Add parents to struct pmu -> dev
Subsystems (26)
cover Add parents to struct pmu -> dev
These are the low hanging fruit following GregKH's feedback that all the devices registered via perf_pmu_register() should have parents.
Note that this causes potential ABI breakage.
It may fall in the category of it isn't breakage if no one notices but I can't be certain of that. Whilst it is arguable that no one should be been accessing PMUs except via the event_source bus, there was documentation suggesting /sys/devices/ for particular PMUs (because it was a shorter path?)
The first patch is pulled out of the series:
https://lore.kernel.org/linux-cxl/[email protected]/
[PATCH v3 0/5] CXL 3.0 Performance Monitoring Unit support
In that particular case it is very useful to be able to figure out which CXL device the PMU device is associated with and looking at it's parents in the device model as shown with ls -lh /sys/bus/event_sources/devices/ is a very easy way to do this (once it is correctly parented).
Addressing all the other instances of struct pmu not covered by this series is likely to be a more complex discussion but unlikely to have an affect on what is proposed here.
Documentation updates deliberately 'fixed' in separate patches before changing the path to highlight that using /sys/bus/event_source/devices path is unchanged by this series and that is presumed to be the most common way these files are accessed.
Jonathan Cameron (32):
perf: Allow a PMU to have a parent
perf/hisi-pcie: Assign parent for event_source device
Documentation: hisi-pmu: Drop reference to /sys/devices path perf/hisi-uncore: Assign parents for event_source devices
Documentation: hns-pmu: Use /sys/bus/event_source/devices paths perf/hisi-hns3: Assign parents for event_source device perf/amlogic: Assign parents for event_source devices perf/arm_cspmu: Assign parents for event_source devices
Documentation: xgene-pmu: Use /sys/bus/event_source/devices paths perf/xgene: Assign parents for event_source devices
Documentation: thunderx2-pmu: Use /sys/bus/event_source/devices paths perf/thunderx2: Assign parents for event_source devices perf/riscv: Assign parents for event_source devices
Documentation: qcom-pmu: Use /sys/bus/event_source/devices paths perf/qcom: Assign parents for event_source devices perf/imx_ddr: Assign parents for event_source devices perf/arm_pmu: Assign parents for event_source devices perf/alibaba_uncore: Assign parents for event_source device perf/arm-cci: Assign parents for event_source device perf/arm-ccn: Assign parents for event_source device perf/arm-cmn: Assign parents for event_source device perf/arm-dmc620: Assign parents for event_source device perf/arm-dsu: Assign parents for event_source device perf/arm-smmuv3: Assign parents for event_source device perf/arm-spe: Assign parents for event_source device
arc: Assign parents for event_source devices
ARM: imx: Assign parents for mmdc event_source devices
dmaengine: idxd: Assign parent for event_source device
fpga: dfl: Assign parent for event_source device drivers/nvdimm: Assign parent for event_source device
Documentation: ABI + trace: hisi_ptt: update paths to bus/event_source
hwtracing: hisi_ptt: Assign parent for event_source device
...i_ptt => sysfs-bus-event_source-devices-hisi_ptt} | 12 ++++++------
Documentation/admin-guide/perf/hisi-pmu.rst | 1 -
Documentation/admin-guide/perf/hns3-pmu.rst | 8 ++++----
Documentation/admin-guide/perf/qcom_l2_pmu.rst | 2 +-
Documentation/admin-guide/perf/qcom_l3_pmu.rst | 2 +-
Documentation/admin-guide/perf/thunderx2-pmu.rst | 2 +-
Documentation/admin-guide/perf/xgene-pmu.rst | 2 +-
Documentation/trace/hisi-ptt.rst | 4 ++--
MAINTAINERS | 2 +-
arch/arc/kernel/perf_event.c | 1 +
arch/arm/mach-imx/mmdc.c | 1 +
drivers/dma/idxd/perfmon.c | 1 +
drivers/fpga/dfl-fme-perf.c | 1 +
drivers/hwtracing/ptt/hisi_ptt.c | 1 +
drivers/nvdimm/nd_perf.c | 1 +
drivers/perf/alibaba_uncore_drw_pmu.c | 1 +
drivers/perf/amlogic/meson_ddr_pmu_core.c | 1 +
drivers/perf/arm-cci.c | 1 +
drivers/perf/arm-ccn.c | 1 +
drivers/perf/arm-cmn.c | 1 +
drivers/perf/arm_cspmu/arm_cspmu.c | 1 +
drivers/perf/arm_dmc620_pmu.c | 1 +
drivers/perf/arm_dsu_pmu.c | 1 +
drivers/perf/arm_pmu_platform.c | 1 +
drivers/perf/arm_smmuv3_pmu.c | 1 +
drivers/perf/arm_spe_pmu.c | 1 +
drivers/perf/fsl_imx8_ddr_perf.c | 1 +
drivers/perf/hisilicon/hisi_pcie_pmu.c | 1 +
drivers/perf/hisilicon/hisi_uncore_pmu.c | 1 +
drivers/perf/hisilicon/hns3_pmu.c | 1 +
drivers/perf/qcom_l2_pmu.c | 1 +
drivers/perf/qcom_l3_pmu.c | 1 +
drivers/perf/riscv_pmu_legacy.c | 1 +
drivers/perf/riscv_pmu_sbi.c | 1 +
drivers/perf/thunderx2_pmu.c | 1 +
drivers/perf/xgene_pmu.c | 1 +
include/linux/perf_event.h | 1 +
kernel/events/core.c | 1 +
38 files changed, 46 insertions(+), 18 deletions(-)
rename Documentation/ABI/testing/{sysfs-devices-hisi_ptt => sysfs-bus-event_source-devices-hisi_ptt} (83%)Signature · 1 line
2.37.2
PATCH 1/32 perf: Allow a PMU to have a parent
Some PMUs have well defined parents such as PCI devices. As the device_initialize() and device_add() are all within pmu_dev_alloc() which is called from perf_pmu_register() there is no opportunity to set the parent from within a driver.
Add a struct device *parent field to struct pmu and use that to set the parent.
---
Previously posted in CPMU series hence the change log.
v3: No change
--- include/linux/perf_event.h | 1 + kernel/events/core.c | 1 + 2 files changed, 2 insertions(+)
include/linux/perf_event.h +1 -0
@@ -303,6 +303,7 @@ struct pmu {struct module *module;struct device *dev; +struct device *parent;const struct attribute_group **attr_groups;const struct attribute_group **attr_update;const char *name;
kernel/events/core.c +1 -0
@@ -11367,6 +11367,7 @@ static int pmu_dev_alloc(struct pmu *pmu)dev_set_drvdata (pmu->dev, pmu); pmu->dev->bus = &pmu_bus; + pmu->dev->parent = pmu->parent; pmu->dev->release = pmu_dev_release; ret =dev_set_name (pmu->dev,"%s" , pmu->name);
Signature · 1 line
2.37.2
Re: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 11 lines · Jonathan Cameron
Some PMUs have well defined parents such as PCI devices. As the
device_initialize()anddevice_add()are all withinpmu_dev_alloc()which is called fromperf_pmu_register()there is no opportunity to set the parent from within a driver.Add a struct device *parent field to struct pmu and use that to set the parent.
Signed-off-by Jonathan Cameron <[email protected]>Reviewed-by Dan Williams <[email protected]>
Re: [PATCH 01/32] perf: Allow a PMU to have a parent
In reply to Jonathan CameronQuoted 36 lines · Jonathan Cameron
Some PMUs have well defined parents such as PCI devices. As the
device_initialize()anddevice_add()are all withinpmu_dev_alloc()which is called fromperf_pmu_register()there is no opportunity to set the parent from within a driver.Add a struct device *parent field to struct pmu and use that to set the parent.
Signed-off-by Jonathan Cameron <[email protected]>Reviewed-by Dan Williams <[email protected]>---
Previously posted in CPMU series hence the change log.
v3: No change--- include/linux/perf_event.h | 1 + kernel/events/core.c | 1 + 2 files changed, 2 insertions(+)include/linux/perf_event.h +1
-0@@ -303,6 +303,7 @@ struct pmu {struct module *module;struct device *dev; +struct device *parent;const struct attribute_group **attr_groups;const struct attribute_group **attr_update;const char *name;kernel/events/core.c +0
-0
kernel/events/core.c static int pmu_dev_alloc(struct pmu *pmu) · line 11370
Quoted 4 lines
kernel/events/core.c +1
-0@@ -11367,6 +11367,7 @@ static int pmu_dev_alloc(struct pmu *pmu)dev_set_drvdata (pmu->dev, pmu); pmu->dev->bus = &pmu_bus; + pmu->dev->parent = pmu->parent;
If there's no parent assigned, is it ok to add some check here? Then we can find it earlier maybe at develop stage.
Thanks.
Quoted 4 lines
kernel/events/core.c +0
-0pmu->dev->release = pmu_dev_release; ret =dev_set_name (pmu->dev,"%s" , pmu->name);
Re: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 38 lines · On Thu, 6 Apr 2023 12:03:27 +0800 Yicong Yang <[email protected]> wrote:
Quoted 37 lines · Jonathan Cameron
Some PMUs have well defined parents such as PCI devices. As the
device_initialize()anddevice_add()are all withinpmu_dev_alloc()which is called fromperf_pmu_register()there is no opportunity to set the parent from within a driver.Add a struct device *parent field to struct pmu and use that to set the parent.
Signed-off-by Jonathan Cameron <[email protected]>Reviewed-by Dan Williams <[email protected]>---
Previously posted in CPMU series hence the change log.
v3: No change--- include/linux/perf_event.h | 1 + kernel/events/core.c | 1 + 2 files changed, 2 insertions(+)include/linux/perf_event.h +1
-0@@ -303,6 +303,7 @@ struct pmu {struct module *module;struct device *dev; +struct device *parent;const struct attribute_group **attr_groups;const struct attribute_group **attr_update;const char *name;kernel/events/core.c +0
-0@@ -11367,6 +11367,7 @@ static int pmu_dev_alloc(struct pmu *pmu)
Quoted 6 lines
kernel/events/core.c static int pmu_dev_alloc(struct pmu *pmu) · line 11370
Quoted 3 lines
kernel/events/core.c +1
-0dev_set_drvdata (pmu->dev, pmu); pmu->dev->bus = &pmu_bus; + pmu->dev->parent = pmu->parent;If there's no parent assigned, is it ok to add some check here? Then we can find it earlier maybe at develop stage.
In the long run I agree it would be good. Short term there are more instances of struct pmu that don't have parents than those that do (even after this series). We need to figure out what to do about those before adding checks on it being set.
Jonathan
Quoted 7 lines
Thanks.
Quoted 4 lines
kernel/events/core.c +0
-0pmu->dev->release = pmu_dev_release; ret =dev_set_name (pmu->dev,"%s" , pmu->name);
Peter Zijlstra
maintainerRe: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 4 lines · Jonathan Cameron
In the long run I agree it would be good. Short term there are more instances of struct pmu that don't have parents than those that do (even after this series). We need to figure out what to do about those before adding checks on it being set.
Right, I don't think you've touched *any* of the x86 PMUs for example, and getting everybody that boots an x86 kernel a warning isn't going to go over well :-)
Re: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 11 lines · On Thu, 6 Apr 2023 14:40:40 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 4 lines · Jonathan Cameron
In the long run I agree it would be good. Short term there are more instances of struct pmu that don't have parents than those that do (even after this series). We need to figure out what to do about those before adding checks on it being set.
Right, I don't think you've touched *any* of the x86 PMUs for example, and getting everybody that boots an x86 kernel a warning isn't going to go over well :-)
It was tempting :) "Warning: Parentless PMU: try a different architecture."
I'd love some inputs on what the x86 PMU devices parents should be? CPU counters in general tend to just spin out of deep in the architecture code.
My overall favorite is an l2 cache related PMU that is spun up in arch/arm/kernel/irq.c init_IRQ()
I'm just not going to try and figure out why...
Jonathan
Re: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 17 lines · Jonathan Cameron
Quoted 11 lines · On Thu, 6 Apr 2023 14:40:40 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 4 lines · Jonathan Cameron
In the long run I agree it would be good. Short term there are more instances of struct pmu that don't have parents than those that do (even after this series). We need to figure out what to do about those before adding checks on it being set.
Right, I don't think you've touched *any* of the x86 PMUs for example, and getting everybody that boots an x86 kernel a warning isn't going to go over well :-)
It was tempting :) "Warning: Parentless PMU: try a different architecture."
Quoted 7 lines
I'd love some inputs on what the x86 PMU devices parents should be? CPU counters in general tend to just spin out of deep in the architecture code.
My overall favorite is an l2 cache related PMU that is spun up in
arch/arm/kernel/irq.cinit_IRQ()I'm just not going to try and figure out why...
Why not change the api to force a parent to be passed in? And if one isn't, we make it a "virtual" device and throw it in the class for them?
thanks,
greg k-h
Re: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 21 lines · On Thu, 6 Apr 2023 19:08:45 +0200 Greg KH <[email protected]> wrote:
Quoted 20 lines · Jonathan Cameron
Quoted 11 lines · On Thu, 6 Apr 2023 14:40:40 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 4 lines · Jonathan Cameron
In the long run I agree it would be good. Short term there are more instances of struct pmu that don't have parents than those that do (even after this series). We need to figure out what to do about those before adding checks on it being set.
Right, I don't think you've touched *any* of the x86 PMUs for example, and getting everybody that boots an x86 kernel a warning isn't going to go over well :-)
It was tempting :) "Warning: Parentless PMU: try a different architecture."
I'd love some inputs on what the x86 PMU devices parents should be? CPU counters in general tend to just spin out of deep in the architecture code.
Quoted 7 lines
Quoted 4 lines
My overall favorite is an l2 cache related PMU that is spun up in
arch/arm/kernel/irq.cinit_IRQ()I'm just not going to try and figure out why...
Why not change the api to force a parent to be passed in? And if one isn't, we make it a "virtual" device and throw it in the class for them?
Longer term I'd be fine doing that, but I'd like to identify the right parents rather than end up sweeping it under the carpet. Anything we either get completely stuck on (or decide we don't care about) could indeed fall back to a virtual device.
Jonathan
Quoted 4 lines
thanks,
greg k-h
Peter Zijlstra
maintainerRe: [PATCH 01/32] perf: Allow a PMU to have a parent
In reply to Jonathan CameronQuoted 10 lines · Jonathan Cameron
Quoted 7 lines · On Thu, 6 Apr 2023 14:40:40 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 4 lines · Jonathan Cameron
In the long run I agree it would be good. Short term there are more instances of struct pmu that don't have parents than those that do (even after this series). We need to figure out what to do about those before adding checks on it being set.
Quoted 6 lines
Quoted 4 lines
Right, I don't think you've touched *any* of the x86 PMUs for example, and getting everybody that boots an x86 kernel a warning isn't going to go over well :-)
It was tempting :) "Warning: Parentless PMU: try a different architecture."
Haha!
Quoted 2 lines
I'd love some inputs on what the x86 PMU devices parents should be? CPU counters in general tend to just spin out of deep in the architecture code.
For the 'simple' ones I suppose we can use the CPU device.
Quoted 2 lines
My overall favorite is an l2 cache related PMU that is spun up in
arch/arm/kernel/irq.cinit_IRQ()
Yeah, we're going to have a ton of them as well. Some of them are PCI devices and have a clear parent, others, not so much :/
Mark Rutland
maintainerRe: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 16 lines · Peter Zijlstra
Quoted 15 lines · Jonathan Cameron
Quoted 11 lines · On Thu, 6 Apr 2023 14:40:40 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 4 lines · Jonathan Cameron
In the long run I agree it would be good. Short term there are more instances of struct pmu that don't have parents than those that do (even after this series). We need to figure out what to do about those before adding checks on it being set.
Right, I don't think you've touched *any* of the x86 PMUs for example, and getting everybody that boots an x86 kernel a warning isn't going to go over well :-)
Quoted 8 lines
Quoted 1 line
It was tempting :) "Warning: Parentless PMU: try a different architecture."
Haha!
Quoted 2 lines
I'd love some inputs on what the x86 PMU devices parents should be? CPU counters in general tend to just spin out of deep in the architecture code.
For the 'simple' ones I suppose we can use the CPU device.
Uh, *which* CPU device? Do we have a container device for all CPUs?
Quoted 2 lines
Quoted 2 lines
My overall favorite is an l2 cache related PMU that is spun up in
arch/arm/kernel/irq.cinit_IRQ()
That's an artifact of the L2 cache controller driver getting initialized there; ideally we'd have a device for the L2 cache itself (which presumably should hang off an aggregate CPU device).
Quoted 2 lines
Yeah, we're going to have a ton of them as well. Some of them are PCI devices and have a clear parent, others, not so much :/
In a number of places the only thing we have is the PMU driver, and we don't have a driver (or device) for the HW block it's a part of. Largely that's interconnect PMUs; we could create container devices there.
Mark.
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Peter Zijlstra
maintainerRe: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 19 lines · Mark Rutland
Quoted 18 lines · Peter Zijlstra
Quoted 16 lines · Jonathan Cameron
Quoted 11 lines · On Thu, 6 Apr 2023 14:40:40 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 4 lines · Jonathan Cameron
In the long run I agree it would be good. Short term there are more instances of struct pmu that don't have parents than those that do (even after this series). We need to figure out what to do about those before adding checks on it being set.
Right, I don't think you've touched *any* of the x86 PMUs for example, and getting everybody that boots an x86 kernel a warning isn't going to go over well :-)
It was tempting :) "Warning: Parentless PMU: try a different architecture."
Quoted 8 lines
Quoted 6 lines
Haha!
Quoted 2 lines
I'd love some inputs on what the x86 PMU devices parents should be? CPU counters in general tend to just spin out of deep in the architecture code.
For the 'simple' ones I suppose we can use the CPU device.
Uh, *which* CPU device? Do we have a container device for all CPUs?
drivers/base/cpu.c:per_cpu(cpu_sys_devices, cpu) for whatever the core pmu is for that cpu ?
Quoted 6 lines
Quoted 2 lines
Quoted 2 lines
My overall favorite is an l2 cache related PMU that is spun up in
arch/arm/kernel/irq.cinit_IRQ()That's an artifact of the L2 cache controller driver getting initialized there; ideally we'd have a device for the L2 cache itself (which presumably should hang off an aggregate CPU device).
/sys/devices/system/cpu/cpuN/cache/indexM
has a struct device somewhere in drivers/base/cacheinfo.c:ci_index_dev or somesuch.
Quoted 6 lines
Quoted 2 lines
Yeah, we're going to have a ton of them as well. Some of them are PCI devices and have a clear parent, others, not so much :/
In a number of places the only thing we have is the PMU driver, and we don't have a driver (or device) for the HW block it's a part of. Largely that's interconnect PMUs; we could create container devices there.
Dont they have a PCI device? But yeah, some are going to be a wee bit challenging.
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Mark Rutland
maintainerRe: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 25 lines · Peter Zijlstra
Quoted 24 lines · Mark Rutland
Quoted 23 lines · Peter Zijlstra
Quoted 10 lines · Jonathan Cameron
Quoted 7 lines · On Thu, 6 Apr 2023 14:40:40 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 4 lines · Jonathan Cameron
In the long run I agree it would be good. Short term there are more instances of struct pmu that don't have parents than those that do (even after this series). We need to figure out what to do about those before adding checks on it being set.
Quoted 6 lines
Quoted 4 lines
Right, I don't think you've touched *any* of the x86 PMUs for example, and getting everybody that boots an x86 kernel a warning isn't going to go over well :-)
It was tempting :) "Warning: Parentless PMU: try a different architecture."
Haha!
Quoted 2 lines
I'd love some inputs on what the x86 PMU devices parents should be? CPU counters in general tend to just spin out of deep in the architecture code.
Quoted 6 lines
Quoted 3 lines
Quoted 1 line
For the 'simple' ones I suppose we can use the CPU device.
Uh, *which* CPU device? Do we have a container device for all CPUs?
drivers/base/cpu.c:per_cpu(cpu_sys_devices, cpu) for whatever the core pmu is for that cpu ?
... but the struct pmu covers several CPUs, so I don't have a single 'cpu', no?
If I have a system where cpu{0,1,2} are Cortex-A53 and cpu{3,4} are Cortex-A57, I have two struct pmu instances, each associated with several CPUs. When I probe each of those I determine a cpumask for each.
Quoted 11 lines
Quoted 6 lines
Quoted 2 lines
Quoted 2 lines
My overall favorite is an l2 cache related PMU that is spun up in
arch/arm/kernel/irq.cinit_IRQ()That's an artifact of the L2 cache controller driver getting initialized there; ideally we'd have a device for the L2 cache itself (which presumably should hang off an aggregate CPU device).
/sys/devices/system/cpu/cpuN/cache/indexMhas a struct device somewhere in drivers/base/cacheinfo.c:ci_index_dev or somesuch.
I guess, but I don't think the L2 cache controller (the PL310) is actually tied to that today.
Quoted 9 lines
Quoted 6 lines
Quoted 2 lines
Yeah, we're going to have a ton of them as well. Some of them are PCI devices and have a clear parent, others, not so much :/
In a number of places the only thing we have is the PMU driver, and we don't have a driver (or device) for the HW block it's a part of. Largely that's interconnect PMUs; we could create container devices there.
Dont they have a PCI device? But yeah, some are going to be a wee bit challenging.
The system might not even have PCI, so it's arguable that they should just hang off an MMIO bus (which is effectively what the platofrm bus is).
Thanks,
Mark.
Peter Zijlstra
maintainerRe: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 10 lines · Mark Rutland
Quoted 4 lines
Quoted 1 line
Uh, *which* CPU device? Do we have a container device for all CPUs?
drivers/base/cpu.c:per_cpu(cpu_sys_devices, cpu) for whatever the core pmu is for that cpu ?
... but the struct pmu covers several CPUs, so I don't have a single 'cpu', no?
If I have a system where cpu{0,1,2} are Cortex-A53 and cpu{3,4} are Cortex-A57, I have two struct pmu instances, each associated with several CPUs. When I probe each of those I determine a cpumask for each.
Bah :/ Clearly I overlooked the disparity there.
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines
Quoted 2 lines
My overall favorite is an l2 cache related PMU that is spun up in
arch/arm/kernel/irq.cinit_IRQ()That's an artifact of the L2 cache controller driver getting initialized there; ideally we'd have a device for the L2 cache itself (which presumably should hang off an aggregate CPU device).
Quoted 7 lines
Quoted 4 lines
/sys/devices/system/cpu/cpuN/cache/indexMhas a struct device somewhere in drivers/base/cacheinfo.c:ci_index_dev or somesuch.
I guess, but I don't think the L2 cache controller (the PL310) is actually tied to that today.
All it would do is make fancy links in sysfs I think, who cares ;-)
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines
Yeah, we're going to have a ton of them as well. Some of them are PCI devices and have a clear parent, others, not so much :/
In a number of places the only thing we have is the PMU driver, and we don't have a driver (or device) for the HW block it's a part of. Largely that's interconnect PMUs; we could create container devices there.
Quoted 5 lines
Quoted 2 lines
Dont they have a PCI device? But yeah, some are going to be a wee bit challenging.
The system might not even have PCI, so it's arguable that they should just hang off an MMIO bus (which is effectively what the platofrm bus is).
You and your dodgy platforms :-)
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Robin Murphy
maintainerRe: [PATCH 01/32] perf: Allow a PMU to have a parent
Quoted 39 lines · Peter Zijlstra
Quoted 10 lines · Mark Rutland
Quoted 4 lines
Quoted 1 line
Uh, *which* CPU device? Do we have a container device for all CPUs?
drivers/base/cpu.c:per_cpu(cpu_sys_devices, cpu) for whatever the core pmu is for that cpu ?
... but the struct pmu covers several CPUs, so I don't have a single 'cpu', no?
If I have a system where cpu{0,1,2} are Cortex-A53 and cpu{3,4} are Cortex-A57, I have two struct pmu instances, each associated with several CPUs. When I probe each of those I determine a cpumask for each.
Bah :/ Clearly I overlooked the disparity there.
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines
Quoted 2 lines
My overall favorite is an l2 cache related PMU that is spun up in
arch/arm/kernel/irq.cinit_IRQ()That's an artifact of the L2 cache controller driver getting initialized there; ideally we'd have a device for the L2 cache itself (which presumably should hang off an aggregate CPU device).
Quoted 7 lines
Quoted 4 lines
/sys/devices/system/cpu/cpuN/cache/indexMhas a struct device somewhere in drivers/base/cacheinfo.c:ci_index_dev or somesuch.
I guess, but I don't think the L2 cache controller (the PL310) is actually tied to that today.
All it would do is make fancy links in sysfs I think, who cares ;-)
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines
Yeah, we're going to have a ton of them as well. Some of them are PCI devices and have a clear parent, others, not so much :/
In a number of places the only thing we have is the PMU driver, and we don't have a driver (or device) for the HW block it's a part of. Largely that's interconnect PMUs; we could create container devices there.
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines
Dont they have a PCI device? But yeah, some are going to be a wee bit challenging.
The system might not even have PCI, so it's arguable that they should just hang off an MMIO bus (which is effectively what the platofrm bus is).
You and your dodgy platforms :-)
For system PMUs we'll pretty much always have a platform device corresponding to a DT/ACPI entry used to describe MMIO registers and/or interrupts. In many cases the PMU is going to be the only part of the underlying device that is meaningful to Linux anyway, so I don't see any issue with just hanging the PMU device off its corresponding platform device - it still gives the user a way to map a PMU instance back to some understandable system topology (i.e. ACPI/DT) to disambiguate, and that's the most important thing.
Thanks,
Robin.
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Robin Murphy
maintainerRe: [PATCH 01/32] perf: Allow a PMU to have a parent
In reply to Jonathan CameronQuoted 17 lines · Jonathan Cameron
Quoted 11 lines · On Thu, 6 Apr 2023 14:40:40 +0200 Peter Zijlstra <[email protected]> wrote:
Quoted 4 lines · Jonathan Cameron
In the long run I agree it would be good. Short term there are more instances of struct pmu that don't have parents than those that do (even after this series). We need to figure out what to do about those before adding checks on it being set.
Right, I don't think you've touched *any* of the x86 PMUs for example, and getting everybody that boots an x86 kernel a warning isn't going to go over well :-)
It was tempting :) "Warning: Parentless PMU: try a different architecture."
Quoted 7 lines
I'd love some inputs on what the x86 PMU devices parents should be? CPU counters in general tend to just spin out of deep in the architecture code.
My overall favorite is an l2 cache related PMU that is spun up in
arch/arm/kernel/irq.cinit_IRQ()I'm just not going to try and figure out why...
I think that's simply because the PMU support was hung off the existing PL310 configuration code, which still supports non-DT boardfiles. The PMU shouldn't strictly need to be registered that early, it would just be a bunch more work to ensure that a platform device is available for it to bind to as a regular driver model driver, which wasn't justifiable at the time.
Thanks,
Robin.
PATCH 2/32 perf/hisi-pcie: Assign parent for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the PCI device.
--- drivers/perf/hisilicon/hisi_pcie_pmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/hisilicon/hisi_pcie_pmu.c +1 -0
@@ -793,6 +793,7 @@ static int hisi_pcie_alloc_pmu(struct pci_dev *pdev, struct hisi_pcie_pmu *pcie_ pcie_pmu->pmu = (struct pmu ) { .name = name, .module =THIS_MODULE , + .parent = &pdev->dev, .event_init = hisi_pcie_pmu_event_init, .pmu_enable = hisi_pcie_pmu_enable, .pmu_disable = hisi_pcie_pmu_disable,
Signature · 1 line
2.37.2
Re: [PATCH 02/32] perf/hisi-pcie: Assign parent for event_source device
Quoted 9 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the PCI device.Signed-off-by Jonathan Cameron <[email protected]>--- drivers/perf/hisilicon/hisi_pcie_pmu.c | 1 + 1 file changed, 1 insertion(+)
Re: [PATCH 02/32] perf/hisi-pcie: Assign parent for event_source device
In reply to Jonathan CameronQuoted 6 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the PCI device.
Quoted 17 lines
--- drivers/perf/hisilicon/hisi_pcie_pmu.c | 1 + 1 file changed, 1 insertion(+)drivers/perf/hisilicon/hisi_pcie_pmu.c +1
-0@@ -793,6 +793,7 @@ static int hisi_pcie_alloc_pmu(struct pci_dev *pdev, struct hisi_pcie_pmu *pcie_ pcie_pmu->pmu = (struct pmu ) { .name = name, .module =THIS_MODULE , + .parent = &pdev->dev, .event_init = hisi_pcie_pmu_event_init, .pmu_enable = hisi_pcie_pmu_enable, .pmu_disable = hisi_pcie_pmu_disable,
PATCH 3/32 Documentation: hisi-pmu: Drop reference to /sys/devices path
In reply to Jonathan CameronHaving assigned a parent to the device, the suggested path is no longer valid. As /sys/bus/event_sources based path is also provided, simply drop mention of alternative.
--- Documentation/admin-guide/perf/hisi-pmu.rst | 1 - 1 file changed, 1 deletion(-)
Documentation/admin-guide/perf/hisi-pmu.rst +0 -1
@@ -20,7 +20,6 @@ interrupt, and the PMU driver shall register perf PMU drivers like L3C, HHA and DDRC etc. The available events and configuration options shall be described in the sysfs, see: -/sys/devices/hisi_sccl{X}_<l3c{Y}/hha{Y}/ddrc{Y}>/, or /sys/bus/event_source/devices/hisi_sccl{X}_<l3c{Y}/hha{Y}/ddrc{Y}>. The "perf list" command shall list the available events from sysfs.
Signature · 1 line
2.37.2
Re: [PATCH 03/32] Documentation: hisi-pmu: Drop reference to /sys/devices path
Quoted 5 lines · Jonathan Cameron
Having assigned a parent to the device, the suggested path is no longer valid. As
/sys/bus/event_sourcesbased path is also provided, simply drop mention of alternative.Signed-off-by Jonathan Cameron <[email protected]>
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 03/32] Documentation: hisi-pmu: Drop reference to /sys/devices path
In reply to Jonathan CameronQuoted 5 lines · Jonathan Cameron
Having assigned a parent to the device, the suggested path is no longer valid. As
/sys/bus/event_sourcesbased path is also provided, simply drop mention of alternative.Signed-off-by Jonathan Cameron <[email protected]>
Quoted 17 lines
--- Documentation/admin-guide/perf/hisi-pmu.rst | 1 - 1 file changed, 1 deletion(-)Documentation/admin-guide/perf/hisi-pmu.rst +0
-1@@ -20,7 +20,6 @@ interrupt, and the PMU driver shall register perf PMU drivers like L3C, HHA and DDRC etc. The available events and configuration options shall be described in the sysfs, see: -/sys/devices/hisi_sccl{X}_<l3c{Y}/hha{Y}/ddrc{Y}>/, or /sys/bus/event_source/devices/hisi_sccl{X}_<l3c{Y}/hha{Y}/ddrc{Y}>. The "perf list" command shall list the available events from sysfs.
PATCH 4/32 perf/hisi-uncore: Assign parents for event_source devices
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/perf/hisilicon/hisi_uncore_pmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/hisilicon/hisi_uncore_pmu.c +1 -0
@@ -538,6 +538,7 @@ void hisi_pmu_init(struct hisi_pmu *hisi_pmu, const char *name, pmu->name = name; pmu->module = module; + pmu->parent = hisi_pmu->dev; pmu->task_ctx_nr = perf_invalid_context; pmu->event_init = hisi_uncore_pmu_event_init; pmu->pmu_enable = hisi_uncore_pmu_enable;
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 04/32] perf/hisi-uncore: Assign parents for event_source devices
Quoted 6 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the platform device.
Re: [PATCH 04/32] perf/hisi-uncore: Assign parents for event_source devices
In reply to Jonathan CameronQuoted 6 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the platform device.
Quoted 17 lines
--- drivers/perf/hisilicon/hisi_uncore_pmu.c | 1 + 1 file changed, 1 insertion(+)drivers/perf/hisilicon/hisi_uncore_pmu.c +1
-0@@ -538,6 +538,7 @@ void hisi_pmu_init(struct hisi_pmu *hisi_pmu, const char *name, pmu->name = name; pmu->module = module; + pmu->parent = hisi_pmu->dev; pmu->task_ctx_nr = perf_invalid_context; pmu->event_init = hisi_uncore_pmu_event_init; pmu->pmu_enable = hisi_uncore_pmu_enable;
PATCH 5/32 Documentation: hns-pmu: Use /sys/bus/event_source/devices paths
In reply to Jonathan CameronTo allow setting an appropriate parent for the struct pmu device remove existing references to /sys/devices/ path.
--- Documentation/admin-guide/perf/hns3-pmu.rst | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
Documentation/admin-guide/perf/hns3-pmu.rst +4 -4
@@ -16,7 +16,7 @@ HNS3 PMU driver The HNS3 PMU driver registers a perf PMU with the name of its sicl id.:: - /sys/devices/hns3_pmu_sicl_<sicl_id> + /sys/bus/event_source/devices/hns3_pmu_sicl_<sicl_id> PMU driver provides description of available events, filter modes, format, identifier and cpumask in sysfs.
@@ -40,9 +40,9 @@ device. Example usage of checking event code and subevent code:: - $# cat /sys/devices/hns3_pmu_sicl_0/events/dly_tx_normal_to_mac_time + $# cat /sys/bus/event_source/devices/hns3_pmu_sicl_0/events/dly_tx_normal_to_mac_time config=0x00204 - $# cat /sys/devices/hns3_pmu_sicl_0/events/dly_tx_normal_to_mac_packet_num + $# cat /sys/bus/event_source/devices/hns3_pmu_sicl_0/events/dly_tx_normal_to_mac_packet_num config=0x10204 Each performance statistic has a pair of events to get two values to
@@ -60,7 +60,7 @@ computation to calculate real performance data is::: Example usage of checking supported filter mode:: - $# cat /sys/devices/hns3_pmu_sicl_0/filtermode/bw_ssu_rpu_byte_num + $# cat /sys/bus/event_source/devices/hns3_pmu_sicl_0/filtermode/bw_ssu_rpu_byte_num filter mode supported: global/port/port-tc/func/func-queue/ Example usage of perf::
Signature · 1 line
2.37.2
Re: [PATCH 05/32] Documentation: hns-pmu: Use /sys/bus/event_source/devices paths
Quoted 4 lines · Jonathan Cameron
To allow setting an appropriate parent for the struct pmu device remove existing references to
/sys/devices/path.Signed-off-by Jonathan Cameron <[email protected]>
PATCH 6/32 perf/hisi-hns3: Assign parents for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the PCI device.
--- drivers/perf/hisilicon/hns3_pmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/hisilicon/hns3_pmu.c +1 -0
@@ -1419,6 +1419,7 @@ static int hns3_pmu_alloc_pmu(struct pci_dev *pdev, struct hns3_pmu *hns3_pmu) hns3_pmu->pmu = (struct pmu ) { .name = name, .module =THIS_MODULE , + .parent = &pdev->dev, .event_init = hns3_pmu_event_init, .pmu_enable = hns3_pmu_enable, .pmu_disable = hns3_pmu_disable,
Signature · 1 line
2.37.2
PATCH 7/32 perf/amlogic: Assign parents for event_source devices
In reply to Jonathan CameronCurrently all these devices appear directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parents to be the platform device.
--- drivers/perf/amlogic/meson_ddr_pmu_core.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/amlogic/meson_ddr_pmu_core.c +1 -0
@@ -490,6 +490,7 @@ int meson_ddr_pmu_create(struct platform_device *pdev) *pmu = (struct ddr_pmu ) { .pmu = { .module =THIS_MODULE , + .parent = &pdev->dev, .capabilities =PERF_PMU_CAP_NO_EXCLUDE , .task_ctx_nr = perf_invalid_context, .attr_groups = attr_groups,
Signature · 1 line
2.37.2
Re: [PATCH 07/32] perf/amlogic: Assign parents for event_source devices
Quoted 18 lines · Jonathan Cameron
[ EXTERNAL EMAIL ]
Currently all these devices appear directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparents to be the platform device.Cc Jiucheng Xu <[email protected]>Signed-off-by Jonathan Cameron <[email protected]>--- drivers/perf/amlogic/meson_ddr_pmu_core.c | 1 + 1 file changed, 1 insertion(+)drivers/perf/amlogic/meson_ddr_pmu_core.c +0
-0
drivers/perf/amlogic/meson_ddr_pmu_core.c int meson_ddr_pmu_create(struct platform_device *pdev) · line 493
Quoted 4 lines
drivers/perf/amlogic/meson_ddr_pmu_core.c +1
-0@@ -490,6 +490,7 @@ int meson_ddr_pmu_create(struct platform_device *pdev) *pmu = (struct ddr_pmu ) { .pmu = { .module =THIS_MODULE , + .parent = &pdev->dev,
Quoted 3 lines
drivers/perf/amlogic/meson_ddr_pmu_core.c +0
-0.capabilities = PERF_PMU_CAP_NO_EXCLUDE, .task_ctx_nr = perf_invalid_context, .attr_groups = attr_groups,
PATCH 8/32 perf/arm_cspmu: Assign parents for event_source devices
In reply to Jonathan CameronCurrently all these devices appear directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parents to be the platform device.
--- drivers/perf/arm_cspmu/arm_cspmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/arm_cspmu/arm_cspmu.c +1 -0
@@ -1155,6 +1155,7 @@ static int arm_cspmu_register_pmu(struct arm_cspmu *cspmu) cspmu->pmu = (struct pmu ){ .task_ctx_nr = perf_invalid_context, .module =THIS_MODULE , + .parent = cspmu->dev, .pmu_enable = arm_cspmu_enable, .pmu_disable = arm_cspmu_disable, .event_init = arm_cspmu_event_init,
Signature · 1 line
2.37.2
Re: [PATCH 08/32] perf/arm_cspmu: Assign parents for event_source devices
Quoted 15 lines · Jonathan Cameron
Currently all these devices appear directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparents to be the platform device.Signed-off-by Jonathan Cameron <[email protected]>--- drivers/perf/arm_cspmu/arm_cspmu.c | 1 + 1 file changed, 1 insertion(+)drivers/perf/arm_cspmu/arm_cspmu.c +0
-0
drivers/perf/arm_cspmu/arm_cspmu.c static int arm_cspmu_register_pmu(struct arm_cspmu *cspmu) · line 1161
Quoted 7 lines
drivers/perf/arm_cspmu/arm_cspmu.c +1
-0@@ -1155,6 +1155,7 @@ static int arm_cspmu_register_pmu(struct arm_cspmu *cspmu) cspmu->pmu = (struct pmu ){ .task_ctx_nr = perf_invalid_context, .module =THIS_MODULE , + .parent = cspmu->dev, .pmu_enable = arm_cspmu_enable, .pmu_disable = arm_cspmu_disable, .event_init = arm_cspmu_event_init,
PATCH 9/32 Documentation: xgene-pmu: Use /sys/bus/event_source/devices paths
In reply to Jonathan CameronTo allow setting an appropriate parent for the struct pmu device remove existing references to /sys/devices/ path.
--- Documentation/admin-guide/perf/xgene-pmu.rst | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Documentation/admin-guide/perf/xgene-pmu.rst +1 -1
@@ -13,7 +13,7 @@ PMU (perf) driver The xgene-pmu driver registers several perf PMU drivers. Each of the perf driver provides description of its available events and configuration options -in sysfs, see /sys/devices/<l3cX/iobX/mcbX/mcX>/. +in sysfs, see /sys/bus/event_source/devices/<l3cX/iobX/mcbX/mcX>/. The "format" directory describes format of the config (event ID), config1 (agent ID) fields of the perf_event_attr structure. The "events"
Signature · 1 line
2.37.2
PATCH 10/32 perf/xgene: Assign parents for event_source devices
In reply to Jonathan CameronCurrently all these devices appear directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parents to be the hardware related struct device.
--- drivers/perf/xgene_pmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/xgene_pmu.c +1 -0
@@ -1104,6 +1104,7 @@ static int xgene_init_perf(struct xgene_pmu_dev *pmu_dev, char *name)/* Perf driver registration */ pmu_dev->pmu = (struct pmu ) { + .parent = pmu_dev->parent->dev, .attr_groups = pmu_dev->attr_groups, .task_ctx_nr = perf_invalid_context, .pmu_enable = xgene_perf_pmu_enable,
Signature · 1 line
2.37.2
PATCH 11/32 Documentation: thunderx2-pmu: Use /sys/bus/event_source/devices paths
In reply to Jonathan CameronTo allow setting an appropriate parent for the struct pmu device remove existing references to /sys/devices/ path.
--- Documentation/admin-guide/perf/thunderx2-pmu.rst | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Documentation/admin-guide/perf/thunderx2-pmu.rst +1 -1
@@ -22,7 +22,7 @@ The thunderx2_pmu driver registers per-socket perf PMUs for the DMC and L3C devices. Each PMU can be used to count up to 4 (DMC/L3C) or up to 8 (CCPI2) events simultaneously. The PMUs provide a description of their available events and configuration options under sysfs, see -/sys/devices/uncore_<l3c_S/dmc_S/ccpi2_S/>; S is the socket id. +/sys/bus/event_source/devices/uncore_<l3c_S/dmc_S/ccpi2_S/>; S is the socket id. The driver does not support sampling, therefore "perf record" will not work. Per-task perf sessions are also not supported.
Signature · 1 line
2.37.2
PATCH 12/32 perf/thunderx2: Assign parents for event_source devices
In reply to Jonathan CameronCurrently all these devices appear directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parents to be the platform device.
--- drivers/perf/thunderx2_pmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/thunderx2_pmu.c +1 -0
@@ -729,6 +729,7 @@ static int tx2_uncore_pmu_register(/* Perf event registration */ tx2_pmu->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = tx2_pmu->dev, .attr_groups = tx2_pmu->attr_groups, .task_ctx_nr = perf_invalid_context, .event_init = tx2_uncore_event_init,
Signature · 1 line
2.37.2
PATCH 13/32 perf/riscv: Assign parents for event_source devices
In reply to Jonathan CameronCurrently all these devices appear directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parents to be the appropriate platform devices.
---
Untested.
This is minimal change, but could be made more elegant either by making one of the existing initialization functions take a struct device * or by embedding one in struct riscv_pmu
--- drivers/perf/riscv_pmu_legacy.c | 1 + drivers/perf/riscv_pmu_sbi.c | 1 + 2 files changed, 2 insertions(+)
drivers/perf/riscv_pmu_legacy.c +1 -0
@@ -102,6 +102,7 @@ static int pmu_legacy_device_probe(struct platform_device *pdev) pmu =riscv_pmu_alloc ();if (!pmu)return -ENOMEM ; + pmu->pmu.parent = &pdev->dev;pmu_legacy_init (pmu);return 0 ;
drivers/perf/riscv_pmu_sbi.c +1 -0
@@ -880,6 +880,7 @@ static int pmu_sbi_device_probe(struct platform_device *pdev) } pmu->pmu.attr_groups = riscv_pmu_attr_groups; + pmu->pmu.parent = &pdev->dev; pmu->cmask = cmask; pmu->ctr_start = pmu_sbi_ctr_start; pmu->ctr_stop = pmu_sbi_ctr_stop;
Signature · 1 line
2.37.2
PATCH 14/32 Documentation: qcom-pmu: Use /sys/bus/event_source/devices paths
In reply to Jonathan CameronTo allow setting an appropriate parent for the struct pmu device remove existing references to /sys/devices/ path.
--- Documentation/admin-guide/perf/qcom_l2_pmu.rst | 2 +- Documentation/admin-guide/perf/qcom_l3_pmu.rst | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
Documentation/admin-guide/perf/qcom_l2_pmu.rst +1 -1
@@ -10,7 +10,7 @@ There is one logical L2 PMU exposed, which aggregates the results from the physical PMUs. The driver provides a description of its available events and configuration -options in sysfs, see /sys/devices/l2cache_0. +options in sysfs, see /sys/bus/event_source/devices/l2cache_0. The "format" directory describes the format of the events.
Documentation/admin-guide/perf/qcom_l3_pmu.rst +1 -1
@@ -9,7 +9,7 @@ PMU with device name l3cache_<socket>_<instance>. User space is responsible for aggregating across slices. The driver provides a description of its available events and configuration -options in sysfs, see /sys/devices/l3cache*. Given that these are uncore PMUs +options in sysfs, see /sys/bus/event_source/devices/l3cache*. Given that these are uncore PMUs the driver also exposes a "cpumask" sysfs attribute which contains a mask consisting of one CPU per socket which will be used to handle all the PMU events on that socket.
Signature · 1 line
2.37.2
PATCH 15/32 perf/qcom: Assign parents for event_source devices
In reply to Jonathan CameronCurrently all these devices appear directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parents to be the platform devices.
--- drivers/perf/qcom_l2_pmu.c | 1 + drivers/perf/qcom_l3_pmu.c | 1 + 2 files changed, 2 insertions(+)
drivers/perf/qcom_l2_pmu.c +1 -0
@@ -904,6 +904,7 @@ static int l2_cache_pmu_probe(struct platform_device *pdev) l2cache_pmu->pmu = (struct pmu ) {/* suffix is instance id for future use with multiple sockets */ .name ="l2cache_0" , + .parent = &pdev->dev, .task_ctx_nr = perf_invalid_context, .pmu_enable = l2_cache_pmu_enable, .pmu_disable = l2_cache_pmu_disable,
drivers/perf/qcom_l3_pmu.c +1 -0
@@ -748,6 +748,7 @@ static int qcom_l3_cache_pmu_probe(struct platform_device *pdev)return -ENOMEM ; l3pmu->pmu = (struct pmu ) { + .parent = &pdev->dev, .task_ctx_nr = perf_invalid_context, .pmu_enable = qcom_l3_cache__pmu_enable,
Signature · 1 line
2.37.2
PATCH 16/32 perf/imx_ddr: Assign parents for event_source devices
In reply to Jonathan CameronCurrently all this device appear directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/perf/fsl_imx8_ddr_perf.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/fsl_imx8_ddr_perf.c +1 -0
@@ -592,6 +592,7 @@ static int ddr_perf_init(struct ddr_pmu *pmu, void __iomem *base, *pmu = (struct ddr_pmu ) { .pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = dev, .capabilities =PERF_PMU_CAP_NO_EXCLUDE , .task_ctx_nr = perf_invalid_context, .attr_groups = attr_groups,
Signature · 1 line
2.37.2
PATCH 17/32 perf/arm_pmu: Assign parents for event_source devices
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
---
Note that the ACPI path for this driver has no obvious 'parent' to use so I've left that for now. See armv8_pmu_driver_init() in arch/arm64/kernel/perf_event.c Suggestions welcome!
--- drivers/perf/arm_pmu_platform.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/arm_pmu_platform.c +1 -0
@@ -197,6 +197,7 @@ int arm_pmu_device_probe(struct platform_device *pdev,if (!pmu)return -ENOMEM ; + pmu->pmu.parent = &pdev->dev; pmu->plat_device = pdev; ret =pmu_parse_irqs (pmu);
Signature · 1 line
2.37.2
Quoted 10 lines · Jonathan Cameron
These are the low hanging fruit following GregKH's feedback that all the devices registered via
perf_pmu_register()should have parents.Note that this causes potential ABI breakage.
It may fall in the category of it isn't breakage if no one notices but I can't be certain of that. Whilst it is arguable that no one should be been accessing PMUs except via the
event_sourcebus, there was documentation suggesting/sys/devices/for particular PMUs (because it was a shorter path?)
devices can always move around /sys/devices/ as there is not a guarantee that they will ever be in the same place. That's what /sys/class/ is used to find (and /sys/bus/ in some cases.)
And even then, the naming scheme is variable, and can and will change (i.e. bus ids), so that too is not required to stay the same.
thanks for doing this work, I'll add it to my review queue...
greg k-h
PATCH 18/32 perf/alibaba_uncore: Assign parents for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/perf/alibaba_uncore_drw_pmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/alibaba_uncore_drw_pmu.c +1 -0
@@ -683,6 +683,7 @@ static int ali_drw_pmu_probe(struct platform_device *pdev) drw_pmu->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &pdev->dev, .task_ctx_nr = perf_invalid_context, .event_init = ali_drw_pmu_event_init, .add = ali_drw_pmu_add,
Signature · 1 line
2.37.2
Re: [PATCH 18/32] perf/alibaba_uncore: Assign parents for event_source device
Quoted 16 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the platform device.Cc Shuai Xue <[email protected]>Signed-off-by Jonathan Cameron <[email protected]>--- drivers/perf/alibaba_uncore_drw_pmu.c | 1 + 1 file changed, 1 insertion(+)drivers/perf/alibaba_uncore_drw_pmu.c +0
-0
drivers/perf/alibaba_uncore_drw_pmu.c static int ali_drw_pmu_probe(struct platform_device *pdev) · line 689
Quoted 7 lines
drivers/perf/alibaba_uncore_drw_pmu.c +1
-0@@ -683,6 +683,7 @@ static int ali_drw_pmu_probe(struct platform_device *pdev) drw_pmu->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &pdev->dev, .task_ctx_nr = perf_invalid_context, .event_init = ali_drw_pmu_event_init, .add = ali_drw_pmu_add,
Thank you.
Best Regards,
Shuai
PATCH 19/32 perf/arm-cci: Assign parents for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/perf/arm-cci.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/arm-cci.c +1 -0
@@ -1412,6 +1412,7 @@ static int cci_pmu_init(struct cci_pmu *cci_pmu, struct platform_device *pdev) cci_pmu->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &pdev->dev, .name = cci_pmu->model->name, .task_ctx_nr = perf_invalid_context, .pmu_enable = cci_pmu_enable,
Signature · 1 line
2.37.2
Re: [PATCH 19/32] perf/arm-cci: Assign parents for event_source device
Quoted 6 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the platform device.
Quoted 16 lines
--- drivers/perf/arm-cci.c | 1 + 1 file changed, 1 insertion(+)drivers/perf/arm-cci.c +1
-0@@ -1412,6 +1412,7 @@ static int cci_pmu_init(struct cci_pmu *cci_pmu, struct platform_device *pdev) cci_pmu->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &pdev->dev, .name = cci_pmu->model->name, .task_ctx_nr = perf_invalid_context, .pmu_enable = cci_pmu_enable,
PATCH 20/32 perf/arm-ccn: Assign parents for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/perf/arm-ccn.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/arm-ccn.c +1 -0
@@ -1265,6 +1265,7 @@ static int arm_ccn_pmu_init(struct arm_ccn *ccn)/* Perf driver registration */ ccn->dt.pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = ccn->dev, .attr_groups = arm_ccn_pmu_attr_groups, .task_ctx_nr = perf_invalid_context, .event_init = arm_ccn_pmu_event_init,
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 20/32] perf/arm-ccn: Assign parents for event_source device
Quoted 15 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the platform device.Signed-off-by Jonathan Cameron <[email protected]>--- drivers/perf/arm-ccn.c | 1 + 1 file changed, 1 insertion(+)drivers/perf/arm-ccn.c +0
-0
drivers/perf/arm-ccn.c static int arm_ccn_pmu_init(struct arm_ccn *ccn) · line 1271
Quoted 7 lines
drivers/perf/arm-ccn.c +1
-0@@ -1265,6 +1265,7 @@ static int arm_ccn_pmu_init(struct arm_ccn *ccn)/* Perf driver registration */ ccn->dt.pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = ccn->dev, .attr_groups = arm_ccn_pmu_attr_groups, .task_ctx_nr = perf_invalid_context, .event_init = arm_ccn_pmu_event_init,
PATCH 21/32 perf/arm-cmn: Assign parents for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/perf/arm-cmn.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/arm-cmn.c +1 -0
@@ -2284,6 +2284,7 @@ static int arm_cmn_probe(struct platform_device *pdev) cmn->cpu =cpumask_local_spread (0 ,dev_to_node (cmn->dev)); cmn->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &pdev->dev, .attr_groups = arm_cmn_attr_groups, .capabilities =PERF_PMU_CAP_NO_EXCLUDE , .task_ctx_nr = perf_invalid_context,
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 21/32] perf/arm-cmn: Assign parents for event_source device
Quoted 3 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the platform device.
Oh, fab! Just the other week I wrote up a patch doing this after the fact with device_move() on the grounds of making the PMU instances easier to identify (and with an equivalent cleanup of Documentation/admin-guide/perf), which I was close to getting round to sending as an RFC. Thus I thoroughly approve :D
Quoted 18 lines
Signed-off-by Jonathan Cameron <[email protected]>--- drivers/perf/arm-cmn.c | 1 + 1 file changed, 1 insertion(+)drivers/perf/arm-cmn.c +1
-0@@ -2284,6 +2284,7 @@ static int arm_cmn_probe(struct platform_device *pdev) cmn->cpu =cpumask_local_spread (0 ,dev_to_node (cmn->dev)); cmn->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &pdev->dev, .attr_groups = arm_cmn_attr_groups, .capabilities =PERF_PMU_CAP_NO_EXCLUDE , .task_ctx_nr = perf_invalid_context,
PATCH 22/32 perf/arm-dmc620: Assign parents for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/perf/arm_dmc620_pmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/arm_dmc620_pmu.c +1 -0
@@ -644,6 +644,7 @@ static int dmc620_pmu_device_probe(struct platform_device *pdev) dmc620_pmu->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &pdev->dev, .capabilities =PERF_PMU_CAP_NO_EXCLUDE , .task_ctx_nr = perf_invalid_context, .event_init = dmc620_pmu_event_init,
Signature · 1 line
2.37.2
PATCH 23/32 perf/arm-dsu: Assign parents for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/perf/arm_dsu_pmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/arm_dsu_pmu.c +1 -0
@@ -751,6 +751,7 @@ static int dsu_pmu_device_probe(struct platform_device *pdev) dsu_pmu->pmu = (struct pmu ) { .task_ctx_nr = perf_invalid_context, + .parent = &pdev->dev, .module =THIS_MODULE , .pmu_enable = dsu_pmu_enable, .pmu_disable = dsu_pmu_disable,
Signature · 1 line
2.37.2
Re: [PATCH 23/32] perf/arm-dsu: Assign parents for event_source device
Quoted 15 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the platform device.Signed-off-by Jonathan Cameron <[email protected]>--- drivers/perf/arm_dsu_pmu.c | 1 + 1 file changed, 1 insertion(+)drivers/perf/arm_dsu_pmu.c +0
-0
drivers/perf/arm_dsu_pmu.c static int dsu_pmu_device_probe(struct platform_device *pdev) · line 757
Quoted 7 lines
drivers/perf/arm_dsu_pmu.c +1
-0@@ -751,6 +751,7 @@ static int dsu_pmu_device_probe(struct platform_device *pdev) dsu_pmu->pmu = (struct pmu ) { .task_ctx_nr = perf_invalid_context, + .parent = &pdev->dev, .module =THIS_MODULE , .pmu_enable = dsu_pmu_enable, .pmu_disable = dsu_pmu_disable,
PATCH 24/32 perf/arm-smmuv3: Assign parents for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/perf/arm_smmuv3_pmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/arm_smmuv3_pmu.c +1 -0
@@ -826,6 +826,7 @@ static int smmu_pmu_probe(struct platform_device *pdev) smmu_pmu->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &pdev->dev, .task_ctx_nr = perf_invalid_context, .pmu_enable = smmu_pmu_enable, .pmu_disable = smmu_pmu_disable,
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
PATCH 25/32 perf/arm-spe: Assign parents for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/perf/arm_spe_pmu.c | 1 + 1 file changed, 1 insertion(+)
drivers/perf/arm_spe_pmu.c +1 -0
@@ -955,6 +955,7 @@ static int arm_spe_pmu_perf_init(struct arm_spe_pmu *spe_pmu) spe_pmu->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &spe_pmu->pdev->dev, .capabilities =PERF_PMU_CAP_EXCLUSIVE |PERF_PMU_CAP_ITRACE , .attr_groups = arm_spe_pmu_attr_groups,/*
Signature · 1 line
2.37.2
Re: [PATCH 25/32] perf/arm-spe: Assign parents for event_source device
Quoted 15 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the platform device.Signed-off-by Jonathan Cameron <[email protected]>--- drivers/perf/arm_spe_pmu.c | 1 + 1 file changed, 1 insertion(+)drivers/perf/arm_spe_pmu.c +0
-0
drivers/perf/arm_spe_pmu.c static int arm_spe_pmu_perf_init(struct arm_spe_pmu *spe_pmu) · line 961
Quoted 7 lines
drivers/perf/arm_spe_pmu.c +1
-0@@ -955,6 +955,7 @@ static int arm_spe_pmu_perf_init(struct arm_spe_pmu *spe_pmu) spe_pmu->pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &spe_pmu->pdev->dev, .capabilities =PERF_PMU_CAP_EXCLUSIVE |PERF_PMU_CAP_ITRACE , .attr_groups = arm_spe_pmu_attr_groups,/*
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
PATCH 26/32 arc: Assign parents for event_source devices
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- arch/arc/kernel/perf_event.c | 1 + 1 file changed, 1 insertion(+)
arch/arc/kernel/perf_event.c +1 -0
@@ -788,6 +788,7 @@ static int arc_pmu_device_probe(struct platform_device *pdev) arc_pmu->attr_groups[ARCPMU_ATTR_GR_FORMATS ] = &arc_pmu_format_attr_gr; arc_pmu->pmu = (struct pmu ) { + .parent = &pdev->dev, .pmu_enable = arc_pmu_enable, .pmu_disable = arc_pmu_disable, .event_init = arc_pmu_event_init,
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
PATCH 27/32 ARM: imx: Assign parents for mmdc event_source devices
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- arch/arm/mach-imx/mmdc.c | 1 + 1 file changed, 1 insertion(+)
arch/arm/mach-imx/mmdc.c +1 -0
@@ -436,6 +436,7 @@ static int mmdc_pmu_init(struct mmdc_pmu *pmu_mmdc, { *pmu_mmdc = (struct mmdc_pmu ) { .pmu = (struct pmu ) { + .parent = dev, .task_ctx_nr = perf_invalid_context, .attr_groups = attr_groups, .event_init = mmdc_pmu_event_init,
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
PATCH 28/32 dmaengine: idxd: Assign parent for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the iDXD PCI Device.
--- drivers/dma/idxd/perfmon.c | 1 + 1 file changed, 1 insertion(+)
drivers/dma/idxd/perfmon.c +1 -0
@@ -478,6 +478,7 @@ static void idxd_pmu_init(struct idxd_pmu *idxd_pmu) } idxd_pmu->pmu.name = idxd_pmu->name; + idxd_pmu->pmu.parent = &idxd_pmu->idxd->pdev->dev; idxd_pmu->pmu.attr_groups = perfmon_attr_groups; idxd_pmu->pmu.task_ctx_nr = perf_invalid_context; idxd_pmu->pmu.event_init = perfmon_pmu_event_init;
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
PATCH 29/32 fpga: dfl: Assign parent for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the Platform device.
--- drivers/fpga/dfl-fme-perf.c | 1 + 1 file changed, 1 insertion(+)
drivers/fpga/dfl-fme-perf.c +1 -0
@@ -912,6 +912,7 @@ static int fme_perf_pmu_register(struct platform_device *pdev,fme_perf_setup_hardware (priv); + pmu->parent = &pdev->dev; pmu->task_ctx_nr = perf_invalid_context; pmu->attr_groups = fme_perf_groups; pmu->attr_update = fme_perf_events_groups;
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 29/32] fpga: dfl: Assign parent for event_source device
Quoted 9 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the Platform device.Cc Wu Hao <[email protected]>Cc Tom Rix <[email protected]>Signed-off-by Jonathan Cameron <[email protected]>
Quoted 19 lines
--- drivers/fpga/dfl-fme-perf.c | 1 + 1 file changed, 1 insertion(+)drivers/fpga/dfl-fme-perf.c +1
-0@@ -912,6 +912,7 @@ static int fme_perf_pmu_register(struct platform_device *pdev,fme_perf_setup_hardware (priv); + pmu->parent = &pdev->dev; pmu->task_ctx_nr = perf_invalid_context; pmu->attr_groups = fme_perf_groups; pmu->attr_update = fme_perf_events_groups;Signature · 1 line
2.37.2
PATCH 30/32 drivers/nvdimm: Assign parent for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the platform device.
--- drivers/nvdimm/nd_perf.c | 1 + 1 file changed, 1 insertion(+)
drivers/nvdimm/nd_perf.c +1 -0
@@ -292,6 +292,7 @@ int register_nvdimm_pmu(struct nvdimm_pmu *nd_pmu, struct platform_device *pdev)* device data in events functions. */ nd_pmu->dev = &pdev->dev; + nd_pmu->pmu.parent = &pdev->dev;/* Fill attribute groups for the nvdimm pmu device */ nd_pmu->pmu.attr_groups[NVDIMM_PMU_FORMAT_ATTR ] = &nvdimm_pmu_format_group;
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
RE: [PATCH 30/32] drivers/nvdimm: Assign parent for event_source device
Quoted 8 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the platform device.Cc Dan Williams <[email protected]>Signed-off-by Jonathan Cameron <[email protected]>
PATCH 31/32 Documentation: ABI + trace: hisi_ptt: update paths to bus/event_source
In reply to Jonathan CameronTo allow for assigning a suitable parent to the struct pmu device update the documentation to describe the device via the event_source bus where it will remain accessible.
For the ABI documention file also rename the file as it is named after the path.
--- ...i_ptt => sysfs-bus-event_source-devices-hisi_ptt} | 12 ++++++------ Documentation/trace/hisi-ptt.rst | 4 ++-- MAINTAINERS | 2 +- 3 files changed, 9 insertions(+), 9 deletions(-)
Documentation/ABI/testing/sysfs-bus-event_source-devices-hisi_ptt renamed from Documentation/ABI/testing/sysfs-devices-hisi_ptt +6 -6
@@ -1,4 +1,4 @@ -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>
@@ -8,7 +8,7 @@ Description: This directory contains files for tuning the PCIe link See Documentation/trace/hisi-ptt.rst for more information. -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_cpl +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_cpl Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>
@@ -18,7 +18,7 @@ Description: (RW) Controls the weight of Tx completion TLPs, which influence will return an error, and out of range values will be converted to 2. The value indicates a probable level of the event. -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_np +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_np Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>
@@ -28,7 +28,7 @@ Description: (RW) Controls the weight of Tx non-posted TLPs, which influence will return an error, and out of range values will be converted to 2. The value indicates a probable level of the event. -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_p +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_p Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>
@@ -38,7 +38,7 @@ Description: (RW) Controls the weight of Tx posted TLPs, which influence the will return an error, and out of range values will be converted to 2. The value indicates a probable level of the event. -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune/rx_alloc_buf_level +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune/rx_alloc_buf_level Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>
@@ -49,7 +49,7 @@ Description: (RW) Control the allocated buffer watermark for inbound packets. will return an error, and out of range values will be converted to 2. The value indicates a probable level of the event. -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune/tx_alloc_buf_level +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune/tx_alloc_buf_level Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>
Documentation/trace/hisi-ptt.rst +2 -2
@@ -40,7 +40,7 @@ IO dies (SICL, Super I/O Cluster), where there's one PCIe Root Complex for each SICL. :: - /sys/devices/hisi_ptt<sicl_id>_<core_id> + /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id> Tune ====
@@ -53,7 +53,7 @@ Each event is presented as a file under $(PTT PMU dir)/tune, and a simple open/read/write/close cycle will be used to tune the event. :: - $ cd /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune + $ cd /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune $ ls qos_tx_cpl qos_tx_np qos_tx_p tx_path_rx_req_alloc_buf_level
MAINTAINERS +1 -1
@@ -9266,7 +9266,7 @@ M: Yicong Yang <[email protected]> M: Jonathan Cameron <[email protected]> L: [email protected] S: Maintained -F: Documentation/ABI/testing/sysfs-devices-hisi_ptt +F: Documentation/ABI/testing/sysfs-bus-event_source-devices-hisi_ptt F: Documentation/trace/hisi-ptt.rst F: drivers/hwtracing/ptt/ F: tools/perf/arch/arm64/util/hisi-ptt.c
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 31/32] Documentation: ABI + trace: hisi_ptt: update paths to bus/event_source
Quoted 8 lines · Jonathan Cameron
To allow for assigning a suitable parent to the struct pmu device update the documentation to describe the device via the
event_sourcebus where it will remain accessible.For the ABI documention file also rename the file as it is named after the path.
Signed-off-by Jonathan Cameron <[email protected]>
Quoted 100 lines
--- ...i_ptt => sysfs-bus-event_source-devices-hisi_ptt} | 12 ++++++------ Documentation/trace/hisi-ptt.rst | 4 ++-- MAINTAINERS | 2 +- 3 files changed, 9 insertions(+), 9 deletions(-)Documentation/ABI/testing/sysfs-bus-event_source-devices-hisi_ptt renamed from Documentation/ABI/testing/sysfs-devices-hisi_ptt +6
-6@@ -1,4 +1,4 @@ -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>@@ -8,7 +8,7 @@ Description: This directory contains files for tuning the PCIe link See Documentation/trace/hisi-ptt.rst for more information. -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_cpl +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_cpl Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>@@ -18,7 +18,7 @@ Description: (RW) Controls the weight of Tx completion TLPs, which influence will return an error, and out of range values will be converted to 2. The value indicates a probable level of the event. -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_np +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_np Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>@@ -28,7 +28,7 @@ Description: (RW) Controls the weight of Tx non-posted TLPs, which influence will return an error, and out of range values will be converted to 2. The value indicates a probable level of the event. -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_p +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune/qos_tx_p Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>@@ -38,7 +38,7 @@ Description: (RW) Controls the weight of Tx posted TLPs, which influence the will return an error, and out of range values will be converted to 2. The value indicates a probable level of the event. -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune/rx_alloc_buf_level +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune/rx_alloc_buf_level Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>@@ -49,7 +49,7 @@ Description: (RW) Control the allocated buffer watermark for inbound packets. will return an error, and out of range values will be converted to 2. The value indicates a probable level of the event. -What: /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune/tx_alloc_buf_level +What: /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune/tx_alloc_buf_level Date: October 2022 KernelVersion: 6.1 Contact: Yicong Yang <[email protected]>Documentation/trace/hisi-ptt.rst +2
-2@@ -40,7 +40,7 @@ IO dies (SICL, Super I/O Cluster), where there's one PCIe Root Complex for each SICL. :: - /sys/devices/hisi_ptt<sicl_id>_<core_id> + /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id> Tune ====@@ -53,7 +53,7 @@ Each event is presented as a file under $(PTT PMU dir)/tune, and a simple open/read/write/close cycle will be used to tune the event. :: - $ cd /sys/devices/hisi_ptt<sicl_id>_<core_id>/tune + $ cd /sys/bus/event_source/devices/hisi_ptt<sicl_id>_<core_id>/tune $ ls qos_tx_cpl qos_tx_np qos_tx_p tx_path_rx_req_alloc_buf_levelMAINTAINERS +1
-1@@ -9266,7 +9266,7 @@ M: Yicong Yang <[email protected]> M: Jonathan Cameron <[email protected]> L: [email protected] S: Maintained -F: Documentation/ABI/testing/sysfs-devices-hisi_ptt +F: Documentation/ABI/testing/sysfs-bus-event_source-devices-hisi_ptt F: Documentation/trace/hisi-ptt.rst F: drivers/hwtracing/ptt/ F: tools/perf/arch/arm64/util/hisi-ptt.c
PATCH 32/32 hwtracing: hisi_ptt: Assign parent for event_source device
In reply to Jonathan CameronCurrently the PMU device appears directly under /sys/devices/ Only root busses should appear there, so instead assign the pmu->dev parent to be the PCI device.
--- drivers/hwtracing/ptt/hisi_ptt.c | 1 + 1 file changed, 1 insertion(+)
drivers/hwtracing/ptt/hisi_ptt.c +1 -0
@@ -871,6 +871,7 @@ static int hisi_ptt_register_pmu(struct hisi_ptt *hisi_ptt) hisi_ptt->hisi_ptt_pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &hisi_ptt->pdev->dev, .capabilities =PERF_PMU_CAP_EXCLUSIVE |PERF_PMU_CAP_ITRACE , .task_ctx_nr = perf_sw_context, .attr_groups = hisi_ptt_pmu_groups,
Signature · 7 lines
2.37.2 _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Re: [PATCH 32/32] hwtracing: hisi_ptt: Assign parent for event_source device
Quoted 8 lines · Jonathan Cameron
Currently the PMU device appears directly under
/sys/devices/Only root busses should appear there, so instead assign thepmu->devparent to be the PCI device.Cc Yicong Yang <[email protected]>Cc Suzuki K Poulose <[email protected]>Signed-off-by Jonathan Cameron <[email protected]>
Quoted 17 lines
--- drivers/hwtracing/ptt/hisi_ptt.c | 1 + 1 file changed, 1 insertion(+)drivers/hwtracing/ptt/hisi_ptt.c +1
-0@@ -871,6 +871,7 @@ static int hisi_ptt_register_pmu(struct hisi_ptt *hisi_ptt) hisi_ptt->hisi_ptt_pmu = (struct pmu ) { .module =THIS_MODULE , + .parent = &hisi_ptt->pdev->dev, .capabilities =PERF_PMU_CAP_EXCLUSIVE |PERF_PMU_CAP_ITRACE , .task_ctx_nr = perf_sw_context, .attr_groups = hisi_ptt_pmu_groups,
Quoted 20 lines · Jonathan Cameron
These are the low hanging fruit following GregKH's feedback that all the devices registered via
perf_pmu_register()should have parents.Note that this causes potential ABI breakage.
It may fall in the category of it isn't breakage if no one notices but I can't be certain of that. Whilst it is arguable that no one should be been accessing PMUs except via the
event_sourcebus, there was documentation suggesting/sys/devices/for particular PMUs (because it was a shorter path?)The first patch is pulled out of the series:
https://lore.kernel.org/linux-cxl/[email protected]/
[PATCH v3 0/5] CXL 3.0 Performance Monitoring Unit supportIn that particular case it is very useful to be able to figure out which CXL device the PMU device is associated with and looking at it's parents in the device model as shown with ls -lh
/sys/bus/event_sources/devices/is a very easy way to do this (once it is correctly parented).
Quoted 8 lines
Addressing all the other instances of struct pmu not covered by this series is likely to be a more complex discussion but unlikely to have an affect on what is proposed here.
Documentation updates deliberately 'fixed' in separate patches before changing the path to highlight that using
/sys/bus/event_source/devicespath is unchanged by this series and that is presumed to be the most common way these files are accessed.
For the whole series, looks good: