If you click on a filename you can then see all the commits and
changes that have been made to it over time. You can also download any
version of the file or find the differences between any two versions.
All of this is not dependent on the state of your local CVS sandbox,
so it can be used to find out the "true" state of the tree at any
time.
If you suspect some sort of CVS problem (like a change getting
reverted) then please use the above URL to work out what has
happened. You should be able to see exactly who made what changes and
when.
Luke Leighton [Sun, 12 Oct 1997 14:17:55 +0000 (14:17 +0000)]
ipc.c:
debugging info. found that data = NULL because of short packet length
indicated from the ntlsaRPC pipe _royally_ stuffs NT's packet handling.
maybe this should go down as a service denial bug to the ntbugtraq list.
pipes.c lsaparse.c smbparse.c :
added more debug stuff. added length of header to data_len in MSRPC
fragment_length field (0x18 bytes short) which caused the above bug
from NT 4.0. oops.
Updated mirrors, combined Sources and Sources&Binaries as our binaries
are now with our sources.
Have requested that mirror sites remove the old 'binaries' tree.
Paul.
I have to admit that I'm not very good at writing Makefiles.
This one will simply create an object file from each of the .c files
in the directory. These could be linked into a static library, but
I'm not sure that this is worthwhile. If anyone wants to improved on
what I've got here, please do! crh -)-----
After generating some discussion, listening to the opinions, and thinking
about it for a while, I've decided to move the tree & list code into a
separate subdirectory.
Jeremy Allison [Fri, 10 Oct 1997 01:32:26 +0000 (01:32 +0000)]
nmblookup.c: Added -A ability to do status on ip address.
smb.h: Added defines we will need for NT SMB calls.
trans2.c: Fixed SMB_QUERY_FILE_ALT_NAME_INFO return - this is only
for short name returns (and only used when you negotiate NT SMB calls
to boot !).
Jeremy (jallison@whistle.com)
Jeremy Allison [Thu, 9 Oct 1997 18:40:52 +0000 (18:40 +0000)]
local.h: Fix spelling mistake :-).
namedbsubnet.c: Stop registering 1x name unless we can be a local master.
reply.c: Remove ERRbaddirectory code.
server.c: Remove abort() - use exit_server() instead.
trans2.c: Remove ERRbaddirectory code.
Jeremy (jallison@whistle.com)
This is my library of lists and trees. My hope is to replace all of the
hard coded linked lists that are currently used in Samba with calls to
these modules. This should make the code simpler, smaller, and (I hope)
faster. The tree code, in particular, should speed up processing where
large lists are involved.
Luke Leighton [Wed, 8 Oct 1997 21:36:42 +0000 (21:36 +0000)]
added a dummy function (space in front so make proto doesn't pick it up)
to call the unused static functions in pipes.c. avoids need to move
#if UNDEFINED_NTDOMAIN about, to stop compiler warnings while code is
being developed, but might be released as-is.
Jeremy Allison [Wed, 8 Oct 1997 20:34:13 +0000 (20:34 +0000)]
Makefile: Split definitions for SGI4,5,6.
includes.h: Split definitions for SGI4,5,6.
pipes.c: Moved Luke's #ifdef to remove warnings.
quotas.c: Two changes for FreeBSD and SGI.
server.c: Quota changes for large filesystems.
Jeremy (jallison@whistle.com)
Jeremy Allison [Wed, 8 Oct 1997 00:21:39 +0000 (00:21 +0000)]
ipc.c: Added ERROR_MORE_DATA error if client buffer too small.
server.c: Allow admin_user on read only shares. I think this is
safe but it needs looking at.
Jeremy (jallison@whistle.com)
I rewrote the name_mangle() and name_len().
The interface is the same, but I've simplified the code a bit and added
a few comments. I tested the changes and the output is the same as that
produced by the original (at least, it was in all of my tests). CRH
Jeremy Allison [Tue, 7 Oct 1997 18:46:19 +0000 (18:46 +0000)]
locking.c: Added fix for race condition in slow share mode code.
lsaparse.c: #ifdef'ed out code so this will compile - LUKE PLEASE CHECK THIS.
pipes.c: #ifdef'ed out code so this will compile - LUKE PLEASE CHECK THIS.
server.c: Fixed last known oplock race condition.
smb.h: Re-removed USE_OPLOCK defines - someone checked in an old version.
smbparse.c: #ifdef'ed out code so this will compile - LUKE PLEASE CHECK THIS.
Jeremy (jallison@whistle.com)
Jeremy Allison [Mon, 6 Oct 1997 19:50:51 +0000 (19:50 +0000)]
Finally ! Found & fixed crash bug with logging message when deleting
invalid share mode entries. Thanks to berg@wienrg.aut.alcatel.at (Dietmar Berg)
for the stack backtrace.
Jeremy (jallison@whistle.com)
Jeremy Allison [Mon, 6 Oct 1997 17:52:25 +0000 (17:52 +0000)]
client.c: Changed shadowed variable.
locking.c: Removed USE_OPLOCKS - now the default.
params.c: Removed unused variable.
proto.h: Updated.
reply.c: Removed USE_OPLOCKS - now the default.
server.c: Removed USE_OPLOCKS - now the default.
smb.h: Removed USE_OPLOCKS - now the default.
smbparse.c: Changed shadowed variable.
status.c: Removed USE_OPLOCKS - now the default.
util.c: Removed USE_OPLOCKS - now the default.
Jeremy (jallison@whistle.com)
Luke Leighton [Sun, 5 Oct 1997 11:56:08 +0000 (11:56 +0000)]
smb.h:
added more structures, this time for the use in the query and response
of the LSA_XXXX functions. next target: the NTLOGON mailslots (10 minutes
work).
smbparse.c:
tidying. adding some more sub-structure functions.
Luke Leighton [Sat, 4 Oct 1997 16:51:43 +0000 (16:51 +0000)]
proto.h:
- recreated, as usual.
smb.h:
- added RPC_HDR structure - the 18 byte MSRPC header
smbparse.c:
- added smb_io_rpc_hdr() function to read/write the RPC_HDR structure.
util.c:
- added align2, align4, align_offset functions.
- added skip_unicode_string, unistrcpy, unistrncpy functions.
- modified unistrcpy and unistrncpy to return the number of unicode
characters returned, effectively making skip_unicode_string redundant.
nmblookup.c:
I've added a -r option. When specified, nmblookup will attempt to bind
to port 137. In order for this to work, the user must be root and the
port must be available (i.e., nmbd not running). If either condition is
not met, nmblookup will default to normal behavior; it will try to bind
to a free port above 7999. I also removed an artifact. The -p <port>
option was being accepted but falling through to the default, which was
the usage message. I removed "p:" from the getopt() list.
params.c:
Rewritten. The scanning is much more linear, and I found and fixed a
few bugs. I don't like the way that the scratch buffer is handled,
but I believe that it is an improvement over the previous version.
Jeremy Allison [Sat, 4 Oct 1997 00:23:57 +0000 (00:23 +0000)]
Race condition with multiple oplock break requests happens
more often than you might think (reproduced here with 4 clients
and netbench :-).
Raising debug log level for report from 1 to 3.
Jeremy (jallison@whistle.com)
Jeremy Allison [Fri, 3 Oct 1997 20:36:06 +0000 (20:36 +0000)]
locking.c: Fixed incorrect parameter count in debug statements. May explain
solaris crashes.
reply.c: Added NT specific error code. Put oplock break code in correct place
in reply_lockingX.
server.c: Removed unneeded error mapping stuff.
Fixed race condition in oplock code.
trans2.c: Added NT specific error code.
util.c: Added paranoia check in interpret_addr. Some core dumps
reported here. Upped fcntl debug levels.
Andrew. Please check the NT specific error code handling (search
for the string "/* Ugly - NT specific hack - but needed (JRA) */",
this makes NT and 95 clients behave correctly here - please check
your Visual Basic apps with this code.
Jeremy Allison [Fri, 3 Oct 1997 03:15:24 +0000 (03:15 +0000)]
proto.h: Updated.
server.c: Updated after netbench observation. Oplocks must be
broken *before* share modes are checked, not after. Netbench seems
to be working now.
smb.h: Added offsets for oplock break time fields.
trans2.c: Upped debug messages.
util.c: Upped debug messages.
Jeremy (jallison@whistle.com)
Andrew Tridgell [Thu, 2 Oct 1997 14:11:34 +0000 (14:11 +0000)]
change a debug level in reply.c
change from ERRbaddirectory to ERRbadpath for ENOTDIR errors. This
reverts to the old Samba code. I've done quite a bit of testing
against NT4 and have yet to get it to produce the ERRbaddirectory
error code. Producing ERRbaddirectory made a visual basic application
that was sent to me not run. This might explain some of the "it
doesn't work any more" complaints we've got about 1.9.17.
Jeremy, can you remember how you got NT to produce ERRbaddirectory?
There might be some specific circumstances we need to cover.
Andrew Tridgell [Thu, 2 Oct 1997 03:14:32 +0000 (03:14 +0000)]
change the semantics of hosts allow/hosts deny so that a global
setting applies to all shares regardless of any settings on other
shares. This allows us to immediately drop a connection if it does not
come from a allowed host, without even parsing the first SMB
packet. The next time we get a nasty security hole we can offer people
the option of just setting their hosts allow line.
If we drop a connection in this way we generate a "Not listening for
calling name" response and then exit.
add a per share "oplocks" option in smb.conf. I think its important to
be able to disable oplocks on a per-share basis as there are occasions
then they are definately not wanted, for example when sharing data
between a windows box and a unix application. This also allows us to
tell people "try disabling oplocks" when diagnosing problems.
fix a bug in process_smb(). It was taking the length of the packet
from outbuf, not inbuf (this bug was introduced with the oplocks
code). Jeremy, I assume this wasn't deliberate?
Jeremy Allison [Wed, 1 Oct 1997 23:32:22 +0000 (23:32 +0000)]
OPLOCK CHECK-IN - oplocks are now *OPERATIONAL* !!!!
Yipeee. At least as far as I can check in a short time :-).
local.h: Changed OPLOCK_BREAK_TIMEOUT to 30 seconds.
locking.c: Big changes to delete oplocks on a share mode entry.
proto.h: updated.
reply.c: Added oplock break code in lockingX reply & readbraw reply.
server.c: Add batch oplock code. Force server shutdown if client fails
to respond to oplock break.
smb.h: Fix silly slow share mode oplock define bug.
status.c: Add oplock status info.