3 *) Support generic wrappers such as cgiwrap by eliminating dependencies
4 on Apache's SUEXEC, renaming the FastCgiSuexec directive
5 FastCgiWrapper and eliminating any checks regarding the target
6 application (this is the repsonibility of the wrapper).
8 *) Fix a nasty bug that occurred when a client aborted a request
9 before the connection to a dynamic FastCGI application was opened.
10 The application's lock file descriptor wasn't setup but was being
11 closed which resulted in FD0 being closed. Normally this open to
12 /dev/null and should pose no problems except that because the FD
13 was available it was being returned by Apache's accept(). This
14 caused it to be registered for pool cleanup. mod_cgi though moves
15 the CGI stdin pipe to FD0 and thus it was getting waxed during pool
16 cleanup. Problem identified by checksum@tebie.com.
18 Changes with mod_fastcgi 2.2.8
20 *) Eliminate the concept of disabled applications. If a failure
21 occurs trying to setup an application (e.g. bind() error) its
22 tried repeatedly every init-start-delay seconds.
24 *) Tweak to Makefile.tmpl to support DSOs. Dave Hill [ddhill@zk3.dec.com]
26 Changes with mod_fastcgi 2.2.6
28 *) Shutdown the PM when Apache appears to have disappeared.
30 *) seteuid() tweak for HP-UX 11. Milton L. Hankins [mlh@swl.msd.ray.com]
32 *) (Win32) More dynamic fixes.
34 *) (Win32) Eliminate the per application named pipe mutex. Its now
35 per process to keep the lib happy (the use _FCGI_MUTEX_ should
36 removed from the lib). This allows pipe-based applications to
37 now handle multiple simoultanous requests (one per process).
39 *) Increase the number of open FDs we look to close when spawning apps.
41 *) Prevent an assert from popping unnecessarily.
43 *) (Win32) Add support for interpretter scripts.
45 *) (Win32) Fix named pipe handling (problems with large responses).
47 *) (Win32) Remove the "can exec" check.
49 Changes with mod_fastcgi 2.2.4
51 *) Beta WinNT support. David Allen [djallen@raleigh.ibm.com] and
52 Rob Saccoccio [robs@chelsea.net]
54 *) Remove request type restriction (GET and POST only) in order to support
55 Web DAV requests. Scott Robertson [sroberts@codeit.com]
57 *) Allow requests for URLs such as /server/some/other/qualifier to match a
58 FastCGI server defined as /server. This was done primarily for Java
59 Servlets, but is generically useful.
61 *) Change the comm between the PM and the request handlers from a regular
62 file to a pipe and an assortment of other dynamic fixes.
64 *) Log dynamic process termination scheduling and the resulting
65 process exit notification.
67 *) Change the default singleThreshhold from 10 to 0 to prevent the last dynamic
68 application process from being killed off due to low demand.
70 *) Clean up FastCGI application pathnames (e.g. remove duplicate slashes).
72 *) Fix bugs that prevented dynamic processes from being shutdown when
73 the load subsided. Lars Heete [heete@do.isst.fhg.de]
75 *) Prevent dynamic processes from being scheduled to be started before
76 the init-start-delay or restart-delay period expires. This prevents
77 the queuing of process start requests while an application which has
78 a long initialization period starts up.
80 *) When suexec is enabled, allow the process manager to become root again in
81 order to signal applications it spawned.
83 *) Change the "server started" log message level from INFO to WARN.
85 *) Fix dynamic server "Connection Refused" handling.
87 *) Fix demand-based dynamic process spawing (uses a more portable approach).
89 *) Add -idle-timeout arg to FastCgiServer, FastCgiExternalServer, and
90 FastCgiConfig. mod_fastcgi will now abort a connection if inactive for
91 longer than this period. It applies to the initiation of connections as
92 well (and thus is similar to appConnTimeout). Default is 30 seconds.
94 *) appConnTimeout is 0 by default now resulting in blocking connect()s. This
95 is more platform portable/predicable.
97 *) Leave STDOUT and STDERR open to the main server error_log. This should
98 eliminate a pile of protocol errors and having to track down 3rd party
99 libs writing to stderr inadvertantly. This doesn't mean these shouldn't
100 be fixed in the application to use the FastCGI I/O, it just allows these
101 applications to run when previously they'd crash and burn.
103 *) Add the pass-header arg to server directives.
105 *) Allow supplementary groups to be passed to spawned FastCGI applications (as
106 is done for CGI) - at least until PR2580 is resolved.
108 *) Initialize the default (empty) environment in a more socially acceptable
111 *) Miscellaneous doc improvements based on notes I've collected up over the
114 *) Fix a bug in the stderr handling introduced in 2.2.2. There were conditions
115 that resulted in not or improperly terminated strings.
117 *) Fix the call to setsockopt() to disable Nagle. [based on a bug report by
118 Johannes Plassmann <johannes.plassmann@pdb.siemens.de>]
120 Changes with mod_fastcgi 2.2.2
122 *) Added support for blocking connect()s by setting appConnTimeout to 0.
123 Non-blocking connect()s (the default) can be troublesome on some platforms.
124 Its expected that blocking connect()s will become the default (and
125 non-blocking connect()s will become optional) in the next release.
127 *) Dump a compile time error if the version of Apache is too old.
129 *) Wrap the SIGPIPE handler manipulation code such that it is only applicable
130 to Apache releases prior to 1.3.6.
132 *) Minor tweaks for RUSSIAN_APACHE. [Sergey Gershtein <sg@mplik.ru>]
136 Always restart a failed dynamic application if it is the last instance.
137 This means there once started there will always be at least one process
138 instance of a dynamic application.
140 Send PLEASE_START to the PM when a connect() results in an ECONNREFUSED.
141 ECONNREFUSED means the listen queue is full (or there isn't one). Asking
142 the PM to start (another) application instance may help empty it faster.
144 Change two sleep() calls to select() based snoozes because alarm() is in
145 effect and sleep() and alarm() don't always play nice together.
147 Fixed a couple of error messages.
149 *) Fix -listen-queue-depth arg on FastCgiConfig (dynamic). Previously it
150 was ignored and the default was always used.
152 *) Allow the -initial-env argument to be used to pass variables from the
153 Apache process environment to the FastCGI server (by specifying a
154 variable name without the "=" or a value). [suggested by
155 Martin Lichtin <lichtin@oanda.com>]
157 *) Cleanup some debug macros.
159 *) Improved script stderr handling. [based on suggestions by
160 David Birnbaum <davidb@chelsea.net>]
162 *) Added IRIX and FreeBSD to the list of supported platforms and other minor
163 updates to the INSTALL doc.
165 *) Changed the default listen-queue-depth (FCGI_DEFAULT_LISTEN_Q) from
166 5 to 100. Its still configurable with the -listen-queue-depth option.
167 This should help eliminate the FAQ - Why do I see "Connection Refused"
170 *) Fix a bug in FastCgiExternalServer that broke support for external servers
171 on other hosts. [Dave Neuer <dneuer@futuristics.net>]
173 Changes with mod_fastcgi 2.2.1
175 *) Updates to the INSTALL doc to describe building as a DSO.
177 *) If the FastCgiIpcDir directive was in use and "httpd -t" was issued while
178 Apache was running ("apachectl restart" does this), the contents of the
179 dynamic directory (dynamic sockets and the mbox) would be blown away
182 *) Add an extern declaration for ap_sys_siglist (its exposed but not in any
183 Apache header file) to prevent a compile error on systems without
186 *) During auth requests, the subprocess_env table was left holding variables
187 sent to the auth FastCGI server (including REMOTE_PASSWD!) which means they
188 were passed to other processing phases (such as CGI/FastCGI). The
189 subprocess_env is now restored to its pre-auth condition.
191 *) Added a FastCGI Authorizer Role compatibility mode which implements the
192 specification to the tee. Use -compat arg with any of the Auth server
193 directives. This is intended for new and existing FastCGI authorizer
194 applications that require compatibility with other server implementations.
196 *) Fix logging from the process manager. Previously, all calls to the
197 logging routines used NULL for the server_rec. This works fine in a dev
198 environment, but results in most of these messages being tossed by
199 log_error_core() because DEFAULT_LOGLEVEL is too low. Odd logic.
201 *) Always setup fr->header so that in the event the FastCGI server
202 doesn't get a valid FastCGI protocol header over the wire, we
203 can still print our error message to the log from do_work().
205 *) Fix a compile error in open_connection_to_fs() on systems with TCP_NODELAY
206 defined (Don Locrasto [locrasto@iceminer.com]).
208 Changes with mod_fastcgi 2.2.0
210 *) Serious rewrite of mod_fastcgi.html. An example conf will have
211 to wait until another time - I'm burned out on docs updates.
213 *) Add to the mod_fastcgi to the server version string.
215 *) SCO doesn't like the const in the arg to inet_addr() and
218 *) Use Apache's NET_SIZE_T to get the correct arg type for getsockopt().
220 *) Deleted redundant header files from fcgi.h.
222 *) Full path names are no longer required for directives which take
223 file names as arguments. Like other Apache directives you can
224 specify paths from server_root (don't start with a "/").
226 *) The fcgiKillMgr is gone. This intermediate process sat between
227 the Apache parent and the FastCGI Process Manager under Apache
228 1.2. The process manager tries to name itself fcgi_pm (it used
231 *) Authentication, Authorization, and Access phases are now
232 supported. I've bent the FastCgi spec a bit: as many of the standard
233 environment variables (that were easy) are sent (the spec says don't
234 send some by name), all headers (except Status) sent by the auth
235 FastCgi server are passed to subprocesses (CGI/FastCGI invocations)
236 as environment variables rather than just those prefixed by
237 "Variable-". Custom responses for auth failures aren't supported
238 (yet.. we'll see what the demand for this is like). In addition to
239 the FastCGI protocol defined environment variable "FCGI_ROLE" being
240 set to "AUTHORIZER" an environment variable "FCGI_APACHE_ROLE" is
241 set indicating which of the three phases is being processed. See
242 the docs for info on the new per-directory directives:
244 FastCgiAuthenticator, FastCgiAuthenticatorAuthoritative,
245 FastCgiAuthorizer, FastCgiAuthorizerAuthoritative,
246 FastCgiAccessChecker, FastCgiAccessCheckerAuthoritative
248 *) The code has been broken into logical chunks making figuring it
249 out much easier. All of the #defines a user may want to change are
250 now in mod_fastcgi.h.
252 *) All the logging has been converted to the "supported" Apache
253 logging routines, ap_log_error() and ap_log_rerror(). Log entries
254 for request specific errors should now be directed to the correct
255 server errorlog (if your running more than one). Some attempt has
256 been made at setting appropriate log levels (comments of course are
257 welcome) and printing errno info when its of value. All of log
258 messages now have "FastCGI" in them (nice for grep).
260 *) All of the calls to Apache routines have been updated to the ap_
261 convention, eliminating the need for the compat header.
263 *) By default we no longer flush every piece of data we get from the
264 FastCGI server to the client. This allows the FastCGI server to
265 release without having to wait for the client flush to complete.
266 The old behaviour can revived with the -flush argument to AppClass.
268 *) I renamed some of the directives for consistency (the old names
271 AppClass -> FastCgiServer,
272 FCGIConfig -> FastCgiConfig,
273 ExternalAppClass -> FastCgiExternalServer
275 *) Directive arguments are no longer case sensitive.
277 *) The module now uses hard_timeout() rather than soft_timeout. This
278 means that the module will longjump out of the request if the
279 drop-dead timeout expires (set with Apache's Timeout directive, the
280 default is 5min) or if the client closes the connection (SIGPIPE).
281 This is more typical for a Apache modules and I'm not convinced we
282 properly handle all of the error cases well enough to use
283 soft_timeout (I can think of one place we don't anyway). This means
284 your FastCGI application can see SIGPIPE.
286 *) SIGPIPE is now ignored by default in FastCGI servers spawned by
287 Apache. Without this the default behaviour is exit().
289 *) It should (I think) run now under DSO now.
291 *) The bug which prevented sending of binary data in 2.1b1 is fixed.
293 *) The broken -initial-env bug ins 2.1b1 is fixed.
295 *) Maybe some other stuff.
299 *** Well there's a big gap here. Maybe I'll go back and extract the CVS
300 commit notes and stick 'em in here.. when I get some free time. ;)
305 *** Originally from docs/README..
307 Apache[X1.03.01]/mod_fastcgi[02.00.05] 19970909 unsupported
308 From: Stanley Gambarin <stanleyg@cs.bu.edu>
310 *) Yet more changes to the source distribution. Separated out
311 the Tcl dynamic string and buffer libraries into separate files.
312 Did the same for the OS library, however, it is not really an
313 abstraction... more like a bunch of wrappers. Added to DEVNOTES
314 a note about the problem of Apache not allowing including header
317 Apache[X1.03.01]/mod_fastcgi[02.00.04] 19970908 unsupported
318 From: Stanley Gambarin <stanleyg@cs.bu.edu>
320 *) More changes to the installation scripts. Theoretically we
321 should be able to to use the same Makefile/installation script
322 for both Apache 1.2.x and 1.3.x. Need some sort of the abstraction
323 layer to use the same source code for both 1.2.x and 1.3.x sources.
325 Apache[X1.03.01]/mod_fastcgi[02.00.03] 19970905 unsupported
326 From: Stanley Gambarin <stanleyg@cs.bu.edu>
328 *) Created Makefile to compile the module into the archive library,
329 just like the proxy module. Also initial draft of the installation
330 script, which should work for both 1.2 and 1.3 sources of Apache.
332 Apache[01.02.04]/mod_fastcgi[02.00.02] 19970903 unsupported
333 From: Stanley Gambarin <stanleyg@cs.bu.edu>
335 *) More source reorg. Separated out some header files. Need much
336 more work on this. Updated the TODO file. Added fcgivers.h and
337 CHANGES file to track the history.
339 Apache[01.02.04]/mod_fastcgi[02.00.01] 19970902 unsupported
340 From: Stanley Gambarin <stanleyg@cs.bu.edu>
342 *) Source reorganization. This is done to provide a basis for later
343 functionality implementation. See TODO file for more information
344 of what possible future enhancements are possible/desired.
346 Apache[01.02.04]/mod_fastcgi[02.00.00] 19970902 unsupported
347 From: David MacKenzie <djm@va.pubnix.com>
349 *) Create the "dynamic" dir and "mbox" as the user specified in the
350 User and Group directives instead of as root. They are created with
351 restrictive permissions, and then the module checks to see whether the
352 user specified in the User and Group directives has read, write, and
353 execute permissions on them. Those permissions had been explicitly
354 denied by creating them as "root".
356 *) Some documentation was garbled or incomplete.
358 *) When a FastCGI application can't be execl'd, the code used a
359 value of errno that may have been stomped on by intervening system
360 calls. That happened on BSD/OS 2.1, where the error_log was
361 reporting errno=25 (NOTTY) instead of the correct errno=13 (EACCES).
362 An intervening stdio call was setting errno as a side-effect, which
363 it has the right to do.
365 *) Erroneous fprintf arguments, a missing return value, an unused
366 function, missing declarations, and other problems detected by gcc
369 *) Duplicated code to create the lock file name merged into a single
370 function, closing several memory leaks.
372 *) FCGIConfig -minProcesses didn't allow a value of 0, so the last
373 FCGI app in a given dynamic class would live forever, even if it
374 hadn't been requested for weeks.
376 *) The test for whether to keep looking for dynamic app victims to
377 kill was backwards, so no dynamic apps were ever selected as
380 *) The killInterval and updateInterval were being ignored; the
381 sigsuspend() forced recalculations only when a child died,
382 instead of at the intervals specified.
384 *) When a dynamic app couldn't be started, the program SEGV'd when
385 trying to free the ipcAddrPtr twice.
387 *) Remove the dead lock file and socket when cleaning up after a
388 server whose last child has been killed, so the FastCGIHandler
389 isn't fooled into thinking there is still a process serving
392 *) Make file locking robust in the presence of signals.
394 *) Rename some badly named variables.
396 *) Fix typos in many comments.
398 *) Fix some memory and file descriptor leaks.
400 *) Make the blocking kill of a server closer to working.
402 *) More fixes to calculations in dynamic application management.
405 *** Originally from docs/README.OMI..
407 What's New: Version 2.0b1, 16 Apr 1997
409 *) Implemented a mechanism by which FastCGI applications are started by
410 the web server on the first request and continue running. Also, a
411 heuristics have been implemented to allow for dynamic killing of the
412 running FastCGI apps. The configuration options are supplied via
413 the new FCGIConfig directive, that is described in the
416 *) When performing internal redirect, since the original request's body
417 has already been read, do not allow the redirected request to think
418 that is has one [body].
420 *) More conditional compilation for OS/2.
422 *) Fixed occassional "Assertion failed: len > 0, file mod_fastcgi.c"
424 *) Fixed "failed assertion `count >= 0 && count <= bufPtr->length'"
426 *) Added bflush() call after bwrite() in DrainReqOutBuf function to
427 force some output to be written in the period of inactivity. This
428 circumvents the Apache's buffering during the server-push.
430 *) Added a header-parser function placeholder for compatibility with an
433 *) Additional checks have been placed with regard to permissions of the
434 FastCGI processes. These include the existence of the file,
435 ExecCGI and IncludesNOEXEC checks, disallowing of nph- scripts, etc.
437 *) Implemented a restart cleanup, so that no parentlesses FastCGI
438 applications are left after Apache is restarted and/or terminated.
439 This was accomplished via two independent mechanisms, where the
440 first one prevents a start of the process manager on the first
441 reading of the configuration files. The second mechanism deals away
442 from Apache's implementation of fork() (via spawn_child()) and
443 implements its own forking, thus removing any dependencies on the
444 Apache to cleanup processes during termination phase.
446 *) Removed the definition of the Sigfunc, as Apache 2.0b10 defines it
447 itself in the file conf.h
449 What's New: Version 1.4.3, 15 Jan 1997
451 *) Fixed compilation warnings for various platforms, as well as
452 conditional compilation for OS/2.
454 What's New: Version 1.4.2, 12 Dec 1996
456 *) mod_fastcgi.c is ported to Apache 1.2b. Any further development
457 will proceed under this version of Apache web server.
459 *) As the result of porting, the "include virtual" construct of SSI
460 will now work correctly using either <Location> or <Directory>
463 What's New: Version 1.4.1, 4 Dec 1996
465 *) Checks have been removed from the ScanCGIHeaders that provided
466 for the presence of both Status and Location headers as being
467 an error. Contradictory to CGI/1.1 Internet Draft, both of
468 these headers are used by the current CGI applications.
470 What's New: Version 1.4, 22 Nov 1996
472 *) Added the -port option to AppClass, allowing TCP/IP communication.
473 Added the -socket option to AppClass, allowing Unix domain
474 communication via a configurable pathname.
476 *) Added the ExternalAppClass directive, allowing TCP/IP
477 communication with remote FastCGI applications.
479 *) The handler had its own code for generating HTTP response headers;
480 now it uses Apache's. This reduces the size of the module.
481 More importantly, it fixes the bug in which "include virtual"
482 sees the HTTP response headers.
484 *) The response header parser performed very little checking.
485 Now the parser enforces the guidelines in the CGI/1.1
486 Internet-Draft: Status and Location are mutually exclusive,
487 Location can only be a response to GET or HEAD,
488 CGI response headers can't be repeated, etc. (The CGI response
489 headers are the ones the handler interprets: Status, Location,
492 *) The response header parser used to miss CGI headers
493 with no whitespace after the colon, e.g. "Status:200 OK".
495 *) The response header parser sometimes interpreted the first line
496 of content as an RFC822 continuation line.
498 *) The handler implemented a nonstandard version of Location
499 which never used internal redirects. The handler
500 now attempts to implement Location as specified in the
501 CGI/1.1 Internet-Draft. The module documentation explains
504 *) Error log entries from the response header parser were
505 pretty uninformative; they are better now. If a header is
506 malformed, the log entry includes it. If the headers are
507 unterminated rather than malformed, the log entry says that,
508 and says how many bytes were received from the app.
510 *) When the application manager forked a new process, and that
511 process ran into trouble before executing the first instruction
512 of application code, the process used to exit with status =
513 errno. This made certain configuration problems (e.g. incorrect
514 file permissions when server parent is root) quite difficult
515 to pin down. Now the failing child process opens up the error log
516 and writes an informative entry before exiting.
518 *) The module now correctly handles a slash at the end of the
519 DocumentRoot directive. This was a one-line fix.
521 What's New: Version 1.3.3, 17 Oct 1996
523 *) The module now registers its request handler under the name
524 fastcgi-script in addition to the name application/x-httpd-fcgi.
525 This was a one-line addition to the module, but had quite
526 a large impact on the documentation and sample configuration.
528 What's New: Version 1.3.2, 27 Sept 1996
530 *) On some systems (SunOS, Linux?), fopen for append has a bug
531 that strikes when two processes append to the same file.
532 This bug causes the process manager to corrupt the error log.
533 Work around the bug by calling open, then fdopen. A patch
534 to Apache 1.1.1 is also required, as described in mod_fastcgi.html.
535 (Reported by Bob Ramstad.)
537 What's New: Version 1.3.1, 17 Sept 1996
539 *) On Linux and SunOS, the Apache default user_id (-1) is not a
540 legal value of uid_t, so cast it. (Reported by Bob Ramstad,
541 Scott Langley, others.)
543 *) On some systems (Linux, some Solaris?, Irix?), connect requires
544 write access to a Unix Domain socket, so provide it. (Reported
545 by Scott Langley, freeform@wired.com, others.)
547 *) If you hit a .fcg file, but there's no AppClass defined, the
548 error should be NOT_FOUND instead of SERVER_ERROR. (Reported
551 What's New: Version 1.3, 4 Sept 1996
553 *) Module sometimes busy-waited in FastCgiDoWork. Fixed.
555 *) Module violated the Apache buff abstraction.
556 Fixed. Might now work with SSL (not tested).
557 As part of the fix, eliminated support for
558 Apache 1.0x versions.
560 *) Module failed to chmod the Unix domain listening sockets
561 it created, so protections were set according to the current
564 *) Module forked too many process manager processes: One
565 per AppClass. When the Apache parent ran as root, these
566 process manager processes ran as root. New process
567 manager is a single process (only started if AppClass is
568 used) and runs with same privileges as other children.
569 The new process manager doesn't do any polling, so
570 there's less system overhead than before.
572 *) AppClass insisted on getting at least two arguments. Fixed.
574 *) Module used setjmp/longjmp, causing compiler warnings
575 on some platforms (e.g. Linux.) The module no longer
578 *) Module created listening sockets in /tmp, where they were
579 sometimes wiped out by cleanup scripts. Added FastCgiIpcDir
580 directive to give control over the location of listening
583 *) Module wrote error log entries without a timestamp. Fixed.
585 *) AppClass directive wrote error messages to stderr in addition
586 to returning a char * message to the Apache core. Fixed.
588 What's New: Version 1.2, 3 June 1996
590 *) Ported from Apache-1.0-based code to Apache-1.1b2 internals by
593 Add version string: APACHEVERSION macro to mod_fastcgi.c
595 *) chown FastCGI socket to user_id and setuid to user_id for app
598 *) Modify GetFromStream() by having it call an OS dependent
599 function GetStreamSize(FILE *) which uses FILE internal data member.
600 Linux users might need to modify GetStreamSize().
602 What's New: Version 1.1, 10 May 1996
604 *) If you specify a non-existent executable in the AppClass directive,
605 or if the file exists and it does not have execute permission you
606 get a constant stream of error messages telling you that "program
607 terminated due to a signal".
609 *) The mod_fastcgi module should use the standard Apache error logging
610 facility instead of writing to stderr.
612 Version 1.0, 30 April 1996