Glenn Strauss
aba45f68d8
[core] perf: simpler buffer_string_space() (tests)
4 years ago
Glenn Strauss
473ce93c14
[core] perf: faster HTTP pipelined requests
...
separate reading header from connection_handle_read_state() into
connection_read_header() so that HTTP pipelined requests can skip
optimistic read()/recv()
4 years ago
Glenn Strauss
d1bf435573
[core] perf: simpler buffer_string_space() (fixed)
4 years ago
Glenn Strauss
70d7d0a0a1
[core] server.chunkqueue-chunk-sz = 4096
...
server.chunkqueue-chunk-sz = 4096 (default)
Can be configured any integral value greater than 0.
Value is rounded up to next 1024 if not an even multiple of 1k.
Sites with large request or response headers may benefit from
an 8k or 16k chunk size. Embedded systems might choose to minimize
memory use by using a 1k chunk size.
4 years ago
Glenn Strauss
cc1c2f0e37
[core] server.compat-module-load = "disable"
...
controls whether default and compatibility modules are loaded,
including default modules mod_indexfile, mod_staticfile, mod_dirlisting
and if mod_authn* modules are loaded when mod_auth in server.modules
and if mod_openssl is loaded when ssl. directives are seen
default: server.compat-module-load = "enable"
(preserve existing behavior)
recommended config: explicitly load desired modules in desired order
4 years ago
Glenn Strauss
1b8453886f
[mod_webdav] quiet coverity false positive
4 years ago
Glenn Strauss
f13db69012
[core] fix setting of headers previously reset ( fixes #2919 )
...
bug may result in long delays when using mod_deflate on connections
with keep-alive, as the result is sent without Content-Length or
Transfer-Encoding
(regression in lighttpd 1.4.51)
(thx GilGalaad)
x-ref:
"high latency on 1.4.51 + proxy + deflate"
https://redmine.lighttpd.net/boards/2/topics/8365
https://redmine.lighttpd.net/issues/2919
4 years ago
Glenn Strauss
41b50cfa71
[core] fix 301 -> 302 overwrite with Location ( fixes #2918 )
...
fix bug resulting in status code being overwritten from 301 to 302 if
Location response header is returned by a backend which is not mod_proxy
(regression in lighttpd 1.4.51)
(thx shochodev)
x-ref:
"lighttpd is not obeying the http status code sent by our application (302 sent by lighttpd when 301 is sent by application)"
https://redmine.lighttpd.net/issues/2918
4 years ago
Glenn Strauss
2d91475828
[mod_cgi] perf: cache getenv() results at start up
4 years ago
Glenn Strauss
5ff1ddeb66
[mod_wstunnel] perf: reuse large buffers
4 years ago
Glenn Strauss
88ee73d0a2
[multiple] perf: simplify chunkqueue_get_memory()
4 years ago
Glenn Strauss
f512bfa4af
[core] use chunk_buf_sz instead of hard-coded num
4 years ago
Glenn Strauss
9f6a46738d
[multiple] dynamic handlers hint backend header sz
4 years ago
Glenn Strauss
31d16af78b
[core] perf: simpler buffer_string_space()
4 years ago
Glenn Strauss
1539cbfdc1
[core] perf: chunkqueue buffers already sized up
...
provide new interface for sizing to even larger if very large request
(chunkqueue_prepend_buffer_open_sz())
4 years ago
Glenn Strauss
f1e5f74f8f
[mod_webdav] return 403 if file should exist
...
return 403 if file should exist for PROPPATCH or LOCK, but stat fails
4 years ago
Glenn Strauss
a1cda77f9b
[mod_openssl] prefer using TLS_server_method()
...
prefer TLS_server_method() instead of SSLv23_server_method()
(SSLv23_server_method() is deprecated in openssl 1.1.0)
4 years ago
Glenn Strauss
aa00359e6c
[mod_cgi] perf: use stat_cache for cgi handler
4 years ago
Glenn Strauss
d182e90bb1
[mod_openssl] SSL_read before second SSL_shutdown
...
call SSL_read() before calling SSL_shutdown() a second time.
Drain SSL read buffers in case pending records need processing.
x-ref:
"New session ticket breaks bidirectional shutdown of TLS 1.3 connection"
https://github.com/openssl/openssl/issues/6262
reported to lighttpd by intexk
https://www.lighttpd.net/2018/10/14/1.4.51/
4 years ago
Glenn Strauss
89062d8134
[mod_openssl] move SSL_shutdown() to separate func
...
mod_openssl_close_notify()
4 years ago
bc91bbd0c6
[core] Don't call RAND_cleanup with OpenSSL 1.1.x
...
RAND_cleanup is deprecated and does nothing with 1.1.x.
It also breaks with OpenSSL compiled with no deprecated APIs.
(-DOPENSSL_API_COMPAT=<version>)
github: closes #93
4 years ago
Glenn Strauss
6dcdc1af18
[mod_fastcgi,mod_scgi] perf: env accumulation
4 years ago
Glenn Strauss
407b4d140b
[mod_cgi] perf: reuse buffers for creating CGI env
4 years ago
Glenn Strauss
e0dbff4750
[core] slightly simpler flag append to string
4 years ago
Glenn Strauss
5fe9339942
[core] perf: inline buffer_append_string_buffer()
4 years ago
Glenn Strauss
8c0792a58c
[multiple] better packing of struct chunk
...
remove c->file.name
previous commit modified c->file.name to share buffer with c->mem
4 years ago
Glenn Strauss
3d8d56d8de
[multiple] perf: reuse large buffers w/ backend
4 years ago
Glenn Strauss
fccc7fc607
[core] perf: chunk.c chunk pool
...
reuse chunks across chunkqueues; remove cq->unused, cq->unused_chunks
reuse same buffer c->mem and c->file.name in chunk
4 years ago
Glenn Strauss
c79bc31609
[mod_fastcgi] perf: reduce data copies
...
http_chunk_transfer_cqlen()
4 years ago
Glenn Strauss
2df8f9ebf4
[mod_fastcgi] perf: reduce data copies
...
reduce data copies while reading FastCGI packets from FastCGI backend
4 years ago
Glenn Strauss
f934615466
[core] permit env vars to be set with blank value
...
(regression in lighttpd 1.4.51)
4 years ago
Glenn Strauss
f410431b58
[multiple] perf: use larger initial backend buffer
...
use larger initial backend buffer if the client request is large
4 years ago
Glenn Strauss
7e8beee577
[multiple] perf: power-2 alloc large headers
...
allocate large headers by power-2 as headers to backend is generated
(mod_proxy, mod_fastcgi, mod_scgi)
4 years ago
Glenn Strauss
610d5c03ab
[multiple] perf: write headers to backend write cq
...
write headers directly to dynamic backend write chunkqueue
(mod_proxy, mod_fastcgi, mod_scgi)
4 years ago
Glenn Strauss
a9ae35128d
[core] chunk.c code reuse
4 years ago
Glenn Strauss
61e507bfe5
[core] perf: better buf reuse reading from backend
4 years ago
Glenn Strauss
40d3408328
[core] perf: append response directly into write q
...
prepend non-dynamic response directly into write queue chunkqueue
4 years ago
Glenn Strauss
758174ecbc
[core] perf: specialized func for array sorting
...
specialized func for array sorting for use in binary search
4 years ago
Glenn Strauss
59a71815dc
[core] perf: copy small strings; extend last chunk
...
copy small strings; extend last chunk
(instead of creating new, tiny chunk)
4 years ago
Glenn Strauss
e7c840502a
[core] perf: copy small strings; better buf reuse
...
copy small strings to write queue for better buffer reuse
(instead of swapping with larger buffers in write chunkqueue)
4 years ago
Glenn Strauss
23c72fc606
[core] perf: prepend headers directly into write q
...
prepend response headers directly into write queue chunkqueue
4 years ago
Glenn Strauss
5a159e828d
[core] perf: size write buffers for reuse
...
size write buffers for reuse in common cases
(so that write buffers are BUFFER_MAX_REUSE_SIZE)
4 years ago
Glenn Strauss
2e23b43d12
[core] perf: buffer_align_size() identity if align
...
use identity if requested size is already aligned to BUFFER_PIECE_SIZE
4 years ago
Glenn Strauss
929909a999
[core] perf: skip redundant prepare copy calls
...
skip redundant calls to buffer_string_prepare_copy() when simply
clearing buffer is sufficient, e.g. when a blanked buffer is desired
before calling buffer_append_strftime(), which internally prepares
buffer for the resultant formatted string
4 years ago
Glenn Strauss
d4a93a531d
[core] perf: buffer_copy_string_len()
...
buffer_copy_string_len() short-circuit common case,
but preserve blank-string initialization side-effect
if buffer is empty
4 years ago
Glenn Strauss
15277bf9e1
[core] perf: one-element cache for host normalize
...
one-element cache for http_request_host_normalize() on IP address
4 years ago
Glenn Strauss
babf8112f7
[core] perf: buffer_free_ptr() __attribute__ cold
4 years ago
Glenn Strauss
062089ff14
[core] perf: array_reset_data_strings()
...
array_reset_data_strings() specialization
4 years ago
Glenn Strauss
29dd9379ff
[core] perf: call connection_reset() fewer times
...
connection_reset() is called at end of request to prepare for keep-alive
or to free resources while waiting for client to close connection.
This is the reason why it is called earlier towards end of request
rather than always waiting until the very end of request when socket
is closed.
connection_reset() used to additionally be called for every new socket
connection, but now is only called when new conns are allocated and
initialized.
4 years ago
Glenn Strauss
3d60d8535f
[build] Fix unportable test(1) operator
...
(thx Thomas Klausner)
4 years ago
Rosen Penev