commit 2d1f7b65f5 ("dm cache policy smq: fix missing locks in
invalidating cache blocks") added mq->lock around the destructive part of
smq_invalidate_mapping(), but left the e->allocated check outside the
critical section.
That leaves a check-then-act race. Two concurrent invalidators can both
observe e->allocated as true before either of them takes mq->lock. The
first invalidator that acquires the lock removes the entry from the
queues and hash table and then calls free_entry(), which clears
e->allocated and puts the entry back on the free list. The second
invalidator can then acquire mq->lock and continue with the stale result
of the unlocked check.
This can corrupt the SMQ queues or hash table by deleting an entry that
is no longer on those structures. It can also hit the allocation check in
free_entry() when the same entry is freed again.
Move the allocation check under mq->lock so the predicate and the
destructive operations are serialized by the same lock.
Fixes: 2d1f7b65f5 ("dm cache policy smq: fix missing locks in invalidating cache blocks")
Signed-off-by: Guangshuo Li <lgs201920130244@gmail.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Since describing the PCIe PERST# property under Host Bridge node is now
deprecated, it is recommended to add it to the Root Port node, so
creating the Root Port node and add the reset-gpios property in Root
Port.
Signed-off-by: Sherry Sun <sherry.sun@nxp.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Since describing the PCIe PERST# property under Host Bridge node is now
deprecated, it is recommended to add it to the Root Port node, so
creating the Root Port node and add the reset-gpios property in Root
Port.
Signed-off-by: Sherry Sun <sherry.sun@nxp.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Since describing the PCIe PERST# property under Host Bridge node is now
deprecated, it is recommended to add it to the Root Port node, so
creating the Root Port node and add the reset-gpios property in Root
Port.
Signed-off-by: Sherry Sun <sherry.sun@nxp.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
The new get-reqs dump operations added to sunrpc_cache.yaml and
nfsd.yaml place the "requests" nested attribute under dump.request.
A netlink dump carries an empty request; its payload travels back
in the reply. Because the spec names no reply attributes, the YNL
C code generator synthesizes a forward reference to a
<op>_rsp struct that is never defined, breaking any consumer of
these specs.
This first surfaced when Thorsten Leemhuis built tools/net/ynl
against -next:
nfsd-user.h:746: error: field 'obj' has incomplete type
struct nfsd_svc_export_get_reqs_rsp obj ...
nfsd-user.h:826: error: field 'obj' has incomplete type
struct nfsd_expkey_get_reqs_rsp obj ...
nfsd-user.c:1211: error: 'nfsd_svc_export_get_reqs_rsp_parse'
undeclared
sunrpc_cache.yaml has the same defect in ip-map-get-reqs and
unix-gid-get-reqs, but nfsd.yaml errors out first in the Makefile's
alphabetical build order and hides the sunrpc failures.
These bugs were introduced by incorrect merge conflict resolution.
Reported-by: Thorsten Leemhuis <linux@leemhuis.info>
Closes: https://lore.kernel.org/linux-nfs/f6a3ca6d-e5cb-4a5c-9af2-8d2b1ce33ef0@leemhuis.info/
Fixes: 1045ccf519ce30 ("sunrpc: add netlink upcall for the auth.unix.ip cache")
Tested-by: Thorsten Leemhuis <linux@leemhuis.info>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Add a new NFSD_CMD_CACHE_FLUSH generic netlink command that allows
userspace to flush the nfsd export caches (svc_export and expkey)
without writing to /proc/net/rpc/*/flush.
An optional NFSD_A_CACHE_FLUSH_MASK u32 attribute selects which caches
to flush (bit 1 = svc_export, bit 2 = expkey). If the attribute is
omitted, all nfsd caches are flushed.
This is used by exportfs to replace its /proc-based cache_flush() with a
netlink equivalent, with /proc fallback for older kernels.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Add a new SUNRPC_CMD_CACHE_FLUSH generic netlink command that allows
userspace to flush the sunrpc auth caches (ip_map and unix_gid) without
writing to /proc/net/rpc/*/flush.
An optional SUNRPC_A_CACHE_FLUSH_MASK u32 attribute selects which caches
to flush (bit 1 = ip_map, bit 2 = unix_gid). If the attribute is
omitted, all sunrpc caches are flushed.
This is used by exportfs to replace its /proc-based cache_flush() with a
netlink equivalent, with /proc fallback for older kernels.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Add netlink-based cache upcall support for the expkey (nfsd.fh) cache,
following the same pattern as the existing svc_export netlink support.
Add expkey to the cache-type enum, a new expkey attribute-set with
client, fsidtype, fsid, negative, expiry, and path fields, and the
expkey-get-reqs / expkey-set-reqs operations to the nfsd YAML spec
and generated headers.
Implement nfsd_nl_expkey_get_reqs_dumpit() which snapshots pending
expkey cache requests and sends each entry's seqno, client name,
fsidtype, and fsid over netlink.
Implement nfsd_nl_expkey_set_reqs_doit() which parses expkey cache
responses from userspace (client, fsidtype, fsid, expiry, and path
or negative flag) and updates the cache via svc_expkey_lookup() /
svc_expkey_update().
Wire up the expkey_notify() callback in svc_expkey_cache_template
so cache misses trigger NFSD_CMD_CACHE_NOTIFY multicast events with
NFSD_CACHE_TYPE_EXPKEY.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Add netlink-based cache upcall support for the svc_export (nfsd.export)
cache to Documentation/netlink/specs/nfsd.yaml and regenerate the
resulting files.
Implement nfsd_cache_notify() which sends a NFSD_CMD_CACHE_NOTIFY
multicast event to the "exportd" group, carrying the cache type so
userspace knows which cache has pending requests.
Implement nfsd_nl_svc_export_get_reqs_dumpit() which snapshots
pending svc_export cache requests and sends each entry's seqno,
client name, and path over netlink.
Implement nfsd_nl_svc_export_set_reqs_doit() which parses svc_export
cache responses from userspace (client, path, expiry, flags, anon
uid/gid, fslocations, uuid, secinfo, xprtsec, fsid, or negative
flag) and updates the cache via svc_export_lookup() /
svc_export_update().
Wire up the svc_export_notify() callback in svc_export_cache_template
so cache misses trigger NFSD_CMD_CACHE_NOTIFY multicast events with
NFSD_CACHE_TYPE_SVC_EXPORT.
Note that the export-flags and xprtsec-mode enums are organized to match
their counterparts in include/uapi/linux/nfsd/export.h. The intent is
that future export options will only be added to the netlink headers,
which should eliminate the need to keep so much in sync.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Add netlink-based cache upcall support for the unix_gid (auth.unix.gid)
cache, using the sunrpc generic netlink family.
Add unix-gid attribute-set (seqno, uid, gids multi-attr, negative,
expiry), unix-gid-reqs wrapper, and unix-gid-get-reqs /
unix-gid-set-reqs operations to the sunrpc_cache YAML spec and
generated headers.
Implement sunrpc_nl_unix_gid_get_reqs_dumpit() which snapshots pending
unix_gid cache requests and sends each entry's seqno and uid over
netlink.
Implement sunrpc_nl_unix_gid_set_reqs_doit() which parses unix_gid
cache responses from userspace (uid, expiry, gids as u32 multi-attr
or negative flag) and updates the cache via unix_gid_lookup() /
sunrpc_cache_update().
Wire up unix_gid_notify() callback in unix_gid_cache_template so
cache misses trigger SUNRPC_CMD_CACHE_NOTIFY multicast events with
SUNRPC_CACHE_TYPE_UNIX_GID.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Add netlink-based cache upcall support for the ip_map (auth.unix.ip)
cache, using the sunrpc generic netlink family.
Add ip-map attribute-set (seqno, class, addr, domain, negative, expiry),
ip-map-reqs wrapper, and ip-map-get-reqs / ip-map-set-reqs operations
to the sunrpc_cache YAML spec and generated headers.
Implement sunrpc_nl_ip_map_get_reqs_dumpit() which snapshots pending
ip_map cache requests and sends each entry's seqno, class name, and
IP address over netlink.
Implement sunrpc_nl_ip_map_set_reqs_doit() which parses ip_map cache
responses from userspace (class, addr, expiry, and domain name or
negative flag) and updates the cache via __ip_map_lookup() /
__ip_map_update().
Wire up ip_map_notify() callback in ip_map_cache_template so cache
misses trigger SUNRPC_CMD_CACHE_NOTIFY multicast events with
SUNRPC_CACHE_TYPE_IP_MAP.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
The auth.unix.ip and auth.unix.gid caches live in the sunrpc module,
so they cannot use the nfsd generic netlink family. Create a new
"sunrpc" generic netlink family with its own "exportd" multicast
group to support cache upcall notifications for sunrpc-resident
caches.
Define a YAML spec (sunrpc_cache.yaml) with a cache-type enum
(ip_map, unix_gid), a cache-notify multicast event, and the
corresponding uapi header.
Implement sunrpc_cache_notify() in cache.c, which checks for
listeners on the exportd multicast group, builds and sends a
SUNRPC_CMD_CACHE_NOTIFY message with the cache-type attribute.
Register/unregister the sunrpc_nl_family in init_sunrpc() and
cleanup_sunrpc().
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Add sunrpc_cache_requests_count() and sunrpc_cache_requests_snapshot()
to allow callers to count and snapshot the pending upcall request list
without exposing struct cache_request outside of cache.c.
Both functions skip entries that no longer have CACHE_PENDING set.
The snapshot function takes a cache_get() reference on each item so the
caller can safely use them after the queue_lock is released.
These will be used by the nfsd generic netlink dumpit handler for
svc_export upcall requests.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
A later patch will be changing the kernel to send a netlink notification
when there is a pending cache_request. Add a new cache_notify operation
to struct cache_detail for this purpose.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Rename cache_pipe_upcall() to cache_do_upcall() in anticipation of the
addition of a netlink-based upcall mechanism.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
This function doesn't have anything to do with a timeout. The only
difference is that it warns if there are no listeners. Rename it to
sunrpc_cache_upcall_warn().
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Since it will soon also send an upcall via netlink, if configured.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
When revoking delegation state, nfsd4_revoke_states() takes an extra
reference on the stid before calling unhash_delegation_locked(). If
unhash_delegation_locked() returns false (the delegation was already
unhashed by a concurrent path), dp is set to NULL and
revoke_delegation() is skipped, but the extra reference is never
released. Each occurrence permanently pins the stid in memory. The
leaked reference also prevents nfs4_put_stid() from decrementing
cl_admin_revoked, leaving the counter permanently inflated.
Drop the extra reference in the failure path.
Fixes: 8dd91e8d31 ("nfsd: fix race between laundromat and free_stateid")
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
svc_xprt_resource_released() calls svc_xprt_enqueue()
whenever XPT_DATA or XPT_DEFERRED is set. During RPC
processing, svc_reserve_auth() reduces the reservation
counter and triggers this path while the current thread
still holds XPT_BUSY. The enqueue enters svc_xprt_ready(),
executes an smp_rmb(), READ_ONCE(), and tracepoint, then
returns false on seeing XPT_BUSY.
Trace data from a 256KB NFSv3 WRITE workload over TCP
shows this pattern generates roughly 195,000 wasted
enqueue calls -- approximately one per RPC -- each
paying the full svc_xprt_ready() cost for no benefit.
Add a BUSY check alongside the existing DATA|DEFERRED
check in svc_xprt_resource_released(). When the
transport is BUSY, the holder will call
svc_xprt_received() upon completion, which already
checks for pending work flags and re-enqueues.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
svc_xprt_received() unconditionally calls
svc_xprt_enqueue() after clearing XPT_BUSY. When no
work flags are pending, the enqueue traverses
svc_xprt_ready() -- executing an smp_rmb(), READ_ONCE(),
and tracepoint -- before returning false.
Trace data from a 256KB NFSv3 workload over RDMA shows
85% of svc_xprt_received() invocations reach
svc_xprt_enqueue() with no pending work flags. In the
WRITE phase, 167,335 of 196,420 calls find no work; in
the READ phase, 97,165 of 98,276. Each unnecessary call
executes a memory barrier, a flags read, and (when
tracing is active) fires the svc_xprt_enqueue
tracepoint.
Add a flags pre-check between clear_bit(XPT_BUSY) and
svc_xprt_enqueue(). Both the clear and the subsequent
READ_ONCE operate on the same xpt_flags word, so
cache-line serialization of the atomic bitops ensures
the read observes any flag set by a concurrent producer
before the line was acquired for the clear. If a
producer's set_bit occurs after the clear_bit, that
producer's own svc_xprt_enqueue() call observes
!XPT_BUSY and dispatches the transport.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
svc_reserve() and svc_xprt_release_slot() call
svc_xprt_enqueue() after modifying xpt_reserved or
xpt_nr_rqsts. The purpose is to re-dispatch the
transport when write-space or a slot becomes available.
However, when neither XPT_DATA nor XPT_DEFERRED is
set, no thread can make progress on the transport and
the enqueue accomplishes nothing.
Trace data from a 256KB NFSv3 WRITE workload over RDMA
shows 11.2 svc_xprt_enqueue() calls per RPC. Of these,
6.9 per RPC lack XPT_DATA and exit svc_xprt_ready()
immediately after executing the smp_rmb(), READ_ONCE(),
and tracepoint. svc_reserve() and svc_xprt_release_slot()
account for roughly five of these per RPC.
A new helper, svc_xprt_resource_released(), checks
XPT_DATA | XPT_DEFERRED before calling
svc_xprt_enqueue(). The existing smp_wmb() barriers
are upgraded to smp_mb() to ensure the flags check
observes a concurrent producer's set_bit(XPT_DATA).
Each producer (svc_rdma_wc_receive, etc.) both sets
XPT_DATA and calls svc_xprt_enqueue(), so even if the
check reads a stale value, the producer's own enqueue
provides a fallback path.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Troy Mitchell <troy.mitchell@linux.spacemit.com> says:
This series adds K3 SoC support for the SpacemiT I2S controller driver.
Patch 1 adds the dt-bindings for the spacemit,k3-i2s compatible. The K3
SoC uses the same I2S IP as K1 but requires additional clocks
(sysclk_div, c_sysclk, c_bclk) that are shared across multiple I2S
controllers.
Patch 2 adds driver support for K3 SoC with additional clocks.
Link: https://patch.msgid.link/20260522-k3-i2s-v3-0-08e3eb811d93@linux.spacemit.com
Add support for the SpacemiT K3 SoC I2S controller, which shares the
same IP as K1 but requires additional clocks: sysclk_div, c_sysclk,
and c_bclk. These clocks only exist on K3 and are not present on K1.
The sysclk_div clock is present on most K3 I2S controllers except I2S1.
The c_sysclk and c_bclk clocks are shared across multiple I2S
controllers on K3.
Use devm_clk_get_optional_enabled() to acquire these clocks so that
the driver works on both K1 (where they are absent) and K3 without
needing SoC-specific match data. For K3, the sysclk_div rate is set
before sysclk in set_sysclk, and the common clock rates are configured
in hw_params based on the sample rate.
Signed-off-by: Troy Mitchell <troy.mitchell@linux.spacemit.com>
Link: https://patch.msgid.link/20260522-k3-i2s-v3-2-08e3eb811d93@linux.spacemit.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Add the spacemit,k3-i2s compatible string for the K3 SoC I2S
controller. The K3 I2S IP is the same as K1 but requires additional
clocks: a dedicated sysclk_div clock, along with common_sysclk and
common_bclk which are shared across multiple I2S controllers on K3.
Acked-by: Rob Herring (Arm) <robh@kernel.org>
Signed-off-by: Troy Mitchell <troy.mitchell@linux.spacemit.com>
Link: https://patch.msgid.link/20260522-k3-i2s-v3-1-08e3eb811d93@linux.spacemit.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The firmware container allocations passed cont->size + sizeof(int) to
kzalloc(), which was wrong: the struct contains an int len followed by a
u8 data[] flexible array. It ended up being the same as the struct's
size is only the int member but still wrong.
Use the modern struct_size helper for this.
Add __counted_by for extra runtime analysis.
Assisted-by: Claude:Opus-4.7
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20260523011749.101555-1-rosenp@gmail.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The Technologic Systems TS-4800 watchdog node was previously named 'wdt',
which violates the core watchdog.yaml schema expecting generic node names.
Rename the node to 'watchdog' to fix the following dtbs_check warning:
'wdt' does not match '^(pmic|timer|watchdog)(@.*|-([0-9]|[1-9][0-9]+))?$'
Signed-off-by: Eduard Bostina <egbostina@gmail.com>
Reviewed-by: Daniel Baluta <daniel.baluta@nxp.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Convert the Technologic Systems TS-4800 watchdog timer bindings
to DT schema.
Signed-off-by: Eduard Bostina <egbostina@gmail.com>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Cezary Rojewski <cezary.rojewski@intel.com> says:
Set of changes composed of one fix and two improvements.
The fix leads the series and addresses "unbalanced disables" coming from
the regulator core during S4 (hibernation) scenario.
The SLEEP_PM_OPS are unset for the driver. Hibernation (S4) causes no
resume (skipped thanks to smart_suspend=true) yet still performs the
suspend sequence unconditionally, see device_complete() in
drivers/base/power/main.c. In essence, we end up with double suspend
(double disable) and thus the warning. Exemplary stack:
unbalanced disables for regulator-dummy
WARNING: drivers/regulator/core.c:3244 at _regulator_disable+0x161/0x270, CPU#6: kworker/u64:2/120
(...)
Workqueue: pm pm_runtime_work
RIP: 0010:_regulator_disable+0x161/0x270
(...)
Call Trace:
<TASK>
regulator_bulk_disable+0xe4/0x220
? __pfx_regulator_bulk_disable+0x10/0x10
? pick_next_task_fair+0x36f/0x11f0
? __pfx_regmap_unlock_mutex+0x10/0x10
? regmap_unlock_mutex+0xe/0x20
pcm3168a_rt_suspend+0x57/0x90 [snd_soc_pcm3168a]
? __pfx_pcm3168a_rt_suspend+0x10/0x10 [snd_soc_pcm3168a]
__rpm_callback+0x69/0x200
? __pfx_pcm3168a_rt_suspend+0x10/0x10 [snd_soc_pcm3168a]
rpm_callback+0xbc/0xd0
rpm_suspend+0x225/0x960
? __pfx_rpm_suspend+0x10/0x10
? __kasan_check_write+0x14/0x20
? _raw_spin_lock_irq+0x8d/0xe0
? __pfx__raw_spin_lock_irq+0x10/0x10
? pwq_dec_nr_in_flight+0x63/0x4f0
pm_runtime_work+0xd0/0xf0
In regard to the improvements, both aim to drop redundant operations.
One targets pm_runtime_idle() - no need to fire it manually,
device-driver core will do that for us - while the second replaces
preprocessor directive with pm_runtime_status_suspended() check.
No !CONFIG_PM dependency equals better code coverege with default
kconfigs.
Link: https://patch.msgid.link/20260525201801.1336936-1-cezary.rojewski@intel.com
Revert changes done in commit 489db5d941 ("ASoC: pcm3168a:
Don't disable pcm3168a when CONFIG_PM defined") and add
pm_runtime_status_suspended() check.
The suspended-check addresses regulator's "unbalanced disables"
warning during driver removal even when CONFIG_PM is enabled.
Signed-off-by: Cezary Rojewski <cezary.rojewski@intel.com>
Link: https://patch.msgid.link/20260525201801.1336936-4-cezary.rojewski@intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Device-driver core runs pm_request_idle() right after the probing
sequence already, see __driver_probe_device().
Signed-off-by: Cezary Rojewski <cezary.rojewski@intel.com>
Link: https://patch.msgid.link/20260525201801.1336936-3-cezary.rojewski@intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The SLEEP_PM_OPS are unset for the driver. Hibernation (S4) causes no
resume (skipped thanks to smart_suspend=true) yet still performs the
suspend sequence unconditionally, see device_complete() in
drivers/base/power/main.c.
If S4 runs for already suspended pcm3168a device, we end up with
"unbalanced disables" warning from the regulators. Assigning the
operations fixes the problem.
Signed-off-by: Cezary Rojewski <cezary.rojewski@intel.com>
Link: https://patch.msgid.link/20260525201801.1336936-2-cezary.rojewski@intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Replace BUG_ON(!mutex_is_locked(&event_mutex)) with
lockdep_assert_held(&event_mutex) in uprobe_buffer_enable() and
uprobe_buffer_disable().
BUG_ON() will crash the kernel. mutex_is_locked() only checks
if any task holds lock,but not the caller task. lockdep_assert_held()
also check current task for lock and no crash on true condition.
Link: https://lore.kernel.org/all/20260521192846.8306-1-yashsuthar983@gmail.com/
Signed-off-by: Yash Suthar <yashsuthar983@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Store probe entry fetch instructions in the probe_entry_arg
allocation instead of allocating a separate instruction array.
This keeps the entry fetch code tied to the entry argument lifetime while
leaving regular probe_arg instruction arrays separately allocated and
freed.
Assisted-by: Codex:GPT-5.5
Link: https://lore.kernel.org/all/20260520215817.16560-1-rosenp@gmail.com/
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
John Madieu <john.madieu.xa@bp.renesas.com> says:
Add audio support for the Renesas RZ/G3E SoC to the R-Car Sound
driver. The RZ/G3E audio subsystem is based on R-Car Sound IP but
has several differences requiring dedicated handling:
- SSI operates exclusively in BUSIF mode (no PIO)
- 2-4 BUSIF channels per SSI (layout differs from R-Car)
- Separate register regions for SCU, ADG, SSIU, SSI accessed by name
- Per-SSI ADG and SSIF supply clocks
- Dedicated audmapp clock/reset for Audio DMAC peri-peri
- Per-SSI and per-module reset controllers via CPG
- Unprefixed DT sub-node names (ssi, ssiu, src, ...) instead of
rcar_sound,xxx
- Hyphenated indexed clock/reset names (ssi-0, src-0, adg-ssi-0,
audio-clka, ...) instead of the legacy dotted form
Link: https://patch.msgid.link/20260525110230.4014435-1-john.madieu.xa@bp.renesas.com
Add system suspend/resume support for the ASoC rsnd driver, required
for RZ/G3E platforms. Distribute the per-module suspend/resume work
across the relevant files (adg.c, ssi.c, ssiu.c, src.c, ctu.c, mix.c,
dvc.c, dma.c) rather than centralising it in core.c.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-19-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The RZ/G3E device tree binding uses standard unprefixed node names
("ssi", "ssiu", "src", "dvc", "mix", "ctu") instead of the legacy
"rcar_sound," prefixed names used by R-Car bindings.
Convert rsnd_parse_of_node() from a macro into a function that tries
the legacy prefixed name first, then falls back to the unprefixed name
by stripping the "rcar_sound," prefix. This makes the driver work
transparently with both old and new bindings.
While at it, update the related comments in dma.c, ssi.c and ssiu.c
that reference the hardcoded "rcar_sound,ssiu" / "rcar_sound,ssi"
names to note that the driver now accepts both the prefixed and the
unprefixed forms.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-18-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The RZ/G3E SoC exposes a shared SCU reset controller used by all SRC
modules. Acquire it once and pass it through per-instance's
rsnd_mod_init() so it is wired into the rsnd_mod->rstc plumbing.
devm_reset_control_get_optional_shared() returns NULL when no reset
is described in DT, leaving existing R-Car generations unaffected.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-17-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The RZ/G3E SoC requires explicit SCU (Sampling Rate Converter Unit)
clock management:
- scu: SCU top-level module clock (CPG_CLKON_15.CLK6_ON)
- scu_x2: SCU top-level double-rate clock (CPG_CLKON_15.CLK7_ON)
- scu_supply: SCU register-access / housekeeping clock
(CPG_CLKON_23.CLK14_ON)
Without every one of them enabled, no SCU register is reachable.
Hold these in a new struct rsnd_src_ctrl and acquire them with
devm_clk_get_optional_enabled(). scu_supply is intentionally left
untouched by the system suspend/resume path added later in the
series, so SCU registers stay reachable across PM transitions.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-16-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The R-Car Sound ADG block has up to four external master-clock inputs
named CLKA, CLKB, CLKC and CLKI by the silicon. On Gen2 R-Car these
come from DT under the legacy names "clk_a", "clk_b", "clk_c", "clk_i"
defined by renesas,rsnd.yaml. Gen4 collapses them to a single "clkin".
The new standalone RZ/G3E sound binding (renesas,r9a09g047-sound.yaml)
uses the standard DT naming convention with a vendor-meaningful prefix
that matches the SoC datasheet pin labels: "audio-clka", "audio-clkb",
"audio-clkc", "audio-clki".
Add a third clkin name table for RZ/G3E and dispatch to it from
rsnd_adg_get_clkin() in the same style as the existing Gen4 branch.
The CLKA/B/C/I enum values, the clkin[] array, and the BRGA/BRGB
derivation are unchanged - only the DT lookup names differ.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-15-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
RZ/G3E's ADG module requires explicit clock management for SSI audio
interfaces that differs from R-Car Gen2/Gen3/Gen4:
- Per-SSI ADG clocks (adg-ssi-N, or adg.ssi.N in legacy bindings)
for each SSI module
- A shared SSIF supply clock for the SSI subsystem
These clocks are acquired using optional APIs, making them transparent
to platforms that do not require them.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-14-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
RZ/G3E requires the ADG reset line to be deasserted for the audio
subsystem to operate. The ADG module clock is already managed via
rsnd_adg_clk_enable/disable() through adg->adg, so no additional
clock handling is needed.
Add support for the optional "adg" reset control on Renesas RZ/G3E SoC.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-13-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Acquire the per-SSI reset controller and pass it through
rsnd_mod_init() so it is wired into the rsnd_mod->rstc plumbing.
The RZ/G3E SoC exposes one reset line per SSI instance. Use the
indexed-name rsnd_devm_reset_control_get_optional_indexed() helper
so the same code accepts both the hyphenated RZ/G3E names
("ssi-0", "ssi-1", ...) and the legacy dotted names used by R-Car
("ssi.0", ...).
The helper returns NULL when no reset is described in DT, leaving
existing R-Car generations unaffected.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-12-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Add support for the SSIU found on the Renesas RZ/G3E SoC, which
provides a different BUSIF layout compared to earlier generations:
- SSI0-SSI4: 4 BUSIF instances each (BUSIF0-3)
- SSI5-SSI8: 1 BUSIF instance each (BUSIF0 only)
- SSI9: 4 BUSIF instances (BUSIF0-3)
- Total: 28 BUSIFs
The RZ/G3E also has only two pairs of BUSIF error-status registers
instead of four, and the SSI always operates in BUSIF mode: the
SSI_MODE0 BUSIF/PIO select bit is not implemented and must not be written.
While at it, add RSND_SSIU_BUSIF_STATUS_COUNT_2 as a capability flag in
the match data, consumed via struct rsnd_ssiu_ctrl, to parametrise the two
BUSIF error-status loops.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-11-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The RZ/G3E SoC exposes a single shared "ssi-all" reset that gates all
SSI/SSIU modules. Acquire it at SSIU probe and pass it through
rsnd_mod_init() so it is wired into the rsnd_mod->rstc plumbing.
devm_reset_control_get_optional_shared() returns NULL when no reset is
described in DT, leaving existing R-Car generations unaffected.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-10-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
RZ/G3E has different DMA register base addresses and offset
calculations compared to R-Car platforms.
Add dedicated rsnd_rzg3e_dma_addr() function with dispatch from
rsnd_dma_addr(), following the existing per-generation pattern.
The function reuses rsnd_dma_addr_lookup() and rsnd_dma_addr_map.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-9-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Replace the raw multi-dimensional array used for DMA address lookup in
rsnd_gen2_dma_addr() with properly named structs: rsnd_dma_addr (in/out
pair), rsnd_dma_addr_dir (capture/playback arrays), and
rsnd_dma_addr_map (src/ssi/ssiu module sets).
While at it, extract the common lookup logic (is_ssi / use_src / use_cmd
evaluation and table indexing) into a shared rsnd_dma_addr_lookup()
function.
No functional change. This is a preparatory refactor for upcoming RZ/G3E
support which will add its own DMA address map using the same struct and
lookup function.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-8-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
RZ/G3E requires additional audmapp clock and reset lines for
Audio DMA-PP operation.
Add global audmapp clock/reset management in rsnd_dma_probe()
using optional APIs to remain transparent to other platforms.
Signed-off-by: John Madieu <john.madieu.xa@bp.renesas.com>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Link: https://patch.msgid.link/20260525110230.4014435-7-john.madieu.xa@bp.renesas.com
Signed-off-by: Mark Brown <broonie@kernel.org>
