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