March 2023 - Alsa-devel - mailman.alsa-project.org

[PATCH V5 0/8] Add SoundWire support for AMD platforms
by Vijendar Mukunda 06 Mar '23

06 Mar '23

ACP IP(v6.x) block has two SoundWire manager instance support. This patchset adds support for AMD SoundWire manager driver. Signed-off-by: Vijendar Mukunda <Vijendar.Mukunda(a)amd.com> changes since v4: - fix nit-picks in the code. - update naming convention for control word and response buffer vairables. - drop pm_suspend check in prepare callback. - use return statement instead of returning ret variable. changes since v3: - add usleep_range() in command/response implementation. - add usleep_range() in clock stop sequence. - modify usleep_range() values throughout the code. - remove unncessary debug statement from code. - update comment in probe() call. - fix timeout condition checks in the code. - drop "ret" variable in amd_disable_sdw_manager(). Changes since v2: - Remove useless variable initializations. - Add helper function to interpret peripheral status. - Move runtime pm sequence to probe_work workqueue. - Use string "SoundWire" instead of "soundwire" in code. - Update comments in interrupt handler and probe sequence. - Rename "sdw_lock" as "acp_sdw_lock". - Remove __func__ from dev_dbg statements. Changes since v1: - Drop asoc tree based patches. will send asoc patches as a separate series. - Fixed double space errors. - Use dev instead of pci->dev. - Use SoundWire manager terminology. - Remove amd_sdw_compute_slave_ports() function and use exported sdw_compute_slave_ports() function. - Remove unused variable "num_ports" from amd_manager structure. - Drop startup and shutdown dai callbacks. - Drop reset_page_addr callback. - Use relative address offset to program SoundWire manager registers throughout the code. - Separate wake enable interrupt handling from slave status handling logic. - Use acp_mmio to program ACP common registers. - Use dai_runtime_array implementation in dai_ops. - Refactor port_ops callbacks. - Add comments in port_ops callbacks. - Add retry count logic in irq thread to address faulty case. - Add helper function to interpret command response. - Add generic bandwidth allocation dependency in Kconfig options. - Add comments for AMD SoundWire power modes. - Add missing timeout check in amd_init_sdw_manager callback. - Declare frameshape parameters in probe call. - Handle error case in clock stop sequence. - Add comments in pm_prepare and pm_ops callbacks. Vijendar Mukunda (8): soundwire: export sdw_compute_slave_ports() function soundwire: amd: Add support for AMD Manager driver soundwire: amd: register SoundWire manager dai ops soundwire: amd: enable build for AMD SoundWire manager driver soundwire: amd: add SoundWire manager interrupt handling soundwire: amd: add runtime pm ops for AMD SoundWire manager driver soundwire: amd: handle SoundWire wake enable interrupt soundwire: amd: add pm_prepare callback and pm ops support drivers/soundwire/Kconfig | 10 + drivers/soundwire/Makefile | 4 + drivers/soundwire/amd_manager.c | 1324 +++++++++++++++++ drivers/soundwire/amd_manager.h | 275 ++++ drivers/soundwire/bus.h | 9 + .../soundwire/generic_bandwidth_allocation.c | 12 +- include/linux/soundwire/sdw_amd.h | 109 ++ 7 files changed, 1734 insertions(+), 9 deletions(-) create mode 100644 drivers/soundwire/amd_manager.c create mode 100644 drivers/soundwire/amd_manager.h create mode 100644 include/linux/soundwire/sdw_amd.h -- 2.34.1

3 10

[PATCH AUTOSEL 5.10 30/30] soundwire: cadence: Drain the RX FIFO after an IO timeout
by Sasha Levin 03 Mar '23

03 Mar '23

From: Richard Fitzgerald <rf(a)opensource.cirrus.com> [ Upstream commit 0603a47bd3a8f439d7844b841eee1819353063e0 ] If wait_for_completion_timeout() times-out in _cdns_xfer_msg() it is possible that something could have been written to the RX FIFO. In this case, we should drain the RX FIFO so that anything in it doesn't carry over and mess up the next transfer. Obviously, if we got to this state something went wrong, and we don't really know the state of everything. The cleanup in this situation cannot be bullet-proof but we should attempt to avoid breaking future transaction, if only to reduce the amount of error noise when debugging the failure from a kernel log. Note that this patch only implements the draining for blocking (non-deferred) transfers. The deferred API doesn't have any proper handling of error conditions and would need some re-design before implementing cleanup. That is a task for a separate patch... Signed-off-by: Richard Fitzgerald <rf(a)opensource.cirrus.com> Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart(a)linux.intel.com> Link: https://lore.kernel.org/r/20221202161812.4186897-4-rf@opensource.cirrus.com Signed-off-by: Vinod Koul <vkoul(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- drivers/soundwire/cadence_master.c | 50 ++++++++++++++++-------------- 1 file changed, 27 insertions(+), 23 deletions(-) diff --git a/drivers/soundwire/cadence_master.c b/drivers/soundwire/cadence_master.c index 292c4460eaaa3..18e7d158fcca4 100644 --- a/drivers/soundwire/cadence_master.c +++ b/drivers/soundwire/cadence_master.c @@ -511,6 +511,29 @@ cdns_fill_msg_resp(struct sdw_cdns *cdns, return SDW_CMD_OK; } +static void cdns_read_response(struct sdw_cdns *cdns) +{ + u32 num_resp, cmd_base; + int i; + + /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ + BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); + + num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); + num_resp &= CDNS_MCP_RX_FIFO_AVAIL; + if (num_resp > ARRAY_SIZE(cdns->response_buf)) { + dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); + num_resp = ARRAY_SIZE(cdns->response_buf); + } + + cmd_base = CDNS_MCP_CMD_BASE; + + for (i = 0; i < num_resp; i++) { + cdns->response_buf[i] = cdns_readl(cdns, cmd_base); + cmd_base += CDNS_MCP_CMD_WORD_LEN; + } +} + static enum sdw_command_response _cdns_xfer_msg(struct sdw_cdns *cdns, struct sdw_msg *msg, int cmd, int offset, int count, bool defer) @@ -552,6 +575,10 @@ _cdns_xfer_msg(struct sdw_cdns *cdns, struct sdw_msg *msg, int cmd, dev_err(cdns->dev, "IO transfer timed out, cmd %d device %d addr %x len %d\n", cmd, msg->dev_num, msg->addr, msg->len); msg->len = 0; + + /* Drain anything in the RX_FIFO */ + cdns_read_response(cdns); + return SDW_CMD_TIMEOUT; } @@ -720,29 +747,6 @@ EXPORT_SYMBOL(cdns_reset_page_addr); * IRQ handling */ -static void cdns_read_response(struct sdw_cdns *cdns) -{ - u32 num_resp, cmd_base; - int i; - - /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ - BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); - - num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); - num_resp &= CDNS_MCP_RX_FIFO_AVAIL; - if (num_resp > ARRAY_SIZE(cdns->response_buf)) { - dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); - num_resp = ARRAY_SIZE(cdns->response_buf); - } - - cmd_base = CDNS_MCP_CMD_BASE; - - for (i = 0; i < num_resp; i++) { - cdns->response_buf[i] = cdns_readl(cdns, cmd_base); - cmd_base += CDNS_MCP_CMD_WORD_LEN; - } -} - static int cdns_update_slave_status(struct sdw_cdns *cdns, u32 slave0, u32 slave1) { -- 2.39.2

1 0

[PATCH AUTOSEL 5.10 29/30] soundwire: cadence: Remove wasted space in response_buf
by Sasha Levin 03 Mar '23

03 Mar '23

From: Richard Fitzgerald <rf(a)opensource.cirrus.com> [ Upstream commit 827c32d0df4bbe0d1c47d79f6a5eabfe9ac75216 ] The response_buf was declared much larger (128 entries) than the number of responses that could ever be written into it. The Cadence IP is configurable up to a maximum of 32 entries, and the datasheet says that RX_FIFO_AVAIL can be 2 larger than this. So allow up to 34 responses. Also add checking in cdns_read_response() to prevent overflowing reponse_buf if RX_FIFO_AVAIL contains an unexpectedly large number. Signed-off-by: Richard Fitzgerald <rf(a)opensource.cirrus.com> Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart(a)linux.intel.com> Link: https://lore.kernel.org/r/20221202161812.4186897-3-rf@opensource.cirrus.com Signed-off-by: Vinod Koul <vkoul(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- drivers/soundwire/cadence_master.c | 7 +++++++ drivers/soundwire/cadence_master.h | 13 ++++++++++++- 2 files changed, 19 insertions(+), 1 deletion(-) diff --git a/drivers/soundwire/cadence_master.c b/drivers/soundwire/cadence_master.c index a3247692ddc07..292c4460eaaa3 100644 --- a/drivers/soundwire/cadence_master.c +++ b/drivers/soundwire/cadence_master.c @@ -725,8 +725,15 @@ static void cdns_read_response(struct sdw_cdns *cdns) u32 num_resp, cmd_base; int i; + /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ + BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); + num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); num_resp &= CDNS_MCP_RX_FIFO_AVAIL; + if (num_resp > ARRAY_SIZE(cdns->response_buf)) { + dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); + num_resp = ARRAY_SIZE(cdns->response_buf); + } cmd_base = CDNS_MCP_CMD_BASE; diff --git a/drivers/soundwire/cadence_master.h b/drivers/soundwire/cadence_master.h index 4d1aab5b5ec2d..e7f0108d417ca 100644 --- a/drivers/soundwire/cadence_master.h +++ b/drivers/soundwire/cadence_master.h @@ -8,6 +8,12 @@ #define SDW_CADENCE_GSYNC_KHZ 4 /* 4 kHz */ #define SDW_CADENCE_GSYNC_HZ (SDW_CADENCE_GSYNC_KHZ * 1000) +/* + * The Cadence IP supports up to 32 entries in the FIFO, though implementations + * can configure the IP to have a smaller FIFO. + */ +#define CDNS_MCP_IP_MAX_CMD_LEN 32 + /** * struct sdw_cdns_pdi: PDI (Physical Data Interface) instance * @@ -119,7 +125,12 @@ struct sdw_cdns { struct sdw_bus bus; unsigned int instance; - u32 response_buf[0x80]; + /* + * The datasheet says the RX FIFO AVAIL can be 2 entries more + * than the FIFO capacity, so allow for this. + */ + u32 response_buf[CDNS_MCP_IP_MAX_CMD_LEN + 2]; + struct completion tx_complete; struct sdw_defer *defer; -- 2.39.2

1 0

[PATCH AUTOSEL 5.15 50/50] soundwire: cadence: Drain the RX FIFO after an IO timeout
by Sasha Levin 03 Mar '23

03 Mar '23

From: Richard Fitzgerald <rf(a)opensource.cirrus.com> [ Upstream commit 0603a47bd3a8f439d7844b841eee1819353063e0 ] If wait_for_completion_timeout() times-out in _cdns_xfer_msg() it is possible that something could have been written to the RX FIFO. In this case, we should drain the RX FIFO so that anything in it doesn't carry over and mess up the next transfer. Obviously, if we got to this state something went wrong, and we don't really know the state of everything. The cleanup in this situation cannot be bullet-proof but we should attempt to avoid breaking future transaction, if only to reduce the amount of error noise when debugging the failure from a kernel log. Note that this patch only implements the draining for blocking (non-deferred) transfers. The deferred API doesn't have any proper handling of error conditions and would need some re-design before implementing cleanup. That is a task for a separate patch... Signed-off-by: Richard Fitzgerald <rf(a)opensource.cirrus.com> Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart(a)linux.intel.com> Link: https://lore.kernel.org/r/20221202161812.4186897-4-rf@opensource.cirrus.com Signed-off-by: Vinod Koul <vkoul(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- drivers/soundwire/cadence_master.c | 50 ++++++++++++++++-------------- 1 file changed, 27 insertions(+), 23 deletions(-) diff --git a/drivers/soundwire/cadence_master.c b/drivers/soundwire/cadence_master.c index b229e1e0d7d97..842e08c56796b 100644 --- a/drivers/soundwire/cadence_master.c +++ b/drivers/soundwire/cadence_master.c @@ -555,6 +555,29 @@ cdns_fill_msg_resp(struct sdw_cdns *cdns, return SDW_CMD_OK; } +static void cdns_read_response(struct sdw_cdns *cdns) +{ + u32 num_resp, cmd_base; + int i; + + /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ + BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); + + num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); + num_resp &= CDNS_MCP_RX_FIFO_AVAIL; + if (num_resp > ARRAY_SIZE(cdns->response_buf)) { + dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); + num_resp = ARRAY_SIZE(cdns->response_buf); + } + + cmd_base = CDNS_MCP_CMD_BASE; + + for (i = 0; i < num_resp; i++) { + cdns->response_buf[i] = cdns_readl(cdns, cmd_base); + cmd_base += CDNS_MCP_CMD_WORD_LEN; + } +} + static enum sdw_command_response _cdns_xfer_msg(struct sdw_cdns *cdns, struct sdw_msg *msg, int cmd, int offset, int count, bool defer) @@ -596,6 +619,10 @@ _cdns_xfer_msg(struct sdw_cdns *cdns, struct sdw_msg *msg, int cmd, dev_err(cdns->dev, "IO transfer timed out, cmd %d device %d addr %x len %d\n", cmd, msg->dev_num, msg->addr, msg->len); msg->len = 0; + + /* Drain anything in the RX_FIFO */ + cdns_read_response(cdns); + return SDW_CMD_TIMEOUT; } @@ -764,29 +791,6 @@ EXPORT_SYMBOL(cdns_reset_page_addr); * IRQ handling */ -static void cdns_read_response(struct sdw_cdns *cdns) -{ - u32 num_resp, cmd_base; - int i; - - /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ - BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); - - num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); - num_resp &= CDNS_MCP_RX_FIFO_AVAIL; - if (num_resp > ARRAY_SIZE(cdns->response_buf)) { - dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); - num_resp = ARRAY_SIZE(cdns->response_buf); - } - - cmd_base = CDNS_MCP_CMD_BASE; - - for (i = 0; i < num_resp; i++) { - cdns->response_buf[i] = cdns_readl(cdns, cmd_base); - cmd_base += CDNS_MCP_CMD_WORD_LEN; - } -} - static int cdns_update_slave_status(struct sdw_cdns *cdns, u64 slave_intstat) { -- 2.39.2

1 0

[PATCH AUTOSEL 5.15 49/50] soundwire: cadence: Remove wasted space in response_buf
by Sasha Levin 03 Mar '23

03 Mar '23

From: Richard Fitzgerald <rf(a)opensource.cirrus.com> [ Upstream commit 827c32d0df4bbe0d1c47d79f6a5eabfe9ac75216 ] The response_buf was declared much larger (128 entries) than the number of responses that could ever be written into it. The Cadence IP is configurable up to a maximum of 32 entries, and the datasheet says that RX_FIFO_AVAIL can be 2 larger than this. So allow up to 34 responses. Also add checking in cdns_read_response() to prevent overflowing reponse_buf if RX_FIFO_AVAIL contains an unexpectedly large number. Signed-off-by: Richard Fitzgerald <rf(a)opensource.cirrus.com> Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart(a)linux.intel.com> Link: https://lore.kernel.org/r/20221202161812.4186897-3-rf@opensource.cirrus.com Signed-off-by: Vinod Koul <vkoul(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- drivers/soundwire/cadence_master.c | 7 +++++++ drivers/soundwire/cadence_master.h | 13 ++++++++++++- 2 files changed, 19 insertions(+), 1 deletion(-) diff --git a/drivers/soundwire/cadence_master.c b/drivers/soundwire/cadence_master.c index 18d2f9b3e2010..b229e1e0d7d97 100644 --- a/drivers/soundwire/cadence_master.c +++ b/drivers/soundwire/cadence_master.c @@ -769,8 +769,15 @@ static void cdns_read_response(struct sdw_cdns *cdns) u32 num_resp, cmd_base; int i; + /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ + BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); + num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); num_resp &= CDNS_MCP_RX_FIFO_AVAIL; + if (num_resp > ARRAY_SIZE(cdns->response_buf)) { + dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); + num_resp = ARRAY_SIZE(cdns->response_buf); + } cmd_base = CDNS_MCP_CMD_BASE; diff --git a/drivers/soundwire/cadence_master.h b/drivers/soundwire/cadence_master.h index e587aede63bf0..e437a604429fa 100644 --- a/drivers/soundwire/cadence_master.h +++ b/drivers/soundwire/cadence_master.h @@ -8,6 +8,12 @@ #define SDW_CADENCE_GSYNC_KHZ 4 /* 4 kHz */ #define SDW_CADENCE_GSYNC_HZ (SDW_CADENCE_GSYNC_KHZ * 1000) +/* + * The Cadence IP supports up to 32 entries in the FIFO, though implementations + * can configure the IP to have a smaller FIFO. + */ +#define CDNS_MCP_IP_MAX_CMD_LEN 32 + /** * struct sdw_cdns_pdi: PDI (Physical Data Interface) instance * @@ -119,7 +125,12 @@ struct sdw_cdns { struct sdw_bus bus; unsigned int instance; - u32 response_buf[0x80]; + /* + * The datasheet says the RX FIFO AVAIL can be 2 entries more + * than the FIFO capacity, so allow for this. + */ + u32 response_buf[CDNS_MCP_IP_MAX_CMD_LEN + 2]; + struct completion tx_complete; struct sdw_defer *defer; -- 2.39.2

1 0

[PATCH AUTOSEL 5.15 19/50] soundwire: bus_type: Avoid lockdep assert in sdw_drv_probe()
by Sasha Levin 03 Mar '23

03 Mar '23

From: Richard Fitzgerald <rf(a)opensource.cirrus.com> [ Upstream commit 3dca1f89ae3455963d7b53245ecf298ea9bae857 ] Don't hold sdw_dev_lock while calling the peripheral driver probe() and remove() callbacks. Holding sdw_dev_lock around the probe() and remove() calls causes a theoretical mutex inversion which lockdep will assert on. During probe() the sdw_dev_lock mutex is taken first and then ASoC/ALSA locks are taken by the probe() implementation. During normal operation ASoC can take its locks and then trigger a runtime resume of the component. The SoundWire resume will then take sdw_dev_lock. This is the reverse order compared to probe(). It's not necessary to hold sdw_dev_lock when calling the probe() and remove(), it is only used to prevent the bus core calling the driver callbacks if there isn't a driver or the driver is removing. All calls to the driver callbacks are guarded by the 'probed' flag. So if sdw_dev_lock is held while setting and clearing the 'probed' flag this is sufficient to guarantee the safety of callback functions. Removing the mutex from around the call to probe() means that it is now possible for a bus event (PING response) to be handled in parallel with the probe(). But sdw_bus_probe() already has handling for this by calling the device update_status() after the probe() has completed. Example lockdep assert: [ 46.098514] ====================================================== [ 46.104736] WARNING: possible circular locking dependency detected [ 46.110961] 6.1.0-rc4-jamerson #1 Tainted: G E [ 46.116842] ------------------------------------------------------ [ 46.123063] mpg123/1130 is trying to acquire lock: [ 46.127883] ffff8b445031fb80 (&slave->sdw_dev_lock){+.+.}-{3:3}, at: sdw_update_slave_status+0x26/0x70 [ 46.137225] but task is already holding lock: [ 46.143074] ffffffffc1455310 (&card->pcm_mutex){+.+.}-{3:3}, at: dpcm_fe_dai_open+0x49/0x830 [ 46.151536] which lock already depends on the new lock.[ 46.159732] the existing dependency chain (in reverse order) is: [ 46.167231] -> #4 (&card->pcm_mutex){+.+.}-{3:3}: [ 46.173428] __mutex_lock+0x94/0x920 [ 46.177542] snd_soc_dpcm_runtime_update+0x2e/0x100 [ 46.182958] snd_soc_dapm_put_enum_double+0x1c2/0x200 [ 46.188548] snd_ctl_elem_write+0x10c/0x1d0 [ 46.193268] snd_ctl_ioctl+0x126/0x850 [ 46.197556] __x64_sys_ioctl+0x87/0xc0 [ 46.201845] do_syscall_64+0x38/0x90 [ 46.205959] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.211553] -> #3 (&card->controls_rwsem){++++}-{3:3}: [ 46.218188] down_write+0x2b/0xd0 [ 46.222038] snd_ctl_add_replace+0x39/0xb0 [ 46.226672] snd_soc_add_controls+0x53/0x80 [ 46.231393] soc_probe_component+0x1e4/0x2a0 [ 46.236202] snd_soc_bind_card+0x51a/0xc80 [ 46.240836] devm_snd_soc_register_card+0x43/0x90 [ 46.246079] mc_probe+0x982/0xfe0 [snd_soc_sof_sdw] [ 46.251500] platform_probe+0x3c/0xa0 [ 46.255700] really_probe+0xde/0x390 [ 46.259814] __driver_probe_device+0x78/0x180 [ 46.264710] driver_probe_device+0x1e/0x90 [ 46.269347] __driver_attach+0x9f/0x1f0 [ 46.273721] bus_for_each_dev+0x78/0xc0 [ 46.278098] bus_add_driver+0x1ac/0x200 [ 46.282473] driver_register+0x8f/0xf0 [ 46.286759] do_one_initcall+0x58/0x310 [ 46.291136] do_init_module+0x4c/0x1f0 [ 46.295422] __do_sys_finit_module+0xb4/0x130 [ 46.300321] do_syscall_64+0x38/0x90 [ 46.304434] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.310027] -> #2 (&card->mutex){+.+.}-{3:3}: [ 46.315883] __mutex_lock+0x94/0x920 [ 46.320000] snd_soc_bind_card+0x3e/0xc80 [ 46.324551] devm_snd_soc_register_card+0x43/0x90 [ 46.329798] mc_probe+0x982/0xfe0 [snd_soc_sof_sdw] [ 46.335219] platform_probe+0x3c/0xa0 [ 46.339420] really_probe+0xde/0x390 [ 46.343532] __driver_probe_device+0x78/0x180 [ 46.348430] driver_probe_device+0x1e/0x90 [ 46.353065] __driver_attach+0x9f/0x1f0 [ 46.357437] bus_for_each_dev+0x78/0xc0 [ 46.361812] bus_add_driver+0x1ac/0x200 [ 46.366716] driver_register+0x8f/0xf0 [ 46.371528] do_one_initcall+0x58/0x310 [ 46.376424] do_init_module+0x4c/0x1f0 [ 46.381239] __do_sys_finit_module+0xb4/0x130 [ 46.386665] do_syscall_64+0x38/0x90 [ 46.391299] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.397416] -> #1 (client_mutex){+.+.}-{3:3}: [ 46.404307] __mutex_lock+0x94/0x920 [ 46.408941] snd_soc_add_component+0x24/0x2c0 [ 46.414345] devm_snd_soc_register_component+0x54/0xa0 [ 46.420522] cs35l56_common_probe+0x280/0x370 [snd_soc_cs35l56] [ 46.427487] cs35l56_sdw_probe+0xf4/0x170 [snd_soc_cs35l56_sdw] [ 46.434442] sdw_drv_probe+0x80/0x1a0 [ 46.439136] really_probe+0xde/0x390 [ 46.443738] __driver_probe_device+0x78/0x180 [ 46.449120] driver_probe_device+0x1e/0x90 [ 46.454247] __driver_attach+0x9f/0x1f0 [ 46.459106] bus_for_each_dev+0x78/0xc0 [ 46.463971] bus_add_driver+0x1ac/0x200 [ 46.468825] driver_register+0x8f/0xf0 [ 46.473592] do_one_initcall+0x58/0x310 [ 46.478441] do_init_module+0x4c/0x1f0 [ 46.483202] __do_sys_finit_module+0xb4/0x130 [ 46.488572] do_syscall_64+0x38/0x90 [ 46.493158] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.499229] -> #0 (&slave->sdw_dev_lock){+.+.}-{3:3}: [ 46.506737] __lock_acquire+0x1121/0x1df0 [ 46.511765] lock_acquire+0xd5/0x300 [ 46.516360] __mutex_lock+0x94/0x920 [ 46.520949] sdw_update_slave_status+0x26/0x70 [ 46.526409] sdw_clear_slave_status+0xd8/0xe0 [ 46.531783] intel_resume_runtime+0x139/0x2a0 [ 46.537155] __rpm_callback+0x41/0x120 [ 46.541919] rpm_callback+0x5d/0x70 [ 46.546422] rpm_resume+0x531/0x7e0 [ 46.550920] __pm_runtime_resume+0x4a/0x80 [ 46.556024] snd_soc_pcm_component_pm_runtime_get+0x2f/0xc0 [ 46.562611] __soc_pcm_open+0x62/0x520 [ 46.567375] dpcm_be_dai_startup+0x116/0x210 [ 46.572661] dpcm_fe_dai_open+0xf7/0x830 [ 46.577597] snd_pcm_open_substream+0x54a/0x8b0 [ 46.583145] snd_pcm_open.part.0+0xdc/0x200 [ 46.588341] snd_pcm_playback_open+0x51/0x80 [ 46.593625] chrdev_open+0xc0/0x250 [ 46.598129] do_dentry_open+0x15f/0x430 [ 46.602981] path_openat+0x75e/0xa80 [ 46.607575] do_filp_open+0xb2/0x160 [ 46.612162] do_sys_openat2+0x9a/0x160 [ 46.616922] __x64_sys_openat+0x53/0xa0 [ 46.621767] do_syscall_64+0x38/0x90 [ 46.626352] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.632414] other info that might help us debug this:[ 46.641862] Chain exists of: &slave->sdw_dev_lock --> &card->controls_rwsem --> &card->pcm_mutex[ 46.655145] Possible unsafe locking scenario:[ 46.662048] CPU0 CPU1 [ 46.667080] ---- ---- [ 46.672108] lock(&card->pcm_mutex); [ 46.676267] lock(&card->controls_rwsem); [ 46.683382] lock(&card->pcm_mutex); [ 46.690063] lock(&slave->sdw_dev_lock); [ 46.694574] *** DEADLOCK ***[ 46.701942] 2 locks held by mpg123/1130: [ 46.706356] #0: ffff8b4457b22b90 (&pcm->open_mutex){+.+.}-{3:3}, at: snd_pcm_open.part.0+0xc9/0x200 [ 46.715999] #1: ffffffffc1455310 (&card->pcm_mutex){+.+.}-{3:3}, at: dpcm_fe_dai_open+0x49/0x830 [ 46.725390] stack backtrace: [ 46.730752] CPU: 0 PID: 1130 Comm: mpg123 Tainted: G E 6.1.0-rc4-jamerson #1 [ 46.739703] Hardware name: AAEON UP-WHL01/UP-WHL01, BIOS UPW1AM19 11/10/2020 [ 46.747270] Call Trace: [ 46.750239] <TASK> [ 46.752857] dump_stack_lvl+0x56/0x73 [ 46.757045] check_noncircular+0x102/0x120 [ 46.761664] __lock_acquire+0x1121/0x1df0 [ 46.766197] lock_acquire+0xd5/0x300 [ 46.770292] ? sdw_update_slave_status+0x26/0x70 [ 46.775432] ? lock_is_held_type+0xe2/0x140 [ 46.780143] __mutex_lock+0x94/0x920 [ 46.784241] ? sdw_update_slave_status+0x26/0x70 [ 46.789387] ? find_held_lock+0x2b/0x80 [ 46.793750] ? sdw_update_slave_status+0x26/0x70 [ 46.798894] ? lock_release+0x147/0x2f0 [ 46.803262] ? lockdep_init_map_type+0x47/0x250 [ 46.808315] ? sdw_update_slave_status+0x26/0x70 [ 46.813456] sdw_update_slave_status+0x26/0x70 [ 46.818422] sdw_clear_slave_status+0xd8/0xe0 [ 46.823302] ? pm_generic_runtime_suspend+0x30/0x30 [ 46.828706] intel_resume_runtime+0x139/0x2a0 [ 46.833583] ? _raw_spin_unlock_irq+0x24/0x50 [ 46.838462] ? pm_generic_runtime_suspend+0x30/0x30 [ 46.843866] __rpm_callback+0x41/0x120 [ 46.848142] ? pm_generic_runtime_suspend+0x30/0x30 [ 46.853550] rpm_callback+0x5d/0x70 [ 46.857568] rpm_resume+0x531/0x7e0 [ 46.861578] ? _raw_spin_lock_irqsave+0x62/0x70 [ 46.866634] __pm_runtime_resume+0x4a/0x80 [ 46.871258] snd_soc_pcm_component_pm_runtime_get+0x2f/0xc0 [ 46.877358] __soc_pcm_open+0x62/0x520 [ 46.881634] ? dpcm_add_paths.isra.0+0x35d/0x4c0 [ 46.886784] dpcm_be_dai_startup+0x116/0x210 [ 46.891592] dpcm_fe_dai_open+0xf7/0x830 [ 46.896046] ? debug_mutex_init+0x33/0x50 [ 46.900591] snd_pcm_open_substream+0x54a/0x8b0 [ 46.905658] snd_pcm_open.part.0+0xdc/0x200 [ 46.910376] ? wake_up_q+0x90/0x90 [ 46.914312] snd_pcm_playback_open+0x51/0x80 [ 46.919118] chrdev_open+0xc0/0x250 [ 46.923147] ? cdev_device_add+0x90/0x90 [ 46.927608] do_dentry_open+0x15f/0x430 [ 46.931976] path_openat+0x75e/0xa80 [ 46.936086] do_filp_open+0xb2/0x160 [ 46.940194] ? lock_release+0x147/0x2f0 [ 46.944563] ? _raw_spin_unlock+0x29/0x50 [ 46.949101] do_sys_openat2+0x9a/0x160 [ 46.953377] __x64_sys_openat+0x53/0xa0 [ 46.957733] do_syscall_64+0x38/0x90 [ 46.961829] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.967402] RIP: 0033:0x7fa6397ccd3b [ 46.971506] Code: 25 00 00 41 00 3d 00 00 41 00 74 4b 64 8b 04 25 18 00 00 00 85 c0 75 67 44 89 e2 48 89 ee bf 9c ff ff ff b8 01 01 00 00 0f 05 <48> 3d 00 f0 ff ff 0f 87 91 00 00 00 48 8b 4c 24 28 64 48 33 0c 25 [ 46.991413] RSP: 002b:00007fff838e8990 EFLAGS: 00000246 ORIG_RAX: 0000000000000101 [ 46.999580] RAX: ffffffffffffffda RBX: 0000000000080802 RCX: 00007fa6397ccd3b [ 47.007311] RDX: 0000000000080802 RSI: 00007fff838e8b50 RDI: 00000000ffffff9c [ 47.015047] RBP: 00007fff838e8b50 R08: 0000000000000000 R09: 0000000000000011 [ 47.022787] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000080802 [ 47.030539] R13: 0000000000000004 R14: 0000000000000000 R15: 00007fff838e8b50 [ 47.038289] </TASK> Signed-off-by: Richard Fitzgerald <rf(a)opensource.cirrus.com> Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart(a)linux.intel.com> Link: https://lore.kernel.org/r/20230123172520.339367-1-rf@opensource.cirrus.com Signed-off-by: Vinod Koul <vkoul(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- drivers/soundwire/bus_type.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) diff --git a/drivers/soundwire/bus_type.c b/drivers/soundwire/bus_type.c index 04b3529f89293..963498db0fd22 100644 --- a/drivers/soundwire/bus_type.c +++ b/drivers/soundwire/bus_type.c @@ -105,20 +105,19 @@ static int sdw_drv_probe(struct device *dev) if (ret) return ret; - mutex_lock(&slave->sdw_dev_lock); - ret = drv->probe(slave, id); if (ret) { name = drv->name; if (!name) name = drv->driver.name; - mutex_unlock(&slave->sdw_dev_lock); dev_err(dev, "Probe of %s failed: %d\n", name, ret); dev_pm_domain_detach(dev, false); return ret; } + mutex_lock(&slave->sdw_dev_lock); + /* device is probed so let's read the properties now */ if (drv->ops && drv->ops->read_prop) drv->ops->read_prop(slave); @@ -167,14 +166,12 @@ static int sdw_drv_remove(struct device *dev) int ret = 0; mutex_lock(&slave->sdw_dev_lock); - slave->probed = false; + mutex_unlock(&slave->sdw_dev_lock); if (drv->remove) ret = drv->remove(slave); - mutex_unlock(&slave->sdw_dev_lock); - dev_pm_domain_detach(dev, false); return ret; -- 2.39.2

1 0

[PATCH AUTOSEL 6.1 60/60] soundwire: cadence: Drain the RX FIFO after an IO timeout
by Sasha Levin 03 Mar '23

03 Mar '23

From: Richard Fitzgerald <rf(a)opensource.cirrus.com> [ Upstream commit 0603a47bd3a8f439d7844b841eee1819353063e0 ] If wait_for_completion_timeout() times-out in _cdns_xfer_msg() it is possible that something could have been written to the RX FIFO. In this case, we should drain the RX FIFO so that anything in it doesn't carry over and mess up the next transfer. Obviously, if we got to this state something went wrong, and we don't really know the state of everything. The cleanup in this situation cannot be bullet-proof but we should attempt to avoid breaking future transaction, if only to reduce the amount of error noise when debugging the failure from a kernel log. Note that this patch only implements the draining for blocking (non-deferred) transfers. The deferred API doesn't have any proper handling of error conditions and would need some re-design before implementing cleanup. That is a task for a separate patch... Signed-off-by: Richard Fitzgerald <rf(a)opensource.cirrus.com> Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart(a)linux.intel.com> Link: https://lore.kernel.org/r/20221202161812.4186897-4-rf@opensource.cirrus.com Signed-off-by: Vinod Koul <vkoul(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- drivers/soundwire/cadence_master.c | 50 ++++++++++++++++-------------- 1 file changed, 27 insertions(+), 23 deletions(-) diff --git a/drivers/soundwire/cadence_master.c b/drivers/soundwire/cadence_master.c index 4e92a6b1ad626..cb59dc634658c 100644 --- a/drivers/soundwire/cadence_master.c +++ b/drivers/soundwire/cadence_master.c @@ -554,6 +554,29 @@ cdns_fill_msg_resp(struct sdw_cdns *cdns, return SDW_CMD_OK; } +static void cdns_read_response(struct sdw_cdns *cdns) +{ + u32 num_resp, cmd_base; + int i; + + /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ + BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); + + num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); + num_resp &= CDNS_MCP_RX_FIFO_AVAIL; + if (num_resp > ARRAY_SIZE(cdns->response_buf)) { + dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); + num_resp = ARRAY_SIZE(cdns->response_buf); + } + + cmd_base = CDNS_MCP_CMD_BASE; + + for (i = 0; i < num_resp; i++) { + cdns->response_buf[i] = cdns_readl(cdns, cmd_base); + cmd_base += CDNS_MCP_CMD_WORD_LEN; + } +} + static enum sdw_command_response _cdns_xfer_msg(struct sdw_cdns *cdns, struct sdw_msg *msg, int cmd, int offset, int count, bool defer) @@ -595,6 +618,10 @@ _cdns_xfer_msg(struct sdw_cdns *cdns, struct sdw_msg *msg, int cmd, dev_err(cdns->dev, "IO transfer timed out, cmd %d device %d addr %x len %d\n", cmd, msg->dev_num, msg->addr, msg->len); msg->len = 0; + + /* Drain anything in the RX_FIFO */ + cdns_read_response(cdns); + return SDW_CMD_TIMEOUT; } @@ -768,29 +795,6 @@ EXPORT_SYMBOL(cdns_read_ping_status); * IRQ handling */ -static void cdns_read_response(struct sdw_cdns *cdns) -{ - u32 num_resp, cmd_base; - int i; - - /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ - BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); - - num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); - num_resp &= CDNS_MCP_RX_FIFO_AVAIL; - if (num_resp > ARRAY_SIZE(cdns->response_buf)) { - dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); - num_resp = ARRAY_SIZE(cdns->response_buf); - } - - cmd_base = CDNS_MCP_CMD_BASE; - - for (i = 0; i < num_resp; i++) { - cdns->response_buf[i] = cdns_readl(cdns, cmd_base); - cmd_base += CDNS_MCP_CMD_WORD_LEN; - } -} - static int cdns_update_slave_status(struct sdw_cdns *cdns, u64 slave_intstat) { -- 2.39.2

1 0

[PATCH AUTOSEL 6.1 59/60] soundwire: cadence: Remove wasted space in response_buf
by Sasha Levin 03 Mar '23

03 Mar '23

From: Richard Fitzgerald <rf(a)opensource.cirrus.com> [ Upstream commit 827c32d0df4bbe0d1c47d79f6a5eabfe9ac75216 ] The response_buf was declared much larger (128 entries) than the number of responses that could ever be written into it. The Cadence IP is configurable up to a maximum of 32 entries, and the datasheet says that RX_FIFO_AVAIL can be 2 larger than this. So allow up to 34 responses. Also add checking in cdns_read_response() to prevent overflowing reponse_buf if RX_FIFO_AVAIL contains an unexpectedly large number. Signed-off-by: Richard Fitzgerald <rf(a)opensource.cirrus.com> Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart(a)linux.intel.com> Link: https://lore.kernel.org/r/20221202161812.4186897-3-rf@opensource.cirrus.com Signed-off-by: Vinod Koul <vkoul(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- drivers/soundwire/cadence_master.c | 7 +++++++ drivers/soundwire/cadence_master.h | 13 ++++++++++++- 2 files changed, 19 insertions(+), 1 deletion(-) diff --git a/drivers/soundwire/cadence_master.c b/drivers/soundwire/cadence_master.c index 93929f19d0831..4e92a6b1ad626 100644 --- a/drivers/soundwire/cadence_master.c +++ b/drivers/soundwire/cadence_master.c @@ -773,8 +773,15 @@ static void cdns_read_response(struct sdw_cdns *cdns) u32 num_resp, cmd_base; int i; + /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ + BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); + num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); num_resp &= CDNS_MCP_RX_FIFO_AVAIL; + if (num_resp > ARRAY_SIZE(cdns->response_buf)) { + dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); + num_resp = ARRAY_SIZE(cdns->response_buf); + } cmd_base = CDNS_MCP_CMD_BASE; diff --git a/drivers/soundwire/cadence_master.h b/drivers/soundwire/cadence_master.h index ca9e805bab88f..51e6ecc027cbc 100644 --- a/drivers/soundwire/cadence_master.h +++ b/drivers/soundwire/cadence_master.h @@ -8,6 +8,12 @@ #define SDW_CADENCE_GSYNC_KHZ 4 /* 4 kHz */ #define SDW_CADENCE_GSYNC_HZ (SDW_CADENCE_GSYNC_KHZ * 1000) +/* + * The Cadence IP supports up to 32 entries in the FIFO, though implementations + * can configure the IP to have a smaller FIFO. + */ +#define CDNS_MCP_IP_MAX_CMD_LEN 32 + /** * struct sdw_cdns_pdi: PDI (Physical Data Interface) instance * @@ -114,7 +120,12 @@ struct sdw_cdns { struct sdw_bus bus; unsigned int instance; - u32 response_buf[0x80]; + /* + * The datasheet says the RX FIFO AVAIL can be 2 entries more + * than the FIFO capacity, so allow for this. + */ + u32 response_buf[CDNS_MCP_IP_MAX_CMD_LEN + 2]; + struct completion tx_complete; struct sdw_defer *defer; -- 2.39.2

1 0

[PATCH AUTOSEL 6.1 23/60] soundwire: bus_type: Avoid lockdep assert in sdw_drv_probe()
by Sasha Levin 03 Mar '23

03 Mar '23

From: Richard Fitzgerald <rf(a)opensource.cirrus.com> [ Upstream commit 3dca1f89ae3455963d7b53245ecf298ea9bae857 ] Don't hold sdw_dev_lock while calling the peripheral driver probe() and remove() callbacks. Holding sdw_dev_lock around the probe() and remove() calls causes a theoretical mutex inversion which lockdep will assert on. During probe() the sdw_dev_lock mutex is taken first and then ASoC/ALSA locks are taken by the probe() implementation. During normal operation ASoC can take its locks and then trigger a runtime resume of the component. The SoundWire resume will then take sdw_dev_lock. This is the reverse order compared to probe(). It's not necessary to hold sdw_dev_lock when calling the probe() and remove(), it is only used to prevent the bus core calling the driver callbacks if there isn't a driver or the driver is removing. All calls to the driver callbacks are guarded by the 'probed' flag. So if sdw_dev_lock is held while setting and clearing the 'probed' flag this is sufficient to guarantee the safety of callback functions. Removing the mutex from around the call to probe() means that it is now possible for a bus event (PING response) to be handled in parallel with the probe(). But sdw_bus_probe() already has handling for this by calling the device update_status() after the probe() has completed. Example lockdep assert: [ 46.098514] ====================================================== [ 46.104736] WARNING: possible circular locking dependency detected [ 46.110961] 6.1.0-rc4-jamerson #1 Tainted: G E [ 46.116842] ------------------------------------------------------ [ 46.123063] mpg123/1130 is trying to acquire lock: [ 46.127883] ffff8b445031fb80 (&slave->sdw_dev_lock){+.+.}-{3:3}, at: sdw_update_slave_status+0x26/0x70 [ 46.137225] but task is already holding lock: [ 46.143074] ffffffffc1455310 (&card->pcm_mutex){+.+.}-{3:3}, at: dpcm_fe_dai_open+0x49/0x830 [ 46.151536] which lock already depends on the new lock.[ 46.159732] the existing dependency chain (in reverse order) is: [ 46.167231] -> #4 (&card->pcm_mutex){+.+.}-{3:3}: [ 46.173428] __mutex_lock+0x94/0x920 [ 46.177542] snd_soc_dpcm_runtime_update+0x2e/0x100 [ 46.182958] snd_soc_dapm_put_enum_double+0x1c2/0x200 [ 46.188548] snd_ctl_elem_write+0x10c/0x1d0 [ 46.193268] snd_ctl_ioctl+0x126/0x850 [ 46.197556] __x64_sys_ioctl+0x87/0xc0 [ 46.201845] do_syscall_64+0x38/0x90 [ 46.205959] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.211553] -> #3 (&card->controls_rwsem){++++}-{3:3}: [ 46.218188] down_write+0x2b/0xd0 [ 46.222038] snd_ctl_add_replace+0x39/0xb0 [ 46.226672] snd_soc_add_controls+0x53/0x80 [ 46.231393] soc_probe_component+0x1e4/0x2a0 [ 46.236202] snd_soc_bind_card+0x51a/0xc80 [ 46.240836] devm_snd_soc_register_card+0x43/0x90 [ 46.246079] mc_probe+0x982/0xfe0 [snd_soc_sof_sdw] [ 46.251500] platform_probe+0x3c/0xa0 [ 46.255700] really_probe+0xde/0x390 [ 46.259814] __driver_probe_device+0x78/0x180 [ 46.264710] driver_probe_device+0x1e/0x90 [ 46.269347] __driver_attach+0x9f/0x1f0 [ 46.273721] bus_for_each_dev+0x78/0xc0 [ 46.278098] bus_add_driver+0x1ac/0x200 [ 46.282473] driver_register+0x8f/0xf0 [ 46.286759] do_one_initcall+0x58/0x310 [ 46.291136] do_init_module+0x4c/0x1f0 [ 46.295422] __do_sys_finit_module+0xb4/0x130 [ 46.300321] do_syscall_64+0x38/0x90 [ 46.304434] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.310027] -> #2 (&card->mutex){+.+.}-{3:3}: [ 46.315883] __mutex_lock+0x94/0x920 [ 46.320000] snd_soc_bind_card+0x3e/0xc80 [ 46.324551] devm_snd_soc_register_card+0x43/0x90 [ 46.329798] mc_probe+0x982/0xfe0 [snd_soc_sof_sdw] [ 46.335219] platform_probe+0x3c/0xa0 [ 46.339420] really_probe+0xde/0x390 [ 46.343532] __driver_probe_device+0x78/0x180 [ 46.348430] driver_probe_device+0x1e/0x90 [ 46.353065] __driver_attach+0x9f/0x1f0 [ 46.357437] bus_for_each_dev+0x78/0xc0 [ 46.361812] bus_add_driver+0x1ac/0x200 [ 46.366716] driver_register+0x8f/0xf0 [ 46.371528] do_one_initcall+0x58/0x310 [ 46.376424] do_init_module+0x4c/0x1f0 [ 46.381239] __do_sys_finit_module+0xb4/0x130 [ 46.386665] do_syscall_64+0x38/0x90 [ 46.391299] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.397416] -> #1 (client_mutex){+.+.}-{3:3}: [ 46.404307] __mutex_lock+0x94/0x920 [ 46.408941] snd_soc_add_component+0x24/0x2c0 [ 46.414345] devm_snd_soc_register_component+0x54/0xa0 [ 46.420522] cs35l56_common_probe+0x280/0x370 [snd_soc_cs35l56] [ 46.427487] cs35l56_sdw_probe+0xf4/0x170 [snd_soc_cs35l56_sdw] [ 46.434442] sdw_drv_probe+0x80/0x1a0 [ 46.439136] really_probe+0xde/0x390 [ 46.443738] __driver_probe_device+0x78/0x180 [ 46.449120] driver_probe_device+0x1e/0x90 [ 46.454247] __driver_attach+0x9f/0x1f0 [ 46.459106] bus_for_each_dev+0x78/0xc0 [ 46.463971] bus_add_driver+0x1ac/0x200 [ 46.468825] driver_register+0x8f/0xf0 [ 46.473592] do_one_initcall+0x58/0x310 [ 46.478441] do_init_module+0x4c/0x1f0 [ 46.483202] __do_sys_finit_module+0xb4/0x130 [ 46.488572] do_syscall_64+0x38/0x90 [ 46.493158] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.499229] -> #0 (&slave->sdw_dev_lock){+.+.}-{3:3}: [ 46.506737] __lock_acquire+0x1121/0x1df0 [ 46.511765] lock_acquire+0xd5/0x300 [ 46.516360] __mutex_lock+0x94/0x920 [ 46.520949] sdw_update_slave_status+0x26/0x70 [ 46.526409] sdw_clear_slave_status+0xd8/0xe0 [ 46.531783] intel_resume_runtime+0x139/0x2a0 [ 46.537155] __rpm_callback+0x41/0x120 [ 46.541919] rpm_callback+0x5d/0x70 [ 46.546422] rpm_resume+0x531/0x7e0 [ 46.550920] __pm_runtime_resume+0x4a/0x80 [ 46.556024] snd_soc_pcm_component_pm_runtime_get+0x2f/0xc0 [ 46.562611] __soc_pcm_open+0x62/0x520 [ 46.567375] dpcm_be_dai_startup+0x116/0x210 [ 46.572661] dpcm_fe_dai_open+0xf7/0x830 [ 46.577597] snd_pcm_open_substream+0x54a/0x8b0 [ 46.583145] snd_pcm_open.part.0+0xdc/0x200 [ 46.588341] snd_pcm_playback_open+0x51/0x80 [ 46.593625] chrdev_open+0xc0/0x250 [ 46.598129] do_dentry_open+0x15f/0x430 [ 46.602981] path_openat+0x75e/0xa80 [ 46.607575] do_filp_open+0xb2/0x160 [ 46.612162] do_sys_openat2+0x9a/0x160 [ 46.616922] __x64_sys_openat+0x53/0xa0 [ 46.621767] do_syscall_64+0x38/0x90 [ 46.626352] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.632414] other info that might help us debug this:[ 46.641862] Chain exists of: &slave->sdw_dev_lock --> &card->controls_rwsem --> &card->pcm_mutex[ 46.655145] Possible unsafe locking scenario:[ 46.662048] CPU0 CPU1 [ 46.667080] ---- ---- [ 46.672108] lock(&card->pcm_mutex); [ 46.676267] lock(&card->controls_rwsem); [ 46.683382] lock(&card->pcm_mutex); [ 46.690063] lock(&slave->sdw_dev_lock); [ 46.694574] *** DEADLOCK ***[ 46.701942] 2 locks held by mpg123/1130: [ 46.706356] #0: ffff8b4457b22b90 (&pcm->open_mutex){+.+.}-{3:3}, at: snd_pcm_open.part.0+0xc9/0x200 [ 46.715999] #1: ffffffffc1455310 (&card->pcm_mutex){+.+.}-{3:3}, at: dpcm_fe_dai_open+0x49/0x830 [ 46.725390] stack backtrace: [ 46.730752] CPU: 0 PID: 1130 Comm: mpg123 Tainted: G E 6.1.0-rc4-jamerson #1 [ 46.739703] Hardware name: AAEON UP-WHL01/UP-WHL01, BIOS UPW1AM19 11/10/2020 [ 46.747270] Call Trace: [ 46.750239] <TASK> [ 46.752857] dump_stack_lvl+0x56/0x73 [ 46.757045] check_noncircular+0x102/0x120 [ 46.761664] __lock_acquire+0x1121/0x1df0 [ 46.766197] lock_acquire+0xd5/0x300 [ 46.770292] ? sdw_update_slave_status+0x26/0x70 [ 46.775432] ? lock_is_held_type+0xe2/0x140 [ 46.780143] __mutex_lock+0x94/0x920 [ 46.784241] ? sdw_update_slave_status+0x26/0x70 [ 46.789387] ? find_held_lock+0x2b/0x80 [ 46.793750] ? sdw_update_slave_status+0x26/0x70 [ 46.798894] ? lock_release+0x147/0x2f0 [ 46.803262] ? lockdep_init_map_type+0x47/0x250 [ 46.808315] ? sdw_update_slave_status+0x26/0x70 [ 46.813456] sdw_update_slave_status+0x26/0x70 [ 46.818422] sdw_clear_slave_status+0xd8/0xe0 [ 46.823302] ? pm_generic_runtime_suspend+0x30/0x30 [ 46.828706] intel_resume_runtime+0x139/0x2a0 [ 46.833583] ? _raw_spin_unlock_irq+0x24/0x50 [ 46.838462] ? pm_generic_runtime_suspend+0x30/0x30 [ 46.843866] __rpm_callback+0x41/0x120 [ 46.848142] ? pm_generic_runtime_suspend+0x30/0x30 [ 46.853550] rpm_callback+0x5d/0x70 [ 46.857568] rpm_resume+0x531/0x7e0 [ 46.861578] ? _raw_spin_lock_irqsave+0x62/0x70 [ 46.866634] __pm_runtime_resume+0x4a/0x80 [ 46.871258] snd_soc_pcm_component_pm_runtime_get+0x2f/0xc0 [ 46.877358] __soc_pcm_open+0x62/0x520 [ 46.881634] ? dpcm_add_paths.isra.0+0x35d/0x4c0 [ 46.886784] dpcm_be_dai_startup+0x116/0x210 [ 46.891592] dpcm_fe_dai_open+0xf7/0x830 [ 46.896046] ? debug_mutex_init+0x33/0x50 [ 46.900591] snd_pcm_open_substream+0x54a/0x8b0 [ 46.905658] snd_pcm_open.part.0+0xdc/0x200 [ 46.910376] ? wake_up_q+0x90/0x90 [ 46.914312] snd_pcm_playback_open+0x51/0x80 [ 46.919118] chrdev_open+0xc0/0x250 [ 46.923147] ? cdev_device_add+0x90/0x90 [ 46.927608] do_dentry_open+0x15f/0x430 [ 46.931976] path_openat+0x75e/0xa80 [ 46.936086] do_filp_open+0xb2/0x160 [ 46.940194] ? lock_release+0x147/0x2f0 [ 46.944563] ? _raw_spin_unlock+0x29/0x50 [ 46.949101] do_sys_openat2+0x9a/0x160 [ 46.953377] __x64_sys_openat+0x53/0xa0 [ 46.957733] do_syscall_64+0x38/0x90 [ 46.961829] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 46.967402] RIP: 0033:0x7fa6397ccd3b [ 46.971506] Code: 25 00 00 41 00 3d 00 00 41 00 74 4b 64 8b 04 25 18 00 00 00 85 c0 75 67 44 89 e2 48 89 ee bf 9c ff ff ff b8 01 01 00 00 0f 05 <48> 3d 00 f0 ff ff 0f 87 91 00 00 00 48 8b 4c 24 28 64 48 33 0c 25 [ 46.991413] RSP: 002b:00007fff838e8990 EFLAGS: 00000246 ORIG_RAX: 0000000000000101 [ 46.999580] RAX: ffffffffffffffda RBX: 0000000000080802 RCX: 00007fa6397ccd3b [ 47.007311] RDX: 0000000000080802 RSI: 00007fff838e8b50 RDI: 00000000ffffff9c [ 47.015047] RBP: 00007fff838e8b50 R08: 0000000000000000 R09: 0000000000000011 [ 47.022787] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000080802 [ 47.030539] R13: 0000000000000004 R14: 0000000000000000 R15: 00007fff838e8b50 [ 47.038289] </TASK> Signed-off-by: Richard Fitzgerald <rf(a)opensource.cirrus.com> Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart(a)linux.intel.com> Link: https://lore.kernel.org/r/20230123172520.339367-1-rf@opensource.cirrus.com Signed-off-by: Vinod Koul <vkoul(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- drivers/soundwire/bus_type.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) diff --git a/drivers/soundwire/bus_type.c b/drivers/soundwire/bus_type.c index 04b3529f89293..963498db0fd22 100644 --- a/drivers/soundwire/bus_type.c +++ b/drivers/soundwire/bus_type.c @@ -105,20 +105,19 @@ static int sdw_drv_probe(struct device *dev) if (ret) return ret; - mutex_lock(&slave->sdw_dev_lock); - ret = drv->probe(slave, id); if (ret) { name = drv->name; if (!name) name = drv->driver.name; - mutex_unlock(&slave->sdw_dev_lock); dev_err(dev, "Probe of %s failed: %d\n", name, ret); dev_pm_domain_detach(dev, false); return ret; } + mutex_lock(&slave->sdw_dev_lock); + /* device is probed so let's read the properties now */ if (drv->ops && drv->ops->read_prop) drv->ops->read_prop(slave); @@ -167,14 +166,12 @@ static int sdw_drv_remove(struct device *dev) int ret = 0; mutex_lock(&slave->sdw_dev_lock); - slave->probed = false; + mutex_unlock(&slave->sdw_dev_lock); if (drv->remove) ret = drv->remove(slave); - mutex_unlock(&slave->sdw_dev_lock); - dev_pm_domain_detach(dev, false); return ret; -- 2.39.2

1 0

[PATCH AUTOSEL 6.2 64/64] soundwire: cadence: Drain the RX FIFO after an IO timeout
by Sasha Levin 03 Mar '23

03 Mar '23

From: Richard Fitzgerald <rf(a)opensource.cirrus.com> [ Upstream commit 0603a47bd3a8f439d7844b841eee1819353063e0 ] If wait_for_completion_timeout() times-out in _cdns_xfer_msg() it is possible that something could have been written to the RX FIFO. In this case, we should drain the RX FIFO so that anything in it doesn't carry over and mess up the next transfer. Obviously, if we got to this state something went wrong, and we don't really know the state of everything. The cleanup in this situation cannot be bullet-proof but we should attempt to avoid breaking future transaction, if only to reduce the amount of error noise when debugging the failure from a kernel log. Note that this patch only implements the draining for blocking (non-deferred) transfers. The deferred API doesn't have any proper handling of error conditions and would need some re-design before implementing cleanup. That is a task for a separate patch... Signed-off-by: Richard Fitzgerald <rf(a)opensource.cirrus.com> Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart(a)linux.intel.com> Link: https://lore.kernel.org/r/20221202161812.4186897-4-rf@opensource.cirrus.com Signed-off-by: Vinod Koul <vkoul(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- drivers/soundwire/cadence_master.c | 50 ++++++++++++++++-------------- 1 file changed, 27 insertions(+), 23 deletions(-) diff --git a/drivers/soundwire/cadence_master.c b/drivers/soundwire/cadence_master.c index 37eef80939475..901052dfe420a 100644 --- a/drivers/soundwire/cadence_master.c +++ b/drivers/soundwire/cadence_master.c @@ -554,6 +554,29 @@ cdns_fill_msg_resp(struct sdw_cdns *cdns, return SDW_CMD_OK; } +static void cdns_read_response(struct sdw_cdns *cdns) +{ + u32 num_resp, cmd_base; + int i; + + /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ + BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); + + num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); + num_resp &= CDNS_MCP_RX_FIFO_AVAIL; + if (num_resp > ARRAY_SIZE(cdns->response_buf)) { + dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); + num_resp = ARRAY_SIZE(cdns->response_buf); + } + + cmd_base = CDNS_MCP_CMD_BASE; + + for (i = 0; i < num_resp; i++) { + cdns->response_buf[i] = cdns_readl(cdns, cmd_base); + cmd_base += CDNS_MCP_CMD_WORD_LEN; + } +} + static enum sdw_command_response _cdns_xfer_msg(struct sdw_cdns *cdns, struct sdw_msg *msg, int cmd, int offset, int count, bool defer) @@ -595,6 +618,10 @@ _cdns_xfer_msg(struct sdw_cdns *cdns, struct sdw_msg *msg, int cmd, dev_err(cdns->dev, "IO transfer timed out, cmd %d device %d addr %x len %d\n", cmd, msg->dev_num, msg->addr, msg->len); msg->len = 0; + + /* Drain anything in the RX_FIFO */ + cdns_read_response(cdns); + return SDW_CMD_TIMEOUT; } @@ -768,29 +795,6 @@ EXPORT_SYMBOL(cdns_read_ping_status); * IRQ handling */ -static void cdns_read_response(struct sdw_cdns *cdns) -{ - u32 num_resp, cmd_base; - int i; - - /* RX_FIFO_AVAIL can be 2 entries more than the FIFO size */ - BUILD_BUG_ON(ARRAY_SIZE(cdns->response_buf) < CDNS_MCP_CMD_LEN + 2); - - num_resp = cdns_readl(cdns, CDNS_MCP_FIFOSTAT); - num_resp &= CDNS_MCP_RX_FIFO_AVAIL; - if (num_resp > ARRAY_SIZE(cdns->response_buf)) { - dev_warn(cdns->dev, "RX AVAIL %d too long\n", num_resp); - num_resp = ARRAY_SIZE(cdns->response_buf); - } - - cmd_base = CDNS_MCP_CMD_BASE; - - for (i = 0; i < num_resp; i++) { - cdns->response_buf[i] = cdns_readl(cdns, cmd_base); - cmd_base += CDNS_MCP_CMD_WORD_LEN; - } -} - static int cdns_update_slave_status(struct sdw_cdns *cdns, u64 slave_intstat) { -- 2.39.2

1 0