reduce code duplication in list resizing
realloc() of NULL ptr has behavior similar to malloc()
Note that if initial size == 0, then code used to adjust size
must be += x to ensure the size is non-zero for reallocation.
(Multiplying 0 * x, e.g. power-2 resizing, will result in 0.)
quickly clear buffer instead of buffer_string_set_length(b, 0) or
buffer_reset(b). Avoids free() of large buffers about to be reused,
or buffers that are module-scoped, persistent, and reused.
(buffer_reset() should still be used with buffers in connection *con
when the data in the buffers is supplied by external, untrusted source)
provide standard types in first.h instead of base.h
provide lighttpd types in base_decls.h instead of settings.h
reduce headers exposed by headers for core data structures
do not expose <pcre.h> or <stdlib.h> in headers
move stat_cache_entry to stat_cache.h
reduce use of "server.h" and "base.h" in headers
fix rare race condition from backends with server.stream-response-body=2
(thx abelbeck)
x-ref:
"fastcgi and stream-response-body=2 hangs on last chunk"
https://redmine.lighttpd.net/issues/2878
Omit calling cgi_handle_fdevent() after CGI process exit.
Another (sub)process may be holding pipe fd open and might write
response instead of the initial CGI process.
add FDEVENT_IN in addition to FDEVENT_HUP when triggering
cgi_handle_fdevent() after the CGI process exits.
(This helps improve reliability when running tests under Cygwin)
centralize most waitpid() handling in core server, with hooks for
modules to be informed of pid and status when a process exits.
This enables faster discovery (and restart) of exited processes,
and also allows for lighttpd to manage backend processes in the
parent (master) process when server.max-worker > 0.
common codebase for socket backends, based off mod_fastcgi with
some features added for mod_proxy
(mostly intended to reduce code duplication and enhance code isolation)
mod_fastcgi and mod_scgi can now use fastcgi.balance and scgi.balance
for similar behavior as proxy.balance, but the balancing is per-host
and not per-proc. proxy.balance is also per-host and not per-proc.
mod_proxy and mod_scgi can now use proxy.map-extensions and
scgi.map-extensions, similar to fastcgi.map-extensions.
mod_fastcgi behavior change (affects only mod_status):
- statistics tags have been renamed from "fastcgi.*" to "gw.*"
"fastcgi.backend.*" -> "gw.backend.*"
"fastcgi.active-requests" -> "gw.active-requests"
("fastcgi.requests" remains "fastcgi.requests")
("proxy.requests" is new)
("scgi.requests" is new)
mod_scgi behavior change (likely minor):
- removed scgi_proclist_sort_down() and scgi_proclist_sort_up().
procs now chosen based on load as measured by num socket connnections
Note:
modules using gw_backend.[ch] are currently still independent modules.
If it had been written as a single module with fastcgi, scgi, proxy
implementations, then there would have been a chance of breaking some
existing user configurations where module ordering made a difference
for which module handled a given request, though for most people, this
would have made no difference.
Details about mod_fastcgi code transformations:
unsigned int debug -> int debug
fastcgi_env member removed from plugin_config
renamed "fcgi" and "fastcgi" to "gw", and "FCGI" to "GW"
reorganize routines for high-level and lower-level interfaces
some lower-level internal interfaces changed to use host,proc,debug
args rather than knowing about higher-level (app) hctx and plugin_data
tabs->spaces and reformatting
fix streaming response when server.stream-response-body = 2
and client catches up to stream from backend
(thx horgh)
x-ref:
"mod_fastcgi can fail to read entire response from server"
https://redmine.lighttpd.net/issues/2796
More specific checks on contents of array lists. Each module using
lists now does better checking on the types of values in the list
(strings, integers, arrays/lists)
This helps prevent misconfiguration of things like cgi.assign,
fastcgi.server, and scgi.server, where source code might be
served as static files if parenthesis are misplaced.
x-ref:
https://redmine.lighttpd.net/boards/2/topics/6571
new directive cgi.local-redir = [enable|disable]
*disable* RFC3875 6.2.2 local-redir by default.
(behavior change from when local-redir support added in lighttpd 1.4.40)
The reason for this behavior change is that CGI local-redir support
(RFC3875 6.2.2) is an optimization. Absence of support may result in
additional latency in servicing a request due the additional round-trip
to the client, but that was the prior behavior (before lighttpd 1.4.40)
and is the behavior of web servers which do not support CGI local-redir.
However, enabling CGI local-redir by default may result in broken links
in the case where a user config (unaware of CGI local-redir behavior)
returns HTML pages containing *relative* paths (not root-relative paths)
which are relative to the location of the local-redir target document,
and the local-redir target document is located at a different URL-path
from the original CGI request.
x-ref:
RFC3875 CGI 1.1 specification section 6.2.2 Local Redirect Response
http://www.ietf.org/rfc/rfc3875
"CGI local redirect not implemented correctly"
https://redmine.lighttpd.net/issues/2108
"1.4.40 regression: broken redirect (using Location) between url.rewrite-once URLs"
https://redmine.lighttpd.net/issues/2793
set status 200 OK if CGI does not return CGI headers
Note:
This mode in lighttpd is deprecated and may be removed in the next major
release of lighttpd. CGI scripts should return a proper CGI header in
the response, even if that header is empty and followed by a blank line,
before return response body.
Without a proper CGI response header, the first line(s) of the response
might be incorrectly construed as being CGI response headers, especially
if they contain ':', and response may be corrupted. That is why this
mode is deprecated (and not supported in numerous other web servers).
The minimal valid CGI response header is "\n", which lighttpd will treat
as equivalent to "Status: 200\n\n"
x-ref:
"error 500 (mod_cgi.c.601) cgi died"
https://redmine.lighttpd.net/issues/2786
It is still not a good idea for backend to send Transfer-Encoding unless
backend is mod_proxy, and mod_proxy should not currently receive chunked
response since mod_proxy sends HTTP/1.0 request.
If mod_proxy is changed to sent HTTP/1.1 request, then lighttpd would
need to check if client is HTTP/1.0 and would need to de-chunk and
remove any other transfer-codings if not supported by next-hop.
x-ref:
"error 500 (mod_cgi.c.601) cgi died"
https://redmine.lighttpd.net/issues/2786
"Status" from CGI/1.1 environment should not be sent back to client.
Also, do not send "Status" back to client in mod_scgi
and more precisely parse for "Status" in mod_fastcgi
RFC3875 CGI local-redir stricter adherence
do not apply local-redir if any response headers besides "Location"
do not apply local-redir if any response body has been received
(though it might not have been received yet, and we do not wait to find
out, if lighttpd is configured to stream response body back to client)
x-ref:
RFC3875 CGI 1.1 specification section 6.2.2 Local Redirect Response
http://www.ietf.org/rfc/rfc3875
"CGI local redirect not implemented correctly"
https://redmine.lighttpd.net/issues/2108
Loosen local redirect handling in mod_cgi to skip handling as local
redirect if the Location matches con->uri.path, since if the request
is intended to redirect back to the same CGI using the same request
method, path info, and query string, the CGI would logically just
return the final intended response. Loosening this handling avoids a
problem with applications (potentially) accessible through multiple
gateways, where the application is not aware of this specific handling
of Location in the Common Gateway Interface (CGI/1.1), the application
sends abs-path in the Location response header instead of absoluteURI,
and the application expects the client to receive this Location response
header instead of the server to process as a CGI local redirect.
One example of such an application is LuCI,
which sends Set-Cookie with Location: /abs-path
https://github.com/openwrt/luci
(Note that this loose check for matching con->uri.path is not perfect
and might not match if the CGI returned a path with a different case
and the server is on a case-insensitive filesystem, or if the path
returned by the CGI is rewritten elsewhere to a different con->uri.path
before getting to mod_cgi.)
RFC3875 CGI 1.1 specification section 6.2.2 Local Redirect Response
http://www.ietf.org/rfc/rfc3875
x-ref:
"CGI local-redir handling conflicts with LuCI redirect w/ Set-Cookie"
https://redmine.lighttpd.net/issues/2779
"CGI local redirect not implemented correctly"
https://redmine.lighttpd.net/issues/2108
This change should fix an issue with lighttpd on Debian kfreebsd-* arch
(kfreebsd-amd64 and kfreebsd-i386)
.libs/mod_cgi.o: In function `cgi_create_env':
./src/mod_cgi.c:1103: warning: pipe2 is not implemented and will always fail
lighttpd is single-threaded so there is no race with pipe()
and then fcntl() F_SETFD FD_CLOEXEC on the pair of pipe fds.
Using pipe2() where available is still slightly more efficient
by eliding the syscalls to set FD_CLOEXEC.
Lack of pipe2() on relic Unix as well as missing on Mac OSX is likely
one reason why threaded web servers such as nginx choose not to support
CGI except via an external service to the process. Without pipe2(),
race conditions exist and it is not safe for a threaded server to use
pipe() and fork() when the server also does not want to potentially leak
open file descriptors to various unrelated CGI scripts.
support Transfer-Encoding: chunked request body in conjunction with
server.stream-request-body = 0
dynamic handlers will still return 411 Length Required if
server.stream-request-body = 1 or 2 (!= 0)
since CGI-like env requires CONTENT_LENGTH be set
(and mod_proxy currently sends HTTP/1.0 requests to backends,
and Content-Length recommended for robust interaction with backend)
x-ref:
"request: support Chunked Transfer Coding for HTTP PUT"
https://redmine.lighttpd.net/issues/2156
