1 Changes with mod_fastcgi 2.2.3
3 *) Beta WinNT support. David Allen [djallen@raleigh.ibm.com] and
4 Rob Saccoccio [robs@chelsea.net]
6 *) Remove request type restriction (GET and POST only) in order to support
7 Web DAV requests. Scott Robertson [sroberts@codeit.com]
9 *) Allow requests for URLs such as /server/some/other/qualifier to match a
10 FastCGI server defined as /server. This was done primarily for Java
11 Servlets, but is generically useful.
13 *) Change the comm between the PM and the request handlers from a regular
14 file to a pipe and an assortment of other dynamic fixes.
16 *) Log dynamic process termination scheduling and the resulting
17 process exit notification.
19 *) Change the default singleThreshhold from 10 to 0 to prevent the last dynamic
20 application process from being killed off due to low demand.
22 *) Clean up FastCGI application pathnames (e.g. remove duplicate slashes).
24 *) Fix bugs that prevented dynamic processes from being shutdown when
25 the load subsided. Lars Heete [heete@do.isst.fhg.de]
27 *) Prevent dynamic processes from being scheduled to be started before
28 the init-start-delay or restart-delay period expires. This prevents
29 the queuing of process start requests while an application which has
30 a long initialization period starts up.
32 *) When suexec is enabled, allow the process manager to become root again in
33 order to signal applications it spawned.
35 *) Change the "server started" log message level from INFO to WARN.
37 *) Fix dynamic server "Connection Refused" handling.
39 *) Fix demand-based dynamic process spawing (uses a more portable approach).
41 *) Add -idle-timeout arg to FastCgiServer, FastCgiExternalServer, and
42 FastCgiConfig. mod_fastcgi will now abort a connection if inactive for
43 longer than this period. It applies to the initiation of connections as
44 well (and thus is similar to appConnTimeout). Default is 30 seconds.
46 *) appConnTimeout is 0 by default now resulting in blocking connect()s. This
47 is more platform portable/predicable.
49 *) Leave STDOUT and STDERR open to the main server error_log. This should
50 eliminate a pile of protocol errors and having to track down 3rd party
51 libs writing to stderr inadvertantly. This doesn't mean these shouldn't
52 be fixed in the application to use the FastCGI I/O, it just allows these
53 applications to run when previously they'd crash and burn.
55 *) Add the pass-header arg to server directives.
57 *) Allow supplementary groups to be passed to spawned FastCGI applications (as
58 is done for CGI) - at least until PR2580 is resolved.
60 *) Initialize the default (empty) environment in a more socially acceptable
63 *) Miscellaneous doc improvements based on notes I've collected up over the
66 *) Fix a bug in the stderr handling introduced in 2.2.2. There were conditions
67 that resulted in not or improperly terminated strings.
69 *) Fix the call to setsockopt() to disable Nagle. [based on a bug report by
70 Johannes Plassmann <johannes.plassmann@pdb.siemens.de>]
72 Changes with mod_fastcgi 2.2.2
74 *) Added support for blocking connect()s by setting appConnTimeout to 0.
75 Non-blocking connect()s (the default) can be troublesome on some platforms.
76 Its expected that blocking connect()s will become the default (and
77 non-blocking connect()s will become optional) in the next release.
79 *) Dump a compile time error if the version of Apache is too old.
81 *) Wrap the SIGPIPE handler manipulation code such that it is only applicable
82 to Apache releases prior to 1.3.6.
84 *) Minor tweaks for RUSSIAN_APACHE. [Sergey Gershtein <sg@mplik.ru>]
88 Always restart a failed dynamic application if it is the last instance.
89 This means there once started there will always be at least one process
90 instance of a dynamic application.
92 Send PLEASE_START to the PM when a connect() results in an ECONNREFUSED.
93 ECONNREFUSED means the listen queue is full (or there isn't one). Asking
94 the PM to start (another) application instance may help empty it faster.
96 Change two sleep() calls to select() based snoozes because alarm() is in
97 effect and sleep() and alarm() don't always play nice together.
99 Fixed a couple of error messages.
101 *) Fix -listen-queue-depth arg on FastCgiConfig (dynamic). Previously it
102 was ignored and the default was always used.
104 *) Allow the -initial-env argument to be used to pass variables from the
105 Apache process environment to the FastCGI server (by specifying a
106 variable name without the "=" or a value). [suggested by
107 Martin Lichtin <lichtin@oanda.com>]
109 *) Cleanup some debug macros.
111 *) Improved script stderr handling. [based on suggestions by
112 David Birnbaum <davidb@chelsea.net>]
114 *) Added IRIX and FreeBSD to the list of supported platforms and other minor
115 updates to the INSTALL doc.
117 *) Changed the default listen-queue-depth (FCGI_DEFAULT_LISTEN_Q) from
118 5 to 100. Its still configurable with the -listen-queue-depth option.
119 This should help eliminate the FAQ - Why do I see "Connection Refused"
122 *) Fix a bug in FastCgiExternalServer that broke support for external servers
123 on other hosts. [Dave Neuer <dneuer@futuristics.net>]
125 Changes with mod_fastcgi 2.2.1
127 *) Updates to the INSTALL doc to describe building as a DSO.
129 *) If the FastCgiIpcDir directive was in use and "httpd -t" was issued while
130 Apache was running ("apachectl restart" does this), the contents of the
131 dynamic directory (dynamic sockets and the mbox) would be blown away
134 *) Add an extern declaration for ap_sys_siglist (its exposed but not in any
135 Apache header file) to prevent a compile error on systems without
138 *) During auth requests, the subprocess_env table was left holding variables
139 sent to the auth FastCGI server (including REMOTE_PASSWD!) which means they
140 were passed to other processing phases (such as CGI/FastCGI). The
141 subprocess_env is now restored to its pre-auth condition.
143 *) Added a FastCGI Authorizer Role compatibility mode which implements the
144 specification to the tee. Use -compat arg with any of the Auth server
145 directives. This is intended for new and existing FastCGI authorizer
146 applications that require compatibility with other server implementations.
148 *) Fix logging from the process manager. Previously, all calls to the
149 logging routines used NULL for the server_rec. This works fine in a dev
150 environment, but results in most of these messages being tossed by
151 log_error_core() because DEFAULT_LOGLEVEL is too low. Odd logic.
153 *) Always setup fr->header so that in the event the FastCGI server
154 doesn't get a valid FastCGI protocol header over the wire, we
155 can still print our error message to the log from do_work().
157 *) Fix a compile error in open_connection_to_fs() on systems with TCP_NODELAY
158 defined (Don Locrasto [locrasto@iceminer.com]).
160 Changes with mod_fastcgi 2.2.0
162 *) Serious rewrite of mod_fastcgi.html. An example conf will have
163 to wait until another time - I'm burned out on docs updates.
165 *) Add to the mod_fastcgi to the server version string.
167 *) SCO doesn't like the const in the arg to inet_addr() and
170 *) Use Apache's NET_SIZE_T to get the correct arg type for getsockopt().
172 *) Deleted redundant header files from fcgi.h.
174 *) Full path names are no longer required for directives which take
175 file names as arguments. Like other Apache directives you can
176 specify paths from server_root (don't start with a "/").
178 *) The fcgiKillMgr is gone. This intermediate process sat between
179 the Apache parent and the FastCGI Process Manager under Apache
180 1.2. The process manager tries to name itself fcgi_pm (it used
183 *) Authentication, Authorization, and Access phases are now
184 supported. I've bent the FastCgi spec a bit: as many of the standard
185 environment variables (that were easy) are sent (the spec says don't
186 send some by name), all headers (except Status) sent by the auth
187 FastCgi server are passed to subprocesses (CGI/FastCGI invocations)
188 as environment variables rather than just those prefixed by
189 "Variable-". Custom responses for auth failures aren't supported
190 (yet.. we'll see what the demand for this is like). In addition to
191 the FastCGI protocol defined environment variable "FCGI_ROLE" being
192 set to "AUTHORIZER" an environment variable "FCGI_APACHE_ROLE" is
193 set indicating which of the three phases is being processed. See
194 the docs for info on the new per-directory directives:
196 FastCgiAuthenticator, FastCgiAuthenticatorAuthoritative,
197 FastCgiAuthorizer, FastCgiAuthorizerAuthoritative,
198 FastCgiAccessChecker, FastCgiAccessCheckerAuthoritative
200 *) The code has been broken into logical chunks making figuring it
201 out much easier. All of the #defines a user may want to change are
202 now in mod_fastcgi.h.
204 *) All the logging has been converted to the "supported" Apache
205 logging routines, ap_log_error() and ap_log_rerror(). Log entries
206 for request specific errors should now be directed to the correct
207 server errorlog (if your running more than one). Some attempt has
208 been made at setting appropriate log levels (comments of course are
209 welcome) and printing errno info when its of value. All of log
210 messages now have "FastCGI" in them (nice for grep).
212 *) All of the calls to Apache routines have been updated to the ap_
213 convention, eliminating the need for the compat header.
215 *) By default we no longer flush every piece of data we get from the
216 FastCGI server to the client. This allows the FastCGI server to
217 release without having to wait for the client flush to complete.
218 The old behaviour can revived with the -flush argument to AppClass.
220 *) I renamed some of the directives for consistency (the old names
223 AppClass -> FastCgiServer,
224 FCGIConfig -> FastCgiConfig,
225 ExternalAppClass -> FastCgiExternalServer
227 *) Directive arguments are no longer case sensitive.
229 *) The module now uses hard_timeout() rather than soft_timeout. This
230 means that the module will longjump out of the request if the
231 drop-dead timeout expires (set with Apache's Timeout directive, the
232 default is 5min) or if the client closes the connection (SIGPIPE).
233 This is more typical for a Apache modules and I'm not convinced we
234 properly handle all of the error cases well enough to use
235 soft_timeout (I can think of one place we don't anyway). This means
236 your FastCGI application can see SIGPIPE.
238 *) SIGPIPE is now ignored by default in FastCGI servers spawned by
239 Apache. Without this the default behaviour is exit().
241 *) It should (I think) run now under DSO now.
243 *) The bug which prevented sending of binary data in 2.1b1 is fixed.
245 *) The broken -initial-env bug ins 2.1b1 is fixed.
247 *) Maybe some other stuff.
251 *** Well there's a big gap here. Maybe I'll go back and extract the CVS
252 commit notes and stick 'em in here.. when I get some free time. ;)
257 *** Originally from docs/README..
259 Apache[X1.03.01]/mod_fastcgi[02.00.05] 19970909 unsupported
260 From: Stanley Gambarin <stanleyg@cs.bu.edu>
262 *) Yet more changes to the source distribution. Separated out
263 the Tcl dynamic string and buffer libraries into separate files.
264 Did the same for the OS library, however, it is not really an
265 abstraction... more like a bunch of wrappers. Added to DEVNOTES
266 a note about the problem of Apache not allowing including header
269 Apache[X1.03.01]/mod_fastcgi[02.00.04] 19970908 unsupported
270 From: Stanley Gambarin <stanleyg@cs.bu.edu>
272 *) More changes to the installation scripts. Theoretically we
273 should be able to to use the same Makefile/installation script
274 for both Apache 1.2.x and 1.3.x. Need some sort of the abstraction
275 layer to use the same source code for both 1.2.x and 1.3.x sources.
277 Apache[X1.03.01]/mod_fastcgi[02.00.03] 19970905 unsupported
278 From: Stanley Gambarin <stanleyg@cs.bu.edu>
280 *) Created Makefile to compile the module into the archive library,
281 just like the proxy module. Also initial draft of the installation
282 script, which should work for both 1.2 and 1.3 sources of Apache.
284 Apache[01.02.04]/mod_fastcgi[02.00.02] 19970903 unsupported
285 From: Stanley Gambarin <stanleyg@cs.bu.edu>
287 *) More source reorg. Separated out some header files. Need much
288 more work on this. Updated the TODO file. Added fcgivers.h and
289 CHANGES file to track the history.
291 Apache[01.02.04]/mod_fastcgi[02.00.01] 19970902 unsupported
292 From: Stanley Gambarin <stanleyg@cs.bu.edu>
294 *) Source reorganization. This is done to provide a basis for later
295 functionality implementation. See TODO file for more information
296 of what possible future enhancements are possible/desired.
298 Apache[01.02.04]/mod_fastcgi[02.00.00] 19970902 unsupported
299 From: David MacKenzie <djm@va.pubnix.com>
301 *) Create the "dynamic" dir and "mbox" as the user specified in the
302 User and Group directives instead of as root. They are created with
303 restrictive permissions, and then the module checks to see whether the
304 user specified in the User and Group directives has read, write, and
305 execute permissions on them. Those permissions had been explicitly
306 denied by creating them as "root".
308 *) Some documentation was garbled or incomplete.
310 *) When a FastCGI application can't be execl'd, the code used a
311 value of errno that may have been stomped on by intervening system
312 calls. That happened on BSD/OS 2.1, where the error_log was
313 reporting errno=25 (NOTTY) instead of the correct errno=13 (EACCES).
314 An intervening stdio call was setting errno as a side-effect, which
315 it has the right to do.
317 *) Erroneous fprintf arguments, a missing return value, an unused
318 function, missing declarations, and other problems detected by gcc
321 *) Duplicated code to create the lock file name merged into a single
322 function, closing several memory leaks.
324 *) FCGIConfig -minProcesses didn't allow a value of 0, so the last
325 FCGI app in a given dynamic class would live forever, even if it
326 hadn't been requested for weeks.
328 *) The test for whether to keep looking for dynamic app victims to
329 kill was backwards, so no dynamic apps were ever selected as
332 *) The killInterval and updateInterval were being ignored; the
333 sigsuspend() forced recalculations only when a child died,
334 instead of at the intervals specified.
336 *) When a dynamic app couldn't be started, the program SEGV'd when
337 trying to free the ipcAddrPtr twice.
339 *) Remove the dead lock file and socket when cleaning up after a
340 server whose last child has been killed, so the FastCGIHandler
341 isn't fooled into thinking there is still a process serving
344 *) Make file locking robust in the presence of signals.
346 *) Rename some badly named variables.
348 *) Fix typos in many comments.
350 *) Fix some memory and file descriptor leaks.
352 *) Make the blocking kill of a server closer to working.
354 *) More fixes to calculations in dynamic application management.
357 *** Originally from docs/README.OMI..
359 What's New: Version 2.0b1, 16 Apr 1997
361 *) Implemented a mechanism by which FastCGI applications are started by
362 the web server on the first request and continue running. Also, a
363 heuristics have been implemented to allow for dynamic killing of the
364 running FastCGI apps. The configuration options are supplied via
365 the new FCGIConfig directive, that is described in the
368 *) When performing internal redirect, since the original request's body
369 has already been read, do not allow the redirected request to think
370 that is has one [body].
372 *) More conditional compilation for OS/2.
374 *) Fixed occassional "Assertion failed: len > 0, file mod_fastcgi.c"
376 *) Fixed "failed assertion `count >= 0 && count <= bufPtr->length'"
378 *) Added bflush() call after bwrite() in DrainReqOutBuf function to
379 force some output to be written in the period of inactivity. This
380 circumvents the Apache's buffering during the server-push.
382 *) Added a header-parser function placeholder for compatibility with an
385 *) Additional checks have been placed with regard to permissions of the
386 FastCGI processes. These include the existence of the file,
387 ExecCGI and IncludesNOEXEC checks, disallowing of nph- scripts, etc.
389 *) Implemented a restart cleanup, so that no parentlesses FastCGI
390 applications are left after Apache is restarted and/or terminated.
391 This was accomplished via two independent mechanisms, where the
392 first one prevents a start of the process manager on the first
393 reading of the configuration files. The second mechanism deals away
394 from Apache's implementation of fork() (via spawn_child()) and
395 implements its own forking, thus removing any dependencies on the
396 Apache to cleanup processes during termination phase.
398 *) Removed the definition of the Sigfunc, as Apache 2.0b10 defines it
399 itself in the file conf.h
401 What's New: Version 1.4.3, 15 Jan 1997
403 *) Fixed compilation warnings for various platforms, as well as
404 conditional compilation for OS/2.
406 What's New: Version 1.4.2, 12 Dec 1996
408 *) mod_fastcgi.c is ported to Apache 1.2b. Any further development
409 will proceed under this version of Apache web server.
411 *) As the result of porting, the "include virtual" construct of SSI
412 will now work correctly using either <Location> or <Directory>
415 What's New: Version 1.4.1, 4 Dec 1996
417 *) Checks have been removed from the ScanCGIHeaders that provided
418 for the presence of both Status and Location headers as being
419 an error. Contradictory to CGI/1.1 Internet Draft, both of
420 these headers are used by the current CGI applications.
422 What's New: Version 1.4, 22 Nov 1996
424 *) Added the -port option to AppClass, allowing TCP/IP communication.
425 Added the -socket option to AppClass, allowing Unix domain
426 communication via a configurable pathname.
428 *) Added the ExternalAppClass directive, allowing TCP/IP
429 communication with remote FastCGI applications.
431 *) The handler had its own code for generating HTTP response headers;
432 now it uses Apache's. This reduces the size of the module.
433 More importantly, it fixes the bug in which "include virtual"
434 sees the HTTP response headers.
436 *) The response header parser performed very little checking.
437 Now the parser enforces the guidelines in the CGI/1.1
438 Internet-Draft: Status and Location are mutually exclusive,
439 Location can only be a response to GET or HEAD,
440 CGI response headers can't be repeated, etc. (The CGI response
441 headers are the ones the handler interprets: Status, Location,
444 *) The response header parser used to miss CGI headers
445 with no whitespace after the colon, e.g. "Status:200 OK".
447 *) The response header parser sometimes interpreted the first line
448 of content as an RFC822 continuation line.
450 *) The handler implemented a nonstandard version of Location
451 which never used internal redirects. The handler
452 now attempts to implement Location as specified in the
453 CGI/1.1 Internet-Draft. The module documentation explains
456 *) Error log entries from the response header parser were
457 pretty uninformative; they are better now. If a header is
458 malformed, the log entry includes it. If the headers are
459 unterminated rather than malformed, the log entry says that,
460 and says how many bytes were received from the app.
462 *) When the application manager forked a new process, and that
463 process ran into trouble before executing the first instruction
464 of application code, the process used to exit with status =
465 errno. This made certain configuration problems (e.g. incorrect
466 file permissions when server parent is root) quite difficult
467 to pin down. Now the failing child process opens up the error log
468 and writes an informative entry before exiting.
470 *) The module now correctly handles a slash at the end of the
471 DocumentRoot directive. This was a one-line fix.
473 What's New: Version 1.3.3, 17 Oct 1996
475 *) The module now registers its request handler under the name
476 fastcgi-script in addition to the name application/x-httpd-fcgi.
477 This was a one-line addition to the module, but had quite
478 a large impact on the documentation and sample configuration.
480 What's New: Version 1.3.2, 27 Sept 1996
482 *) On some systems (SunOS, Linux?), fopen for append has a bug
483 that strikes when two processes append to the same file.
484 This bug causes the process manager to corrupt the error log.
485 Work around the bug by calling open, then fdopen. A patch
486 to Apache 1.1.1 is also required, as described in mod_fastcgi.html.
487 (Reported by Bob Ramstad.)
489 What's New: Version 1.3.1, 17 Sept 1996
491 *) On Linux and SunOS, the Apache default user_id (-1) is not a
492 legal value of uid_t, so cast it. (Reported by Bob Ramstad,
493 Scott Langley, others.)
495 *) On some systems (Linux, some Solaris?, Irix?), connect requires
496 write access to a Unix Domain socket, so provide it. (Reported
497 by Scott Langley, freeform@wired.com, others.)
499 *) If you hit a .fcg file, but there's no AppClass defined, the
500 error should be NOT_FOUND instead of SERVER_ERROR. (Reported
503 What's New: Version 1.3, 4 Sept 1996
505 *) Module sometimes busy-waited in FastCgiDoWork. Fixed.
507 *) Module violated the Apache buff abstraction.
508 Fixed. Might now work with SSL (not tested).
509 As part of the fix, eliminated support for
510 Apache 1.0x versions.
512 *) Module failed to chmod the Unix domain listening sockets
513 it created, so protections were set according to the current
516 *) Module forked too many process manager processes: One
517 per AppClass. When the Apache parent ran as root, these
518 process manager processes ran as root. New process
519 manager is a single process (only started if AppClass is
520 used) and runs with same privileges as other children.
521 The new process manager doesn't do any polling, so
522 there's less system overhead than before.
524 *) AppClass insisted on getting at least two arguments. Fixed.
526 *) Module used setjmp/longjmp, causing compiler warnings
527 on some platforms (e.g. Linux.) The module no longer
530 *) Module created listening sockets in /tmp, where they were
531 sometimes wiped out by cleanup scripts. Added FastCgiIpcDir
532 directive to give control over the location of listening
535 *) Module wrote error log entries without a timestamp. Fixed.
537 *) AppClass directive wrote error messages to stderr in addition
538 to returning a char * message to the Apache core. Fixed.
540 What's New: Version 1.2, 3 June 1996
542 *) Ported from Apache-1.0-based code to Apache-1.1b2 internals by
545 Add version string: APACHEVERSION macro to mod_fastcgi.c
547 *) chown FastCGI socket to user_id and setuid to user_id for app
550 *) Modify GetFromStream() by having it call an OS dependent
551 function GetStreamSize(FILE *) which uses FILE internal data member.
552 Linux users might need to modify GetStreamSize().
554 What's New: Version 1.1, 10 May 1996
556 *) If you specify a non-existent executable in the AppClass directive,
557 or if the file exists and it does not have execute permission you
558 get a constant stream of error messages telling you that "program
559 terminated due to a signal".
561 *) The mod_fastcgi module should use the standard Apache error logging
562 facility instead of writing to stderr.
564 Version 1.0, 30 April 1996