linux-2.6-block.git
3 days agoSUNRPC: add svcauth_map_clnt_to_svc_cred_local
Weston Andros Adamson [Thu, 5 Sep 2024 19:09:46 +0000 (15:09 -0400)]
SUNRPC: add svcauth_map_clnt_to_svc_cred_local

Add new funtion svcauth_map_clnt_to_svc_cred_local which maps a
generic cred to a svc_cred suitable for use in nfsd.

This is needed by the localio code to map nfs client creds to nfs
server credentials.

Following from net/sunrpc/auth_unix.c:unx_marshal() it is clear that
->fsuid and ->fsgid must be used (rather than ->uid and ->gid).  In
addition, these uid and gid must be translated with from_kuid_munged()
so local client uses correct uid and gid when acting as local server.

Jeff Layton noted:
  This is where the magic happens. Since we're working in
  kuid_t/kgid_t, we don't need to worry about further idmapping.

Suggested-by: NeilBrown <neilb@suse.de> # to approximate unx_marshal()
Signed-off-by: Weston Andros Adamson <dros@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Co-developed-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoSUNRPC: remove call_allocate() BUG_ONs
Mike Snitzer [Thu, 5 Sep 2024 19:09:45 +0000 (15:09 -0400)]
SUNRPC: remove call_allocate() BUG_ONs

Remove BUG_ON if p_arglen=0 to allow RPC with void arg.
Remove BUG_ON if p_replen=0 to allow RPC with void return.

The former was needed for the first revision of the LOCALIO protocol
which had an RPC that took a void arg:

    /* raw RFC 9562 UUID */
    typedef u8 uuid_t<UUID_SIZE>;

    program NFS_LOCALIO_PROGRAM {
        version LOCALIO_V1 {
            void
                NULL(void) = 0;

            uuid_t
                GETUUID(void) = 1;
        } = 1;
    } = 400122;

The latter is needed for the final revision of the LOCALIO protocol
which has a UUID_IS_LOCAL RPC which returns a void:

    /* raw RFC 9562 UUID */
    typedef u8 uuid_t<UUID_SIZE>;

    program NFS_LOCALIO_PROGRAM {
        version LOCALIO_V1 {
            void
                NULL(void) = 0;

            void
                UUID_IS_LOCAL(uuid_t) = 1;
        } = 1;
    } = 400122;

There is really no value in triggering a BUG_ON in response to either
of these previously unsupported conditions.

NeilBrown would like the entire 'if (proc->p_proc != 0)' branch
removed (not just the one BUG_ON that must be removed for LOCALIO's
immediate needs of returning void).

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Reviewed-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfsd: add nfsd_serv_try_get and nfsd_serv_put
Mike Snitzer [Thu, 5 Sep 2024 19:09:44 +0000 (15:09 -0400)]
nfsd: add nfsd_serv_try_get and nfsd_serv_put

Introduce nfsd_serv_try_get and nfsd_serv_put and update the nfsd code
to prevent nfsd_destroy_serv from destroying nn->nfsd_serv until any
caller of nfsd_serv_try_get releases their reference using nfsd_serv_put.

A percpu_ref is used to implement the interlock between
nfsd_destroy_serv and any caller of nfsd_serv_try_get.

This interlock is needed to properly wait for the completion of client
initiated localio calls to nfsd (that are _not_ in the context of nfsd).

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfsd: add nfsd_file_acquire_local()
NeilBrown [Thu, 5 Sep 2024 19:09:43 +0000 (15:09 -0400)]
nfsd: add nfsd_file_acquire_local()

nfsd_file_acquire_local() can be used to look up a file by filehandle
without having a struct svc_rqst.  This can be used by NFS LOCALIO to
allow the NFS client to bypass the NFS protocol to directly access a
file provided by the NFS server which is running in the same kernel.

In nfsd_file_do_acquire() care is taken to always use fh_verify() if
rqstp is not NULL (as is the case for non-LOCALIO callers).  Otherwise
the non-LOCALIO callers will not supply the correct and required
arguments to __fh_verify (e.g. gssclient isn't passed).

Introduce fh_verify_local() wrapper around __fh_verify to make it
clear that LOCALIO is intended caller.

Also, use GC for nfsd_file returned by nfsd_file_acquire_local.  GC
offers performance improvements if/when a file is reopened before
launderette cleans it from the filecache's LRU.

Suggested-by: Jeff Layton <jlayton@kernel.org> # use filecache's GC
Signed-off-by: NeilBrown <neilb@suse.de>
Co-developed-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfsd: factor out __fh_verify to allow NULL rqstp to be passed
NeilBrown [Thu, 5 Sep 2024 19:09:42 +0000 (15:09 -0400)]
nfsd: factor out __fh_verify to allow NULL rqstp to be passed

__fh_verify() offers an interface like fh_verify() but doesn't require
a struct svc_rqst *, instead it also takes the specific parts as
explicit required arguments.  So it is safe to call __fh_verify() with
a NULL rqstp, but the net, cred, and client args must not be NULL.

__fh_verify() does not use SVC_NET(), nor does the functions it calls.

Rather than using rqstp->rq_client pass the client and gssclient
explicitly to __fh_verify and then to nfsd_set_fh_dentry().

Lastly, it should be noted that the previous commit prepared for 4
associated tracepoints to only be used if rqstp is not NULL (this is a
stop-gap that should be properly fixed so localio also benefits from
the utility these tracepoints provide when debugging fh_verify
issues).

Signed-off-by: NeilBrown <neilb@suse.de>
Co-developed-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoNFSD: Short-circuit fh_verify tracepoints for LOCALIO
Chuck Lever [Thu, 5 Sep 2024 19:09:41 +0000 (15:09 -0400)]
NFSD: Short-circuit fh_verify tracepoints for LOCALIO

LOCALIO will be able to call fh_verify() with a NULL rqstp. In this
case, the existing trace points need to be skipped because they
want to dereference the address fields in the passed-in rqstp.

Temporarily make these trace points conditional to avoid a seg
fault in this case. Putting the "rqstp != NULL" check in the trace
points themselves makes the check more efficient.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Acked-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoNFSD: Avoid using rqstp->rq_vers in nfsd_set_fh_dentry()
Chuck Lever [Thu, 5 Sep 2024 19:09:40 +0000 (15:09 -0400)]
NFSD: Avoid using rqstp->rq_vers in nfsd_set_fh_dentry()

Currently, fh_verify() makes some daring assumptions about which
version of file handle the caller wants, based on the things it can
find in the passed-in rqstp. The about-to-be-introduced LOCALIO use
case sometimes has no svc_rqst context, so this logic won't work in
that case.

Instead, examine the passed-in file handle. It's .max_size field
should carry information to allow nfsd_set_fh_dentry() to initialize
the file handle appropriately.

The file handle used by lockd and the one created by write_filehandle
never need any of the version-specific fields (which affect things
like write and getattr requests and pre/post attributes).

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoNFSD: Refactor nfsd_setuser_and_check_port()
NeilBrown [Thu, 5 Sep 2024 19:09:39 +0000 (15:09 -0400)]
NFSD: Refactor nfsd_setuser_and_check_port()

There are several places where __fh_verify unconditionally dereferences
rqstp to check that the connection is suitably secure.  They look at
rqstp->rq_xprt which is not meaningful in the target use case of
"localio" NFS in which the client talks directly to the local server.

Prepare these to always succeed when rqstp is NULL.

Signed-off-by: NeilBrown <neilb@suse.de>
Co-developed-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoNFSD: Handle @rqstp == NULL in check_nfsd_access()
NeilBrown [Thu, 5 Sep 2024 19:09:38 +0000 (15:09 -0400)]
NFSD: Handle @rqstp == NULL in check_nfsd_access()

LOCALIO-initiated open operations are not running in an nfsd thread
and thus do not have an associated svc_rqst context.

Signed-off-by: NeilBrown <neilb@suse.de>
Co-developed-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfs: factor out {encode,decode}_opaque_fixed to nfs_xdr.h
Mike Snitzer [Thu, 5 Sep 2024 19:09:37 +0000 (15:09 -0400)]
nfs: factor out {encode,decode}_opaque_fixed to nfs_xdr.h

Eliminates duplicate functions in various files to allow for
additional callers.

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Reviewed-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfs_common: factor out nfs4_errtbl and nfs4_stat_to_errno
Mike Snitzer [Thu, 5 Sep 2024 19:09:36 +0000 (15:09 -0400)]
nfs_common: factor out nfs4_errtbl and nfs4_stat_to_errno

Common nfs4_stat_to_errno() is used by fs/nfs/nfs4xdr.c and will be
used by fs/nfs/localio.c

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfs_common: factor out nfs_errtbl and nfs_stat_to_errno
Mike Snitzer [Thu, 5 Sep 2024 19:09:35 +0000 (15:09 -0400)]
nfs_common: factor out nfs_errtbl and nfs_stat_to_errno

Common nfs_stat_to_errno() is used by both fs/nfs/nfs2xdr.c and
fs/nfs/nfs3xdr.c

Will also be used by fs/nfsd/localio.c

Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfs: add 'noalignwrite' option for lock-less 'lost writes' prevention
Dan Aloni [Wed, 24 Jul 2024 11:07:12 +0000 (14:07 +0300)]
nfs: add 'noalignwrite' option for lock-less 'lost writes' prevention

There are some applications that write to predefined non-overlapping
file offsets from multiple clients and therefore don't need to rely on
file locking. However, if these applications want non-aligned offsets
and sizes they need to either use locks or risk data corruption, as the
NFS client defaults to extending writes to whole pages.

This commit adds a new mount option `noalignwrite`, which allows to turn
that off and avoid the need of locking, as long as these applications
don't overlap on offsets.

Signed-off-by: Dan Aloni <dan.aloni@vastdata.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfs: fix the comment of nfs_get_root
Li Lingfeng [Sat, 24 Aug 2024 01:43:35 +0000 (09:43 +0800)]
nfs: fix the comment of nfs_get_root

The comment for nfs_get_root() needs to be updated as it would also be
used by NFS4 as follows:
@x[
    nfs_get_root+1
    nfs_get_tree_common+1819
    nfs_get_tree+2594
    vfs_get_tree+73
    fc_mount+23
    do_nfs4_mount+498
    nfs4_try_get_tree+134
    nfs_get_tree+2562
    vfs_get_tree+73
    path_mount+2776
    do_mount+226
    __se_sys_mount+343
    __x64_sys_mount+106
    do_syscall_64+69
    entry_SYSCALL_64_after_hwframe+97
, mount.nfs4]: 1

Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Acked-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoNFSv4.2: Fix detection of "Proxying of Times" server support
Roi Azarzar [Sun, 15 Sep 2024 10:27:35 +0000 (10:27 +0000)]
NFSv4.2: Fix detection of "Proxying of Times" server support

According to draft-ietf-nfsv4-delstid-07:
   If a server informs the client via the fattr4_open_arguments
   attribute that it supports
   OPEN_ARGS_SHARE_ACCESS_WANT_DELEG_TIMESTAMPS and it returns a valid
   delegation stateid for an OPEN operation which sets the
   OPEN4_SHARE_ACCESS_WANT_DELEG_TIMESTAMPS flag, then it MUST query the
   client via a CB_GETATTR for the fattr4_time_deleg_access (see
   Section 5.2) attribute and fattr4_time_deleg_modify attribute (see
   Section 5.2).

Thus, we should look that the server supports proxying of times via
OPEN4_SHARE_ACCESS_WANT_DELEG_TIMESTAMPS.

We want to be extra pedantic and continue to check that FATTR4_TIME_DELEG_ACCESS
and FATTR4_TIME_DELEG_MODIFY are set. The server needs to expose both for the
client to correctly detect "Proxying of Times" support.

Signed-off-by: Roi Azarzar <roi.azarzar@vastdata.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Fixes: dcb3c20f7419 ("NFSv4: Add a capability for delegated attributes")
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoNFSv4: Fail mounts if the lease setup times out
Trond Myklebust [Mon, 9 Sep 2024 17:47:07 +0000 (13:47 -0400)]
NFSv4: Fail mounts if the lease setup times out

If the server is down when the client is trying to mount, so that the
calls to exchange_id or create_session fail, then we should allow the
mount system call to fail rather than hang and block other mount/umount
calls.

Reported-by: Oleksandr Tymoshenko <ovt@google.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agofs: nfs: fix missing refcnt by replacing folio_set_private by folio_attach_private
Zhaoyang Huang [Fri, 30 Aug 2024 03:27:47 +0000 (11:27 +0800)]
fs: nfs: fix missing refcnt by replacing folio_set_private by folio_attach_private

This patch is inspired by a code review of fs codes which aims at
folio's extra refcnt that could introduce unwanted behavious when
judging refcnt, such as[1].That is, the folio passed to
mapping_evict_folio carries the refcnts from find_lock_entries,
page_cache, corresponding to PTEs and folio's private if has. However,
current code doesn't take the refcnt for folio's private which could
have mapping_evict_folio miss the one to only PTE and lead to
call filemap_release_folio wrongly.

[1]
long mapping_evict_folio(struct address_space *mapping, struct folio *folio)
{
...
//current code will misjudge here if there is one pte on the folio which
is be deemed as the one as folio's private
        if (folio_ref_count(folio) >
                        folio_nr_pages(folio) + folio_has_private(folio) + 1)
                return 0;
        if (!filemap_release_folio(folio, 0))
                return 0;

        return remove_mapping(mapping, folio);
}

Signed-off-by: Zhaoyang Huang <zhaoyang.huang@unisoc.com>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfs: Remove obsoleted declaration for nfs_read_prepare
Gaosheng Cui [Mon, 26 Aug 2024 03:21:57 +0000 (11:21 +0800)]
nfs: Remove obsoleted declaration for nfs_read_prepare

The nfs_read_prepare() have been removed since
commit a4cdda59111f ("NFS: Create a common pgio_rpc_prepare function"),
and now it is useless, so remove it.

Signed-off-by: Gaosheng Cui <cuigaosheng1@huawei.com>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonet/sunrpc: make use of the helper macro LIST_HEAD()
Hongbo Li [Tue, 27 Aug 2024 09:52:18 +0000 (17:52 +0800)]
net/sunrpc: make use of the helper macro LIST_HEAD()

list_head can be initialized automatically with LIST_HEAD()
instead of calling INIT_LIST_HEAD(). Here we can simplify
the code.

Signed-off-by: Hongbo Li <lihongbo22@huawei.com>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoSUNRPC: clnt.c: Remove misleading comment
Siddh Raman Pant [Mon, 2 Sep 2024 09:32:48 +0000 (15:02 +0530)]
SUNRPC: clnt.c: Remove misleading comment

destroy_wait doesn't store all RPC clients. There was a list named
"all_clients" above it, which got moved to struct sunrpc_net in 2012,
but the comment was never removed.

Fixes: 70abc49b4f4a ("SUNRPC: make SUNPRC clients list per network namespace context")
Signed-off-by: Siddh Raman Pant <siddh.raman.pant@oracle.com>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoSUNRPC: convert RPC_TASK_* constants to enum
Stephen Brennan [Mon, 19 Aug 2024 15:58:59 +0000 (08:58 -0700)]
SUNRPC: convert RPC_TASK_* constants to enum

The RPC_TASK_* constants are defined as macros, which means that most
kernel builds will not contain their definitions in the debuginfo.
However, it's quite useful for debuggers to be able to view the task
state constant and interpret it correctly. Conversion to an enum will
ensure the constants are present in debuginfo and can be interpreted by
debuggers without needing to hard-code them and track their changes.

Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoSUNRPC: Fix -Wformat-truncation warning
Kunwu Chan [Fri, 16 Aug 2024 02:07:40 +0000 (10:07 +0800)]
SUNRPC: Fix -Wformat-truncation warning

Increase size of the servername array to avoid truncated output warning.

net/sunrpc/clnt.c:582:75: error:‘%s’ directive output may be truncated
writing up to 107 bytes into a region of size 48
[-Werror=format-truncation=]
  582 |                   snprintf(servername, sizeof(servername), "%s",
      |                                                             ^~

net/sunrpc/clnt.c:582:33: note:‘snprintf’ output
between 1 and 108 bytes into a destination of size 48
  582 |                     snprintf(servername, sizeof(servername), "%s",
      |                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  583 |                                          sun->sun_path);

Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
Suggested-by: NeilBrown <neilb@suse.de>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfs: Remove unnecessary NULL check before kfree()
Thorsten Blum [Sun, 1 Sep 2024 16:19:13 +0000 (18:19 +0200)]
nfs: Remove unnecessary NULL check before kfree()

Since kfree() already checks if its argument is NULL, an additional
check before calling kfree() is unnecessary and can be removed.

Remove it and thus also the following Coccinelle/coccicheck warning
reported by ifnullfree.cocci:

  WARNING: NULL check before some freeing functions is not needed

Reviewed-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Thorsten Blum <thorsten.blum@toblux.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfs: Annotate struct nfs_cache_array with __counted_by()
Thorsten Blum [Wed, 14 Aug 2024 10:01:28 +0000 (12:01 +0200)]
nfs: Annotate struct nfs_cache_array with __counted_by()

Add the __counted_by compiler attribute to the flexible array member
array to improve access bounds-checking via CONFIG_UBSAN_BOUNDS and
CONFIG_FORTIFY_SOURCE.

Increment size before adding a new struct to the array.

Signed-off-by: Thorsten Blum <thorsten.blum@toblux.com>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfs: simplify and guarantee owner uniqueness.
NeilBrown [Fri, 6 Sep 2024 02:32:03 +0000 (12:32 +1000)]
nfs: simplify and guarantee owner uniqueness.

I have evidence of an Linux NFS client getting NFS4ERR_BAD_SEQID to a
v4.0 LOCK request to a Linux server (which had fixed the problem with
RELEASE_LOCKOWNER bug fixed).
The LOCK request presented a "new" lock owner so there are two seq ids
in the request: that for the open file, and that for the new lock.
Given the context I am confident that the new lock owner was reported to
have the wrong seqid.  As lock owner identifiers are reused, the server
must still have a lock owner active which the client thinks is no longer
active.

I wasn't able to determine a root-cause but the simplest fix seems to be
to ensure lock owners are always unique much as open owners are (thanks
to a time stamp).  The easiest way to ensure uniqueness is with a 64bit
counter for each server.  That will never cycle (if updated once a
nanosecond the last 584 years.  A single NFS server would not handle
open/lock requests nearly that fast, and a Linux node is unlikely to
have an uptime approaching that).

This patch removes the 2 ida and instead uses a per-server
atomic64_t to provide uniqueness.

Note that the lock owner already encodes the id as 64 bits even though
it is a 32bit value.  So changing to a 64bit value does not change the
encoding of the lock owner.  The open owner encoding is now 4 bytes
larger.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agonfs: fix memory leak in error path of nfs4_do_reclaim
Li Lingfeng [Wed, 4 Sep 2024 12:34:57 +0000 (20:34 +0800)]
nfs: fix memory leak in error path of nfs4_do_reclaim

Commit c77e22834ae9 ("NFSv4: Fix a potential sleep while atomic in
nfs4_do_reclaim()") separate out the freeing of the state owners from
nfs4_purge_state_owners() and finish it outside the rcu lock.
However, the error path is omitted. As a result, the state owners in
"freeme" will not be released.
Fix it by adding freeing in the error path.

Fixes: c77e22834ae9 ("NFSv4: Fix a potential sleep while atomic in nfs4_do_reclaim()")
Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Cc: stable@vger.kernel.org # v5.3+
Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com>
3 days agoMerge tag 'nfsd-6.12' into linux-next-with-localio
Anna Schumaker [Mon, 23 Sep 2024 19:00:07 +0000 (15:00 -0400)]
Merge tag 'nfsd-6.12' into linux-next-with-localio

NFSD 6.12 Release Notes

Notable features of this release include:

- Pre-requisites for automatically determining the RPC server thread
  count
- Clean-up and preparation for supporting LOCALIO, which will be
  merged via the NFS client tree
- Enhancements and fixes to NFSv4.2 COPY offload
- A new Python-based tool for generating kernel SunRPC XDR encoding
  and decoding functions, added as an aid for prototyping features
  in protocols based on the Linux kernel's SunRPC implementation.

As always I am grateful to the NFSD contributors, reviewers,
testers, and bug reporters who participated during this cycle.

6 days agoxdrgen: Prevent reordering of encoder and decoder functions
Chuck Lever [Fri, 13 Sep 2024 17:50:56 +0000 (13:50 -0400)]
xdrgen: Prevent reordering of encoder and decoder functions

I noticed that "xdrgen source" reorders the procedure encoder and
decoder functions every time it is run. I would prefer that the
generated code be more deterministic: it enables a reader to better
see exactly what has changed between runs of the tool.

The problem is that Python sets are not ordered. I use a Python set
to ensure that, when multiple procedures use a particular argument or
result type, the encoder/decoder for that type is emitted only once.

Sets aren't ordered, but I can use Python dictionaries for this
purpose to ensure the procedure functions are always emitted in the
same order if the .x file does not change.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoxdrgen: typedefs should use the built-in string and opaque functions
Chuck Lever [Tue, 10 Sep 2024 19:51:46 +0000 (15:51 -0400)]
xdrgen: typedefs should use the built-in string and opaque functions

'typedef opaque yada<XYZ>' should use xdrgen's built-in opaque
encoder and decoder, to enable better compiler optimization.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoxdrgen: Fix return code checking in built-in XDR decoders
Chuck Lever [Tue, 10 Sep 2024 19:31:19 +0000 (15:31 -0400)]
xdrgen: Fix return code checking in built-in XDR decoders

xdr_stream_encode_u32() returns XDR_UNIT on success.
xdr_stream_decode_u32() returns zero or -EMSGSIZE, but never
XDR_UNIT.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agotools: Add xdrgen
Chuck Lever [Fri, 13 Sep 2024 18:08:13 +0000 (14:08 -0400)]
tools: Add xdrgen

Add a Python-based tool for translating XDR specifications into XDR
encoder and decoder functions written in the Linux kernel's C coding
style. The generator attempts to match the usual C coding style of
the Linux kernel's SunRPC consumers.

This approach is similar to the netlink code generator in
tools/net/ynl .

The maintainability benefits of machine-generated XDR code include:

- Stronger type checking
- Reduces the number of bugs introduced by human error
- Makes the XDR code easier to audit and analyze
- Enables rapid prototyping of new RPC-based protocols
- Hardens the layering between protocol logic and marshaling
- Makes it easier to add observability on demand
- Unit tests might be built for both the tool and (automatically)
  for the generated code

In addition, converting the XDR layer to use memory-safe languages
such as Rust will be easier if much of the code can be converted
automatically.

Tested-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: fix delegation_blocked() to block correctly for at least 30 seconds
NeilBrown [Mon, 9 Sep 2024 05:06:36 +0000 (15:06 +1000)]
nfsd: fix delegation_blocked() to block correctly for at least 30 seconds

The pair of bloom filtered used by delegation_blocked() was intended to
block delegations on given filehandles for between 30 and 60 seconds.  A
new filehandle would be recorded in the "new" bit set.  That would then
be switch to the "old" bit set between 0 and 30 seconds later, and it
would remain as the "old" bit set for 30 seconds.

Unfortunately the code intended to clear the old bit set once it reached
30 seconds old, preparing it to be the next new bit set, instead cleared
the *new* bit set before switching it to be the old bit set.  This means
that the "old" bit set is always empty and delegations are blocked
between 0 and 30 seconds.

This patch updates bd->new before clearing the set with that index,
instead of afterwards.

Reported-by: Olga Kornievskaia <okorniev@redhat.com>
Cc: stable@vger.kernel.org
Fixes: 6282cd565553 ("NFSD: Don't hand out delegations for 30 seconds after recalling them.")
Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Benjamin Coddington <bcodding@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: fix initial getattr on write delegation
Jeff Layton [Mon, 9 Sep 2024 14:40:53 +0000 (10:40 -0400)]
nfsd: fix initial getattr on write delegation

At this point in compound processing, currentfh refers to the parent of
the file, not the file itself. Get the correct dentry from the delegation
stateid instead.

Fixes: c5967721e106 ("NFSD: handle GETATTR conflict with write delegation")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: untangle code in nfsd4_deleg_getattr_conflict()
NeilBrown [Thu, 29 Aug 2024 13:26:40 +0000 (09:26 -0400)]
nfsd: untangle code in nfsd4_deleg_getattr_conflict()

The code in nfsd4_deleg_getattr_conflict() is convoluted and buggy.

With this patch we:
 - properly handle non-nfsd leases.  We must not assume flc_owner is a
    delegation unless fl_lmops == &nfsd_lease_mng_ops
 - move the main code out of the for loop
 - have a single exit which calls nfs4_put_stid()
   (and other exits which don't need to call that)

[ jlayton: refactored on top of Neil's other patch: nfsd: fix
   nfsd4_deleg_getattr_conflict in presence of third party lease ]

Fixes: c5967721e106 ("NFSD: handle GETATTR conflict with write delegation")
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: enforce upper limit for namelen in __cld_pipe_inprogress_downcall()
Scott Mayhew [Mon, 9 Sep 2024 20:28:54 +0000 (16:28 -0400)]
nfsd: enforce upper limit for namelen in __cld_pipe_inprogress_downcall()

This patch is intended to go on top of "nfsd: return -EINVAL when
namelen is 0" from Li Lingfeng.  Li's patch checks for 0, but we should
be enforcing an upper bound as well.

Note that if nfsdcld somehow gets an id > NFS4_OPAQUE_LIMIT in its
database, it'll truncate it to NFS4_OPAQUE_LIMIT when it does the
downcall anyway.

Signed-off-by: Scott Mayhew <smayhew@redhat.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: return -EINVAL when namelen is 0
Li Lingfeng [Tue, 3 Sep 2024 11:14:46 +0000 (19:14 +0800)]
nfsd: return -EINVAL when namelen is 0

When we have a corrupted main.sqlite in /var/lib/nfs/nfsdcld/, it may
result in namelen being 0, which will cause memdup_user() to return
ZERO_SIZE_PTR.
When we access the name.data that has been assigned the value of
ZERO_SIZE_PTR in nfs4_client_to_reclaim(), null pointer dereference is
triggered.

[ T1205] ==================================================================
[ T1205] BUG: KASAN: null-ptr-deref in nfs4_client_to_reclaim+0xe9/0x260
[ T1205] Read of size 1 at addr 0000000000000010 by task nfsdcld/1205
[ T1205]
[ T1205] CPU: 11 PID: 1205 Comm: nfsdcld Not tainted 5.10.0-00003-g2c1423731b8d #406
[ T1205] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS ?-20190727_073836-buildvm-ppc64le-16.ppc.fedoraproject.org-3.fc31 04/01/2014
[ T1205] Call Trace:
[ T1205]  dump_stack+0x9a/0xd0
[ T1205]  ? nfs4_client_to_reclaim+0xe9/0x260
[ T1205]  __kasan_report.cold+0x34/0x84
[ T1205]  ? nfs4_client_to_reclaim+0xe9/0x260
[ T1205]  kasan_report+0x3a/0x50
[ T1205]  nfs4_client_to_reclaim+0xe9/0x260
[ T1205]  ? nfsd4_release_lockowner+0x410/0x410
[ T1205]  cld_pipe_downcall+0x5ca/0x760
[ T1205]  ? nfsd4_cld_tracking_exit+0x1d0/0x1d0
[ T1205]  ? down_write_killable_nested+0x170/0x170
[ T1205]  ? avc_policy_seqno+0x28/0x40
[ T1205]  ? selinux_file_permission+0x1b4/0x1e0
[ T1205]  rpc_pipe_write+0x84/0xb0
[ T1205]  vfs_write+0x143/0x520
[ T1205]  ksys_write+0xc9/0x170
[ T1205]  ? __ia32_sys_read+0x50/0x50
[ T1205]  ? ktime_get_coarse_real_ts64+0xfe/0x110
[ T1205]  ? ktime_get_coarse_real_ts64+0xa2/0x110
[ T1205]  do_syscall_64+0x33/0x40
[ T1205]  entry_SYSCALL_64_after_hwframe+0x67/0xd1
[ T1205] RIP: 0033:0x7fdbdb761bc7
[ T1205] Code: 0f 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b7 0f 1f 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 514
[ T1205] RSP: 002b:00007fff8c4b7248 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
[ T1205] RAX: ffffffffffffffda RBX: 000000000000042b RCX: 00007fdbdb761bc7
[ T1205] RDX: 000000000000042b RSI: 00007fff8c4b75f0 RDI: 0000000000000008
[ T1205] RBP: 00007fdbdb761bb0 R08: 0000000000000000 R09: 0000000000000001
[ T1205] R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000042b
[ T1205] R13: 0000000000000008 R14: 00007fff8c4b75f0 R15: 0000000000000000
[ T1205] ==================================================================

Fix it by checking namelen.

Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Fixes: 74725959c33c ("nfsd: un-deprecate nfsdcld")
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Scott Mayhew <smayhew@redhat.com>
Tested-by: Scott Mayhew <smayhew@redhat.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoNFSD: Wrap async copy operations with trace points
Chuck Lever [Wed, 28 Aug 2024 17:40:09 +0000 (13:40 -0400)]
NFSD: Wrap async copy operations with trace points

Add an nfsd_copy_async_done to record the timestamp, the final
status code, and the callback stateid of an async copy.

Rename the nfsd_copy_do_async tracepoint to match that naming
convention to make it easier to enable both of these with a
single glob.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoNFSD: Clean up extra whitespace in trace_nfsd_copy_done
Chuck Lever [Wed, 28 Aug 2024 17:40:07 +0000 (13:40 -0400)]
NFSD: Clean up extra whitespace in trace_nfsd_copy_done

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoNFSD: Record the callback stateid in copy tracepoints
Chuck Lever [Wed, 28 Aug 2024 17:40:06 +0000 (13:40 -0400)]
NFSD: Record the callback stateid in copy tracepoints

Match COPY operations up with CB_OFFLOAD operations.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoNFSD: Display copy stateids with conventional print formatting
Chuck Lever [Wed, 28 Aug 2024 17:40:05 +0000 (13:40 -0400)]
NFSD: Display copy stateids with conventional print formatting

Make it easier to grep for s2s COPY stateids in trace logs: Use the
same display format in nfsd_copy_class as is used to display other
stateids.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoNFSD: Limit the number of concurrent async COPY operations
Chuck Lever [Wed, 28 Aug 2024 17:40:04 +0000 (13:40 -0400)]
NFSD: Limit the number of concurrent async COPY operations

Nothing appears to limit the number of concurrent async COPY
operations that clients can start. In addition, AFAICT each async
COPY can copy an unlimited number of 4MB chunks, so can run for a
long time. Thus IMO async COPY can become a DoS vector.

Add a restriction mechanism that bounds the number of concurrent
background COPY operations. Start simple and try to be fair -- this
patch implements a per-namespace limit.

An async COPY request that occurs while this limit is exceeded gets
NFS4ERR_DELAY. The requesting client can choose to send the request
again after a delay or fall back to a traditional read/write style
copy.

If there is need to make the mechanism more sophisticated, we can
visit that in future patches.

Cc: stable@vger.kernel.org
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoNFSD: Async COPY result needs to return a write verifier
Chuck Lever [Wed, 28 Aug 2024 17:40:03 +0000 (13:40 -0400)]
NFSD: Async COPY result needs to return a write verifier

Currently, when NFSD handles an asynchronous COPY, it returns a
zero write verifier, relying on the subsequent CB_OFFLOAD callback
to pass the write verifier and a stable_how4 value to the client.

However, if the CB_OFFLOAD never arrives at the client (for example,
if a network partition occurs just as the server sends the
CB_OFFLOAD operation), the client will never receive this verifier.
Thus, if the client sends a follow-up COMMIT, there is no way for
the client to assess the COMMIT result.

The usual recovery for a missing CB_OFFLOAD is for the client to
send an OFFLOAD_STATUS operation, but that operation does not carry
a write verifier in its result. Neither does it carry a stable_how4
value, so the client /must/ send a COMMIT in this case -- which will
always fail because currently there's still no write verifier in the
COPY result.

Thus the server needs to return a normal write verifier in its COPY
result even if the COPY operation is to be performed asynchronously.

If the server recognizes the callback stateid in subsequent
OFFLOAD_STATUS operations, then obviously it has not restarted, and
the write verifier the client received in the COPY result is still
valid and can be used to assess a COMMIT of the copied data, if one
is needed.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: avoid races with wake_up_var()
NeilBrown [Fri, 30 Aug 2024 07:03:17 +0000 (17:03 +1000)]
nfsd: avoid races with wake_up_var()

wake_up_var() needs a barrier after the important change is made in the
var and before wake_up_var() is called, else it is possible that a wake
up won't be sent when it should.

In each case here the var is changed in an "atomic" manner, so
smb_mb__after_atomic() is sufficient.

In one case the important change (removing the lease) is performed
*after* the wake_up, which is backwards.  The code survives in part
because the wait_var_event is given a timeout.

This patch adds the required barriers and calls destroy_delegation()
*before* waking any threads waiting for the delegation to be destroyed.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: use clear_and_wake_up_bit()
NeilBrown [Fri, 30 Aug 2024 07:03:16 +0000 (17:03 +1000)]
nfsd: use clear_and_wake_up_bit()

nfsd has two places that open-code clear_and_wake_up_bit().  One has
the required memory barriers.  The other does not.

Change both to use clear_and_wake_up_bit() so we have the barriers
without the noise.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agosunrpc: xprtrdma: Use ERR_CAST() to return
Yan Zhen [Fri, 30 Aug 2024 01:43:56 +0000 (09:43 +0800)]
sunrpc: xprtrdma: Use ERR_CAST() to return

Using ERR_CAST() is more reasonable and safer, When it is necessary
to convert the type of an error pointer and return it.

Signed-off-by: Yan Zhen <yanzhen@vivo.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoNFSD: Annotate struct pnfs_block_deviceaddr with __counted_by()
Thorsten Blum [Wed, 28 Aug 2024 21:42:55 +0000 (23:42 +0200)]
NFSD: Annotate struct pnfs_block_deviceaddr with __counted_by()

Add the __counted_by compiler attribute to the flexible array member
volumes to improve access bounds-checking via CONFIG_UBSAN_BOUNDS and
CONFIG_FORTIFY_SOURCE.

Use struct_size() instead of manually calculating the number of bytes to
allocate for a pnfs_block_deviceaddr with a single volume.

Signed-off-by: Thorsten Blum <thorsten.blum@toblux.com>
Reviewed-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Acked-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: call cache_put if xdr_reserve_space returns NULL
Guoqing Jiang [Wed, 21 Aug 2024 14:03:18 +0000 (22:03 +0800)]
nfsd: call cache_put if xdr_reserve_space returns NULL

If not enough buffer space available, but idmap_lookup has triggered
lookup_fn which calls cache_get and returns successfully. Then we
missed to call cache_put here which pairs with cache_get.

Fixes: ddd1ea563672 ("nfsd4: use xdr_reserve_space in attribute encoding")
Signed-off-by: Guoqing Jiang <guoqing.jiang@linux.dev>
Reviwed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: add more nfsd_cb tracepoints
Jeff Layton [Mon, 26 Aug 2024 12:50:13 +0000 (08:50 -0400)]
nfsd: add more nfsd_cb tracepoints

Add some tracepoints in the callback client RPC operations. Also
add a tracepoint to nfsd4_cb_getattr_done.

Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: track the main opcode for callbacks
Jeff Layton [Mon, 26 Aug 2024 12:50:12 +0000 (08:50 -0400)]
nfsd: track the main opcode for callbacks

Keep track of the "main" opcode for the callback, and display it in the
tracepoint. This makes it simpler to discern what's happening when there
is more than one callback in flight.

The one special case is the CB_NULL RPC. That's not a CB_COMPOUND
opcode, so designate the value 0 for that.

Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: add more info to WARN_ON_ONCE on failed callbacks
Jeff Layton [Mon, 26 Aug 2024 12:50:11 +0000 (08:50 -0400)]
nfsd: add more info to WARN_ON_ONCE on failed callbacks

Currently, you get the warning and stack trace, but nothing is printed
about the relevant error codes. Add that in.

Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: fix some spelling errors in comments
Li Lingfeng [Sat, 24 Aug 2024 01:43:36 +0000 (09:43 +0800)]
nfsd: fix some spelling errors in comments

Fix spelling errors in comments of nfsd4_release_lockowner and
nfs4_set_delegation.

Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: remove unused parameter of nfsd_file_mark_find_or_create
Li Lingfeng [Fri, 23 Aug 2024 07:00:49 +0000 (15:00 +0800)]
nfsd: remove unused parameter of nfsd_file_mark_find_or_create

Commit 427f5f83a319 ("NFSD: Ensure nf_inode is never dereferenced") passes
inode directly to nfsd_file_mark_find_or_create instead of getting it from
nf, so there is no need to pass nf.

Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: use LIST_HEAD() to simplify code
Hongbo Li [Wed, 21 Aug 2024 06:53:26 +0000 (14:53 +0800)]
nfsd: use LIST_HEAD() to simplify code

list_head can be initialized automatically with LIST_HEAD()
instead of calling INIT_LIST_HEAD().

Signed-off-by: Hongbo Li <lihongbo22@huawei.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: map the EBADMSG to nfserr_io to avoid warning
Li Lingfeng [Sat, 17 Aug 2024 06:27:13 +0000 (14:27 +0800)]
nfsd: map the EBADMSG to nfserr_io to avoid warning

Ext4 will throw -EBADMSG through ext4_readdir when a checksum error
occurs, resulting in the following WARNING.

Fix it by mapping EBADMSG to nfserr_io.

nfsd_buffered_readdir
 iterate_dir // -EBADMSG -74
  ext4_readdir // .iterate_shared
   ext4_dx_readdir
    ext4_htree_fill_tree
     htree_dirblock_to_tree
      ext4_read_dirblock
       __ext4_read_dirblock
        ext4_dirblock_csum_verify
         warn_no_space_for_csum
          __warn_no_space_for_csum
        return ERR_PTR(-EFSBADCRC) // -EBADMSG -74
 nfserrno // WARNING

[  161.115610] ------------[ cut here ]------------
[  161.116465] nfsd: non-standard errno: -74
[  161.117315] WARNING: CPU: 1 PID: 780 at fs/nfsd/nfsproc.c:878 nfserrno+0x9d/0xd0
[  161.118596] Modules linked in:
[  161.119243] CPU: 1 PID: 780 Comm: nfsd Not tainted 5.10.0-00014-g79679361fd5d #138
[  161.120684] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qe
mu.org 04/01/2014
[  161.123601] RIP: 0010:nfserrno+0x9d/0xd0
[  161.124676] Code: 0f 87 da 30 dd 00 83 e3 01 b8 00 00 00 05 75 d7 44 89 ee 48 c7 c7 c0 57 24 98 89 44 24 04 c6
 05 ce 2b 61 03 01 e8 99 20 d8 00 <0f> 0b 8b 44 24 04 eb b5 4c 89 e6 48 c7 c7 a0 6d a4 99 e8 cc 15 33
[  161.127797] RSP: 0018:ffffc90000e2f9c0 EFLAGS: 00010286
[  161.128794] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
[  161.130089] RDX: 1ffff1103ee16f6d RSI: 0000000000000008 RDI: fffff520001c5f2a
[  161.131379] RBP: 0000000000000022 R08: 0000000000000001 R09: ffff8881f70c1827
[  161.132664] R10: ffffed103ee18304 R11: 0000000000000001 R12: 0000000000000021
[  161.133949] R13: 00000000ffffffb6 R14: ffff8881317c0000 R15: ffffc90000e2fbd8
[  161.135244] FS:  0000000000000000(0000) GS:ffff8881f7080000(0000) knlGS:0000000000000000
[  161.136695] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  161.137761] CR2: 00007fcaad70b348 CR3: 0000000144256006 CR4: 0000000000770ee0
[  161.139041] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  161.140291] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[  161.141519] PKRU: 55555554
[  161.142076] Call Trace:
[  161.142575]  ? __warn+0x9b/0x140
[  161.143229]  ? nfserrno+0x9d/0xd0
[  161.143872]  ? report_bug+0x125/0x150
[  161.144595]  ? handle_bug+0x41/0x90
[  161.145284]  ? exc_invalid_op+0x14/0x70
[  161.146009]  ? asm_exc_invalid_op+0x12/0x20
[  161.146816]  ? nfserrno+0x9d/0xd0
[  161.147487]  nfsd_buffered_readdir+0x28b/0x2b0
[  161.148333]  ? nfsd4_encode_dirent_fattr+0x380/0x380
[  161.149258]  ? nfsd_buffered_filldir+0xf0/0xf0
[  161.150093]  ? wait_for_concurrent_writes+0x170/0x170
[  161.151004]  ? generic_file_llseek_size+0x48/0x160
[  161.151895]  nfsd_readdir+0x132/0x190
[  161.152606]  ? nfsd4_encode_dirent_fattr+0x380/0x380
[  161.153516]  ? nfsd_unlink+0x380/0x380
[  161.154256]  ? override_creds+0x45/0x60
[  161.155006]  nfsd4_encode_readdir+0x21a/0x3d0
[  161.155850]  ? nfsd4_encode_readlink+0x210/0x210
[  161.156731]  ? write_bytes_to_xdr_buf+0x97/0xe0
[  161.157598]  ? __write_bytes_to_xdr_buf+0xd0/0xd0
[  161.158494]  ? lock_downgrade+0x90/0x90
[  161.159232]  ? nfs4svc_decode_voidarg+0x10/0x10
[  161.160092]  nfsd4_encode_operation+0x15a/0x440
[  161.160959]  nfsd4_proc_compound+0x718/0xe90
[  161.161818]  nfsd_dispatch+0x18e/0x2c0
[  161.162586]  svc_process_common+0x786/0xc50
[  161.163403]  ? nfsd_svc+0x380/0x380
[  161.164137]  ? svc_printk+0x160/0x160
[  161.164846]  ? svc_xprt_do_enqueue.part.0+0x365/0x380
[  161.165808]  ? nfsd_svc+0x380/0x380
[  161.166523]  ? rcu_is_watching+0x23/0x40
[  161.167309]  svc_process+0x1a5/0x200
[  161.168019]  nfsd+0x1f5/0x380
[  161.168663]  ? nfsd_shutdown_threads+0x260/0x260
[  161.169554]  kthread+0x1c4/0x210
[  161.170224]  ? kthread_insert_work_sanity_check+0x80/0x80
[  161.171246]  ret_from_fork+0x1f/0x30

Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Cc: stable@vger.kernel.org
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoNFSD: remove redundant assignment operation
Li Lingfeng [Wed, 14 Aug 2024 11:29:07 +0000 (19:29 +0800)]
NFSD: remove redundant assignment operation

Commit 5826e09bf3dd ("NFSD: OP_CB_RECALL_ANY should recall both read and
write delegations") added a new assignment statement to add
RCA4_TYPE_MASK_WDATA_DLG to ra_bmval bitmask of OP_CB_RECALL_ANY. So the
old one should be removed.

Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days ago.mailmap: Add an entry for my work email address
Chuck Lever [Tue, 13 Aug 2024 21:55:45 +0000 (17:55 -0400)]
.mailmap: Add an entry for my work email address

Collect a few very old previous employers as well.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agoNFSD: Fix NFSv4's PUTPUBFH operation
Chuck Lever [Sun, 11 Aug 2024 17:11:07 +0000 (13:11 -0400)]
NFSD: Fix NFSv4's PUTPUBFH operation

According to RFC 8881, all minor versions of NFSv4 support PUTPUBFH.

Replace the XDR decoder for PUTPUBFH with a "noop" since we no
longer want the minorversion check, and PUTPUBFH has no arguments to
decode. (Ideally nfsd4_decode_noop should really be called
nfsd4_decode_void).

PUTPUBFH should now behave just like PUTROOTFH.

Reported-by: Cedric Blancher <cedric.blancher@gmail.com>
Fixes: e1a90ebd8b23 ("NFSD: Combine decode operations for v4 and v4.1")
Cc: Dan Shelton <dan.f.shelton@gmail.com>
Cc: Roland Mainz <roland.mainz@nrubsig.org>
Cc: stable@vger.kernel.org
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: Add quotes to client info 'callback address'
Mark Grimes [Wed, 7 Aug 2024 01:58:34 +0000 (18:58 -0700)]
nfsd: Add quotes to client info 'callback address'

The 'callback address' in client_info_show is output without quotes
causing yaml parsers to fail on processing IPv6 addresses.
Adding quotes to 'callback address' also matches that used by
the 'address' field.

Signed-off-by: Mark Grimes <mark.grimes@ixsystems.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agosvcrdma: Handle device removal outside of the CM event handler
Chuck Lever [Mon, 29 Jul 2024 20:52:32 +0000 (16:52 -0400)]
svcrdma: Handle device removal outside of the CM event handler

Synchronously wait for all disconnects to complete to ensure the
transports have divested all hardware resources before the
underlying RDMA device can safely be removed.

Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: move error choice for incorrect object types to version-specific code.
NeilBrown [Wed, 14 Aug 2024 13:21:01 +0000 (09:21 -0400)]
nfsd: move error choice for incorrect object types to version-specific code.

If an NFS operation expects a particular sort of object (file, dir, link,
etc) but gets a file handle for a different sort of object, it must
return an error.  The actual error varies among NFS versions in non-trivial
ways.

For v2 and v3 there are ISDIR and NOTDIR errors and, for NFSv4 only,
INVAL is suitable.

For v4.0 there is also NFS4ERR_SYMLINK which should be used if a SYMLINK
was found when not expected.  This take precedence over NOTDIR.

For v4.1+ there is also NFS4ERR_WRONG_TYPE which should be used in
preference to EINVAL when none of the specific error codes apply.

When nfsd_mode_check() finds a symlink where it expected a directory it
needs to return an error code that can be converted to NOTDIR for v2 or
v3 but will be SYMLINK for v4.  It must be different from the error
code returns when it finds a symlink but expects a regular file - that
must be converted to EINVAL or SYMLINK.

So we introduce an internal error code nfserr_symlink_not_dir which each
version converts as appropriate.

nfsd_check_obj_isreg() is similar to nfsd_mode_check() except that it is
only used by NFSv4 and only for OPEN.  NFSERR_INVAL is never a suitable
error if the object is the wrong time.  For v4.0 we use nfserr_symlink
for non-dirs even if not a symlink.  For v4.1 we have nfserr_wrong_type.
We handle this difference in-place in nfsd_check_obj_isreg() as there is
nothing to be gained by delaying the choice to nfsd4_map_status().

As a result of these changes, nfsd_mode_check() doesn't need an rqstp
arg any more.

Note that NFSv4 operations are actually performed in the xdr code(!!!)
so to the only place that we can map the status code successfully is in
nfsd4_encode_operation().

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: be more systematic about selecting error codes for internal use.
NeilBrown [Mon, 29 Jul 2024 01:47:23 +0000 (11:47 +1000)]
nfsd: be more systematic about selecting error codes for internal use.

Rather than using ad hoc values for internal errors (30000, 11000, ...)
use 'enum' to sequentially allocate numbers starting from the first
known available number - now visible as NFS4ERR_FIRST_FREE.

The goal is values that are distinct from all be32 error codes.  To get
those we must first select integers that are not already used, then
convert them with cpu_to_be32().

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: Move error code mapping to per-version proc code.
NeilBrown [Mon, 29 Jul 2024 01:47:22 +0000 (11:47 +1000)]
nfsd: Move error code mapping to per-version proc code.

There is code scattered around nfsd which chooses an error status based
on the particular version of nfs being used.  It is cleaner to have the
version specific choices in version specific code.

With this patch common code returns the most specific error code
possible and the version specific code maps that if necessary.

Both v2 (nfsproc.c) and v3 (nfs3proc.c) now have a "map_status()"
function which is called to map the resp->status before each non-trivial
nfsd_proc_* or nfsd3_proc_* function returns.

NFS4ERR_SYMLINK and NFS4ERR_WRONG_TYPE introduce extra complications and
are left for a later patch.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: move V4ROOT version check to nfsd_set_fh_dentry()
NeilBrown [Fri, 26 Jul 2024 02:21:35 +0000 (12:21 +1000)]
nfsd: move V4ROOT version check to nfsd_set_fh_dentry()

This further centralizes version number checks.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: further centralize protocol version checks.
NeilBrown [Fri, 26 Jul 2024 02:21:34 +0000 (12:21 +1000)]
nfsd: further centralize protocol version checks.

With this patch the only places that test ->rq_vers against a specific
version are nfsd_v4client() and nfsd_set_fh_dentry().
The latter sets some flags in the svc_fh, which now includes:
  fh_64bit_cookies
  fh_use_wgather

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: use nfsd_v4client() in nfsd_breaker_owns_lease()
NeilBrown [Fri, 26 Jul 2024 02:21:33 +0000 (12:21 +1000)]
nfsd: use nfsd_v4client() in nfsd_breaker_owns_lease()

nfsd_breaker_owns_lease() currently open-codes the same test that
nfsd_v4client() performs.

With this patch we use nfsd_v4client() instead.

Also as i_am_nfsd() is only used in combination with kthread_data(),
replace it with nfsd_current_rqst() which combines the two and returns a
valid svc_rqst, or NULL.

The test for NULL is moved into nfsd_v4client() for code clarity.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: Pass 'cred' instead of 'rqstp' to some functions.
NeilBrown [Fri, 26 Jul 2024 02:21:31 +0000 (12:21 +1000)]
nfsd: Pass 'cred' instead of 'rqstp' to some functions.

nfsd_permission(), exp_rdonly(), nfsd_setuser(), and nfsexp_flags()
only ever need the cred out of rqstp, so pass it explicitly instead of
the whole rqstp.

This makes the interfaces cleaner.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: Don't pass all of rqst into rqst_exp_find()
NeilBrown [Fri, 26 Jul 2024 02:21:30 +0000 (12:21 +1000)]
nfsd: Don't pass all of rqst into rqst_exp_find()

Rather than passing the whole rqst, pass the pieces that are actually
needed.  This makes the inputs to rqst_exp_find() more obvious.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: don't assume copy notify when preprocessing the stateid
Sagi Grimberg [Wed, 24 Jul 2024 17:01:37 +0000 (10:01 -0700)]
nfsd: don't assume copy notify when preprocessing the stateid

Move the stateid handling to nfsd4_copy_notify.
If nfs4_preprocess_stateid_op did not produce an output stateid, error out.

Copy notify specifically does not permit the use of special stateids,
so enforce that outside generic stateid pre-processing.

Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Olga Kornievskaia <aglo@umich.edu>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agosunrpc: allow svc threads to fail initialisation cleanly
NeilBrown [Sun, 15 Sep 2024 23:45:40 +0000 (09:45 +1000)]
sunrpc: allow svc threads to fail initialisation cleanly

If an svc thread needs to perform some initialisation that might fail,
it has no good way to handle the failure.

Before the thread can exit it must call svc_exit_thread(), but that
requires the service mutex to be held.  The thread cannot simply take
the mutex as that could deadlock if there is a concurrent attempt to
shut down all threads (which is unlikely, but not impossible).

nfsd currently call svc_exit_thread() unprotected in the unlikely event
that unshare_fs_struct() fails.

We can clean this up by introducing svc_thread_init_status() by which an
svc thread can report whether initialisation has succeeded.  If it has,
it continues normally into the action loop.  If it has not,
svc_thread_init_status() immediately aborts the thread.
svc_start_kthread() waits for either of these to happen, and calls
svc_exit_thread() (under the mutex) if the thread aborted.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agosunrpc: merge svc_rqst_alloc() into svc_prepare_thread()
NeilBrown [Mon, 29 Jul 2024 21:19:41 +0000 (07:19 +1000)]
sunrpc: merge svc_rqst_alloc() into svc_prepare_thread()

The only caller of svc_rqst_alloc() is svc_prepare_thread().  So merge
the one into the other and simplify.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agosunrpc: don't take ->sv_lock when updating ->sv_nrthreads.
NeilBrown [Mon, 15 Jul 2024 07:14:19 +0000 (17:14 +1000)]
sunrpc: don't take ->sv_lock when updating ->sv_nrthreads.

As documented in svc_xprt.c, sv_nrthreads is protected by the service
mutex, and it does not need ->sv_lock.
(->sv_lock is needed only for sv_permsocks, sv_tempsocks, and
sv_tmpcnt).

So remove the unnecessary locking.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agosunrpc: change sp_nrthreads from atomic_t to unsigned int.
NeilBrown [Mon, 15 Jul 2024 07:14:18 +0000 (17:14 +1000)]
sunrpc: change sp_nrthreads from atomic_t to unsigned int.

sp_nrthreads is only ever accessed under the service mutex
  nlmsvc_mutex nfs_callback_mutex nfsd_mutex
so these is no need for it to be an atomic_t.

The fact that all code using it is single-threaded means that we can
simplify svc_pool_victim and remove the temporary elevation of
sp_nrthreads.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agosunrpc: document locking rules for svc_exit_thread()
NeilBrown [Fri, 26 Jul 2024 03:59:55 +0000 (13:59 +1000)]
sunrpc: document locking rules for svc_exit_thread()

The locking required for svc_exit_thread() is not obvious, so document
it in a kdoc comment.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
6 days agonfsd: don't allocate the versions array.
NeilBrown [Wed, 7 Aug 2024 01:11:32 +0000 (21:11 -0400)]
nfsd: don't allocate the versions array.

Instead of using kmalloc to allocate an array for storing active version
info, just declare an array to the max size - it is only 5 or so.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
2 weeks agoLinux 6.11-rc7
Linus Torvalds [Sun, 8 Sep 2024 21:50:28 +0000 (14:50 -0700)]
Linux 6.11-rc7

2 weeks agoMerge tag 'timers_urgent_for_v6.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 8 Sep 2024 17:25:25 +0000 (10:25 -0700)]
Merge tag 'timers_urgent_for_v6.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull timer fixes from Borislav Petkov:

 - Remove percpu irq related code in the timer-of initialization routine
   as it is broken but also unused (Daniel Lezcano)

 - Fix return -ETIME when delta exceeds INT_MAX and the next event not
   taking effect sometimes (Jacky Bai)

* tag 'timers_urgent_for_v6.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  clocksource/drivers/imx-tpm: Fix next event not taking effect sometime
  clocksource/drivers/imx-tpm: Fix return -ETIME when delta exceeds INT_MAX
  clocksource/drivers/timer-of: Remove percpu irq related code

2 weeks agoMerge tag 'perf_urgent_for_v6.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 8 Sep 2024 17:20:44 +0000 (10:20 -0700)]
Merge tag 'perf_urgent_for_v6.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Borislav Petkov:

 - Fix perf's AUX buffer serialization

 - Prevent uninitialized struct members in perf's uprobes handling

* tag 'perf_urgent_for_v6.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/aux: Fix AUX buffer serialization
  uprobes: Use kzalloc to allocate xol area

2 weeks agoMerge tag 'char-misc-6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregk...
Linus Torvalds [Sun, 8 Sep 2024 17:13:39 +0000 (10:13 -0700)]
Merge tag 'char-misc-6.11-rc7' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are some small char/misc/other driver fixes for 6.11-rc7. It's
  nothing huge, just a bunch of small fixes of reported problems,
  including:

   - lots of tiny iio driver fixes

   - nvmem driver fixex

   - binder UAF bugfix

   - uio driver crash fix

   - other small fixes

  All of these have been in linux-next this week with no reported
  problems"

* tag 'char-misc-6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc: (21 commits)
  VMCI: Fix use-after-free when removing resource in vmci_resource_remove()
  Drivers: hv: vmbus: Fix rescind handling in uio_hv_generic
  uio_hv_generic: Fix kernel NULL pointer dereference in hv_uio_rescind
  misc: keba: Fix sysfs group creation
  dt-bindings: nvmem: Use soc-nvmem node name instead of nvmem
  nvmem: Fix return type of devm_nvmem_device_get() in kerneldoc
  nvmem: u-boot-env: error if NVMEM device is too small
  misc: fastrpc: Fix double free of 'buf' in error path
  binder: fix UAF caused by offsets overwrite
  iio: imu: inv_mpu6050: fix interrupt status read for old buggy chips
  iio: adc: ad7173: fix GPIO device info
  iio: adc: ad7124: fix DT configuration parsing
  iio: adc: ad_sigma_delta: fix irq_flags on irq request
  iio: adc: ads1119: Fix IRQ flags
  iio: fix scale application in iio_convert_raw_to_processed_unlocked
  iio: adc: ad7124: fix config comparison
  iio: adc: ad7124: fix chip ID mismatch
  iio: adc: ad7173: Fix incorrect compatible string
  iio: buffer-dmaengine: fix releasing dma channel on error
  iio: adc: ad7606: remove frstdata check for serial mode
  ...

2 weeks agoMerge tag 'usb-6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sun, 8 Sep 2024 17:07:28 +0000 (10:07 -0700)]
Merge tag 'usb-6.11-rc7' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are a handful of small USB fixes for 6.11-rc7. Included in here
  are:

   - dwc3 driver fixes for two reported problems

   - two typec ucsi driver fixes

   - cdns2 controller reset fix

  All of these have been in linux-next this week with no reported
  problems"

* tag 'usb-6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: typec: ucsi: Fix cable registration
  usb: typec: ucsi: Fix the partner PD revision
  usb: cdns2: Fix controller reset issue
  usb: dwc3: core: update LC timer as per USB Spec V3.2
  usb: dwc3: Avoid waking up gadget during startxfer

2 weeks agoMerge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 7 Sep 2024 18:29:13 +0000 (11:29 -0700)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fixes from Stephen Boyd:
 "A pile of Qualcomm clk driver fixes with two main themes: the alpha
  PLL driver and shared RCGs, and one fix for the Starfive JH7110 SoC.

   - The Alpha PLL clk_ops had multiple problems around setting rates.

     There are a handful of patches here that fix masks and skip
     enabling the clk from set_rate() when the PLL is disabled. The PLLs
     are crucial to operation of the system as almost all frequencies in
     the system are derived from them.

   - Parking shared RCGs at a slow always on clk at registration time
     breaks stuff.

     USB host mode can't handle such a slow frequency and the serial
     console gets all garbled when the UART clk is handed over to the
     kernel. There's a few patches that don't use the shared clk_ops for
     the UART clks and another one to skip parking the USB clk at
     registration time.

   - The Starfive PLL driver used for the CPU was busted causing cpufreq
     to fail because the clk didn't change to a safe parent during
     set_rate().

     The fix is to register a notifier and switch to a safe parent so
     the PLL can change rate in a glitch free manner"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: qcom: gcc-sc8280xp: don't use parking clk_ops for QUPs
  clk: starfive: jh7110-sys: Add notifier for PLL0 clock
  clk: qcom: gcc-sm8650: Don't use shared clk_ops for QUPs
  clk: qcom: gcc-sm8550: Don't park the USB RCG at registration time
  clk: qcom: gcc-sm8550: Don't use parking clk_ops for QUPs
  clk: qcom: gcc-x1e80100: Don't use parking clk_ops for QUPs
  clk: qcom: ipq9574: Update the alpha PLL type for GPLLs
  clk: qcom: gcc-x1e80100: Fix USB 0 and 1 PHY GDSC pwrsts flags
  clk: qcom: clk-alpha-pll: Update set_rate for Zonda PLL
  clk: qcom: clk-alpha-pll: Fix zonda set_rate failure when PLL is disabled
  clk: qcom: clk-alpha-pll: Fix the trion pll postdiv set rate API
  clk: qcom: clk-alpha-pll: Fix the pll post div mask

2 weeks agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 7 Sep 2024 17:25:39 +0000 (10:25 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fix from James Bottomley:
 "Single ufs driver fix quirking around another device spec violation"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: ufs: ufs-mediatek: Add UFSHCD_QUIRK_BROKEN_LSDBS_CAP

2 weeks agoMerge tag 'pinctrl-v6.11-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Sat, 7 Sep 2024 16:59:57 +0000 (09:59 -0700)]
Merge tag 'pinctrl-v6.11-3' of git://git./linux/kernel/git/linusw/linux-pinctrl

Pull pin control fix from Linus Walleij:
 "A single fix for Qualcomm laptops that are affected by
  missing wakeup IRQs"

* tag 'pinctrl-v6.11-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  pinctrl: qcom: x1e80100: Bypass PDC wakeup parent for now

2 weeks agoMerge tag 'linux_kselftest-kunit-fixes-6.11-rc7-2' of git://git.kernel.org/pub/scm...
Linus Torvalds [Sat, 7 Sep 2024 00:48:24 +0000 (17:48 -0700)]
Merge tag 'linux_kselftest-kunit-fixes-6.11-rc7-2' of git://git./linux/kernel/git/shuah/linux-kselftest

PullKUnit fix from Shuah Khan:
 "Fix to a missing function parameter warning found during documentation
  build in linux-next"

* tag 'linux_kselftest-kunit-fixes-6.11-rc7-2' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
  kunit: Fix missing kerneldoc comment

2 weeks agoMerge tag 'pci-v6.11-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci
Linus Torvalds [Sat, 7 Sep 2024 00:38:50 +0000 (17:38 -0700)]
Merge tag 'pci-v6.11-fixes-3' of git://git./linux/kernel/git/pci/pci

Pull pci fixes from Bjorn Helgaas:

 - Unregister platform devices for child nodes when stopping a PCI
   device, even if the PCI core has already cleared the OF_POPULATED bit
   and of_platform_depopulate() doesn't do anything (Bartosz
   Golaszewski)

 - Rescan the bus from a separate thread so we don't deadlock when
   triggering rescan from sysfs (Bartosz Golaszewski)

* tag 'pci-v6.11-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci:
  PCI/pwrctl: Rescan bus on a separate thread
  PCI: Don't rely on of_platform_depopulate() for reused OF-nodes

2 weeks agoMerge tag 'v6.11-rc6-cifs-client-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 7 Sep 2024 00:30:33 +0000 (17:30 -0700)]
Merge tag 'v6.11-rc6-cifs-client-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull smb client fixes from Steve French:

 - fix potential mount hang

 - fix retry problem in two types of compound operations

 - important netfs integration fix in SMB1 read paths

 - fix potential uninitialized zero point of inode

 - minor patch to improve debugging for potential crediting problems

* tag 'v6.11-rc6-cifs-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  netfs, cifs: Improve some debugging bits
  cifs: Fix SMB1 readv/writev callback in the same way as SMB2/3
  cifs: Fix zero_point init on inode initialisation
  smb: client: fix double put of @cfile in smb2_set_path_size()
  smb: client: fix double put of @cfile in smb2_rename_path()
  smb: client: fix hang in wait_for_response() for negproto

2 weeks agoKVM: x86: don't fall through case statements without annotations
Linus Torvalds [Fri, 6 Sep 2024 22:23:33 +0000 (15:23 -0700)]
KVM: x86: don't fall through case statements without annotations

clang warns on this because it has an unannotated fall-through between
cases:

   arch/x86/kvm/x86.c:4819:2: error: unannotated fall-through between switch labels [-Werror,-Wimplicit-fallthrough]

and while we could annotate it as a fallthrough, the proper fix is to
just add the break for this case, instead of falling through to the
default case and the break there.

gcc also has that warning, but it looks like gcc only warns for the
cases where they fall through to "real code", rather than to just a
break.  Odd.

Fixes: d30d9ee94cc0 ("KVM: x86: Only advertise KVM_CAP_READONLY_MEM when supported by VM")
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Tom Dohrmann <erbse.13@gmx.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 weeks agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 6 Sep 2024 20:08:38 +0000 (13:08 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fix from Catalin Marinas:
 "Fix the arm64 usage of ftrace_graph_ret_addr() to pass the
  &state->graph_idx pointer instead of NULL, otherwise this function
  just returns early"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: stacktrace: fix the usage of ftrace_graph_ret_addr()

2 weeks agoMerge tag 'riscv-for-linus-6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 6 Sep 2024 20:00:59 +0000 (13:00 -0700)]
Merge tag 'riscv-for-linus-6.11-rc7' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Palmer Dabbelt:

 - A revert for the mmap() change that ties the allocation range to the
   hint adress, as what we tried to do ended up regressing on other
   userspace workloads.

 - A fix to avoid a kernel memory leak when emulating misaligned
   accesses from userspace.

 - A Kconfig fix for toolchain vector detection, which now correctly
   detects vector support on toolchains where the V extension depends on
   the M extension.

 - A fix to avoid failing the linear mapping bootmem bounds check on
   NOMMU systems.

 - A fix for early alternatives on relocatable kernels.

* tag 'riscv-for-linus-6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: Fix RISCV_ALTERNATIVE_EARLY
  riscv: Do not restrict memory size because of linear mapping on nommu
  riscv: Fix toolchain vector detection
  riscv: misaligned: Restrict user access to kernel memory
  riscv: mm: Do not restrict mmap address based on hint
  riscv: selftests: Remove mmap hint address checks
  Revert "RISC-V: mm: Document mmap changes"

2 weeks agoMerge tag 'powerpc-6.11-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Fri, 6 Sep 2024 19:50:03 +0000 (12:50 -0700)]
Merge tag 'powerpc-6.11-3' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - Fix a deadlock in the powerpc qspinlock MCS queue logic

 - Fix the return type of pgd_val() to not truncate 64-bit PTEs on 85xx

 - Allow the check for dynamic relocations in the VDSO to work correctly

 - Make mmu_pte_psize static to fix a build error

Thanks to Christophe Leroy, Nysal Jan K.A., Nicholas Piggin, Geetika
Moolchandani, Jijo Varghese, and Vaishnavi Bhat.

* tag 'powerpc-6.11-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/qspinlock: Fix deadlock in MCS queue
  powerpc/mm: Fix return type of pgd_val()
  powerpc/vdso: Don't discard rela sections
  powerpc/64e: Define mmu_pte_psize static

2 weeks agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 6 Sep 2024 19:45:43 +0000 (12:45 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull x86 kvm fixes from Paolo Bonzini:
 "Many small fixes that accumulated while I was on vacation...

   - Fixup missed comments from the REMOVED_SPTE => FROZEN_SPTE rename

   - Ensure a root is successfully loaded when pre-faulting SPTEs

   - Grab kvm->srcu when handling KVM_SET_VCPU_EVENTS to guard against
     accessing memslots if toggling SMM happens to force a VM-Exit

   - Emulate MSR_{FS,GS}_BASE on SVM even though interception is always
     disabled, so that KVM does the right thing if KVM's emulator
     encounters {RD,WR}MSR

   - Explicitly clear BUS_LOCK_DETECT from KVM's caps on AMD, as KVM
     doesn't yet virtualize BUS_LOCK_DETECT on AMD

   - Cleanup the help message for CONFIG_KVM_AMD_SEV, and call out that
     KVM now supports SEV-SNP too

   - Specialize return value of
     KVM_CHECK_EXTENSION(KVM_CAP_READONLY_MEM), based on VM type

   - Remove unnecessary dependency on CONFIG_HIGH_RES_TIMERS

   - Note an RCU quiescent state on guest exit. This avoids a call to
     rcu_core() if there was a grace period request while guest was
     running"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: Remove HIGH_RES_TIMERS dependency
  kvm: Note an RCU quiescent state on guest exit
  KVM: x86: Only advertise KVM_CAP_READONLY_MEM when supported by VM
  KVM: SEV: Update KVM_AMD_SEV Kconfig entry and mention SEV-SNP
  KVM: SVM: Don't advertise Bus Lock Detect to guest if SVM support is missing
  KVM: SVM: fix emulation of msr reads/writes of MSR_FS_BASE and MSR_GS_BASE
  KVM: x86: Acquire kvm->srcu when handling KVM_SET_VCPU_EVENTS
  KVM: x86/mmu: Check that root is valid/loaded when pre-faulting SPTEs
  KVM: x86/mmu: Fixup comments missed by the REMOVED_SPTE=>FROZEN_SPTE rename

2 weeks agoMerge tag 'pm-6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Linus Torvalds [Fri, 6 Sep 2024 19:17:44 +0000 (12:17 -0700)]
Merge tag 'pm-6.11-rc7' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fix from Rafael Wysocki:
 "Fix an incorrect warning emitted by the amd-pstate driver on
  processors that don't support X86_FEATURE_CPPC (Gautham Shenoy)"

* tag 'pm-6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  cpufreq/amd-pstate: Remove warning for X86_FEATURE_CPPC on certain Zen models

2 weeks agoMerge tag 'block-6.11-20240906' of git://git.kernel.dk/linux
Linus Torvalds [Fri, 6 Sep 2024 19:04:06 +0000 (12:04 -0700)]
Merge tag 'block-6.11-20240906' of git://git.kernel.dk/linux

Pull block fixes from Jens Axboe:
 "Mostly just some fixlets for NVMe, but also a bug fix for the ublk
  driver and an integrity fix"

* tag 'block-6.11-20240906' of git://git.kernel.dk/linux:
  bio-integrity: don't restrict the size of integrity metadata
  ublk_drv: fix NULL pointer dereference in ublk_ctrl_start_recovery()
  nvmet: Identify-Active Namespace ID List command should reject invalid nsid
  nvme: set BLK_FEAT_ZONED for ZNS multipath disks
  nvme-pci: Add sleep quirk for Samsung 990 Evo
  nvme-pci: allocate tagset on reset if necessary
  nvmet-tcp: fix kernel crash if commands allocation fails
  nvme: use better description for async reset reason
  nvmet: Make nvmet_debugfs static

2 weeks agoMerge tag 'sound-6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Fri, 6 Sep 2024 18:56:03 +0000 (11:56 -0700)]
Merge tag 'sound-6.11-rc7' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "Hopefully the last PR for 6.11, at least for this level of amount.

  In addition to the usual HD-audio quirks, there are more changes in
  ASoC, but all look small and device-specific fixes, and nothing stands
  out. The only slightly big change is sunxi I2S fix, which looks quite
  safe to apply, too"

* tag 'sound-6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound: (21 commits)
  ALSA: hda/realtek - Fix inactive headset mic jack for ASUS Vivobook 15 X1504VAP
  ALSA: hda/realtek: Support mute LED on HP Laptop 14-dq2xxx
  ALSA: hda/realtek: Enable Mute Led for HP Victus 15-fb1xxx
  ALSA: hda/realtek: extend quirks for Clevo V5[46]0
  ASoC: codecs: lpass-va-macro: set the default codec version for sm8250
  ALSA: hda: add HDMI codec ID for Intel PTL
  ALSA: hda/realtek: add patch for internal mic in Lenovo V145
  ASoC: sunxi: sun4i-i2s: fix LRCLK polarity in i2s mode
  ASoC: amd: yc: Add a quirk for MSI Bravo 17 (D7VEK)
  ASoC: mediatek: mt8188-mt6359: Modify key
  ASoc: SOF: topology: Clear SOF link platform name upon unload
  ALSA: hda/conexant: Add pincfg quirk to enable top speakers on Sirius devices
  ASoC: SOF: ipc: replace "enum sof_comp_type" field with "uint32_t"
  ASoC: fix module autoloading
  ASoC: tda7419: fix module autoloading
  ASoC: google: fix module autoloading
  ASoC: intel: fix module autoloading
  ASoC: tegra: Fix CBB error during probe()
  ASoC: dapm: Fix UAF for snd_soc_pcm_runtime object
  ASoC: Intel: soc-acpi-cht: Make Lenovo Yoga Tab 3 X90F DMI match less strict
  ...

2 weeks agoMerge tag 'mmc-v6.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Fri, 6 Sep 2024 18:50:55 +0000 (11:50 -0700)]
Merge tag 'mmc-v6.11-rc5' of git://git./linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:
 "MMC core:
   - Apply SD quirks earlier during probe so they become relevant

  MMC host:
   - cqhci: Fix checking of CQHCI_HALT state
   - dw_mmc: Fix IDMAC operation with pages bigger than 4K
   - sdhci-of-aspeed: Fix module autoloading"

* tag 'mmc-v6.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: cqhci: Fix checking of CQHCI_HALT state
  mmc: dw_mmc: Fix IDMAC operation with pages bigger than 4K
  mmc: sdhci-of-aspeed: fix module autoloading
  mmc: core: apply SD quirks earlier during probe

2 weeks agoMerge tag 'gpio-fixes-for-v6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 6 Sep 2024 18:40:42 +0000 (11:40 -0700)]
Merge tag 'gpio-fixes-for-v6.11-rc7' of git://git./linux/kernel/git/brgl/linux

Pull gpio fixes from Bartosz Golaszewski:

 - fix an OF node reference leak in gpio-rockchip

 - add the missing module device table to gpio-modepin

* tag 'gpio-fixes-for-v6.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux:
  gpio: modepin: Enable module autoloading
  gpio: rockchip: fix OF node leak in probe()

2 weeks agoMerge tag 'pmdomain-v6.11-rc2-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 6 Sep 2024 18:32:30 +0000 (11:32 -0700)]
Merge tag 'pmdomain-v6.11-rc2-2' of git://git./linux/kernel/git/ulfh/linux-pm

Pull pmdomain fix from Ulf Hansson:

 - Fix support for required OPPs for multiple PM domains

* tag 'pmdomain-v6.11-rc2-2' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/linux-pm:
  OPP: Fix support for required OPPs for multiple PM domains

2 weeks agoMerge tag 'pwm/for-6.11-rc7-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 6 Sep 2024 18:28:14 +0000 (11:28 -0700)]
Merge tag 'pwm/for-6.11-rc7-fixes' of git://git./linux/kernel/git/ukleinek/linux

Pull pwm fix from Uwe Kleine-König:
 "Fix an off-by-one in the stm32 driver.

  Hardware engineers tend to start counting at 1 while the software guys
  usually start with 0. This isn't so nice because that results in
  drivers where pwm device #2 needs to use the hardware registers with
  index 3.

  This was noticed by Fabrice Gasnier.

  A small patch fixing that mismatch is the only change included here"

* tag 'pwm/for-6.11-rc7-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/ukleinek/linux:
  pwm: stm32: Use the right CCxNP bit in stm32_pwm_enable()

2 weeks agoMerge tag 'drm-fixes-2024-09-06' of https://gitlab.freedesktop.org/drm/kernel
Linus Torvalds [Fri, 6 Sep 2024 18:15:40 +0000 (11:15 -0700)]
Merge tag 'drm-fixes-2024-09-06' of https://gitlab.freedesktop.org/drm/kernel

Pull drm fixes from Dave Airlie:
 "This has a fair few patches in it, but I reviewed them all and they
  seem like real things, amdgpu, i915 and xe each have a bunch of fixes
  for various things, then there is a some bridge suspend/resume
  ordering fixes for a recent rework, and then some single driver
  changes in a few others.

  Nothing looks too serious, hopefully next week is quiet.

  amdgpu:
   - IPS workaround
   - Fix compatibility with older MES firmware
   - Fix CPU spikes when clearing VRAM
   - Backlight fix
   - PMO fix
   - Revert SWSMU change to fix regression

  xe:
   - GSC loading fix
   - PCODE mutex fix
   - Suspend/Resume fixes
   - RPM fixes

  i915:
   - Do not attempt to load the GSC multiple times
   - Fix readout degamma_lut mismatch on ilk/snb
   - Mark debug_fence_init_onstack() with __maybe_unused
   - fence: Mark debug_fence_free() with __maybe_unused
   - display: Add mechanism to use sink model when applying quirk
   - display: Increase Fast Wake Sync length as a quirk

  komeda:
   - zpos normalization fix

  nouveau:
   - incorrect register fix

  imagination:
   - memory leak fix

  bridge:
   - hdmi/bridge rework fixes

  panthor:
   - cache coherency fix
   - hi priority access fix

  panel:
   - change of compatible string

  fbdev:
   - deferred-io init with no struct page fix"

* tag 'drm-fixes-2024-09-06' of https://gitlab.freedesktop.org/drm/kernel: (29 commits)
  Revert "drm/amdgpu: align pp_power_profile_mode with kernel docs"
  drm/fbdev-dma: Only install deferred I/O if necessary
  drm/panthor: flush FW AS caches in slow reset path
  drm: panel: nv3052c: Correct WL-355608-A8 panel compatible
  dt-bindings: display: panel: Rename WL-355608-A8 panel to rg35xx-*-panel
  drm/panthor: Restrict high priorities on group_create
  drm/xe/display: Avoid encoder_suspend at runtime suspend
  drm/xe: Suspend/resume user access only during system s/r
  drm/xe/display: Match i915 driver suspend/resume sequences better
  drm/xe: Add missing runtime reference to wedged upon gt_reset
  drm/xe/pcode: Treat pcode as per-tile rather than per-GT
  drm/xe/gsc: Do not attempt to load the GSC multiple times
  drm/bridge-connector: reset the HDMI connector state
  drm/bridge-connector: move to DRM_DISPLAY_HELPER module
  drm/display: stop depending on DRM_DISPLAY_HELPER
  drm/i915/display: Increase Fast Wake Sync length as a quirk
  drm/i915/display: Add mechanism to use sink model when applying quirk
  drm/amd/display: Block timing sync for different signals in PMO
  drm/amd/display: Lock DC and exit IPS when changing backlight
  drm/amdgpu: always allocate cleared VRAM for GEM allocations
  ...

2 weeks agolibfs: fix get_stashed_dentry()
Christian Brauner [Fri, 6 Sep 2024 16:22:22 +0000 (18:22 +0200)]
libfs: fix get_stashed_dentry()

get_stashed_dentry() tries to optimistically retrieve a stashed dentry
from a provided location.  It needs to ensure to hold rcu lock before it
dereference the stashed location to prevent UAF issues.  Use
rcu_dereference() instead of READ_ONCE() it's effectively equivalent
with some lockdep bells and whistles and it communicates clearly that
this expects rcu protection.

Link: https://lore.kernel.org/r/20240906-vfs-hotfix-5959800ffa68@brauner
Fixes: 07fd7c329839 ("libfs: add path_from_stashed()")
Reported-by: syzbot+f82b36bffae7ef78b6a7@syzkaller.appspotmail.com
Fixes: syzbot+f82b36bffae7ef78b6a7@syzkaller.appspotmail.com
Reported-by: syzbot+cbe4b96e1194b0e34db6@syzkaller.appspotmail.com
Fixes: syzbot+cbe4b96e1194b0e34db6@syzkaller.appspotmail.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 weeks agoMerge tag 'asoc-fix-v6.11-rc6' of https://git.kernel.org/pub/scm/linux/kernel/git...
Takashi Iwai [Fri, 6 Sep 2024 06:24:56 +0000 (08:24 +0200)]
Merge tag 'asoc-fix-v6.11-rc6' of https://git./linux/kernel/git/broonie/sound into for-linus

ASoC: Fixes for v6.11

A larger set of fixes than I'd like at this point, but mainly due to
people working on fixing module autoloading by adding missing exports of
ID tables rather than anything particularly concerning.  There are some
other runtime fixes and quirks, and a tweak to the ABI definition for
SOF which ensures that a struct layout doesn't vary depending on the
architecture of the host.