IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
In cli_echo with more than one response we ended up with more than one read_smb
request. One from the call to cli_smb_req_set_pending called from
cli_smb_received. The other one from cli_smb_received itself. I don't really
see another way to deal with this than to hold the read_smb request in the
cli_state.
Metze, please check!
Volker
We are here only if we have more than one num_pending
Autobuild-User: Volker Lendecke <vlendec@samba.org>
Autobuild-Date: Mon Jun 6 18:21:17 CEST 2011 on sn-devel-104
Several places want "milliseconds from current time", and several were
simply doing "msec * 1000" which can (and does in one place) result in
a usec value over 1 a million.
Using a helper to do this is safer and more readable.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
I have a test failure on my 32-bit Ubuntu system, in that
samba3.smbtorture_s3.plain(s3dc).LOCK9 immediately times out (rather than
waiting 5 seconds for the child).
Debugging revealed this code: timeout is in ms and is set to > 1000 in
various places. The code dates from 2002, and other perturbations didn't
reveal why it breaks now, but fix it anyway.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Autobuild-User: Rusty Russell <rusty@rustcorp.com.au>
Autobuild-Date: Tue May 10 12:09:07 CEST 2011 on sn-devel-104
When winbind sees a signing error on the smb connection to a DC (for whatever
reason, our bug, network glitch, etc) it should recover properly. The "old"
code in clientgen.c just closed the socket in this case. This is the right
thing to do, this connection is spoiled anyway. The new, async code did not do
this so far, which led to the code in winbindd_cm.c not detect that we need to
reconnect.
DOS error codes were being lost with the conversion to async
libsmbclient. If we're passing around NTSTATUS internally,
let's just convert it when we get it.
DOS ACCESS_DENIED on nautilus was not prompting for other credentials,
because it was not being mapped.
I've tried to solve this just within cli_smb_recv(), but I could not find a way
to sanely determine when we are receiving the last entry in the chain just from
looking at the blob. This solves it in an a bit more brutal way...
A socket where the other side has closed only becomes readable. To catch
errors early when sitting in a pure writev, we need to also test for
readability.
Metze is right: If we have *any* error at the socket level, we just can
not continue.
Also, apply some defensive programming: With this async stuff someone else
might already have closed the socket.
This way we can destinguish between requests which failed
because the connection broke after they were triggered
and the requests which are started on an already broken
connection.
This also moves the check to cli_smb_req_iov_send()
where it really belongs.
metze
I know this is just yet another iteration, but I like this one much better than
the one that exists right now :-)
It will do trans and echo requests without a _recv helper and without
unnecessary memcpy().
We store the seqnum/mid mapping in the cli_request structure
for async requests and in the cli_state structure for sync calls.
We skip the signing check for oplock requests while waiting
for async requests coming in.
metze
If it takes more than 10 seconds to understand the code you've written yourself
less than a year ago, it's time for comments or refactoring. I couldn't find a
way to refactor that cleanly, so add comments :-)
If we really want to keep the pipe busy, we need to write everything we have as
early as possible, giving the kernel the chance to get rid of the buffers
quickly :-)
For example open&x and write&x needs the bytes to be aligned relative to the
SMB header. In particular for write&x we should not have to move stuff around.
The async trans calls need this, as for secondary trans calls they have to
modify the MID from what cli_request_chain() gave us.
(This used to be commit c85de4b7b5)
This gives a hint whether a function is called from within the middle of a
chain. In particular the trans calls don't really like this.
(This used to be commit 4252b32db5)
Along the lines of cli_request_send this abstracts away the smb-level buffer
handling when parsing replies we got from the server.
(This used to be commit 253134d3aa)
cli_request_send() is supposed to bundle all generic SMB-header handling. This
makes cli_request_new static to async_smb.c.
(This used to be commit 7e73dd4e76)
dos errors are used and there is no error. It should
be mapped directly to NT_STATUS_OK. smbclient to older
servers didn't work.
(This used to be commit 78f009b7ef)
