January 2025 - Linux-stable-mirror

[PATCH] usb: roles: cache usb roles received during switch registration

by Elson Roy Serrao

The role switch registration and set_role() can happen in parallel as they are invoked independent of each other. There is a possibility that a driver might spend significant amount of time in usb_role_switch_register() API due to the presence of time intensive operations like component_add() which operate under common mutex. This leads to a time window after allocating the switch and before setting the registered flag where the set role notifications are dropped. Below timeline summarizes this behavior Thread1 | Thread2 usb_role_switch_register() | | | ---> allocate switch | | | ---> component_add() | usb_role_switch_set_role() | | | | | --> Drop role notifications | | since sw->registered | | flag is not set. | | --->Set registered flag.| To avoid this, cache the last role received and set it once the switch registration is complete. Since we are now caching the roles based on registered flag, protect this flag with the switch mutex. Fixes: b787a3e78175 ("usb: roles: don't get/set_role() when usb_role_switch is unregistered") cc: stable(a)vger.kernel.org Signed-off-by: Elson Roy Serrao <quic_eserrao(a)quicinc.com> --- drivers/usb/roles/class.c | 45 ++++++++++++++++++++++++++++++++------- 1 file changed, 37 insertions(+), 8 deletions(-) diff --git a/drivers/usb/roles/class.c b/drivers/usb/roles/class.c index c58a12c147f4..c0149c31c01b 100644 --- a/drivers/usb/roles/class.c +++ b/drivers/usb/roles/class.c @@ -26,6 +26,8 @@ struct usb_role_switch { struct mutex lock; /* device lock*/ struct module *module; /* the module this device depends on */ enum usb_role role; + enum usb_role cached_role; + bool cached; bool registered; /* From descriptor */ @@ -65,6 +67,20 @@ static const struct component_ops connector_ops = { .unbind = connector_unbind, }; +static int __usb_role_switch_set_role(struct usb_role_switch *sw, + enum usb_role role) +{ + int ret; + + ret = sw->set(sw, role); + if (!ret) { + sw->role = role; + kobject_uevent(&sw->dev.kobj, KOBJ_CHANGE); + } + + return ret; +} + /** * usb_role_switch_set_role - Set USB role for a switch * @sw: USB role switch @@ -79,17 +95,21 @@ int usb_role_switch_set_role(struct usb_role_switch *sw, enum usb_role role) if (IS_ERR_OR_NULL(sw)) return 0; - if (!sw->registered) - return -EOPNOTSUPP; - + /* + * Since we have a valid sw struct here, role switch registration might + * be in progress. Hence cache the role here and send it out once + * registration is complete. + */ mutex_lock(&sw->lock); - - ret = sw->set(sw, role); - if (!ret) { - sw->role = role; - kobject_uevent(&sw->dev.kobj, KOBJ_CHANGE); + if (!sw->registered) { + sw->cached = true; + sw->cached_role = role; + mutex_unlock(&sw->lock); + return 0; } + ret = __usb_role_switch_set_role(sw, role); + mutex_unlock(&sw->lock); return ret; @@ -399,8 +419,14 @@ usb_role_switch_register(struct device *parent, dev_warn(&sw->dev, "failed to add component\n"); } + mutex_lock(&sw->lock); sw->registered = true; + if (sw->cached) + __usb_role_switch_set_role(sw, sw->cached_role); + + mutex_unlock(&sw->lock); + /* TODO: Symlinks for the host port and the device controller. */ return sw; @@ -417,7 +443,10 @@ void usb_role_switch_unregister(struct usb_role_switch *sw) { if (IS_ERR_OR_NULL(sw)) return; + mutex_lock(&sw->lock); sw->registered = false; + sw->cached = false; + mutex_unlock(&sw->lock); if (dev_fwnode(&sw->dev)) component_del(&sw->dev, &connector_ops); device_unregister(&sw->dev); -- 2.17.1

3 months, 4 weeks

3
2
0 0

[PATCH] soc: qcom: mark pd-mapper as broken

by Johan Hovold

When using the in-kernel pd-mapper on x1e80100, client drivers often fail to communicate with the firmware during boot, which specifically breaks battery and USB-C altmode notifications. This has been observed to happen on almost every second boot (41%) but likely depends on probe order: pmic_glink_altmode.pmic_glink_altmode pmic_glink.altmode.0: failed to send altmode request: 0x10 (-125) pmic_glink_altmode.pmic_glink_altmode pmic_glink.altmode.0: failed to request altmode notifications: -125 ucsi_glink.pmic_glink_ucsi pmic_glink.ucsi.0: failed to send UCSI read request: -125 qcom_battmgr.pmic_glink_power_supply pmic_glink.power-supply.0: failed to request power notifications In the same setup audio also fails to probe albeit much more rarely: PDR: avs/audio get domain list txn wait failed: -110 PDR: service lookup for avs/audio failed: -110 Chris Lew has provided an analysis and is working on a fix for the ECANCELED (125) errors, but it is not yet clear whether this will also address the audio regression. Even if this was first observed on x1e80100 there is currently no reason to believe that these issues are specific to that platform. Disable the in-kernel pd-mapper for now, and make sure to backport this to stable to prevent users and distros from migrating away from the user-space service. Fixes: 1ebcde047c54 ("soc: qcom: add pd-mapper implementation") Cc: stable(a)vger.kernel.org # 6.11 Link: https://lore.kernel.org/lkml/Zqet8iInnDhnxkT9@hovoldconsulting.com/ Signed-off-by: Johan Hovold <johan+linaro(a)kernel.org> --- It's now been over two months since I reported this regression, and even if we seem to be making some progress on at least some of these issues I think we need disable the pd-mapper temporarily until the fixes are in place (e.g. to prevent distros from dropping the user-space service). Johan #regzbot introduced: 1ebcde047c54 drivers/soc/qcom/Kconfig | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/soc/qcom/Kconfig b/drivers/soc/qcom/Kconfig index 74b9121240f8..35ddab9338d4 100644 --- a/drivers/soc/qcom/Kconfig +++ b/drivers/soc/qcom/Kconfig @@ -78,6 +78,7 @@ config QCOM_PD_MAPPER select QCOM_PDR_MSG select AUXILIARY_BUS depends on NET && QRTR && (ARCH_QCOM || COMPILE_TEST) + depends on BROKEN default QCOM_RPROC_COMMON help The Protection Domain Mapper maps registered services to the domains -- 2.45.2

3 months, 4 weeks

6
18
0 0

[PATCH AUTOSEL 6.13 01/16] ASoC: SOF: Intel: hda-dai: Ensure DAI widget is valid during params

by Sasha Levin

From: Bard Liao <yung-chuan.liao(a)linux.intel.com> [ Upstream commit 569922b82ca660f8b24e705f6cf674e6b1f99cc7 ] Each cpu DAI should associate with a widget. However, the topology might not create the right number of DAI widgets for aggregated amps. And it will cause NULL pointer deference. Check that the DAI widget associated with the CPU DAI is valid to prevent NULL pointer deference due to missing DAI widgets in topologies with aggregated amps. Signed-off-by: Bard Liao <yung-chuan.liao(a)linux.intel.com> Reviewed-by: Ranjani Sridharan <ranjani.sridharan(a)linux.intel.com> Reviewed-by: Péter Ujfalusi <peter.ujfalusi(a)linux.intel.com> Reviewed-by: Liam Girdwood <liam.r.girdwood(a)intel.com> Link: https://patch.msgid.link/20241203104853.56956-1-yung-chuan.liao@linux.intel… Signed-off-by: Mark Brown <broonie(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- sound/soc/sof/intel/hda-dai.c | 12 ++++++++++++ sound/soc/sof/intel/hda.c | 5 +++++ 2 files changed, 17 insertions(+) diff --git a/sound/soc/sof/intel/hda-dai.c b/sound/soc/sof/intel/hda-dai.c index 0db2a3e554fb2..da12aabc1bb85 100644 --- a/sound/soc/sof/intel/hda-dai.c +++ b/sound/soc/sof/intel/hda-dai.c @@ -503,6 +503,12 @@ int sdw_hda_dai_hw_params(struct snd_pcm_substream *substream, int ret; int i; + if (!w) { + dev_err(cpu_dai->dev, "%s widget not found, check amp link num in the topology\n", + cpu_dai->name); + return -EINVAL; + } + ops = hda_dai_get_ops(substream, cpu_dai); if (!ops) { dev_err(cpu_dai->dev, "DAI widget ops not set\n"); @@ -582,6 +588,12 @@ int sdw_hda_dai_hw_params(struct snd_pcm_substream *substream, */ for_each_rtd_cpu_dais(rtd, i, dai) { w = snd_soc_dai_get_widget(dai, substream->stream); + if (!w) { + dev_err(cpu_dai->dev, + "%s widget not found, check amp link num in the topology\n", + dai->name); + return -EINVAL; + } ipc4_copier = widget_to_copier(w); memcpy(&ipc4_copier->dma_config_tlv[cpu_dai_id], dma_config_tlv, sizeof(*dma_config_tlv)); diff --git a/sound/soc/sof/intel/hda.c b/sound/soc/sof/intel/hda.c index f991785f727e9..be689f6e10c81 100644 --- a/sound/soc/sof/intel/hda.c +++ b/sound/soc/sof/intel/hda.c @@ -63,6 +63,11 @@ static int sdw_params_stream(struct device *dev, struct snd_soc_dapm_widget *w = snd_soc_dai_get_widget(d, params_data->substream->stream); struct snd_sof_dai_config_data data = { 0 }; + if (!w) { + dev_err(dev, "%s widget not found, check amp link num in the topology\n", + d->name); + return -EINVAL; + } data.dai_index = (params_data->link_id << 8) | d->id; data.dai_data = params_data->alh_stream_id; data.dai_node_id = data.dai_data; -- 2.39.5

3 months, 4 weeks

3
18
0 0

[PATCH] ACPI: PRM: Remove unnecessary strict handler address checks

by Aubrey Li

Commit 088984c8d54c ("ACPI: PRM: Find EFI_MEMORY_RUNTIME block for PRM handler and context") added unnecessary strict handler address checks, caused the PRM module to fail in translating memory error addresses. Both static data buffer address and acpi parameter buffer address may be NULL if they are not needed, as described in section 4.1.2 PRM Handler Information Structure of Platform Runtime Mechanism specification [1]. Here are two examples from real hardware: ----PRMT.dsl---- - staic data address is not used [10Ch 0268 2] Revision : 0000 [10Eh 0270 2] Length : 002C [110h 0272 16] Handler GUID : F6A58D47-E04F-4F5A-86B8-2A50D4AA109B [120h 0288 8] Handler address : 0000000065CE51F4 [128h 0296 8] Satic Data Address : 0000000000000000 [130h 0304 8] ACPI Parameter Address : 000000006522A718 - ACPI parameter address is not used [1B0h 0432 2] Revision : 0000 [1B2h 0434 2] Length : 002C [1B4h 0436 16] Handler GUID : 657E8AE6-A8FC-4877-BB28-42E7DE1899A5 [1C4h 0452 8] Handler address : 0000000065C567C8 [1CCh 0460 8] Satic Data Address : 000000006113FB98 [1D4h 0468 8] ACPI Parameter Address : 0000000000000000 Fixes: 088984c8d54c ("ACPI: PRM: Find EFI_MEMORY_RUNTIME block for PRM handler and context") Reported-and-tested-by: Shi Liu <aurelianliu(a)tencent.com> Cc: All applicable <stable(a)vger.kernel.org> Signed-off-by: Aubrey Li <aubrey.li(a)linux.intel.com> Link: https://uefi.org/sites/default/files/resources/Platform%20Runtime%20Mechani… # [1] --- drivers/acpi/prmt.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/drivers/acpi/prmt.c b/drivers/acpi/prmt.c index 747f83f7114d..e549914a636c 100644 --- a/drivers/acpi/prmt.c +++ b/drivers/acpi/prmt.c @@ -287,9 +287,7 @@ static acpi_status acpi_platformrt_space_handler(u32 function, if (!handler || !module) goto invalid_guid; - if (!handler->handler_addr || - !handler->static_data_buffer_addr || - !handler->acpi_param_buffer_addr) { + if (!handler->handler_addr) { buffer->prm_status = PRM_HANDLER_ERROR; return AE_OK; } -- 2.34.1

3 months, 4 weeks

4
3
0 0

[PATCH v2] scsi: core: Do not retry I/Os during depopulation

by Igor Pylypiv

Fail I/Os instead of retry to prevent user space processes from being blocked on the I/O completion for several minutes. Retrying I/Os during "depopulation in progress" or "depopulation restore in progress" results in a continuous retry loop until the depopulation completes or until the I/O retry loop is aborted due to a timeout by the scsi_cmd_runtime_exceeced(). Depopulation is slow and can take 24+ hours to complete on 20+ TB HDDs. Most I/Os in the depopulation retry loop end up taking several minutes before returning the failure to user space. Cc: <stable(a)vger.kernel.org> # 4.18.x: 2bbeb8d scsi: core: Handle depopulation and restoration in progress Cc: <stable(a)vger.kernel.org> # 4.18.x Fixes: e37c7d9a0341 ("scsi: core: sanitize++ in progress") Signed-off-by: Igor Pylypiv <ipylypiv(a)google.com> --- Changes in v2: - Added Fixes: and Cc: stable tags. drivers/scsi/scsi_lib.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c index e7ea1f04164a..3ab4c958da45 100644 --- a/drivers/scsi/scsi_lib.c +++ b/drivers/scsi/scsi_lib.c @@ -872,13 +872,18 @@ static void scsi_io_completion_action(struct scsi_cmnd *cmd, int result) case 0x1a: /* start stop unit in progress */ case 0x1b: /* sanitize in progress */ case 0x1d: /* configuration in progress */ - case 0x24: /* depopulation in progress */ - case 0x25: /* depopulation restore in progress */ action = ACTION_DELAYED_RETRY; break; case 0x0a: /* ALUA state transition */ action = ACTION_DELAYED_REPREP; break; + /* + * Depopulation might take many hours, + * thus it is not worthwhile to retry. + */ + case 0x24: /* depopulation in progress */ + case 0x25: /* depopulation restore in progress */ + fallthrough; default: action = ACTION_FAIL; break; -- 2.48.1.362.g079036d154-goog

3 months, 4 weeks

2
1
0 0

[PATCH V10 1/4] perf/x86/intel: Apply static call for drain_pebs

by kan.liang＠linux.intel.com

From: "Peter Zijlstra (Intel)" <peterz(a)infradead.org> The x86_pmu_drain_pebs static call was introduced in commit 7c9903c9bf71 ("x86/perf, static_call: Optimize x86_pmu methods"), but it's not really used to replace the old method. Apply the static call for drain_pebs. Fixes: 7c9903c9bf71 ("x86/perf, static_call: Optimize x86_pmu methods") Signed-off-by: Kan Liang <kan.liang(a)linux.intel.com> Signed-off-by: Peter Zijlstra (Intel) <peterz(a)infradead.org> Cc: stable(a)vger.kernel.org --- New for V10 arch/x86/events/intel/core.c | 2 +- arch/x86/events/intel/ds.c | 2 +- arch/x86/events/perf_event.h | 1 + 3 files changed, 3 insertions(+), 2 deletions(-) diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c index 2a2824e9c50d..4daa45ae9bd2 100644 --- a/arch/x86/events/intel/core.c +++ b/arch/x86/events/intel/core.c @@ -3066,7 +3066,7 @@ static int handle_pmi_common(struct pt_regs *regs, u64 status) handled++; x86_pmu_handle_guest_pebs(regs, &data); - x86_pmu.drain_pebs(regs, &data); + static_call(x86_pmu_drain_pebs)(regs, &data); status &= intel_ctrl | GLOBAL_STATUS_TRACE_TOPAPMI; /* diff --git a/arch/x86/events/intel/ds.c b/arch/x86/events/intel/ds.c index ba74e1198328..322963b02a91 100644 --- a/arch/x86/events/intel/ds.c +++ b/arch/x86/events/intel/ds.c @@ -957,7 +957,7 @@ static inline void intel_pmu_drain_pebs_buffer(void) { struct perf_sample_data data; - x86_pmu.drain_pebs(NULL, &data); + static_call(x86_pmu_drain_pebs)(NULL, &data); } /* diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h index 31c2771545a6..084e9196b458 100644 --- a/arch/x86/events/perf_event.h +++ b/arch/x86/events/perf_event.h @@ -1107,6 +1107,7 @@ extern struct x86_pmu x86_pmu __read_mostly; DECLARE_STATIC_CALL(x86_pmu_set_period, *x86_pmu.set_period); DECLARE_STATIC_CALL(x86_pmu_update, *x86_pmu.update); +DECLARE_STATIC_CALL(x86_pmu_drain_pebs, *x86_pmu.drain_pebs); static __always_inline struct x86_perf_task_context_opt *task_context_opt(void *ctx) { -- 2.38.1

4 months

3
5
0 0

[PATCH v3 0/2] Tegra ADMA fixes

by Mohan Kumar D

- Fix build error due to 64-by-32 division - Additional check for adma max page Mohan Kumar D (2): dmaengine: tegra210-adma: Fix build error due to 64-by-32 division dmaengine: tegra210-adma: check for adma max page drivers/dma/tegra210-adma.c | 19 ++++++++++++++++--- 1 file changed, 16 insertions(+), 3 deletions(-) -- 2.25.1

4 months

4
11
0 0

[PATCH AUTOSEL 6.13 1/8] soc: qcom: pd-mapper: Add X1P42100

by Sasha Levin

From: Konrad Dybcio <konrad.dybcio(a)oss.qualcomm.com> [ Upstream commit e7282bf8a0e9bb8a4cb1be406674ff7bb7b264f2 ] X1P42100 is a cousin of X1E80100, and hence can make use of the latter's configuration. Do so. Signed-off-by: Konrad Dybcio <konrad.dybcio(a)oss.qualcomm.com> Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov(a)linaro.org> Link: https://lore.kernel.org/r/20241221-topic-x1p4_soc-v1-3-55347831d73c@oss.qua… Signed-off-by: Bjorn Andersson <andersson(a)kernel.org> Signed-off-by: Sasha Levin <sashal(a)kernel.org> --- drivers/soc/qcom/qcom_pd_mapper.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/soc/qcom/qcom_pd_mapper.c b/drivers/soc/qcom/qcom_pd_mapper.c index 6e30f08761aa4..50aa54996901f 100644 --- a/drivers/soc/qcom/qcom_pd_mapper.c +++ b/drivers/soc/qcom/qcom_pd_mapper.c @@ -561,6 +561,7 @@ static const struct of_device_id qcom_pdm_domains[] __maybe_unused = { { .compatible = "qcom,sm8550", .data = sm8550_domains, }, { .compatible = "qcom,sm8650", .data = sm8550_domains, }, { .compatible = "qcom,x1e80100", .data = x1e80100_domains, }, + { .compatible = "qcom,x1p42100", .data = x1e80100_domains, }, {}, }; -- 2.39.5

4 months

2
9
0 0

[PATCH V2 1/1] x86/tdx: Route safe halt execution via tdx_safe_halt()

by Vishal Annapurve

Direct HLT instruction execution causes #VEs for TDX VMs which is routed to hypervisor via tdvmcall. This process renders HLT instruction execution inatomic, so any preceding instructions like STI/MOV SS will end up enabling interrupts before the HLT instruction is routed to the hypervisor. This creates scenarios where interrupts could land during HLT instruction emulation without aborting halt operation leading to idefinite halt wait times. Commit bfe6ed0c6727 ("x86/tdx: Add HLT support for TDX guests") already upgraded x86_idle() to invoke tdvmcall to avoid such scenarios, but it didn't cover pv_native_safe_halt() which can be invoked using raw_safe_halt() from call sites like acpi_safe_halt(). raw_safe_halt() also returns with interrupts enabled so upgrade tdx_safe_halt() to enable interrupts by default and ensure that paravirt safe_halt() executions invoke tdx_safe_halt(). Earlier x86_idle() is now handled via tdx_idle() which simply invokes tdvmcall while preserving irq state. To avoid future call sites which cause HLT instruction emulation with irqs enabled, add a warn and fail the HLT instruction emulation. Cc: stable(a)vger.kernel.org Fixes: bfe6ed0c6727 ("x86/tdx: Add HLT support for TDX guests") Signed-off-by: Vishal Annapurve <vannapurve(a)google.com> --- Changes since V1: 1) Addressed comments from Dave H - Comment regarding adding a check for TDX VMs in halt path is not resolved in v2, would like feedback around better place to do so, maybe in pv_native_safe_halt (?). 2) Added a new version of tdx_safe_halt() that will enable interrupts. 3) Previous tdx_safe_halt() implementation is moved to newly introduced tdx_idle(). V1: https://lore.kernel.org/lkml/Z5l6L3Hen9_Y3SGC@google.com/T/ arch/x86/coco/tdx/tdx.c | 23 ++++++++++++++++++++++- arch/x86/include/asm/tdx.h | 2 +- arch/x86/kernel/process.c | 2 +- 3 files changed, 24 insertions(+), 3 deletions(-) diff --git a/arch/x86/coco/tdx/tdx.c b/arch/x86/coco/tdx/tdx.c index 0d9b090b4880..cc2a637dca15 100644 --- a/arch/x86/coco/tdx/tdx.c +++ b/arch/x86/coco/tdx/tdx.c @@ -14,6 +14,7 @@ #include <asm/ia32.h> #include <asm/insn.h> #include <asm/insn-eval.h> +#include <asm/paravirt_types.h> #include <asm/pgtable.h> #include <asm/set_memory.h> #include <asm/traps.h> @@ -380,13 +381,18 @@ static int handle_halt(struct ve_info *ve) { const bool irq_disabled = irqs_disabled(); + if (!irq_disabled) { + WARN_ONCE(1, "HLT instruction emulation unsafe with irqs enabled\n"); + return -EIO; + } + if (__halt(irq_disabled)) return -EIO; return ve_instr_len(ve); } -void __cpuidle tdx_safe_halt(void) +void __cpuidle tdx_idle(void) { const bool irq_disabled = false; @@ -397,6 +403,12 @@ void __cpuidle tdx_safe_halt(void) WARN_ONCE(1, "HLT instruction emulation failed\n"); } +static void __cpuidle tdx_safe_halt(void) +{ + tdx_idle(); + raw_local_irq_enable(); +} + static int read_msr(struct pt_regs *regs, struct ve_info *ve) { struct tdx_module_args args = { @@ -1083,6 +1095,15 @@ void __init tdx_early_init(void) x86_platform.guest.enc_kexec_begin = tdx_kexec_begin; x86_platform.guest.enc_kexec_finish = tdx_kexec_finish; +#ifdef CONFIG_PARAVIRT_XXL + /* + * halt instruction execution is not atomic for TDX VMs as it generates + * #VEs, so otherwise "safe" halt invocations which cause interrupts to + * get enabled right after halt instruction don't work for TDX VMs. + */ + pv_ops.irq.safe_halt = tdx_safe_halt; +#endif + /* * TDX intercepts the RDMSR to read the X2APIC ID in the parallel * bringup low level code. That raises #VE which cannot be handled diff --git a/arch/x86/include/asm/tdx.h b/arch/x86/include/asm/tdx.h index eba178996d84..dd386500ab1c 100644 --- a/arch/x86/include/asm/tdx.h +++ b/arch/x86/include/asm/tdx.h @@ -58,7 +58,7 @@ void tdx_get_ve_info(struct ve_info *ve); bool tdx_handle_virt_exception(struct pt_regs *regs, struct ve_info *ve); -void tdx_safe_halt(void); +void tdx_idle(void); bool tdx_early_handle_ve(struct pt_regs *regs); diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c index f63f8fd00a91..4083838fe4a0 100644 --- a/arch/x86/kernel/process.c +++ b/arch/x86/kernel/process.c @@ -933,7 +933,7 @@ void __init select_idle_routine(void) static_call_update(x86_idle, mwait_idle); } else if (cpu_feature_enabled(X86_FEATURE_TDX_GUEST)) { pr_info("using TDX aware idle routine\n"); - static_call_update(x86_idle, tdx_safe_halt); + static_call_update(x86_idle, tdx_idle); } else { static_call_update(x86_idle, default_idle); } -- 2.48.1.262.g85cc9f2d1e-goog

4 months

3
15
0 0

FAILED: patch "[PATCH] drm/amd/display: Reduce accessing remote DPCD overhead" failed to apply to 6.12-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 6.12-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.12.y git checkout FETCH_HEAD git cherry-pick -x adb4998f4928a17d91be054218a902ba9f8c1f93 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025012032-phoenix-crushing-da7a@gregkh' --subject-prefix 'PATCH 6.12.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From adb4998f4928a17d91be054218a902ba9f8c1f93 Mon Sep 17 00:00:00 2001 From: Wayne Lin <Wayne.Lin(a)amd.com> Date: Mon, 9 Dec 2024 15:25:35 +0800 Subject: [PATCH] drm/amd/display: Reduce accessing remote DPCD overhead [Why] Observed frame rate get dropped by tool like glxgear. Even though the output to monitor is 60Hz, the rendered frame rate drops to 30Hz lower. It's due to code path in some cases will trigger dm_dp_mst_is_port_support_mode() to read out remote Link status to assess the available bandwidth for dsc maniplation. Overhead of keep reading remote DPCD is considerable. [How] Store the remote link BW in mst_local_bw and use end-to-end full_pbn as an indicator to decide whether update the remote link bw or not. Whenever we need the info to assess the BW, visit the stored one first. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3720 Fixes: fa57924c76d9 ("drm/amd/display: Refactor function dm_dp_mst_is_port_support_mode()") Cc: Mario Limonciello <mario.limonciello(a)amd.com> Cc: Alex Deucher <alexander.deucher(a)amd.com> Reviewed-by: Jerry Zuo <jerry.zuo(a)amd.com> Signed-off-by: Wayne Lin <Wayne.Lin(a)amd.com> Signed-off-by: Tom Chung <chiahsuan.chung(a)amd.com> Tested-by: Daniel Wheeler <daniel.wheeler(a)amd.com> Signed-off-by: Alex Deucher <alexander.deucher(a)amd.com> (cherry picked from commit 4a9a918545455a5979c6232fcf61ed3d8f0db3ae) Cc: stable(a)vger.kernel.org diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h index 6464a8378387..2227cd8e4a89 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h @@ -697,6 +697,8 @@ struct amdgpu_dm_connector { struct drm_dp_mst_port *mst_output_port; struct amdgpu_dm_connector *mst_root; struct drm_dp_aux *dsc_aux; + uint32_t mst_local_bw; + uint16_t vc_full_pbn; struct mutex handle_mst_msg_ready; /* TODO see if we can merge with ddc_bus or make a dm_connector */ diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c index aadaa61ac5ac..1080075ccb17 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c @@ -155,6 +155,17 @@ amdgpu_dm_mst_connector_late_register(struct drm_connector *connector) return 0; } + +static inline void +amdgpu_dm_mst_reset_mst_connector_setting(struct amdgpu_dm_connector *aconnector) +{ + aconnector->drm_edid = NULL; + aconnector->dsc_aux = NULL; + aconnector->mst_output_port->passthrough_aux = NULL; + aconnector->mst_local_bw = 0; + aconnector->vc_full_pbn = 0; +} + static void amdgpu_dm_mst_connector_early_unregister(struct drm_connector *connector) { @@ -182,9 +193,7 @@ amdgpu_dm_mst_connector_early_unregister(struct drm_connector *connector) dc_sink_release(dc_sink); aconnector->dc_sink = NULL; - aconnector->drm_edid = NULL; - aconnector->dsc_aux = NULL; - port->passthrough_aux = NULL; + amdgpu_dm_mst_reset_mst_connector_setting(aconnector); } aconnector->mst_status = MST_STATUS_DEFAULT; @@ -504,9 +513,7 @@ dm_dp_mst_detect(struct drm_connector *connector, dc_sink_release(aconnector->dc_sink); aconnector->dc_sink = NULL; - aconnector->drm_edid = NULL; - aconnector->dsc_aux = NULL; - port->passthrough_aux = NULL; + amdgpu_dm_mst_reset_mst_connector_setting(aconnector); amdgpu_dm_set_mst_status(&aconnector->mst_status, MST_REMOTE_EDID | MST_ALLOCATE_NEW_PAYLOAD | MST_CLEAR_ALLOCATED_PAYLOAD, @@ -1819,9 +1826,18 @@ enum dc_status dm_dp_mst_is_port_support_mode( struct drm_dp_mst_port *immediate_upstream_port = NULL; uint32_t end_link_bw = 0; - /*Get last DP link BW capability*/ - if (dp_get_link_current_set_bw(&aconnector->mst_output_port->aux, &end_link_bw)) { - if (stream_kbps > end_link_bw) { + /*Get last DP link BW capability. Mode shall be supported by Legacy peer*/ + if (aconnector->mst_output_port->pdt != DP_PEER_DEVICE_DP_LEGACY_CONV && + aconnector->mst_output_port->pdt != DP_PEER_DEVICE_NONE) { + if (aconnector->vc_full_pbn != aconnector->mst_output_port->full_pbn) { + dp_get_link_current_set_bw(&aconnector->mst_output_port->aux, &end_link_bw); + aconnector->vc_full_pbn = aconnector->mst_output_port->full_pbn; + aconnector->mst_local_bw = end_link_bw; + } else { + end_link_bw = aconnector->mst_local_bw; + } + + if (end_link_bw > 0 && stream_kbps > end_link_bw) { DRM_DEBUG_DRIVER("MST_DSC dsc decode at last link." "Mode required bw can't fit into last link\n"); return DC_FAIL_BANDWIDTH_VALIDATE;

4 months

5
4
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror January 2025