perf/x86/intel/uncore: Fix the index of PCU.3 Broadwell CPUs
Subsystems (2)
PATCH RESEND 1/1 perf/x86/intel/uncore: Fix the index of PCU.3 Broadwell CPUs
From: Masayoshi Mizuma <[email protected]>
commit 15a3e845b01c ("perf/x86/intel/uncore: Fix SBOX support for Broadwell CPUs") introduced PCU.3 for Broadwell CPU. Unfortunately, the driver_data of PCU.3 conflicts to QPI Port 2 filter.
{ /* QPI Port 2 filter */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f46), .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 2),
{ /* PCU.3 (for Capability registers) */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6fc0), .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV,
HSWEP_PCI_PCU_3),
// HSWEP_PCI_PCU_3 == 2
As the result, the index for each device gets same, so the pci device is overwrited in uncore_extra_pci_dev[pkg].dev[idx] when it is probed.
static int uncore_pci_probe(struct pci_dev *pdev,... {
...
if (UNCORE_PCI_DEV_TYPE(id->driver_data) == UNCORE_EXTRA_PCI_DEV) {
int idx = UNCORE_PCI_DEV_IDX(id->driver_data); // HERE!!
uncore_extra_pci_dev[pkg].dev[idx] = pdev; // HERE!!
pci_set_drvdata(pdev, NULL);
return 0;
}Due to the overwriting, the following warning message are shown while CPU hot-removing.
WARNING: CPU: 126 PID: 6 at arch/x86/events/intel/uncore.c:988 uncore_pci_remove+0x10b/0x150 Call Trace: pci_device_remove+0x42/0xd0 device_release_driver_internal+0x148/0x220 pci_stop_bus_device+0x76/0xa0 pci_stop_root_bus+0x44/0x60 acpi_pci_root_remove+0x1f/0x80 acpi_bus_trim+0x57/0x90 acpi_bus_trim+0x2e/0x90 acpi_device_hotplug+0x2bc/0x4b0 acpi_hotplug_work_fn+0x1a/0x30 process_one_work+0x174/0x3a0 worker_thread+0x4c/0x3d0 kthread+0xf8/0x130
To avoid the conflict, this patch changes the PCU.3 driver_data. And also increase UNCORE_EXTRA_PCI_DEV_MAX to handle the new index.
--- arch/x86/events/intel/uncore.h | 2 +- arch/x86/events/intel/uncore_snbep.c | 7 ++++--- 2 files changed, 5 insertions(+), 4 deletions(-)
arch/x86/events/intel/uncore.h +1 -1
@@ -28,7 +28,7 @@#define UNCORE_PCI_DEV_TYPE (data) ((data >> 8) & 0xff)#define UNCORE_PCI_DEV_IDX (data) (data & 0xff)#define UNCORE_EXTRA_PCI_DEV 0xff -#define UNCORE_EXTRA_PCI_DEV_MAX 3 +#define UNCORE_EXTRA_PCI_DEV_MAX 4#define UNCORE_EVENT_CONSTRAINT (c, n) EVENT_CONSTRAINT(c, n, 0xff)
arch/x86/events/intel/uncore_snbep.c +4 -3
@@ -1030,6 +1030,7 @@ enum {SNBEP_PCI_QPI_PORT0_FILTER ,SNBEP_PCI_QPI_PORT1_FILTER ,HSWEP_PCI_PCU_3 , + BDX_PCI_PCU_3, };static int snbep_qpi_hw_config (struct intel_uncore_box *box,struct perf_event *event)
@@ -3070,11 +3071,11 @@ void bdx_uncore_cpu_init(void)if (boot_cpu_data.x86_model ==86 ) { uncore_msr_uncores[BDX_MSR_UNCORE_SBOX ] =NULL ;/* Detect systems with no SBOXes */ - }else if (uncore_extra_pci_dev[pkg].dev[HSWEP_PCI_PCU_3 ]) { + }else if (uncore_extra_pci_dev[pkg].dev[BDX_PCI_PCU_3 ]) {struct pci_dev *pdev;u32 capid4; - pdev = uncore_extra_pci_dev[pkg].dev[HSWEP_PCI_PCU_3 ]; + pdev = uncore_extra_pci_dev[pkg].dev[BDX_PCI_PCU_3 ];pci_read_config_dword (pdev,0x94 , &capid4);if (((capid4 >>6 ) &0x3 ) ==0 ) bdx_msr_uncores[BDX_MSR_UNCORE_SBOX ] =NULL ;
@@ -3299,7 +3300,7 @@ static const struct pci_device_id bdx_uncore_pci_ids[] = { {/* PCU.3 (for Capability registers) */ PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6fc0), .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, - HSWEP_PCI_PCU_3), + BDX_PCI_PCU_3), }, {/* end: all zeroes */ } };
Signature · 1 line
2.16.1
Quoted 19 lines · Masayoshi Mizuma
From: Masayoshi Mizuma <[email protected]>
commit
15a3e845b01c("perf/x86/intel/uncore: Fix SBOX support for Broadwell CPUs") introduced PCU.3 for Broadwell CPU. Unfortunately, thedriver_dataof PCU.3 conflicts to QPI Port 2 filter.{ /* QPI Port 2 filter */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f46), .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 2),{ /* PCU.3 (for Capability registers) */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6fc0), .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV,
HSWEP_PCI_PCU_3),
//HSWEP_PCI_PCU_3== 2
arch/x86/events/intel/uncore_snbep.c enum { · line 1034
Quoted 5 lines
arch/x86/events/intel/uncore_snbep.c +1
-0@@ -1030,6 +1030,7 @@ enum {SNBEP_PCI_QPI_PORT0_FILTER ,SNBEP_PCI_QPI_PORT1_FILTER ,HSWEP_PCI_PCU_3 , + BDX_PCI_PCU_3, };
So we use a magic '2' enumerator in the 'QPI Port 2 filter', and that overlaps with HSWEP_PCI_PCU_3, right?
Shouldn't we clean up all the enumerators and not use magic numbers, and this fix the conflict?
Thanks,
Ingo
Quoted 31 lines · Ingo Molnar
Quoted 19 lines · Masayoshi Mizuma
From: Masayoshi Mizuma <[email protected]>
commit
15a3e845b01c("perf/x86/intel/uncore: Fix SBOX support for Broadwell CPUs") introduced PCU.3 for Broadwell CPU. Unfortunately, thedriver_dataof PCU.3 conflicts to QPI Port 2 filter.{ /* QPI Port 2 filter */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f46), .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 2),{ /* PCU.3 (for Capability registers) */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6fc0), .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV,
HSWEP_PCI_PCU_3),
//HSWEP_PCI_PCU_3== 2arch/x86/events/intel/uncore_snbep.c enum { · line 1034
Quoted 5 lines
arch/x86/events/intel/uncore_snbep.c +1
-0@@ -1030,6 +1030,7 @@ enum {SNBEP_PCI_QPI_PORT0_FILTER ,SNBEP_PCI_QPI_PORT1_FILTER ,HSWEP_PCI_PCU_3 , + BDX_PCI_PCU_3, };So we use a magic '2' enumerator in the 'QPI Port 2 filter', and that overlaps with
HSWEP_PCI_PCU_3, right?
Quoted 3 lines
Shouldn't we clean up all the enumerators and not use magic numbers, and this fix the conflict?
Yes, it should fix the conflict. I will clean up the code.
Thanks,
Kan
Quoted 4 lines
Thanks,
Ingo
Quoted 34 lines · Kan
Quoted 31 lines · Ingo Molnar
Quoted 19 lines · Masayoshi Mizuma
From: Masayoshi Mizuma <[email protected]>
commit
15a3e845b01c("perf/x86/intel/uncore: Fix SBOX support for Broadwell CPUs") introduced PCU.3 for Broadwell CPU. Unfortunately, thedriver_dataof PCU.3 conflicts to QPI Port 2 filter.{ /* QPI Port 2 filter */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f46), .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 2),{ /* PCU.3 (for Capability registers) */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6fc0), .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, HSWEP_PCI_PCU_3), //HSWEP_PCI_PCU_3== 2arch/x86/events/intel/uncore_snbep.c enum { · line 1034
Quoted 5 lines
arch/x86/events/intel/uncore_snbep.c +1
-0@@ -1030,6 +1030,7 @@ enum { SNBEP_PCI_QPI_PORT0_FILTER, SNBEP_PCI_QPI_PORT1_FILTER, HSWEP_PCI_PCU_3, + BDX_PCI_PCU_3, };So we use a magic '2' enumerator in the 'QPI Port 2 filter', and that overlaps with
HSWEP_PCI_PCU_3, right?
Quoted 5 lines
Quoted 3 lines
Shouldn't we clean up all the enumerators and not use magic numbers, and this fix the conflict?
Yes, it should fix the conflict. I will clean up the code.
Thanks a lot!
I would appreciate if you could add CC to me when you post the patch.
Thanks,
Masa
Quoted 9 lines
Thanks,
KanQuoted 4 lines
Thanks,
Ingo
Quoted 37 lines · Masayoshi Mizuma
Quoted 34 lines · Kan
Quoted 31 lines · Ingo Molnar
Quoted 19 lines · Masayoshi Mizuma
From: Masayoshi Mizuma <[email protected]>
commit
15a3e845b01c("perf/x86/intel/uncore: Fix SBOX support for Broadwell CPUs") introduced PCU.3 for Broadwell CPU. Unfortunately, thedriver_dataof PCU.3 conflicts to QPI Port 2 filter.{ /* QPI Port 2 filter */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f46), .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 2),{ /* PCU.3 (for Capability registers) */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6fc0), .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, HSWEP_PCI_PCU_3), //HSWEP_PCI_PCU_3== 2arch/x86/events/intel/uncore_snbep.c enum { · line 1034
Quoted 5 lines
arch/x86/events/intel/uncore_snbep.c +1
-0@@ -1030,6 +1030,7 @@ enum { SNBEP_PCI_QPI_PORT0_FILTER, SNBEP_PCI_QPI_PORT1_FILTER, HSWEP_PCI_PCU_3, + BDX_PCI_PCU_3, };So we use a magic '2' enumerator in the 'QPI Port 2 filter', and that overlaps with
HSWEP_PCI_PCU_3, right?
Quoted 9 lines
Quoted 5 lines
Quoted 3 lines
Shouldn't we clean up all the enumerators and not use magic numbers, and this fix the conflict?
Yes, it should fix the conflict. I will clean up the code.
Thanks a lot!
I would appreciate if you could add CC to me when you post the patch.
Here is the patch.
Masa, could you please give it a try?
Thanks,
Kan
From 688378a4003ec33156958a52dc822105c18075af Mon Sep 17 00:00:00 2001
From: Kan Liang <[email protected]>
Date: Mon, 16 Jul 2018 04:57:51 -0400
Subject: [PATCH] perf/x86/intel/uncore: Fix hardcode index of Broadwell extra PCI DEV
Masa reports that a warning message is shown while CPU hot-removing on Broadwell server.
WARNING: CPU: 126 PID: 6 at arch/x86/events/intel/uncore.c:988 uncore_pci_remove+0x10b/0x150 Call Trace: pci_device_remove+0x42/0xd0 device_release_driver_internal+0x148/0x220 pci_stop_bus_device+0x76/0xa0 pci_stop_root_bus+0x44/0x60 acpi_pci_root_remove+0x1f/0x80 acpi_bus_trim+0x57/0x90 acpi_bus_trim+0x2e/0x90 acpi_device_hotplug+0x2bc/0x4b0 acpi_hotplug_work_fn+0x1a/0x30 process_one_work+0x174/0x3a0 worker_thread+0x4c/0x3d0 kthread+0xf8/0x130
This bug was introduced in:
commit 15a3e845b01c ("perf/x86/intel/uncore: Fix SBOX support for
Broadwell CPUs")
The index of "QPI Port 2 filter" was hardcode to 2. The index of "PCU.3" used enumerator "HSWEP_PCI_PCU_3", which equals to 2 as well.
To fix the conflict, the hardcode index needs to be cleaned up. Introduce a new enumerator "BDX_PCI_QPI_PORT2_FILTER" for "QPI Port 2 filter" on Broadwell, and increase the UNCORE_EXTRA_PCI_DEV_MAX. Clean up hardcode index.
Broadwell CPUs")
--- arch/x86/events/intel/uncore.h | 2 +- arch/x86/events/intel/uncore_snbep.c | 10 +++++++--- 2 files changed, 8 insertions(+), 4 deletions(-)
arch/x86/events/intel/uncore.h +1 -1
@@ -28,7 +28,7 @@#define UNCORE_PCI_DEV_TYPE (data) ((data >> 8) & 0xff)#define UNCORE_PCI_DEV_IDX (data) (data & 0xff)#define UNCORE_EXTRA_PCI_DEV 0xff -#define UNCORE_EXTRA_PCI_DEV_MAX 3 +#define UNCORE_EXTRA_PCI_DEV_MAX 4#define UNCORE_EVENT_CONSTRAINT (c, n) EVENT_CONSTRAINT(c, n, 0xff)
arch/x86/events/intel/uncore_snbep.c +7 -3
@@ -1029,6 +1029,7 @@ void snbep_uncore_cpu_init(void)enum {SNBEP_PCI_QPI_PORT0_FILTER ,SNBEP_PCI_QPI_PORT1_FILTER , +BDX_PCI_QPI_PORT2_FILTER ,HSWEP_PCI_PCU_3 , };
@@ -3286,15 +3287,18 @@ static const struct pci_device_id bdx_uncore_pci_ids[] = { }, {/* QPI Port 0 filter */ PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f86), - .driver_data =UNCORE_PCI_DEV_DATA (UNCORE_EXTRA_PCI_DEV ,0 ), + .driver_data =UNCORE_PCI_DEV_DATA (UNCORE_EXTRA_PCI_DEV , +SNBEP_PCI_QPI_PORT0_FILTER ), }, { /* QPI Port 1 filter */PCI_DEVICE (PCI_VENDOR_ID_INTEL ,0x6f96 ), - .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 1), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, +SNBEP_PCI_QPI_PORT1_FILTER ), }, { /* QPI Port 2 filter */PCI_DEVICE (PCI_VENDOR_ID_INTEL ,0x6f46 ), - .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 2), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, +BDX_PCI_QPI_PORT2_FILTER ), }, { /* PCU.3 (for Capability registers) */ PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6fc0),
Signature · 1 line
2.4.11
Masayoshi Mizuma
Tested-byQuoted 46 lines · Kan
Quoted 43 lines · Masayoshi Mizuma
Quoted 39 lines · Kan
Quoted 31 lines · Ingo Molnar
Quoted 19 lines · Masayoshi Mizuma
From: Masayoshi Mizuma <[email protected]>
commit
15a3e845b01c("perf/x86/intel/uncore: Fix SBOX support for Broadwell CPUs") introduced PCU.3 for Broadwell CPU. Unfortunately, thedriver_dataof PCU.3 conflicts to QPI Port 2 filter.{ /* QPI Port 2 filter */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f46), .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 2),{ /* PCU.3 (for Capability registers) */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6fc0), .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, HSWEP_PCI_PCU_3), //HSWEP_PCI_PCU_3== 2arch/x86/events/intel/uncore_snbep.c enum { · line 1034
Quoted 5 lines
arch/x86/events/intel/uncore_snbep.c +1
-0@@ -1030,6 +1030,7 @@ enum { SNBEP_PCI_QPI_PORT0_FILTER, SNBEP_PCI_QPI_PORT1_FILTER, HSWEP_PCI_PCU_3, + BDX_PCI_PCU_3, };So we use a magic '2' enumerator in the 'QPI Port 2 filter', and that overlaps with
HSWEP_PCI_PCU_3, right?Quoted 3 lines
Shouldn't we clean up all the enumerators and not use magic numbers, and this fix the conflict?
Yes, it should fix the conflict. I will clean up the code.
Quoted 7 lines
Quoted 3 lines
Thanks a lot!
I would appreciate if you could add CC to me when you post the patch.Here is the patch.
Masa, could you please give it a try?
Thank you for the patch, it works well!
Please feel free to add:
Thanks,
Masa
Quoted 98 lines
Thanks,
KanFrom
688378a4003ec33156958a52dc822105c18075afMon Sep 17 00:00:00 2001
From: Kan Liang <[email protected]>
Date: Mon, 16 Jul 2018 04:57:51 -0400
Subject: [PATCH] perf/x86/intel/uncore: Fix hardcode index of Broadwell extra PCI DEVMasa reports that a warning message is shown while CPU hot-removing on Broadwell server.
WARNING: CPU: 126 PID: 6 at arch/x86/events/intel/uncore.c:988 uncore_pci_remove+0x10b/0x150 Call Trace: pci_device_remove+0x42/0xd0 device_release_driver_internal+0x148/0x220 pci_stop_bus_device+0x76/0xa0 pci_stop_root_bus+0x44/0x60 acpi_pci_root_remove+0x1f/0x80 acpi_bus_trim+0x57/0x90 acpi_bus_trim+0x2e/0x90 acpi_device_hotplug+0x2bc/0x4b0 acpi_hotplug_work_fn+0x1a/0x30 process_one_work+0x174/0x3a0 worker_thread+0x4c/0x3d0 kthread+0xf8/0x130This bug was introduced in:
commit
15a3e845b01c("perf/x86/intel/uncore: Fix SBOX support for Broadwell CPUs")The index of "QPI Port 2 filter" was hardcode to 2. The index of "PCU.3" used enumerator "
HSWEP_PCI_PCU_3", which equals to 2 as well.To fix the conflict, the hardcode index needs to be cleaned up. Introduce a new enumerator "
BDX_PCI_QPI_PORT2_FILTER" for "QPI Port 2 filter" on Broadwell, and increase theUNCORE_EXTRA_PCI_DEV_MAX. Clean up hardcode index.Reported-by Masayoshi Mizuma <[email protected]>Suggested-by Ingo Molnar <[email protected]>Signed-off-by Kan Liang <[email protected]>Fixes 15a3e845b01c ("perf/x86/intel/uncore: Fix SBOX support forBroadwell CPUs")
---
arch/x86/events/intel/uncore.h | 2 +-arch/x86/events/intel/uncore_snbep.c| 10 +++++++---2 files changed, 8 insertions(+), 4 deletions(-)arch/x86/events/intel/uncore.h +1
-1@@ -28,7 +28,7 @@ #define UNCORE_PCI_DEV_TYPE(data) ((data >> 8) & 0xff) #define UNCORE_PCI_DEV_IDX(data) (data & 0xff) #define UNCORE_EXTRA_PCI_DEV 0xff -#define UNCORE_EXTRA_PCI_DEV_MAX 3 +#define UNCORE_EXTRA_PCI_DEV_MAX 4 #define UNCORE_EVENT_CONSTRAINT(c, n) EVENT_CONSTRAINT(c, n, 0xff)arch/x86/events/intel/uncore_snbep.c +1
-0@@ -1029,6 +1029,7 @@ void snbep_uncore_cpu_init(void) enum { SNBEP_PCI_QPI_PORT0_FILTER, SNBEP_PCI_QPI_PORT1_FILTER, + BDX_PCI_QPI_PORT2_FILTER, HSWEP_PCI_PCU_3, };@@ -3286,15 +3287,18 @@ static const struct pci_device_id bdx_uncore_pci_ids[] = { },{ /* QPI Port 0 filter */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f86),
- .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 0), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, +SNBEP_PCI_QPI_PORT0_FILTER), },
{ /* QPI Port 1 filter */PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f96),
- .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 1), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, +SNBEP_PCI_QPI_PORT1_FILTER), },
{ /* QPI Port 2 filter */PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f46),
- .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 2), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, +BDX_PCI_QPI_PORT2_FILTER),
},
{ /* PCU.3 (for Capability registers) */PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6fc0),
Hi Ingo,
Is the following Kan's patch ready to merge...?
Thanks,
Masa
On 07/17/2018 12:37 PM, Masayoshi Mizuma wrote:
...
Quoted 111 lines
Quoted 3 lines
Here is the patch.
Masa, could you please give it a try?
Thank you for the patch, it works well!
Please feel free to add:
Tested-by Masayoshi Mizuma <[email protected]>Thanks,
MasaQuoted 98 lines
Thanks,
KanFrom
688378a4003ec33156958a52dc822105c18075afMon Sep 17 00:00:00 2001
From: Kan Liang <[email protected]>
Date: Mon, 16 Jul 2018 04:57:51 -0400
Subject: [PATCH] perf/x86/intel/uncore: Fix hardcode index of Broadwell extra PCI DEVMasa reports that a warning message is shown while CPU hot-removing on Broadwell server.
WARNING: CPU: 126 PID: 6 at arch/x86/events/intel/uncore.c:988 uncore_pci_remove+0x10b/0x150 Call Trace: pci_device_remove+0x42/0xd0 device_release_driver_internal+0x148/0x220 pci_stop_bus_device+0x76/0xa0 pci_stop_root_bus+0x44/0x60 acpi_pci_root_remove+0x1f/0x80 acpi_bus_trim+0x57/0x90 acpi_bus_trim+0x2e/0x90 acpi_device_hotplug+0x2bc/0x4b0 acpi_hotplug_work_fn+0x1a/0x30 process_one_work+0x174/0x3a0 worker_thread+0x4c/0x3d0 kthread+0xf8/0x130This bug was introduced in:
commit
15a3e845b01c("perf/x86/intel/uncore: Fix SBOX support for Broadwell CPUs")The index of "QPI Port 2 filter" was hardcode to 2. The index of "PCU.3" used enumerator "
HSWEP_PCI_PCU_3", which equals to 2 as well.To fix the conflict, the hardcode index needs to be cleaned up. Introduce a new enumerator "
BDX_PCI_QPI_PORT2_FILTER" for "QPI Port 2 filter" on Broadwell, and increase theUNCORE_EXTRA_PCI_DEV_MAX. Clean up hardcode index.Reported-by Masayoshi Mizuma <[email protected]>Suggested-by Ingo Molnar <[email protected]>Signed-off-by Kan Liang <[email protected]>Fixes 15a3e845b01c ("perf/x86/intel/uncore: Fix SBOX support forBroadwell CPUs")
---
arch/x86/events/intel/uncore.h | 2 +-arch/x86/events/intel/uncore_snbep.c| 10 +++++++---2 files changed, 8 insertions(+), 4 deletions(-)arch/x86/events/intel/uncore.h +1
-1@@ -28,7 +28,7 @@ #define UNCORE_PCI_DEV_TYPE(data) ((data >> 8) & 0xff) #define UNCORE_PCI_DEV_IDX(data) (data & 0xff) #define UNCORE_EXTRA_PCI_DEV 0xff -#define UNCORE_EXTRA_PCI_DEV_MAX 3 +#define UNCORE_EXTRA_PCI_DEV_MAX 4 #define UNCORE_EVENT_CONSTRAINT(c, n) EVENT_CONSTRAINT(c, n, 0xff)arch/x86/events/intel/uncore_snbep.c +1
-0@@ -1029,6 +1029,7 @@ void snbep_uncore_cpu_init(void) enum { SNBEP_PCI_QPI_PORT0_FILTER, SNBEP_PCI_QPI_PORT1_FILTER, + BDX_PCI_QPI_PORT2_FILTER, HSWEP_PCI_PCU_3, };@@ -3286,15 +3287,18 @@ static const struct pci_device_id bdx_uncore_pci_ids[] = { },{ /* QPI Port 0 filter */
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f86),
- .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 0), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, +SNBEP_PCI_QPI_PORT0_FILTER), },
{ /* QPI Port 1 filter */PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f96),
- .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 1), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, +SNBEP_PCI_QPI_PORT1_FILTER), },
{ /* QPI Port 2 filter */PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6f46),
- .driver_data =UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, 2), + .driver_data = UNCORE_PCI_DEV_DATA(UNCORE_EXTRA_PCI_DEV, +BDX_PCI_QPI_PORT2_FILTER),
},
{ /* PCU.3 (for Capability registers) */PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x6fc0),
Quoted 3 lines · Masayoshi Mizuma
Hi Ingo,
Is the following Kan's patch ready to merge...?
Looks good at first sight - but it was whitespace damaged here so I couldn't apply it.
Kan, mind re-sending it properly as a standalone patch, with Masayoshi's Reported-by, Debugged-by and Tested-by tags added?
Thanks,
Ingo
Quoted 12 lines · Ingo Molnar
Quoted 3 lines · Masayoshi Mizuma
Hi Ingo,
Is the following Kan's patch ready to merge...?
Looks good at first sight - but it was whitespace damaged here so I couldn't apply it.
Kan, mind re-sending it properly as a standalone patch, with Masayoshi's Reported-by, Debugged-by and Tested-by tags added?
Sure, I will re-send the patch.
Thanks,
Kan