]> git.ipfire.org Git - thirdparty/squid.git/log
thirdparty/squid.git
26 years agoFrom: Henrik Nordstrom <hno@hem.passagen.se>
wessels [Fri, 22 Jan 1999 06:15:35 +0000 (06:15 +0000)] 
From:    Henrik Nordstrom <hno@hem.passagen.se>
This patch removes the last traces of cache_mem high/low water marks.
Keep hot object store at cache_mem instead of cache_mem_high.

26 years agoFrom: Henrik Nordstrom <hno@hem.passagen.se>
wessels [Fri, 22 Jan 1999 04:13:00 +0000 (04:13 +0000)] 
From:    Henrik Nordstrom <hno@hem.passagen.se>

* Changed async_io.c and aiops.c to use MemPool pools for their
"request" structures.

* Fix for a odd situation where I/O operations could get stuck in the
queue if there was no new I/O activity. Symptoms seen was: 1. The
shutdown code could trigger some assertions which seemed impossible. 2.
Some times a number of files could be kept open until the next swapout.

* Changed the log message shown when running low on I/O threads, and
also changed logging of this message to only occur if out of threads for
more than one second.

* Moved some aiops.c log messages to lower priorities.

* Removed that special low-priority queue for unlinks. I had some
strange effects from it, and the number of queued unlinks fluctuaded a
bit to much when under heavy load. If store maintaince is a problem then
this should it be corrected at a higher level, and not by delaying the
unlinks.

* Minor preparations for making NUMTHREADS a squid.conf tunable.

* moved memClean later in the shutdown procedure, to allow async-io to
use MemPool..

26 years agochanged open_disk_fd to store_open_disk_fd.
wessels [Fri, 22 Jan 1999 04:10:32 +0000 (04:10 +0000)] 
changed open_disk_fd to store_open_disk_fd.

1) we want to track only STORE files, not necessarily log files.  Log files
   don't contribute much to disk load.

2) this accounting belongs in the "store filesystem" layer, not
   the "fd layer".  Mostly motivated by Henrik's addition of
   Opening_FD counter in the async code.

26 years agoclear ctrl.fd timeout when setting data.fd timeout
wessels [Thu, 21 Jan 1999 12:26:14 +0000 (12:26 +0000)] 
clear ctrl.fd timeout when setting data.fd timeout

26 years agoportability
wessels [Thu, 21 Jan 1999 02:43:00 +0000 (02:43 +0000)] 
portability

26 years agoFixed up *Timeout() functions. We might get fooTimeout() called
wessels [Thu, 21 Jan 1999 02:27:08 +0000 (02:27 +0000)] 
Fixed up *Timeout() functions.  We might get fooTimeout() called
when the entry has been ENTRY_ABORTED because the abort callback
comes in an event, not immediately.  Also, fooTimeout() now uses
fwdFail instead of building its own error page.

26 years agoclear ftpState->data.fd after closing it
wessels [Wed, 20 Jan 1999 10:02:48 +0000 (10:02 +0000)] 
clear ftpState->data.fd after closing it

26 years agohenrik sez don't cache objects that we can't refresh later
wessels [Wed, 20 Jan 1999 06:17:58 +0000 (06:17 +0000)] 
henrik sez don't cache objects that we can't refresh later

26 years agogindent
wessels [Wed, 20 Jan 1999 06:16:48 +0000 (06:16 +0000)] 
gindent

26 years agorestart bug fixes
wessels [Wed, 20 Jan 1999 06:16:05 +0000 (06:16 +0000)] 
restart bug fixes

26 years agoFrom: Tony Finch <fanf@demon.net>
wessels [Wed, 20 Jan 1999 05:00:57 +0000 (05:00 +0000)] 
From:    Tony Finch <fanf@demon.net>
The patch below adds accelerator mode functionality to tcp-banger2. In
this mode it makes requests like
        PURGE /path/to/file HTTP/1.0
        Host: www.example.com
instead of
        PURGE http://www.example.com/path/to/file HTTP/1.0

We use it for keeping our accelerators up-to-date with respect to
customer changes on their sites.

26 years agoFrom: Robert Side <rside@aiinc.bc.ca>
wessels [Wed, 20 Jan 1999 04:55:54 +0000 (04:55 +0000)] 
From:    Robert Side <rside@aiinc.bc.ca>
Date:    Mon, 18 Jan 1999 17:48:00 GMT

To get squid version 2.1.Patch1 to compile and run on SCO Unix Version
3.2v4.2 you need to use this configure line (from a csh like shell)

env LIBS=-lintl ./configure --disable-poll

The env LIBS=-lintl is required so squid can find strftime. Should
configure check for the existence of strftime in the standard
libaries and if it can't find it, check libintl.a on SCO systems?

26 years agoFTP REST support
wessels [Wed, 20 Jan 1999 04:40:39 +0000 (04:40 +0000)] 
FTP REST support

26 years agoHenrik Added a (local) hit rate to mempool metering.
wessels [Wed, 20 Jan 1999 03:43:46 +0000 (03:43 +0000)] 
Henrik Added a (local) hit rate to mempool metering.

26 years agoWe had false loop detections because "ee.lbl.gov" matched "ren.ee.lbl.gov".
wessels [Wed, 20 Jan 1999 03:26:32 +0000 (03:26 +0000)] 
We had false loop detections because "ee.lbl.gov" matched "ren.ee.lbl.gov".
Added variable (ThisCache2) that has a space prepended to the hostname
so we now search for " ee.lbl.gov".

26 years agoslightly more intelligent loop detector
wessels [Wed, 20 Jan 1999 02:23:27 +0000 (02:23 +0000)] 
slightly more intelligent loop detector

26 years agoENTRY_SPECIAL objects should never be aborted
wessels [Wed, 20 Jan 1999 02:03:05 +0000 (02:03 +0000)] 
ENTRY_SPECIAL objects should never be aborted

26 years agochanged assertion. in fwdStateFree() it is okay if ENTRY_FWD_HDR_WAIT
wessels [Wed, 20 Jan 1999 01:11:16 +0000 (01:11 +0000)] 
changed assertion.   in fwdStateFree() it is okay if ENTRY_FWD_HDR_WAIT
is not set, so long as there are no clients reading the entry.

26 years agoremoved fdTouch debugging
wessels [Wed, 20 Jan 1999 00:41:02 +0000 (00:41 +0000)] 
removed fdTouch debugging

26 years agobraino, typo
wessels [Tue, 19 Jan 1999 23:35:13 +0000 (23:35 +0000)] 
braino, typo

26 years agodamn; I guess we need this too
wessels [Tue, 19 Jan 1999 13:38:09 +0000 (13:38 +0000)] 
damn; I guess we need this too

26 years agofwdStateFree will call storeComplete() if the entry is in
wessels [Tue, 19 Jan 1999 13:31:42 +0000 (13:31 +0000)] 
fwdStateFree will call storeComplete() if the entry is in
STORE_PENDING state.  This can happen in the server-side just
chooses to close the server socket without setting up any
sort of failure information (fwdFail).

also added / changed debugging

26 years agoerrorpage.c: clear ENTRY_FWD_HDR_WAIT after creating an error page,
wessels [Tue, 19 Jan 1999 12:24:47 +0000 (12:24 +0000)] 
errorpage.c: clear ENTRY_FWD_HDR_WAIT after creating an error page,
     before calling storeComplete

forward.c: break some calling loops for things like failed DNS lookups.
   added FwdState->flags.dont_retry

   assert that ENTRY_FWD_HDR_WAIT is clear when we free
   the FwdState

   added loop detector to fwdConnectDone

26 years ago2.1 branch merge and gindent
wessels [Tue, 19 Jan 1999 09:24:19 +0000 (09:24 +0000)] 
2.1 branch merge and gindent

26 years agocast
wessels [Tue, 19 Jan 1999 08:13:47 +0000 (08:13 +0000)] 
cast

26 years agoupdate
wessels [Tue, 19 Jan 1999 06:21:31 +0000 (06:21 +0000)] 
update

26 years agoFD leak for 504 replies because we leaked "http->entry" when creating
wessels [Tue, 19 Jan 1999 06:15:42 +0000 (06:15 +0000)] 
FD leak for 504 replies because we leaked "http->entry" when creating
a new reply.

26 years agosigh. Okay, this might work better.
wessels [Tue, 19 Jan 1999 05:23:32 +0000 (05:23 +0000)] 
sigh.  Okay, this might work better.

To "decouple" storeAbort calling into the server side (fwdAbort) and
wreaking havoc, lets have the storeAbort callback done as an event,
thus outside of comm_select loop.

Hopefully this means we don't need the gross "flags.delayed_comm_close"
and current_hdl_fd crap.

Also, since we moved the abort callback to an event, we might end up
doing a server read after the entry has been aborted, but before
the event is called.  Thus, the server modules need to check
for ENTRY_ABORTED before reading.

this commit also adds some (temp?) debugging to track open
disk FDs

26 years agooops
wessels [Mon, 18 Jan 1999 02:46:28 +0000 (02:46 +0000)] 
oops

26 years agoPid file now written on SIGHUP too.
glenn [Sat, 16 Jan 1999 07:03:38 +0000 (07:03 +0000)] 
Pid file now written on SIGHUP too.

26 years agoSNMP Agent now returns object instances correctly instead of objects
glenn [Sat, 16 Jan 1999 03:48:15 +0000 (03:48 +0000)] 
SNMP Agent now returns object instances correctly instead of objects

26 years agoprobably multicast-miss-stream buffer overrun bug causing lots of coredumps
wessels [Fri, 15 Jan 1999 13:36:35 +0000 (13:36 +0000)] 
probably multicast-miss-stream buffer overrun bug causing lots of coredumps

26 years agochange debug level for storeLateRelease
wessels [Fri, 15 Jan 1999 13:30:10 +0000 (13:30 +0000)] 
change debug level for storeLateRelease

26 years agoonly call errorAppendEntry in one spot
wessels [Fri, 15 Jan 1999 13:11:42 +0000 (13:11 +0000)] 
only call errorAppendEntry in one spot

26 years agoFD leak for unauthenticated, non-anonymous FTP requests
wessels [Fri, 15 Jan 1999 07:16:19 +0000 (07:16 +0000)] 
FD leak for unauthenticated, non-anonymous FTP requests

26 years agoadding
wessels [Fri, 15 Jan 1999 06:58:52 +0000 (06:58 +0000)] 
adding

26 years agodon't expose given passwords in error messages
wessels [Fri, 15 Jan 1999 06:55:40 +0000 (06:55 +0000)] 
don't expose given passwords in error messages

26 years agoHere is a small addon to tcp-banger2: Warn if an unexpected HTTP status
wessels [Fri, 15 Jan 1999 06:45:09 +0000 (06:45 +0000)] 
Here is a small addon to tcp-banger2: Warn if an unexpected HTTP status
is received.

26 years agoWe used to close persistent connections if the number of free
wessels [Fri, 15 Jan 1999 06:15:43 +0000 (06:15 +0000)] 
We used to close persistent connections if the number of free
descriptors was less than four times the reservedFD limit.  On
small-FD systems, we will never keep a pconn open because reservedFD
is set to numberFD/4.

Lets try twice the reserved FD limit.

26 years agoFrom: Martin Hamilton <martin@net.lut.ac.uk>
wessels [Thu, 14 Jan 1999 23:22:58 +0000 (23:22 +0000)] 
From:    Martin Hamilton <martin@net.lut.ac.uk>
The line which reads:
                  send($sock, $request, 0, $them);
should say
                  send($sock, $request, 0);

26 years agoadding
wessels [Thu, 14 Jan 1999 06:49:18 +0000 (06:49 +0000)] 
adding

26 years agoOS/2 fixes from Doug Nazar nazard@man-assoc.on.ca
wessels [Thu, 14 Jan 1999 06:24:09 +0000 (06:24 +0000)] 
OS/2 fixes from Doug Nazar nazard@man-assoc.on.ca

26 years agoupdate
wessels [Thu, 14 Jan 1999 05:17:06 +0000 (05:17 +0000)] 
update

26 years agoFrom: Dancer <dancer@zeor.simegen.com>
wessels [Thu, 14 Jan 1999 05:13:38 +0000 (05:13 +0000)] 
From:    Dancer <dancer@zeor.simegen.com>

It turns out that if you are using an external authenticator, the
username is only logged on _subsequent_ (that is, cached) authentication
hits. If the user's authentication credentials expire from the cache,
their next (initially authenticated) request doesn't log any user info.

I'm pretty sure that this behaviour is unintentional.
Here's a patch:

26 years agochange error message for fwdConnectTimeout to ERR_CONNECT_FAIL/ETIMEDOUT
wessels [Thu, 14 Jan 1999 04:59:52 +0000 (04:59 +0000)] 
change error message for fwdConnectTimeout to ERR_CONNECT_FAIL/ETIMEDOUT

26 years agomore fixes for assertions on HP-UX/GCC
wessels [Thu, 14 Jan 1999 04:55:23 +0000 (04:55 +0000)] 
more fixes for assertions on HP-UX/GCC

26 years agoFreeBSD has a linuxthreads package
wessels [Thu, 14 Jan 1999 04:00:10 +0000 (04:00 +0000)] 
FreeBSD has a linuxthreads package

26 years agoCheckQuickAbort2 debug levels
wessels [Wed, 13 Jan 1999 12:56:08 +0000 (12:56 +0000)] 
CheckQuickAbort2 debug levels

26 years agostoreAppend() might abort an entry in httpReadReply()
wessels [Wed, 13 Jan 1999 12:54:48 +0000 (12:54 +0000)] 
storeAppend() might abort an entry in httpReadReply()

26 years agoopen --> flags.open
wessels [Wed, 13 Jan 1999 06:52:08 +0000 (06:52 +0000)] 
open --> flags.open

26 years agogindent
wessels [Wed, 13 Jan 1999 06:38:35 +0000 (06:38 +0000)] 
gindent

26 years ago- removed storeAbort callback-server side argument
wessels [Wed, 13 Jan 1999 06:37:41 +0000 (06:37 +0000)] 
- removed storeAbort callback-server side argument
- removed _fde->mode, I guess it was a bad idea anyway
- removed recently added ENTRY_ABORTED checks in server modules
- replaced fwdAbort()
- fixed a mem_obj->request == NULL bug when fwdStart() returns
    503 service unavailable.

26 years agoupdate
wessels [Wed, 13 Jan 1999 06:32:55 +0000 (06:32 +0000)] 
update

26 years agocheck _SQUID_NEWSOS6_ instead
wessels [Wed, 13 Jan 1999 06:22:28 +0000 (06:22 +0000)] 
check _SQUID_NEWSOS6_ instead

26 years agochanged struct _fde->open to _fde->flags.open to save a few bytes
wessels [Tue, 12 Jan 1999 23:42:15 +0000 (23:42 +0000)] 
changed struct _fde->open to _fde->flags.open to save a few bytes

26 years agoin errorAppendEntry:
wessels [Tue, 12 Jan 1999 23:40:35 +0000 (23:40 +0000)] 
in errorAppendEntry:
    If the entry is not STORE_PENDING, then no clients
    care about it, and we don't need to generate an
    error message

26 years agotypo
wessels [Tue, 12 Jan 1999 23:23:22 +0000 (23:23 +0000)] 
typo

26 years agobe more flexible in storeComplete(). instead of asserting that store_status
wessels [Tue, 12 Jan 1999 23:22:16 +0000 (23:22 +0000)] 
be more flexible in storeComplete().  instead of asserting that store_status
is STORE_PENDING, if its not, assert that the entry was aborted and
there are zero clients.

26 years agoserver-side handlers check for ENTRY_ABORTED and close their FD
wessels [Tue, 12 Jan 1999 22:47:51 +0000 (22:47 +0000)] 
server-side handlers check for ENTRY_ABORTED and close their FD

pump.c needs FwdState to set fwdFail()

26 years agooops, we didn't cache cache-control: no-cache
wessels [Tue, 12 Jan 1999 07:20:42 +0000 (07:20 +0000)] 
oops, we didn't cache cache-control: no-cache

26 years agodeleted ERR_CLIENT_ABORT
wessels [Tue, 12 Jan 1999 06:29:41 +0000 (06:29 +0000)] 
deleted ERR_CLIENT_ABORT
added ERR_SHUTTING_DOWN
set theOutIcpConnect = -1 when we close that FD
have fwdStart() return ERR_SHUTTING_DOWN if we are shutting down
the cache

26 years agodeleted ERR_CLIENT_ABORT; added ERR_SHUTTING_DOWN
wessels [Tue, 12 Jan 1999 06:26:42 +0000 (06:26 +0000)] 
deleted ERR_CLIENT_ABORT; added ERR_SHUTTING_DOWN

26 years agoadding
wessels [Tue, 12 Jan 1999 05:56:38 +0000 (05:56 +0000)] 
adding

26 years agoThis patch fully transforms proxy_auth into a ACL type, allowing both
wessels [Tue, 12 Jan 1999 05:54:15 +0000 (05:54 +0000)] 
This patch fully transforms proxy_auth into a ACL type, allowing both
allow and deny in any order suitable.

Authentication is requested from the user if:
* proxy_auth is used and no valid authentication header is present.
* wrong password is used
* denied by a proxy_auth ACL

A positive sideeffect is that is is also possible to use deny_info on
proxy_auth ACLs to change the message returned. It also fixes the
problem where proxy_auth user wasn't logged if the proxy_auth user info
wasn't cached.

This patch fully replaces my previous attemt.

There is a small collision with one of my other changes in proto.h, but
I assume that you know how to resolve this.

26 years agomake commResetFD() more like comm_open(); set outgoing address, etc.
wessels [Tue, 12 Jan 1999 05:46:16 +0000 (05:46 +0000)] 
make commResetFD() more like comm_open(); set outgoing address, etc.

26 years agofixed client_side cache-control max-age (henrik)
wessels [Tue, 12 Jan 1999 05:23:07 +0000 (05:23 +0000)] 
fixed client_side cache-control max-age (henrik)

26 years agouse EUC encoding
wessels [Tue, 12 Jan 1999 05:13:07 +0000 (05:13 +0000)] 
use EUC encoding

26 years agouse STDC_HEADERS instead of __STDC__ as recommend by autoconf documentation
wessels [Tue, 12 Jan 1999 04:55:33 +0000 (04:55 +0000)] 
use STDC_HEADERS instead of __STDC__ as recommend by autoconf documentation

26 years agoadding
wessels [Tue, 12 Jan 1999 04:32:16 +0000 (04:32 +0000)] 
adding

26 years agoportability problem for non-blocking connect() with sony newsos
wessels [Tue, 12 Jan 1999 04:29:46 +0000 (04:29 +0000)] 
portability problem for non-blocking connect() with sony newsos

26 years agoIRIX inline IO optimization ( -OPT:fast_io=ON) fubars this:
wessels [Tue, 12 Jan 1999 03:34:14 +0000 (03:34 +0000)] 
IRIX inline IO optimization ( -OPT:fast_io=ON) fubars this:

fprintf(fopen("foo", "w"), "%d", val);

26 years agocomment out unsupported FTP restart code
wessels [Tue, 12 Jan 1999 03:28:02 +0000 (03:28 +0000)] 
comment out unsupported FTP restart code

26 years agoadding
wessels [Tue, 12 Jan 1999 03:22:23 +0000 (03:22 +0000)] 
adding

26 years agosend old entry if IMS request results in 500's status
wessels [Tue, 12 Jan 1999 03:16:04 +0000 (03:16 +0000)] 
send old entry if IMS request results in 500's status

26 years ago// fix from Henrik
wessels [Tue, 12 Jan 1999 03:09:38 +0000 (03:09 +0000)] 
// fix from Henrik

26 years agoSmall change of Squid output for FTP
wessels [Tue, 12 Jan 1999 03:05:50 +0000 (03:05 +0000)] 
Small change of Squid output for FTP
Henrik Nordstrom
Andrew Filonov <aef@shu.smolensk.su>

26 years agoprint ENTRY_ABORTED flag
wessels [Tue, 12 Jan 1999 02:42:22 +0000 (02:42 +0000)] 
print ENTRY_ABORTED flag

26 years agoclear current_hdl_fd after each handler call
wessels [Tue, 12 Jan 1999 02:40:29 +0000 (02:40 +0000)] 
clear current_hdl_fd after each handler call

26 years agochanged store_status STORE_ABORTED to entry->flags ENTRY_ABORTED
wessels [Mon, 11 Jan 1999 23:50:27 +0000 (23:50 +0000)] 
changed store_status STORE_ABORTED to entry->flags ENTRY_ABORTED

26 years agoNPR bug
wessels [Sat, 9 Jan 1999 05:07:54 +0000 (05:07 +0000)] 
NPR bug

26 years agotypo
wessels [Sat, 9 Jan 1999 04:19:43 +0000 (04:19 +0000)] 
typo

26 years agoupdate
wessels [Sat, 9 Jan 1999 04:14:49 +0000 (04:14 +0000)] 
update

26 years agoMoved storeAbort() call to store_client.c, where it really belongs
wessels [Sat, 9 Jan 1999 04:12:06 +0000 (04:12 +0000)] 
Moved storeAbort() call to store_client.c, where it really belongs

26 years agoadded HDR_WAIT to storeEntryFlags() description
wessels [Sat, 9 Jan 1999 03:39:32 +0000 (03:39 +0000)] 
added HDR_WAIT to storeEntryFlags() description

26 years agofixed bug relating to searching ASN radix tree for INADDR_ANY.
wessels [Fri, 8 Jan 1999 05:13:35 +0000 (05:13 +0000)] 
fixed bug relating to searching ASN radix tree for INADDR_ANY.

26 years agodamn newline
wessels [Fri, 1 Jan 1999 06:50:25 +0000 (06:50 +0000)] 
damn newline

26 years agofwdReforward() needs to check pumpRestart() for PUMP methods
wessels [Wed, 30 Dec 1998 05:51:11 +0000 (05:51 +0000)] 
fwdReforward() needs to check pumpRestart() for PUMP methods

26 years agobetter debugging
wessels [Wed, 16 Dec 1998 13:34:55 +0000 (13:34 +0000)] 
better debugging

26 years agothese functions can be static
wessels [Wed, 16 Dec 1998 13:34:31 +0000 (13:34 +0000)] 
these functions can be static

26 years agofwd stats polishing
wessels [Wed, 16 Dec 1998 13:25:11 +0000 (13:25 +0000)] 
fwd stats polishing

26 years agoforwarding statistics
wessels [Wed, 16 Dec 1998 13:04:15 +0000 (13:04 +0000)] 
forwarding statistics

26 years agomore cases where it is wrong to call cbdataValid
wessels [Wed, 16 Dec 1998 12:38:57 +0000 (12:38 +0000)] 
more cases where it is wrong to call cbdataValid

26 years agoits wrong to assert cbdataValid(foo) when 'foo' is the callback data
wessels [Wed, 16 Dec 1998 12:21:37 +0000 (12:21 +0000)] 
its wrong to assert cbdataValid(foo) when 'foo' is the callback data
in the callback function itself.

26 years agomore peer / peer_digest locking fixes
wessels [Wed, 16 Dec 1998 07:07:17 +0000 (07:07 +0000)] 
more peer / peer_digest locking fixes

26 years agocleaning up peer and peer_digest interlocking goop
wessels [Wed, 16 Dec 1998 06:51:19 +0000 (06:51 +0000)] 
cleaning up peer and peer_digest interlocking goop

26 years agosigh, we don't really need to lock/unlock peer here
wessels [Wed, 16 Dec 1998 01:22:40 +0000 (01:22 +0000)] 
sigh, we don't really need to lock/unlock peer here

26 years agoforgot to call cbdataFree() on the peer!!
wessels [Wed, 16 Dec 1998 01:19:26 +0000 (01:19 +0000)] 
forgot to call cbdataFree() on the peer!!

26 years agostrip_query_terms option
wessels [Wed, 16 Dec 1998 00:33:58 +0000 (00:33 +0000)] 
strip_query_terms option

26 years agolookup FQDN if missed for access log
wessels [Wed, 16 Dec 1998 00:33:33 +0000 (00:33 +0000)] 
lookup FQDN if missed for access log

26 years ago- do not lock event cbdata
rousskov [Tue, 15 Dec 1998 06:45:15 +0000 (06:45 +0000)] 
- do not lock event cbdata