needed to build on Solaris using Oracle Solaris Studio (thx petrs)
x-ref:
"vector_realloc missing when linking test_configfile"
https://redmine.lighttpd.net/issues/2752
lots of fixes and improvements
limitations: see comments at top of mod_deflate.c
missing functionality: encode streaming response
(module currently requires response be collected before being sent)
potential functionality: addition of compressed file cache would
allow mod_deflate to fully supplant mod_compress in lighttpd 1.4.x
x-ref:
"Adding mod_deflate to 1.4.xx"
https://redmine.lighttpd.net/issues/1824
"mod_deflate backport compile error if ENABLE_MMAP not defined"
https://redmine.lighttpd.net/issues/2753
github: closes #67
allow authorizer and responder to be configured for same path or ext
x-ref:
"mod_fastcgi authorizers cannot protect fastcgi responders"
https://redmine.lighttpd.net/issues/321
(for persistence across multiple re-entries into routines upon
receiving fdevent)
(setting module debug flag in global scope is still recommended
since there are places where p->conf.debug is checked since
handler_ctx might not be available at all points)
import Variable-* from FastCGI authorizer response into con->environment
restart request after FastCGI authorizer if no fastcgi.server docroot
(thx Christoph Kreutzer for initial patch attempt)
x-ref:
"mod_fastcgi authorizers cannot protect fastcgi responders"
http://redmine.lighttpd.net/issues/321
x-ref:
"FastCGI Authorizer support for Variable-name variable passing"
http://redmine.lighttpd.net/issues/322
github: closes #70
Note: http_auth_backend_t digest interface returns result as a
binary MD5 (16-bytes) so that caller consistently converts to
lowercase before using it in further digest calculation.
(Alternatively, the http_auth_backend_t digest interface could have
taken a 33-char buffer and returned an explicitly lowercased hex str)
create new, extensible interface for (additional) auth backends
attempt to handle HANDLER_WAIT_FOR_EVENT returned by auth backends
to allow for async auth backends (e.g. to mysql database)
separate auth backends from mod_auth and http_auth
mod_authn_file.c htdigest, htpasswd, plain auth backends
mod_authn_ldap.c ldap auth backend
add http_auth.c to common_sources for auth backend registration
(mod_authn_file could be three separate modules, but no need for now)
copied javascript from mod_status and from lighttpd2 mod_dirlist
modified and specialized for stable dirlist sorting by name
Partial implementation of Apache autoindex request query arguments
(https://httpd.apache.org/docs/current/en/mod/mod_autoindex.html)
If query string is supplied, allow specifying initial column to sort
?C=N name (default)
?C=M last-modified, then by name
?C=S size, then by name
?C=T type, then by name
?C=D type, then by name
and O=[AD] can be added for descending or ascending order, e.g.
?C=N&O=D descending (default)
?C=N&O=A ascending
(While functional, no effort was made on js performance.
Patches welcome)
New directive dir-listing.external-js for user to replace sorting js
Note: dir-listing.external-js or default js sorting is enabled only
if dir-listing.auto-layout = "enable" (which is the default)
x-ref:
"client-selectable directory list sorting"
https://redmine.lighttpd.net/issues/613
"dir-listing.external-js"
https://redmine.lighttpd.net/issues/2315
Due to the POLLHUP behavior triggered on *BSD/Darwin, the shutdown()
had previously been limited to local connections. If interested in
squeezing the last bits of performance out of a machine, an admin
should configure local connections to be AF_UNIX instead of AF_INET
or AF_INET6 to localhost. The reason the shutdown() was originally
added in mod_proxy and mod_scgi was to aggressively reduce the number
of potential sockets in TIME_WAIT held by lighttpd.
(See commit:923688d2 "drain backend socket/pipe bufs upon FDEVENT_HUP",
done for reliability given the aforementioned *BSD/Darwin behavior.)
When using AF_UNIX, the TIME_WAIT issue does not exist, ergo, the
recommendation is to use AF_UNIX for local sockets, when available.
Using AF_UNIX sockets is a better solution to eliminate TIME_WAIT
than is TCP shutdown() half-close which, as we have seen, might not
be handled well by frameworks which are more complex than basic read
request, send response, and close.
x-ref:
"1.4.40/41 mod_proxy, mod_scgi may trigger POLLHUP on *BSD,Darwin"
https://redmine.lighttpd.net/issues/2743
do not enter handler a second time in connection_state_machine()
after read body completes if dynamic handler is still waiting for event
and CON_STATE_READ_POST transitioned to CON_STATE_HANDLE_REQUEST
mod_fastcgi, mod_scgi, and mod_proxy must enforce wait for POLLWR
after EINPROGRESS or else getsockopt(fd, SOL_SOCKET, SO_ERROR, ...)
may succeed even though socket connection is not yet established,
and subsequent writev() will fail ENOTCONN.
(thx pkubaj)
x-ref:
"1.4.40/41 writev failed: Socket is not connected (fastcgi,scgi,proxy)"
https://redmine.lighttpd.net/issues/2744
Check if client half-closed TCP connection if POLLHUP is received.
This more robustly handles if client called shutdown(fd, SHUT_WR).
This patch reverts commit:ab05eb7c which should now be handled properly.
(Time will tell.)
x-ref:
"1.4.40/41 mod_proxy, mod_scgi may trigger POLLHUP on *BSD,Darwin"
https://redmine.lighttpd.net/issues/2743
shutdown(fd, SHUT_WR) after sending request to proxy or SCGI
only if remote is local and platform is not *BSD or Darwin.
The reason this fix is special-casing *BSD and Darwin is that the Single
Unix Specification and POSIX.1-2013 clearly specify that POLLHUP event
should be returned by poll only when the stream is no longer writable.
A half-closed socket that is still writable clearly does not match that
condition, yet that is what I am seeing on Darwin (El Capitan), and
presumably what others are seeing on *BSD, from which Apple originally
inherited the Darwin TCP stack.
Single Unix Specification (SUSv2) from 1997
(yes, that is nearly 20 years ago):
http://pubs.opengroup.org/onlinepubs/007908799/xsh/poll.html
POLLHUP
The device has been disconnected. This event and POLLOUT are
mutually exclusive; a stream can never be writable if a hangup has
occurred. However, this event and POLLIN, POLLRDNORM, POLLRDBAND or
POLLPRI are not mutually exclusive. This flag is only valid in the
revents bitmask; it is ignored in the events member.
Updated version of The Open Group Base Specifications Issue 7
(published in 2013):
http://pubs.opengroup.org/onlinepubs/9699919799/
POLLHUP
A device has been disconnected, or a pipe or FIFO has been closed
by the last process that had it open for writing. Once set, the
hangup state of a FIFO shall persist until some process opens the
FIFO for writing or until all read-only file descriptors for the
FIFO are closed. This event and POLLOUT are mutually-exclusive;
a stream can never be writable if a hangup has occurred. However,
this event and POLLIN, POLLRDNORM, POLLRDBAND, or POLLPRI are not
mutually-exclusive. This flag is only valid in the revents bitmask;
it shall be ignored in the events member.
x-ref:
"1.4.40/41 mod_proxy, mod_scgi may trigger POLLHUP on *BSD,Darwin"
https://redmine.lighttpd.net/issues/2743
disable stat_cache if server.follow-symlink = "disable"
if server.stat-cache-engine = "simple". Caching is still enabled
for server.stat-cache-engine = "fam" since the FAM notification is
almost immediate, however there is still a small race condition.
NOTE: server.follow-symlink = "disable" implementation still has
time-of-check versus time-of-use (ToC-ToU) race conditions and
its use is *not recommended* except to discourage symlinking.
It *does not* prevent symlinking by a determined attacker with
the ability to create files on the server.
server.stat-cache-engine = "disable" can also be used to discourage
symlinking, and also does not eliminate ToC-ToU race conditions.
While more modern systems might use openat() and other *at() routines
to eliminate the ToC-ToU race conditions, this is not currently
implemented in lighttpd. Besides, for systems needing such
protections against actors able to modify local files, it would be
better to set up multiple lighttpd servers running in separate user
contexts with filesystem permissions preventing access, rather than
giving a single lighttpd server running under a single lighttpd user
access to files across security boundaries, and trying to prevent
access by lighttpd user if a file is a symlink.
Note that there are performance implications to setting either of
server.follow-symlink = "disable"
server.stat-cache-engine = "disable"
since stat cache normally reduces filesystem overhead for
frequently-accessed files.
x-ref:
"security: stat cache *very large* race condition if caching when
follow_symlink disabled"
https://redmine.lighttpd.net/issues/2724
server.username can not be root or 0.
server.groupname can not be root or 0.
If server.username is set, previous behavior might retain gid 0
if server.groupname was not set.
New behavior calls setgid() on server.username primary gid, and
then initgroups on server.username if server.username is set but
server.groupname is not set.
x-ref:
"server.groupname not required with server.username"
https://redmine.lighttpd.net/issues/2725
If server.upload-dirs is not configured, then attempt to use TMPDIR
from the environment, if set, or else use /var/tmp which is not often
a tmpfs, unlike /tmp. Warn at startup if tempdirs are not present.
If server.upload-dirs is not configured, then attempt to use TMPDIR
from the environment, if set, or else use /tmp. Warn at startup if
tempdirs are not present.
try AF_INET after AF_INET6 if server.use-ipv6 = "enable" and
getaddrinfo() fails EAI_ADDRFAMILY when hints.ai_family is AF_INET6.
(Prefer IPv6 instead of setting hinst.ai_family to AF_UNSPEC since
lighttpd only uses the first address returned)
read all available records from SSL_read(), even if larger than
MAX_READ_LIMIT, since the data is already in memory. openssl is
configured with SSL_MODE_RELEASE_BUFFERS and will release openssl
buffers once records have been read.
Without reading available data, there was a chance that the connection
would hang waiting for a read event on the fd, even though all the
data had already been read from kernel socket buffers and was in openssl
memory waiting to be read with SSL_read().
(thx glen and avij)
Christoph Kreutzer