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