Pull nfsd updates from Chuck Lever:
"Neil Brown and Jeff Layton contributed a dynamic thread pool sizing
mechanism for NFSD. The sunrpc layer now tracks minimum and maximum
thread counts per pool, and NFSD adjusts running thread counts based
on workload: idle threads exit after a timeout when the pool exceeds
its minimum, and new threads spawn automatically when all threads are
busy. Administrators control this behavior via the nfsdctl netlink
interface.
Rick Macklem, FreeBSD NFS maintainer, generously contributed server-
side support for the POSIX ACL extension to NFSv4, as specified in
draft-ietf-nfsv4-posix-acls. This extension allows NFSv4 clients to
get and set POSIX access and default ACLs using native NFSv4
operations, eliminating the need for sideband protocols. The feature
is gated by a Kconfig option since the IETF draft has not yet been
ratified.
Chuck Lever delivered numerous improvements to the xdrgen tool. Error
reporting now covers parsing, AST transformation, and invalid
declarations. Generated enum decoders validate incoming values against
valid enumerator lists. New features include pass-through line support
for embedding C directives in XDR specifications, 16-bit integer
types, and program number definitions. Several code generation issues
were also addressed.
When an administrator revokes NFSv4 state for a filesystem via the
unlock_fs interface, ongoing async COPY operations referencing that
filesystem are now cancelled, with CB_OFFLOAD callbacks notifying
affected clients.
The remaining patches in this pull request are clean-ups and minor
optimizations. Sincere thanks to all contributors, reviewers, testers,
and bug reporters who participated in the v7.0 NFSD development cycle"
* tag 'nfsd-7.0' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux: (45 commits)
NFSD: Add POSIX ACL file attributes to SUPPATTR bitmasks
NFSD: Add POSIX draft ACL support to the NFSv4 SETATTR operation
NFSD: Add support for POSIX draft ACLs for file creation
NFSD: Add support for XDR decoding POSIX draft ACLs
NFSD: Refactor nfsd_setattr()'s ACL error reporting
NFSD: Do not allow NFSv4 (N)VERIFY to check POSIX ACL attributes
NFSD: Add nfsd4_encode_fattr4_posix_access_acl
NFSD: Add nfsd4_encode_fattr4_posix_default_acl
NFSD: Add nfsd4_encode_fattr4_acl_trueform_scope
NFSD: Add nfsd4_encode_fattr4_acl_trueform
Add RPC language definition of NFSv4 POSIX ACL extension
NFSD: Add a Kconfig setting to enable support for NFSv4 POSIX ACLs
xdrgen: Implement pass-through lines in specifications
nfsd: cancel async COPY operations when admin revokes filesystem state
nfsd: add controls to set the minimum number of threads per pool
nfsd: adjust number of running nfsd threads based on activity
sunrpc: allow svc_recv() to return -ETIMEDOUT and -EBUSY
sunrpc: split new thread creation into a separate function
sunrpc: introduce the concept of a minimum number of threads per pool
sunrpc: track the max number of requested threads in a pool
...
Extend ynltool to compute HW GRO savings metric - how many
packets has HW GRO been able to save the kernel from seeing.
Note that this definition does not actually take into account
whether the segments were or weren't eligible for HW GRO.
If a machine is receiving all-UDP traffic - new metric will show
HW-GRO savings of 0%. Conversely since the super-packet still
counts as a received packet, savings of 100% is not achievable.
Perfect HW-GRO on a machine with 4k MTU and 64kB super-frames
would show ~93.75% savings. With 1.5k MTU we may see up to
~97.8% savings (if my math is right).
Example after 10 sec of iperf on a freshly booted machine
with 1.5k MTU:
$ ynltool qstats show
eth0 rx-packets: 40681280 rx-bytes: 61575208437
rx-alloc-fail: 0 rx-hw-gro-packets: 1225133
rx-hw-gro-wire-packets: 40656633
$ ynltool qstats hw-gro
eth0: 96.9% savings
None of the NICs I have access to can report "missed" HW-GRO
opportunities so computing a true "effectiveness" metric
is not possible. One could also argue that effectiveness metric
is inferior in environments where we control both senders and
receivers, the savings metrics will capture both regressions
in receiver's HW GRO effectiveness but also regressions in senders
sending smaller TSO trains. And we care about both. The main
downside is that it's hard to tell at a glance how well the NIC
is doing because the savings will be dependent on traffic patterns.
Reviewed-by: Petr Machata <petrm@nvidia.com>
Link: https://patch.msgid.link/20260207003509.3927744-4-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Make the default (non-JSON) output more compact. Looking at RSS
context dumps is pretty much impossible without this, because
default print shows the indirection table with line per entry:
'indir': [0,
1,
2,
...
And indirection tables have 100-200 entries each.
The compact output is far more readable:
'indir': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29,
Reviewed-by: Donald Hunter <donald.hunter@gmail.com>
Link: https://patch.msgid.link/20260131203029.1173492-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
XDR specification files can contain lines prefixed with '%' that
pass through unchanged to generated output. Traditional rpcgen
removes the '%' and emits the remainder verbatim, allowing direct
insertion of C includes, pragma directives, or other language-
specific content into the generated code.
Until now, xdrgen silently discarded these lines during parsing.
This prevented specifications from including necessary headers or
preprocessor directives that might be required for the generated
code to compile correctly.
The grammar now captures pass-through lines instead of ignoring
them. A new AST node type represents pass-through content, and
the AST transformer strips the leading '%' character. Definition
and source generators emit pass-through content in document order,
preserving the original placement within the specification.
This brings xdrgen closer to feature parity with traditional
rpcgen while maintaining the existing document-order processing
model.
Existing generated xdrgen source code has been regenerated.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
XDR enum decoders generated by xdrgen do not verify that incoming
values are valid members of the enum. Incoming out-of-range values
from malicious or buggy peers propagate through the system
unchecked.
Add validation logic to generated enum decoders using a switch
statement that explicitly lists valid enumerator values. The
compiler optimizes this to a simple range check when enum values
are dense (contiguous), while correctly rejecting invalid values
for sparse enums with gaps in their value ranges.
The --no-enum-validation option on the source subcommand disables
this validation when not needed.
The minimum and maximum fields in _XdrEnum, which were previously
unused placeholders for a range-based validation approach, have
been removed since the switch-based validation handles both dense
and sparse enums correctly.
Because the new mechanism results in substantive changes to
generated code, existing .x files are regenerated. Unrelated white
space and semicolon changes in the generated code are due to recent
commit 1c873a2fd1 ("xdrgen: Don't generate unnecessary semicolon")
and commit 38c4df91242b ("xdrgen: Address some checkpatch whitespace
complaints").
Reviewed-by: NeilBrown <neil@brown.name>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
struct svc_service has a .vs_xdrsize field that is filled in by
servers for each of their RPC programs. This field is supposed to
contain the size of the largest procedure argument in the RPC
program. This value is also sometimes used to size network
transport buffers.
Currently, server implementations must manually calculate and
hard-code this value, which is error-prone and requires updates
when procedure arguments change.
Update xdrgen to determine which procedure argument structure is
largest, and emit a macro with a well-known name that contains
the size of that structure. Server code then uses this macro when
initializing the .vs_xdrsize field.
For NLM version 4, xdrgen now emits:
#define NLM4_MAX_ARGS_SZ (NLM4_nlm4_lockargs_sz)
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Commit 277df18d7df9 ("xdrgen: Improve parse error reporting") added
clean, compiler-style error messages for syntax errors detected during
parsing. However, semantic errors discovered during AST transformation
still produce verbose Python stack traces.
When an XDR specification references an undefined type, the transformer
raises a VisitError wrapping a KeyError. Before this change:
Traceback (most recent call last):
File ".../lark/visitors.py", line 124, in _call_userfunc
return f(children)
...
KeyError: 'fsh4_mode'
...
lark.exceptions.VisitError: Error trying to process rule "basic":
'fsh4_mode'
After this change:
file.x:156:2: semantic error
Undefined type 'fsh4_mode'
fsh4_mode mode;
^
The new handle_transform_error() function extracts position information
from the Lark tree node metadata and formats the error consistently with
parse error messages.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
The current verbose Lark exception output makes it difficult to
quickly identify and fix syntax errors in XDR specifications. Users
must wade through hundreds of lines of cascading errors to find the
root cause.
Replace this with concise, compiler-style error messages showing
file, line, column, the unexpected token, and the source line with
a caret pointing to the error location.
Before:
Unexpected token Token('__ANON_1', '+1') at line 14, column 35.
Expected one of:
* SEMICOLON
Previous tokens: [Token('__ANON_0', 'LM_MAXSTRLEN')]
[hundreds more cascading errors...]
After:
file.x:14:35: parse error
Unexpected number '+1'
const LM_MAXNAMELEN = LM_MAXSTRLEN+1;
^
The error handler now raises XdrParseError on the first error,
preventing cascading messages that obscure the root cause.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
The client-side source code template mistakenly includes the
nlm4.h header file, which is specific to the NLM protocol and
should not be present in the generic template that generates
client stubs for all XDR-based protocols.
Fixes: 903a7d37d9 ("xdrgen: Update the files included in client-side source code")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
"short" and "unsigned short" types are not defined in RFC 4506, but
are supported by the rpcgen program. An upcoming protocol
specification includes at least one "unsigned short" field, so xdrgen
needs to implement support for these types.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
"xdrgen definitions" was not providing a definition of a symbolic
constant for the RPC program number being defined.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
The program templates for decoder/argument.j2 and encoder/result.j2
unconditionally add 'struct' prefix to all types. This is incorrect
when an RPC protocol specification lists a typedef'd basic type or
an enum as a procedure argument or result (e.g., NFSv2's fhandle or
stat), resulting in compiler errors when building generated C code.
Fixes: 4b132aacb0 ("tools: Add xdrgen")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
This is a roll-up of three template fixes that eliminate noise from
checkpatch output so that it's easier to spot non-trivial problems.
To follow conventional kernel C style, when a union declaration is
marked with "pragma public", there should be a blank line between
the emitted "union xxx { ... };" and the decoder and encoder
function declarations.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Eliminate this warning in code generated by xdrgen:
fs/nfsd/nfs3xdr_gen.c:220:2: warning: switch condition has boolean value [-Wswitch-bool]
220 | switch (ptr->attributes_follow) {
| ^ ~~~~~~~~~~~~~~~~~~~~~~
No more -Wswitch-bool warnings when compiling with W=1.
The generated code is functionally equivalent but somewhat more
idiomatic.
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202511172336.Y75zj4v6-lkp@intel.com/
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
RFC 4506 defines void as a zero-length type that may appear only as
union arms or as program argument/result types. It cannot be declared
with an identifier, so constructs like "typedef void temp;" are not
valid XDR.
Previously, xdrgen raised a NotImplementedError when it encountered a
void declaration in a typedef. Which was misleading, as the problem is an
invalid RPC specification rather than missing functionality in xdrgen.
This patch replaces the NotImplementedError for _XdrVoid in typedef
handling with a clearer ValueError that specifies incorrect use of void
in the XDR input, making it clear that the issue lies in the RPC
specification being parsed.
Signed-off-by: Khushal Chitturi <kc9282016@gmail.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
When building ynltool with parallel make (-jN), a warning is emitted:
make[1]: warning: jobserver unavailable: using -j1.
Add '+' to parent make rule.
The warning trips up local runs of NIPA's ingest_mdir.py, which
correctly fails on make warnings.
This occurs because SRC_VERSION uses $(shell make ...) to make
kernelversion. The $(shell) function inherits make's MAKEFLAGS env var
which specifies "--jobserver-auth=R,W" pointing to file descriptors that
the invoked make sub-shell does not have access to.
Observed with:
$ make --version | head -1
GNU Make 4.3
Instead of suppressing MAKEFLAGS and foregoing all future MAKEFLAGS
(some of which may be desirable, such as variable overrides) or
introducing a new make target, we instead just ignore the warning by
piping stderr to /dev/null. If 'make kernelversion' fails, the ' || echo
"unknown"' phrase will catch the failure.
Before:
NIPA ingest_mdir.py:
ynl
Full series FAIL (1)
Generated files up to date; build has 1 warnings/errors; no diff in
generated;
After:
NIPA ingest_mdir.py:
Series level tests:
ynl OKAY
Validated output:
$ ./ynltool/ynltool --version
ynltool 6.19.0-rc4
Signed-off-by: Bobby Eshleman <bobbyeshleman@meta.com>
Link: https://patch.msgid.link/20260112-ynl-make-fix-v1-1-c399e76925ad@meta.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
As pointed out during review of the --list-attrs support the GET
ops very often return the same attrs from do and dump. Make the
output more readable by combining the reply information, from:
Do request attributes:
- ifindex: u32
netdev ifindex
Do reply attributes:
- ifindex: u32
netdev ifindex
[ .. other attrs .. ]
Dump reply attributes:
- ifindex: u32
netdev ifindex
[ .. other attrs .. ]
To, after:
Do request attributes:
- ifindex: u32
netdev ifindex
Do and Dump reply attributes:
- ifindex: u32
netdev ifindex
[ .. other attrs .. ]
Tested-by: Gal Pressman <gal@nvidia.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Reviewed-by: Donald Hunter <donald.hunter@gmail.com>
Link: https://patch.msgid.link/20260110233142.3921386-8-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Event and notify handling is quite different from do / dump
handling. Forcing it into print_mode_attrs() doesn't really
buy us anything as events and notifications do not have requests.
Call print_attr_list() directly. Apart form subjective code
clarity this also removes the word "reply" from the output:
Before:
Event reply attributes:
Now:
Event attributes:
Tested-by: Gal Pressman <gal@nvidia.com>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Reviewed-by: Donald Hunter <donald.hunter@gmail.com>
Link: https://patch.msgid.link/20260110233142.3921386-7-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Linus Torvalds
Jakub Kicinski
Chuck Lever
Khushal Chitturi
Michel Lind
Kuniyuki Iwashima
Bobby Eshleman
Donald Hunter