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.
Make the real data dependencies clearer, we don't need a full
share_entry, only its access and sharemode masks
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
It was hard for me to understand share_conflict(), so once I understood
it I thought this version would be easier to follow. It violates
README.Coding (one argument per line), but grouping the parameters to
mask_conflict makes it clearer to me what belongs together.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Less bytes .text, the compiler now does not have to expect other threads to
alter the pointers.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Initialize the File-ID in fsp->fsp_name->st, any subsequent metadata fetch on
this file-handle needs this, eg QFID SMB2 Create-Context or GETINFO SMB
requests.
It would be nice if SMB_VFS_SET_DOS_ATTRIBUTE() would do this, unfortunately it
gets a const struct smb_filename.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14121
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Walk the share_modes array only once.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Tue Aug 20 00:33:12 UTC 2019 on sn-devel-184
This is a preparation for the next commit: Only walk the share_modes[]
array once when handling oplocks and leases.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
We'll merge grant_fsp_oplock_type() into this function. This makes the next
commit smaller, and the newly extended delay_for_oplocks will for example
reference static file_has_brlocks() above. Make forward declarations
unnecessary.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
A further commit will merge grant_fsp_oplock_type (returning NTSTATUS)
into delay_for_oplock(). Make that commit smaller.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This is never called for INTERNAL_OPENs anymore, see
handle_share_mode_lease()
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This consolidates the core share_mode_lock access of open_file_ntcreate
into one routine.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This shows that "req", "share_access" and "access_mask" are not needed
for the core logic of grant_fsp_oplock_type() and it separates
concerns a bit: open_directory() also does the set_share_mode() in the
main open routine, not in a helper like grant_fsp_oplock_type()
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Note that this is not a cut&paste: Instead of fsp->access_mask we use
the access_mask the client requested. At the new code location
fsp->access_mask (a.k.a. open_access_mask) might have FILE_WRITE_DATA
from O_TRUNC (a.k.a. FILE_OVERWRITE).
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
There is no reference to "file_existed" after this point anymore
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
If you follow "existing_dos_attributes" through the routine, this can
only ever be !=0 if SMB_VFS_GET_DOS_ATTRIBUTES() was successful. This
can only have been successful if the file existed.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
It seems to make little sense to me to do the oplock break with one
setting and then later on grant_fsp_oplock_type with another
one. Survives tests, I can't think of any scenario where this (to me)
simplification would break anything
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
The kernel has already sent the signal. We don't have to send another
message. Instead, just directly wait for the downgrade to happen via
the watch on the share mode lock assuming it's there. Also setup the
polling interval: I could imagine that in some race situation the file
has already been closed and re-opened by a nonsamba process while we
were waiting.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This will be used when waiting for a oplock break that has been
signalled via a kernel oplock break.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
get_lease_type() can involve a database access. Do that only if
necessary, and that is at most once in this loop.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Böhme <slow@samba.org>
* Assign *file_created on every exit.
* Directly assign curr_flags without &= / |=
Both of these changes make the routine easier to understand for me,
less jumping around in the code to see where the values came from.
* Do the retry in a "positive" if-clause
Normally I'm a big fan of early returns, but this single retry is so
simple that to me it's easier to understand this way.
Overall, 13 lines less code. YMMV :-)
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Sat Aug 10 00:07:28 UTC 2019 on sn-devel-184
I don't really have a test case, but to me a positive test for a
regular file makes more sense here than just ruling out FIFOs. While
we probably only ever hit regular files (or FIFOs), there might be
more that we catch and don't properly handle.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This is the one place where *lease actually got modified. We can
easily make a copy, "struct smb2_lease" is not too large, and this
case is pretty rare anyway.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Simplify the flow in open_file_ntcreate, streamline it for SMB2
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This is close to what Windows SMB1 does: Instead of waiting for the
share entry causing the SHARING_VIOLATION to disappear, retry every
200msec up to one second. Windows does it a little differently: Retry
up to 5 times. But up to one second should be close enough.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This is necessary for the following case:
We want to delete a file with an open stream that is not open with
FILE_SHARE_DELETE. In SMB1, we need to defer the sharing violation
reply (we don't do that right now, test to follow). However, when we
move that sharing violation delay to where it belongs, into the outer
layers, only very deep in the nested open_streams_for_delete smb1
sharing violation delay handling call we will hit the sharing
violation in the 1-second retry case. However, that
open_streams_for_delete itself is INTERNAL_OPEN_ONLY and thus not
deferred itself. This means that it will not overwrite
req->request_time at all.
Exec summary: We only have one request_time now, set it properly as
early as possible.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This will become important in the next commits when the SMB1 sharing
violation delay will use this. We want to be able to reduce the
timeout to less than 200msec, see the next commits.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This is for callers who don't want to call open_was_deferred()
afterwards
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This is a generic "open retry without locking.tdb waiter" loop: Take
the specific timeouts as parameters.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Failure to postpone a request is not really fatal: We just don't retry
as wanted but return an error to the client that might have resolved
itself after a few seconds. From my point of view such a spurious and
rare error, which is highly unlikely anyway does not justify to kill
that client's connection.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This piece of code predates our user-space access checks, which we
nowadays always do in open_file()
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Slightly simplify assumptions in the code
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Tue Aug 6 23:06:41 UTC 2019 on sn-devel-184
This hides a talloc off the NULL context: The caller needs to make sure
this is put on a real talloc context later. Make that more
obvious. Also, it passes down a boolean flag, making its purpose a bit
opaque to the caller sites.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
The fsp carries all required information also for leases. There's no
need to pass that as additional parameters
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Nobody used this (except vfs_gpfs, which did not need it really). If
you *really* need this, you can always look in locking.tdb, but this
should never happen in any hot code path, as no runtime decisions are
made on the share access after the open is done.
Bump VFS interface number to 42.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Why? While restructuring open_file_ntcreate() I found the data flow for
these values confusing: grant_fsp_oplock_type() depends on
fsp->access_mask, which changes its value inside
open_file_ntcreate(). I find the data flow easier to follow if it
happens in explicit variables.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Makes the interface more obvious to me. Also, I want to remove
fsp->share_access, which is not really used anywhere after the fsp has
been fully established.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
For whatever reason, "st_size" in "struct stat" is an off_t, which is a
signed integer. Negative sizes don't really make sense, so this cast
should be safe.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
In the 2nd for-loop we need a signed int as we are comparing to >=0.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
It seems more natural to pass in a request to a routine called
request_timed_out(), and it's a few bytes less .text
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
We have "smb_request->request_time" that is already set up by
init_smb_request().
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This now removed comment describes the bug correctly:
/*
* As this timer event is owned by req, it will
* disappear if req it talloc_freed.
*/
In smb1, "req" disappears once the reply_whatever routine is done. Thus
the timer goes away and we never look at "req" again.
This change moves the valid data (xconn and mid) to
deferred_open_record, and changes the talloc hierarchy such that the
timer is now a child of open_rec, which is a child of the deferred
message.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=14060
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Wed Jul 31 00:12:34 UTC 2019 on sn-devel-184
It's perfectly possible that someone else takes a kernel oplock and
makes us block, independent of our own kernel oplock setting.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=14060
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This is SMB1-only and pre-ntcreate with only 3 callers that look at
NTCREATEX_OPTIONS_PRIVATE_DENY_[DOS|FCB]. It is a bit less efficient
if it kicks in (we have to recreate the fsp), but SMB1 is less and
less popular, and this particular share mode combination from the
open&x family of calls might not be worth optimizing for.
This adds smb1_utils.[ch] as a kitchen sink for functions that can go
away once we drop SMB1.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
If there are no share modes, we'll just not enter the loop.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
This removes a kludgy implementation that worked around a locking
hierarchy problem: Setting a byte range lock had to contend the level2
oplocks, which are stored in locking.tdb/leases.tdb. We could not
access locking.tdb in the brlock.tdb code, as brlock.tdb might have
been locked first without locking.tdb, violating the locking hierarchy
locking.tdb->brlock.tdb. Now that that problem is gone (see the commit
wrapping do_lock() in share_mode_do_locked()), we can remove this
kludge.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Lazy update of the flag: Whenever we add a read lease, we have to set
the flag. Nobody except contend_level2_oplocks_begin will remove that
flag again, as this would mean a full lease traverse when removing
one. And contend_level2_oplocks_begin traverses the leases anyway
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
As we never use kernel oplocks on directory handles,
there is no reason not to always open file descriptors (no
more "stat" opens on directories).
Preparing to have SMB1search use real directory
opens.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Change test to check two things:
1) Open a symlink for SD read or write access should fail.
2) Request attribute open. Getsd/Setsd on this handle should
fail.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
We don't need a share mode lock from a data dependency point of view
anymore, the leases data moved to leases.tdb. However, from a
coherency point of view it's probably wise to do this under a share
mode lock.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
"const" ist just a hint to make sure it's actually not modified inside
the loop
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Opening a file with a stale (smbd died) LEVEL_II oplock makes
vfs_set_filelen-> ... ->contend_level2_oplocks_begin_default
trigger the immediate leading to do_break_to_none. This goes through
because fsp->oplock_type is not initialized yet, thus 0. Also,
file_has_read_oplocks is still valid, because the smbd that has died
could not clean up the brlock.tdb entry.
Later in the code the exclusive oplock is granted, which is then found
by do_break_to_none, making it panic.
This patch just runs the direct FTRUNCATE instead of vfs_set_filelen.
This means the contend_level2_oplock code is skipped.
The relevant break (LEVEL_II to NONE) is now done in delay_for_oplock()
with the nice effect of removing a comment that was very confusing to
me.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=13957
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Wed May 22 20:09:29 UTC 2019 on sn-devel-184
file_id plus share_file_id remotely specify the fsp. This avoids the
explicit loop in the receiver.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Sat May 18 20:18:55 UTC 2019 on sn-devel-184
The previous scheme was overloaded, a idl definition is easier to
print, and it clarifies what data is actually needed
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Thu May 16 23:48:18 UTC 2019 on sn-devel-184
delay_for_batch_oplocks() is no more. Also, open_mode_check (which
calls into this routine) is called before delay_for_oplock.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
We don't have to look at the leases.tdb record if it's our own lease.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Thu May 16 07:59:52 UTC 2019 on sn-devel-184
This also removes the temporary functions introduced during the patchset.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
Autobuild-User(master): Christof Schmitt <cs@samba.org>
Autobuild-Date(master): Sun Apr 14 05:18:14 UTC 2019 on sn-devel-144
Simple refactoring into simpler routines. View best with "git show -b"
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
This is an interim function supposed to be around for just a few patches as
long as we have both the leases.tdb entries and the leases[] in
share_mode_entries around. It makes it easier to transition to just use
leases.tdb while keeping the code running.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
The only reason for grant_fsp_lease to return the lease_idx was to pass it down
to set_share_mode. That does not need it anymore.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
Temporary patch to keep the code running. The new code in set_share_mode() will
leave again once the patchset to remove share_mode_lease and thus the lease_idx
in share_mode_entry goes away.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
This was the last "share_mode_lease" reference in this function, remove
variable "l".
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
Whenever we update the share_mode_lease struct, also update the leases.tdb
entry.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
This is the data stored in share_mode_lease inside the leases[] array in
locking.tdb. This and all the following patches move all leases array to
looking at the leases.tdb.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
Why? I am preparing a patchset that will remove
"share_mode_lease". This patch is a micro-step towards that, removing
a set of references to this struct.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
grant_fsp_oplock_type has enough complex logic, make this a bit shorter
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
check_parent_access() currently leaks a number of allocations onto the
talloc_tos() context in both success and error paths.
Signed-off-by: David Disseldorp <ddiss@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Autobuild-User(master): David Disseldorp <ddiss@samba.org>
Autobuild-Date(master): Fri Mar 15 11:32:04 UTC 2019 on sn-devel-144
All existing callers pass NULL, no change in behaviour.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=13455
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
MS-FSA states that a CREATE with FILE_DELETE_ON_CLOSE on an existing
file with READ_ONLY attribute has to return STATUS_CANNOT_DELETE. This
was missing in smbd as the check used the DOS attributes from the CREATE
instead of the DOS attributes on the existing file.
We need to handle the new file and existing file cases separately.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13673
Signed-off-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
We used %llu as conversion specifier which results in a decimal number
being printed, so remove the misleading "0x" prefix.
While at it, I'll change %llu to the terse %ju.
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Thu Oct 25 21:44:17 CEST 2018 on sn-devel-144
This has been implicitly initialized to 0 with the explicit struct
initializer above.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
In future this will an impersonation wrapper tevent_context based on the
user session.
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
This will be used to mark basefile opens of streams opens. This is
needed to later implement a function that can determine if a file has
stream opens.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=13451
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This is a behavior change, it will modify the POSIX ACL mask
from a value of rwx instead of modifying the existing ACE
entries to be ANDed with the passed in mode. However it
will have no effect on the underlying permissions, and
better reflects the proper use of POSIX ACLs (i.e. I
didn't understand the use of the mask entry in the
ACL when I first wrote the POSIX ACL code).
In addition, the vfs_acl_common.c module already
filters these calls for all but POSIX opens, which
means the only place this change is exposed to the
client would be a cifsfs unix extensions client doing
posix acl calls (and they would expect the mask to
be set like this on chmod).
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
We have potentially called SMB_VFS_FCHMOD() here in
the file_set_dosmode() call associated with the comment
/* Overwritten files should be initially set as archive */
at line 3755 above, so there is no need to do any POSIX ACL
mask protection.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Signed-off-by: Timur I. Bakeyev <timur@iXsystems.com>
Reviewed-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Martin Schwenke <martin@meltin.net>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Wed May 16 21:29:24 CEST 2018 on sn-devel-144
This implements a check to test the delete-on-close flag of a directory
for requests to create files in this directory.
Windows server implement this check, Samba doesn't as it has performance
implications.
This commit implements the check and a new option to control it. By
default the check is skipped, setting "check parent directory delete on
close = yes" enables it.
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Sat Feb 3 23:42:16 CET 2018 on sn-devel-144
The initial idea was to have some "atomicity" in this API. Every
caller interested in a record would have to do something with
it once it changes. However, only one caller really used this
feature, and that is easily changed to not use it. So
remove the complexity.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
If a O_NONBLOCK open fails with EWOULDBLOCK, this code changes smbd to
do a retry open every second, until either the timeout or we get a successful
open. If we're opening a file that has a kernel lease set by a non-smbd
process, this is the best we can do.
Prior to this, smbd would block on the second open on such a leased file
(not using O_NONBLOCK) which freezes active clients.
Regression test to follow.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13121
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
This reverts commit b35a296a27.
This was the cause of
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13058
1. client of smbd-1 opens the file and sets the oplock.
2. client of smbd-2 tries to open the file. open() fails(EAGAIN) and open is deferred.
3. client of smbd-1 sends oplock break request to the client.
4. client of smbd-1 closes the file.
5. client of smbd-1 opens the file and sets the oplock.
6. client of smbd-2 calls defer_open_done(), sees that the file lease was not changed
and does not reschedule open.
and is no longer needed now vfs_streams_xattr.c no longer opens
the base file internally.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Böhme <slow@samba.org>
If the caller passes O_DIRECTORY we just try to chdir() to smb_fname
directly, not to the parent directory.
The security check in check_reduced_name() will continue to work, but
this fixes the case of an open() for a previous version of a
subdirectory that contains snapshopt.
Eg:
[share]
path = /shares/test
vfs objects = shadow_copy2
shadow:snapdir = .snapshots
shadow:snapdirseverywhere = yes
Directory tree with fake snapshots:
$ tree -a /shares/test/
/shares/test/
├── dir
│ ├── file
│ └── .snapshots
│ └── @GMT-2017.07.04-04.30.12
│ └── file
├── dir2
│ └── file
├── file
├── .snapshots
│ └── @GMT-2001.01.01-00.00.00
│ ├── dir2
│ │ └── file
│ └── file
└── testfsctl.dat
./bin/smbclient -U slow%x //localhost/share -c 'ls @GMT-2017.07.04-04.30.12/dir/*'
NT_STATUS_OBJECT_NAME_NOT_FOUND listing \@GMT-2017.07.04-04.30.12\dir\*
Bug: https://bugzilla.samba.org/show_bug.cgi?id=12885
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
We need to migrate all pathname based VFS calls to use a struct
to finish modernising the VFS with extra timestamp and flags parameters.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Richard Sharpe <realrichardsharpe@gmail.com>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Sat Jul 1 07:20:28 CEST 2017 on sn-devel-144
We need to migrate all pathname based VFS calls to use a struct
to finish modernising the VFS with extra timestamp and flags parameters.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Richard Sharpe <realrichardsharpe@gmail.com>
We need to migrate all pathname based VFS calls to use a struct
to finish modernising the VFS with extra timestamp and flags parameters.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Richard Sharpe <realrichardsharpe@gmail.com>
We need to migrate all pathname based VFS calls to use a struct
to finish modernising the VFS with extra timestamp and flags parameters.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Richard Sharpe <realrichardsharpe@gmail.com>
The errno returned by open() is ambiguous when called with flags O_NOFOLLOW and
O_DIRECTORY on a symlink. With ELOOP, we know for certain that we've tried to
open a symlink. With ENOTDIR, we might have hit a symlink, and need to perform
further checks to be sure. Adjust non_widelink_open() accordingly. This fixes
a regression where symlinks to directories within the same share were no
longer followed for some call paths on systems returning ENOTDIR in the above
case.
Also remove the knownfail added in previous commit.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=12860
Signed-off-by: Daniel Kobras <d.kobras@science-computing.de>
Reviewed-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
We need to migrate all pathname based VFS calls to use a struct
to finish modernising the VFS with extra timestamp and flags parameters.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Richard Sharpe <realrichardsharpe@gmail.com>
We need to migrate all pathname based VFS calls to use a struct
to finish modernising the VFS with extra timestamp and flags parameters.
Requires a few extra cleanups in calling code.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Add an extra paramter to cwd_name to check_reduced_name().
If cwd_name == NULL then fname is a client given path relative
to the root path of the share.
If cwd_name != NULL then fname is a client given path relative
to cwd_name. cwd_name is relative to the root path of the share.
Not yet used, logic added in the next commit.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=12721
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
BUG: https://bugzilla.samba.org/show_bug.cgi?id=12496
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Uri Simchoni <uri@samba.org>
Autobuild-User(master): Karolin Seeger <kseeger@samba.org>
Autobuild-Date(master): Thu Mar 23 22:55:04 CET 2017 on sn-devel-144
I noticed smbd can get stuck in an open() call with kernel oplocks
enabled and named streams (provided by vfs_streams_xattr):
- client opens a file and with an exclusive oplock
- client starts writing to the file
- client opens an existing stream of the file
- the smbd process gets stuck in an open()
What happens is:
we had setup a locking.tdb record watch in defer_open(), the watch was
triggered, we reattempted the open and got stuck in a blocking open
because the oplock holder (ourselves) hadn't given up the oplock yet.
Cf e576bf5310 for the commit that added
the kernel oplock retry logic. tldr: with kernel oplocks the first open
is non-blocking, but the second one is blocking.
Detailed analysis follows.
When opening a named stream of a file, Samba internally opens the
underlying "base" file first. This internal open of the basefile suceeds
and does *not* trigger an oplock break (because it is an internal open
that doesn't call open() at all) but it is added as an entry to the
locking.tdb record of the file.
Next, the stream open ends up in streams_xattr where a non-blocking
open() on the base file is called. This open fails with EWOULDBLOCK
because we have another fd with a kernel oplock on the file.
So we call defer_open() which sets up a watch on the locking.tdb record.
In the subsequent error unwinding code in open_file_ntcreate() and
callers we close the internal open file handle of the basefile which
also removes the entry from the locking.tdb record and so *changes the
record*.
This fires the record watch and in the callback defer_open_done() we
don't check whether the condition (oplock gone) we're interested in is
actually met. The callback blindly reschedules the open request with
schedule_deferred_open_message_smb().
schedule_deferred_open_message_smb() schedules an immediate tevent event
which has precedence over the IPC fd events in messaging, so the open is
always (!) reattempted before processing the oplock break message.
As explained above, this second open will be a blocking one so we get
stuck in a blocking open.
It doesn't help to make all opens non-blocking, that would just result
in a busy loop failing the open, as we never process the oplock break
message (remember, schedule_deferred_open_message_smb() used immediate
tevent events).
To fix this we must add some logic to the record watch callback to check
whether the record watch was done for a kernel oplock file and if yes,
check if the oplock state changed. If not, simply reschedule the
deferred open and keep waiting.
This logic is only needed for kernel oplocks, not for Samba-level
oplocks, because there's no risk of deadlocking, the worst that can
happen is a rescheduled open that fails again in the oplock checks and
gets deferred again.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=7537
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
No change in behaviour. Update the function comment explaining how it
works and relies on lck for a record watch.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=7537
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
schedule_async_open() was calling defer_open with sharemode lock = NULL,
as a result there was never an active 20 s timeout.
This has been broken since the commits in
$ git log --reverse -p -10 8283fd0e00
Just roll our own deferred record instead of calling defer_open() and
also set up timer that, as a last resort, catches stuck opens and just
exits for now.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=7537
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Add a new function that does an immediate open rescheduling.
The first deferred open this commit changes was never scheduled, as the
scheduling relies on a timeout of the watch on the sharemode lock.
This has been broken since the commits in
$ git log --reverse -p -10 8283fd0e00
That patchset added the dbwrap watch record logic to defer_open() and
removed the timers.
I'm doing this mainly to untangle the defer_open() logic which is
complicated by the lck arg.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=7537
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Add a helper function deferred_open_record_create() that creates a
deferred_open_record and let all callers pass all needed arguments
individually.
While we're at it, enhance the debug message in defer_open() to print
all variables.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=7537
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
req can't be NULL because the if condition surrounding this code checks
!(oplock_request & INTERNAL_OPEN_ONLY).
Bug: https://bugzilla.samba.org/show_bug.cgi?id=7537
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Signed-off-by: Chris Lamb <chris@chris-lamb.co.uk>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
In the FILE_OPEN_IF case we have O_CREAT, but not
O_EXCL. Previously we went into a loop trying first
~(O_CREAT|O_EXCL), and if that returned ENOENT
try (O_CREAT|O_EXCL). We kept looping indefinately
until we got an error, or the file was created or
opened.
The big problem here is dangling symlinks. Opening
without O_NOFOLLOW means both bad symlink
and missing path return -1, ENOENT from open(). As POSIX
is pathname based it's not possible to tell
the difference between these two cases in a
non-racy way, so change to try only two attempts before
giving up.
We don't have this problem for the O_NOFOLLOW
case as we just return NT_STATUS_OBJECT_PATH_NOT_FOUND
mapped from the ELOOP POSIX error and immediately
returned.
Unroll the loop logic to two tries instead.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=12572
Pair-programmed-with: Ralph Boehme <slow@samba.org>
Signed-off-by: Jeremy Allison <jra@samba.org>
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
This can be used to emulate folder quotas, as explained in the
modified manpage.
Signed-off-by: Uri Simchoni <uri@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
When using UNIX extensions to delete a file containing streams,
the open for delete and close operations need to enumerate the
contained streams and do CREATE and UNLINK operations on the
stream names. These must always be done as Windows operations
(remove the SMB_FILENAME_POSIX_PATH flag) as the stream names
are Windows paths.
Without this the create operation under the unlink will
recurse and cause the client to time out (or a server crash).
BUG: https://bugzilla.samba.org/show_bug.cgi?id=12021
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Typicall, when we watch a record, we wait for a process to give up some
resource. Be it an oplock, a share mode or the g_lock. If everything goes well,
the blocker sends us a message. If the blocker dies hard, we want to also be
informed immediately.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This changes the way we check for old vs new DOS attributes on open with
overwrite: only check against the DOS attributes actually set by a
client and stored in the DOS attributes xattr.
With this change "hide dot files" and "hide files" continue to work with
"store dos attributes = yes".
Bug: https://bugzilla.samba.org/show_bug.cgi?id=11992
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Tue Jun 21 22:22:03 CEST 2016 on sn-devel-144
This avoids errors due to 'not implemented' for SMB_VFS_KERNEL_FLOCK
on some file systems like glusterfs (with the vfs module). The only
other code path where SMB_VFS_KERNEL_FLOCK is called, is already protected.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=11919
Signed-off-by: Michael Adam <obnox@samba.org>
Reviewed-by: Christian Ambach <ambi@samba.org>
Autobuild-User(master): Christian Ambach <ambi@samba.org>
Autobuild-Date(master): Thu May 19 02:34:36 CEST 2016 on sn-devel-144
Get it from parent/deriving smb_filename if present.
Use 0 (as usually this a Windows-style lookup) if
not.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Uri Simchoni <uri@samba.org>
This is needed because the smb2.create.mkdir-dup test creates a race,
and against an AD DC this can cause a flapping test if the lstat() and
stat() calls are made either side of the chown() due to creation of a
file by administrator.
Fix based on original patches by myself, by Douglas Bagnall
<douglas.bagnall@catalyst.net.nz>. and Jeremy Allison <jra@samba.org>
BUG: https://bugzilla.samba.org/show_bug.cgi?id=11780
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Sat Mar 12 09:43:21 CET 2016 on sn-devel-144
Prepare for changing vfs_streaminfo to do the same.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Mon Mar 7 21:12:56 CET 2016 on sn-devel-144
Preparing to reduce use of lp_posix_pathnames().
Uses the same techniques as commit 616d068f0c
(synthetic_smb_fname()) to cope with modules that
modify the incoming pathname.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Uri Simchoni <uri@samba.org>
Autobuild-User(master): Uri Simchoni <uri@samba.org>
Autobuild-Date(master): Wed Feb 24 16:05:55 CET 2016 on sn-devel-144
Bumps VFS version to 35.
Preparing to reduce use of lp_posix_pathnames().
Most of this is boilerplate, the only subtleties are in
the modules:
vfs_catia.c
vfs_media_harmony.c
vfs_shadow_copy2.c
vfs_unityed_media.c
Where the path is modified then passed to SMB_VFS_NEXT_GET_NT_ACL().
In these cases the change uses synthetic_smb_fname() to
create a new struct smb_filename from the modified path.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <rb@sernet.de>
Signed-off-by: Jose A. Rivera <jarrpa@samba.org>
Pair-programmed-with: Ira Cooper <ira@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Fri Feb 5 04:37:43 CET 2016 on sn-devel-144
We reject directory creation with an initial allocation size > 0 with
NT_STATUS_ACCESS_DENIED. Windows servers ignore the initial allocation
size on directories.
Bug: https://bugzilla.samba.org/show_bug.cgi?id=11684
Pair-Programmed-With: Volker Lendecke <vl@samba.org>
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Start using FSP_POSIX_FLAGS_PATHNAMES instead of the kitchen sink
FSP_POSIX_FLAGS_OPEN.
Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Wed Dec 23 10:37:07 CET 2015 on sn-devel-144
This will allow us to move lp_posix_pathnames() out of unix_convert()
later.
Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <rb@sernet.de>
Fixes POSIX rename problem introduced in d698cec1c7
Signed-off-by: Ralph Boehme <rb@sernet.de>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Mon Dec 14 02:03:12 CET 2015 on sn-devel-104
