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