vh_listen_sockfd;
#else
sockfd = dup(a->info->vh_listen_sockfd);
#endif
#endif
}
else
sockfd = lws_fi(&a->vhost->fic, "listenskt") ?
LWS_SOCK_INVALID :
socket(a->af, SOCK_STREAM, 0);
if (sockfd == LWS_SOCK_INVALID) {
lwsl_err("ERROR opening socket\n");
return 1;
}
#if !defined(LWS_PLAT_FREERTOS)
#if (defined(WIN32) || defined(_WIN32)) && defined(SO_EXCLUSIVEADDRUSE)
/*
* only accept that we are the only listener on the port
* https://msdn.microsoft.com/zh-tw/library/
* windows/desktop/ms740621(v=vs.85).aspx
*
* for lws, to match Linux, we default to exclusive listen
*/
if (!lws_check_opt(a->vhost->options,
LWS_SERVER_OPTION_ALLOW_LISTEN_SHARE)) {
if (setsockopt(sockfd, SOL_SOCKET, SO_EXCLUSIVEADDRUSE,
(const void *)&opt, sizeof(opt)) < 0) {
lwsl_err("reuseaddr failed\n");
compatible_close(sockfd);
return -1;
}
} else
#endif
/*
* allow us to restart even if old sockets in TIME_WAIT
*/
if (setsockopt(sockfd, SOL_SOCKET, SO_REUSEADDR,
(const void *)&opt, sizeof(opt)) < 0) {
lwsl_err("reuseaddr failed\n");
compatible_close(sockfd);
return -1;
}
#if defined(LWS_WITH_IPV6) && defined(IPV6_V6ONLY)
/*
* If we have an ipv6 listen socket, it only accepts ipv6.
*
* There will be a separate ipv4 listen socket if that's
* enabled.
*/
if (a->af == AF_INET6 && (!a->info || !a->info->vh_listen_sockfd) &&
setsockopt(sockfd, IPPROTO_IPV6, IPV6_V6ONLY,
(const void*)&value, sizeof(value)) < 0) {
lwsl_err("ipv6 only failed\n");
compatible_close(sockfd);
return -1;
}
#endif
#if defined(__linux__) && defined(SO_REUSEPORT)
/* keep coverity happy */
#if LWS_MAX_SMP > 1
n = 1;
#else
n = lws_check_opt(a->vhost->options,
LWS_SERVER_OPTION_ALLOW_LISTEN_SHARE);
#endif
if (n || cx->count_threads > 1) /* ... also implied by threads > 1 */
if (setsockopt(sockfd, SOL_SOCKET, SO_REUSEPORT,
(const void *)&opt, sizeof(opt)) < 0) {
lwsl_err("reuseport failed\n");
compatible_close(sockfd);
return -1;
}
#endif
#endif
lws_plat_set_socket_options(a->vhost, sockfd, 0);
if (!a->info || !a->info->vh_listen_sockfd) {
is = lws_socket_bind(a->vhost, NULL, sockfd,
a->vhost->listen_port,
a->vhost->iface, a->af);
if (is == LWS_ITOSA_BUSY) {
/* treat as fatal */
compatible_close(sockfd);
return -1;
}
/*
* There is a race where the network device may come up and then
* go away and fail here. So correctly handle unexpected failure
* here despite we earlier confirmed it.
*/
if (is < 0) {
lwsl_info("%s: lws_socket_bind says %d\n", __func__, is);
compatible_close(sockfd);
if (a->vhost->iface)
goto deal;
return -1;
}
}
/*
* Create the listen wsi and customize it
*/
lws_context_lock(cx, __func__);
wsi = __lws_wsi_create_with_role(cx, m, &role_ops_listen, NULL);
lws_context_unlock(cx);
if (wsi == NULL) {
lwsl_err("Out of mem\n");
goto bail;
}
wsi->af = (uint8_t)a->af;
#ifdef LWS_WITH_UNIX_SOCK
if (!LWS_UNIX_SOCK_ENABLED(a->vhost))
#endif
{
wsi->unix_skt = 1;
a->vhost->listen_port = is;
lwsl_debug("%s: lws_socket_bind says %d\n", __func__, is);
}
wsi->desc.sockfd = sockfd;
wsi->a.protocol = a->vhost->protocols;
lws_vhost_bind_wsi(a->vhost, wsi);
wsi->listener = 1;
if (wsi->a.context->event_loop_ops->init_vhost_listen_wsi)
wsi->a.context->event_loop_ops->init_vhost_listen_wsi(wsi);
pt = &cx->pt[m];
lws_pt_lock(pt, __func__);
if (__insert_wsi_socket_into_fds(cx, wsi)) {
lwsl_notice("inserting wsi socket into fds failed\n");
lws_pt_unlock(pt);
goto bail;
}
lws_dll2_add_tail(&wsi->listen_list, &a->vhost->listen_wsi);
lws_pt_unlock(pt);
#if defined(WIN32) && defined(TCP_FASTOPEN)
if (a->vhost->fo_listen_queue) {
int optval = 1;
if (setsockopt(wsi->desc.sockfd, IPPROTO_TCP,
TCP_FASTOPEN,
(const char*)&optval, sizeof(optval)) < 0) {
#if (_LWS_ENABLED_LOGS & LLL_WARN)
int error = LWS_ERRNO;
lwsl_warn("%s: TCP_NODELAY failed with error %d\n",
__func__, error);
#endif
}
}
#else
#if defined(TCP_FASTOPEN)
if (a->vhost->fo_listen_queue) {
int qlen = a->vhost->fo_listen_queue;
if (setsockopt(wsi->desc.sockfd, SOL_TCP, TCP_FASTOPEN,
&qlen, sizeof(qlen)))
lwsl_warn("%s: TCP_FASTOPEN failed\n", __func__);
}
#endif
#endif
n = listen(wsi->desc.sockfd, LWS_SOMAXCONN);
if (n < 0) {
lwsl_err("listen failed with error %d\n", LWS_ERRNO);
lws_dll2_remove(&wsi->listen_list);
__remove_wsi_socket_from_fds(wsi);
goto bail;
}
if (wsi) {
if (a->info && a->info->vh_listen_sockfd)
a->vhost->listen_port = a->info->port;
__lws_lc_tag(a->vhost->context,
&a->vhost->context->lcg[LWSLCG_WSI],
&wsi->lc, "listen|%s|%s|%d",
a->vhost->name,
a->vhost->iface ? a->vhost->iface : "",
(int)a->vhost->listen_port);
}
} /* for each thread able to independently listen */
if (!lws_check_opt(cx->options, LWS_SERVER_OPTION_EXPLICIT_VHOSTS)) {
#ifdef LWS_WITH_UNIX_SOCK
if (a->af == AF_UNIX)
lwsl_info(" Listening on \"%s\"\n", a->vhost->iface);
else
#endif
lwsl_info(" Listening on %s:%d\n",
a->vhost->iface,
a->vhost->listen_port);
}
// info->port = vhost->listen_port;
return 0;
bail:
lwsl_err("%s: bailing\n", __func__);
compatible_close(sockfd);
return -1;
}
int
_lws_vhost_init_server(const struct lws_context_creation_info *info,
struct lws_vhost *vhost)
{
struct vh_sock_args a;
int n;
a.info = info;
a.vhost = vhost;
if (info) {
vhost->iface = info->iface;
vhost->listen_port = info->port;
}
/* set up our external listening socket we serve on */
if (vhost->listen_port == CONTEXT_PORT_NO_LISTEN ||
vhost->listen_port == CONTEXT_PORT_NO_LISTEN_SERVER)
return 0;
if (info && info->vh_listen_sockfd) {
a.af = AF_UNSPEC;
goto single;
}
/*
* Let's figure out what AF(s) we want this vhost to listen on.
*
* We want AF_UNIX alone if that's what's told
*/
#if defined(LWS_WITH_UNIX_SOCK)
/*
* If unix socket, ask for that and we are done
*/
if (LWS_UNIX_SOCK_ENABLED(vhost)) {
a.af = AF_UNIX;
goto single;
}
#endif
/*
* We may support both ipv4 and ipv6, but get a numeric vhost listen
* iface that is unambiguously ipv4 or ipv6, meaning we can only listen
* for the related AF then.
*/
if (vhost->iface) {
uint8_t buf[16];
int q;
q = lws_parse_numeric_address(vhost->iface, buf, sizeof(buf));
if (q == 4) {
a.af = AF_INET;
goto single;
}
if (q == 16) {
#if defined(LWS_WITH_IPV6)
if (LWS_IPV6_ENABLED(vhost)) {
a.af = AF_INET6;
goto single;
}
#endif
lwsl_err("%s: ipv6 not supported on %s\n", __func__,
vhost->name);
return 1;
}
}
/*
* ... if we make it here, we would want to listen on AF_INET and
* AF_INET6 unless one or the other is forbidden
*/
#if defined(LWS_WITH_IPV6)
if (!(LWS_IPV6_ENABLED(vhost) &&
(vhost->options & LWS_SERVER_OPTION_IPV6_V6ONLY_MODIFY) &&
(vhost->options & LWS_SERVER_OPTION_IPV6_V6ONLY_VALUE))) {
#endif
a.af = AF_INET;
n = _lws_vhost_init_server_af(&a);
if (n)
return n;
#if defined(LWS_WITH_IPV6)
}
if (LWS_IPV6_ENABLED(vhost)) {
a.af = AF_INET6;
goto single;
}
#endif
return 0;
single:
return _lws_vhost_init_server_af(&a);
}
#endif
struct lws_vhost *
lws_select_vhost(struct lws_context *context, int port, const char *servername)
{
struct lws_vhost *vhost = context->vhost_list;
const char *p;
int n, colon;
n = (int)strlen(servername);
colon = n;
p = strchr(servername, ':');
if (p)
colon = lws_ptr_diff(p, servername);
/* Priotity 1: first try exact matches */
while (vhost) {
if (port == vhost->listen_port &&
!strncmp(vhost->name, servername, (unsigned int)colon)) {
lwsl_info("SNI: Found: %s\n", servername);
return vhost;
}
vhost = vhost->vhost_next;
}
/*
* Priority 2: if no exact matches, try matching *.vhost-name
* unintentional matches are possible but resolve to x.com for *.x.com
* which is reasonable. If exact match exists we already chose it and
* never reach here. SSL will still fail it if the cert doesn't allow
* *.x.com.
*/
vhost = context->vhost_list;
while (vhost) {
int m = (int)strlen(vhost->name);
if (port && port == vhost->listen_port &&
m <= (colon - 2) &&
servername[colon - m - 1] == '.' &&
!strncmp(vhost->name, servername + colon - m, (unsigned int)m)) {
lwsl_info("SNI: Found %s on wildcard: %s\n",
servername, vhost->name);
return vhost;
}
vhost = vhost->vhost_next;
}
/* Priority 3: match the first vhost on our port */
vhost = context->vhost_list;
while (vhost) {
if (port && port == vhost->listen_port) {
lwsl_info("%s: vhost match to %s based on port %d\n",
__func__, vhost->name, port);
return vhost;
}
vhost = vhost->vhost_next;
}
/* no match */
return NULL;
}
static const struct lws_mimetype {
const char *extension;
const char *mimetype;
} server_mimetypes[] = {
{ ".html", "text/html" },
{ ".htm", "text/html" },
{ ".js", "text/javascript" },
{ ".css", "text/css" },
{ ".png", "image/png" },
{ ".jpg", "image/jpeg" },
{ ".jpeg", "image/jpeg" },
{ ".ico", "image/x-icon" },
{ ".gif", "image/gif" },
{ ".svg", "image/svg+xml" },
{ ".ttf", "application/x-font-ttf" },
{ ".otf", "application/font-woff" },
{ ".woff", "application/font-woff" },
{ ".woff2", "application/font-woff2" },
{ ".gz", "application/gzip" },
{ ".txt", "text/plain" },
{ ".xml", "application/xml" },
{ ".json", "application/json" },
{ ".mjs", "text/javascript" },
};
const char *
lws_get_mimetype(const char *file, const struct lws_http_mount *m)
{
const struct lws_protocol_vhost_options *pvo;
size_t n = strlen(file), len, i;
const char *fallback_mimetype = NULL;
const struct lws_mimetype *mt;
/* prioritize user-defined mimetypes */
for (pvo = m ? m->extra_mimetypes : NULL; pvo; pvo = pvo->next) {
/* ie, match anything */
if (!fallback_mimetype && pvo->name[0] == '*') {
fallback_mimetype = pvo->value;
continue;
}
len = strlen(pvo->name);
if (n > len && !strcasecmp(&file[n - len], pvo->name)) {
lwsl_info("%s: match to user mimetype: %s\n", __func__,
pvo->value);
return pvo->value;
}
}
/* fallback to server-defined mimetypes */
for (i = 0; i < LWS_ARRAY_SIZE(server_mimetypes); ++i) {
mt = &server_mimetypes[i];
len = strlen(mt->extension);
if (n > len && !strcasecmp(&file[n - len], mt->extension)) {
lwsl_info("%s: match to server mimetype: %s\n", __func__,
mt->mimetype);
return mt->mimetype;
}
}
/* fallback to '*' if defined */
if (fallback_mimetype) {
lwsl_info("%s: match to any mimetype: %s\n", __func__,
fallback_mimetype);
return fallback_mimetype;
}
return NULL;
}
#if defined(LWS_WITH_FILE_OPS)
static lws_fop_flags_t
lws_vfs_prepare_flags(struct lws *wsi)
{
lws_fop_flags_t f = 0;
if (!lws_hdr_total_length(wsi, WSI_TOKEN_HTTP_ACCEPT_ENCODING))
return f;
if (strstr(lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP_ACCEPT_ENCODING),
"gzip")) {
lwsl_info("client indicates GZIP is acceptable\n");
f |= LWS_FOP_FLAG_COMPR_ACCEPTABLE_GZIP;
}
return f;
}
static int
lws_http_serve(struct lws *wsi, char *uri, const char *origin,
const struct lws_http_mount *m)
{
const struct lws_protocol_vhost_options *pvo = m->interpret;
struct lws_process_html_args args;
const char *mimetype;
#if !defined(_WIN32_WCE)
const struct lws_plat_file_ops *fops;
const char *vpath;
lws_fop_flags_t fflags = LWS_O_RDONLY;
#if defined(WIN32) && defined(LWS_HAVE__STAT32I64)
struct _stat32i64 st;
#else
struct stat st;
#endif
int spin = 0;
#endif
char path[256], sym[2048];
unsigned char *p = (unsigned char *)sym + 32 + LWS_PRE, *start = p;
unsigned char *end = p + sizeof(sym) - 32 - LWS_PRE;
#if !defined(WIN32) && !defined(LWS_PLAT_FREERTOS)
size_t len;
#endif
int n;
wsi->handling_404 = 0;
if (!wsi->a.vhost)
return -1;
#if defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2)
if (wsi->a.vhost->http.error_document_404 &&
!strcmp(uri, wsi->a.vhost->http.error_document_404))
wsi->handling_404 = 1;
#endif
lws_snprintf(path, sizeof(path) - 1, "%s/%s", origin, uri);
#if !defined(_WIN32_WCE)
fflags |= lws_vfs_prepare_flags(wsi);
do {
spin++;
fops = lws_vfs_select_fops(wsi->a.context->fops, path, &vpath);
if (wsi->http.fop_fd)
lws_vfs_file_close(&wsi->http.fop_fd);
wsi->http.fop_fd = fops->LWS_FOP_OPEN(fops, wsi->a.context->fops,
path, vpath, &fflags);
if (!wsi->http.fop_fd) {
lwsl_info("%s: Unable to open '%s': errno %d\n",
__func__, path, errno);
return 1;
}
/* if it can't be statted, don't try */
if (fflags & LWS_FOP_FLAG_VIRTUAL)
break;
#if defined(LWS_PLAT_FREERTOS)
break;
#endif
#if !defined(WIN32)
if (fstat(wsi->http.fop_fd->fd, &st)) {
lwsl_info("unable to stat %s\n", path);
goto notfound;
}
#else
#if defined(LWS_HAVE__STAT32I64)
{
WCHAR buf[MAX_PATH];
MultiByteToWideChar(CP_UTF8, 0, path, -1, buf, LWS_ARRAY_SIZE(buf));
if (_wstat32i64(buf, &st)) {
lwsl_info("unable to stat %s\n", path);
goto notfound;
}
}
#else
if (stat(path, &st)) {
lwsl_info("unable to stat %s\n", path);
goto notfound;
}
#endif
#endif
wsi->http.fop_fd->mod_time = (uint32_t)st.st_mtime;
fflags |= LWS_FOP_FLAG_MOD_TIME_VALID;
#if !defined(WIN32) && !defined(LWS_PLAT_FREERTOS)
if ((S_IFMT & st.st_mode) == S_IFLNK) {
len = (size_t)readlink(path, sym, sizeof(sym) - 1);
if (len) {
lwsl_err("Failed to read link %s\n", path);
goto notfound;
}
sym[len] = '\0';
lwsl_debug("symlink %s -> %s\n", path, sym);
lws_snprintf(path, sizeof(path) - 1, "%s", sym);
}
#endif
if ((S_IFMT & st.st_mode) == S_IFDIR) {
lwsl_debug("default filename append to dir\n");
lws_snprintf(path, sizeof(path) - 1, "%s/%s/%s",
origin, uri, m->def ? m->def : "index.html");
}
} while ((S_IFMT & st.st_mode) != S_IFREG && spin < 5);
if (spin == 5)
lwsl_err("symlink loop %s \n", path);
n = sprintf(sym, "%08llX%08lX",
(unsigned long long)lws_vfs_get_length(wsi->http.fop_fd),
(unsigned long)lws_vfs_get_mod_time(wsi->http.fop_fd));
/* disable ranges if IF_RANGE token invalid */
if (lws_hdr_total_length(wsi, WSI_TOKEN_HTTP_IF_RANGE))
if (strcmp(sym, lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP_IF_RANGE)))
/* differs - defeat Range: */
wsi->http.ah->frag_index[WSI_TOKEN_HTTP_RANGE] = 0;
if (lws_hdr_total_length(wsi, WSI_TOKEN_HTTP_IF_NONE_MATCH)) {
/*
* he thinks he has some version of it already,
* check if the tag matches
*/
if (!strcmp(sym, lws_hdr_simple_ptr(wsi,
WSI_TOKEN_HTTP_IF_NONE_MATCH))) {
char cache_control[50], *cc = "no-store";
int cclen = 8;
lwsl_debug("%s: ETAG match %s %s\n", __func__,
uri, origin);
/* we don't need to send the payload */
if (lws_add_http_header_status(wsi,
HTTP_STATUS_NOT_MODIFIED, &p, end)) {
lwsl_err("%s: failed adding not modified\n",
__func__);
return -1;
}
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_ETAG,
(unsigned char *)sym, n, &p, end))
return -1;
/* but we still need to send cache control... */
if (m->cache_max_age && m->cache_reusable) {
if (!m->cache_revalidate) {
cc = cache_control;
cclen = sprintf(cache_control,
"%s, max-age=%u",
intermediates[wsi->cache_intermediaries],
m->cache_max_age);
} else {
cc = cache_control;
cclen = sprintf(cache_control,
"must-revalidate, %s, max-age=%u",
intermediates[wsi->cache_intermediaries],
m->cache_max_age);
}
}
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_CACHE_CONTROL,
(unsigned char *)cc, cclen, &p, end))
return -1;
if (lws_finalize_http_header(wsi, &p, end))
return -1;
n = lws_write(wsi, start, lws_ptr_diff_size_t(p, start),
LWS_WRITE_HTTP_HEADERS |
LWS_WRITE_H2_STREAM_END);
if (n != lws_ptr_diff(p, start)) {
lwsl_err("_write returned %d from %ld\n", n,
(long)(p - start));
return -1;
}
lws_vfs_file_close(&wsi->http.fop_fd);
if (lws_http_transaction_completed(wsi))
return -1;
return 0;
}
}
if (lws_add_http_header_by_token(wsi, WSI_TOKEN_HTTP_ETAG,
(unsigned char *)sym, n, &p, end))
return -1;
#endif
mimetype = lws_get_mimetype(path, m);
if (!mimetype) {
lwsl_info("unknown mimetype for %s\n", path);
if (lws_return_http_status(wsi,
HTTP_STATUS_UNSUPPORTED_MEDIA_TYPE, NULL) ||
lws_http_transaction_completed(wsi))
return -1;
return 0;
}
if (!mimetype[0])
lwsl_debug("sending no mimetype for %s\n", path);
wsi->sending_chunked = 0;
wsi->interpreting = 0;
/*
* check if this is in the list of file suffixes to be interpreted by
* a protocol
*/
while (pvo) {
n = (int)strlen(path);
if (n > (int)strlen(pvo->name) &&
!strcmp(&path[(unsigned int)n - strlen(pvo->name)], pvo->name)) {
wsi->interpreting = 1;
if (!wsi->mux_substream)
wsi->sending_chunked = 1;
wsi->protocol_interpret_idx = (char)(
lws_vhost_name_to_protocol(wsi->a.vhost,
pvo->value) -
&lws_get_vhost(wsi)->protocols[0]);
lwsl_debug("want %s interpreted by %s (pcol is %s)\n", path,
wsi->a.vhost->protocols[
(int)wsi->protocol_interpret_idx].name,
wsi->a.protocol->name);
if (lws_bind_protocol(wsi, &wsi->a.vhost->protocols[
(int)wsi->protocol_interpret_idx], __func__))
return -1;
if (lws_ensure_user_space(wsi))
return -1;
break;
}
pvo = pvo->next;
}
if (wsi->sending_chunked) {
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_TRANSFER_ENCODING,
(unsigned char *)"chunked", 7,
&p, end))
return -1;
}
if (m->protocol) {
const struct lws_protocols *pp = lws_vhost_name_to_protocol(
wsi->a.vhost, m->protocol);
if (lws_bind_protocol(wsi, pp, __func__))
return -1;
args.p = (char *)p;
args.max_len = lws_ptr_diff(end, p);
if (pp->callback(wsi, LWS_CALLBACK_ADD_HEADERS,
wsi->user_space, &args, 0))
return -1;
p = (unsigned char *)args.p;
}
*p = '\0';
n = lws_serve_http_file(wsi, path, mimetype, (char *)start,
lws_ptr_diff(p, start));
if (n < 0 || ((n > 0) && lws_http_transaction_completed(wsi)))
return -1; /* error or can't reuse connection: close the socket */
return 0;
notfound:
return 1;
}
#endif
#if defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2)
const struct lws_http_mount *
lws_find_mount(struct lws *wsi, const char *uri_ptr, int uri_len)
{
const struct lws_http_mount *hm, *hit = NULL;
int best = 0;
hm = wsi->a.vhost->http.mount_list;
while (hm) {
if (uri_len >= hm->mountpoint_len &&
!strncmp(uri_ptr, hm->mountpoint, hm->mountpoint_len) &&
(uri_ptr[hm->mountpoint_len] == '\0' ||
uri_ptr[hm->mountpoint_len] == '/' ||
hm->mountpoint_len == 1)
) {
#if defined(LWS_WITH_SYS_METRICS)
lws_metrics_tag_wsi_add(wsi, "mnt", hm->mountpoint);
#endif
if (hm->origin_protocol == LWSMPRO_NO_MOUNT)
return NULL;
if (hm->origin_protocol == LWSMPRO_CALLBACK ||
((hm->origin_protocol == LWSMPRO_CGI ||
lws_hdr_total_length(wsi, WSI_TOKEN_GET_URI) ||
lws_hdr_total_length(wsi, WSI_TOKEN_POST_URI) ||
#if defined(LWS_WITH_HTTP_UNCOMMON_HEADERS)
lws_hdr_total_length(wsi, WSI_TOKEN_PUT_URI) ||
lws_hdr_total_length(wsi, WSI_TOKEN_PATCH_URI) ||
lws_hdr_total_length(wsi, WSI_TOKEN_DELETE_URI) ||
#endif
lws_hdr_total_length(wsi, WSI_TOKEN_HEAD_URI) ||
#if defined(LWS_ROLE_H2)
(wsi->mux_substream &&
lws_hdr_total_length(wsi,
WSI_TOKEN_HTTP_COLON_PATH)) ||
#endif
hm->protocol) &&
hm->mountpoint_len > best)) {
best = hm->mountpoint_len;
hit = hm;
}
}
hm = hm->mount_next;
}
return hit;
}
#endif
#if defined(LWS_WITH_HTTP_BASIC_AUTH) && !defined(LWS_PLAT_FREERTOS) && defined(LWS_WITH_FILE_OPS)
static int
lws_find_string_in_file(const char *filename, const char *string, int stringlen)
{
char buf[128];
int fd, match = 0, pos = 0, n = 0, hit = 0;
fd = lws_open(filename, O_RDONLY);
if (fd < 0) {
lwsl_err("can't open auth file: %s\n", filename);
return 0;
}
while (1) {
if (pos == n) {
n = (int)read(fd, buf, sizeof(buf));
if (n <= 0) {
if (match == stringlen)
hit = 1;
break;
}
pos = 0;
}
if (match == stringlen) {
if (buf[pos] == '\r' || buf[pos] == '\n') {
hit = 1;
break;
}
match = 0;
}
if (buf[pos] == string[match])
match++;
else
match = 0;
pos++;
}
close(fd);
return hit;
}
#endif
#if defined(LWS_WITH_HTTP_BASIC_AUTH)
int
lws_unauthorised_basic_auth(struct lws *wsi)
{
struct lws_context_per_thread *pt = &wsi->a.context->pt[(int)wsi->tsi];
unsigned char *start = pt->serv_buf + LWS_PRE,
*p = start, *end = p + 2048;
char buf[64];
int n;
/* no auth... tell him it is required */
if (lws_add_http_header_status(wsi, HTTP_STATUS_UNAUTHORIZED, &p, end))
return -1;
n = lws_snprintf(buf, sizeof(buf), "Basic realm=\"lwsws\"");
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_WWW_AUTHENTICATE,
(unsigned char *)buf, n, &p, end))
return -1;
if (lws_add_http_header_content_length(wsi, 0, &p, end))
return -1;
if (lws_finalize_http_header(wsi, &p, end))
return -1;
n = lws_write(wsi, start, lws_ptr_diff_size_t(p, start), LWS_WRITE_HTTP_HEADERS |
LWS_WRITE_H2_STREAM_END);
if (n < 0)
return -1;
return lws_http_transaction_completed(wsi);
}
#endif
int lws_clean_url(char *p)
{
if (p[0] == 'h' && p[1] == 't' && p[2] == 't' && p[3] == 'p') {
p += 4;
if (*p == 's')
p++;
if (*p == ':') {
p++;
if (*p == '/')
p++;
}
}
while (*p) {
if (p[0] == '/' && p[1] == '/') {
char *p1 = p;
while (*p1) {
*p1 = p1[1];
p1++;
}
continue;
}
p++;
}
return 0;
}
static const unsigned char methods[] = {
WSI_TOKEN_GET_URI,
WSI_TOKEN_POST_URI,
#if defined(LWS_WITH_HTTP_UNCOMMON_HEADERS)
WSI_TOKEN_OPTIONS_URI,
WSI_TOKEN_PUT_URI,
WSI_TOKEN_PATCH_URI,
WSI_TOKEN_DELETE_URI,
#endif
WSI_TOKEN_CONNECT,
WSI_TOKEN_HEAD_URI,
#ifdef LWS_WITH_HTTP2
WSI_TOKEN_HTTP_COLON_PATH,
#endif
};
int
lws_http_get_uri_and_method(struct lws *wsi, char **puri_ptr, int *puri_len)
{
int n, count = 0;
for (n = 0; n < (int)LWS_ARRAY_SIZE(methods); n++)
if (lws_hdr_total_length(wsi, methods[n]))
count++;
if (!count) {
lwsl_warn("Missing URI in HTTP request\n");
return -1;
}
if (count != 1 &&
!((wsi->mux_substream || wsi->h2_stream_carries_ws)
#if defined(LWS_ROLE_H2)
&&
lws_hdr_total_length(wsi, WSI_TOKEN_HTTP_COLON_PATH)
#endif
)) {
lwsl_warn("multiple methods?\n");
return -1;
}
for (n = 0; n < (int)LWS_ARRAY_SIZE(methods); n++)
if (lws_hdr_total_length(wsi, methods[n])) {
*puri_ptr = lws_hdr_simple_ptr(wsi, methods[n]);
*puri_len = lws_hdr_total_length(wsi, methods[n]);
return n;
}
return -1;
}
#if defined(LWS_WITH_HTTP_BASIC_AUTH)
enum lws_check_basic_auth_results
lws_check_basic_auth(struct lws *wsi, const char *basic_auth_login_file,
unsigned int auth_mode)
{
#if defined(LWS_WITH_FILE_OPS)
char b64[160], plain[(sizeof(b64) * 3) / 4], *pcolon;
int m, ml, fi, bar;
if (!basic_auth_login_file && auth_mode == LWSAUTHM_DEFAULT)
return LCBA_CONTINUE;
/* Did he send auth? */
ml = lws_hdr_total_length(wsi, WSI_TOKEN_HTTP_AUTHORIZATION);
if (!ml)
return LCBA_FAILED_AUTH;
/* Disallow fragmentation monkey business */
fi = wsi->http.ah->frag_index[WSI_TOKEN_HTTP_AUTHORIZATION];
if (wsi->http.ah->frags[fi].nfrag) {
lwsl_err("fragmented basic auth header not allowed\n");
return LCBA_FAILED_AUTH;
}
m = lws_hdr_copy(wsi, b64, sizeof(b64),
WSI_TOKEN_HTTP_AUTHORIZATION);
if (m < 7) {
lwsl_err("b64 auth too long\n");
return LCBA_END_TRANSACTION;
}
b64[5] = '\0';
if (strcasecmp(b64, "Basic")) {
lwsl_err("auth missing basic: %s\n", b64);
return LCBA_END_TRANSACTION;
}
/* It'll be like Authorization: Basic QWxhZGRpbjpPcGVuU2VzYW1l */
m = lws_b64_decode_string(b64 + 6, plain, sizeof(plain) - 1);
if (m < 0) {
lwsl_err("plain auth too long\n");
return LCBA_END_TRANSACTION;
}
plain[m] = '\0';
pcolon = strchr(plain, ':');
if (!pcolon) {
lwsl_err("basic auth format broken\n");
return LCBA_END_TRANSACTION;
}
switch (auth_mode) {
case LWSAUTHM_DEFAULT:
if (lws_find_string_in_file(basic_auth_login_file, plain, m))
break;
lwsl_err("%s: basic auth lookup failed\n", __func__);
return LCBA_FAILED_AUTH;
case LWSAUTHM_BASIC_AUTH_CALLBACK:
bar = wsi->a.protocol->callback(wsi,
LWS_CALLBACK_VERIFY_BASIC_AUTHORIZATION,
wsi->user_space, plain, (unsigned int)m);
if (!bar)
return LCBA_FAILED_AUTH;
break;
default:
/* Invalid auth mode so lets fail all authentication attempts */
return LCBA_FAILED_AUTH;
}
/*
* Rewrite WSI_TOKEN_HTTP_AUTHORIZATION so it is just the
* authorized username
*/
*pcolon = '\0';
wsi->http.ah->frags[fi].len = (uint16_t)lws_ptr_diff_size_t(pcolon, &plain[0]);
pcolon = lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP_AUTHORIZATION);
strncpy(pcolon, plain, (unsigned int)(ml - 1));
pcolon[ml - 1] = '\0';
lwsl_info("%s: basic auth accepted for %s\n", __func__,
lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP_AUTHORIZATION));
return LCBA_CONTINUE;
#else
if (!basic_auth_login_file && auth_mode == LWSAUTHM_DEFAULT)
return LCBA_CONTINUE;
return LCBA_FAILED_AUTH;
#endif
}
#endif
#if defined(LWS_WITH_HTTP_PROXY)
/*
* Set up an onward http proxy connection according to the mount this
* uri falls under. Notice this can also be starting the proxying of what was
* originally an incoming h1 upgrade, or an h2 ws "upgrade".
*/
int
lws_http_proxy_start(struct lws *wsi, const struct lws_http_mount *hit,
char *uri_ptr, char ws)
{
char ads[96], host[96], *pcolon, *pslash, unix_skt = 0;
struct lws_client_connect_info i;
struct lws *cwsi;
int n, na;
unsigned int max_http_header_data = wsi->a.context->max_http_header_data > 256 ?
wsi->a.context->max_http_header_data : 256;
char *rpath = NULL;
#if defined(LWS_ROLE_WS)
if (ws)
/*
* Neither our inbound ws upgrade request side, nor our onward
* ws client connection on our side can bind to the actual
* protocol that only the remote inbound side and the remote
* onward side understand.
*
* Instead these are both bound to our built-in "lws-ws-proxy"
* protocol, which understands how to proxy between the two
* sides.
*
* We bind the parent, inbound part here and our side of the
* onward client connection is bound to the same handler using
* the .local_protocol_name.
*/
lws_bind_protocol(wsi, &lws_ws_proxy, __func__);
#endif
memset(&i, 0, sizeof(i));
i.context = lws_get_context(wsi);
if (hit->origin[0] == '+')
unix_skt = 1;
pcolon = strchr(hit->origin, ':');
pslash = strchr(hit->origin, '/');
if (!pslash) {
lwsl_err("Proxy mount origin '%s' must have /\n", hit->origin);
return -1;
}
if (unix_skt) {
if (!pcolon) {
lwsl_err("Proxy mount origin for unix skt must "
"have address delimited by :\n");
return -1;
}
n = lws_ptr_diff(pcolon, hit->origin);
pslash = pcolon;
} else {
if (pcolon > pslash)
pcolon = NULL;
if (pcolon)
n = (int)(pcolon - hit->origin);
else
n = (int)(pslash - hit->origin);
if (n >= (int)sizeof(ads) - 2)
n = sizeof(ads) - 2;
}
memcpy(ads, hit->origin, (unsigned int)n);
ads[n] = '\0';
i.address = ads;
i.port = 80;
if (hit->origin_protocol == LWSMPRO_HTTPS) {
i.port = 443;
i.ssl_connection = 1;
}
if (pcolon)
i.port = atoi(pcolon + 1);
rpath = lws_malloc(max_http_header_data, __func__);
if (!rpath)
return -1;
/* rpath needs cleaning after this... ---> */
n = lws_snprintf(rpath, max_http_header_data - 1, "/%s/%s",
pslash + 1, uri_ptr + hit->mountpoint_len) - 1;
lws_clean_url(rpath);
n = (int)strlen(rpath);
if (n && rpath[n - 1] == '/')
n--;
na = lws_hdr_total_length(wsi, WSI_TOKEN_HTTP_URI_ARGS);
if (na) {
char *p;
int budg;
if (!n) /* don't start with the ?... use the first / if so */
n++;
p = rpath + n;
if (na >= (int)max_http_header_data - n - 2) {
lwsl_info("%s: query string %d longer "
"than we can handle\n", __func__,
na);
lws_free(rpath);
return -1;
}
*p++ = '?';
budg = lws_hdr_copy(wsi, p,
(int)(&rpath[max_http_header_data - 1] - p),
WSI_TOKEN_HTTP_URI_ARGS);
if (budg > 0)
p += budg;
*p = '\0';
}
i.path = rpath;
lwsl_notice("%s: proxied path '%s'\n", __func__, i.path);
/* incoming may be h1 or h2... if he sends h1 HOST, use that
* directly, otherwise we must convert h2 :authority to h1
* host */
#if 0
i.host = NULL;
#if defined(LWS_ROLE_H2)
n = lws_hdr_total_length(wsi, WSI_TOKEN_HTTP_COLON_AUTHORITY);
if (n > 0)
i.host = lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP_COLON_AUTHORITY);
else
#endif
{
n = lws_hdr_total_length(wsi, WSI_TOKEN_HOST);
if (n > 0)
i.host = lws_hdr_simple_ptr(wsi, WSI_TOKEN_HOST);
}
#endif
i.host = i.address;
#if 0
if (i.address[0] != '+' ||
!lws_hdr_simple_ptr(wsi, WSI_TOKEN_HOST))
i.host = i.address;
else
i.host = lws_hdr_simple_ptr(wsi, WSI_TOKEN_HOST);
#endif
i.origin = NULL;
if (!ws) {
if (lws_hdr_simple_ptr(wsi, WSI_TOKEN_POST_URI)
#if defined(LWS_WITH_HTTP2)
|| (
lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP_COLON_METHOD) &&
!strcmp(lws_hdr_simple_ptr(wsi,
WSI_TOKEN_HTTP_COLON_METHOD), "post")
)
#endif
)
i.method = "POST";
else if (lws_hdr_simple_ptr(wsi, WSI_TOKEN_PUT_URI)
#if defined(LWS_WITH_HTTP2)
|| (
lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP_COLON_METHOD) &&
!strcmp(lws_hdr_simple_ptr(wsi,
WSI_TOKEN_HTTP_COLON_METHOD), "put")
)
#endif
)
i.method = "PUT";
else if (lws_hdr_simple_ptr(wsi, WSI_TOKEN_PATCH_URI)
#if defined(LWS_WITH_HTTP2)
|| (
lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP_COLON_METHOD) &&
!strcmp(lws_hdr_simple_ptr(wsi,
WSI_TOKEN_HTTP_COLON_METHOD), "patch")
)
#endif
)
i.method = "PATCH";
else if (lws_hdr_simple_ptr(wsi, WSI_TOKEN_DELETE_URI)
#if defined(LWS_WITH_HTTP2)
|| (
lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP_COLON_METHOD) &&
!strcmp(lws_hdr_simple_ptr(wsi,
WSI_TOKEN_HTTP_COLON_METHOD), "delete")
)
#endif
)
i.method = "DELETE";
else
i.method = "GET";
}
if (i.host)
lws_snprintf(host, sizeof(host), "%s:%u", i.host,
wsi->a.vhost->listen_port);
else
lws_snprintf(host, sizeof(host), "%s:%d", i.address, i.port);
i.host = host;
i.alpn = "http/1.1";
i.parent_wsi = wsi;
i.pwsi = &cwsi;
#if defined(LWS_ROLE_WS)
i.protocol = lws_hdr_simple_ptr(wsi, WSI_TOKEN_PROTOCOL);
if (ws)
i.local_protocol_name = "lws-ws-proxy";
#endif
// i.uri_replace_from = hit->origin;
// i.uri_replace_to = hit->mountpoint;
lwsl_info("proxying to %s port %d url %s, ssl %d, from %s, to %s\n",
i.address, i.port, i.path, i.ssl_connection,
i.uri_replace_from, i.uri_replace_to);
if (!lws_client_connect_via_info(&i)) {
lwsl_err("proxy connect fail\n");
/*
* ... we can't do the proxy action, but we can
* cleanly return him a 503 and a description
*/
lws_return_http_status(wsi,
HTTP_STATUS_SERVICE_UNAVAILABLE,
"<h1>Service Temporarily Unavailable</h1>"
"The server is temporarily unable to service "
"your request due to maintenance downtime or "
"capacity problems. Please try again later.");
lws_free(rpath);
return 1;
}
lws_free(rpath);
lwsl_info("%s: setting proxy clientside on %s (parent %s)\n",
__func__, lws_wsi_tag(cwsi), lws_wsi_tag(lws_get_parent(cwsi)));
cwsi->http.proxy_clientside = 1;
if (ws) {
wsi->proxied_ws_parent = 1;
cwsi->h1_ws_proxied = 1;
if (i.protocol) {
lwsl_debug("%s: (requesting '%s')\n",
__func__, i.protocol);
}
}
return 0;
}
#endif
static const char * const oprot[] = {
"http://", "https://"
};
static int
lws_http_redirect_hit(struct lws_context_per_thread *pt, struct lws *wsi,
const struct lws_http_mount *hit, char *uri_ptr,
int uri_len, int *h)
{
char *s;
int n;
*h = 0;
s = uri_ptr + hit->mountpoint_len;
/*
* if we have a mountpoint like https://xxx.com/yyy
* there is an implied / at the end for our purposes since
* we can only mount on a "directory".
*
* But if we just go with that, the browser cannot understand
* that he is actually looking down one "directory level", so
* even though we give him /yyy/abc.html he acts like the
* current directory level is /. So relative urls like "x.png"
* wrongly look outside the mountpoint.
*
* Therefore if we didn't come in on a url with an explicit
* / at the end, we must redirect to add it so the browser
* understands he is one "directory level" down.
*/
if ((hit->mountpoint_len > 1 ||
(hit->origin_protocol == LWSMPRO_REDIR_HTTP ||
hit->origin_protocol == LWSMPRO_REDIR_HTTPS)) &&
(*s != '/' ||
(hit->origin_protocol == LWSMPRO_REDIR_HTTP ||
hit->origin_protocol == LWSMPRO_REDIR_HTTPS)) &&
(hit->origin_protocol != LWSMPRO_CGI &&
hit->origin_protocol != LWSMPRO_CALLBACK)) {
unsigned char *start = pt->serv_buf + LWS_PRE, *p = start,
*end = p + wsi->a.context->pt_serv_buf_size -
LWS_PRE - 512;
*h = 1;
lwsl_info("Doing 301 '%s' org %s\n", s, hit->origin);
/* > at start indicates deal with by redirect */
if (hit->origin_protocol == LWSMPRO_REDIR_HTTP ||
hit->origin_protocol == LWSMPRO_REDIR_HTTPS)
n = lws_snprintf((char *)end, 256, "%s%s",
oprot[hit->origin_protocol & 1],
hit->origin);
else {
if (!lws_hdr_total_length(wsi, WSI_TOKEN_HOST)) {
#if defined(LWS_ROLE_H2)
if (!lws_hdr_total_length(wsi,
WSI_TOKEN_HTTP_COLON_AUTHORITY))
#endif
goto bail_nuke_ah;
#if defined(LWS_ROLE_H2)
n = lws_snprintf((char *)end, 256,
"%s%s%s/", oprot[!!lws_is_ssl(wsi)],
lws_hdr_simple_ptr(wsi,
WSI_TOKEN_HTTP_COLON_AUTHORITY),
uri_ptr);
#endif
} else
n = lws_snprintf((char *)end, 256,
"%s%s%s/", oprot[!!lws_is_ssl(wsi)],
lws_hdr_simple_ptr(wsi, WSI_TOKEN_HOST),
uri_ptr);
}
lws_clean_url((char *)end);
n = lws_http_redirect(wsi, HTTP_STATUS_MOVED_PERMANENTLY,
end, n, &p, end);
if ((int)n < 0)
goto bail_nuke_ah;
return lws_http_transaction_completed(wsi);
}
return 0;
bail_nuke_ah:
lws_header_table_detach(wsi, 1);
return 1;
}
int
lws_http_action(struct lws *wsi)
{
struct lws_context_per_thread *pt = &wsi->a.context->pt[(int)wsi->tsi];
int uri_len = 0, meth, m, http_version_len, ha;
const struct lws_http_mount *hit = NULL;
enum http_version request_version;
struct lws_process_html_args args;
enum http_conn_type conn_type;
char content_length_str[32];
char http_version_str[12];
char http_conn_str[25];
char *uri_ptr = NULL;
#if defined(LWS_WITH_FILE_OPS)
char *s;
#endif
unsigned int n;
meth = lws_http_get_uri_and_method(wsi, &uri_ptr, &uri_len);
if (meth < 0 || meth >= (int)LWS_ARRAY_SIZE(method_names))
goto bail_nuke_ah;
lws_metrics_tag_wsi_add(wsi, "vh", wsi->a.vhost->name);
lws_metrics_tag_wsi_add(wsi, "meth", method_names[meth]);
/* we insist on absolute paths */
if (!uri_ptr || uri_ptr[0] != '/') {
lws_return_http_status(wsi, HTTP_STATUS_FORBIDDEN, NULL);
goto bail_nuke_ah;
}
lwsl_info("Method: '%s' (%d), request for '%s'\n", method_names[meth],
meth, uri_ptr);
if (wsi->role_ops &&
lws_rops_fidx(wsi->role_ops, LWS_ROPS_check_upgrades))
switch (lws_rops_func_fidx(wsi->role_ops,
LWS_ROPS_check_upgrades).
check_upgrades(wsi)) {
case LWS_UPG_RET_DONE:
return 0;
case LWS_UPG_RET_CONTINUE:
break;
case LWS_UPG_RET_BAIL:
goto bail_nuke_ah;
}
if (lws_ensure_user_space(wsi))
goto bail_nuke_ah;
/* HTTP header had a content length? */
wsi->http.rx_content_length = 0;
wsi->http.content_length_explicitly_zero = 0;
if (lws_hdr_total_length(wsi, WSI_TOKEN_POST_URI)
#if defined(LWS_WITH_HTTP_UNCOMMON_HEADERS)
||
lws_hdr_total_length(wsi, WSI_TOKEN_PATCH_URI) ||
lws_hdr_total_length(wsi, WSI_TOKEN_PUT_URI)
#endif
)
wsi->http.rx_content_length = 100 * 1024 * 1024;
if (lws_hdr_total_length(wsi, WSI_TOKEN_HTTP_CONTENT_LENGTH) &&
lws_hdr_copy(wsi, content_length_str,
sizeof(content_length_str) - 1,
WSI_TOKEN_HTTP_CONTENT_LENGTH) > 0) {
wsi->http.rx_content_remain = wsi->http.rx_content_length =
(lws_filepos_t)atoll(content_length_str);
if (!wsi->http.rx_content_length) {
wsi->http.content_length_explicitly_zero = 1;
lwsl_debug("%s: explicit 0 content-length\n", __func__);
}
}
if (wsi->mux_substream) {
wsi->http.request_version = HTTP_VERSION_2;
} else {
/* http_version? Default to 1.0, override with token: */
request_version = HTTP_VERSION_1_0;
/* Works for single digit HTTP versions. : */
http_version_len = lws_hdr_total_length(wsi, WSI_TOKEN_HTTP);
if (http_version_len > 7 &&
lws_hdr_copy(wsi, http_version_str,
sizeof(http_version_str) - 1,
WSI_TOKEN_HTTP) > 0 &&
http_version_str[5] == '1' && http_version_str[7] == '1')
request_version = HTTP_VERSION_1_1;
wsi->http.request_version = request_version;
/* HTTP/1.1 defaults to "keep-alive", 1.0 to "close" */
if (request_version == HTTP_VERSION_1_1)
conn_type = HTTP_CONNECTION_KEEP_ALIVE;
else
conn_type = HTTP_CONNECTION_CLOSE;
/* Override default if http "Connection:" header: */
if (lws_hdr_total_length(wsi, WSI_TOKEN_CONNECTION) &&
lws_hdr_copy(wsi, http_conn_str, sizeof(http_conn_str) - 1,
WSI_TOKEN_CONNECTION) > 0) {
http_conn_str[sizeof(http_conn_str) - 1] = '\0';
if (!strcasecmp(http_conn_str, "keep-alive"))
conn_type = HTTP_CONNECTION_KEEP_ALIVE;
else
if (!strcasecmp(http_conn_str, "close"))
conn_type = HTTP_CONNECTION_CLOSE;
}
wsi->http.conn_type = conn_type;
}
n = (unsigned int)wsi->a.protocol->callback(wsi, LWS_CALLBACK_FILTER_HTTP_CONNECTION,
wsi->user_space, uri_ptr, (unsigned int)uri_len);
if (n) {
lwsl_info("LWS_CALLBACK_HTTP closing\n");
return 1;
}
/*
* if there is content supposed to be coming,
* put a timeout on it having arrived
*/
if (!wsi->mux_stream_immortal)
lws_set_timeout(wsi, PENDING_TIMEOUT_HTTP_CONTENT,
(int)wsi->a.context->timeout_secs);
#if defined(LWS_WITH_TLS)
if (wsi->tls.redirect_to_https) {
/*
* We accepted http:// only so we could redirect to
* https://, so issue the redirect. Create the redirection
* URI from the host: header, and regenerate the path part from
* the parsed pieces
*/
unsigned char *start = pt->serv_buf + LWS_PRE, *p = start,
*end = p + wsi->a.context->pt_serv_buf_size -
LWS_PRE;
n = (unsigned int)lws_hdr_total_length(wsi, WSI_TOKEN_HOST);
if (!n || n > 128)
goto bail_nuke_ah;
if (!lws_hdr_simple_ptr(wsi, WSI_TOKEN_HOST))
goto bail_nuke_ah;
p += lws_snprintf((char *)p, lws_ptr_diff_size_t(end, p), "https://");
memcpy(p, lws_hdr_simple_ptr(wsi, WSI_TOKEN_HOST), n);
p += n;
*p++ = '/';
if (uri_len >= lws_ptr_diff(end, p))
goto bail_nuke_ah;
if (uri_ptr[0])
p--;
memcpy(p, uri_ptr, (unsigned int)uri_len);
p += uri_len;
n = 0;
while (lws_hdr_copy_fragment(wsi, (char *)p + 1,
lws_ptr_diff(end, p) - 2,
WSI_TOKEN_HTTP_URI_ARGS, (int)n) > 0) {
*p = n ? '&' : '?';
p += strlen((char *)p);
if (p >= end - 2)
goto bail_nuke_ah;
n++;
}
n = (unsigned int)lws_ptr_diff(p, start);
p += LWS_PRE;
n = (unsigned int)lws_http_redirect(wsi, HTTP_STATUS_MOVED_PERMANENTLY,
start, (int)n, &p, end);
if ((int)n < 0)
goto bail_nuke_ah;
return lws_http_transaction_completed(wsi);
}
#endif
#ifdef LWS_WITH_ACCESS_LOG
lws_prepare_access_log_info(wsi, uri_ptr, uri_len, meth);
#endif
/* can we serve it from the mount list? */
wsi->mount_hit = 0;
hit = lws_find_mount(wsi, uri_ptr, uri_len);
if (!hit) {
/* deferred cleanup and reset to protocols[0] */
lwsl_info("no hit\n");
if (lws_bind_protocol(wsi, &wsi->a.vhost->protocols[0],
"no mount hit"))
return 1;
lwsi_set_state(wsi, LRS_DOING_TRANSACTION);
m = wsi->a.protocol->callback(wsi, LWS_CALLBACK_HTTP,
wsi->user_space, uri_ptr, (unsigned int)uri_len);
goto after;
}
wsi->mount_hit = 1;
#if defined(LWS_WITH_FILE_OPS)
s = uri_ptr + hit->mountpoint_len;
#endif
n = (unsigned int)lws_http_redirect_hit(pt, wsi, hit, uri_ptr, uri_len, &ha);
if (ha)
return (int)n;
#if defined(LWS_WITH_HTTP_BASIC_AUTH)
/* basic auth? */
switch (lws_check_basic_auth(wsi, hit->basic_auth_login_file,
hit->auth_mask & AUTH_MODE_MASK)) {
case LCBA_CONTINUE:
break;
case LCBA_FAILED_AUTH:
return lws_unauthorised_basic_auth(wsi);
case LCBA_END_TRANSACTION:
lws_return_http_status(wsi, HTTP_STATUS_FORBIDDEN, NULL);
return lws_http_transaction_completed(wsi);
}
#endif
#if defined(LWS_WITH_HTTP_PROXY)
/*
* The mount is a reverse proxy?
*/
// if (hit)
// lwsl_notice("%s: origin_protocol: %d\n", __func__, hit->origin_protocol);
//else
// lwsl_notice("%s: no hit\n", __func__);
if (hit->origin_protocol == LWSMPRO_HTTPS ||
hit->origin_protocol == LWSMPRO_HTTP) {
n = (unsigned int)lws_http_proxy_start(wsi, hit, uri_ptr, 0);
// lwsl_notice("proxy start says %d\n", n);
if (n)
return (int)n;
goto deal_body;
}
#endif
/*
* A particular protocol callback is mounted here?
*
* For the duration of this http transaction, bind us to the
* associated protocol
*/
if (hit->origin_protocol == LWSMPRO_CALLBACK || hit->protocol) {
const struct lws_protocols *pp;
const char *name = hit->origin;
if (hit->protocol)
name = hit->protocol;
pp = lws_vhost_name_to_protocol(wsi->a.vhost, name);
if (!pp) {
lwsl_err("Unable to find plugin '%s'\n",
name);
return 1;
}
if (lws_bind_protocol(wsi, pp, "http action CALLBACK bind"))
return 1;
lwsl_debug("%s: %s, checking access rights for mask 0x%x\n",
__func__, hit->origin, hit->auth_mask);
args.p = uri_ptr;
args.len = uri_len;
args.max_len = hit->auth_mask & ~AUTH_MODE_MASK;
args.final = 0; /* used to signal callback dealt with it */
args.chunked = 0;
n = (unsigned int)wsi->a.protocol->callback(wsi,
LWS_CALLBACK_CHECK_ACCESS_RIGHTS,
wsi->user_space, &args, 0);
if (n) {
lws_return_http_status(wsi, HTTP_STATUS_UNAUTHORIZED,
NULL);
goto bail_nuke_ah;
}
if (args.final) /* callback completely handled it well */
return 0;
if (hit->cgienv && wsi->a.protocol->callback(wsi,
LWS_CALLBACK_HTTP_PMO,
wsi->user_space, (void *)hit->cgienv, 0))
return 1;
if (lws_hdr_total_length(wsi, WSI_TOKEN_POST_URI)) {
m = wsi->a.protocol->callback(wsi, LWS_CALLBACK_HTTP,
wsi->user_space,
uri_ptr + hit->mountpoint_len,
(unsigned int)uri_len - hit->mountpoint_len);
goto after;
}
}
#ifdef LWS_WITH_CGI
/* did we hit something with a cgi:// origin? */
if (hit->origin_protocol == LWSMPRO_CGI) {
const char *cmd[] = {
NULL, /* replace with cgi path */
NULL
};
lwsl_debug("%s: cgi\n", __func__);
cmd[0] = hit->origin;
n = 5;
if (hit->cgi_timeout)
n = (unsigned int)hit->cgi_timeout;
n = (unsigned int)lws_cgi(wsi, cmd, hit->mountpoint_len, (int)n,
hit->cgienv);
if (n) {
lwsl_err("%s: cgi failed\n", __func__);
return -1;
}
goto deal_body;
}
#endif
#if defined(LWS_WITH_FILE_OPS)
n = (unsigned int)(uri_len - lws_ptr_diff(s, uri_ptr));
if (s[0] == '\0' || (n == 1 && s[n - 1] == '/'))
s = (char *)hit->def;
if (!s)
s = "index.html";
#endif
wsi->cache_secs = (unsigned int)hit->cache_max_age;
wsi->cache_reuse = hit->cache_reusable;
wsi->cache_revalidate = hit->cache_revalidate;
wsi->cache_no = hit->cache_no;
wsi->cache_intermediaries = hit->cache_intermediaries;
#if defined(LWS_WITH_FILE_OPS)
m = 1;
if (hit->origin_protocol == LWSMPRO_FILE)
m = lws_http_serve(wsi, s, hit->origin, hit);
if (m > 0)
#endif
{
lwsl_notice("%s: hit->protocol %s\n", __func__, hit->protocol);
/*
* lws_return_http_status(wsi, HTTP_STATUS_NOT_FOUND, NULL);
*/
if (hit->protocol) {
const struct lws_protocols *pp =
lws_vhost_name_to_protocol(
wsi->a.vhost, hit->protocol);
lwsl_notice("%s: pp %p\n", __func__, pp);
/* coverity */
if (!pp)
return 1;
lwsi_set_state(wsi, LRS_DOING_TRANSACTION);
if (lws_bind_protocol(wsi, pp, "http_action HTTP"))
return 1;
m = pp->callback(wsi, LWS_CALLBACK_HTTP,
wsi->user_space,
uri_ptr + hit->mountpoint_len,
(size_t)(uri_len - hit->mountpoint_len));
} else
m = wsi->a.protocol->callback(wsi, LWS_CALLBACK_HTTP,
wsi->user_space, uri_ptr, (size_t)uri_len);
}
after:
if (m) {
lwsl_info("LWS_CALLBACK_HTTP closing\n");
return 1;
}
#if defined(LWS_WITH_CGI) || defined(LWS_WITH_HTTP_PROXY)
deal_body:
#endif
/*
* If we're not issuing a file, check for content_length or
* HTTP keep-alive. No keep-alive header allocation for
* ISSUING_FILE, as this uses HTTP/1.0.
*
* In any case, return 0 and let lws_read decide how to
* proceed based on state
*/
if (lwsi_state(wsi) == LRS_ISSUING_FILE)
return 0;
/* Prepare to read body if we have a content length: */
lwsl_debug("wsi->http.rx_content_length %lld %d %d\n",
(long long)wsi->http.rx_content_length,
wsi->upgraded_to_http2, wsi->mux_substream);
if (wsi->http.content_length_explicitly_zero &&
lws_hdr_total_length(wsi, WSI_TOKEN_POST_URI)) {
/*
* POST with an explicit content-length of zero
*
* If we don't give the user code the empty HTTP_BODY callback,
* he may become confused to hear the HTTP_BODY_COMPLETION (due
* to, eg, instantiation of lws_spa never happened).
*
* HTTP_BODY_COMPLETION is responsible for sending the result
* status code and result body if any, and to do the transaction
* complete processing.
*/
if (wsi->a.protocol->callback(wsi, LWS_CALLBACK_HTTP_BODY,
wsi->user_space, NULL, 0))
return 1;
if (wsi->a.protocol->callback(wsi, LWS_CALLBACK_HTTP_BODY_COMPLETION,
wsi->user_space, NULL, 0))
return 1;
return 0;
}
if (wsi->http.rx_content_length <= 0)
return 0;
if (lwsi_state(wsi) != LRS_DISCARD_BODY) {
lwsi_set_state(wsi, LRS_BODY);
lwsl_info("%s: %s: LRS_BODY state set (0x%x)\n", __func__,
lws_wsi_tag(wsi), (int)wsi->wsistate);
}
wsi->http.rx_content_remain = wsi->http.rx_content_length;
/*
* At this point we have transitioned from deferred
* action to expecting BODY on the stream wsi, if it's
* in a bundle like h2. So if the stream wsi has its
* own buflist, we need to deal with that first.
*/
while (1) {
struct lws_tokens ebuf;
int m;
ebuf.len = (int)lws_buflist_next_segment_len(&wsi->buflist,
&ebuf.token);
if (!ebuf.len)
break;
lwsl_debug("%s: consuming %d\n", __func__, (int)ebuf.len);
m = lws_read_h1(wsi, ebuf.token, (lws_filepos_t)ebuf.len);
if (m < 0)
return -1;
if (lws_buflist_aware_finished_consuming(wsi, &ebuf, m, 1,
__func__))
return -1;
}
return 0;
bail_nuke_ah:
lws_header_table_detach(wsi, 1);
return 1;
}
int
lws_confirm_host_header(struct lws *wsi)
{
struct lws_tokenize ts;
lws_tokenize_elem e;
int port = 80, n;
char buf[128];
/*
* this vhost wants us to validate what the
* client sent against our vhost name
*/
if (!lws_hdr_total_length(wsi, WSI_TOKEN_HOST)) {
lwsl_info("%s: missing host on upgrade\n", __func__);
return 1;
}
#if defined(LWS_WITH_TLS)
if (wsi->tls.ssl)
port = 443;
#endif
n = lws_hdr_copy(wsi, buf, sizeof(buf) - 1, WSI_TOKEN_HOST);
if (n <= 0) {
lwsl_info("%s: missing or oversize host header\n", __func__);
return 1;
}
ts.len = (size_t)n;
lws_tokenize_init(&ts, buf, LWS_TOKENIZE_F_DOT_NONTERM /* server.com */|
LWS_TOKENIZE_F_NO_FLOATS /* 1.server.com */|
LWS_TOKENIZE_F_MINUS_NONTERM /* a-b.com */);
if (lws_tokenize(&ts) != LWS_TOKZE_TOKEN)
goto bad_format;
if (strncmp(ts.token, wsi->a.vhost->name, ts.token_len)) {
buf[(size_t)(ts.token - buf) + ts.token_len] = '\0';
lwsl_info("%s: '%s' in host hdr but vhost name %s\n",
__func__, ts.token, wsi->a.vhost->name);
return 1;
}
e = lws_tokenize(&ts);
if (e == LWS_TOKZE_DELIMITER && ts.token[0] == ':') {
if (lws_tokenize(&ts) != LWS_TOKZE_INTEGER)
goto bad_format;
else
port = atoi(ts.token);
} else
if (e != LWS_TOKZE_ENDED)
goto bad_format;
if (wsi->a.vhost->listen_port != port &&
wsi->a.vhost->listen_port != CONTEXT_PORT_NO_LISTEN_SERVER) {
lwsl_info("%s: host port %d mismatches vhost port %d\n",
__func__, port, wsi->a.vhost->listen_port);
return 1;
}
lwsl_debug("%s: host header OK\n", __func__);
return 0;
bad_format:
lwsl_info("%s: bad host header format\n", __func__);
return 1;
}
#if defined(LWS_WITH_SERVER)
int
lws_http_to_fallback(struct lws *wsi, unsigned char *obuf, size_t olen)
{
const struct lws_role_ops *role = &role_ops_raw_skt;
const struct lws_protocols *p1, *protocol =
&wsi->a.vhost->protocols[wsi->a.vhost->raw_protocol_index];
char ipbuf[64];
int n;
if (wsi->a.vhost->listen_accept_role &&
lws_role_by_name(wsi->a.vhost->listen_accept_role))
role = lws_role_by_name(wsi->a.vhost->listen_accept_role);
if (wsi->a.vhost->listen_accept_protocol) {
p1 = lws_vhost_name_to_protocol(wsi->a.vhost,
wsi->a.vhost->listen_accept_protocol);
if (p1)
protocol = p1;
}
lws_bind_protocol(wsi, protocol, __func__);
lws_role_transition(wsi, LWSIFR_SERVER, LRS_ESTABLISHED, role);
lws_header_table_detach(wsi, 0);
lws_set_timeout(wsi, NO_PENDING_TIMEOUT, 0);
n = LWS_CALLBACK_SERVER_NEW_CLIENT_INSTANTIATED;
if (wsi->role_ops->adoption_cb[1])
n = wsi->role_ops->adoption_cb[1];
ipbuf[0] = '\0';
#if !defined(LWS_PLAT_OPTEE)
lws_get_peer_simple(wsi, ipbuf, sizeof(ipbuf));
#endif
lwsl_notice("%s: vh %s, peer: %s, role %s, "
"protocol %s, cb %d, ah %p\n", __func__, wsi->a.vhost->name,
ipbuf, role ? role->name : "null", protocol->name, n,
wsi->http.ah);
if ((wsi->a.protocol->callback)(wsi, (enum lws_callback_reasons)n, wsi->user_space, NULL, 0))
return 1;
n = LWS_CALLBACK_RAW_RX;
if (wsi->role_ops->rx_cb[lwsi_role_server(wsi)])
n = wsi->role_ops->rx_cb[lwsi_role_server(wsi)];
if (wsi->a.protocol->callback(wsi, (enum lws_callback_reasons)n, wsi->user_space, obuf, olen))
return 1;
return 0;
}
int
lws_handshake_server(struct lws *wsi, unsigned char **buf, size_t len)
{
struct lws_context *context = lws_get_context(wsi);
struct lws_context_per_thread *pt = &context->pt[(int)wsi->tsi];
#if defined(LWS_WITH_HTTP2)
struct allocated_headers *ah;
#endif
unsigned char *obuf = *buf;
#if defined(LWS_WITH_HTTP2)
char tbuf[128], *p;
#endif
size_t olen = len;
int n = 0, m, i;
if (len >= 10000000) {
lwsl_err("%s: assert: len %ld\n", __func__, (long)len);
assert(0);
}
if (!wsi->http.ah) {
lwsl_err("%s: assert: NULL ah\n", __func__);
assert(0);
}
while (len) {
if (!lwsi_role_server(wsi) || !lwsi_role_http(wsi)) {
lwsl_err("%s: bad wsi role 0x%x\n", __func__,
(int)lwsi_role(wsi));
goto bail_nuke_ah;
}
i = (int)len;
m = lws_parse(wsi, *buf, &i);
lwsl_info("%s: parsed count %d\n", __func__, (int)len - i);
(*buf) += (int)len - i;
len = (unsigned int)i;
if (m == LPR_DO_FALLBACK) {
/*
* http parser went off the rails and
* LWS_SERVER_OPTION_FALLBACK_TO_APPLY_LISTEN_
* ACCEPT_CONFIG is set on this vhost.
*
* We are transitioning from http with an AH, to
* a backup role (raw-skt, by default). Drop
* the ah, bind to the role with mode as
* ESTABLISHED.
*/
raw_transition:
if (lws_http_to_fallback(wsi, obuf, olen)) {
lwsl_info("%s: fallback -> close\n", __func__);
goto bail_nuke_ah;
}
(*buf) = obuf + olen;
return 0;
}
if (m) {
lwsl_info("lws_parse failed\n");
goto bail_nuke_ah;
}
/* coverity... */
if (!wsi->http.ah)
goto bail_nuke_ah;
if (wsi->http.ah->parser_state != WSI_PARSING_COMPLETE)
continue;
lwsl_parser("%s: lws_parse sees parsing complete\n", __func__);
/* select vhost */
if (wsi->a.vhost->listen_port &&
lws_hdr_total_length(wsi, WSI_TOKEN_HOST)) {
struct lws_vhost *vhost = lws_select_vhost(
context, wsi->a.vhost->listen_port,
lws_hdr_simple_ptr(wsi, WSI_TOKEN_HOST));
if (vhost)
lws_vhost_bind_wsi(vhost, wsi);
} else
lwsl_info("no host\n");
if ((!lwsi_role_h2(wsi) || !lwsi_role_server(wsi)) &&
(!wsi->conn_stat_done))
wsi->conn_stat_done = 1;
/* check for unwelcome guests */
#if defined(LWS_WITH_HTTP_UNCOMMON_HEADERS)
if (wsi->a.context->reject_service_keywords) {
const struct lws_protocol_vhost_options *rej =
wsi->a.context->reject_service_keywords;
char ua[384], *msg = NULL;
if (lws_hdr_copy(wsi, ua, sizeof(ua) - 1,
WSI_TOKEN_HTTP_USER_AGENT) > 0) {
#ifdef LWS_WITH_ACCESS_LOG
char *uri_ptr = NULL;
int meth, uri_len;
#endif
ua[sizeof(ua) - 1] = '\0';
while (rej) {
if (!strstr(ua, rej->name)) {
rej = rej->next;
continue;
}
msg = strchr(rej->value, ' ');
if (msg)
msg++;
lws_return_http_status(wsi,
(unsigned int)atoi(rej->value), msg);
#ifdef LWS_WITH_ACCESS_LOG
meth = lws_http_get_uri_and_method(wsi,
&uri_ptr, &uri_len);
if (meth >= 0)
lws_prepare_access_log_info(wsi,
uri_ptr, uri_len, meth);
/* wsi close will do the log */
#endif
/*
* We don't want anything from
* this rejected guy. Follow
* the close flow, not the
* transaction complete flow.
*/
goto bail_nuke_ah;
}
}
}
#endif
/*
* So he may have come to us requesting one or another kind
* of upgrade from http... but we may want to redirect him at
* http level. In that case, we need to check the redirect
* situation even though he's not actually wanting http and
* prioritize returning that if there is one.
*/
{
const struct lws_http_mount *hit = NULL;
int uri_len = 0, ha, n;
char *uri_ptr = NULL;
n = lws_http_get_uri_and_method(wsi, &uri_ptr, &uri_len);
if (n >= 0) {
hit = lws_find_mount(wsi, uri_ptr, uri_len);
if (hit) {
n = lws_http_redirect_hit(pt, wsi, hit, uri_ptr,
uri_len, &ha);
if (ha)
return n;
}
}
}
if (lws_hdr_total_length(wsi, WSI_TOKEN_CONNECT)) {
lwsl_info("Changing to RAW mode\n");
goto raw_transition;
}
lwsi_set_state(wsi, LRS_PRE_WS_SERVING_ACCEPT);
lws_set_timeout(wsi, NO_PENDING_TIMEOUT, 0);
if (lws_hdr_total_length(wsi, WSI_TOKEN_UPGRADE)) {
const char *up = lws_hdr_simple_ptr(wsi,
WSI_TOKEN_UPGRADE);
if (strcasecmp(up, "websocket") &&
strcasecmp(up, "h2c")) {
lwsl_info("Unknown upgrade '%s'\n", up);
if (lws_return_http_status(wsi,
HTTP_STATUS_FORBIDDEN, NULL) ||
lws_http_transaction_completed(wsi))
goto bail_nuke_ah;
}
n = user_callback_handle_rxflow(wsi->a.protocol->callback,
wsi, LWS_CALLBACK_HTTP_CONFIRM_UPGRADE,
wsi->user_space, (char *)up, 0);
/* just hang up? */
if (n < 0)
goto bail_nuke_ah;
/* callback returned headers already, do t_c? */
if (n > 0) {
if (lws_http_transaction_completed(wsi))
goto bail_nuke_ah;
/* continue on */
return 0;
}
/* callback said 0, it was allowed */
if (wsi->a.vhost->options &
LWS_SERVER_OPTION_VHOST_UPG_STRICT_HOST_CHECK &&
lws_confirm_host_header(wsi))
goto bail_nuke_ah;
if (!strcasecmp(up, "websocket")) {
#if defined(LWS_ROLE_WS)
lws_metrics_tag_wsi_add(wsi, "upg", "ws");
lwsl_info("Upgrade to ws\n");
goto upgrade_ws;
#endif
}
#if defined(LWS_WITH_HTTP2)
if (!strcasecmp(up, "h2c")) {
lws_metrics_tag_wsi_add(wsi, "upg", "h2c");
lwsl_info("Upgrade to h2c\n");
goto upgrade_h2c;
}
#endif
}
/* no upgrade ack... he remained as HTTP */
lwsl_info("%s: %s: No upgrade\n", __func__, lws_wsi_tag(wsi));
lwsi_set_state(wsi, LRS_ESTABLISHED);
#if defined(LWS_WITH_FILE_OPS)
wsi->http.fop_fd = NULL;
#endif
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
lws_http_compression_validate(wsi);
#endif
lwsl_debug("%s: %s: ah %p\n", __func__, lws_wsi_tag(wsi),
(void *)wsi->http.ah);
n = lws_http_action(wsi);
return n;
#if defined(LWS_WITH_HTTP2)
upgrade_h2c:
if (!lws_hdr_total_length(wsi, WSI_TOKEN_HTTP2_SETTINGS)) {
lwsl_info("missing http2_settings\n");
goto bail_nuke_ah;
}
lwsl_info("h2c upgrade...\n");
p = lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP2_SETTINGS);
/* convert the peer's HTTP-Settings */
n = lws_b64_decode_string(p, tbuf, sizeof(tbuf));
if (n < 0) {
lwsl_parser("HTTP2_SETTINGS too long\n");
return 1;
}
wsi->upgraded_to_http2 = 1;
/* adopt the header info */
ah = wsi->http.ah;
lws_role_transition(wsi, LWSIFR_SERVER, LRS_H2_AWAIT_PREFACE,
&role_ops_h2);
/* http2 union member has http union struct at start */
wsi->http.ah = ah;
if (!wsi->h2.h2n) {
wsi->h2.h2n = lws_zalloc(sizeof(*wsi->h2.h2n), "h2n");
if (!wsi->h2.h2n)
return 1;
}
lws_h2_init(wsi);
/* HTTP2 union */
lws_h2_settings(wsi, &wsi->h2.h2n->peer_set, (uint8_t *)tbuf, n);
if (lws_hpack_dynamic_size(wsi, (int)wsi->h2.h2n->peer_set.s[
H2SET_HEADER_TABLE_SIZE]))
return 1;
strcpy(tbuf, "HTTP/1.1 101 Switching Protocols\x0d\x0a"
"Connection: Upgrade\x0d\x0a"
"Upgrade: h2c\x0d\x0a\x0d\x0a");
m = (int)strlen(tbuf);
n = lws_issue_raw(wsi, (unsigned char *)tbuf, (unsigned int)m);
if (n != m) {
lwsl_debug("http2 switch: ERROR writing to socket\n");
return 1;
}
return 0;
#endif
#if defined(LWS_ROLE_WS)
upgrade_ws:
if (lws_process_ws_upgrade(wsi))
goto bail_nuke_ah;
return 0;
#endif
} /* while all chars are handled */
return 0;
bail_nuke_ah:
/* drop the header info */
lws_header_table_detach(wsi, 1);
return 1;
}
#endif
int LWS_WARN_UNUSED_RESULT
lws_http_transaction_completed(struct lws *wsi)
{
int n;
if (wsi->http.cgi_transaction_complete)
return 0;
if (lws_has_buffered_out(wsi)
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
|| wsi->http.comp_ctx.buflist_comp ||
wsi->http.comp_ctx.may_have_more
#endif
) {
/*
* ...so he tried to send something large as the http reply,
* it went as a partial, but he immediately said the
* transaction was completed.
*
* Defer the transaction completed until the last part of the
* partial is sent.
*/
lwsl_debug("%s: %s: deferring due to partial\n", __func__,
lws_wsi_tag(wsi));
wsi->http.deferred_transaction_completed = 1;
lws_callback_on_writable(wsi);
return 0;
}
/*
* Are we finishing the transaction before we have consumed any body?
*
* For h1 this would kill keepalive pipelining, and for h2, considering
* it can extend over multiple DATA frames, it would kill the network
* connection.
*/
if (wsi->http.rx_content_length && wsi->http.rx_content_remain) {
/*
* are we already in LRS_DISCARD_BODY and didn't clear the
* remaining before trying to complete the transaction again?
*/
if (lwsi_state(wsi) == LRS_DISCARD_BODY)
return -1;
/*
* let's defer transaction completed processing until we
* discarded the remaining body
*/
lwsi_set_state(wsi, LRS_DISCARD_BODY);
return 0;
}
#if defined(LWS_WITH_SYS_METRICS)
{
char tmp[10];
lws_snprintf(tmp, sizeof(tmp), "%u", wsi->http.response_code);
lws_metrics_tag_wsi_add(wsi, "status", tmp);
}
#endif
lwsl_info("%s: %s\n", __func__, lws_wsi_tag(wsi));
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
lws_http_compression_destroy(wsi);
#endif
lws_access_log(wsi);
if (!wsi->hdr_parsing_completed
#if defined(LWS_WITH_CGI)
&& !wsi->http.cgi
#endif
) {
char peer[64];
#if !defined(LWS_PLAT_OPTEE)
lws_get_peer_simple(wsi, peer, sizeof(peer) - 1);
#else
peer[0] = '\0';
#endif
peer[sizeof(peer) - 1] = '\0';
lwsl_info("%s: (from %s) ignoring, ah parsing incomplete\n",
__func__, peer);
return 0;
}
#if defined(LWS_WITH_CGI)
if (wsi->http.cgi) {
lwsl_debug("%s: cleaning cgi\n", __func__);
wsi->http.cgi_transaction_complete = 1;
lws_cgi_remove_and_kill(wsi);
lws_spawn_piped_destroy(&wsi->http.cgi->lsp);
lws_sul_cancel(&wsi->http.cgi->sul_grace);
lws_free_set_NULL(wsi->http.cgi);
wsi->http.cgi_transaction_complete = 0;
}
#endif
/* if we can't go back to accept new headers, drop the connection */
if (wsi->mux_substream)
return 1;
if (wsi->seen_zero_length_recv)
return 1;
if (wsi->http.conn_type != HTTP_CONNECTION_KEEP_ALIVE) {
lwsl_info("%s: %s: close connection\n", __func__, lws_wsi_tag(wsi));
return 1;
}
if (lws_bind_protocol(wsi, &wsi->a.vhost->protocols[0], __func__))
return 1;
/*
* otherwise set ourselves up ready to go again, but because we have no
* idea about the wsi writability, we make put it in a holding state
* until we can verify POLLOUT. The part of this that confirms POLLOUT
* with no partials is in lws_server_socket_service() below.
*/
lwsl_debug("%s: %s: setting DEF_ACT from 0x%x: %p\n", __func__,
lws_wsi_tag(wsi), (int)wsi->wsistate, wsi->buflist);
lwsi_set_state(wsi, LRS_DEFERRING_ACTION);
wsi->http.tx_content_length = 0;
wsi->http.tx_content_remain = 0;
wsi->hdr_parsing_completed = 0;
wsi->sending_chunked = 0;
#ifdef LWS_WITH_ACCESS_LOG
wsi->http.access_log.sent = 0;
#endif
#if defined(LWS_WITH_FILE_OPS) && (defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2))
if (lwsi_role_http(wsi) && lwsi_role_server(wsi) &&
wsi->http.fop_fd != NULL)
lws_vfs_file_close(&wsi->http.fop_fd);
#endif
n = NO_PENDING_TIMEOUT;
if (wsi->a.vhost->keepalive_timeout)
n = PENDING_TIMEOUT_HTTP_KEEPALIVE_IDLE;
lws_set_timeout(wsi, (enum pending_timeout)n, wsi->a.vhost->keepalive_timeout);
/*
* We already know we are on http1.1 / keepalive and the next thing
* coming will be another header set.
*
* If there is no pending rx and we still have the ah, drop it and
* reacquire a new ah when the new headers start to arrive. (Otherwise
* we needlessly hog an ah indefinitely.)
*
* However if there is pending rx and we know from the keepalive state
* that is already at least the start of another header set, simply
* reset the existing header table and keep it.
*/
if (wsi->http.ah) {
// lws_buflist_describe(&wsi->buflist, wsi, __func__);
if (!lws_buflist_next_segment_len(&wsi->buflist, NULL)) {
lwsl_debug("%s: %s: nothing in buflist, detaching ah\n",
__func__, lws_wsi_tag(wsi));
lws_header_table_detach(wsi, 1);
#ifdef LWS_WITH_TLS
/*
* additionally... if we are hogging an SSL instance
* with no pending pipelined headers (or ah now), and
* SSL is scarce, drop this connection without waiting
*/
if (wsi->a.vhost->tls.use_ssl &&
wsi->a.context->simultaneous_ssl_restriction &&
wsi->a.context->simultaneous_ssl ==
wsi->a.context->simultaneous_ssl_restriction) {
lwsl_info("%s: simultaneous_ssl_restriction\n",
__func__);
return 1;
}
#endif
} else {
lwsl_info("%s: %s: resetting/keeping ah as pipeline\n",
__func__, lws_wsi_tag(wsi));
lws_header_table_reset(wsi, 0);
/*
* If we kept the ah, we should restrict the amount
* of time we are willing to keep it. Otherwise it
* will be bound the whole time the connection remains
* open.
*/
lws_set_timeout(wsi, PENDING_TIMEOUT_HOLDING_AH,
wsi->a.vhost->keepalive_timeout);
}
/* If we're (re)starting on headers, need other implied init */
if (wsi->http.ah)
wsi->http.ah->ues = URIES_IDLE;
//lwsi_set_state(wsi, LRS_ESTABLISHED); // !!!
} else
if (lws_buflist_next_segment_len(&wsi->buflist, NULL))
if (lws_header_table_attach(wsi, 0))
lwsl_debug("acquired ah\n");
lwsl_debug("%s: %s: keep-alive await new transaction (state 0x%x)\n",
__func__, lws_wsi_tag(wsi), (int)wsi->wsistate);
lws_callback_on_writable(wsi);
return 0;
}
#if defined(LWS_WITH_FILE_OPS)
int
lws_serve_http_file(struct lws *wsi, const char *file, const char *content_type,
const char *other_headers, int other_headers_len)
{
struct lws_context *context = lws_get_context(wsi);
struct lws_context_per_thread *pt = &context->pt[(int)wsi->tsi];
unsigned char *response = pt->serv_buf + LWS_PRE;
#if defined(LWS_WITH_RANGES)
struct lws_range_parsing *rp = &wsi->http.range;
#endif
int ret = 0, cclen = 8, n = HTTP_STATUS_OK;
char cache_control[50], *cc = "no-store";
lws_fop_flags_t fflags = LWS_O_RDONLY;
const struct lws_plat_file_ops *fops;
lws_filepos_t total_content_length;
unsigned char *p = response;
unsigned char *end = p + context->pt_serv_buf_size - LWS_PRE;
const char *vpath;
#if defined(LWS_WITH_RANGES)
int ranges;
#endif
if (wsi->handling_404)
n = HTTP_STATUS_NOT_FOUND;
/*
* We either call the platform fops .open with first arg platform fops,
* or we call fops_zip .open with first arg platform fops, and fops_zip
* open will decide whether to switch to fops_zip or stay with fops_def.
*
* If wsi->http.fop_fd is already set, the caller already opened it
*/
if (!wsi->http.fop_fd) {
fops = lws_vfs_select_fops(wsi->a.context->fops, file, &vpath);
fflags |= lws_vfs_prepare_flags(wsi);
wsi->http.fop_fd = fops->LWS_FOP_OPEN(fops, wsi->a.context->fops,
file, vpath, &fflags);
if (!wsi->http.fop_fd) {
lwsl_info("%s: Unable to open: '%s': errno %d\n",
__func__, file, errno);
if (lws_return_http_status(wsi, HTTP_STATUS_NOT_FOUND,
NULL))
return -1;
return !wsi->mux_substream;
}
}
/*
* Caution... wsi->http.fop_fd is live from here
*/
wsi->http.filelen = lws_vfs_get_length(wsi->http.fop_fd);
total_content_length = wsi->http.filelen;
#if defined(LWS_WITH_RANGES)
ranges = lws_ranges_init(wsi, rp, wsi->http.filelen);
lwsl_debug("Range count %d\n", ranges);
/*
* no ranges -> 200;
* 1 range -> 206 + Content-Type: normal; Content-Range;
* more -> 206 + Content-Type: multipart/byteranges
* Repeat the true Content-Type in each multipart header
* along with Content-Range
*/
if (ranges < 0) {
/* it means he expressed a range in Range:, but it was illegal */
lws_return_http_status(wsi,
HTTP_STATUS_REQ_RANGE_NOT_SATISFIABLE, NULL);
if (lws_http_transaction_completed(wsi))
goto bail; /* <0 means just hang up */
lws_vfs_file_close(&wsi->http.fop_fd);
return 0; /* == 0 means we did the transaction complete */
}
if (ranges)
n = HTTP_STATUS_PARTIAL_CONTENT;
#endif
if (lws_add_http_header_status(wsi, (unsigned int)n, &p, end))
goto bail;
if ((wsi->http.fop_fd->flags & (LWS_FOP_FLAG_COMPR_ACCEPTABLE_GZIP |
LWS_FOP_FLAG_COMPR_IS_GZIP)) ==
(LWS_FOP_FLAG_COMPR_ACCEPTABLE_GZIP | LWS_FOP_FLAG_COMPR_IS_GZIP)) {
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_CONTENT_ENCODING,
(unsigned char *)"gzip", 4, &p, end))
goto bail;
lwsl_info("file is being provided in gzip\n");
}
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
else {
/*
* if we know its very compressible, and we can use
* compression, then use the most preferred compression
* method that the client said he will accept
*/
if (!wsi->interpreting && (
!strncmp(content_type, "text/", 5) ||
!strcmp(content_type, "application/javascript") ||
!strcmp(content_type, "image/svg+xml")))
lws_http_compression_apply(wsi, NULL, &p, end, 0);
}
#endif
if (
#if defined(LWS_WITH_RANGES)
ranges < 2 &&
#endif
content_type && content_type[0])
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_CONTENT_TYPE,
(unsigned char *)content_type,
(int)strlen(content_type),
&p, end))
goto bail;
#if defined(LWS_WITH_RANGES)
if (ranges >= 2) { /* multipart byteranges */
lws_strncpy(wsi->http.multipart_content_type, content_type,
sizeof(wsi->http.multipart_content_type));
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_CONTENT_TYPE,
(unsigned char *)
"multipart/byteranges; "
"boundary=_lws",
20, &p, end))
goto bail;
/*
* our overall content length has to include
*
* - (n + 1) x "_lws\r\n"
* - n x Content-Type: xxx/xxx\r\n
* - n x Content-Range: bytes xxx-yyy/zzz\r\n
* - n x /r/n
* - the actual payloads (aggregated in rp->agg)
*
* Precompute it for the main response header
*/
total_content_length = (lws_filepos_t)rp->agg +
6 /* final _lws\r\n */;
lws_ranges_reset(rp);
while (lws_ranges_next(rp)) {
n = lws_snprintf(cache_control, sizeof(cache_control),
"bytes %llu-%llu/%llu",
rp->start, rp->end, rp->extent);
total_content_length = total_content_length +
(lws_filepos_t)(
6 /* header _lws\r\n */ +
/* Content-Type: xxx/xxx\r\n */
14 + (int)strlen(content_type) + 2 +
/* Content-Range: xxxx\r\n */
15 + n + 2 +
2); /* /r/n */
}
lws_ranges_reset(rp);
lws_ranges_next(rp);
}
if (ranges == 1) {
total_content_length = (lws_filepos_t)rp->agg;
n = lws_snprintf(cache_control, sizeof(cache_control),
"bytes %llu-%llu/%llu",
rp->start, rp->end, rp->extent);
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_CONTENT_RANGE,
(unsigned char *)cache_control,
n, &p, end))
goto bail;
}
wsi->http.range.inside = 0;
if (lws_add_http_header_by_token(wsi, WSI_TOKEN_HTTP_ACCEPT_RANGES,
(unsigned char *)"bytes", 5, &p, end))
goto bail;
#endif
if (!wsi->mux_substream) {
/* for http/1.1 ... */
if (!wsi->sending_chunked
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
&& !wsi->http.lcs
#endif
) {
/* ... if not already using chunked and not using an
* http compression translation, then send the naive
* content length
*/
if (lws_add_http_header_content_length(wsi,
total_content_length, &p, end))
goto bail;
} else {
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
if (wsi->http.lcs) {
/* ...otherwise, for http 1 it must go chunked.
* For the compression case, the reason is we
* compress on the fly and do not know the
* compressed content-length until it has all
* been sent. Http/1.1 pipelining must be able
* to know where the transaction boundaries are
* ... so chunking...
*/
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_TRANSFER_ENCODING,
(unsigned char *)"chunked", 7,
&p, end))
goto bail;
/*
* ...this is fun, isn't it :-) For h1 that is
* using an http compression translation, the
* compressor must chunk its output privately.
*
* h2 doesn't need (or support) any of this
* crap.
*/
lwsl_debug("setting chunking\n");
wsi->http.comp_ctx.chunking = 1;
}
#endif
}
}
if (wsi->cache_no) {
cc = cache_control;
cclen = sprintf(cache_control, "no-cache");
}
else if (wsi->cache_secs && wsi->cache_reuse) {
if (!wsi->cache_revalidate) {
cc = cache_control;
cclen = sprintf(cache_control, "%s, max-age=%u",
intermediates[wsi->cache_intermediaries],
wsi->cache_secs);
} else {
cc = cache_control;
cclen = sprintf(cache_control,
"must-revalidate, %s, max-age=%u",
intermediates[wsi->cache_intermediaries],
wsi->cache_secs);
}
}
/* Only add cache control if its not specified by any other_headers. */
if (!other_headers ||
(!strstr(other_headers, "cache-control") &&
!strstr(other_headers, "Cache-Control"))) {
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_CACHE_CONTROL,
(unsigned char *)cc, cclen, &p, end))
goto bail;
}
if (other_headers) {
if ((end - p) < other_headers_len)
goto bail;
memcpy(p, other_headers, (unsigned int)other_headers_len);
p += other_headers_len;
}
if (lws_finalize_http_header(wsi, &p, end))
goto bail;
ret = lws_write(wsi, response, lws_ptr_diff_size_t(p, response), LWS_WRITE_HTTP_HEADERS);
if (ret != (p - response)) {
lwsl_err("_write returned %d from %ld\n", ret,
(long)(p - response));
goto bail;
}
wsi->http.filepos = 0;
lwsi_set_state(wsi, LRS_ISSUING_FILE);
if (lws_hdr_total_length(wsi, WSI_TOKEN_HEAD_URI)) {
/* we do not emit the body */
lws_vfs_file_close(&wsi->http.fop_fd);
if (lws_http_transaction_completed(wsi))
goto bail;
return 0;
}
lws_callback_on_writable(wsi);
return 0;
bail:
lws_vfs_file_close(&wsi->http.fop_fd);
return -1;
}
#endif
#if defined(LWS_WITH_FILE_OPS)
int lws_serve_http_file_fragment(struct lws *wsi)
{
struct lws_context *context = wsi->a.context;
struct lws_context_per_thread *pt = &context->pt[(int)wsi->tsi];
struct lws_process_html_args args;
lws_filepos_t amount, poss;
unsigned char *p, *pstart;
#if defined(LWS_WITH_RANGES)
unsigned char finished = 0;
#endif
#if defined(LWS_ROLE_H2)
struct lws *nwsi;
#endif
int n, m;
lwsl_debug("wsi->mux_substream %d\n", wsi->mux_substream);
do {
/* priority 1: buffered output */
if (lws_has_buffered_out(wsi)) {
if (lws_issue_raw(wsi, NULL, 0) < 0) {
lwsl_info("%s: closing\n", __func__);
goto file_had_it;
}
break;
}
/* priority 2: buffered pre-compression-transform */
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
if (wsi->http.comp_ctx.buflist_comp ||
wsi->http.comp_ctx.may_have_more) {
enum lws_write_protocol wp = LWS_WRITE_HTTP;
lwsl_info("%s: completing comp partial (buflist %p, may %d)\n",
__func__, wsi->http.comp_ctx.buflist_comp,
wsi->http.comp_ctx.may_have_more);
if (lws_rops_fidx(wsi->role_ops, LWS_ROPS_write_role_protocol) &&
lws_rops_func_fidx(wsi->role_ops, LWS_ROPS_write_role_protocol).
write_role_protocol(wsi, NULL, 0, &wp) < 0) {
lwsl_info("%s signalling to close\n", __func__);
goto file_had_it;
}
lws_callback_on_writable(wsi);
break;
}
#endif
if (wsi->http.filepos == wsi->http.filelen)
goto all_sent;
n = 0;
p = pstart = pt->serv_buf + LWS_H2_FRAME_HEADER_LENGTH;
#if defined(LWS_WITH_RANGES)
if (wsi->http.range.count_ranges && !wsi->http.range.inside) {
lwsl_notice("%s: doing range start %llu\n", __func__,
wsi->http.range.start);
if ((long long)lws_vfs_file_seek_cur(wsi->http.fop_fd,
(lws_fileofs_t)wsi->http.range.start -
(lws_fileofs_t)wsi->http.filepos) < 0)
goto file_had_it;
wsi->http.filepos = wsi->http.range.start;
if (wsi->http.range.count_ranges > 1) {
n = lws_snprintf((char *)p,
context->pt_serv_buf_size -
LWS_H2_FRAME_HEADER_LENGTH,
"_lws\x0d\x0a"
"Content-Type: %s\x0d\x0a"
"Content-Range: bytes "
"%llu-%llu/%llu\x0d\x0a"
"\x0d\x0a",
wsi->http.multipart_content_type,
wsi->http.range.start,
wsi->http.range.end,
wsi->http.range.extent);
p += n;
}
wsi->http.range.budget = wsi->http.range.end -
wsi->http.range.start + 1;
wsi->http.range.inside = 1;
}
#endif
poss = context->pt_serv_buf_size;
#if defined(LWS_ROLE_H2)
/*
* If it's h2, restrict any lump that we are sending to the
* max h2 frame size the peer indicated he could handle in
* his SETTINGS
*/
nwsi = lws_get_network_wsi(wsi);
if (nwsi->h2.h2n &&
poss > (lws_filepos_t)nwsi->h2.h2n->peer_set.s[H2SET_MAX_FRAME_SIZE])
poss = (lws_filepos_t)nwsi->h2.h2n->peer_set.s[H2SET_MAX_FRAME_SIZE];
#endif
poss = poss - (lws_filepos_t)(n + LWS_H2_FRAME_HEADER_LENGTH);
if (wsi->http.tx_content_length)
if (poss > wsi->http.tx_content_remain)
poss = wsi->http.tx_content_remain;
/*
* If there is a hint about how much we will do well to send at
* one time, restrict ourselves to only trying to send that.
*/
if (wsi->a.protocol->tx_packet_size &&
poss > wsi->a.protocol->tx_packet_size)
poss = wsi->a.protocol->tx_packet_size;
if (lws_rops_fidx(wsi->role_ops, LWS_ROPS_tx_credit)) {
lws_filepos_t txc = (unsigned int)lws_rops_func_fidx(wsi->role_ops,
LWS_ROPS_tx_credit).
tx_credit(wsi, LWSTXCR_US_TO_PEER, 0);
if (!txc) {
/*
* We shouldn't've been able to get the
* WRITEABLE if we are skint
*/
lwsl_notice("%s: %s: no tx credit\n", __func__,
lws_wsi_tag(wsi));
return 0;
}
if (txc < poss)
poss = txc;
/*
* Tracking consumption of the actual payload amount
* will be handled when the role data frame is sent...
*/
}
#if defined(LWS_WITH_RANGES)
if (wsi->http.range.count_ranges) {
if (wsi->http.range.count_ranges > 1)
poss -= 7; /* allow for final boundary */
if (poss > wsi->http.range.budget)
poss = wsi->http.range.budget;
}
#endif
if (wsi->sending_chunked) {
/* we need to drop the chunk size in here */
p += 10;
/* allow for the chunk to grow by 128 in translation */
poss -= 10 + 128;
}
amount = 0;
if (lws_vfs_file_read(wsi->http.fop_fd, &amount, p, poss) < 0)
goto file_had_it; /* caller will close */
if (wsi->sending_chunked)
n = (int)amount;
else
n = lws_ptr_diff(p, pstart) + (int)amount;
lwsl_debug("%s: sending %d\n", __func__, n);
if (n) {
lws_set_timeout(wsi, PENDING_TIMEOUT_HTTP_CONTENT,
(int)context->timeout_secs);
if (wsi->interpreting) {
args.p = (char *)p;
args.len = n;
args.max_len = (int)(unsigned int)poss + 128;
args.final = wsi->http.filepos + (unsigned int)n ==
wsi->http.filelen;
args.chunked = wsi->sending_chunked;
if (user_callback_handle_rxflow(
wsi->a.vhost->protocols[
(int)wsi->protocol_interpret_idx].callback,
wsi, LWS_CALLBACK_PROCESS_HTML,
wsi->user_space, &args, 0) < 0)
goto file_had_it;
n = args.len;
p = (unsigned char *)args.p;
} else
p = pstart;
#if defined(LWS_WITH_RANGES)
if (wsi->http.range.send_ctr + 1 ==
wsi->http.range.count_ranges && // last range
wsi->http.range.count_ranges > 1 && // was 2+ ranges (ie, multipart)
wsi->http.range.budget - amount == 0) {// final part
n += lws_snprintf((char *)pstart + n, 6,
"_lws\x0d\x0a"); // append trailing boundary
lwsl_debug("added trailing boundary\n");
}
#endif
m = lws_write(wsi, p, (unsigned int)n, wsi->http.filepos + amount ==
wsi->http.filelen ?
LWS_WRITE_HTTP_FINAL : LWS_WRITE_HTTP);
if (m < 0)
goto file_had_it;
wsi->http.filepos += amount;
#if defined(LWS_WITH_RANGES)
if (wsi->http.range.count_ranges >= 1) {
wsi->http.range.budget -= amount;
if (wsi->http.range.budget == 0) {
lwsl_notice("range budget exhausted\n");
wsi->http.range.inside = 0;
wsi->http.range.send_ctr++;
if (lws_ranges_next(&wsi->http.range) < 1) {
finished = 1;
goto all_sent;
}
}
}
#endif
if (m != n) {
/* adjust for what was not sent */
if (lws_vfs_file_seek_cur(wsi->http.fop_fd,
m - n) ==
(lws_fileofs_t)-1)
goto file_had_it;
}
}
all_sent:
if ((!lws_has_buffered_out(wsi)
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
&& !wsi->http.comp_ctx.buflist_comp &&
!wsi->http.comp_ctx.may_have_more
#endif
) && (wsi->http.filepos >= wsi->http.filelen
#if defined(LWS_WITH_RANGES)
|| finished)
#else
)
#endif
) {
lwsi_set_state(wsi, LRS_ESTABLISHED);
/* we might be in keepalive, so close it off here */
lws_vfs_file_close(&wsi->http.fop_fd);
lwsl_debug("file completed\n");
if (wsi->a.protocol->callback &&
user_callback_handle_rxflow(wsi->a.protocol->callback,
wsi, LWS_CALLBACK_HTTP_FILE_COMPLETION,
wsi->user_space, NULL, 0) < 0) {
/*
* For http/1.x, the choices from
* transaction_completed are either
* 0 to use the connection for pipelined
* or nonzero to hang it up.
*
* However for http/2. while we are
* still interested in hanging up the
* nwsi if there was a network-level
* fatal error, simply completing the
* transaction is a matter of the stream
* state, not the root connection at the
* network level
*/
if (wsi->mux_substream)
return 1;
else
return -1;
}
if (wsi->http.ah)
lws_header_table_reset(wsi, 0);
return 1; /* >0 indicates completed */
}
/*
* while(1) here causes us to spam the whole file contents into
* a hugely bloated output buffer if it ever can't send the
* whole chunk...
*/
} while (!lws_send_pipe_choked(wsi));
lws_callback_on_writable(wsi);
return 0; /* indicates further processing must be done */
file_had_it:
lws_vfs_file_close(&wsi->http.fop_fd);
return -1;
}
#endif
#if defined(LWS_WITH_SERVER)
void
lws_server_get_canonical_hostname(struct lws_context *context,
const struct lws_context_creation_info *info)
{
if (lws_check_opt(info->options,
LWS_SERVER_OPTION_SKIP_SERVER_CANONICAL_NAME))
return;
#if !defined(LWS_PLAT_FREERTOS)
/* find canonical hostname */
if (gethostname((char *)context->canonical_hostname,
sizeof(context->canonical_hostname) - 1))
lws_strncpy((char *)context->canonical_hostname, "unknown",
sizeof(context->canonical_hostname));
lwsl_cx_info(context, " canonical_hostname = %s\n",
context->canonical_hostname);
#else
(void)context;
#endif
}
#endif
int
lws_chunked_html_process(struct lws_process_html_args *args,
struct lws_process_html_state *s)
{
char *sp, buffer[32];
const char *pc;
int old_len, n;
/* do replacements */
sp = args->p;
old_len = args->len;
args->len = 0;
s->start = sp;
while (sp < args->p + old_len) {
if (args->len + 7 >= args->max_len) {
lwsl_err("Used up interpret padding\n");
return -1;
}
if ((!s->pos && *sp == '$') || s->pos) {
int hits = 0, hit = 0;
if (!s->pos)
s->start = sp;
s->swallow[s->pos++] = *sp;
if (s->pos == sizeof(s->swallow) - 1)
goto skip;
for (n = 0; n < s->count_vars; n++)
if (!strncmp(s->swallow, s->vars[n], (unsigned int)s->pos)) {
hits++;
hit = n;
}
if (!hits) {
skip:
s->swallow[s->pos] = '\0';
memcpy(s->start, s->swallow, (unsigned int)s->pos);
args->len++;
s->pos = 0;
sp = s->start + 1;
continue;
}
if (hits == 1 && s->pos == (int)strlen(s->vars[hit])) {
pc = s->replace(s->data, hit);
if (!pc)
pc = "NULL";
n = (int)strlen(pc);
s->swallow[s->pos] = '\0';
if (n != s->pos) {
memmove(s->start + n, s->start + s->pos,
(unsigned int)(old_len - (sp - args->p) - 1));
old_len += (n - s->pos) + 1;
}
memcpy(s->start, pc, (unsigned int)n);
args->len++;
sp = s->start + 1;
s->pos = 0;
}
sp++;
continue;
}
args->len++;
sp++;
}
if (args->chunked) {
/* no space left for final chunk trailer */
if (args->final && args->len + 7 >= args->max_len)
return -1;
n = sprintf(buffer, "%X\x0d\x0a", args->len);
args->p -= n;
memcpy(args->p, buffer, (unsigned int)n);
args->len += n;
if (args->final) {
sp = args->p + args->len;
*sp++ = '\x0d';
*sp++ = '\x0a';
*sp++ = '0';
*sp++ = '\x0d';
*sp++ = '\x0a';
*sp++ = '\x0d';
*sp++ = '\x0a';
args->len += 7;
} else {
sp = args->p + args->len;
*sp++ = '\x0d';
*sp++ = '\x0a';
args->len += 2;
}
}
return 0;
}