2019-05-06 21:30:11

by Len Brown

[permalink] [raw]
Subject: [PATCH 16/22] perf/x86/intel/uncore: Support multi-die/package

From: Kan Liang <[email protected]>

Uncore becomes die-scope on Xeon Cascade Lake-AP. Uncore driver needs to
support die-scope uncore units.

Use topology_logical_die_id() to replace topology_logical_package_id().
For previous platforms which doesn't have multi-die,
topology_logical_die_id() is identical as topology_logical_package_id().

In pci_probe/remove, the group id reads from PCI BUS is logical die id
for multi-die systems.

Use topology_die_cpumask() to replace topology_core_cpumask().
For previous platforms which doesn't have multi-die,
topology_die_cpumask() is identical as topology_core_cpumask().

There is no functional change for previous platforms.

Signed-off-by: Kan Liang <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Len Brown <[email protected]>
---
arch/x86/events/intel/uncore.c | 20 +++++++++++---------
1 file changed, 11 insertions(+), 9 deletions(-)

diff --git a/arch/x86/events/intel/uncore.c b/arch/x86/events/intel/uncore.c
index 9fe64c01a2e5..a6461ff85a32 100644
--- a/arch/x86/events/intel/uncore.c
+++ b/arch/x86/events/intel/uncore.c
@@ -100,7 +100,7 @@ ssize_t uncore_event_show(struct kobject *kobj,

struct intel_uncore_box *uncore_pmu_to_box(struct intel_uncore_pmu *pmu, int cpu)
{
- unsigned int pkgid = topology_logical_package_id(cpu);
+ unsigned int pkgid = topology_logical_die_id(cpu);

/*
* The unsigned check also catches the '-1' return value for non
@@ -942,7 +942,8 @@ static int uncore_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id
if (phys_id < 0)
return -ENODEV;

- pkg = topology_phys_to_logical_pkg(phys_id);
+ pkg = (topology_max_die_per_package() > 1) ? phys_id :
+ topology_phys_to_logical_pkg(phys_id);
if (pkg < 0)
return -EINVAL;

@@ -1033,7 +1034,8 @@ static void uncore_pci_remove(struct pci_dev *pdev)

box = pci_get_drvdata(pdev);
if (!box) {
- pkg = topology_phys_to_logical_pkg(phys_id);
+ pkg = (topology_max_die_per_package() > 1) ? phys_id :
+ topology_phys_to_logical_pkg(phys_id);
for (i = 0; i < UNCORE_EXTRA_PCI_DEV_MAX; i++) {
if (uncore_extra_pci_dev[pkg].dev[i] == pdev) {
uncore_extra_pci_dev[pkg].dev[i] = NULL;
@@ -1110,7 +1112,7 @@ static void uncore_change_type_ctx(struct intel_uncore_type *type, int old_cpu,
struct intel_uncore_box *box;
int i, pkg;

- pkg = topology_logical_package_id(old_cpu < 0 ? new_cpu : old_cpu);
+ pkg = topology_logical_die_id(old_cpu < 0 ? new_cpu : old_cpu);
for (i = 0; i < type->num_boxes; i++, pmu++) {
box = pmu->boxes[pkg];
if (!box)
@@ -1151,7 +1153,7 @@ static int uncore_event_cpu_offline(unsigned int cpu)
if (!cpumask_test_and_clear_cpu(cpu, &uncore_cpu_mask))
goto unref;
/* Find a new cpu to collect uncore events */
- target = cpumask_any_but(topology_core_cpumask(cpu), cpu);
+ target = cpumask_any_but(topology_die_cpumask(cpu), cpu);

/* Migrate uncore events to the new target */
if (target < nr_cpu_ids)
@@ -1164,7 +1166,7 @@ static int uncore_event_cpu_offline(unsigned int cpu)

unref:
/* Clear the references */
- pkg = topology_logical_package_id(cpu);
+ pkg = topology_logical_die_id(cpu);
for (; *types; types++) {
type = *types;
pmu = type->pmus;
@@ -1223,7 +1225,7 @@ static int uncore_event_cpu_online(unsigned int cpu)
struct intel_uncore_box *box;
int i, ret, pkg, target;

- pkg = topology_logical_package_id(cpu);
+ pkg = topology_logical_die_id(cpu);
ret = allocate_boxes(types, pkg, cpu);
if (ret)
return ret;
@@ -1242,7 +1244,7 @@ static int uncore_event_cpu_online(unsigned int cpu)
* Check if there is an online cpu in the package
* which collects uncore events already.
*/
- target = cpumask_any_and(&uncore_cpu_mask, topology_core_cpumask(cpu));
+ target = cpumask_any_and(&uncore_cpu_mask, topology_die_cpumask(cpu));
if (target < nr_cpu_ids)
return 0;

@@ -1411,7 +1413,7 @@ static int __init intel_uncore_init(void)
if (boot_cpu_has(X86_FEATURE_HYPERVISOR))
return -ENODEV;

- max_packages = topology_max_packages();
+ max_packages = topology_max_packages() * topology_max_die_per_package();

uncore_init = (struct intel_uncore_init_fun *)id->driver_data;
if (uncore_init->pci_init) {
--
2.18.0-rc0


2019-05-07 12:22:58

by Peter Zijlstra

[permalink] [raw]
Subject: Re: [PATCH 16/22] perf/x86/intel/uncore: Support multi-die/package

On Mon, May 06, 2019 at 05:26:11PM -0400, Len Brown wrote:
> @@ -1411,7 +1413,7 @@ static int __init intel_uncore_init(void)
> if (boot_cpu_has(X86_FEATURE_HYPERVISOR))
> return -ENODEV;
>
> - max_packages = topology_max_packages();
> + max_packages = topology_max_packages() * topology_max_die_per_package();

That's max_dies..

> uncore_init = (struct intel_uncore_init_fun *)id->driver_data;
> if (uncore_init->pci_init) {
> --
> 2.18.0-rc0
>

2019-05-07 12:24:09

by Peter Zijlstra

[permalink] [raw]
Subject: Re: [PATCH 16/22] perf/x86/intel/uncore: Support multi-die/package

On Mon, May 06, 2019 at 05:26:11PM -0400, Len Brown wrote:
> @@ -1223,7 +1225,7 @@ static int uncore_event_cpu_online(unsigned int cpu)
> struct intel_uncore_box *box;
> int i, ret, pkg, target;
>
> - pkg = topology_logical_package_id(cpu);
> + pkg = topology_logical_die_id(cpu);
> ret = allocate_boxes(types, pkg, cpu);
> if (ret)
> return ret;

'pkg' != 'die'

2019-05-08 20:02:40

by Len Brown

[permalink] [raw]
Subject: Re: [PATCH 16/22] perf/x86/intel/uncore: Support multi-die/package

On Tue, May 7, 2019 at 8:21 AM Peter Zijlstra <[email protected]> wrote:
>
> On Mon, May 06, 2019 at 05:26:11PM -0400, Len Brown wrote:
> > @@ -1411,7 +1413,7 @@ static int __init intel_uncore_init(void)
> > if (boot_cpu_has(X86_FEATURE_HYPERVISOR))
> > return -ENODEV;
> >
> > - max_packages = topology_max_packages();
> > + max_packages = topology_max_packages() * topology_max_die_per_package();
>
> That's max_dies..

To keep the patch with this functional change minimal and obvious,
the rename of this variable was segregated into patch 22, which is
re-names only.

Len Brown, Intel Open Source Technology Center

2019-05-08 20:02:40

by Len Brown

[permalink] [raw]
Subject: Re: [PATCH 16/22] perf/x86/intel/uncore: Support multi-die/package

On Tue, May 7, 2019 at 8:22 AM Peter Zijlstra <[email protected]> wrote:
>
> On Mon, May 06, 2019 at 05:26:11PM -0400, Len Brown wrote:
> > @@ -1223,7 +1225,7 @@ static int uncore_event_cpu_online(unsigned int cpu)
> > struct intel_uncore_box *box;
> > int i, ret, pkg, target;
> >
> > - pkg = topology_logical_package_id(cpu);
> > + pkg = topology_logical_die_id(cpu);
> > ret = allocate_boxes(types, pkg, cpu);
> > if (ret)
> > return ret;
>
> 'pkg' != 'die'

To keep the patch with this functional change minimal and obvious,
the rename of this variable was segregated into patch 22, which is
re-names only.

Len Brown, Intel Open Source Technology Center