3 *) Don't invoke suexec when the user/group for the fastcgi application
4 is the same as the apache main server. This is consistent with
5 apache's suexec handling.
7 *) Reset the apache drop dead timer upon successful read or writes
8 to/from the client. This eliminates timeouts that were occuring
9 during the large file transfers to/from slow clients.
11 *) Support generic wrappers such as cgiwrap by eliminating dependencies
12 on Apache's SUEXEC, renaming the FastCgiSuexec directive
13 FastCgiWrapper and eliminating any checks regarding the target
14 application (this is the repsonibility of the wrapper).
16 *) Fix a nasty bug that occurred when a client aborted a request
17 before the connection to a dynamic FastCGI application was opened.
18 The application's lock file descriptor wasn't setup but was being
19 closed which resulted in FD0 being closed. Normally this open to
20 /dev/null and should pose no problems except that because the FD
21 was available it was being returned by Apache's accept(). This
22 caused it to be registered for pool cleanup. mod_cgi though moves
23 the CGI stdin pipe to FD0 and thus it was getting waxed during pool
24 cleanup. Problem identified by checksum@163.net.
26 Changes with mod_fastcgi 2.2.8
28 *) Eliminate the concept of disabled applications. If a failure
29 occurs trying to setup an application (e.g. bind() error) its
30 tried repeatedly every init-start-delay seconds.
32 *) Tweak to Makefile.tmpl to support DSOs. Dave Hill [ddhill@zk3.dec.com]
34 Changes with mod_fastcgi 2.2.6
36 *) Shutdown the PM when Apache appears to have disappeared.
38 *) seteuid() tweak for HP-UX 11. Milton L. Hankins [mlh@swl.msd.ray.com]
40 *) (Win32) More dynamic fixes.
42 *) (Win32) Eliminate the per application named pipe mutex. Its now
43 per process to keep the lib happy (the use _FCGI_MUTEX_ should
44 removed from the lib). This allows pipe-based applications to
45 now handle multiple simoultanous requests (one per process).
47 *) Increase the number of open FDs we look to close when spawning apps.
49 *) Prevent an assert from popping unnecessarily.
51 *) (Win32) Add support for interpretter scripts.
53 *) (Win32) Fix named pipe handling (problems with large responses).
55 *) (Win32) Remove the "can exec" check.
57 Changes with mod_fastcgi 2.2.4
59 *) Beta WinNT support. David Allen [djallen@raleigh.ibm.com] and
60 Rob Saccoccio [robs@chelsea.net]
62 *) Remove request type restriction (GET and POST only) in order to support
63 Web DAV requests. Scott Robertson [sroberts@codeit.com]
65 *) Allow requests for URLs such as /server/some/other/qualifier to match a
66 FastCGI server defined as /server. This was done primarily for Java
67 Servlets, but is generically useful.
69 *) Change the comm between the PM and the request handlers from a regular
70 file to a pipe and an assortment of other dynamic fixes.
72 *) Log dynamic process termination scheduling and the resulting
73 process exit notification.
75 *) Change the default singleThreshhold from 10 to 0 to prevent the last dynamic
76 application process from being killed off due to low demand.
78 *) Clean up FastCGI application pathnames (e.g. remove duplicate slashes).
80 *) Fix bugs that prevented dynamic processes from being shutdown when
81 the load subsided. Lars Heete [heete@do.isst.fhg.de]
83 *) Prevent dynamic processes from being scheduled to be started before
84 the init-start-delay or restart-delay period expires. This prevents
85 the queuing of process start requests while an application which has
86 a long initialization period starts up.
88 *) When suexec is enabled, allow the process manager to become root again in
89 order to signal applications it spawned.
91 *) Change the "server started" log message level from INFO to WARN.
93 *) Fix dynamic server "Connection Refused" handling.
95 *) Fix demand-based dynamic process spawing (uses a more portable approach).
97 *) Add -idle-timeout arg to FastCgiServer, FastCgiExternalServer, and
98 FastCgiConfig. mod_fastcgi will now abort a connection if inactive for
99 longer than this period. It applies to the initiation of connections as
100 well (and thus is similar to appConnTimeout). Default is 30 seconds.
102 *) appConnTimeout is 0 by default now resulting in blocking connect()s. This
103 is more platform portable/predicable.
105 *) Leave STDOUT and STDERR open to the main server error_log. This should
106 eliminate a pile of protocol errors and having to track down 3rd party
107 libs writing to stderr inadvertantly. This doesn't mean these shouldn't
108 be fixed in the application to use the FastCGI I/O, it just allows these
109 applications to run when previously they'd crash and burn.
111 *) Add the pass-header arg to server directives.
113 *) Allow supplementary groups to be passed to spawned FastCGI applications (as
114 is done for CGI) - at least until PR2580 is resolved.
116 *) Initialize the default (empty) environment in a more socially acceptable
119 *) Miscellaneous doc improvements based on notes I've collected up over the
122 *) Fix a bug in the stderr handling introduced in 2.2.2. There were conditions
123 that resulted in not or improperly terminated strings.
125 *) Fix the call to setsockopt() to disable Nagle. [based on a bug report by
126 Johannes Plassmann <johannes.plassmann@pdb.siemens.de>]
128 Changes with mod_fastcgi 2.2.2
130 *) Added support for blocking connect()s by setting appConnTimeout to 0.
131 Non-blocking connect()s (the default) can be troublesome on some platforms.
132 Its expected that blocking connect()s will become the default (and
133 non-blocking connect()s will become optional) in the next release.
135 *) Dump a compile time error if the version of Apache is too old.
137 *) Wrap the SIGPIPE handler manipulation code such that it is only applicable
138 to Apache releases prior to 1.3.6.
140 *) Minor tweaks for RUSSIAN_APACHE. [Sergey Gershtein <sg@mplik.ru>]
144 Always restart a failed dynamic application if it is the last instance.
145 This means there once started there will always be at least one process
146 instance of a dynamic application.
148 Send PLEASE_START to the PM when a connect() results in an ECONNREFUSED.
149 ECONNREFUSED means the listen queue is full (or there isn't one). Asking
150 the PM to start (another) application instance may help empty it faster.
152 Change two sleep() calls to select() based snoozes because alarm() is in
153 effect and sleep() and alarm() don't always play nice together.
155 Fixed a couple of error messages.
157 *) Fix -listen-queue-depth arg on FastCgiConfig (dynamic). Previously it
158 was ignored and the default was always used.
160 *) Allow the -initial-env argument to be used to pass variables from the
161 Apache process environment to the FastCGI server (by specifying a
162 variable name without the "=" or a value). [suggested by
163 Martin Lichtin <lichtin@oanda.com>]
165 *) Cleanup some debug macros.
167 *) Improved script stderr handling. [based on suggestions by
168 David Birnbaum <davidb@chelsea.net>]
170 *) Added IRIX and FreeBSD to the list of supported platforms and other minor
171 updates to the INSTALL doc.
173 *) Changed the default listen-queue-depth (FCGI_DEFAULT_LISTEN_Q) from
174 5 to 100. Its still configurable with the -listen-queue-depth option.
175 This should help eliminate the FAQ - Why do I see "Connection Refused"
178 *) Fix a bug in FastCgiExternalServer that broke support for external servers
179 on other hosts. [Dave Neuer <dneuer@futuristics.net>]
181 Changes with mod_fastcgi 2.2.1
183 *) Updates to the INSTALL doc to describe building as a DSO.
185 *) If the FastCgiIpcDir directive was in use and "httpd -t" was issued while
186 Apache was running ("apachectl restart" does this), the contents of the
187 dynamic directory (dynamic sockets and the mbox) would be blown away
190 *) Add an extern declaration for ap_sys_siglist (its exposed but not in any
191 Apache header file) to prevent a compile error on systems without
194 *) During auth requests, the subprocess_env table was left holding variables
195 sent to the auth FastCGI server (including REMOTE_PASSWD!) which means they
196 were passed to other processing phases (such as CGI/FastCGI). The
197 subprocess_env is now restored to its pre-auth condition.
199 *) Added a FastCGI Authorizer Role compatibility mode which implements the
200 specification to the tee. Use -compat arg with any of the Auth server
201 directives. This is intended for new and existing FastCGI authorizer
202 applications that require compatibility with other server implementations.
204 *) Fix logging from the process manager. Previously, all calls to the
205 logging routines used NULL for the server_rec. This works fine in a dev
206 environment, but results in most of these messages being tossed by
207 log_error_core() because DEFAULT_LOGLEVEL is too low. Odd logic.
209 *) Always setup fr->header so that in the event the FastCGI server
210 doesn't get a valid FastCGI protocol header over the wire, we
211 can still print our error message to the log from do_work().
213 *) Fix a compile error in open_connection_to_fs() on systems with TCP_NODELAY
214 defined (Don Locrasto [locrasto@iceminer.com]).
216 Changes with mod_fastcgi 2.2.0
218 *) Serious rewrite of mod_fastcgi.html. An example conf will have
219 to wait until another time - I'm burned out on docs updates.
221 *) Add to the mod_fastcgi to the server version string.
223 *) SCO doesn't like the const in the arg to inet_addr() and
226 *) Use Apache's NET_SIZE_T to get the correct arg type for getsockopt().
228 *) Deleted redundant header files from fcgi.h.
230 *) Full path names are no longer required for directives which take
231 file names as arguments. Like other Apache directives you can
232 specify paths from server_root (don't start with a "/").
234 *) The fcgiKillMgr is gone. This intermediate process sat between
235 the Apache parent and the FastCGI Process Manager under Apache
236 1.2. The process manager tries to name itself fcgi_pm (it used
239 *) Authentication, Authorization, and Access phases are now
240 supported. I've bent the FastCgi spec a bit: as many of the standard
241 environment variables (that were easy) are sent (the spec says don't
242 send some by name), all headers (except Status) sent by the auth
243 FastCgi server are passed to subprocesses (CGI/FastCGI invocations)
244 as environment variables rather than just those prefixed by
245 "Variable-". Custom responses for auth failures aren't supported
246 (yet.. we'll see what the demand for this is like). In addition to
247 the FastCGI protocol defined environment variable "FCGI_ROLE" being
248 set to "AUTHORIZER" an environment variable "FCGI_APACHE_ROLE" is
249 set indicating which of the three phases is being processed. See
250 the docs for info on the new per-directory directives:
252 FastCgiAuthenticator, FastCgiAuthenticatorAuthoritative,
253 FastCgiAuthorizer, FastCgiAuthorizerAuthoritative,
254 FastCgiAccessChecker, FastCgiAccessCheckerAuthoritative
256 *) The code has been broken into logical chunks making figuring it
257 out much easier. All of the #defines a user may want to change are
258 now in mod_fastcgi.h.
260 *) All the logging has been converted to the "supported" Apache
261 logging routines, ap_log_error() and ap_log_rerror(). Log entries
262 for request specific errors should now be directed to the correct
263 server errorlog (if your running more than one). Some attempt has
264 been made at setting appropriate log levels (comments of course are
265 welcome) and printing errno info when its of value. All of log
266 messages now have "FastCGI" in them (nice for grep).
268 *) All of the calls to Apache routines have been updated to the ap_
269 convention, eliminating the need for the compat header.
271 *) By default we no longer flush every piece of data we get from the
272 FastCGI server to the client. This allows the FastCGI server to
273 release without having to wait for the client flush to complete.
274 The old behaviour can revived with the -flush argument to AppClass.
276 *) I renamed some of the directives for consistency (the old names
279 AppClass -> FastCgiServer,
280 FCGIConfig -> FastCgiConfig,
281 ExternalAppClass -> FastCgiExternalServer
283 *) Directive arguments are no longer case sensitive.
285 *) The module now uses hard_timeout() rather than soft_timeout. This
286 means that the module will longjump out of the request if the
287 drop-dead timeout expires (set with Apache's Timeout directive, the
288 default is 5min) or if the client closes the connection (SIGPIPE).
289 This is more typical for a Apache modules and I'm not convinced we
290 properly handle all of the error cases well enough to use
291 soft_timeout (I can think of one place we don't anyway). This means
292 your FastCGI application can see SIGPIPE.
294 *) SIGPIPE is now ignored by default in FastCGI servers spawned by
295 Apache. Without this the default behaviour is exit().
297 *) It should (I think) run now under DSO now.
299 *) The bug which prevented sending of binary data in 2.1b1 is fixed.
301 *) The broken -initial-env bug ins 2.1b1 is fixed.
303 *) Maybe some other stuff.
307 *** Well there's a big gap here. Maybe I'll go back and extract the CVS
308 commit notes and stick 'em in here.. when I get some free time. ;)
313 *** Originally from docs/README..
315 Apache[X1.03.01]/mod_fastcgi[02.00.05] 19970909 unsupported
316 From: Stanley Gambarin <stanleyg@cs.bu.edu>
318 *) Yet more changes to the source distribution. Separated out
319 the Tcl dynamic string and buffer libraries into separate files.
320 Did the same for the OS library, however, it is not really an
321 abstraction... more like a bunch of wrappers. Added to DEVNOTES
322 a note about the problem of Apache not allowing including header
325 Apache[X1.03.01]/mod_fastcgi[02.00.04] 19970908 unsupported
326 From: Stanley Gambarin <stanleyg@cs.bu.edu>
328 *) More changes to the installation scripts. Theoretically we
329 should be able to to use the same Makefile/installation script
330 for both Apache 1.2.x and 1.3.x. Need some sort of the abstraction
331 layer to use the same source code for both 1.2.x and 1.3.x sources.
333 Apache[X1.03.01]/mod_fastcgi[02.00.03] 19970905 unsupported
334 From: Stanley Gambarin <stanleyg@cs.bu.edu>
336 *) Created Makefile to compile the module into the archive library,
337 just like the proxy module. Also initial draft of the installation
338 script, which should work for both 1.2 and 1.3 sources of Apache.
340 Apache[01.02.04]/mod_fastcgi[02.00.02] 19970903 unsupported
341 From: Stanley Gambarin <stanleyg@cs.bu.edu>
343 *) More source reorg. Separated out some header files. Need much
344 more work on this. Updated the TODO file. Added fcgivers.h and
345 CHANGES file to track the history.
347 Apache[01.02.04]/mod_fastcgi[02.00.01] 19970902 unsupported
348 From: Stanley Gambarin <stanleyg@cs.bu.edu>
350 *) Source reorganization. This is done to provide a basis for later
351 functionality implementation. See TODO file for more information
352 of what possible future enhancements are possible/desired.
354 Apache[01.02.04]/mod_fastcgi[02.00.00] 19970902 unsupported
355 From: David MacKenzie <djm@va.pubnix.com>
357 *) Create the "dynamic" dir and "mbox" as the user specified in the
358 User and Group directives instead of as root. They are created with
359 restrictive permissions, and then the module checks to see whether the
360 user specified in the User and Group directives has read, write, and
361 execute permissions on them. Those permissions had been explicitly
362 denied by creating them as "root".
364 *) Some documentation was garbled or incomplete.
366 *) When a FastCGI application can't be execl'd, the code used a
367 value of errno that may have been stomped on by intervening system
368 calls. That happened on BSD/OS 2.1, where the error_log was
369 reporting errno=25 (NOTTY) instead of the correct errno=13 (EACCES).
370 An intervening stdio call was setting errno as a side-effect, which
371 it has the right to do.
373 *) Erroneous fprintf arguments, a missing return value, an unused
374 function, missing declarations, and other problems detected by gcc
377 *) Duplicated code to create the lock file name merged into a single
378 function, closing several memory leaks.
380 *) FCGIConfig -minProcesses didn't allow a value of 0, so the last
381 FCGI app in a given dynamic class would live forever, even if it
382 hadn't been requested for weeks.
384 *) The test for whether to keep looking for dynamic app victims to
385 kill was backwards, so no dynamic apps were ever selected as
388 *) The killInterval and updateInterval were being ignored; the
389 sigsuspend() forced recalculations only when a child died,
390 instead of at the intervals specified.
392 *) When a dynamic app couldn't be started, the program SEGV'd when
393 trying to free the ipcAddrPtr twice.
395 *) Remove the dead lock file and socket when cleaning up after a
396 server whose last child has been killed, so the FastCGIHandler
397 isn't fooled into thinking there is still a process serving
400 *) Make file locking robust in the presence of signals.
402 *) Rename some badly named variables.
404 *) Fix typos in many comments.
406 *) Fix some memory and file descriptor leaks.
408 *) Make the blocking kill of a server closer to working.
410 *) More fixes to calculations in dynamic application management.
413 *** Originally from docs/README.OMI..
415 What's New: Version 2.0b1, 16 Apr 1997
417 *) Implemented a mechanism by which FastCGI applications are started by
418 the web server on the first request and continue running. Also, a
419 heuristics have been implemented to allow for dynamic killing of the
420 running FastCGI apps. The configuration options are supplied via
421 the new FCGIConfig directive, that is described in the
424 *) When performing internal redirect, since the original request's body
425 has already been read, do not allow the redirected request to think
426 that is has one [body].
428 *) More conditional compilation for OS/2.
430 *) Fixed occassional "Assertion failed: len > 0, file mod_fastcgi.c"
432 *) Fixed "failed assertion `count >= 0 && count <= bufPtr->length'"
434 *) Added bflush() call after bwrite() in DrainReqOutBuf function to
435 force some output to be written in the period of inactivity. This
436 circumvents the Apache's buffering during the server-push.
438 *) Added a header-parser function placeholder for compatibility with an
441 *) Additional checks have been placed with regard to permissions of the
442 FastCGI processes. These include the existence of the file,
443 ExecCGI and IncludesNOEXEC checks, disallowing of nph- scripts, etc.
445 *) Implemented a restart cleanup, so that no parentlesses FastCGI
446 applications are left after Apache is restarted and/or terminated.
447 This was accomplished via two independent mechanisms, where the
448 first one prevents a start of the process manager on the first
449 reading of the configuration files. The second mechanism deals away
450 from Apache's implementation of fork() (via spawn_child()) and
451 implements its own forking, thus removing any dependencies on the
452 Apache to cleanup processes during termination phase.
454 *) Removed the definition of the Sigfunc, as Apache 2.0b10 defines it
455 itself in the file conf.h
457 What's New: Version 1.4.3, 15 Jan 1997
459 *) Fixed compilation warnings for various platforms, as well as
460 conditional compilation for OS/2.
462 What's New: Version 1.4.2, 12 Dec 1996
464 *) mod_fastcgi.c is ported to Apache 1.2b. Any further development
465 will proceed under this version of Apache web server.
467 *) As the result of porting, the "include virtual" construct of SSI
468 will now work correctly using either <Location> or <Directory>
471 What's New: Version 1.4.1, 4 Dec 1996
473 *) Checks have been removed from the ScanCGIHeaders that provided
474 for the presence of both Status and Location headers as being
475 an error. Contradictory to CGI/1.1 Internet Draft, both of
476 these headers are used by the current CGI applications.
478 What's New: Version 1.4, 22 Nov 1996
480 *) Added the -port option to AppClass, allowing TCP/IP communication.
481 Added the -socket option to AppClass, allowing Unix domain
482 communication via a configurable pathname.
484 *) Added the ExternalAppClass directive, allowing TCP/IP
485 communication with remote FastCGI applications.
487 *) The handler had its own code for generating HTTP response headers;
488 now it uses Apache's. This reduces the size of the module.
489 More importantly, it fixes the bug in which "include virtual"
490 sees the HTTP response headers.
492 *) The response header parser performed very little checking.
493 Now the parser enforces the guidelines in the CGI/1.1
494 Internet-Draft: Status and Location are mutually exclusive,
495 Location can only be a response to GET or HEAD,
496 CGI response headers can't be repeated, etc. (The CGI response
497 headers are the ones the handler interprets: Status, Location,
500 *) The response header parser used to miss CGI headers
501 with no whitespace after the colon, e.g. "Status:200 OK".
503 *) The response header parser sometimes interpreted the first line
504 of content as an RFC822 continuation line.
506 *) The handler implemented a nonstandard version of Location
507 which never used internal redirects. The handler
508 now attempts to implement Location as specified in the
509 CGI/1.1 Internet-Draft. The module documentation explains
512 *) Error log entries from the response header parser were
513 pretty uninformative; they are better now. If a header is
514 malformed, the log entry includes it. If the headers are
515 unterminated rather than malformed, the log entry says that,
516 and says how many bytes were received from the app.
518 *) When the application manager forked a new process, and that
519 process ran into trouble before executing the first instruction
520 of application code, the process used to exit with status =
521 errno. This made certain configuration problems (e.g. incorrect
522 file permissions when server parent is root) quite difficult
523 to pin down. Now the failing child process opens up the error log
524 and writes an informative entry before exiting.
526 *) The module now correctly handles a slash at the end of the
527 DocumentRoot directive. This was a one-line fix.
529 What's New: Version 1.3.3, 17 Oct 1996
531 *) The module now registers its request handler under the name
532 fastcgi-script in addition to the name application/x-httpd-fcgi.
533 This was a one-line addition to the module, but had quite
534 a large impact on the documentation and sample configuration.
536 What's New: Version 1.3.2, 27 Sept 1996
538 *) On some systems (SunOS, Linux?), fopen for append has a bug
539 that strikes when two processes append to the same file.
540 This bug causes the process manager to corrupt the error log.
541 Work around the bug by calling open, then fdopen. A patch
542 to Apache 1.1.1 is also required, as described in mod_fastcgi.html.
543 (Reported by Bob Ramstad.)
545 What's New: Version 1.3.1, 17 Sept 1996
547 *) On Linux and SunOS, the Apache default user_id (-1) is not a
548 legal value of uid_t, so cast it. (Reported by Bob Ramstad,
549 Scott Langley, others.)
551 *) On some systems (Linux, some Solaris?, Irix?), connect requires
552 write access to a Unix Domain socket, so provide it. (Reported
553 by Scott Langley, freeform@wired.com, others.)
555 *) If you hit a .fcg file, but there's no AppClass defined, the
556 error should be NOT_FOUND instead of SERVER_ERROR. (Reported
559 What's New: Version 1.3, 4 Sept 1996
561 *) Module sometimes busy-waited in FastCgiDoWork. Fixed.
563 *) Module violated the Apache buff abstraction.
564 Fixed. Might now work with SSL (not tested).
565 As part of the fix, eliminated support for
566 Apache 1.0x versions.
568 *) Module failed to chmod the Unix domain listening sockets
569 it created, so protections were set according to the current
572 *) Module forked too many process manager processes: One
573 per AppClass. When the Apache parent ran as root, these
574 process manager processes ran as root. New process
575 manager is a single process (only started if AppClass is
576 used) and runs with same privileges as other children.
577 The new process manager doesn't do any polling, so
578 there's less system overhead than before.
580 *) AppClass insisted on getting at least two arguments. Fixed.
582 *) Module used setjmp/longjmp, causing compiler warnings
583 on some platforms (e.g. Linux.) The module no longer
586 *) Module created listening sockets in /tmp, where they were
587 sometimes wiped out by cleanup scripts. Added FastCgiIpcDir
588 directive to give control over the location of listening
591 *) Module wrote error log entries without a timestamp. Fixed.
593 *) AppClass directive wrote error messages to stderr in addition
594 to returning a char * message to the Apache core. Fixed.
596 What's New: Version 1.2, 3 June 1996
598 *) Ported from Apache-1.0-based code to Apache-1.1b2 internals by
601 Add version string: APACHEVERSION macro to mod_fastcgi.c
603 *) chown FastCGI socket to user_id and setuid to user_id for app
606 *) Modify GetFromStream() by having it call an OS dependent
607 function GetStreamSize(FILE *) which uses FILE internal data member.
608 Linux users might need to modify GetStreamSize().
610 What's New: Version 1.1, 10 May 1996
612 *) If you specify a non-existent executable in the AppClass directive,
613 or if the file exists and it does not have execute permission you
614 get a constant stream of error messages telling you that "program
615 terminated due to a signal".
617 *) The mod_fastcgi module should use the standard Apache error logging
618 facility instead of writing to stderr.
620 Version 1.0, 30 April 1996