lighttpd
/
lighttpd1.4
2
0
Fork 0
Code Issues Releases Wiki Activity

rewrote the event and error handling with the help of the fcgi-responder.c from the tests.

Browse Source 
fixed #38 and other issues this way


git-svn-id: svn://svn.lighttpd.net/lighttpd/trunk@63 152afb58-edef-0310-8abb-c4023f1b3aa9
svn/tags/release-1.3.12
Jan Kneschke 19 years ago
parent 2420fd96d2
commit 7725e92ff6
4 changed files with 141 additions and 105 deletions
Show Stats Download Patch File Download Diff File

5
src/connections.c
Unescape Escape View File

@ -182,7 +182,7 @@ static void dump_packet(const unsigned char *data, size_t len) {
fprintf(stderr, "\n");
}
}
#endif // #if 0
#endif
static int connection_handle_read(server *srv, connection *con) {
int len;
@ -1256,12 +1256,15 @@ int connection_state_machine(server *srv, connection *con) {
connection_set_state(srv, con, CON_STATE_HANDLE_REQUEST);
done = 1; /* is this neccesary ? */
break;
case HANDLER_COMEBACK:
done = -1;
case HANDLER_WAIT_FOR_EVENT:
/* come back here */
connection_set_state(srv, con, CON_STATE_HANDLE_REQUEST);
done = 1; /* is this neccesary ? */
break;
case HANDLER_ERROR:
/* something went wrong */

239
src/mod_fastcgi.c
Unescape Escape View File

@ -323,6 +323,9 @@ typedef struct {
size_t path_info_offset; /* start of path_info in uri.path */
pid_t pid;
int got_proc;
plugin_config conf;
connection *remote_conn; /* dumb pointer */
@ -1256,8 +1259,8 @@ void fcgi_connection_cleanup(server *srv, handler_ctx *hctx) {
if (hctx->host && hctx->proc) {
hctx->host->load--;
if (hctx->state != FCGI_STATE_INIT &&
hctx->state != FCGI_STATE_CONNECT) {
if (hctx->got_proc) {
/* after the connect the process gets a load */
hctx->proc->load--;
@ -2246,6 +2249,92 @@ int fcgi_proclist_sort_down(server *srv, fcgi_extension_host *host, fcgi_proc *p
return 0;
}
static int fcgi_restart_dead_procs(server *srv, plugin_data *p, fcgi_extension_host *host) {
fcgi_proc *proc;
for (proc = host->first; proc; proc = proc->next) {
if (p->conf.debug) {
log_error_write(srv, __FILE__, __LINE__, "sbdbdddd",
"proc:",
host->host, proc->port,
proc->socket,
proc->state,
proc->is_local,
proc->load,
proc->pid);
}
if (0 == proc->is_local) {
/*
* external servers might get disabled
*
* enable the server again, perhaps it is back again
*/
if ((proc->state == PROC_STATE_DISABLED) &&
(srv->cur_ts - proc->disable_ts > FCGI_RETRY_TIMEOUT)) {
proc->state = PROC_STATE_RUNNING;
host->active_procs++;
log_error_write(srv, __FILE__, __LINE__, "sbdb",
"fcgi-server re-enabled:",
host->host, host->port,
host->unixsocket);
}
} else {
/* the child should not terminate at all */
int status;
if (proc->state == PROC_STATE_DIED_WAIT_FOR_PID) {
switch(waitpid(proc->pid, &status, WNOHANG)) {
case 0:
/* child is still alive */
break;
case -1:
break;
default:
if (WIFEXITED(status)) {
#if 0
log_error_write(srv, __FILE__, __LINE__, "sdsd",
"child exited, pid:", proc->pid,
"status:", WEXITSTATUS(status));
#endif
} else if (WIFSIGNALED(status)) {
log_error_write(srv, __FILE__, __LINE__, "sd",
"child signaled:",
WTERMSIG(status));
} else {
log_error_write(srv, __FILE__, __LINE__, "sd",
"child died somehow:",
status);
}
proc->state = PROC_STATE_DIED;
break;
}
}
/*
* local servers might died, but we restart them
*
*/
if (proc->state == PROC_STATE_DIED &&
proc->load == 0) {
/* restart the child */
if (fcgi_spawn_connection(srv, p, host, proc)) {
log_error_write(srv, __FILE__, __LINE__, "s",
"ERROR: spawning fcgi failed.");
return HANDLER_ERROR;
}
fcgi_proclist_sort_down(srv, host, proc);
}
}
}
return 0;
}
static handler_t fcgi_write_request(server *srv, handler_ctx *hctx) {
@ -2303,6 +2392,9 @@ static handler_t fcgi_write_request(server *srv, handler_ctx *hctx) {
return HANDLER_ERROR;
}
if (hctx->proc->is_local) {
hctx->pid = hctx->proc->pid;
}
switch (fcgi_establish_connection(srv, hctx)) {
case 1:
@ -2310,6 +2402,8 @@ static handler_t fcgi_write_request(server *srv, handler_ctx *hctx) {
/* connection is in progress, wait for an event and call getsockopt() below */
fdevent_event_add(srv->ev, &(hctx->fde_ndx), hctx->fd, FDEVENT_OUT);
return HANDLER_WAIT_FOR_EVENT;
case -1:
/* if ECONNREFUSED choose another connection -> FIXME */
@ -2334,9 +2428,13 @@ static handler_t fcgi_write_request(server *srv, handler_ctx *hctx) {
return HANDLER_ERROR;
}
if (socket_error != 0) {
log_error_write(srv, __FILE__, __LINE__, "ss",
"establishing connection failed:", strerror(socket_error),
"port:", hctx->proc->port);
if (!hctx->proc->is_local) {
/* local procs get restarted */
log_error_write(srv, __FILE__, __LINE__, "ss",
"establishing connection failed:", strerror(socket_error),
"port:", hctx->proc->port);
}
return HANDLER_ERROR;
}
@ -2346,6 +2444,7 @@ static handler_t fcgi_write_request(server *srv, handler_ctx *hctx) {
hctx->proc->load++;
hctx->proc->last_used = srv->cur_ts;
hctx->got_proc = 1;
if (p->conf.debug) {
log_error_write(srv, __FILE__, __LINE__, "sddbdd",
@ -2377,11 +2476,13 @@ static handler_t fcgi_write_request(server *srv, handler_ctx *hctx) {
/* fall through */
case FCGI_STATE_WRITE:
/* continue with the code after the switch */
if (-1 == (r = write(hctx->fd,
hctx->write_buffer->ptr + hctx->write_offset,
hctx->write_buffer->used - hctx->write_offset))) {
/* why aren't we using the network_ interface here ? */
r = write(hctx->fd,
hctx->write_buffer->ptr + hctx->write_offset,
hctx->write_buffer->used - hctx->write_offset);
if (-1 == r) {
if (errno == ENOTCONN) {
/* the connection got dropped after accept()
*
@ -2411,8 +2512,6 @@ static handler_t fcgi_write_request(server *srv, handler_ctx *hctx) {
"write-offset:", hctx->write_offset,
"reconnect attempts:", hctx->reconnects);
return HANDLER_ERROR;
}
@ -2424,6 +2523,8 @@ static handler_t fcgi_write_request(server *srv, handler_ctx *hctx) {
return HANDLER_ERROR;
} else {
fdevent_event_add(srv->ev, &(hctx->fde_ndx), hctx->fd, FDEVENT_OUT);
return HANDLER_WAIT_FOR_EVENT;
}
}
@ -2437,6 +2538,9 @@ static handler_t fcgi_write_request(server *srv, handler_ctx *hctx) {
break;
case FCGI_STATE_READ:
/* waiting for a response */
fdevent_event_add(srv->ev, &(hctx->fde_ndx), hctx->fd, FDEVENT_IN);
break;
default:
log_error_write(srv, __FILE__, __LINE__, "s", "(debug) unknown state");
@ -2446,91 +2550,6 @@ static handler_t fcgi_write_request(server *srv, handler_ctx *hctx) {
return HANDLER_WAIT_FOR_EVENT;
}
static int fcgi_restart_dead_procs(server *srv, plugin_data *p, fcgi_extension_host *host) {
fcgi_proc *proc;
for (proc = host->first; proc; proc = proc->next) {
if (p->conf.debug) {
log_error_write(srv, __FILE__, __LINE__, "sbdbdddd",
"proc:",
host->host, proc->port,
proc->socket,
proc->state,
proc->is_local,
proc->load,
proc->pid);
}
if (0 == proc->is_local) {
/*
* external servers might get disabled
*
* enable the server again, perhaps it is back again
*/
if ((proc->state == PROC_STATE_DISABLED) &&
(srv->cur_ts - proc->disable_ts > FCGI_RETRY_TIMEOUT)) {
proc->state = PROC_STATE_RUNNING;
host->active_procs++;
log_error_write(srv, __FILE__, __LINE__, "sbdb",
"fcgi-server re-enabled:",
host->host, host->port,
host->unixsocket);
}
} else {
/* the child should not terminate at all */
int status;
if (proc->state == PROC_STATE_DIED_WAIT_FOR_PID) {
switch(waitpid(proc->pid, &status, WNOHANG)) {
case 0:
/* child is still alive */
break;
case -1:
break;
default:
if (WIFEXITED(status)) {
log_error_write(srv, __FILE__, __LINE__, "sdsd",
"child exited, pid:", proc->pid,
"status:", WEXITSTATUS(status));
} else if (WIFSIGNALED(status)) {
log_error_write(srv, __FILE__, __LINE__, "sd",
"child signaled:",
WTERMSIG(status));
} else {
log_error_write(srv, __FILE__, __LINE__, "sd",
"child died somehow:",
status);
}
proc->state = PROC_STATE_DIED;
break;
}
}
/*
* local servers might died, but we restart them
*
*/
if (proc->state == PROC_STATE_DIED &&
proc->load == 0) {
/* restart the child */
if (fcgi_spawn_connection(srv, p, host, proc)) {
log_error_write(srv, __FILE__, __LINE__, "s",
"ERROR: spawning fcgi failed.");
return HANDLER_ERROR;
}
fcgi_proclist_sort_down(srv, host, proc);
}
}
}
return 0;
}
SUBREQUEST_FUNC(mod_fastcgi_handle_subrequest) {
plugin_data *p = p_d;
@ -2571,14 +2590,28 @@ SUBREQUEST_FUNC(mod_fastcgi_handle_subrequest) {
* restart the request-handling
*/
if (proc && proc->is_local) {
#if 0
log_error_write(srv, __FILE__, __LINE__, "sbdb", "connect() to fastcgi failed, restarting the request-handling:",
host->host,
proc->port,
proc->socket);
proc->state = PROC_STATE_DIED_WAIT_FOR_PID;
#endif
/*
* several hctx might reference the same proc
*
* Only one of them should mark the proc as dead all the other
* ones should just take a new one.
*
* If a new proc was started with the old struct this might lead
* the mark a perfect proc as dead otherwise
*
*/
if (proc->state == PROC_STATE_RUNNING &&
hctx->pid == proc->pid) {
proc->state = PROC_STATE_DIED_WAIT_FOR_PID;
}
}
fcgi_restart_dead_procs(srv, p, host);
fcgi_connection_cleanup(srv, hctx);
@ -2586,15 +2619,12 @@ SUBREQUEST_FUNC(mod_fastcgi_handle_subrequest) {
buffer_reset(con->physical.path);
con->mode = DIRECT;
joblist_append(srv, con);
/* mis-using HANDLER_WAIT_FOR_FD to break out of the loop
* and hope that the childs will be restarted
*
*/
return HANDLER_WAIT_FOR_FD;
} else {
fcgi_connection_cleanup(srv, hctx);
buffer_reset(con->physical.path);
@ -2682,8 +2712,11 @@ static handler_t fcgi_handle_fdevent(void *s, void *ctx, int revents) {
return HANDLER_FINISHED;
case -1:
if (proc->pid) {
if (proc->pid && proc->state != PROC_STATE_DIED) {
int status;
/* only fetch the zombie if it is not already done */
switch(waitpid(proc->pid, &status, WNOHANG)) {
case 0:
/* child is still alive */

1
src/response.c
Unescape Escape View File

@ -1242,7 +1242,6 @@ handler_t http_response_prepare(server *srv, connection *con) {
switch(r = plugins_call_handle_subrequest(srv, con)) {
case HANDLER_GO_ON:
/* request was not handled, looks like we are done */
return HANDLER_FINISHED;
case HANDLER_FINISHED:
/* request is finished */

1
src/server.c
Unescape Escape View File

@ -952,6 +952,7 @@ int main (int argc, char **argv) {
case HANDLER_FINISHED:
case HANDLER_GO_ON:
case HANDLER_WAIT_FOR_EVENT:
case HANDLER_WAIT_FOR_FD:
break;
case HANDLER_ERROR:
/* should never happen */

Write Preview
Loading…
Cancel
Save