libwebsockets/lib/output.c
Andy Green d2feeee1d6 recv revert treating zero as hangup
While checking with ab, I found
commit 30cdb3ac8f
Author: Justin Chen <justinchen00@github.invalid.com>
Date:   Thu Apr 14 21:40:53 2016 +0800

    recv treat zero return as error

    https://github.com/warmcat/libwebsockets/issues/475

turned ab performance to crap, reverting it made everything fast again.

recv manpage says there is three ways to get zero returned

1)       When a stream socket peer has performed an orderly shutdown, the return value will be 0  (the  traditional  "end-of-file"
       return).

2)       Datagram  sockets  in  various  domains  (e.g., the UNIX and Internet domains) permit zero-length datagrams.  When such a
       datagram is received, the return value is 0.

3)       The value 0 may also be returned if the requested number of bytes to receive from a stream socket was 0.

we can't just assume it means the peer shut down.

If the peer shut down, then the event loop should get an event on the socket like POLLHUP and deal with it that way.

So the patch mentioned above is simply reverted here.

Signed-off-by: Andy Green <andy@warmcat.com>
2016-05-05 09:23:05 +08:00

673 lines
18 KiB
C

/*
* libwebsockets - small server side websockets and web server implementation
*
* Copyright (C) 2010-2015 Andy Green <andy@warmcat.com>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation:
* version 2.1 of the License.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
* MA 02110-1301 USA
*/
#include "private-libwebsockets.h"
static int
lws_0405_frame_mask_generate(struct lws *wsi)
{
#if 0
wsi->u.ws.mask[0] = 0;
wsi->u.ws.mask[1] = 0;
wsi->u.ws.mask[2] = 0;
wsi->u.ws.mask[3] = 0;
#else
int n;
/* fetch the per-frame nonce */
n = lws_get_random(lws_get_context(wsi), wsi->u.ws.mask, 4);
if (n != 4) {
lwsl_parser("Unable to read from random device %s %d\n",
SYSTEM_RANDOM_FILEPATH, n);
return 1;
}
#endif
/* start masking from first byte of masking key buffer */
wsi->u.ws.mask_idx = 0;
return 0;
}
#ifdef _DEBUG
LWS_VISIBLE void lwsl_hexdump(void *vbuf, size_t len)
{
unsigned char *buf = (unsigned char *)vbuf;
unsigned int n, m, start;
char line[80];
char *p;
lwsl_parser("\n");
for (n = 0; n < len;) {
start = n;
p = line;
p += sprintf(p, "%04X: ", start);
for (m = 0; m < 16 && n < len; m++)
p += sprintf(p, "%02X ", buf[n++]);
while (m++ < 16)
p += sprintf(p, " ");
p += sprintf(p, " ");
for (m = 0; m < 16 && (start + m) < len; m++) {
if (buf[start + m] >= ' ' && buf[start + m] < 127)
*p++ = buf[start + m];
else
*p++ = '.';
}
while (m++ < 16)
*p++ = ' ';
*p++ = '\n';
*p = '\0';
lwsl_debug("%s", line);
}
lwsl_debug("\n");
}
#endif
/*
* notice this returns number of bytes consumed, or -1
*/
int lws_issue_raw(struct lws *wsi, unsigned char *buf, size_t len)
{
struct lws_context *context = lws_get_context(wsi);
size_t real_len = len;
int n, m;
if (!len)
return 0;
/* just ignore sends after we cleared the truncation buffer */
if (wsi->state == LWSS_FLUSHING_STORED_SEND_BEFORE_CLOSE &&
!wsi->trunc_len)
return len;
if (wsi->trunc_len && (buf < wsi->trunc_alloc ||
buf > (wsi->trunc_alloc + wsi->trunc_len +
wsi->trunc_offset))) {
lwsl_err("****** %x Sending new, pending truncated ...\n", wsi);
assert(0);
}
m = lws_ext_cb_active(wsi, LWS_EXT_CB_PACKET_TX_DO_SEND, &buf, len);
if (m < 0)
return -1;
if (m) /* handled */ {
n = m;
goto handle_truncated_send;
}
if (!lws_socket_is_valid(wsi->sock))
lwsl_warn("** error invalid sock but expected to send\n");
/* nope, send it on the socket directly */
lws_latency_pre(context, wsi);
n = lws_ssl_capable_write(wsi, buf, len);
lws_latency(context, wsi, "send lws_issue_raw", n,
(unsigned int)n == len);
switch (n) {
case LWS_SSL_CAPABLE_ERROR:
/* we're going to close, let close know sends aren't possible */
wsi->socket_is_permanently_unusable = 1;
return -1;
case LWS_SSL_CAPABLE_MORE_SERVICE:
/* nothing got sent, not fatal, retry the whole thing later */
n = 0;
break;
}
handle_truncated_send:
/*
* we were already handling a truncated send?
*/
if (wsi->trunc_len) {
lwsl_info("%p partial adv %d (vs %d)\n", wsi, n, real_len);
wsi->trunc_offset += n;
wsi->trunc_len -= n;
if (!wsi->trunc_len) {
lwsl_info("***** %x partial send completed\n", wsi);
/* done with it, but don't free it */
n = real_len;
if (wsi->state == LWSS_FLUSHING_STORED_SEND_BEFORE_CLOSE) {
lwsl_info("***** %x signalling to close now\n", wsi);
return -1; /* retry closing now */
}
}
/* always callback on writeable */
lws_callback_on_writable(wsi);
return n;
}
if ((unsigned int)n == real_len)
/* what we just sent went out cleanly */
return n;
/*
* Newly truncated send. Buffer the remainder (it will get
* first priority next time the socket is writable)
*/
lwsl_info("%p new partial sent %d from %d total\n", wsi, n, real_len);
/*
* - if we still have a suitable malloc lying around, use it
* - or, if too small, reallocate it
* - or, if no buffer, create it
*/
if (!wsi->trunc_alloc || real_len - n > wsi->trunc_alloc_len) {
lws_free(wsi->trunc_alloc);
wsi->trunc_alloc_len = real_len - n;
wsi->trunc_alloc = lws_malloc(real_len - n);
if (!wsi->trunc_alloc) {
lwsl_err("truncated send: unable to malloc %d\n",
real_len - n);
return -1;
}
}
wsi->trunc_offset = 0;
wsi->trunc_len = real_len - n;
memcpy(wsi->trunc_alloc, buf + n, real_len - n);
/* since something buffered, force it to get another chance to send */
lws_callback_on_writable(wsi);
return real_len;
}
/**
* lws_write() - Apply protocol then write data to client
* @wsi: Websocket instance (available from user callback)
* @buf: The data to send. For data being sent on a websocket
* connection (ie, not default http), this buffer MUST have
* LWS_PRE bytes valid BEFORE the pointer.
* This is so the protocol header data can be added in-situ.
* @len: Count of the data bytes in the payload starting from buf
* @protocol: Use LWS_WRITE_HTTP to reply to an http connection, and one
* of LWS_WRITE_BINARY or LWS_WRITE_TEXT to send appropriate
* data on a websockets connection. Remember to allow the extra
* bytes before and after buf if LWS_WRITE_BINARY or LWS_WRITE_TEXT
* are used.
*
* This function provides the way to issue data back to the client
* for both http and websocket protocols.
*
* In the case of sending using websocket protocol, be sure to allocate
* valid storage before and after buf as explained above. This scheme
* allows maximum efficiency of sending data and protocol in a single
* packet while not burdening the user code with any protocol knowledge.
*
* Return may be -1 for a fatal error needing connection close, or a
* positive number reflecting the amount of bytes actually sent. This
* can be less than the requested number of bytes due to OS memory
* pressure at any given time.
*/
LWS_VISIBLE int lws_write(struct lws *wsi, unsigned char *buf, size_t len,
enum lws_write_protocol wp)
{
struct lws_context_per_thread *pt = &wsi->context->pt[(int)wsi->tsi];
int masked7 = (wsi->mode == LWSCM_WS_CLIENT);
unsigned char is_masked_bit = 0;
unsigned char *dropmask = NULL;
struct lws_tokens eff_buf;
int pre = 0, n;
size_t orig_len = len;
if (wsi->state == LWSS_ESTABLISHED && wsi->u.ws.tx_draining_ext) {
/* remove us from the list */
struct lws **w = &pt->tx_draining_ext_list;
lwsl_debug("%s: TX EXT DRAINING: Remove from list\n", __func__);
wsi->u.ws.tx_draining_ext = 0;
/* remove us from context draining ext list */
while (*w) {
if (*w == wsi) {
*w = wsi->u.ws.tx_draining_ext_list;
break;
}
w = &((*w)->u.ws.tx_draining_ext_list);
}
wsi->u.ws.tx_draining_ext_list = NULL;
wp = (wsi->u.ws.tx_draining_stashed_wp & 0xc0) |
LWS_WRITE_CONTINUATION;
lwsl_ext("FORCED draining wp to 0x%02X\n", wp);
}
if (wp == LWS_WRITE_HTTP ||
wp == LWS_WRITE_HTTP_FINAL ||
wp == LWS_WRITE_HTTP_HEADERS)
goto send_raw;
/* if not in a state to send stuff, then just send nothing */
if (wsi->state != LWSS_ESTABLISHED &&
((wsi->state != LWSS_RETURNED_CLOSE_ALREADY &&
wsi->state != LWSS_AWAITING_CLOSE_ACK) ||
wp != LWS_WRITE_CLOSE))
return 0;
/* if we are continuing a frame that already had its header done */
if (wsi->u.ws.inside_frame) {
lwsl_debug("INSIDE FRAME\n");
goto do_more_inside_frame;
}
wsi->u.ws.clean_buffer = 1;
/*
* give a chance to the extensions to modify payload
* the extension may decide to produce unlimited payload erratically
* (eg, compression extension), so we require only that if he produces
* something, it will be a complete fragment of the length known at
* the time (just the fragment length known), and if he has
* more we will come back next time he is writeable and allow him to
* produce more fragments until he's drained.
*
* This allows what is sent each time it is writeable to be limited to
* a size that can be sent without partial sends or blocking, allows
* interleaving of control frames and other connection service.
*/
eff_buf.token = (char *)buf;
eff_buf.token_len = len;
switch ((int)wp) {
case LWS_WRITE_PING:
case LWS_WRITE_PONG:
case LWS_WRITE_CLOSE:
break;
default:
n = lws_ext_cb_active(wsi, LWS_EXT_CB_PAYLOAD_TX, &eff_buf, wp);
if (n < 0)
return -1;
if (n && eff_buf.token_len) {
/* extension requires further draining */
wsi->u.ws.tx_draining_ext = 1;
wsi->u.ws.tx_draining_ext_list = pt->tx_draining_ext_list;
pt->tx_draining_ext_list = wsi;
/* we must come back to do more */
lws_callback_on_writable(wsi);
/*
* keep a copy of the write type for the overall
* action that has provoked generation of these
* fragments, so the last guy can use its FIN state.
*/
wsi->u.ws.tx_draining_stashed_wp = wp;
/* this is definitely not actually the last fragment
* because the extension asserted he has more coming
* So make sure this intermediate one doesn't go out
* with a FIN.
*/
wp |= LWS_WRITE_NO_FIN;
}
if (eff_buf.token_len && wsi->u.ws.stashed_write_pending) {
wsi->u.ws.stashed_write_pending = 0;
wp = (wp &0xc0) | (int)wsi->u.ws.stashed_write_type;
}
}
/*
* an extension did something we need to keep... for example, if
* compression extension, it has already updated its state according
* to this being issued
*/
if ((char *)buf != eff_buf.token) {
/*
* ext might eat it, but no have anything to issue yet
* in that case we have to follow his lead, but stash and
* replace the write type that was lost here the first time.
*/
if (len && !eff_buf.token_len) {
if (!wsi->u.ws.stashed_write_pending)
wsi->u.ws.stashed_write_type = (char)wp & 0x3f;
wsi->u.ws.stashed_write_pending = 1;
return len;
}
/*
* extension recreated it:
* need to buffer this if not all sent
*/
wsi->u.ws.clean_buffer = 0;
}
buf = (unsigned char *)eff_buf.token;
len = eff_buf.token_len;
switch (wsi->ietf_spec_revision) {
case 13:
if (masked7) {
pre += 4;
dropmask = &buf[0 - pre];
is_masked_bit = 0x80;
}
switch (wp & 0xf) {
case LWS_WRITE_TEXT:
n = LWSWSOPC_TEXT_FRAME;
break;
case LWS_WRITE_BINARY:
n = LWSWSOPC_BINARY_FRAME;
break;
case LWS_WRITE_CONTINUATION:
n = LWSWSOPC_CONTINUATION;
break;
case LWS_WRITE_CLOSE:
n = LWSWSOPC_CLOSE;
break;
case LWS_WRITE_PING:
n = LWSWSOPC_PING;
break;
case LWS_WRITE_PONG:
n = LWSWSOPC_PONG;
break;
default:
lwsl_warn("lws_write: unknown write opc / wp\n");
return -1;
}
if (!(wp & LWS_WRITE_NO_FIN))
n |= 1 << 7;
if (len < 126) {
pre += 2;
buf[-pre] = n;
buf[-pre + 1] = (unsigned char)(len | is_masked_bit);
} else {
if (len < 65536) {
pre += 4;
buf[-pre] = n;
buf[-pre + 1] = 126 | is_masked_bit;
buf[-pre + 2] = (unsigned char)(len >> 8);
buf[-pre + 3] = (unsigned char)len;
} else {
pre += 10;
buf[-pre] = n;
buf[-pre + 1] = 127 | is_masked_bit;
#if defined __LP64__
buf[-pre + 2] = (len >> 56) & 0x7f;
buf[-pre + 3] = len >> 48;
buf[-pre + 4] = len >> 40;
buf[-pre + 5] = len >> 32;
#else
buf[-pre + 2] = 0;
buf[-pre + 3] = 0;
buf[-pre + 4] = 0;
buf[-pre + 5] = 0;
#endif
buf[-pre + 6] = (unsigned char)(len >> 24);
buf[-pre + 7] = (unsigned char)(len >> 16);
buf[-pre + 8] = (unsigned char)(len >> 8);
buf[-pre + 9] = (unsigned char)len;
}
}
break;
}
do_more_inside_frame:
/*
* Deal with masking if we are in client -> server direction and
* the wp demands it
*/
if (masked7) {
if (!wsi->u.ws.inside_frame)
if (lws_0405_frame_mask_generate(wsi)) {
lwsl_err("frame mask generation failed\n");
return -1;
}
/*
* in v7, just mask the payload
*/
if (dropmask) { /* never set if already inside frame */
for (n = 4; n < (int)len + 4; n++)
dropmask[n] = dropmask[n] ^ wsi->u.ws.mask[
(wsi->u.ws.mask_idx++) & 3];
/* copy the frame nonce into place */
memcpy(dropmask, wsi->u.ws.mask, 4);
}
}
send_raw:
switch ((int)wp) {
case LWS_WRITE_CLOSE:
/* lwsl_hexdump(&buf[-pre], len); */
case LWS_WRITE_HTTP:
case LWS_WRITE_HTTP_FINAL:
case LWS_WRITE_HTTP_HEADERS:
case LWS_WRITE_PONG:
case LWS_WRITE_PING:
#ifdef LWS_USE_HTTP2
if (wsi->mode == LWSCM_HTTP2_SERVING) {
unsigned char flags = 0;
n = LWS_HTTP2_FRAME_TYPE_DATA;
if (wp == LWS_WRITE_HTTP_HEADERS) {
n = LWS_HTTP2_FRAME_TYPE_HEADERS;
flags = LWS_HTTP2_FLAG_END_HEADERS;
if (wsi->u.http2.send_END_STREAM)
flags |= LWS_HTTP2_FLAG_END_STREAM;
}
if ((wp == LWS_WRITE_HTTP ||
wp == LWS_WRITE_HTTP_FINAL) &&
wsi->u.http.content_length) {
wsi->u.http.content_remain -= len;
lwsl_info("%s: content_remain = %lu\n", __func__,
wsi->u.http.content_remain);
if (!wsi->u.http.content_remain) {
lwsl_info("%s: selecting final write mode\n", __func__);
wp = LWS_WRITE_HTTP_FINAL;
}
}
if (wp == LWS_WRITE_HTTP_FINAL && wsi->u.http2.END_STREAM) {
lwsl_info("%s: setting END_STREAM\n", __func__);
flags |= LWS_HTTP2_FLAG_END_STREAM;
}
return lws_http2_frame_write(wsi, n, flags,
wsi->u.http2.my_stream_id, len, buf);
}
#endif
return lws_issue_raw(wsi, (unsigned char *)buf - pre, len + pre);
default:
break;
}
/*
* give any active extensions a chance to munge the buffer
* before send. We pass in a pointer to an lws_tokens struct
* prepared with the default buffer and content length that's in
* there. Rather than rewrite the default buffer, extensions
* that expect to grow the buffer can adapt .token to
* point to their own per-connection buffer in the extension
* user allocation. By default with no extensions or no
* extension callback handling, just the normal input buffer is
* used then so it is efficient.
*
* callback returns 1 in case it wants to spill more buffers
*
* This takes care of holding the buffer if send is incomplete, ie,
* if wsi->u.ws.clean_buffer is 0 (meaning an extension meddled with
* the buffer). If wsi->u.ws.clean_buffer is 1, it will instead
* return to the user code how much OF THE USER BUFFER was consumed.
*/
n = lws_issue_raw_ext_access(wsi, buf - pre, len + pre);
wsi->u.ws.inside_frame = 1;
if (n <= 0)
return n;
if (n == (int)len + pre) {
/* everything in the buffer was handled (or rebuffered...) */
wsi->u.ws.inside_frame = 0;
return orig_len;
}
/*
* it is how many bytes of user buffer got sent... may be < orig_len
* in which case callback when writable has already been arranged
* and user code can call lws_write() again with the rest
* later.
*/
return n - pre;
}
LWS_VISIBLE int lws_serve_http_file_fragment(struct lws *wsi)
{
struct lws_context *context = wsi->context;
struct lws_context_per_thread *pt = &context->pt[(int)wsi->tsi];
unsigned long amount;
int n, m;
while (!lws_send_pipe_choked(wsi)) {
if (wsi->trunc_len) {
if (lws_issue_raw(wsi, wsi->trunc_alloc +
wsi->trunc_offset,
wsi->trunc_len) < 0) {
lwsl_info("%s: closing\n", __func__);
return -1;
}
continue;
}
if (wsi->u.http.filepos == wsi->u.http.filelen)
goto all_sent;
if (lws_plat_file_read(wsi, wsi->u.http.fd, &amount,
pt->serv_buf,
LWS_MAX_SOCKET_IO_BUF) < 0)
return -1; /* caller will close */
n = (int)amount;
if (n) {
lws_set_timeout(wsi, PENDING_TIMEOUT_HTTP_CONTENT,
context->timeout_secs);
wsi->u.http.filepos += n;
m = lws_write(wsi, pt->serv_buf, n,
wsi->u.http.filepos == wsi->u.http.filelen ?
LWS_WRITE_HTTP_FINAL : LWS_WRITE_HTTP);
if (m < 0)
return -1;
if (m != n)
/* adjust for what was not sent */
if (lws_plat_file_seek_cur(wsi, wsi->u.http.fd,
m - n) ==
(unsigned long)-1)
return -1;
}
all_sent:
if (!wsi->trunc_len && wsi->u.http.filepos == wsi->u.http.filelen) {
wsi->state = LWSS_HTTP;
/* we might be in keepalive, so close it off here */
lws_plat_file_close(wsi, wsi->u.http.fd);
wsi->u.http.fd = LWS_INVALID_FILE;
if (wsi->protocol->callback)
/* ignore callback returned value */
if (user_callback_handle_rxflow(
wsi->protocol->callback, wsi,
LWS_CALLBACK_HTTP_FILE_COMPLETION,
wsi->user_space, NULL, 0) < 0)
return -1;
return 1; /* >0 indicates completed */
}
}
lwsl_info("choked before able to send whole file (post)\n");
lws_callback_on_writable(wsi);
return 0; /* indicates further processing must be done */
}
#if LWS_POSIX
LWS_VISIBLE int
lws_ssl_capable_read_no_ssl(struct lws *wsi, unsigned char *buf, int len)
{
int n;
n = recv(wsi->sock, (char *)buf, len, 0);
if (n >= 0)
return n;
#if LWS_POSIX
if (LWS_ERRNO == LWS_EAGAIN ||
LWS_ERRNO == LWS_EWOULDBLOCK ||
LWS_ERRNO == LWS_EINTR)
return LWS_SSL_CAPABLE_MORE_SERVICE;
#endif
lwsl_warn("error on reading from skt\n");
return LWS_SSL_CAPABLE_ERROR;
}
LWS_VISIBLE int
lws_ssl_capable_write_no_ssl(struct lws *wsi, unsigned char *buf, int len)
{
int n = 0;
#if LWS_POSIX
n = send(wsi->sock, (char *)buf, len, MSG_NOSIGNAL);
// lwsl_info("%s: sent len %d result %d", __func__, len, n);
if (n >= 0)
return n;
if (LWS_ERRNO == LWS_EAGAIN ||
LWS_ERRNO == LWS_EWOULDBLOCK ||
LWS_ERRNO == LWS_EINTR) {
if (LWS_ERRNO == LWS_EWOULDBLOCK)
lws_set_blocking_send(wsi);
return LWS_SSL_CAPABLE_MORE_SERVICE;
}
#else
(void)n;
(void)wsi;
(void)buf;
(void)len;
// !!!
#endif
lwsl_debug("ERROR writing len %d to skt fd %d err %d / errno %d\n", len, wsi->sock, n, LWS_ERRNO);
return LWS_SSL_CAPABLE_ERROR;
}
#endif
LWS_VISIBLE int
lws_ssl_pending_no_ssl(struct lws *wsi)
{
(void)wsi;
return 0;
}