Martin Schwenke [Wed, 15 Oct 2025 23:42:22 +0000 (10:42 +1100)]
ctdb-tests: Add an event script unit test for updateip
This illustrates the current failure where an unassigned public IP
address causes updateip to fail.
After commit 0536d7a98b832fc00d26b09c26bf14fb63dbf5fb (which improves
IP address state checking), this will almost certainly not occur on
platforms with getifaddrs(3) (e.g. Linux). This means it is only
likely to occur in 4.21 when net.ipv4.ip_nonlocal_bind=1.
Martin Schwenke [Wed, 15 Oct 2025 21:17:44 +0000 (08:17 +1100)]
ctdb-daemon: Fix a crash due to a failed updateip
This should really be a takeip. However, CTDB's weak check of the IP
address state (using bind(2)) incorrectly indicates that the IP
address is assigned to an interface so it is converted to an updateip.
After commit 0536d7a98b832fc00d26b09c26bf14fb63dbf5fb (which improves
IP address state checking), this will almost certainly not occur on
platforms with getifaddrs(3) (e.g. Linux). This means it is only
likely to occur in 4.21 when net.ipv4.ip_nonlocal_bind=1.
This commit fixes a situation in which vfs_streams_xattr could
write unitialized memory into alternate data streams if the
user writes to an offset that is beyond the current end of file
to insert a hole in it.
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Stefan Metzmacher <metze@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Thu Oct 16 18:42:05 UTC 2025 on atb-devel-224
If AddressSanitizer is loaded it will detect alignment issues. The tool
is not ours, so ignore it.
Signed-off-by: Andreas Schneider <asn@samba.org> Reviewed-by: Anoop C S <anoopcs@samba.org>
Autobuild-User(master): Anoop C S <anoopcs@samba.org>
Autobuild-Date(master): Thu Oct 16 09:34:58 UTC 2025 on atb-devel-224
Martin Schwenke [Sat, 11 Oct 2025 06:52:32 +0000 (17:52 +1100)]
ctdb-tests: Don't hard code creation of $CTDB_BASE/events/legacy/
setup_ctdb_base() should always create from etc-ctdb/. That directory
may be changed to not include the legacy/ component subdirectory, so
don't hard-code this particular subdirectory.
Signed-off-by: Martin Schwenke <mschwenke@ddn.com> Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Tue Oct 14 17:57:49 UTC 2025 on atb-devel-224
Martin Schwenke [Tue, 7 Oct 2025 11:27:06 +0000 (22:27 +1100)]
ctdb-tests: Improve setup_ctdb_base() to fix local_daemons.sh bug
$ ctdb/tests/local_daemons.sh foo onnode 0 ctdb event script list legacy
Unable to find event script installation directory: foo/node.0/share/events/legacy
This happens if the $CTDB_BASE/share/events/legacy/ directory does not
exist. This directory is (theoretically) for installed but not
necessarily enabled event scripts. ctdb/tests/etc-ctdb/ contains
events/legacy/ but setup_ctdb_base() currently does not ensure
creation of the corresponding directory under $CTDB_BASE/share/.
Ensure that an event script component installation subdirectory exists
for each corresponding configuration subdirectory.
An alternative would be to continue to allow the existing failure, but
create directory ctdb/tests/etc-ctdb/share/events/legacy/ specifically
to make local_daemons.sh work. However, it took 5 years to find the
current bug, so this may be too subtle for other users of
local_daemons.sh. Anyone wanting to test the failure case can remove
unwanted subdirectories after "local_daemons.sh setup".
Signed-off-by: Martin Schwenke <mschwenke@ddn.com> Reviewed-by: Ralph Boehme <slow@samba.org>
0x7d2f14452360 is located 736 bytes inside of 861-byte region [0x7d2f14452080,0x7d2f144523dd)
freed by thread T0 here:
#0 0x7faf16d208eb (/lib64/libasan.so.8+0x1208eb) (BuildId: 61b31c4760766f5f2552c32e175755894d8f6565)
#1 0x7faf14560a72 in _tc_free_poolmem ../../lib/talloc/talloc.c:1080
#2 0x7faf1455f71b in _tc_free_internal ../../lib/talloc/talloc.c:1215
#3 0x7faf1455ee1b in _tc_free_children_internal ../../lib/talloc/talloc.c:1669
#4 0x7faf1455ee1b in _tc_free_internal ../../lib/talloc/talloc.c:1184
#5 0x7faf14560315 in _talloc_free_internal ../../lib/talloc/talloc.c:1248
#6 0x7faf14560315 in _talloc_free ../../lib/talloc/talloc.c:1792
#7 0x7baf0a5c3883 in py_cli_notify_get_changes ../../source3/libsmb/pylibsmb.c:2274
#8 0x7faf165ba239 (/lib64/glibc-hwcaps/x86-64-v3/libpython3.13.so.1.0+0x1ba239) (BuildId: 3925b60e845f4803e4de04e1fdf7845f2e54ecb0)
Signed-off-by: Andreas Schneider <asn@samba.org> Reviewed-by: Noel Power <npower@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Tue Oct 14 12:35:37 UTC 2025 on atb-devel-224
Volker Lendecke [Mon, 13 Oct 2025 09:18:35 +0000 (11:18 +0200)]
vfs_fake_acls: Fix error path return in fake_acls_fstatat()
Thanks to Jeremy for pointing this out to me!
Signed-off-by: Volker Lendecke <vl@samba.org> Reviewed-by: Anoop C S <anoopcs@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Tue Oct 14 08:27:39 UTC 2025 on atb-devel-224
Gary Lockyer [Sun, 12 Oct 2025 20:46:16 +0000 (09:46 +1300)]
s4:torture:vfs:fruit fix Clang compilation issue
Fix:
[4176/4995] Compiling source4/torture/vfs/fruit.c
../../source4/torture/vfs/fruit.c:7900:2: error: variable 'h' is used
uninitialized whenever 'if' condition is true
[-Werror,-Wsometimes-uninitialized]
7900 | CHECK_STATUS(status, NT_STATUS_OK);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz> Reviewed-by: Martin Schwenke <martin@meltin.net>
Autobuild-User(master): Martin Schwenke <martins@samba.org>
Autobuild-Date(master): Mon Oct 13 04:42:16 UTC 2025 on atb-devel-224
Ralph Boehme [Sat, 27 Sep 2025 11:01:32 +0000 (13:01 +0200)]
smbtorture: add test "smb2.lease.lease-epoch"
Verifies the lease epoch is not incremented by the server (returns what the
client sent in the request) if a lease was not granted ie lease_level=NONE.
Test passes against Windows 2025.
From MS-SMB2 3.3.5.9.11 "Handling the SMB2_CREATE_REQUEST_LEASE_V2 Create
Context":
If the object store succeeds this request, Lease.LeaseState MUST be set to the
new caching state. The server MUST increment Lease.Epoch by 1.
Ralph Boehme [Fri, 28 Feb 2025 05:15:26 +0000 (06:15 +0100)]
smbtorture: correctly handle scale-out shares in smb2.durable-v2-open.create-blob
On a share with SMB2_SHARE_CAP_SCALEOUT capability the cluster doesn't
grant batch oplocks, at most level-II oplocks and it also doesn't
grant durable-handles (v1 or v2) as they require an RWH-lease.
With this change the test passes against Windows Server 2025.
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Volker Lendecke <vl@samba.org>
Ralph Boehme [Wed, 9 May 2018 06:25:51 +0000 (08:25 +0200)]
smbtorture: fix smb2.durable-v2-open.open-lease and smb2.durable-v2-open.persistent-open-lease
Only check the resulting lease state if we actually requested a lease and
correctly deal with SMB2_SHARE_CAP_SCALEOUT shares where the server grants at
most R leases and no Durable Handles.
With this change the test passes against Windows Server 2025.
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Volker Lendecke <vl@samba.org>
Ralph Boehme [Thu, 9 Oct 2025 13:27:31 +0000 (15:27 +0200)]
smbd: use op->global->open_global_id for the share_mode_entry.share_file_id
open_global_id is an uint32_t, share_file_id is an uint64_t.
For internal opens that don't have an smbXsrv_open_global.tdb entry, continue to
use a generation counter starting at UINT32_MAX + 1.
We don't expose internal opens via srvsrc, which imposes the requirement for
using an uint32_t, so we can use larger ids for the internal opens as srvsvc
never sees them (as they're not in smbXsrv_open_global.tdb and srvsvc as any
other component listing open files like smbstatus lists smbXsrv_open_global.tdb).
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Volker Lendecke <vl@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Fri Oct 10 11:43:16 UTC 2025 on atb-devel-224
Ralph Boehme [Thu, 9 Oct 2025 13:17:38 +0000 (15:17 +0200)]
smbd: move calling fsp_set_gen_id() to fsp_bind_smb()
No change in behaviour. fsp_set_gen_id() sets the primary key for the
share_mode_entry of the corresponding open file handle. None of the pathref fsp
needs that before being passed to SMB_VFS_CREATE_FILE().
As file_new() calls fsp_bind_smb(), all fsps will get an id when
create_file_unixpath() either calls file_new when creating files or it calls
fsp_bind_smb() when "converting" an pathref fsp that was passed into
SMB_VFS_CREATE_FILE() to an FSA fsp.
Signed-off-by: Ralph Boehme <slow@samba.org> Reviewed-by: Volker Lendecke <vl@samba.org>
Ralph Boehme [Thu, 18 Sep 2025 18:35:22 +0000 (20:35 +0200)]
smbd: hang directory pattern matching case sensitivity on the pathname
For the SMB3 POSIX client both posix_open=true and (fsp->fsp_name->flags &
SMB_FILENAME_POSIX_PATH) will always be the case, so this is no change in
behaviour for that case.
However, for the macOS client fruit will carefully setup both flags as
posix_open=true but SMB_FILENAME_POSIX_PATH will not be set.
This is a deliberate hack to give the macOS client POSIX behaviour for some
operations, but not for others, while also allowing the POSIX-ified macOS client
to continue to get case insensitive behavour.
Ralph Boehme [Thu, 18 Sep 2025 21:44:34 +0000 (23:44 +0200)]
smbd: hang posix brl per-handle check on the pathname
For the SMB3 POSIX client both posix_open=true and (fsp->fsp_name->flags &
SMB_FILENAME_POSIX_PATH) will always be the case, so this is no change in
behaviour for that case.
However, for the macOS client fruit will carefully setup both flags as
posix_open=true but SMB_FILENAME_POSIX_PATH will not be set.
This is a deliberate hack to give the macOS client POSIX behaviour for some
operations, but not for others, while also allowing the POSIX-ified macOS client
to continue to get case insensitive behavour.
Tags alls opens as POSIX by setting fsp_flags.posix_open to true.
POSIX handles have different behaviour compared to Windows:
Behaviour | POSIX | Windows | macOS |fruit:posix_opens = yes
-----------------------------------+------------+----------------------------------------
Deleting files with open handles | yes | no | yes | yes
Moving directories with open files | yes | no | yes | yes
Byterange locks behaviour | POSIX-ish | Window-ish | POSIX-ish | POSIX-ish
Sticky writetime | no | yes | no | no
Case sensitive | no | yes | yes | yes
Streams allowed | no | yes | yes | yes
macOS follows POSIX for the first four, but needs case insensitive behaviour
and needs streams.
By carefully setting fsp_flags.posix_open to true *after* going through the path
resolution logic, but before opens are added to locking.tdb, with
"fruit:posix_opens = yes" we get closest to macOS semantics.
Volker Lendecke [Sat, 4 Oct 2025 16:55:42 +0000 (18:55 +0200)]
vfs_fake_acls: Call fake_acls_fstatat() from fake_acls_[l]stat()
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): Fri Oct 10 09:26:47 UTC 2025 on atb-devel-224
Map the content of sdb_certificate_mappings to the hdb extension
HDB_Ext_CertificateMapping
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz> Reviewed-by: Jennifer Sutton <jennifersutton@catalyst.net.nz>
Autobuild-User(master): Jennifer Sutton <jsutton@samba.org>
Autobuild-Date(master): Fri Oct 10 02:30:06 UTC 2025 on atb-devel-224
Gary Lockyer [Thu, 28 Aug 2025 23:05:10 +0000 (11:05 +1200)]
config: add certificate backdating compensation
When preforming certificate based kerberos authentication (PKINIT) with
"strong certificate binding enforcement = compatibility"", for WEAK mappings
the certificate issue date must be after the date the user record was created.
This parameter relaxes that constraint by allowing the certificate to have been
issued up to the specified number of minutes before the user record was created.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz> Reviewed-by: Jennifer Sutton <jennifersutton@catalyst.net.nz>
This parameter controls the enforcement of Windows Certificate bindings as
outlined in "KB5014754: Certificate-based authentication changes on Windows
domain controllers",
when preforming certificate based kerberos authentication (PKINIT)
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz> Reviewed-by: Jennifer Sutton <jennifersutton@catalyst.net.nz>
Gary Lockyer [Mon, 6 Oct 2025 01:28:19 +0000 (14:28 +1300)]
s4:dsdb:audit_log change action for auth info
Change the action logged for authentication information changes from
"Public key change" to "Auth info change". To reflect that it's not just
changes to public keys that get logged.
This doesn't require a JSON log format version change, because the
version was recently bumped in c9e752ab18f43758d704951f7f31e39dafa6fdb4
and there hasn't been a Samba release in the meantime.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz> Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Gary Lockyer [Mon, 6 Oct 2025 00:06:12 +0000 (13:06 +1300)]
s4:dsdb:audit_log log auth info changes
Log changes to altSecurityIdentities, dNSHostName, msDS-additionalDnsHostNames
and servicePrincipal name in the same way that changes to mdDS-keyCredentialLink
changes are logged.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz> Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Gary Lockyer [Sun, 5 Oct 2025 21:29:28 +0000 (10:29 +1300)]
tests:audit_log_pass_change refactor the tests
Refactor python/samba/tests/audit_log_pass_change.py to make it easier add tests
for the logging of changes to alternateSecurityIdentities, dNSHostName,
servicePrincipleName
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz> Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Changes to heimdal to implement Windows strong and flexible certificate mapping
as outlined in
KB5014754: Certificate-based authentication changes
on Windows domain controllers
https://support.microsoft.com/en-us/topic/kb5014754-certificate-based-
authentication-changes-on-windows-
domain-controllers- ad2c23b0-15d8-4340-a468-4d4f3b188f16
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz> Reviewed-by: Jennifer Sutton <jennifersutton@catalyst.net.nz>
Autobuild-User(master): Jennifer Sutton <jsutton@samba.org>
Autobuild-Date(master): Wed Oct 8 21:12:44 UTC 2025 on atb-devel-224
Volker Lendecke [Thu, 2 Oct 2025 10:45:18 +0000 (12:45 +0200)]
vfs_fruit: Implement fstatat
This violates the abstraction in adouble.[ch], but passing "dirfsp"
and "relname" through ad_get() & friends would have been a more churn,
and with this violation of abstraction we only do fgetxattr once where
with a separate update_btime we do it twice. So in theory it should be
more efficient.
Signed-off-by: Volker Lendecke <vl@samba.org> Reviewed-by: Ralph Boehme <slow@samba.org>
Volker Lendecke [Thu, 2 Oct 2025 08:38:24 +0000 (10:38 +0200)]
lib: Add adouble_buf_parse()
Simplified version of ad_get that takes a buffer and does basic parsing of an
AppleDouble file format. The entries are represented as DATA_BLOBs directly
pointing at "buf" to avoid offset calculations in users of this.
Yes, this is a duplication of logic, but it makes the next patch
possible. Future patches could use this in ad_unpack()
Signed-off-by: Volker Lendecke <vl@samba.org> Reviewed-by: Ralph Boehme <slow@samba.org>
Martin Schwenke [Mon, 29 Sep 2025 07:24:53 +0000 (17:24 +1000)]
ctdb-event: Replace "no data" event scripts with previous data
If an event script list contains results with "no data" (summary ==
-ENODATA), then overwrite these with results from the previous run,
where possible.
To maintain current behaviour in "ctdb event status ...", stop
processing scripts if the begin timestamp in the list goes backwards.
This maintains backward compatibility, where only status for scripts
from the most recent run are shown.
Signed-off-by: Martin Schwenke <mschwenke@ddn.com> Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Tue Oct 7 11:09:41 UTC 2025 on atb-devel-224
Martin Schwenke [Fri, 26 Sep 2025 05:58:26 +0000 (15:58 +1000)]
ctdb-event: Don't replace an existing result with NULL
If script_list is NULL then event->script_list is set to NULL by
TALLOC_FREE(). This seems like the wrong thing to do because NULL
indicates a problem running the current event. We should keep the
previous result because it contains useful information.
In theory, this can't happen because the caller checks for NULL.
However, given that the check is here, it might as well do the right
thing... and this simplifies a subsequent change.
Signed-off-by: Martin Schwenke <mschwenke@ddn.com> Reviewed-by: Ralph Boehme <slow@samba.org>
Martin Schwenke [Mon, 29 Sep 2025 05:24:10 +0000 (15:24 +1000)]
ctdb-common: Make run_event return full script list
run_event currently truncates the resulting script list after the
first failure.
Instead, return the full script list, with scripts that weren't run
flagged with ENODATA in their summary. This structure is zeroed on
allocation, so this is the only field that needs to be set.
For now, have the event tool skip such scripts. Do this in
print_status() instead of print_status_one(), since there will soon be
additional checks in print_status() and it makes sense to put them
together.
Signed-off-by: Martin Schwenke <mschwenke@ddn.com> Reviewed-by: Ralph Boehme <slow@samba.org>
Pawan Sahu [Fri, 20 Jun 2025 03:42:01 +0000 (05:42 +0200)]
gpfs: Fetch GPFS DOS attributes asynchronously
This change introduces asynchronous handling of GPFS DOS attributes
using the pthreadpool infrastructure. This is part of the overall
effort to improve SMB directory listing performance in environments
with GPFS as the backend filesystem.
Move the logic that checks for sufficient threads and per-thread credential
support into a dedicated helper function: vfswrap_check_async_with_thread_creds().
Pawan Sahu [Fri, 9 May 2025 06:19:03 +0000 (23:19 -0700)]
smbd: Introduce struct vfs_pthread_pool_job_state
Refactor the vfswrap_getattrat_state structure by extracting the members
specific to pthreadpool job handling into a new struct vfs_pthread_pool_job_state.
This improves code clarity and allows reuse of the job-related
state in other contexts.
lib: Add a safeguard for misconfigured directory permissions
An installation I've seen had the path to the msg.lock directory set
up such that only root could access it. We're calling
serverid_exists() as non-root, so this is a pretty bad error. Log this
as an error in syslog by default and assume this PID still exists.
Signed-off-by: Volker Lendecke <vl@samba.org> Reviewed-by: Anoop C S <anoopcs@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Mon Sep 29 08:14:22 UTC 2025 on atb-devel-224
If read is failed torture_assert could return without freeing the t.
Fixing the leak.
Signed-off-by: Shweta Sodani <ssodani@redhat.com> Reviewed-by: Volker Lendecke <vl@samba.org> Reviewed-by: Vinit Agnihotri <vagnihot@redhat.com> Reviewed-by: Anoop C S <anoopcs@samba.org>
Autobuild-User(master): Anoop C S <anoopcs@samba.org>
Autobuild-Date(master): Sun Sep 28 09:30:56 UTC 2025 on atb-devel-224
lib:crypto: Use gnutls_cipher_decrypt3() if possible
Signed-off-by: Andreas Schneider <asn@samba.org> Reviewed-by: Stefan Metzmacher <metze@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Fri Sep 26 18:38:57 UTC 2025 on atb-devel-224
The call to get_time_t_max() doesn't work as expected when time_t is a
64-bits type and the returned value is stored into a 32-bits unsigned
integer. Truncating a 64-bits constant to a 32-bits number won't return,
in general, the same value we would get if time_t were a 32-bits type.
It's unsafe and could even return small numbers very far from the
intended maximum.
This patch completely avoids the need to use get_time_t_max() by
assuming that when pwd_max_age is 0 or -1, it means no maximum age is
defined, so the password never expires and we don't need to do any
comparison.
A small adjustment has also been made to avoid calling
pdb_get_account_policy() if it's not necessary.
Signed-off-by: Xavi Hernandez <xhernandez@gmail.com> Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz> Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Douglas Bagnall <dbagnall@samba.org>
Autobuild-Date(master): Fri Sep 26 05:23:30 UTC 2025 on atb-devel-224
Martin Schwenke [Fri, 15 Aug 2025 05:01:58 +0000 (15:01 +1000)]
ctdb-common: Only respect CTDB_SOCKET in CTDB_TEST_MODE
At the moment CTDB_SOCKET can be used outside of test mode even though
nobody should do this. So, no longer allow this.
This means ensuring CTDB_TEST_MODE is set in the in the
"clusteredmember" selftest environment, so that CTDB_SOCKET is
respected there..
Details...
The associated use of chown(2) and chmod(2), used to secure the socket
in ctdb_daemon.c:ux_socket_bind(), potentially enables a symlink race
attack. However, the chown(2) is currently not done in test mode, so
restricting the use of CTDB_SOCKET to test mode solves the potential
security issue.
Also, sprinkle warnings about use of CTDB_TEST_MODE in appropriate
places, just to attempt to limit unwanted behaviour.
An alternative could be to use the socket file descriptor with
fchown(2) and fchmod(2). However, these system calls are not well
defined on sockets. Still, this was previously done in CTDB's early
days (using the poorly documented method where they are allowed in
Linux (only?) before calling bind(2)). It was removed (due to
portability issues, via commits cf1056df94943ddcc3d547d4533b4bc04f57f265 and 2da3fe1b175a468fdff4aa4f65627facd2c28394) and replaced with the
current post-bind chown(2) and chmod(2).
I would like to remove the CTDB_SOCKET environment variable entirely,
since setting CTDB_TEST_MODE and CTDB_BASE covers all reasonable test
environments. However, I have a feeling that people use it for
interactive testing, and that can still be done in CTDB_TEST_MODE.
Martin Schwenke [Thu, 18 Sep 2025 05:24:32 +0000 (15:24 +1000)]
ctdb-ib: Replace uses of sprintf()
An unbounded sprintf() into ibw_lasterr (a 512-byte static data
buffer) can potentially cause overflow into other BSS data.
However, the risk is effectively minimised to zero due to:
* This code not being executed at all in ctdbd. It is only executed
in the accompanying test code:
- The function ibw_process_init_attrs() can cause a buffer overflow
if its 2nd argument, nattr, is non-zero and one of the structs in
the array pointed to by its 1st argument, attr, contains a name
member that is too long.
- ibw_process_init_attrs() is only called by ibw_init(), which also
has attr and nattr as its 1st and 2nd args, and it just passes them
straight through.
- ibw_init() is called in 2 places:
1. In ibwrapper_test.c, which is targeted test code.
2. In ibw_ctdb_init.c:ctdb_ibw_init(), which is the initialisation
function use to initialise the IB transport in ctdbd. Here, NULL
and 0 are passed as the relevant arguments to ibw_init(). Both
arguments are flagged with TODO comments. :-)
* This code is not built by default (--enable-infiniband is
required).
It appears that Debian and Red Hat family Linux distributions have
never distributed binaries with this enabled.
* Documentation (ctdb(7) and the wiki) recommends that private
addresses are configured on a private network that is separate from
client networks. So, even if the TODOs were done and the relevant
arguments could come off the wire, the attack surface should be very
small.
Only the instance with %s in the format is potentially problematic.
The others can not overflow the current 512 byte buffer. However, it
makes sense to change them all in case someone foolishly reduces the
size of the buffer and makes other changes so that the buffer can be
overflowed in ctdbd.
Now, will static analysers complain that the result of snprintf() is
not checked even though snprintf() always NUL-terminates?
Signed-off-by: Martin Schwenke <mschwenke@ddn.com> Reported-by: Marcos “Tr0p” Tolosa <marcos.tolosa@owasp.org> Reviewed-by: Volker Lendecke <vl@samba.org>
Fix:
*** CID 1665466: Resource leaks (RESOURCE_LEAK)
/source4/kdc/sdb_to_hdb.c: 482 in sdb_entry_to_hdb_entry()
476 }
477
478 h->context = ske;
479 if (ske != NULL) {
480 ske->kdc_entry = h;
481 }
>>> CID 1665466: Resource leaks (RESOURCE_LEAK)
>>> Variable "kt" going out of scope leaks the storage "kt.val" points to.
482 return 0;
483 error:
484 free_hdb_entry(h);
485 return rc;
This is almost certainly a false positive as when kt.len == 0, kt.val will be
NULL. But changing the condition to kt.val != NULL, will not do any harm.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz> Reviewed-by: Andreas Schneider <asn@samba.org> Reviewed-by: Anoop C S <anoopcs@samba.org>
Autobuild-User(master): Anoop C S <anoopcs@samba.org>
Autobuild-Date(master): Thu Sep 25 07:13:28 UTC 2025 on atb-devel-224
passdb: Fixing CID 1509029 for time_t in DEBUG statement
In the most likely scenario of time_t being 64 bit signed, we do the
typecast to intmax_t to also log negative values correctly.
Signed-off-by: Rabinarayan Panigrahi <rapanigr@redhat.com> Reviewed-by: Andreas Schneider <asn@samba.org> Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz> Reviewed-by: Anoop C S <anoopcs@samba.org> Reviewed-by: Vinit Agnihotri <vagnihot@redhat.com>
Autobuild-User(master): Anoop C S <anoopcs@samba.org>
Autobuild-Date(master): Wed Sep 24 06:44:50 UTC 2025 on atb-devel-224
Signed-off-by: Shweta Sodani <ssodani@redhat.com> Reviewed-by: Volker Lendecke <vl@samba.org> Reviewed-by: Anoop C S <anoopcs@samba.org>
Autobuild-User(master): Anoop C S <anoopcs@samba.org>
Autobuild-Date(master): Tue Sep 23 12:53:10 UTC 2025 on atb-devel-224
projects / thirdparty / samba.git / log
