1 Changes with mod_fastcgi 2.2.5
3 *) (Win32) More dynamic fixes.
5 *) (Win32) Eliminate the per application named pipe mutex. Its now
6 per process to keep the lib happy (the use _FCGI_MUTEX_ should
7 removed from the lib). This allows pipe-based applications to
8 now handle multiple simoultanous requests (one per process).
10 *) Increase the number of open FDs we look to close when spawning apps.
12 *) Prevent an assert from popping unnecessarily.
14 *) (Win32) Add support for interpretter scripts.
16 *) (Win32) Fix named pipe handling (problems with large responses).
18 *) (Win32) Remove the "can exec" check.
20 Changes with mod_fastcgi 2.2.4
22 *) Beta WinNT support. David Allen [djallen@raleigh.ibm.com] and
23 Rob Saccoccio [robs@chelsea.net]
25 *) Remove request type restriction (GET and POST only) in order to support
26 Web DAV requests. Scott Robertson [sroberts@codeit.com]
28 *) Allow requests for URLs such as /server/some/other/qualifier to match a
29 FastCGI server defined as /server. This was done primarily for Java
30 Servlets, but is generically useful.
32 *) Change the comm between the PM and the request handlers from a regular
33 file to a pipe and an assortment of other dynamic fixes.
35 *) Log dynamic process termination scheduling and the resulting
36 process exit notification.
38 *) Change the default singleThreshhold from 10 to 0 to prevent the last dynamic
39 application process from being killed off due to low demand.
41 *) Clean up FastCGI application pathnames (e.g. remove duplicate slashes).
43 *) Fix bugs that prevented dynamic processes from being shutdown when
44 the load subsided. Lars Heete [heete@do.isst.fhg.de]
46 *) Prevent dynamic processes from being scheduled to be started before
47 the init-start-delay or restart-delay period expires. This prevents
48 the queuing of process start requests while an application which has
49 a long initialization period starts up.
51 *) When suexec is enabled, allow the process manager to become root again in
52 order to signal applications it spawned.
54 *) Change the "server started" log message level from INFO to WARN.
56 *) Fix dynamic server "Connection Refused" handling.
58 *) Fix demand-based dynamic process spawing (uses a more portable approach).
60 *) Add -idle-timeout arg to FastCgiServer, FastCgiExternalServer, and
61 FastCgiConfig. mod_fastcgi will now abort a connection if inactive for
62 longer than this period. It applies to the initiation of connections as
63 well (and thus is similar to appConnTimeout). Default is 30 seconds.
65 *) appConnTimeout is 0 by default now resulting in blocking connect()s. This
66 is more platform portable/predicable.
68 *) Leave STDOUT and STDERR open to the main server error_log. This should
69 eliminate a pile of protocol errors and having to track down 3rd party
70 libs writing to stderr inadvertantly. This doesn't mean these shouldn't
71 be fixed in the application to use the FastCGI I/O, it just allows these
72 applications to run when previously they'd crash and burn.
74 *) Add the pass-header arg to server directives.
76 *) Allow supplementary groups to be passed to spawned FastCGI applications (as
77 is done for CGI) - at least until PR2580 is resolved.
79 *) Initialize the default (empty) environment in a more socially acceptable
82 *) Miscellaneous doc improvements based on notes I've collected up over the
85 *) Fix a bug in the stderr handling introduced in 2.2.2. There were conditions
86 that resulted in not or improperly terminated strings.
88 *) Fix the call to setsockopt() to disable Nagle. [based on a bug report by
89 Johannes Plassmann <johannes.plassmann@pdb.siemens.de>]
91 Changes with mod_fastcgi 2.2.2
93 *) Added support for blocking connect()s by setting appConnTimeout to 0.
94 Non-blocking connect()s (the default) can be troublesome on some platforms.
95 Its expected that blocking connect()s will become the default (and
96 non-blocking connect()s will become optional) in the next release.
98 *) Dump a compile time error if the version of Apache is too old.
100 *) Wrap the SIGPIPE handler manipulation code such that it is only applicable
101 to Apache releases prior to 1.3.6.
103 *) Minor tweaks for RUSSIAN_APACHE. [Sergey Gershtein <sg@mplik.ru>]
107 Always restart a failed dynamic application if it is the last instance.
108 This means there once started there will always be at least one process
109 instance of a dynamic application.
111 Send PLEASE_START to the PM when a connect() results in an ECONNREFUSED.
112 ECONNREFUSED means the listen queue is full (or there isn't one). Asking
113 the PM to start (another) application instance may help empty it faster.
115 Change two sleep() calls to select() based snoozes because alarm() is in
116 effect and sleep() and alarm() don't always play nice together.
118 Fixed a couple of error messages.
120 *) Fix -listen-queue-depth arg on FastCgiConfig (dynamic). Previously it
121 was ignored and the default was always used.
123 *) Allow the -initial-env argument to be used to pass variables from the
124 Apache process environment to the FastCGI server (by specifying a
125 variable name without the "=" or a value). [suggested by
126 Martin Lichtin <lichtin@oanda.com>]
128 *) Cleanup some debug macros.
130 *) Improved script stderr handling. [based on suggestions by
131 David Birnbaum <davidb@chelsea.net>]
133 *) Added IRIX and FreeBSD to the list of supported platforms and other minor
134 updates to the INSTALL doc.
136 *) Changed the default listen-queue-depth (FCGI_DEFAULT_LISTEN_Q) from
137 5 to 100. Its still configurable with the -listen-queue-depth option.
138 This should help eliminate the FAQ - Why do I see "Connection Refused"
141 *) Fix a bug in FastCgiExternalServer that broke support for external servers
142 on other hosts. [Dave Neuer <dneuer@futuristics.net>]
144 Changes with mod_fastcgi 2.2.1
146 *) Updates to the INSTALL doc to describe building as a DSO.
148 *) If the FastCgiIpcDir directive was in use and "httpd -t" was issued while
149 Apache was running ("apachectl restart" does this), the contents of the
150 dynamic directory (dynamic sockets and the mbox) would be blown away
153 *) Add an extern declaration for ap_sys_siglist (its exposed but not in any
154 Apache header file) to prevent a compile error on systems without
157 *) During auth requests, the subprocess_env table was left holding variables
158 sent to the auth FastCGI server (including REMOTE_PASSWD!) which means they
159 were passed to other processing phases (such as CGI/FastCGI). The
160 subprocess_env is now restored to its pre-auth condition.
162 *) Added a FastCGI Authorizer Role compatibility mode which implements the
163 specification to the tee. Use -compat arg with any of the Auth server
164 directives. This is intended for new and existing FastCGI authorizer
165 applications that require compatibility with other server implementations.
167 *) Fix logging from the process manager. Previously, all calls to the
168 logging routines used NULL for the server_rec. This works fine in a dev
169 environment, but results in most of these messages being tossed by
170 log_error_core() because DEFAULT_LOGLEVEL is too low. Odd logic.
172 *) Always setup fr->header so that in the event the FastCGI server
173 doesn't get a valid FastCGI protocol header over the wire, we
174 can still print our error message to the log from do_work().
176 *) Fix a compile error in open_connection_to_fs() on systems with TCP_NODELAY
177 defined (Don Locrasto [locrasto@iceminer.com]).
179 Changes with mod_fastcgi 2.2.0
181 *) Serious rewrite of mod_fastcgi.html. An example conf will have
182 to wait until another time - I'm burned out on docs updates.
184 *) Add to the mod_fastcgi to the server version string.
186 *) SCO doesn't like the const in the arg to inet_addr() and
189 *) Use Apache's NET_SIZE_T to get the correct arg type for getsockopt().
191 *) Deleted redundant header files from fcgi.h.
193 *) Full path names are no longer required for directives which take
194 file names as arguments. Like other Apache directives you can
195 specify paths from server_root (don't start with a "/").
197 *) The fcgiKillMgr is gone. This intermediate process sat between
198 the Apache parent and the FastCGI Process Manager under Apache
199 1.2. The process manager tries to name itself fcgi_pm (it used
202 *) Authentication, Authorization, and Access phases are now
203 supported. I've bent the FastCgi spec a bit: as many of the standard
204 environment variables (that were easy) are sent (the spec says don't
205 send some by name), all headers (except Status) sent by the auth
206 FastCgi server are passed to subprocesses (CGI/FastCGI invocations)
207 as environment variables rather than just those prefixed by
208 "Variable-". Custom responses for auth failures aren't supported
209 (yet.. we'll see what the demand for this is like). In addition to
210 the FastCGI protocol defined environment variable "FCGI_ROLE" being
211 set to "AUTHORIZER" an environment variable "FCGI_APACHE_ROLE" is
212 set indicating which of the three phases is being processed. See
213 the docs for info on the new per-directory directives:
215 FastCgiAuthenticator, FastCgiAuthenticatorAuthoritative,
216 FastCgiAuthorizer, FastCgiAuthorizerAuthoritative,
217 FastCgiAccessChecker, FastCgiAccessCheckerAuthoritative
219 *) The code has been broken into logical chunks making figuring it
220 out much easier. All of the #defines a user may want to change are
221 now in mod_fastcgi.h.
223 *) All the logging has been converted to the "supported" Apache
224 logging routines, ap_log_error() and ap_log_rerror(). Log entries
225 for request specific errors should now be directed to the correct
226 server errorlog (if your running more than one). Some attempt has
227 been made at setting appropriate log levels (comments of course are
228 welcome) and printing errno info when its of value. All of log
229 messages now have "FastCGI" in them (nice for grep).
231 *) All of the calls to Apache routines have been updated to the ap_
232 convention, eliminating the need for the compat header.
234 *) By default we no longer flush every piece of data we get from the
235 FastCGI server to the client. This allows the FastCGI server to
236 release without having to wait for the client flush to complete.
237 The old behaviour can revived with the -flush argument to AppClass.
239 *) I renamed some of the directives for consistency (the old names
242 AppClass -> FastCgiServer,
243 FCGIConfig -> FastCgiConfig,
244 ExternalAppClass -> FastCgiExternalServer
246 *) Directive arguments are no longer case sensitive.
248 *) The module now uses hard_timeout() rather than soft_timeout. This
249 means that the module will longjump out of the request if the
250 drop-dead timeout expires (set with Apache's Timeout directive, the
251 default is 5min) or if the client closes the connection (SIGPIPE).
252 This is more typical for a Apache modules and I'm not convinced we
253 properly handle all of the error cases well enough to use
254 soft_timeout (I can think of one place we don't anyway). This means
255 your FastCGI application can see SIGPIPE.
257 *) SIGPIPE is now ignored by default in FastCGI servers spawned by
258 Apache. Without this the default behaviour is exit().
260 *) It should (I think) run now under DSO now.
262 *) The bug which prevented sending of binary data in 2.1b1 is fixed.
264 *) The broken -initial-env bug ins 2.1b1 is fixed.
266 *) Maybe some other stuff.
270 *** Well there's a big gap here. Maybe I'll go back and extract the CVS
271 commit notes and stick 'em in here.. when I get some free time. ;)
276 *** Originally from docs/README..
278 Apache[X1.03.01]/mod_fastcgi[02.00.05] 19970909 unsupported
279 From: Stanley Gambarin <stanleyg@cs.bu.edu>
281 *) Yet more changes to the source distribution. Separated out
282 the Tcl dynamic string and buffer libraries into separate files.
283 Did the same for the OS library, however, it is not really an
284 abstraction... more like a bunch of wrappers. Added to DEVNOTES
285 a note about the problem of Apache not allowing including header
288 Apache[X1.03.01]/mod_fastcgi[02.00.04] 19970908 unsupported
289 From: Stanley Gambarin <stanleyg@cs.bu.edu>
291 *) More changes to the installation scripts. Theoretically we
292 should be able to to use the same Makefile/installation script
293 for both Apache 1.2.x and 1.3.x. Need some sort of the abstraction
294 layer to use the same source code for both 1.2.x and 1.3.x sources.
296 Apache[X1.03.01]/mod_fastcgi[02.00.03] 19970905 unsupported
297 From: Stanley Gambarin <stanleyg@cs.bu.edu>
299 *) Created Makefile to compile the module into the archive library,
300 just like the proxy module. Also initial draft of the installation
301 script, which should work for both 1.2 and 1.3 sources of Apache.
303 Apache[01.02.04]/mod_fastcgi[02.00.02] 19970903 unsupported
304 From: Stanley Gambarin <stanleyg@cs.bu.edu>
306 *) More source reorg. Separated out some header files. Need much
307 more work on this. Updated the TODO file. Added fcgivers.h and
308 CHANGES file to track the history.
310 Apache[01.02.04]/mod_fastcgi[02.00.01] 19970902 unsupported
311 From: Stanley Gambarin <stanleyg@cs.bu.edu>
313 *) Source reorganization. This is done to provide a basis for later
314 functionality implementation. See TODO file for more information
315 of what possible future enhancements are possible/desired.
317 Apache[01.02.04]/mod_fastcgi[02.00.00] 19970902 unsupported
318 From: David MacKenzie <djm@va.pubnix.com>
320 *) Create the "dynamic" dir and "mbox" as the user specified in the
321 User and Group directives instead of as root. They are created with
322 restrictive permissions, and then the module checks to see whether the
323 user specified in the User and Group directives has read, write, and
324 execute permissions on them. Those permissions had been explicitly
325 denied by creating them as "root".
327 *) Some documentation was garbled or incomplete.
329 *) When a FastCGI application can't be execl'd, the code used a
330 value of errno that may have been stomped on by intervening system
331 calls. That happened on BSD/OS 2.1, where the error_log was
332 reporting errno=25 (NOTTY) instead of the correct errno=13 (EACCES).
333 An intervening stdio call was setting errno as a side-effect, which
334 it has the right to do.
336 *) Erroneous fprintf arguments, a missing return value, an unused
337 function, missing declarations, and other problems detected by gcc
340 *) Duplicated code to create the lock file name merged into a single
341 function, closing several memory leaks.
343 *) FCGIConfig -minProcesses didn't allow a value of 0, so the last
344 FCGI app in a given dynamic class would live forever, even if it
345 hadn't been requested for weeks.
347 *) The test for whether to keep looking for dynamic app victims to
348 kill was backwards, so no dynamic apps were ever selected as
351 *) The killInterval and updateInterval were being ignored; the
352 sigsuspend() forced recalculations only when a child died,
353 instead of at the intervals specified.
355 *) When a dynamic app couldn't be started, the program SEGV'd when
356 trying to free the ipcAddrPtr twice.
358 *) Remove the dead lock file and socket when cleaning up after a
359 server whose last child has been killed, so the FastCGIHandler
360 isn't fooled into thinking there is still a process serving
363 *) Make file locking robust in the presence of signals.
365 *) Rename some badly named variables.
367 *) Fix typos in many comments.
369 *) Fix some memory and file descriptor leaks.
371 *) Make the blocking kill of a server closer to working.
373 *) More fixes to calculations in dynamic application management.
376 *** Originally from docs/README.OMI..
378 What's New: Version 2.0b1, 16 Apr 1997
380 *) Implemented a mechanism by which FastCGI applications are started by
381 the web server on the first request and continue running. Also, a
382 heuristics have been implemented to allow for dynamic killing of the
383 running FastCGI apps. The configuration options are supplied via
384 the new FCGIConfig directive, that is described in the
387 *) When performing internal redirect, since the original request's body
388 has already been read, do not allow the redirected request to think
389 that is has one [body].
391 *) More conditional compilation for OS/2.
393 *) Fixed occassional "Assertion failed: len > 0, file mod_fastcgi.c"
395 *) Fixed "failed assertion `count >= 0 && count <= bufPtr->length'"
397 *) Added bflush() call after bwrite() in DrainReqOutBuf function to
398 force some output to be written in the period of inactivity. This
399 circumvents the Apache's buffering during the server-push.
401 *) Added a header-parser function placeholder for compatibility with an
404 *) Additional checks have been placed with regard to permissions of the
405 FastCGI processes. These include the existence of the file,
406 ExecCGI and IncludesNOEXEC checks, disallowing of nph- scripts, etc.
408 *) Implemented a restart cleanup, so that no parentlesses FastCGI
409 applications are left after Apache is restarted and/or terminated.
410 This was accomplished via two independent mechanisms, where the
411 first one prevents a start of the process manager on the first
412 reading of the configuration files. The second mechanism deals away
413 from Apache's implementation of fork() (via spawn_child()) and
414 implements its own forking, thus removing any dependencies on the
415 Apache to cleanup processes during termination phase.
417 *) Removed the definition of the Sigfunc, as Apache 2.0b10 defines it
418 itself in the file conf.h
420 What's New: Version 1.4.3, 15 Jan 1997
422 *) Fixed compilation warnings for various platforms, as well as
423 conditional compilation for OS/2.
425 What's New: Version 1.4.2, 12 Dec 1996
427 *) mod_fastcgi.c is ported to Apache 1.2b. Any further development
428 will proceed under this version of Apache web server.
430 *) As the result of porting, the "include virtual" construct of SSI
431 will now work correctly using either <Location> or <Directory>
434 What's New: Version 1.4.1, 4 Dec 1996
436 *) Checks have been removed from the ScanCGIHeaders that provided
437 for the presence of both Status and Location headers as being
438 an error. Contradictory to CGI/1.1 Internet Draft, both of
439 these headers are used by the current CGI applications.
441 What's New: Version 1.4, 22 Nov 1996
443 *) Added the -port option to AppClass, allowing TCP/IP communication.
444 Added the -socket option to AppClass, allowing Unix domain
445 communication via a configurable pathname.
447 *) Added the ExternalAppClass directive, allowing TCP/IP
448 communication with remote FastCGI applications.
450 *) The handler had its own code for generating HTTP response headers;
451 now it uses Apache's. This reduces the size of the module.
452 More importantly, it fixes the bug in which "include virtual"
453 sees the HTTP response headers.
455 *) The response header parser performed very little checking.
456 Now the parser enforces the guidelines in the CGI/1.1
457 Internet-Draft: Status and Location are mutually exclusive,
458 Location can only be a response to GET or HEAD,
459 CGI response headers can't be repeated, etc. (The CGI response
460 headers are the ones the handler interprets: Status, Location,
463 *) The response header parser used to miss CGI headers
464 with no whitespace after the colon, e.g. "Status:200 OK".
466 *) The response header parser sometimes interpreted the first line
467 of content as an RFC822 continuation line.
469 *) The handler implemented a nonstandard version of Location
470 which never used internal redirects. The handler
471 now attempts to implement Location as specified in the
472 CGI/1.1 Internet-Draft. The module documentation explains
475 *) Error log entries from the response header parser were
476 pretty uninformative; they are better now. If a header is
477 malformed, the log entry includes it. If the headers are
478 unterminated rather than malformed, the log entry says that,
479 and says how many bytes were received from the app.
481 *) When the application manager forked a new process, and that
482 process ran into trouble before executing the first instruction
483 of application code, the process used to exit with status =
484 errno. This made certain configuration problems (e.g. incorrect
485 file permissions when server parent is root) quite difficult
486 to pin down. Now the failing child process opens up the error log
487 and writes an informative entry before exiting.
489 *) The module now correctly handles a slash at the end of the
490 DocumentRoot directive. This was a one-line fix.
492 What's New: Version 1.3.3, 17 Oct 1996
494 *) The module now registers its request handler under the name
495 fastcgi-script in addition to the name application/x-httpd-fcgi.
496 This was a one-line addition to the module, but had quite
497 a large impact on the documentation and sample configuration.
499 What's New: Version 1.3.2, 27 Sept 1996
501 *) On some systems (SunOS, Linux?), fopen for append has a bug
502 that strikes when two processes append to the same file.
503 This bug causes the process manager to corrupt the error log.
504 Work around the bug by calling open, then fdopen. A patch
505 to Apache 1.1.1 is also required, as described in mod_fastcgi.html.
506 (Reported by Bob Ramstad.)
508 What's New: Version 1.3.1, 17 Sept 1996
510 *) On Linux and SunOS, the Apache default user_id (-1) is not a
511 legal value of uid_t, so cast it. (Reported by Bob Ramstad,
512 Scott Langley, others.)
514 *) On some systems (Linux, some Solaris?, Irix?), connect requires
515 write access to a Unix Domain socket, so provide it. (Reported
516 by Scott Langley, freeform@wired.com, others.)
518 *) If you hit a .fcg file, but there's no AppClass defined, the
519 error should be NOT_FOUND instead of SERVER_ERROR. (Reported
522 What's New: Version 1.3, 4 Sept 1996
524 *) Module sometimes busy-waited in FastCgiDoWork. Fixed.
526 *) Module violated the Apache buff abstraction.
527 Fixed. Might now work with SSL (not tested).
528 As part of the fix, eliminated support for
529 Apache 1.0x versions.
531 *) Module failed to chmod the Unix domain listening sockets
532 it created, so protections were set according to the current
535 *) Module forked too many process manager processes: One
536 per AppClass. When the Apache parent ran as root, these
537 process manager processes ran as root. New process
538 manager is a single process (only started if AppClass is
539 used) and runs with same privileges as other children.
540 The new process manager doesn't do any polling, so
541 there's less system overhead than before.
543 *) AppClass insisted on getting at least two arguments. Fixed.
545 *) Module used setjmp/longjmp, causing compiler warnings
546 on some platforms (e.g. Linux.) The module no longer
549 *) Module created listening sockets in /tmp, where they were
550 sometimes wiped out by cleanup scripts. Added FastCgiIpcDir
551 directive to give control over the location of listening
554 *) Module wrote error log entries without a timestamp. Fixed.
556 *) AppClass directive wrote error messages to stderr in addition
557 to returning a char * message to the Apache core. Fixed.
559 What's New: Version 1.2, 3 June 1996
561 *) Ported from Apache-1.0-based code to Apache-1.1b2 internals by
564 Add version string: APACHEVERSION macro to mod_fastcgi.c
566 *) chown FastCGI socket to user_id and setuid to user_id for app
569 *) Modify GetFromStream() by having it call an OS dependent
570 function GetStreamSize(FILE *) which uses FILE internal data member.
571 Linux users might need to modify GetStreamSize().
573 What's New: Version 1.1, 10 May 1996
575 *) If you specify a non-existent executable in the AppClass directive,
576 or if the file exists and it does not have execute permission you
577 get a constant stream of error messages telling you that "program
578 terminated due to a signal".
580 *) The mod_fastcgi module should use the standard Apache error logging
581 facility instead of writing to stderr.
583 Version 1.0, 30 April 1996