<feed xmlns='http://www.w3.org/2005/Atom'>
<title>linux-toradex.git/fs/smb, branch master</title>
<subtitle>Linux kernel for Apalis and Colibri modules</subtitle>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/'/>
<entry>
<title>Merge tag 'cifs-fixes-7.4-rc4' of https://git.manguebit.org/linux</title>
<updated>2026-11-21T23:23:44+00:00</updated>
<author>
<name>Linus Torvalds</name>
<email>torvalds@linux-foundation.org</email>
</author>
<published>2026-11-21T23:23:44+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=62b13f41be6284250c06ab4d79a7fb13fe93f4d7'/>
<id>62b13f41be6284250c06ab4d79a7fb13fe93f4d7</id>
<content type='text'>
Pull smb client fixes from Paulo Alcantara:
 "The main ones are a server-triggerable use-after-free and two fixes
  that stop the client from accepting SMB2 responses and read data with
  a bad signature; the rest are a couple of busy-inode-on-unmount races
  and some metadata fixes.

   - Fix a use-after-free in the netfs write collector that a malicious
     or broken server can trigger by returning a write count larger than
     the request

   - Reject SMB2 responses whose signature fails verification instead of
     using them, so a tampered response on a signed connection is no
     longer accepted

   - Don't return read data whose signature failed verification when the
     read lands at the remote end of file

   - Fix a busy inode and dentry on unmount caused by a late oplock
     break queued after the workqueues were already flushed

   - Fix a busy dentry on unmount caused by a persistent-handle reopen
     racing the unmount

   - Fix creation and change time assignments in the SMB2 POSIX readdir
     path, which clobbered ctime and could give a hard link a separate
     in-memory inode

   - Normalise pre-epoch timestamps read from the server so stat no
     longer reports a garbled time

   - Fix DFS prefix paths being truncated on non-utf8 mounts, where
     path_consumed was applied in the wrong charset and the mount failed
     with -ENOENT"

* tag 'cifs-fixes-7.4-rc4' of https://git.manguebit.org/linux:
  smb: client: don't use read data that failed signature verification
  smb: client: do not let the server grow a write subrequest
  smb: client: fix busy dentry on unmount from persistent handle reopen
  smb: client: fix busy dentry and inode on unmount from late oplock break
  smb: client: fail SMB2 responses with a bad signature
  smb: client: fix times assignments in cifs_posix_to_fattr()
  smb: client: fix negative timestamps being read from smb
  smb: client: convert path_consumed to mount charset units
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
Pull smb client fixes from Paulo Alcantara:
 "The main ones are a server-triggerable use-after-free and two fixes
  that stop the client from accepting SMB2 responses and read data with
  a bad signature; the rest are a couple of busy-inode-on-unmount races
  and some metadata fixes.

   - Fix a use-after-free in the netfs write collector that a malicious
     or broken server can trigger by returning a write count larger than
     the request

   - Reject SMB2 responses whose signature fails verification instead of
     using them, so a tampered response on a signed connection is no
     longer accepted

   - Don't return read data whose signature failed verification when the
     read lands at the remote end of file

   - Fix a busy inode and dentry on unmount caused by a late oplock
     break queued after the workqueues were already flushed

   - Fix a busy dentry on unmount caused by a persistent-handle reopen
     racing the unmount

   - Fix creation and change time assignments in the SMB2 POSIX readdir
     path, which clobbered ctime and could give a hard link a separate
     in-memory inode

   - Normalise pre-epoch timestamps read from the server so stat no
     longer reports a garbled time

   - Fix DFS prefix paths being truncated on non-utf8 mounts, where
     path_consumed was applied in the wrong charset and the mount failed
     with -ENOENT"

* tag 'cifs-fixes-7.4-rc4' of https://git.manguebit.org/linux:
  smb: client: don't use read data that failed signature verification
  smb: client: do not let the server grow a write subrequest
  smb: client: fix busy dentry on unmount from persistent handle reopen
  smb: client: fix busy dentry and inode on unmount from late oplock break
  smb: client: fail SMB2 responses with a bad signature
  smb: client: fix times assignments in cifs_posix_to_fattr()
  smb: client: fix negative timestamps being read from smb
  smb: client: convert path_consumed to mount charset units
</pre>
</div>
</content>
</entry>
<entry>
<title>smb: client: don't use read data that failed signature verification</title>
<updated>2026-11-19T15:02:54+00:00</updated>
<author>
<name>Suyuan Wang</name>
<email>stevewong@hust.edu.cn</email>
</author>
<published>2026-11-19T04:00:48+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=a1a0ef3ba8e802060bfc2b0ebf131a2cf98c3b71'/>
<id>a1a0ef3ba8e802060bfc2b0ebf131a2cf98c3b71</id>
<content type='text'>
When smb2_readv_callback() finds a bad signature, it sets the error in
rdata-&gt;result but leaves rdata-&gt;got_bytes alone. Further down, a short
read that reaches the remote EOF clears the error again:

    size_t trans = rdata-&gt;subreq.transferred + rdata-&gt;got_bytes;
    if (trans &lt; rdata-&gt;subreq.len &amp;&amp;
        rdata-&gt;subreq.start + trans &gt;= netfs_read_remote_i_size(inode)) {
            rdata-&gt;result = 0;

and got_bytes is then added to subreq.transferred. So the last part of a
file, or all of a small one, is returned even though its signature is
wrong.

Clearing got_bytes is not enough on its own. If the read starts at or
past the cached remote size, for example because the file has grown on
the server since the size was last fetched, the check above still
matches and the error turns into EOF, or into a zero-filled tail when
NETFS_SREQ_CLEAR_TAIL is set.

Clear got_bytes when the signature check fails, as the retry path in the
same function already does, and only treat a short read as hitting the
remote EOF when the read itself succeeded. That is what commit
1da29f2c39b6 ("netfs, cifs: Fix handling of short DIO read") describes:
"clear the error if a short read occurred without error and the read-to
file position is now at the remote inode size".

Found with a static analysis tool and checked by hand.

Fixes: 1da29f2c39b6 ("netfs, cifs: Fix handling of short DIO read")
Cc: stable@vger.kernel.org
Co-developed-by: Kaixuan Li &lt;kaixuan.li@ntu.edu.sg&gt;
Signed-off-by: Kaixuan Li &lt;kaixuan.li@ntu.edu.sg&gt;
Signed-off-by: Suyuan Wang &lt;stevewong@hust.edu.cn&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
When smb2_readv_callback() finds a bad signature, it sets the error in
rdata-&gt;result but leaves rdata-&gt;got_bytes alone. Further down, a short
read that reaches the remote EOF clears the error again:

    size_t trans = rdata-&gt;subreq.transferred + rdata-&gt;got_bytes;
    if (trans &lt; rdata-&gt;subreq.len &amp;&amp;
        rdata-&gt;subreq.start + trans &gt;= netfs_read_remote_i_size(inode)) {
            rdata-&gt;result = 0;

and got_bytes is then added to subreq.transferred. So the last part of a
file, or all of a small one, is returned even though its signature is
wrong.

Clearing got_bytes is not enough on its own. If the read starts at or
past the cached remote size, for example because the file has grown on
the server since the size was last fetched, the check above still
matches and the error turns into EOF, or into a zero-filled tail when
NETFS_SREQ_CLEAR_TAIL is set.

Clear got_bytes when the signature check fails, as the retry path in the
same function already does, and only treat a short read as hitting the
remote EOF when the read itself succeeded. That is what commit
1da29f2c39b6 ("netfs, cifs: Fix handling of short DIO read") describes:
"clear the error if a short read occurred without error and the read-to
file position is now at the remote inode size".

Found with a static analysis tool and checked by hand.

Fixes: 1da29f2c39b6 ("netfs, cifs: Fix handling of short DIO read")
Cc: stable@vger.kernel.org
Co-developed-by: Kaixuan Li &lt;kaixuan.li@ntu.edu.sg&gt;
Signed-off-by: Kaixuan Li &lt;kaixuan.li@ntu.edu.sg&gt;
Signed-off-by: Suyuan Wang &lt;stevewong@hust.edu.cn&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</pre>
</div>
</content>
</entry>
<entry>
<title>smb: client: do not let the server grow a write subrequest</title>
<updated>2026-11-19T13:41:04+00:00</updated>
<author>
<name>ozgun kultekin</name>
<email>ozgunkultekin@gmail.com</email>
</author>
<published>2026-11-18T00:21:59+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=f834c81f108f8bdb0e3aa8375a536fb6dda56c70'/>
<id>f834c81f108f8bdb0e3aa8375a536fb6dda56c70</id>
<content type='text'>
smb2_writev_callback() takes the byte count from the SMB2 WRITE
response, masks it with 0xFFFF if it is larger than the request (for
old OS/2 servers), and then copies it into wdata-&gt;subreq.len.  A server
can return a count that is still larger than the request after the
mask, for example 0x8002 for a 5-byte write or 0xFFFF for a 4 KiB
write, and the subrequest then covers data that was never sent.

netfs uses the subrequest length to decide how far the write collector
may advance.  With a grown subrequest the collector moves collected_to
past the data that was issued: netfs_writeback_unlock_folios() walks
folio_queue slots that were never filled, and
rolling_buffer_delete_spent() frees folio_queue segments that later
in-flight subrequests still use.  This is a use-after-free that a
malicious or broken server triggers from an ordinary write:

  BUG: KASAN: slab-use-after-free in netfs_limit_iter+0xed4/0x1040
  Read of size 1 at addr ff11000010561511 by task kworker/u8:4/63
  Workqueue: events_unbound netfs_write_collection_worker
   netfs_limit_iter+0xed4/0x1040
   netfs_retry_writes+0x1410/0x1e90
   netfs_write_collection+0x2647/0x3880
   netfs_write_collection_worker+0x1cd/0x3e0
  Allocated by task 141:
   netfs_folioq_alloc+0x39a/0x470
   rolling_buffer_init+0x29/0xc0
   netfs_create_write_req+0x12f/0x840
  Freed by task 42:
   rolling_buffer_delete_spent+0xb1/0x220
   netfs_write_collection+0x1f62/0x3880

Never change the subrequest length from the response.  If the count is
still larger than the request after the mask, fail the write with -EIO
and record it with a new write_overlarge trace point, like the read
path does with read_overlarge.  A short count still returns -ENOSPC as
before.

The folio_queue form of the use-after-free exists since
commit cd0277ed0c18 ("netfs: Use new folio_queue data type and
iterator instead of xarray iter").

Fixes: 3ee1a1fc3981 ("cifs: Cut over to using netfslib")
Cc: stable@vger.kernel.org
Signed-off-by: ozgun kultekin &lt;ozgunkultekin@gmail.com&gt;
Reviewed-by: Namjae Jeon &lt;linkinjeon@kernel.org&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
smb2_writev_callback() takes the byte count from the SMB2 WRITE
response, masks it with 0xFFFF if it is larger than the request (for
old OS/2 servers), and then copies it into wdata-&gt;subreq.len.  A server
can return a count that is still larger than the request after the
mask, for example 0x8002 for a 5-byte write or 0xFFFF for a 4 KiB
write, and the subrequest then covers data that was never sent.

netfs uses the subrequest length to decide how far the write collector
may advance.  With a grown subrequest the collector moves collected_to
past the data that was issued: netfs_writeback_unlock_folios() walks
folio_queue slots that were never filled, and
rolling_buffer_delete_spent() frees folio_queue segments that later
in-flight subrequests still use.  This is a use-after-free that a
malicious or broken server triggers from an ordinary write:

  BUG: KASAN: slab-use-after-free in netfs_limit_iter+0xed4/0x1040
  Read of size 1 at addr ff11000010561511 by task kworker/u8:4/63
  Workqueue: events_unbound netfs_write_collection_worker
   netfs_limit_iter+0xed4/0x1040
   netfs_retry_writes+0x1410/0x1e90
   netfs_write_collection+0x2647/0x3880
   netfs_write_collection_worker+0x1cd/0x3e0
  Allocated by task 141:
   netfs_folioq_alloc+0x39a/0x470
   rolling_buffer_init+0x29/0xc0
   netfs_create_write_req+0x12f/0x840
  Freed by task 42:
   rolling_buffer_delete_spent+0xb1/0x220
   netfs_write_collection+0x1f62/0x3880

Never change the subrequest length from the response.  If the count is
still larger than the request after the mask, fail the write with -EIO
and record it with a new write_overlarge trace point, like the read
path does with read_overlarge.  A short count still returns -ENOSPC as
before.

The folio_queue form of the use-after-free exists since
commit cd0277ed0c18 ("netfs: Use new folio_queue data type and
iterator instead of xarray iter").

Fixes: 3ee1a1fc3981 ("cifs: Cut over to using netfslib")
Cc: stable@vger.kernel.org
Signed-off-by: ozgun kultekin &lt;ozgunkultekin@gmail.com&gt;
Reviewed-by: Namjae Jeon &lt;linkinjeon@kernel.org&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</pre>
</div>
</content>
</entry>
<entry>
<title>smb: client: fix busy dentry on unmount from persistent handle reopen</title>
<updated>2026-11-19T13:41:04+00:00</updated>
<author>
<name>Siddharth Chintamaneni</name>
<email>sidchintamaneni@gmail.com</email>
</author>
<published>2026-11-17T00:17:33+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=340ca61e9d31c31403013dc9190550d615844828'/>
<id>340ca61e9d31c31403013dc9190550d615844828</id>
<content type='text'>
During unmount, cifs_kill_sb() waits on cifs_sb-&gt;outstanding_rreq for
in-flight netfs I/O requests and oplock breaks to complete before
flushing serverclose_wq and fileinfo_put_wq.

While cifs_kill_sb() is waiting on outstanding_rreq, any cifsFileInfo
referenced by an in-flight request (req-&gt;cfile) remains on
tcon-&gt;openFileList. On a mount using persistent handles, if a reconnect
runs on cifsiod_wq during this window, smb2_reconnect_server() calls
cifs_reopen_persistent_handles(), which walks tcon-&gt;openFileList and
takes a new reference on open_file via cifsFileInfo_get() to reopen the
handle over the network:

  Thread 1 (umount)      Thread 2 (cifsiod_wq)    Thread 3 (workers)
  -----------------      ---------------------    ------------------
  cifs_kill_sb()
  wait(outstanding_rreq)
    [blocks: rreq == 1]
                         smb2_reconnect_server()
                         cifs_reopen_persistent_handles()
                           finds open_file on list
                           cifsFileInfo_get(open_file)
                           [cfile-&gt;count: 1 -&gt; 2]
                           cifs_reopen_file()
                                                  cifs_free_request()
                                                    cifsFileInfo_put()
                                                    [cfile-&gt;count: 2 -&gt; 1]
                                                    dec(outstanding_rreq)
    [wakes: rreq == 0]
  flush(serverclose_wq)
  flush(fileinfo_put_wq)
  kill_anon_super(sb)
    -&gt; generic_shutdown_super()
       BUG: Dentry still in use (1)
       VFS: Busy inodes after unmount!

Fix this by incrementing cifs_sb-&gt;outstanding_rreq when
cifs_reopen_persistent_handles() takes a reference on open_file and
decrementing it after cifsFileInfo_put(open_file), so wait_var_event()
in cifs_kill_sb() waits for any in-progress persistent handle reopen to
finish and queue its final put before draining fileinfo_put_wq.

Fixes: 340cea84f691 ("cifs: open files should not hold ref on superblock")
Cc: stable@vger.kernel.org
Reported-by: Adam Kim &lt;adamkim@microsoft.com&gt;
Signed-off-by: Siddharth Chintamaneni &lt;sidchintamaneni@gmail.com&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
During unmount, cifs_kill_sb() waits on cifs_sb-&gt;outstanding_rreq for
in-flight netfs I/O requests and oplock breaks to complete before
flushing serverclose_wq and fileinfo_put_wq.

While cifs_kill_sb() is waiting on outstanding_rreq, any cifsFileInfo
referenced by an in-flight request (req-&gt;cfile) remains on
tcon-&gt;openFileList. On a mount using persistent handles, if a reconnect
runs on cifsiod_wq during this window, smb2_reconnect_server() calls
cifs_reopen_persistent_handles(), which walks tcon-&gt;openFileList and
takes a new reference on open_file via cifsFileInfo_get() to reopen the
handle over the network:

  Thread 1 (umount)      Thread 2 (cifsiod_wq)    Thread 3 (workers)
  -----------------      ---------------------    ------------------
  cifs_kill_sb()
  wait(outstanding_rreq)
    [blocks: rreq == 1]
                         smb2_reconnect_server()
                         cifs_reopen_persistent_handles()
                           finds open_file on list
                           cifsFileInfo_get(open_file)
                           [cfile-&gt;count: 1 -&gt; 2]
                           cifs_reopen_file()
                                                  cifs_free_request()
                                                    cifsFileInfo_put()
                                                    [cfile-&gt;count: 2 -&gt; 1]
                                                    dec(outstanding_rreq)
    [wakes: rreq == 0]
  flush(serverclose_wq)
  flush(fileinfo_put_wq)
  kill_anon_super(sb)
    -&gt; generic_shutdown_super()
       BUG: Dentry still in use (1)
       VFS: Busy inodes after unmount!

Fix this by incrementing cifs_sb-&gt;outstanding_rreq when
cifs_reopen_persistent_handles() takes a reference on open_file and
decrementing it after cifsFileInfo_put(open_file), so wait_var_event()
in cifs_kill_sb() waits for any in-progress persistent handle reopen to
finish and queue its final put before draining fileinfo_put_wq.

Fixes: 340cea84f691 ("cifs: open files should not hold ref on superblock")
Cc: stable@vger.kernel.org
Reported-by: Adam Kim &lt;adamkim@microsoft.com&gt;
Signed-off-by: Siddharth Chintamaneni &lt;sidchintamaneni@gmail.com&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</pre>
</div>
</content>
</entry>
<entry>
<title>smb: client: fix busy dentry and inode on unmount from late oplock break</title>
<updated>2026-11-19T13:40:52+00:00</updated>
<author>
<name>Siddharth Chintamaneni</name>
<email>sidchintamaneni@gmail.com</email>
</author>
<published>2026-11-17T00:17:32+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=a10ec8dba77799a0460a4b3dfcac040ac452a325'/>
<id>a10ec8dba77799a0460a4b3dfcac040ac452a325</id>
<content type='text'>
During unmount, cifs_kill_sb() flushes cifsoplockd_wq and deferredclose_wq
before waiting on cifs_sb-&gt;outstanding_rreq for in-flight netfs I/O
requests to finish their cleanup_work in cifs_free_request().

While cifs_kill_sb() is waiting on outstanding_rreq, any cifsFileInfo
referenced by an in-flight request (req-&gt;cfile) still has cfile-&gt;count &gt;= 1
and remains on tcon-&gt;openFileList. If the SMB server sends a lease or
oplock break for that file during this window, the demultiplex thread
(cifsd) finds cfile on tcon-&gt;openFileList and calls
cifs_queue_oplock_break(cfile), which takes a new reference on cfile and
queues cfile-&gt;oplock_break onto cifsoplockd_wq after cifsoplockd_wq was
already flushed:

  Thread 1 (umount)      Thread 2 (cifsd)       Thread 3 (workers)
  -----------------      ----------------       ------------------
  cifs_kill_sb()
  flush(cifsoplockd_wq)
  flush(deferredclose_wq)
  wait(outstanding_rreq)
    [blocks: rreq == 1]
                         SMB2 Lease Break
                         finds cfile on list
                         cifs_queue_oplock_break()
                           queue(cifsoplockd_wq)
                           cifsFileInfo_get(cfile)
                           [cfile-&gt;count: 1 -&gt; 2]
                                                cifs_oplock_break()
                                                  ihold(inode)
                                                cifs_free_request()
                                                  cifsFileInfo_put()
                                                  [cfile-&gt;count: 2 -&gt; 1]
                                                  dec(outstanding_rreq)
    [wakes: rreq == 0]
  flush(serverclose_wq)
  flush(fileinfo_put_wq)
  kill_anon_super(sb)
    -&gt; generic_shutdown_super()
       BUG: Dentry still in use (1)
       VFS: Busy inodes after unmount!

Furthermore, cifs_oplock_break() itself holds the inode via ihold() after
dropping cfile and may initiate writeback via filemap_fdatawrite(),
spawning new netfs I/O requests.

Fix this by counting queued oplock breaks in cifs_sb-&gt;outstanding_rreq
from when cifs_queue_oplock_break() queues the work until
cifs_oplock_break() finishes, so wait_var_event() in cifs_kill_sb() waits
for both in-flight I/O requests and any late oplock breaks to complete.

Fixes: 340cea84f691 ("cifs: open files should not hold ref on superblock")
Cc: stable@vger.kernel.org
Reported-by: Adam Kim &lt;adamkim@microsoft.com&gt;
Signed-off-by: Siddharth Chintamaneni &lt;sidchintamaneni@gmail.com&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
During unmount, cifs_kill_sb() flushes cifsoplockd_wq and deferredclose_wq
before waiting on cifs_sb-&gt;outstanding_rreq for in-flight netfs I/O
requests to finish their cleanup_work in cifs_free_request().

While cifs_kill_sb() is waiting on outstanding_rreq, any cifsFileInfo
referenced by an in-flight request (req-&gt;cfile) still has cfile-&gt;count &gt;= 1
and remains on tcon-&gt;openFileList. If the SMB server sends a lease or
oplock break for that file during this window, the demultiplex thread
(cifsd) finds cfile on tcon-&gt;openFileList and calls
cifs_queue_oplock_break(cfile), which takes a new reference on cfile and
queues cfile-&gt;oplock_break onto cifsoplockd_wq after cifsoplockd_wq was
already flushed:

  Thread 1 (umount)      Thread 2 (cifsd)       Thread 3 (workers)
  -----------------      ----------------       ------------------
  cifs_kill_sb()
  flush(cifsoplockd_wq)
  flush(deferredclose_wq)
  wait(outstanding_rreq)
    [blocks: rreq == 1]
                         SMB2 Lease Break
                         finds cfile on list
                         cifs_queue_oplock_break()
                           queue(cifsoplockd_wq)
                           cifsFileInfo_get(cfile)
                           [cfile-&gt;count: 1 -&gt; 2]
                                                cifs_oplock_break()
                                                  ihold(inode)
                                                cifs_free_request()
                                                  cifsFileInfo_put()
                                                  [cfile-&gt;count: 2 -&gt; 1]
                                                  dec(outstanding_rreq)
    [wakes: rreq == 0]
  flush(serverclose_wq)
  flush(fileinfo_put_wq)
  kill_anon_super(sb)
    -&gt; generic_shutdown_super()
       BUG: Dentry still in use (1)
       VFS: Busy inodes after unmount!

Furthermore, cifs_oplock_break() itself holds the inode via ihold() after
dropping cfile and may initiate writeback via filemap_fdatawrite(),
spawning new netfs I/O requests.

Fix this by counting queued oplock breaks in cifs_sb-&gt;outstanding_rreq
from when cifs_queue_oplock_break() queues the work until
cifs_oplock_break() finishes, so wait_var_event() in cifs_kill_sb() waits
for both in-flight I/O requests and any late oplock breaks to complete.

Fixes: 340cea84f691 ("cifs: open files should not hold ref on superblock")
Cc: stable@vger.kernel.org
Reported-by: Adam Kim &lt;adamkim@microsoft.com&gt;
Signed-off-by: Siddharth Chintamaneni &lt;sidchintamaneni@gmail.com&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</pre>
</div>
</content>
</entry>
<entry>
<title>smb: client: fail SMB2 responses with a bad signature</title>
<updated>2026-11-18T19:45:38+00:00</updated>
<author>
<name>Suyuan Wang</name>
<email>stevewong@hust.edu.cn</email>
</author>
<published>2026-11-16T04:06:23+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=2a5bd7a2bbb787483cd69ecf099e2b591e3633b2'/>
<id>2a5bd7a2bbb787483cd69ecf099e2b591e3633b2</id>
<content type='text'>
smb2_check_receive() checks the response signature when signing is on,
but a failed check is only logged:

    rc = smb2_verify_signature(&amp;rqst, server);
    if (rc)
            cifs_server_dbg(VFS, "SMB signature verification returned ...",
                            rc);
    ...
    return map_smb2_to_linux_error(mid-&gt;resp_buf, log_error);

The status from the response is returned as usual, and the caller goes
on to use a response whose signature is wrong. This is the path for
synchronous and compound requests and for smb2_writev_callback(). On a
signed but unencrypted connection, anyone who can modify the traffic
can change e.g. a directory listing or file attributes, and the client
accepts it.

Return the verification error instead. That alone is not enough for
compound_send_recv(), which still hands the response buffer to the
caller after check_receive() fails. smb2_compound_op() then maps the
status from that buffer again and copies the payload out, and other
callers look at the buffer on error too, e.g. smb2_open_file() parses
a symlink target out of a STATUS_STOPPED_ON_SYMLINK response. So also
mark the mid MID_RESPONSE_MALFORMED, and have compound_send_recv() handle
it like a mid in a bad state: mark the earlier mids cancelled, so that a
CREATE which did succeed gets its handle closed, and don't pass any
response on.

Unlike the SMB1 fix in commit dc96f01d54cc ("smb: client: terminate
session upon failed client required signing"), don't reconnect: this
runs in the thread that sent the request, and only cifsd should call
cifs_reconnect(). And fail the request whichever side required signing,
since a response with a bad signature can't be trusted either way.

Found with a static analysis tool and checked by hand.

Reported-by: Daniel Birtwhistle &lt;daniel.birtwhistle@gehealthcare.com&gt;
Closes: https://lore.kernel.org/linux-cifs/SN4PR22MB28883DAF57E0B4D4BFB1C22AEEC42@SN4PR22MB2888.namprd22.prod.outlook.com/
Fixes: 3c1bf7e48e9e ("CIFS: Enable signing in SMB2")
Cc: stable@vger.kernel.org
Co-developed-by: Kaixuan Li &lt;kaixuan.li@ntu.edu.sg&gt;
Signed-off-by: Kaixuan Li &lt;kaixuan.li@ntu.edu.sg&gt;
Signed-off-by: Suyuan Wang &lt;stevewong@hust.edu.cn&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
smb2_check_receive() checks the response signature when signing is on,
but a failed check is only logged:

    rc = smb2_verify_signature(&amp;rqst, server);
    if (rc)
            cifs_server_dbg(VFS, "SMB signature verification returned ...",
                            rc);
    ...
    return map_smb2_to_linux_error(mid-&gt;resp_buf, log_error);

The status from the response is returned as usual, and the caller goes
on to use a response whose signature is wrong. This is the path for
synchronous and compound requests and for smb2_writev_callback(). On a
signed but unencrypted connection, anyone who can modify the traffic
can change e.g. a directory listing or file attributes, and the client
accepts it.

Return the verification error instead. That alone is not enough for
compound_send_recv(), which still hands the response buffer to the
caller after check_receive() fails. smb2_compound_op() then maps the
status from that buffer again and copies the payload out, and other
callers look at the buffer on error too, e.g. smb2_open_file() parses
a symlink target out of a STATUS_STOPPED_ON_SYMLINK response. So also
mark the mid MID_RESPONSE_MALFORMED, and have compound_send_recv() handle
it like a mid in a bad state: mark the earlier mids cancelled, so that a
CREATE which did succeed gets its handle closed, and don't pass any
response on.

Unlike the SMB1 fix in commit dc96f01d54cc ("smb: client: terminate
session upon failed client required signing"), don't reconnect: this
runs in the thread that sent the request, and only cifsd should call
cifs_reconnect(). And fail the request whichever side required signing,
since a response with a bad signature can't be trusted either way.

Found with a static analysis tool and checked by hand.

Reported-by: Daniel Birtwhistle &lt;daniel.birtwhistle@gehealthcare.com&gt;
Closes: https://lore.kernel.org/linux-cifs/SN4PR22MB28883DAF57E0B4D4BFB1C22AEEC42@SN4PR22MB2888.namprd22.prod.outlook.com/
Fixes: 3c1bf7e48e9e ("CIFS: Enable signing in SMB2")
Cc: stable@vger.kernel.org
Co-developed-by: Kaixuan Li &lt;kaixuan.li@ntu.edu.sg&gt;
Signed-off-by: Kaixuan Li &lt;kaixuan.li@ntu.edu.sg&gt;
Signed-off-by: Suyuan Wang &lt;stevewong@hust.edu.cn&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</pre>
</div>
</content>
</entry>
<entry>
<title>smb: client: fix times assignments in cifs_posix_to_fattr()</title>
<updated>2026-11-17T22:09:47+00:00</updated>
<author>
<name>koraynilay</name>
<email>koray.fra@gmail.com</email>
</author>
<published>2026-11-16T15:45:26+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=8b456f136a76a29fee8d36934740cb7f7fb5c837'/>
<id>8b456f136a76a29fee8d36934740cb7f7fb5c837</id>
<content type='text'>
Currently cifs_posix_to_fattr() wrongly sets SMB's CreationTime to
fattr-&gt;cf_ctime.

Fix this by setting fattr-&gt;cf_createtime to SMB's CreationTime and
fattr-&gt;cf_ctime to SMB's ChangeTime, matching the behaviour of
__dir_info_to_fattr().

This patch also fixes a few related issues reported by sashiko on
another patch (see Sashiko Link tag).

Reproducers for these bugs, with /tmp as the local filesystem folder
exported as [test] in ksmbd.conf and /mnt as the smb folder mounted with
mount -t cifs:

  mount -t cifs //127.0.0.1/test /mnt -o username=ciao,password=c,posix

Issue 1: ctime clobbered by CreationTime during readdir

  mkdir -p /mnt/dir
  touch /mnt/dir/file
  sleep 4
  chmod 644 /mnt/dir/file
  stat -c "ctime and btime before readdir: %Z %W" /mnt/dir/file
  ls -lc /mnt/dir &gt; /dev/null
  stat -c "ctime and btime after readdir:  %Z %W" /mnt/dir/file
  stat -c "ctime and btime on disk:        %Z %W" /tmp/dir/file

output before the fix, notice how ctime is the same as btime after the
readdir, while it's 4 seconds ahead (because of the sleep 4) before the
readdir (on disk timestamps are not touched):

  ctime and btime before readdir: 1794842068 1794842064
  ctime and btime after readdir:  1794842064 1794842064

  ctime and btime on disk:        1794842068 1794842064

output after the fix:

  ctime and btime before readdir: 1794842908 1794842904
  ctime and btime after readdir:  1794842908 1794842904

  ctime and btime on disk:        1794842908 1794842904

Issue 2: btime and ptime wiped out by readdir

[ no reproducer as it only happens after the linked patch series is
  applied, since it needs cifs_i-&gt;createtime = fattr-&gt;cf_createtime in
  cifs_fattr_to_inode() ]

Issue 3: in-memory inode aliasing (flock bypass on hardlink via readdir)

  mkdir -p /mnt/dir1 /mnt/dir2
  echo "test" &gt; /mnt/dir1/h1
  ln /mnt/dir1/h1 /mnt/dir2/h2
  exec 3&lt; /mnt/dir1/h1
  flock -x -n 3 # get exclusive lock
  ls /mnt/dir2 &gt; /dev/null
  exec 4&lt; /mnt/dir2/h2
  flock -x -n 4
  if [ $? -eq 0 ];then
  	echo "acquired: duplicate in-memory inode"
  else
  	echo "blocked: same inode"
  fi

output before the fix:

  acquired: duplicate in-memory inode

output after the fix:

  blocked: same inode

Fixes: 3d519bd1269f ("cifs: plumb smb2 POSIX dir enumeration")
Cc: stable@vger.kernel.org
Reported-by: Sashiko &lt;sashiko-bot@kernel.org&gt;
Link: https://sashiko.dev/#/patchset/20261115180343.576773-1-koray.fra%40gmail.com?part=10
Link: https://lore.kernel.org/all/20261115180343.576773-11-koray.fra@gmail.com/
Assisted-by: Gemini:3.8-flash-high antigravity-cli-1.5.7
Signed-off-by: koraynilay &lt;koray.fra@gmail.com&gt;
Reviewed-by: Namjae Jeon &lt;linkinjeon@kernel.org&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
Currently cifs_posix_to_fattr() wrongly sets SMB's CreationTime to
fattr-&gt;cf_ctime.

Fix this by setting fattr-&gt;cf_createtime to SMB's CreationTime and
fattr-&gt;cf_ctime to SMB's ChangeTime, matching the behaviour of
__dir_info_to_fattr().

This patch also fixes a few related issues reported by sashiko on
another patch (see Sashiko Link tag).

Reproducers for these bugs, with /tmp as the local filesystem folder
exported as [test] in ksmbd.conf and /mnt as the smb folder mounted with
mount -t cifs:

  mount -t cifs //127.0.0.1/test /mnt -o username=ciao,password=c,posix

Issue 1: ctime clobbered by CreationTime during readdir

  mkdir -p /mnt/dir
  touch /mnt/dir/file
  sleep 4
  chmod 644 /mnt/dir/file
  stat -c "ctime and btime before readdir: %Z %W" /mnt/dir/file
  ls -lc /mnt/dir &gt; /dev/null
  stat -c "ctime and btime after readdir:  %Z %W" /mnt/dir/file
  stat -c "ctime and btime on disk:        %Z %W" /tmp/dir/file

output before the fix, notice how ctime is the same as btime after the
readdir, while it's 4 seconds ahead (because of the sleep 4) before the
readdir (on disk timestamps are not touched):

  ctime and btime before readdir: 1794842068 1794842064
  ctime and btime after readdir:  1794842064 1794842064

  ctime and btime on disk:        1794842068 1794842064

output after the fix:

  ctime and btime before readdir: 1794842908 1794842904
  ctime and btime after readdir:  1794842908 1794842904

  ctime and btime on disk:        1794842908 1794842904

Issue 2: btime and ptime wiped out by readdir

[ no reproducer as it only happens after the linked patch series is
  applied, since it needs cifs_i-&gt;createtime = fattr-&gt;cf_createtime in
  cifs_fattr_to_inode() ]

Issue 3: in-memory inode aliasing (flock bypass on hardlink via readdir)

  mkdir -p /mnt/dir1 /mnt/dir2
  echo "test" &gt; /mnt/dir1/h1
  ln /mnt/dir1/h1 /mnt/dir2/h2
  exec 3&lt; /mnt/dir1/h1
  flock -x -n 3 # get exclusive lock
  ls /mnt/dir2 &gt; /dev/null
  exec 4&lt; /mnt/dir2/h2
  flock -x -n 4
  if [ $? -eq 0 ];then
  	echo "acquired: duplicate in-memory inode"
  else
  	echo "blocked: same inode"
  fi

output before the fix:

  acquired: duplicate in-memory inode

output after the fix:

  blocked: same inode

Fixes: 3d519bd1269f ("cifs: plumb smb2 POSIX dir enumeration")
Cc: stable@vger.kernel.org
Reported-by: Sashiko &lt;sashiko-bot@kernel.org&gt;
Link: https://sashiko.dev/#/patchset/20261115180343.576773-1-koray.fra%40gmail.com?part=10
Link: https://lore.kernel.org/all/20261115180343.576773-11-koray.fra@gmail.com/
Assisted-by: Gemini:3.8-flash-high antigravity-cli-1.5.7
Signed-off-by: koraynilay &lt;koray.fra@gmail.com&gt;
Reviewed-by: Namjae Jeon &lt;linkinjeon@kernel.org&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</pre>
</div>
</content>
</entry>
<entry>
<title>smb: client: fix negative timestamps being read from smb</title>
<updated>2026-11-17T16:00:38+00:00</updated>
<author>
<name>koraynilay</name>
<email>koray.fra@gmail.com</email>
</author>
<published>2026-11-12T11:16:37+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=6a75ce0866bb1efe67947fc5536a476381d7a9d9'/>
<id>6a75ce0866bb1efe67947fc5536a476381d7a9d9</id>
<content type='text'>
When reading a timestamp before epoch 0, cifs converts it from NT format
to Unix format without normalizing it, leading to the nanoseconds field
to be used as a negative number, which leads to a garbled stat command
output.

This bug also violates a struct timespec64 invariant, as documented in
the set_normalized_timespec64() kernel-doc:
[...]
 * Note: The tv_nsec part is always in the range of 0 &lt;= tv_nsec &lt;
 * NSEC_PER_SEC.
 * For negative values only the tv_sec field is negative !
 */

Reproducer, with /tmp as the local filesystem folder exported as [test]
in ksmbd.conf and /mnt as the smb folder mounted with mount -t cifs:

  mount -t cifs //127.0.0.1/test /mnt -o username=user,password=pass
  touch -d "1969-07-20 20:17:40.123456789" /tmp/test.txt
  stat -c "Server local (/tmp):   %y" /tmp/test.txt
  stat -c "CIFS client  (/mnt):   %y" /mnt/test.txt --cached=never
  smbclient //127.0.0.1/test -U user%pass -c "allinfo test.txt"

which prints (stat output is garbled because of negative value):

  Server local (/tmp):   1969-07-20 20:17:40.123456789 +0000
  CIFS client  (/mnt):   1969-07-20 20:17:41.()*+,--00 +0000
  altname: TESTT~K6.TXT
  [...]
  access_time:    Sun Jul 20 20:17:40 1969 UTC
  write_time:     Sun Jul 20 20:17:40 1969 UTC
  [...]

Fix this by using set_normalized_timespec64() in cifs_NTtimeToUnix().

Fixes: 2ae83bf93882 ("[CIFS] Fix setting time before epoch (negative time values)")
Cc: stable@vger.kernel.org
Assisted-by: Gemini:3.8-flash-high antigravity-cli-1.5.3
Signed-off-by: koraynilay &lt;koray.fra@gmail.com&gt;
Reviewed-by: Namjae Jeon &lt;linkinjeon@kernel.org&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
When reading a timestamp before epoch 0, cifs converts it from NT format
to Unix format without normalizing it, leading to the nanoseconds field
to be used as a negative number, which leads to a garbled stat command
output.

This bug also violates a struct timespec64 invariant, as documented in
the set_normalized_timespec64() kernel-doc:
[...]
 * Note: The tv_nsec part is always in the range of 0 &lt;= tv_nsec &lt;
 * NSEC_PER_SEC.
 * For negative values only the tv_sec field is negative !
 */

Reproducer, with /tmp as the local filesystem folder exported as [test]
in ksmbd.conf and /mnt as the smb folder mounted with mount -t cifs:

  mount -t cifs //127.0.0.1/test /mnt -o username=user,password=pass
  touch -d "1969-07-20 20:17:40.123456789" /tmp/test.txt
  stat -c "Server local (/tmp):   %y" /tmp/test.txt
  stat -c "CIFS client  (/mnt):   %y" /mnt/test.txt --cached=never
  smbclient //127.0.0.1/test -U user%pass -c "allinfo test.txt"

which prints (stat output is garbled because of negative value):

  Server local (/tmp):   1969-07-20 20:17:40.123456789 +0000
  CIFS client  (/mnt):   1969-07-20 20:17:41.()*+,--00 +0000
  altname: TESTT~K6.TXT
  [...]
  access_time:    Sun Jul 20 20:17:40 1969 UTC
  write_time:     Sun Jul 20 20:17:40 1969 UTC
  [...]

Fix this by using set_normalized_timespec64() in cifs_NTtimeToUnix().

Fixes: 2ae83bf93882 ("[CIFS] Fix setting time before epoch (negative time values)")
Cc: stable@vger.kernel.org
Assisted-by: Gemini:3.8-flash-high antigravity-cli-1.5.3
Signed-off-by: koraynilay &lt;koray.fra@gmail.com&gt;
Reviewed-by: Namjae Jeon &lt;linkinjeon@kernel.org&gt;
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</pre>
</div>
</content>
</entry>
<entry>
<title>smb: client: convert path_consumed to mount charset units</title>
<updated>2026-11-17T15:08:21+00:00</updated>
<author>
<name>Fredric Cover</name>
<email>fredric.cover.lkernel@gmail.com</email>
</author>
<published>2026-11-12T04:21:36+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=3a191eec3627acd558640cf4092e9e337aa15291'/>
<id>3a191eec3627acd558640cf4092e9e337aa15291</id>
<content type='text'>
path_consumed in a DFS referral counts bytes of the canonical path,
which is in the DFS cache charset (utf8). dfs_parse_target_referral()
applies it as an offset into full_path, which is in the mount charset.
When the mount charset is not utf8, each non-ASCII character in the
consumed part that encodes longer in utf8 than in the mount charset
makes the offset too large, so leading characters of the prefix path
are dropped. For example, with iocharset=iso8859-1, mounting
//srv/dfs/café é/sub yields the prefix "ub" instead of "sub" and the
mount fails with -ENOENT. (A single accented character only worked
because the offset happened to land after the delimiter.)

Translate the count in parse_dfs_target(), where both strings are
available, by matching path components: count the delimiters in the
consumed part of the canonical path and find the same position in the
mount charset path. Delimiters are ASCII in both and are not affected
by remapping. If the consumed bytes are identical in both strings,
which is the case for utf8 mounts and ASCII-only prefixes using the
default backslash delimiter, the value is used unchanged.

This assumes the server reports PathConsumed on a component boundary;
a translated value that ends mid-component is rounded up to the next
boundary.

Tested against Samba on a KASAN kernel: with iocharset=iso8859-1, links
with 0 to 3 accented characters followed by a tail, a two-level tail,
an accented tail, a link in a subdirectory, and mapposix with an
accented name; iso8859-15 and cp932 mounts; and iocharset=utf8
controls. Not tested against Windows servers, with posixpaths, or with
reserved characters under mapchars/mapposix.

Signed-off-by: Fredric Cover &lt;fredric.cover.lkernel@gmail.com&gt;
Cc: stable@vger.kernel.org
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
path_consumed in a DFS referral counts bytes of the canonical path,
which is in the DFS cache charset (utf8). dfs_parse_target_referral()
applies it as an offset into full_path, which is in the mount charset.
When the mount charset is not utf8, each non-ASCII character in the
consumed part that encodes longer in utf8 than in the mount charset
makes the offset too large, so leading characters of the prefix path
are dropped. For example, with iocharset=iso8859-1, mounting
//srv/dfs/café é/sub yields the prefix "ub" instead of "sub" and the
mount fails with -ENOENT. (A single accented character only worked
because the offset happened to land after the delimiter.)

Translate the count in parse_dfs_target(), where both strings are
available, by matching path components: count the delimiters in the
consumed part of the canonical path and find the same position in the
mount charset path. Delimiters are ASCII in both and are not affected
by remapping. If the consumed bytes are identical in both strings,
which is the case for utf8 mounts and ASCII-only prefixes using the
default backslash delimiter, the value is used unchanged.

This assumes the server reports PathConsumed on a component boundary;
a translated value that ends mid-component is rounded up to the next
boundary.

Tested against Samba on a KASAN kernel: with iocharset=iso8859-1, links
with 0 to 3 accented characters followed by a tail, a two-level tail,
an accented tail, a link in a subdirectory, and mapposix with an
accented name; iso8859-15 and cp932 mounts; and iocharset=utf8
controls. Not tested against Windows servers, with posixpaths, or with
reserved characters under mapchars/mapposix.

Signed-off-by: Fredric Cover &lt;fredric.cover.lkernel@gmail.com&gt;
Cc: stable@vger.kernel.org
Signed-off-by: Paulo Alcantara &lt;pc@manguebit.org&gt;
</pre>
</div>
</content>
</entry>
<entry>
<title>Merge tag 'ksmbd-for-7.4-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/linkinjeon/smb</title>
<updated>2026-11-14T16:23:39+00:00</updated>
<author>
<name>Linus Torvalds</name>
<email>torvalds@linux-foundation.org</email>
</author>
<published>2026-11-14T16:23:39+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=cba88960471ac67635080708b67ed5c2a2e4e9fd'/>
<id>cba88960471ac67635080708b67ed5c2a2e4e9fd</id>
<content type='text'>
Pull smb server fixes from Namjae Jeon:

 - Evaluate explicit-access requests in DACL order with exact SID
   matching, and prevent guest and anonymous sessions from matching
   Authenticated Users

 - Apply Unix group and OWNER RIGHTS ACEs, and keep their denials when
   POSIX ACL fallback determines access

 - Allow the parent directory’s DACL to grant explicitly requested
   DELETE and FILE_READ_ATTRIBUTES through FILE_DELETE_CHILD and
   FILE_LIST_DIRECTORY

 - Exclude metadata-only file and stream opens from sharing conflicts
   and stream delete-sharing checks, while accounting for writes implied
   by overwrites and named-stream creation

 - Return STATUS_ACCESS_DENIED for reads on files lacking FMODE_READ,
   preventing warnings in __kernel_read() and potential panics

 - Pin parent dentries during SMB2 CREATE permission, status, and ACL
   inheritance operations to prevent use-after-free races with
   concurrent rename and removal

 - Keep expired sessions bound to their original user during NTLM
   reauthentication, preventing identity changes and a user-object leak

* tag 'ksmbd-for-7.4-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/linkinjeon/smb:
  ksmbd: keep group denials when a POSIX ACL entry decides access
  ksmbd: refuse reauthentication of an expired session as another user
  ksmbd: pin parent dentries in SMB2 CREATE ACL path
  ksmbd: let the parent's DACL grant DELETE and FILE_READ_ATTRIBUTES
  ksmbd: evaluate the DACL in order for explicit access requests
  ksmbd: skip metadata-only stream opens in the delete check
  ksmbd: skip metadata-only opens in sharing checks
  ksmbd: refuse to read a handle that is not open for reading
  ksmbd: grant a caller what its groups and OWNER RIGHTS ACEs allow
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
Pull smb server fixes from Namjae Jeon:

 - Evaluate explicit-access requests in DACL order with exact SID
   matching, and prevent guest and anonymous sessions from matching
   Authenticated Users

 - Apply Unix group and OWNER RIGHTS ACEs, and keep their denials when
   POSIX ACL fallback determines access

 - Allow the parent directory’s DACL to grant explicitly requested
   DELETE and FILE_READ_ATTRIBUTES through FILE_DELETE_CHILD and
   FILE_LIST_DIRECTORY

 - Exclude metadata-only file and stream opens from sharing conflicts
   and stream delete-sharing checks, while accounting for writes implied
   by overwrites and named-stream creation

 - Return STATUS_ACCESS_DENIED for reads on files lacking FMODE_READ,
   preventing warnings in __kernel_read() and potential panics

 - Pin parent dentries during SMB2 CREATE permission, status, and ACL
   inheritance operations to prevent use-after-free races with
   concurrent rename and removal

 - Keep expired sessions bound to their original user during NTLM
   reauthentication, preventing identity changes and a user-object leak

* tag 'ksmbd-for-7.4-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/linkinjeon/smb:
  ksmbd: keep group denials when a POSIX ACL entry decides access
  ksmbd: refuse reauthentication of an expired session as another user
  ksmbd: pin parent dentries in SMB2 CREATE ACL path
  ksmbd: let the parent's DACL grant DELETE and FILE_READ_ATTRIBUTES
  ksmbd: evaluate the DACL in order for explicit access requests
  ksmbd: skip metadata-only stream opens in the delete check
  ksmbd: skip metadata-only opens in sharing checks
  ksmbd: refuse to read a handle that is not open for reading
  ksmbd: grant a caller what its groups and OWNER RIGHTS ACEs allow
</pre>
</div>
</content>
</entry>
</feed>
