Bill Stoddard [Sun, 22 Oct 2000 15:20:08 +0000 (15:20 +0000)]
Fix problem where a 304 response was not generating an EOS bucket.
ap_finalize_request_protocol()'s sole purpose in Apache 2.0 is to
send the EOS bucket. Probably need to reimplement ap_send_error_response()
to be consistent with ap_finalize...'s new mission.
Ryan Bloom [Sat, 21 Oct 2000 14:20:14 +0000 (14:20 +0000)]
Hack headers to work semi-properly. This is a hack that will need to be
fixed, but it is good enough for now. The idea is that headers shouldn't
flow through the BUFF anymore. Now, we have a header filter that is
called at the end of the request-filter chain. This filter writes the
headers directly to the connection filters.
Jeff Trawick [Sat, 21 Oct 2000 14:07:41 +0000 (14:07 +0000)]
Get translation of request bodies working.
This code is pretty ugly, but I need to checkpoint (and this stuff
actually works).
The direction: Get xlate_brigade() working for output too, then
kill the analogous code in xlate_out_filter(). At that point, the
same translation code works for input and output.
Tony Finch [Thu, 19 Oct 2000 19:24:51 +0000 (19:24 +0000)]
Belatedly update the documentation to include the NameVirtualHost *
stuff. This isn't quite a sync with the 1.3 docco because we don't
want to mention 1.3 in the 2.0 docs.
Tony Finch [Thu, 19 Oct 2000 19:15:14 +0000 (19:15 +0000)]
Belatedly update the documentation to include the NameVirtualHost *
stuff. This isn't quite a sync with the 1.3 docco because we don't
want to mention 1.3 in the 2.0 docs.
Greg Stein [Thu, 19 Oct 2000 10:43:03 +0000 (10:43 +0000)]
somebody was a Very Bad Boy when they inserted casts into this function.
casting away the const was absolutely wrong... the warnings were saying the
return value type needed to be fixed. did that and torched the casts.
Tony Finch [Thu, 19 Oct 2000 00:40:57 +0000 (00:40 +0000)]
Update to the media types as of 2000-10-19. I also fixed the case of
application/batch-SMTP to match the official name.
Obtained from: ftp://ftp.isi.edu/in-notes/iana/assignments/media-types
PR: 6613
Ryan Bloom [Wed, 18 Oct 2000 22:14:18 +0000 (22:14 +0000)]
Sub-requests have to have a value for their request_config structure,
or we'll seg-fault. This should set the sub-request's request_config
structure to the same exact same structure as the main request is using.
Ryan Bloom [Wed, 18 Oct 2000 19:32:30 +0000 (19:32 +0000)]
Make the core access the socket directly instead of going through the
BUFF. This doesn't actually change any behavior, it just makes the core
access socket directly.
Ryan Bloom [Wed, 18 Oct 2000 19:12:16 +0000 (19:12 +0000)]
This begins to remove BUFF from the server. The idea is to go very slowly
with this. To begin with, we store both the socket and the BUFF in the
conn_rec. Functions are free to use which ever they want, in the end all
of the data goes to the same place. This modifies all of the MPMs except
Windows. All of the Unix MPMs are working, but the others need to be
tested.
Ryan Bloom [Wed, 18 Oct 2000 18:50:11 +0000 (18:50 +0000)]
We have to create the core_request_config structure in the core on internal
redirects. If this isn't done, we'll seg-fault during the AP_DEBUG_ASSERT
that makes sure getline didn't leave any droppings.
Jeff Trawick [Wed, 18 Oct 2000 17:42:44 +0000 (17:42 +0000)]
ap_get_client_block() now uses a brigade in core_request_config
instead of a brigade in core_dir_conf. For now, getline() and
ap_get_client_block() share a brigade.
Ryan Bloom [Wed, 18 Oct 2000 15:58:29 +0000 (15:58 +0000)]
The final line of the config file was not being read if there was
no \n at the end of it. This was caused by apr_fgets returning
APR_EOF even though we had read valid data. This is solved by
making cfg_getline check the buff that was returned from apr_fgets.
If apr_fgets return APR_EOF, but there was data in the buf, then we
return the buf, otherwise we return NULL.
Greg Ames [Wed, 18 Oct 2000 15:21:32 +0000 (15:21 +0000)]
Add a core_request_config so we have a safe semi-hidden place to save core data
that lives across function calls during a single request. Change getline() to
take a request_rec parm (rather than a conn_rec) so we can access the
core_request_config.
This is in preparation for adding look-ahead functionality to getline(), so it
can support header line folding once again. I'm committing these changes first
so the core_request_config can be used elsewhere (i.e. ap_get_client_block).
Ryan Bloom [Wed, 18 Oct 2000 06:09:10 +0000 (06:09 +0000)]
Fix piped logs in 2.0. This basically:
1) cleans up an annoying type that was getting in my way while I was
trying to fix things.
2) Makes some of the allocations pcalloc instead of palloc
3) The arg array passed to create_process is a const *char *, not
const *char [].
PR: 6642
Tony Finch [Wed, 18 Oct 2000 04:50:25 +0000 (04:50 +0000)]
Restore functionality broken by the mod_rewrite security fix:
rewrite map lookup keys and default values are now expanded
so that the lookup can depend on the requested URI etc.
Because the new code is recursive you can now write configurations
like the following (with nested map lookups) which wasn't directly
possible before -- you had to use a RewriteCond and a %N backref
instead, as in http://www.apache.org/docs/vhosts/mass.html#xtra-conf
Tony Finch [Wed, 18 Oct 2000 04:48:34 +0000 (04:48 +0000)]
Tighten up the syntax checking of Host: headers to fix a
security bug in some mass virtual hosting configurations
that can allow a remote attacker to retrieve some files
on the system that should be inaccessible. The problem
occured with requests including the line "Host: ..." --
the last dot is stripped and the remaining ".." then
reveals a parent directory.
Reported by: Peter Christoffersen <pch@mindpass.com>
Message-ID: <8quts6$2el$1@news.inet.tele.dk>
Newsgroups: comp.infosystems.www.servers.unix
Ryan Bloom [Tue, 17 Oct 2000 21:53:41 +0000 (21:53 +0000)]
Add pool buckets to the bucket list. These buckets are not currently
used although I have begun to test them on my machine. The idea behind
these buckets is that data allocated out of a pool and put into a bucket
has to survive the death of the pool until the bucket is destroyed. To
accomplish this, we register a cleanup when we create the bucket. If the
pool is cleared, the cleanup converts the bucket to a heap bucket and
everything is good. If the pool isn't cleared, then we kill the cleanup
when the bucket is destroyed.
With this bucket type, all of the core buckets that were listed have been
implemented.
Jeff Trawick [Tue, 17 Oct 2000 20:43:35 +0000 (20:43 +0000)]
ap_get_client_block() allocates its brigade from r->pool (as it was
before today) so that we don't have a growth of these across many
pipelined requests.
http_filter() uses one brigade per connection which it reads into.
As it needs to deliver buckets to the caller, they are removed from
its brigade into the caller's brigade.
Submitted by: Ryan Bloom (but anything broken is my fault)
Jeff Trawick [Tue, 17 Oct 2000 15:28:57 +0000 (15:28 +0000)]
http_filter():
when delivering body bytes, only deliver one block of data (however
much is returned by bucket read) instead of delivering the entire
body at once; this gets painful with a large body
make a note of an issue with the blocking state of the socket;
currently, the socket is non-blocking, but when reading body bytes
we should feel free to wait for body bytes until a timeout occurs;
Jeff Trawick [Tue, 17 Oct 2000 01:35:12 +0000 (01:35 +0000)]
Rename output filters field ("filters") in core_dir_config to
"output_filters" for consistency with the name of the input
filters field ("input_filters").
Ryan Bloom [Tue, 17 Oct 2000 00:24:36 +0000 (00:24 +0000)]
Add a flush bucket type. This bucket advises filters to flush any data
they are currently storing. There is no way we can force them to flush,
but we can advise. This also adds the code to ap_rflush to use flush
buckets, although it isn't enabled yet. I will enable it once we remove
buff from the code. I also removed all calls to ap_rflush that are either
immediately before or immediately after a call to ap_finalize_protocol.
ap_finalize_protocol sends an EOS bucket, which also advises filters to
flush their data, so having both calls right next to each other is
redundant.
Ryan Bloom [Mon, 16 Oct 2000 23:15:55 +0000 (23:15 +0000)]
Add a sub-request filter. This filter just strips the EOS from the
brigade generated by the sub-request. If this is not done, then the
main-request's core_output_filter will get very confused, as will any other
filter in the main-request filter-stack that looks for EOS.
Jeff Trawick [Mon, 16 Oct 2000 20:08:14 +0000 (20:08 +0000)]
dechunk filter:
Get rid of an assertion which assumed that HTTP_IN is the filter
below us. Some other filter may play the same role.
ap_setup_client_block():
Get rid of a commented-out hack which was used to allow chunked
transport encoding of a request body received by mod_cgi[d].
ap_get_client_block():
Get rid of special handling for zero-length buckets... The main
loop handles that fine.
Submitted by: Ryan Bloom
Jeff Trawick [Mon, 16 Oct 2000 19:11:53 +0000 (19:11 +0000)]
input filtering changes:
get dechunking working
verify that infrastructure for input filters works
(use existing AddInputFilter directive)
Unlike with my previous patch, ap_get_client_block() saves state between
calls in the core's per-request dir config.
Unlike with my previous patch, HTTP_IN keeps a count of remaining bytes
in the conn_rec. Code that needs to prod it to deliver a certain amount
of request body plays with conn_rec->remain directly.
Renamed all MODULE_EXPORT symbols to AP_MODULE_DECLARE and all symbols
for CORE_EXPORT to AP_CORE_DECLARE (namespace protecting the wrapper)
and retitled API_EXPORT as AP_DECLARE and APR_EXPORT as APR_DECLARE.
All _VAR_ flavors changes to _DATA to be absolutely clear.
Thank you Greg, for the most obvious suggestion.
Eliminating a bunch-o-junk from the Win32 os.h header that now lives
in apr or isn't used. Please review closely (every symbol removed was
grep't for, and no references were found.)
Ryan Bloom [Sun, 15 Oct 2000 18:15:13 +0000 (18:15 +0000)]
Stop using the index into the array for the bucket type. Now we just use
a pointer to a static structure. The ap_foo_type functions have also been
replaced with simple macro calls. I am going to replace the
ap_bucket_(read|split|setaside|destroy) functions with macros soon.
Reviewed by: Will Rowe
David Reid [Sun, 15 Oct 2000 11:46:22 +0000 (11:46 +0000)]
Some fixes and a general tidy up of ab. With these changes ab can now
support concurrency levels > 1 against the loopback on BONE and will
actually use all the connections it starts.
Change to buckets for isapi output. Still requires emulation of IO
Completion/Async behavior, but a non-async REQ_HSE_TRANSMIT_FILE is
now implemented.
Ryan Bloom [Sat, 14 Oct 2000 06:53:02 +0000 (06:53 +0000)]
Cleanup ap_get_client_block a bit. All of the code that I am removing
doesn't actually do anything useful, it was meant to remove all zero
length buckets before trying to read from the brigade. However, it is
perfectly fine to not do this, and to just read those buckets along with
the rest of them.
Ryan Bloom [Sat, 14 Oct 2000 06:21:38 +0000 (06:21 +0000)]
My last patch had most of the code to insert the eos bucket, but it was
missing some major pieces that I only found after the commit. This patch
fixes the problem, and makes sure that request bodies are always ended
with an EOS bucket.
Ryan Bloom [Sat, 14 Oct 2000 04:37:22 +0000 (04:37 +0000)]
Get non-chunked input body filtering working with an EOS bucket. The
basic design has ap_setup_client_block setting a field in the conn_rec
which tells http_filter how much data is in the body (with chunking this
will represent how much data is in the chunk). The ap_get_client_block
then calls down the stack with the maximum amount of data that it can
receive back. When http_filter reads all of the data, it adds an eos
bucket to the end of the brigade. ap_get_client_block continues to read
data until it gets the eos bucket. This allows filters to increase the
size of the body data.
Jeff Trawick [Fri, 13 Oct 2000 18:39:18 +0000 (18:39 +0000)]
Introduce ap_debug_assert() macro, like ap_assert() but only active if
AP_DEBUG is defined.
ap_get_client_block():
. avoid some cases where we leak a temporary bucket brigade
. clean up/fix the logic to copy a brigade into the caller's buffer;
the wrong length was used in some cases
. add an AP_DEBUG-only assertion for some assumptions made regarding
the brigade returned by the filters
Submitted by: partly by Greg Stein, but of course anything bad is mine
Jeff Trawick [Fri, 13 Oct 2000 17:54:24 +0000 (17:54 +0000)]
getline() fixes...
If ap_get_brigade() returns APR_SUCCESS but an empty brigade, bail out.
Previously, we kept going and sometimes segfaulted while operating on
what we thought was the first bucket.
Ryan Bloom [Fri, 13 Oct 2000 05:35:57 +0000 (05:35 +0000)]
Remove all function pointers from the ap_bucket type. These function
pointers are replaced with a global table that allows modules to register
their bucket types. Those bucket types are then allowed to be used in
the server processing. This also required removing all direct calls to
those functions. The ap_bucket type has an index into an array, so in
each ap_bucket_* function, we use that index to find the correct set of
functions.
Jeff Trawick [Fri, 13 Oct 2000 03:38:35 +0000 (03:38 +0000)]
mod_cgid: In the handler, shut down the Unix socket (only for write)
once we finish writing the request body to the cgi child process;
otherwise, the client doesn't hit EOF on stdin. Small request bodies
worked without this change (for reasons I don't understand), but large
ones didn't.
Jeff Trawick [Fri, 13 Oct 2000 03:12:54 +0000 (03:12 +0000)]
Fix a couple of problems related to large request bodies:
http_filter() - this was referencing e->length even though e->length
is sometimes -1 (if we have to read the socket bucket to get the
next bit of the request body); now it does e->read() to force a
socket need if required
ap_get_client_block() - this was getting a brigade back with the
desired amount of data (yeah, right) but only copying out the first
bucket