]>
git.ipfire.org Git - thirdparty/squid.git/log
wessels [Wed, 16 Sep 1998 15:24:25 +0000 (15:24 +0000)]
adding
wessels [Wed, 16 Sep 1998 04:36:18 +0000 (04:36 +0000)]
update from Bertold
wessels [Wed, 16 Sep 1998 04:29:30 +0000 (04:29 +0000)]
remove fatal_dump
wessels [Wed, 16 Sep 1998 04:05:10 +0000 (04:05 +0000)]
HEAD fixes (damn I hate HEAD)
on the server-side, HEAD transfers are always done soon as we make it
to the end of the reply headers
on the client-side, we can't do IMS for HEAD. Always return cache HITS,
never a refresh.
httpMaybeRemovePublic() gets bastardized already (thanks henrik). Also
make it remove HEAD objects for new GET requests.
wessels [Wed, 16 Sep 1998 04:05:09 +0000 (04:05 +0000)]
HEAD fixes (damn I hate HEAD)
on the server-side, HEAD transfers are always done soon as we make it
to the end of the reply headers
on the client-side, we can't do IMS for HEAD. Always return cache HITS,
never a refresh.
httpMaybeRemovePublic() gets bastardized already. Also
make it remove HEAD objects for new GET requests.
wessels [Wed, 16 Sep 1998 02:36:15 +0000 (02:36 +0000)]
gindent
wessels [Wed, 16 Sep 1998 02:24:07 +0000 (02:24 +0000)]
look for cached HEAD objects for HEAD requests -- don't always look for
GET objects for HEAD requests
wessels [Wed, 16 Sep 1998 02:16:22 +0000 (02:16 +0000)]
in storeRelease(), move aioCancel() call to after store_rebuilding check
wessels [Wed, 16 Sep 1998 01:37:41 +0000 (01:37 +0000)]
gindent
wessels [Tue, 15 Sep 1998 13:17:50 +0000 (13:17 +0000)]
make x-routines a little bit mt-safer
wessels [Tue, 15 Sep 1998 12:49:58 +0000 (12:49 +0000)]
relocate all-redirectors-have-exited check.
removed NRedirectorsOpen == 0 assertion
wessels [Tue, 15 Sep 1998 12:38:28 +0000 (12:38 +0000)]
orphan the redirector in redirectShutdown; gindent
wessels [Tue, 15 Sep 1998 12:35:24 +0000 (12:35 +0000)]
fixed ENDIAN mixup
wessels [Tue, 15 Sep 1998 12:34:55 +0000 (12:34 +0000)]
compiler
wessels [Tue, 15 Sep 1998 12:25:17 +0000 (12:25 +0000)]
fix select case for stewarts incoming code
wessels [Tue, 15 Sep 1998 12:23:05 +0000 (12:23 +0000)]
removed redirectUnregister
wessels [Tue, 15 Sep 1998 12:22:16 +0000 (12:22 +0000)]
reimplement redirector code to handle reconfigure better
wessels [Tue, 15 Sep 1998 10:23:37 +0000 (10:23 +0000)]
fix new flags endian and size bugs
wessels [Tue, 15 Sep 1998 05:42:37 +0000 (05:42 +0000)]
warning
wessels [Tue, 15 Sep 1998 05:42:07 +0000 (05:42 +0000)]
compiler
wessels [Tue, 15 Sep 1998 05:35:24 +0000 (05:35 +0000)]
new, improved comm_connect_addr():
don't call connect() after the first one would block. Instead,
call getsockopt(SO_ERROR). This should handle the HP-UX and
FreeBSD EALREADY looping.
wessels [Tue, 15 Sep 1998 04:40:11 +0000 (04:40 +0000)]
compiler
wessels [Tue, 15 Sep 1998 04:38:27 +0000 (04:38 +0000)]
make int:1 flags unsigned
wessels [Tue, 15 Sep 1998 04:34:02 +0000 (04:34 +0000)]
define Safe_ports
wessels [Tue, 15 Sep 1998 04:29:11 +0000 (04:29 +0000)]
update
wessels [Tue, 15 Sep 1998 04:27:57 +0000 (04:27 +0000)]
integer range support for PORT ACLs
wessels [Tue, 15 Sep 1998 04:20:00 +0000 (04:20 +0000)]
typo
wessels [Tue, 15 Sep 1998 04:18:04 +0000 (04:18 +0000)]
more EBIT replacements
wessels [Tue, 15 Sep 1998 04:17:56 +0000 (04:17 +0000)]
more EBIT replacements
wessels [Tue, 15 Sep 1998 03:58:45 +0000 (03:58 +0000)]
replaced NEIGHBOR_ and HELPER_ EBIT calls
wessels [Tue, 15 Sep 1998 03:28:00 +0000 (03:28 +0000)]
removed EBIT macros for StoreEntry->flag
wessels [Tue, 15 Sep 1998 02:29:30 +0000 (02:29 +0000)]
timeout calculation bug; off by 1000!
wessels [Tue, 15 Sep 1998 02:29:12 +0000 (02:29 +0000)]
update
wessels [Mon, 14 Sep 1998 21:46:48 +0000 (21:46 +0000)]
prevent NPR
wessels [Mon, 14 Sep 1998 08:33:54 +0000 (08:33 +0000)]
update comments
wessels [Mon, 14 Sep 1998 08:33:01 +0000 (08:33 +0000)]
remove DOUBTFUL code
wessels [Fri, 11 Sep 1998 23:07:41 +0000 (23:07 +0000)]
Removed 'method bits' hack from ICP query/reply. This will break
interoperability. Instead the ICP server treats all queries as
METHOD_GET.
The ICP client code, however, uses an array of the most recent
8192 keys sent in ICP queries. The reqnum is an index to this
array, so we can look up the key for an ICP reply easily.
Also changed storeKeyPublic and storeKeyPrivate to NOT use snprintf, but
just pass integers 'id' and 'method' directly to md5update.
wessels [Fri, 11 Sep 1998 21:56:43 +0000 (21:56 +0000)]
duplicate ICP_SECHO case
wessels [Fri, 11 Sep 1998 11:22:56 +0000 (11:22 +0000)]
update
wessels [Fri, 11 Sep 1998 05:46:09 +0000 (05:46 +0000)]
typo
wessels [Fri, 11 Sep 1998 04:30:19 +0000 (04:30 +0000)]
added asndb cachemgr output
wessels [Fri, 11 Sep 1998 01:56:28 +0000 (01:56 +0000)]
don't close same FD twice if unlinkd_wfd == unlinkd_rfd.
wessels [Fri, 11 Sep 1998 01:50:55 +0000 (01:50 +0000)]
If we are over the FD limit, and we can read more on the server-side
without deferring, then postpone the swapout open until the FD usage
goes down.
wessels [Fri, 11 Sep 1998 01:48:55 +0000 (01:48 +0000)]
further safegurads in httpMaybeRemovePublic. return if the key
is PUBLIC, and assert that we don't remove the "current" entry.
wessels [Fri, 11 Sep 1998 01:45:53 +0000 (01:45 +0000)]
If neighbors don't support our private keys, then we can't use the
httpMaybeRemovePublic() function because ALL entries are public and
we end up removing our own.
wessels [Thu, 10 Sep 1998 22:03:25 +0000 (22:03 +0000)]
ordering typos
wessels [Thu, 10 Sep 1998 08:55:59 +0000 (08:55 +0000)]
debug level
wessels [Thu, 10 Sep 1998 02:05:47 +0000 (02:05 +0000)]
changed no-caching based on http service time to be based on number
of disk files open.
wessels [Thu, 10 Sep 1998 00:21:50 +0000 (00:21 +0000)]
add defaults
wessels [Thu, 10 Sep 1998 00:20:49 +0000 (00:20 +0000)]
more typo
wessels [Thu, 10 Sep 1998 00:19:31 +0000 (00:19 +0000)]
typo
wessels [Thu, 10 Sep 1998 00:18:02 +0000 (00:18 +0000)]
From: Stewart Forster <slf@connect.com.au>
Another important patch for HIGH loads. Someone mentioned once when we were
designing the automatic comm_select_incoming counters tuning code that we
should try not to check too often. I argued against that and I was wrong.
At high loads, the sheer load of polling starts to bite badly, and so the
below patch reduces the amount of polling needed. Now squid slows down more
sanely at really high loads (rather than brick wall). Ideally these values
should be configurable and called:
incoming_icp_average 6
incoming_http_average 4
min_icp_poll_cnt 8
min_http_poll_cnt 8
but I don't have the time to write this up. These values will definately
vary from site to site depending on the ICP/HHTP load ratio. The above
values work okay for us, but maybe not for everyone.....
The lower the poll_cnts will mean the more rapid the polling.
The lower the incoming_*_average will mean the more rapid the polling.
The min_*_poll_cnt will put a maximum limit on the rate of polls.
wessels [Wed, 9 Sep 1998 23:51:05 +0000 (23:51 +0000)]
note when ICP service is disabled
wessels [Wed, 9 Sep 1998 23:47:09 +0000 (23:47 +0000)]
comments
wessels [Wed, 9 Sep 1998 22:47:56 +0000 (22:47 +0000)]
moved neighbor RTT updates out of neighborAlive() because we sometimes
call neighborAlive for ICP replies which don't correspond to the StoreEntry.
wessels [Wed, 9 Sep 1998 22:47:08 +0000 (22:47 +0000)]
make high-http-median-svc-time no-cache WIP
wessels [Sun, 6 Sep 1998 23:11:34 +0000 (23:11 +0000)]
only close "stuck" swapout files in STORE_OK state if there is
not a write already queued
wessels [Sun, 6 Sep 1998 22:20:04 +0000 (22:20 +0000)]
fix swapout FD leak for ENTRY_BAD_LENGTH objects
wessels [Sat, 5 Sep 1998 05:04:38 +0000 (05:04 +0000)]
syscall counters, mostly I/O
wessels [Sat, 5 Sep 1998 03:40:27 +0000 (03:40 +0000)]
skip bogus RTT calculation when start_time == 0
wessels [Fri, 4 Sep 1998 00:39:06 +0000 (00:39 +0000)]
memFree MD5 digests instead of xfree!
wessels [Thu, 3 Sep 1998 09:48:36 +0000 (09:48 +0000)]
comm_select # ready fd's histogram
wessels [Thu, 3 Sep 1998 09:37:33 +0000 (09:37 +0000)]
more select stats to debug poor performance
wessels [Thu, 3 Sep 1998 09:36:24 +0000 (09:36 +0000)]
storeCheckCachable() returns 0 if http_median_svc_time exceeds 2 seconds
wessels [Wed, 2 Sep 1998 23:21:00 +0000 (23:21 +0000)]
debug
wessels [Wed, 2 Sep 1998 11:56:45 +0000 (11:56 +0000)]
It apparently was a BAD idea to have storeHashInsert() move PRIVATE
entries to the tail of the LRU list. Instead, lets try doing that
in storeUnlockObject when the lock count is zero.
wessels [Wed, 2 Sep 1998 10:00:42 +0000 (10:00 +0000)]
make sure MAX_MEM_BUF_SZ is at least twice CLIENT_SOCK_SZ
wessels [Wed, 2 Sep 1998 05:36:13 +0000 (05:36 +0000)]
update
wessels [Wed, 2 Sep 1998 05:32:44 +0000 (05:32 +0000)]
debugging
wessels [Wed, 2 Sep 1998 05:31:22 +0000 (05:31 +0000)]
Internal HTTP requests bypass miss_access rules. Internal requests
are passed to fwdStart() with client_address == no_addr.
wessels [Wed, 2 Sep 1998 04:12:06 +0000 (04:12 +0000)]
spello
wessels [Tue, 1 Sep 1998 11:35:28 +0000 (11:35 +0000)]
check for libbind.a; give libbind.a higher precedence than libresolv.a
wessels [Sun, 30 Aug 1998 11:21:38 +0000 (11:21 +0000)]
- added fqdncache_size configuration option
- added MD5 hashes to mem pools
- removed some DONT code
rousskov [Fri, 28 Aug 1998 04:01:47 +0000 (04:01 +0000)]
- icp_port is now icp.port
rousskov [Fri, 28 Aug 1998 04:01:12 +0000 (04:01 +0000)]
- all bit-fields of size 1 should be unsigned according to some compilers
rousskov [Fri, 28 Aug 1998 04:00:07 +0000 (04:00 +0000)]
- one IRIX an unsigned size_t creates problems when we do (len-1) in snprintf
fixed by cheking that len is positive before subtracting 1.
- return value was sometimes uninitialized when calling snprintf
rousskov [Fri, 28 Aug 1998 02:43:53 +0000 (02:43 +0000)]
- added custom compiler and linker options for mips-sgi-irix6.4
wessels [Thu, 27 Aug 1998 12:29:30 +0000 (12:29 +0000)]
unuse hexdump
wessels [Thu, 27 Aug 1998 12:28:54 +0000 (12:28 +0000)]
renamed icp_ping_data to just ping_data (for HTCP)
wessels [Thu, 27 Aug 1998 12:28:28 +0000 (12:28 +0000)]
HTCP progress
wessels [Thu, 27 Aug 1998 03:42:05 +0000 (03:42 +0000)]
undo debugging which wasnt supposed to be committed
wessels [Thu, 27 Aug 1998 01:53:39 +0000 (01:53 +0000)]
HTCP progress
wessels [Thu, 27 Aug 1998 01:08:54 +0000 (01:08 +0000)]
HTCP progress
wessels [Thu, 27 Aug 1998 01:08:00 +0000 (01:08 +0000)]
progress
wessels [Thu, 27 Aug 1998 00:48:56 +0000 (00:48 +0000)]
removed unused opcode arg of neighborsCountIgnored()
wessels [Wed, 26 Aug 1998 11:36:42 +0000 (11:36 +0000)]
HTCP progress
wessels [Wed, 26 Aug 1998 11:34:51 +0000 (11:34 +0000)]
progress
wessels [Wed, 26 Aug 1998 01:19:05 +0000 (01:19 +0000)]
added hoHtcpReply
wessels [Wed, 26 Aug 1998 01:10:43 +0000 (01:10 +0000)]
added hoHtcpReply
made httpHeaderReset public
wessels [Wed, 26 Aug 1998 01:10:10 +0000 (01:10 +0000)]
update
wessels [Wed, 26 Aug 1998 01:09:20 +0000 (01:09 +0000)]
added netdbHostData as a faster way to get #samples, RTT, and HOPS
wessels [Wed, 26 Aug 1998 01:04:00 +0000 (01:04 +0000)]
Added resp_hdrs, entity_hdrs, cache_hdrs in TST HIT replies
wessels [Tue, 25 Aug 1998 21:58:14 +0000 (21:58 +0000)]
comment typo
wessels [Tue, 25 Aug 1998 13:02:42 +0000 (13:02 +0000)]
progress
wessels [Tue, 25 Aug 1998 12:20:42 +0000 (12:20 +0000)]
progress
wessels [Tue, 25 Aug 1998 10:12:51 +0000 (10:12 +0000)]
update
wessels [Tue, 25 Aug 1998 10:11:35 +0000 (10:11 +0000)]
separate client socket buffers from MEM_4K_BUF so we can change the
socket buffer size.
wessels [Tue, 25 Aug 1998 09:35:18 +0000 (09:35 +0000)]
update
wessels [Tue, 25 Aug 1998 04:06:44 +0000 (04:06 +0000)]
Henrik's fixes for DW adding FD to AIOCB bugs
wessels [Tue, 25 Aug 1998 03:11:30 +0000 (03:11 +0000)]
Do a better job of swapping out uncachable objects. We need to be
smarter about calculating the new inmem_lo value. Even though
the object may not be ENTRY_CACHABLE, we might be currently swapping it
out. We have some assertion failures when we free past the swapout_done
offset.