Jeff Layton [Wed, 3 Jan 2024 13:36:52 +0000 (08:36 -0500)]
nfsd: drop the nfsd_put helper
It's not safe to call nfsd_put once nfsd_last_thread has been called, as
that function will zero out the nn->nfsd_serv pointer.
Drop the nfsd_put helper altogether and open-code the svc_put in its
callers instead. That allows us to not be reliant on the value of that
pointer when handling an error.
Fixes: 2a501f55cd64 ("nfsd: call nfsd_last_thread() before final nfsd_put()") Reported-by: Zhi Li <yieli@redhat.com> Cc: NeilBrown <neilb@suse.de> Signed-off-by: Jeffrey Layton <jlayton@redhat.com> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Guillaume says:
> I believe commit 5f7fc5d69f6e ("SUNRPC: Resupply rq_pages from
> node-local memory") in Linux 6.5+ is incorrect. It passes
> unconditionally rq_pool->sp_id as the NUMA node.
>
> While the comment in the svc_pool declaration in sunrpc/svc.h says
> that sp_id is also the NUMA node id, it might not be the case if
> the svc is created using svc_create_pooled(). svc_created_pooled()
> can use the per-cpu pool mode therefore in this case sp_id would
> be the cpu id.
Fix this by reverting now. At a later point this minor optimization,
and the deceptive labeling of the sp_id field, can be revisited.
Reported-by: Guillaume Morin <guillaume@morinfr.org> Closes: https://lore.kernel.org/linux-nfs/ZYC9rsno8qYggVt9@bender.morinfr.org/T/#u Fixes: 5f7fc5d69f6e ("SUNRPC: Resupply rq_pages from node-local memory") Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
There's nothing wrong with this commit, but this is dead code now
that nothing triggers a CB_GETATTR callback. It can be re-introduced
once the issues with handling conflicting GETATTRs are resolved.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
For some reason, the wait_on_bit() in nfsd4_deleg_getattr_conflict()
is waiting forever, preventing a clean server shutdown. The
requesting client might also hang waiting for a reply to the
conflicting GETATTR.
Invoking wait_on_bit() in an nfsd thread context is a hazard. The
correct fix is to replace this wait_on_bit() call site with a
mechanism that defers the conflicting GETATTR until the CB_GETATTR
completes or is known to have failed.
That will require some surgery and extended testing and it's late
in the v6.7-rc cycle, so I'm reverting now in favor of trying again
in a subsequent kernel release.
This is my fault: I should have recognized the ramifications of
calling wait_on_bit() in here before accepting this patch.
Thanks to Dai Ngo <dai.ngo@oracle.com> for diagnosing the issue.
Reported-by: Wolfgang Walter <linux-nfs@stwm.de> Closes: https://lore.kernel.org/linux-nfs/e3d43ecdad554fbdcaa7181833834f78@stwm.de/ Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
NeilBrown [Fri, 15 Dec 2023 00:56:31 +0000 (11:56 +1100)]
nfsd: call nfsd_last_thread() before final nfsd_put()
If write_ports_addfd or write_ports_addxprt fail, they call nfsd_put()
without calling nfsd_last_thread(). This leaves nn->nfsd_serv pointing
to a structure that has been freed.
So remove 'static' from nfsd_last_thread() and call it when the
nfsd_serv is about to be destroyed.
Fixes: ec52361df99b ("SUNRPC: stop using ->sv_nrthreads as a refcount") Signed-off-by: NeilBrown <neilb@suse.de> Reviewed-by: Jeff Layton <jlayton@kernel.org> Cc: <stable@vger.kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Fri, 10 Nov 2023 16:28:45 +0000 (11:28 -0500)]
NFSD: Fix checksum mismatches in the duplicate reply cache
nfsd_cache_csum() currently assumes that the server's RPC layer has
been advancing rq_arg.head[0].iov_base as it decodes an incoming
request, because that's the way it used to work. On entry, it
expects that buf->head[0].iov_base points to the start of the NFS
header, and excludes the already-decoded RPC header.
These days however, head[0].iov_base now points to the start of the
RPC header during all processing. It no longer points at the NFS
Call header when execution arrives at nfsd_cache_csum().
In a retransmitted RPC the XID and the NFS header are supposed to
be the same as the original message, but the contents of the
retransmitted RPC header can be different. For example, for krb5,
the GSS sequence number will be different between the two. Thus if
the RPC header is always included in the DRC checksum computation,
the checksum of the retransmitted message might not match the
checksum of the original message, even though the NFS part of these
messages is identical.
The result is that, even if a matching XID is found in the DRC,
the checksum mismatch causes the server to execute the
retransmitted RPC transaction again.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Tested-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Fri, 10 Nov 2023 16:28:33 +0000 (11:28 -0500)]
NFSD: Fix "start of NFS reply" pointer passed to nfsd_cache_update()
The "statp + 1" pointer that is passed to nfsd_cache_update() is
supposed to point to the start of the egress NFS Reply header. In
fact, it does point there for AUTH_SYS and RPCSEC_GSS_KRB5 requests.
But both krb5i and krb5p add fields between the RPC header's
accept_stat field and the start of the NFS Reply header. In those
cases, "statp + 1" points at the extra fields instead of the Reply.
The result is that nfsd_cache_update() caches what looks to the
client like garbage.
A connection break can occur for a number of reasons, but the most
common reason when using krb5i/p is a GSS sequence number window
underrun. When an underrun is detected, the server is obliged to
drop the RPC and the connection to force a retransmit with a fresh
GSS sequence number. The client presents the same XID, it hits in
the server's DRC, and the server returns the garbage cache entry.
The "statp + 1" argument has been used since the oldest changeset
in the kernel history repo, so it has been in nfsd_dispatch()
literally since before history began. The problem arose only when
the server-side GSS implementation was added twenty years ago.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Tested-by: Jeff Layton <jlayton@kernel.org Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Fri, 10 Nov 2023 16:28:39 +0000 (11:28 -0500)]
NFSD: Update nfsd_cache_append() to use xdr_stream
When inserting a DRC-cached response into the reply buffer, ensure
that the reply buffer's xdr_stream is updated properly. Otherwise
the server will send a garbage response.
Cc: stable@vger.kernel.org # v6.3+ Reviewed-by: Jeff Layton <jlayton@kernel.org> Tested-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Mahmoud Adam [Fri, 10 Nov 2023 18:21:04 +0000 (19:21 +0100)]
nfsd: fix file memleak on client_opens_release
seq_release should be called to free the allocated seq_file
Cc: stable@vger.kernel.org # v5.3+ Signed-off-by: Mahmoud Adam <mngyadam@amazon.com> Reviewed-by: Jeff Layton <jlayton@kernel.org> Fixes: 78599c42ae3c ("nfsd4: add file to display list of client's opens") Reviewed-by: NeilBrown <neilb@suse.de> Tested-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Fri, 13 Oct 2023 12:22:24 +0000 (08:22 -0400)]
svcrdma: Fix tracepoint printk format
Other tracepoints use "cq.id=" rather than "cq_id=". Let's make it
more reliable to grep for the CQ restracker ID.
Reviewed-by: Benjamin Coddington <bcodding@redhat.com> Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Tue, 10 Oct 2023 17:23:41 +0000 (13:23 -0400)]
svcrdma: Drop connection after an RDMA Read error
When an RPC Call message cannot be pulled from the client, that
is a message loss, by definition. Close the connection to trigger
the client to resend.
Cc: <stable@vger.kernel.org> Reviewed-by: Tom Talpey <tom@talpey.com> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Sicong Huang [Thu, 12 Oct 2023 08:34:58 +0000 (16:34 +0800)]
NFSD: clean up alloc_init_deleg()
Modify the conditional statement for null pointer check in the function
'alloc_init_deleg' to make this function more robust and clear. Otherwise,
this function may have potential pointer dereference problem in the future,
when modifying or expanding the nfs4_delegation structure.
Signed-off-by: Sicong Huang <huangsicong@iie.ac.cn> Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Sun, 1 Oct 2023 17:25:16 +0000 (13:25 -0400)]
NFSD: Fix frame size warning in svc_export_parse()
fs/nfsd/export.c: In function 'svc_export_parse':
fs/nfsd/export.c:737:1: warning: the frame size of 1040 bytes is larger than 1024 bytes [-Wframe-larger-than=]
737 | }
On my systems, svc_export_parse() has a stack frame of over 800
bytes, not 1040, but nonetheless, it could do with some reduction.
When a struct svc_export is on the stack, it's a temporary structure
used as an argument, and not visible as an actual exported FS. No
need to reserve space for export_stats in such cases.
Reported-by: kernel test robot <lkp@intel.com> Closes: https://lore.kernel.org/oe-kbuild-all/202310012359.YEw5IrK6-lkp@intel.com/ Cc: Amir Goldstein <amir73il@gmail.com> Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Sun, 1 Oct 2023 17:25:09 +0000 (13:25 -0400)]
NFSD: Rewrite synopsis of nfsd_percpu_counters_init()
In function ‘export_stats_init’,
inlined from ‘svc_export_alloc’ at fs/nfsd/export.c:866:6:
fs/nfsd/export.c:337:16: warning: ‘nfsd_percpu_counters_init’ accessing 40 bytes in a region of size 0 [-Wstringop-overflow=]
337 | return nfsd_percpu_counters_init(&stats->counter, EXP_STATS_COUNTERS_NUM);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
fs/nfsd/export.c:337:16: note: referencing argument 1 of type ‘struct percpu_counter[0]’
fs/nfsd/stats.h: In function ‘svc_export_alloc’:
fs/nfsd/stats.h:40:5: note: in a call to function ‘nfsd_percpu_counters_init’
40 | int nfsd_percpu_counters_init(struct percpu_counter counters[], int num);
| ^~~~~~~~~~~~~~~~~~~~~~~~~
Cc: Amir Goldstein <amir73il@gmail.com> Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
KaiLong Wang [Thu, 28 Sep 2023 02:43:04 +0000 (10:43 +0800)]
nfsd: Clean up errors in nfs4state.c
Fix the following errors reported by checkpatch:
ERROR: spaces required around that '=' (ctx:VxW)
ERROR: space required after that ',' (ctx:VxO)
ERROR: space required before that '~' (ctx:OxV) Signed-off-by: KaiLong Wang <wangkailong@jari.cn> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
The error paths in nfsd_svc() are needlessly complex and can result in a
final call to svc_put() without nfsd_last_thread() being called. This
results in the listening sockets not being closed properly.
The per-netns setup provided by nfsd_startup_new() and removed by
nfsd_shutdown_net() is needed precisely when there are running threads.
So we don't need nfsd_up_before. We don't need to know if it *was* up.
We only need to know if any threads are left. If none are, then we must
call nfsd_shutdown_net(). But we don't need to do that explicitly as
nfsd_last_thread() does that for us.
So simply call nfsd_last_thread() before the last svc_put() if there are
no running threads. That will always do the right thing.
Also discard:
pr_info("nfsd: last server has exited, flushing export cache\n");
It may not be true if an attempt to start the first server failed, and
it isn't particularly helpful and it simply reports normal behaviour.
Signed-off-by: NeilBrown <neilb@suse.de> Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Mon, 9 Oct 2023 18:30:16 +0000 (14:30 -0400)]
NFSD: Clean up nfsd4_encode_copy_notify()
Replace open-coded encoding logic with the use of conventional XDR
utility functions.
Note that if we replace the cpn_sec and cpn_nsec fields with a
single struct timespec64 field, the encoder can use
nfsd4_encode_nfstime4(), as that is the data type specified by the
XDR spec.
NFS4ERR_INVAL seems inappropriate if the encoder doesn't support
encoding the response. Instead use NFS4ERR_SERVERFAULT, since this
condition is a software bug on the server.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Mon, 9 Oct 2023 18:29:56 +0000 (14:29 -0400)]
NFSD: Clean up nfsd4_encode_exchange_id()
Restructure nfsd4_encode_exchange_id() so that it will be more
straightforward to add support for SSV one day. Also, adopt the use
of the conventional XDR utility functions.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Wed, 4 Oct 2023 13:41:44 +0000 (09:41 -0400)]
NFSD: Rename nfsd4_encode_dirent()
Rename nfsd4_encode_dirent() to match the naming convention already
used in the NFSv2 and NFSv3 readdir paths. The new name reflects the
name of the spec-defined XDR data type for an NFSv4 directory entry.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Fri, 29 Sep 2023 13:59:05 +0000 (09:59 -0400)]
NFSD: Add nfsd4_encode_open_read_delegation4()
Refactor nfsd4_encode_open() so the open_read_delegation4 type is
encoded in a separate function. This makes it more straightforward
to later add support for returning an nfsace4 in OPEN responses that
offer a delegation.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Fri, 29 Sep 2023 13:58:59 +0000 (09:58 -0400)]
NFSD: Refactor nfsd4_encode_lock_denied()
Use the modern XDR utility functions.
The LOCK and LOCKT encoder functions need to return nfserr_denied
when a lock is denied, but nfsd4_encode_lock4denied() should return
a status code that is consistent with other XDR encoders.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Fri, 29 Sep 2023 13:58:53 +0000 (09:58 -0400)]
NFSD: Add nfsd4_encode_lock_owner4()
To improve readability and better align the LOCK encoders with the
XDR specification, add an explicit encoder named for the lock_owner4
type.
In particular, to avoid code duplication, use
nfsd4_encode_clientid4() to encode the clientid in the lock owner
rather than open-coding it.
It looks to me like nfs4_set_lock_denied() already clears the
clientid if it won't return an owner (cf: the nevermind: label). The
code in the XDR encoder appears to be redundant and can safely be
removed.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Thu, 12 Oct 2023 17:46:39 +0000 (13:46 -0400)]
NFSD: Remove a layering violation when encoding lock_denied
An XDR encoder is responsible for marshaling results, not releasing
memory that was allocated by the upper layer. We have .op_release
for that purpose.
Move the release of the ld_owner.data string to op_release functions
for LOCK and LOCKT.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Mon, 25 Sep 2023 13:28:23 +0000 (09:28 -0400)]
NFSD: Clean up nfsd4_encode_getdeviceinfo()
Adopt the conventional XDR utility functions. Also, restructure to
make the function align more closely with the spec -- there doesn't
seem to be a performance need for speciality code, so prioritize
readability.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Mon, 25 Sep 2023 13:27:58 +0000 (09:27 -0400)]
NFSD: Clean up nfsd4_encode_layoutget()
De-duplicate the open-coded stateid4 encoder. Adopt the use of the
conventional current XDR encoding helpers. Refactor the encoder to
align with the XDR specification.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Mon, 25 Sep 2023 13:27:45 +0000 (09:27 -0400)]
NFSD: Clean up nfsd4_encode_stateid()
Update the encoder function name to match the type name, as is the
convention with other such encoder utility functions, and with
nfsd4_decode_stateid4().
Make the @stateid argument a const so that callers of
nfsd4_encode_stateid4() in the future can be passed const pointers
to structures.
Since the compiler is allowed to add padding to structs, use the
wire (spec-defined) size when reserving buffer space.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Mon, 18 Sep 2023 14:02:06 +0000 (10:02 -0400)]
NFSD: Copy FATTR4 bit number definitions from RFCs
I'd like to convert nfsd4_encode_fattr() to rotate through the
attrmask using for_each_bit() instead of explicitly testing the
bitmask for each bit value. This means I need the bit numbers, as
defined in the specs, instead of our internal bitmask constants.
As a clean up, use the new spec-derived values to define the WORD#
bitmask constants.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Mon, 18 Sep 2023 13:57:51 +0000 (09:57 -0400)]
NFSD: Add nfsd4_encode_fattr4_change()
Refactor the encoder for FATTR4_CHANGE into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.
The code is restructured a bit to use the modern xdr_stream flow,
and the encoded cinfo value is made const so that callers of the
encoders can be passed a const cinfo.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Chuck Lever [Mon, 18 Sep 2023 13:57:39 +0000 (09:57 -0400)]
NFSD: Add nfsd4_encode_fattr4_type()
Refactor the encoder for FATTR4_TYPE into a helper. In a subsequent
patch, this helper will be called from a bitmask loop.
In addition, restructure the code so that byte-swapping is done on
constant values rather than at run time. Run-time swapping can be
costly on some platforms, and "type" is a frequently-requested
attribute.
Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>