- 21 Jun, 2023 4 commits
-
-
Richard Fitzgerald authored
Always add buses to the stream->master_list in a fixed order. The unique bus->id is used to order the adding of buses to the list. This prevents lockdep asserts and possible deadlocks on streams that have multiple buses. sdw_acquire_bus_lock() takes bus_lock in the order that buses are listed in stream->master_list. do_bank_switch() takes all the msg_lock in the same order. To prevent a lockdep assert, and a possible real deadlock, the relative order of taking these mutexes must always be the same. For example, if a stream takes the mutexes in the order (bus0, bus1) lockdep will assert if another stream takes them in the order (bus1, bus0). More complex relative ordering will also assert, for example if two streams take (bus0, bus1) and (bus1, bus2), then a third stream takes (bus2, bus0). Previously sdw_stream_add_master() simply added the given bus to the end of the list, requiring the caller to guarantee that buses are added in a fixed order. This isn't reasonable or necessary - it's an internal implementation detail that should not be exposed by the API. It doesn't really make sense when there could be multiple independent calling drivers, to say "you must add your buses in the same order as a different driver, that you don't know about, added them". Signed-off-by:
Richard Fitzgerald <rf@opensource.cirrus.com> Reviewed-by:
Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Link: https://lore.kernel.org/r/20230615141208.679011-2-rf@opensource.cirrus.comSigned-off-by:
Vinod Koul <vkoul@kernel.org>
-
Richard Fitzgerald authored
Give the bus_lock and msg_lock of each bus a different unique key so that it is possible to acquire the locks of multiple buses without lockdep asserting a possible deadlock. Using mutex_init() to initialize a mutex gives all those mutexes the same lock class. Lockdep checking treats it as an error to attempt to take a mutex while already holding a mutex of the same class. This causes a lockdep assert when sdw_acquire_bus_lock() attempts to lock multiple buses, and when do_bank_switch() takes multiple msg_lock. [ 138.697350] WARNING: possible recursive locking detected [ 138.697366] 6.3.0-test #1 Tainted: G E [ 138.697380] -------------------------------------------- [ 138.697394] play/903 is trying to acquire lock: [ 138.697409] ffff99b8c41aa8c8 (&bus->bus_lock){+.+.}-{3:3}, at: sdw_prepare_stream+0x52/0x2e0 [ 138.697443] but task is already holding lock: [ 138.697468] ffff99b8c41af8c8 (&bus->bus_lock){+.+.}-{3:3}, at: sdw_prepare_stream+0x52/0x2e0 [ 138.697493] other info that might help us debug this: [ 138.697521] Possible unsafe locking scenario: [ 138.697540] CPU0 [ 138.697550] ---- [ 138.697559] lock(&bus->bus_lock); [ 138.697570] lock(&bus->bus_lock); [ 138.697581] *** DEADLOCK *** Giving each mutex a unique key allows multiple to be held without triggering a lockdep assert. But note that it does not allow them to be taken in one order then a different order. If two mutexes are taken in the order A, B then they must always be taken in that order otherwise they could deadlock. Signed-off-by: -
Krzysztof Kozlowski authored
The 'qcom_swrm_ctrl->pconfig' has size of QCOM_SDW_MAX_PORTS (14), however we index it starting from 1, not 0, to match real port numbers. This can lead to writing port config past 'pconfig' bounds and overwriting next member of 'qcom_swrm_ctrl' struct. Reported also by smatch: drivers/soundwire/qcom.c:1269 qcom_swrm_get_port_config() error: buffer overflow 'ctrl->pconfig' 14 <= 14 Fixes: 9916c02c ("soundwire: qcom: cleanup internal port config indexing") Cc: <stable@vger.kernel.org> Reported-by:
kernel test robot <lkp@intel.com> Reported-by:
Dan Carpenter <error27@gmail.com> Link: https://lore.kernel.org/r/202305201301.sCJ8UDKV-lkp@intel.com/Signed-off-by:
Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Reviewed-by:
Konrad Dybcio <konrad.dybcio@linaro.org> Link: https://lore.kernel.org/r/20230601102525.609627-1-krzysztof.kozlowski@linaro.orgSigned-off-by:
-
Arnd Bergmann authored
The ace2x driver can be build with or without mlink support, but when SND_SOC_SOF_HDA_MLINK is set to =m and soundwire is built-in, it fails with a link error: ld.lld: error: undefined symbol: hdac_bus_eml_sdw_wait_syncpu_unlocked >>> referenced by intel_ace2x.c >>> drivers/soundwire/intel_ace2x.o:(intel_link_power_up) in archive vmlinux.a ld.lld: error: undefined symbol: hdac_bus_eml_sdw_sync_arm_unlocked >>> referenced by intel_ace2x.c >>> drivers/soundwire/intel_ace2x.o:(intel_sync_arm) in archive vmlinux.a Add a Kconfig dependency that prevents that broken configuration but still allows soundwire to be a loadable module instead. Fixes: 4d1e2464 ("soundwire: intel_ace2x: add sync_arm/sync_go helpers") Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Link: https://lore.kernel.org/r/20230616090932.2714714-1-arnd@kernel.orgSigned-off-by:
-
- 08 Jun, 2023 7 commits
-
-
Uday M Bhat authored
SCP registers needs to be updated to accommodate additional register entries as per the Soundwire 1.2 specification. Signed-off-by:
Uday M Bhat <uday.m.bhat@intel.com> Signed-off-by:
Yong Zhi <yong.zhi@intel.com> Reviewed-by:
Ranjani Sridharan <ranjani.sridharan@linux.intel.com> Signed-off-by:
Bard Liao <yung-chuan.liao@linux.intel.com> Link: https://lore.kernel.org/r/20230607020632.1030309-1-yung-chuan.liao@linux.intel.comSigned-off-by:
-
Charles Keepax authored
There is a lot of code using gotos to skip small sections of code, this is a fairly dubious use of a goto, especially when the level of intentation is really low. Most of this code doesn't even breach 80 characters when naively shifted over. Simplify the code a bit, by replacing these unnecessary gotos with simple ifs. Signed-off-by:
Charles Keepax <ckeepax@opensource.cirrus.com> Reviewed-by:
-
Charles Keepax authored
sdw_stream_add_slave/master have flags to indicate if the master or slave runtime where allocated in that call to the function. Currently these flags are cleared on all the paths where the runtime is not allocated, it is more logic and simpler to set the flag on the one path where the runtime is allocated. Signed-off-by:
-
Charles Keepax authored
Version of the code prior to commit d014688e ("soundwire: stream: remove bus->dev from logs on multiple buses"), used bus->dev in the error message after do_bank_switch, this necessitated some checking to ensure the bus pointer was valid. As the code no longer uses bus->dev said checking is now redundant, so remove it. Signed-off-by:
-
Charles Keepax authored
The block_offset variable in _sdw_compute_port_params adds nothing either functionally or in terms of code clarity, remove it. Reviewed-by:
-
Vinod Koul authored
This resolves dependecy for the series 20230602101140.2040141-1-ckeepax@opensource.cirrus.com
-
Charles Keepax authored
The current path that skips allocating the slave runtime does not clear the alloc_slave_rt flag, this is clearly incorrect. Add the missing clear, so the runtime won't be erroneously cleaned up. Fixes: f3016b89 ("soundwire: stream: sdw_stream_add_ functions can be called multiple times") Reviewed-by:
-
- 29 May, 2023 10 commits
-
-
Pierre-Louis Bossart authored
We want to wait for the CONFIG_UPDATE bit to clear before doing something else. Signed-off-by:
Rander Wang <rander.wang@intel.com> Signed-off-by:
-
Pierre-Louis Bossart authored
Combining hardware reset with the multi-link mode leads to a shortened hardware reset pattern observed on the bus. The updated hardware programming sequence is to first enable the clock with the sync_arm/sync_go pattern, and only in a second step to issue the hardware reset sequence. Since there is no longer a dependency between sync_arm/sync_go and hw_reset, the behavior of sdw_cdns_exit_reset() is changed to wait for the self-clearing CONFIG_UPDATE to go back to zero, Link: https://github.com/thesofproject/linux/issues/4170Signed-off-by:
-
Pierre-Louis Bossart authored
It's not clear why we enabled interrupts in the Cadence IP first. The logical programming sequence should be to first start the bus, and only second to enable the interrupts. Signed-off-by:
-
Pierre-Louis Bossart authored
We use __func__ in all calls of sdw_cdns_check_self_clearing_bits(), except in one case. Likely an editing miss when the code was refactored. Signed-off-by:
-
Uwe Kleine-König authored
Returning an error code in the remove callback yields to an error message remove callback returned a non-zero value. This will be ignored. After that the device is removed anyhow. Improve the error message to at least say what the actual problem is. While touching that code, convert the driver to the .remove_new() callback which returns no value with the same effect as returning zero in a .remove() callback. As the return value is ignored by the core the only effect of this patch is to improve the error message. (And the motivating effect is that there is one less driver using .remove().) Signed-off-by:
Uwe Kleine-König <u.kleine-koenig@pengutronix.de> Link: https://lore.kernel.org/r/20230518200823.249795-1-u.kleine-koenig@pengutronix.deSigned-off-by:
-
Krzysztof Kozlowski authored
If pm_runtime_resume_and_get() failed with -EACCES, the driver continued execution and finally called pm_runtime_put_autosuspend(). Since pm_runtime_resume_and_get() drops the usage counter on every error, this lead to double decrement of that counter. Fixes: b275bf45 ("soundwire: debugfs: Switch to sdw_read_no_pm") Signed-off-by:
-
Krzysztof Kozlowski authored
This reverts commit 57ed510b ("soundwire: qcom: use pm_runtime_resume_and_get()") which introduced unbalanced pm_runtime_put(), when device did not have runtime PM enabled. If pm_runtime_resume_and_get() failed with -EACCES, the driver continued execution and finally called pm_runtime_put_autosuspend(). Since pm_runtime_resume_and_get() drops the usage counter on every error, this lead to double decrement of that counter visible in certain debugfs actions on unattached devices (still in reset state): $ cat /sys/kernel/debug/soundwire/master-0-0/sdw:0:0217:f001:00:0/registers qcom-soundwire 3210000.soundwire-controller: swrm_wait_for_wr_fifo_avail err write overflow soundwire sdw-master-0: trf on Slave 1 failed:-5 read addr e36 count 1 soundwire sdw:0:0217:f001:00:0: Runtime PM usage count underflow! Fixes: 57ed510b ("soundwire: qcom: use pm_runtime_resume_and_get()") Signed-off-by:
-
Srinivas Kandagatla authored
WSA Soundwire controller needs an full reset if clock stop support is not available in slave devices. WSA881x does not support clock stop however WSA883x supports clock stop. Make setting this flag at runtime to address above issue. Signed-off-by:
Srinivas Kandagatla <srinivas.kandagatla@linaro.org> Link: https://lore.kernel.org/r/20230525133812.30841-5-srinivas.kandagatla@linaro.orgSigned-off-by:
-
Srinivas Kandagatla authored
Sometimes Hard reset does not clear some of the registers, this sometimes results in firing a bus clash interrupt. Add workaround for this during power up sequence, as suggested by hardware manual. Signed-off-by:
-
Srinivas Kandagatla authored
Wait for Fifo to be empty before going to suspend or before bank switch happens. Just to make sure that all the reads/writes are done. Signed-off-by:
-
- 27 May, 2023 19 commits
-
-
Krzysztof Kozlowski authored
Reverse actions in qcom_swrm_startup() error paths to avoid leaking stream memory and keeping runtime PM unbalanced. Signed-off-by:
-
Krzysztof Kozlowski authored
Drop unused members from the driver state container: struct qcom_swrm_ctrl. Signed-off-by:
-
Chao Song authored
Start from ACE1.x, DOAISE is added to AC timing control register bit 5, it combines with DOAIS to get effective timing, and has the default value 1. The current code fills DOAIS, DACTQE and DODS bits to a variable initialized to zero, and updates the variable to AC timing control register. With this operation, We change DOAISE to 0, and force a much more aggressive timing. The timing is even unable to form a working waveform on SDA pin on Meteorlake. This patch uses read-modify-write operation for the AC timing control register access, thus makes sure those bits not supposed and intended to change are not touched. Signed-off-by:
Chao Song <chao.song@linux.intel.com> Reviewed-by:
-
Pierre-Louis Bossart authored
A BIOS/DMI update seems to have broken some devices, let's add a new mapping. Link: https://github.com/thesofproject/linux/issues/4323Signed-off-by:
-
Pierre-Louis Bossart authored
The interface is not needed for IPC3 but will be needed for ACE2.x+IPC4 combinations, with the substream information passed as a parameter. Signed-off-by:
-
Pierre-Louis Bossart authored
The interface is not needed for IPC3 solution but will be needed with an updated parameter list for ACE2.x+IPC4 combinations. Signed-off-by:
-
Pierre-Louis Bossart authored
The interface is not needed for IPC3 but will be needed for ACE2.x+IPC4 combinations, with the substream information passed as a parameter. Signed-off-by:
-
Pierre-Louis Bossart authored
The interface is not needed for IPC3 solutions but will be needed with an updated parameter list for ACE2.x+IPC4 combinations. Signed-off-by:
-
Pierre-Louis Bossart authored
An earlier simplification to only pass the direction is no longer suitable, all the ACE2.x HDaudio DMA management relies on access to the substream structure. This patch is an iso-functionality change, the HDaudio DMA parts will be provided separately. Signed-off-by:
-
Pierre-Louis Bossart authored
The .pre_ and .post_switch callbacks are mandatory. Signed-off-by:
-
Pierre-Louis Bossart authored
Add the abstraction needed to only program the LSDIID registers for the HDaudio extended links. It's perfectly fine to program this register multiple times in case devices lose sync and reattach. Signed-off-by:
Péter Ujfalusi <peter.ujfalusi@linux.intel.com> Reviewed-by:
-
Pierre-Louis Bossart authored
When a peripheral reports as ATTACHED, the manager may need to follow a programming sequence, e.g. to assign DMA resources and/or assign a command queue for that peripheral. This patch adds an optional callback, which will be invoked every time the peripheral attaches. This might be overkill in some scenarios, and one could argue that this should be invoked only on the first attachment. The bus does not however track this first attachment with any existing state-mirroring variable, and using dev_num_sticky would not work across suspend-resume cycles. Signed-off-by:
-
Pierre-Louis Bossart authored
This is the last callback needed for all bus management routines on new hardware. Same concept as before, just different register. Signed-off-by:
-
Pierre-Louis Bossart authored
The WAKEEN and WAKESTS registers were moved to the per-link SHIM_VS area. Signed-off-by:
-
Pierre-Louis Bossart authored
The sequences are so far identical, so the abstraction is a bit over-engineered. In time we will simplify if there is no need to special case or work-around programming sequences. Signed-off-by:
-
Pierre-Louis Bossart authored
Same functionality as before, but with the registers moved to the HDaudio multi-link area. Signed-off-by:
-
Pierre-Louis Bossart authored
The code is similar to the previous implementation, the only difference is that the PDI descriptors are now in different areas. Using common helpers proves tricky with multiple changed registers, workarounds that are no longer necessary. It's simpler to duplicate the intel_register_dai() function rather than try to add multiple levels of abstraction and indirections. Signed-off-by:
-
Pierre-Louis Bossart authored
Unlike previous hardware generations, the glue-to-master transition is not managed by software, instead the transitions are managed as part of the power-up/down sequences controlled by SPA/CPA bits. The only thing that's required is to configure the link PHY for 'normal' operation instead of the PHY test mode. Signed-off-by:
-
Pierre-Louis Bossart authored
The registers used for multi-link synchronization are no longer in the SHIM but in the HDaudio multi-link capability space. Use helpers to configure the SYNCPRD value, and wait for SYNCPU to change after powering-up. Note that the SYNCPRD value is shared between all sublinks, for obvious reasons if those links are supposed to be synchronized. The value of SYNCPRD is programmed only once for all sublinks. Signed-off-by:
-
