lib: pmu: allow to use the highest available counter
PATCH 1/1 lib: pmu: allow to use the highest available counter
Both OpenSBI and OS explicitly assume that there is no hardware counter with index 1: hardware uses that bit for TM control. So OpenSBI filters out that index in sanity checks. However the range sanity checks do not treat that index in a special way. As a result, OpenSBI does not allow to use the firmware counter with the highest index. This change modifies range checks to allow access to the highest index firmware counter.
The simple test is to make sure that there is no counter multiplexing in the following command:
$ perf stat -e \ r8000000000000000,r8000000000000001,r8000000000000002,r8000000000000003, \ r8000000000000004,r8000000000000005,r8000000000000006,r8000000000000007, \ r8000000000000008,r8000000000000009,r800000000000000a,r800000000000000b, \ r800000000000000c,r800000000000000d,r800000000000000e,r800000000000000f \ ls
---
Note that perf RISC-V SBI Linux driver is adapted for the current OpenSBI
behavior: it passes counters mask with exluded highest valid index. So the accompanying fix is also required for Linux, see the patches posted to the risc-v mailing list:
https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3c
lib/sbi/sbi_pmu.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)
lib/sbi/sbi_pmu.c +9 -9
@@ -115,7 +115,7 @@ static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) event_idx_val = active_events[hartid][cidx]; -if (cidx >= total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID )) +if (cidx > total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID ))return SBI_EINVAL ; event_idx_type =get_cidx_type (event_idx_val);
@@ -347,13 +347,13 @@ int sbi_pmu_ctr_start(unsigned long cbase, unsigned long cmask,int ret =SBI_EINVAL ;bool bUpdate = FALSE; -if (sbi_fls (ctr_mask) >= total_ctrs) +if (sbi_fls (ctr_mask) > total_ctrs)return ret;if (flags &SBI_PMU_START_FLAG_SET_INIT_VALUE ) bUpdate = TRUE; -for_each_set_bit_from (cbase, &ctr_mask, total_ctrs) { +for_each_set_bit_from (cbase, &ctr_mask, total_ctrs +1 ) { event_idx_type =pmu_ctr_validate (cbase, &event_code);if (event_idx_type <0 )/* Continue the start operation for other counters */
@@ -423,10 +423,10 @@ int sbi_pmu_ctr_stop(unsigned long cbase, unsigned long cmask,uint32_t event_code;unsigned long ctr_mask = cmask << cbase; -if (sbi_fls (ctr_mask) >= total_ctrs) +if (sbi_fls (ctr_mask) > total_ctrs)return SBI_EINVAL ; -for_each_set_bit_from (cbase, &ctr_mask, total_ctrs) { +for_each_set_bit_from (cbase, &ctr_mask, total_ctrs +1 ) { event_idx_type =pmu_ctr_validate (cbase, &event_code);if (event_idx_type <0 )/* Continue the stop operation for other counters */
@@ -598,7 +598,7 @@ static int pmu_ctr_find_fw(unsigned long cbase, unsigned long cmask, u32 hartid)else fw_base = cbase; -for (i = fw_base; i < total_ctrs; i++) +for (i = fw_base; i <= total_ctrs; i++)if ((active_events[hartid][i] ==SBI_PMU_EVENT_IDX_INVALID ) && ((1UL << i) & ctr_mask))return i;
@@ -618,7 +618,7 @@ int sbi_pmu_ctr_cfg_match(unsigned long cidx_base, unsigned long cidx_mask,unsigned long tmp = cidx_mask << cidx_base;/* Do a basic sanity check of counter base & mask */ -if (sbi_fls (tmp) >= total_ctrs || event_type >=SBI_PMU_EVENT_TYPE_MAX ) +if (sbi_fls (tmp) > total_ctrs || event_type >=SBI_PMU_EVENT_TYPE_MAX )return SBI_EINVAL ;if (flags &SBI_PMU_CFG_FLAG_SKIP_MATCH ) {
@@ -690,7 +690,7 @@ int sbi_pmu_ctr_get_info(uint32_t cidx, unsigned long *ctr_info)struct sbi_scratch *scratch =sbi_scratch_thishart_ptr ();/* Sanity check. Counter1 is not mapped at all */ -if (cidx >= total_ctrs || cidx ==1 ) +if (cidx > total_ctrs || cidx ==1 )return SBI_EINVAL ;/* We have 31 HW counters with 31 being the last index(MHPMCOUNTER31) */
@@ -719,7 +719,7 @@ static void pmu_reset_event_map(u32 hartid)int j;/* Initialize the counter to event mapping table */ -for (j =3 ; j < total_ctrs; j++) +for (j =3 ; j <= total_ctrs; j++) active_events[hartid][j] =SBI_PMU_EVENT_IDX_INVALID ;for (j =0 ; j <SBI_PMU_FW_CTR_MAX ; j++) sbi_memset(&fw_event_map[hartid][j], 0,
Signature · 1 line
2.36.1
PATCH 1/1 lib: pmu: allow to use the highest available counter
Quoted 37 lines · Sergey Matyukevich
Both OpenSBI and OS explicitly assume that there is no hardware counter with index 1: hardware uses that bit for TM control. So OpenSBI filters out that index in sanity checks. However the range sanity checks do not treat that index in a special way. As a result, OpenSBI does not allow to use the firmware counter with the highest index. This change modifies range checks to allow access to the highest index firmware counter.
The simple test is to make sure that there is no counter multiplexing in the following command:
$ perf stat -e \ r8000000000000000,r8000000000000001,r8000000000000002,r8000000000000003, \ r8000000000000004,r8000000000000005,r8000000000000006,r8000000000000007, \ r8000000000000008,r8000000000000009,r800000000000000a,r800000000000000b, \ r800000000000000c,r800000000000000d,r800000000000000e,r800000000000000f \ lsSigned-off-by Sergey Matyukevich <[email protected]>---
Note that perf RISC-V SBI Linux driver is adapted for the current OpenSBI
behavior: it passes counters mask with exluded highest valid index. So the accompanying fix is also required for Linux, see the patches posted to the risc-v mailing list:https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3c
lib/sbi/sbi_pmu.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)lib/sbi/sbi_pmu.c +0
-0
lib/sbi/sbi_pmu.c static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) · line 118
Quoted 5 lines
lib/sbi/sbi_pmu.c +1
-1@@ -115,7 +115,7 @@ static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) event_idx_val = active_events[hartid][cidx]; -if (cidx >= total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID )) +if (cidx > total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID ))
Instead of changing all comparisons of total_ctrs, why not set "num_hw_ctrs = sbi_hart_mhpm_count(scratch) + 3;" in sbi_pmu_init() ?
Regards,
Anup
Quoted 76 lines
lib/sbi/sbi_pmu.c +8
-9return SBI_EINVAL ; event_idx_type =get_cidx_type (event_idx_val);@@ -347,13 +347,13 @@ int sbi_pmu_ctr_start(unsigned long cbase, unsigned long cmask,int ret =SBI_EINVAL ;bool bUpdate = FALSE; -if (sbi_fls (ctr_mask) >= total_ctrs) +if (sbi_fls (ctr_mask) > total_ctrs)return ret;if (flags &SBI_PMU_START_FLAG_SET_INIT_VALUE ) bUpdate = TRUE; -for_each_set_bit_from (cbase, &ctr_mask, total_ctrs) { +for_each_set_bit_from (cbase, &ctr_mask, total_ctrs +1 ) { event_idx_type =pmu_ctr_validate (cbase, &event_code);if (event_idx_type <0 )/* Continue the start operation for other counters */ @@ -423,10 +423,10 @@ int sbi_pmu_ctr_stop(unsigned long cbase, unsigned long cmask,uint32_t event_code;unsigned long ctr_mask = cmask << cbase; -if (sbi_fls (ctr_mask) >= total_ctrs) +if (sbi_fls (ctr_mask) > total_ctrs)return SBI_EINVAL ; -for_each_set_bit_from (cbase, &ctr_mask, total_ctrs) { +for_each_set_bit_from (cbase, &ctr_mask, total_ctrs +1 ) { event_idx_type =pmu_ctr_validate (cbase, &event_code);if (event_idx_type <0 )/* Continue the stop operation for other counters */ @@ -598,7 +598,7 @@ static int pmu_ctr_find_fw(unsigned long cbase, unsigned long cmask, u32 hartid)else fw_base = cbase; -for (i = fw_base; i < total_ctrs; i++) +for (i = fw_base; i <= total_ctrs; i++)if ((active_events[hartid][i] ==SBI_PMU_EVENT_IDX_INVALID ) && ((1UL << i) & ctr_mask))return i;@@ -618,7 +618,7 @@ int sbi_pmu_ctr_cfg_match(unsigned long cidx_base, unsigned long cidx_mask,unsigned long tmp = cidx_mask << cidx_base;/* Do a basic sanity check of counter base & mask */ -if (sbi_fls (tmp) >= total_ctrs || event_type >=SBI_PMU_EVENT_TYPE_MAX ) +if (sbi_fls (tmp) > total_ctrs || event_type >=SBI_PMU_EVENT_TYPE_MAX )return SBI_EINVAL ;if (flags &SBI_PMU_CFG_FLAG_SKIP_MATCH ) {@@ -690,7 +690,7 @@ int sbi_pmu_ctr_get_info(uint32_t cidx, unsigned long *ctr_info)struct sbi_scratch *scratch =sbi_scratch_thishart_ptr ();/* Sanity check. Counter1 is not mapped at all */ -if (cidx >= total_ctrs || cidx ==1 ) +if (cidx > total_ctrs || cidx ==1 )return SBI_EINVAL ;/* We have 31 HW counters with 31 being the last index(MHPMCOUNTER31) */ @@ -719,7 +719,7 @@ static void pmu_reset_event_map(u32 hartid)int j;/* Initialize the counter to event mapping table */ -for (j =3 ; j < total_ctrs; j++) +for (j =3 ; j <= total_ctrs; j++) active_events[hartid][j] =SBI_PMU_EVENT_IDX_INVALID ;for (j =0 ; j <SBI_PMU_FW_CTR_MAX ; j++) sbi_memset(&fw_event_map[hartid][j], 0, --2.36.1
--
opensbi mailing list
opensbi at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi
PATCH 1/1 lib: pmu: allow to use the highest available counter
Quoted 40 lines · Anup Patel
Quoted 38 lines · Sergey Matyukevich
Both OpenSBI and OS explicitly assume that there is no hardware counter with index 1: hardware uses that bit for TM control. So OpenSBI filters out that index in sanity checks. However the range sanity checks do not treat that index in a special way. As a result, OpenSBI does not allow to use the firmware counter with the highest index. This change modifies range checks to allow access to the highest index firmware counter.
The simple test is to make sure that there is no counter multiplexing in the following command:
$ perf stat -e \ r8000000000000000,r8000000000000001,r8000000000000002,r8000000000000003, \ r8000000000000004,r8000000000000005,r8000000000000006,r8000000000000007, \ r8000000000000008,r8000000000000009,r800000000000000a,r800000000000000b, \ r800000000000000c,r800000000000000d,r800000000000000e,r800000000000000f \ lsSigned-off-by Sergey Matyukevich <[email protected]>---
Note that perf RISC-V SBI Linux driver is adapted for the current OpenSBI
behavior: it passes counters mask with exluded highest valid index. So the accompanying fix is also required for Linux, see the patches posted to the risc-v mailing list:https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3c
lib/sbi/sbi_pmu.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)lib/sbi/sbi_pmu.c +0
-0@@ -115,7 +115,7 @@ static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code)
Quoted 8 lines
lib/sbi/sbi_pmu.c static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) · line 118
Quoted 4 lines
lib/sbi/sbi_pmu.c +1
-1event_idx_val = active_events[hartid][cidx]; -if (cidx >= total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID )) +if (cidx > total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID ))Instead of changing all comparisons of
total_ctrs, why not set "num_hw_ctrs=sbi_hart_mhpm_count(scratch)+ 3;" insbi_pmu_init()?
Yes. That is simpler. We need a few other fixes along with that as well. Here is the diff
lib/sbi/sbi_pmu.c +0 -0
@@ -594,7 +594,7 @@ static int pmu_ctr_find_fw(unsigned long cbase,
unsigned long cmask, u32 hartid)
unsigned long ctr_mask = cmask << cbase;
if (cbase <= num_hw_ctrs)
- fw_base = num_hw_ctrs + 1;
+ fw_base = num_hw_ctrs;
else
fw_base = cbase;@@ -694,7 +694,7 @@ int sbi_pmu_ctr_get_info(uint32_t cidx, unsigned long *ctr_info)
return SBI_EINVAL;
/* We have 31 HW counters with 31 being the last index(MHPMCOUNTER31) */
- if (cidx <= num_hw_ctrs) {
+ if (cidx < num_hw_ctrs) {
cinfo.type = SBI_PMU_CTR_TYPE_HW;
cinfo.csr = CSR_CYCLE + cidx;
/* mcycle & minstret are always 64 bit */
@@ -749,7 +749,7 @@ int sbi_pmu_init(struct sbi_scratch *scratch, bool
cold_boot)
sbi_platform_pmu_init(plat);
/* mcycle & minstret is available always */
- num_hw_ctrs = sbi_hart_mhpm_count(scratch) + 2;
+ num_hw_ctrs = sbi_hart_mhpm_count(scratch) + 3;
total_ctrs = num_hw_ctrs + SBI_PMU_FW_CTR_MAX;
}As the counters are zero indexed, the firmware counter should start from num_hw_ctrs
E.g. If a hardware has 8 programmable counters, num_hw_ctrs will be set to 11. with counter value (0-10). Firmware counters should range from 11-26.
Quoted 84 lines
Regards,
AnupQuoted 72 lines
lib/sbi/sbi_pmu.c +8
-9return SBI_EINVAL ; event_idx_type =get_cidx_type (event_idx_val);@@ -347,13 +347,13 @@ int sbi_pmu_ctr_start(unsigned long cbase, unsigned long cmask,int ret =SBI_EINVAL ;bool bUpdate = FALSE; -if (sbi_fls (ctr_mask) >= total_ctrs) +if (sbi_fls (ctr_mask) > total_ctrs)return ret;if (flags &SBI_PMU_START_FLAG_SET_INIT_VALUE ) bUpdate = TRUE; -for_each_set_bit_from (cbase, &ctr_mask, total_ctrs) { +for_each_set_bit_from (cbase, &ctr_mask, total_ctrs +1 ) { event_idx_type =pmu_ctr_validate (cbase, &event_code);if (event_idx_type <0 )/* Continue the start operation for other counters */ @@ -423,10 +423,10 @@ int sbi_pmu_ctr_stop(unsigned long cbase, unsigned long cmask,uint32_t event_code;unsigned long ctr_mask = cmask << cbase; -if (sbi_fls (ctr_mask) >= total_ctrs) +if (sbi_fls (ctr_mask) > total_ctrs)return SBI_EINVAL ; -for_each_set_bit_from (cbase, &ctr_mask, total_ctrs) { +for_each_set_bit_from (cbase, &ctr_mask, total_ctrs +1 ) { event_idx_type =pmu_ctr_validate (cbase, &event_code);if (event_idx_type <0 )/* Continue the stop operation for other counters */ @@ -598,7 +598,7 @@ static int pmu_ctr_find_fw(unsigned long cbase, unsigned long cmask, u32 hartid)else fw_base = cbase; -for (i = fw_base; i < total_ctrs; i++) +for (i = fw_base; i <= total_ctrs; i++)if ((active_events[hartid][i] ==SBI_PMU_EVENT_IDX_INVALID ) && ((1UL << i) & ctr_mask))return i;@@ -618,7 +618,7 @@ int sbi_pmu_ctr_cfg_match(unsigned long cidx_base, unsigned long cidx_mask,unsigned long tmp = cidx_mask << cidx_base;/* Do a basic sanity check of counter base & mask */ -if (sbi_fls (tmp) >= total_ctrs || event_type >=SBI_PMU_EVENT_TYPE_MAX ) +if (sbi_fls (tmp) > total_ctrs || event_type >=SBI_PMU_EVENT_TYPE_MAX )return SBI_EINVAL ;if (flags &SBI_PMU_CFG_FLAG_SKIP_MATCH ) {@@ -690,7 +690,7 @@ int sbi_pmu_ctr_get_info(uint32_t cidx, unsigned long *ctr_info)struct sbi_scratch *scratch =sbi_scratch_thishart_ptr ();/* Sanity check. Counter1 is not mapped at all */ -if (cidx >= total_ctrs || cidx ==1 ) +if (cidx > total_ctrs || cidx ==1 )return SBI_EINVAL ;/* We have 31 HW counters with 31 being the last index(MHPMCOUNTER31) */ @@ -719,7 +719,7 @@ static void pmu_reset_event_map(u32 hartid)int j;/* Initialize the counter to event mapping table */ -for (j =3 ; j < total_ctrs; j++) +for (j =3 ; j <= total_ctrs; j++) active_events[hartid][j] =SBI_PMU_EVENT_IDX_INVALID ;for (j =0 ; j <SBI_PMU_FW_CTR_MAX ; j++) sbi_memset(&fw_event_map[hartid][j], 0, --2.36.1
Quoted 4 lines
--
opensbi mailing list
opensbi at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi--
opensbi mailing list
opensbi at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi
Signature · 2 lines
Regards, Atish
PATCH 1/1 lib: pmu: allow to use the highest available counter
Hi Atish and Sergey,
Quoted 88 lines · Atish Patra
Quoted 40 lines · Anup Patel
Quoted 38 lines · Sergey Matyukevich
Both OpenSBI and OS explicitly assume that there is no hardware counter with index 1: hardware uses that bit for TM control. So OpenSBI filters out that index in sanity checks. However the range sanity checks do not treat that index in a special way. As a result, OpenSBI does not allow to use the firmware counter with the highest index. This change modifies range checks to allow access to the highest index firmware counter.
The simple test is to make sure that there is no counter multiplexing in the following command:
$ perf stat -e \ r8000000000000000,r8000000000000001,r8000000000000002,r8000000000000003, \ r8000000000000004,r8000000000000005,r8000000000000006,r8000000000000007, \ r8000000000000008,r8000000000000009,r800000000000000a,r800000000000000b, \ r800000000000000c,r800000000000000d,r800000000000000e,r800000000000000f \ lsSigned-off-by Sergey Matyukevich <[email protected]>---
Note that perf RISC-V SBI Linux driver is adapted for the current OpenSBI
behavior: it passes counters mask with exluded highest valid index. So the accompanying fix is also required for Linux, see the patches posted to the risc-v mailing list:https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3c
lib/sbi/sbi_pmu.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)lib/sbi/sbi_pmu.c +0
-0@@ -115,7 +115,7 @@ static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code)Quoted 8 lines
lib/sbi/sbi_pmu.c static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) · line 118
Quoted 4 lines
lib/sbi/sbi_pmu.c +1
-1event_idx_val = active_events[hartid][cidx]; -if (cidx >= total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID )) +if (cidx > total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID ))Instead of changing all comparisons of
total_ctrs, why not set "num_hw_ctrs=sbi_hart_mhpm_count(scratch)+ 3;" insbi_pmu_init()?Yes. That is simpler. We need a few other fixes along with that as well. Here is the diff
lib/sbi/sbi_pmu.c +2
-2@@ -594,7 +594,7 @@ static int pmu_ctr_find_fw(unsigned long cbase, unsigned long cmask, u32 hartid) unsigned long ctr_mask = cmask << cbase;if (cbase <= num_hw_ctrs) - fw_base = num_hw_ctrs +1 ; + fw_base = num_hw_ctrs;else fw_base = cbase;@@ -694,7 +694,7 @@ int sbi_pmu_ctr_get_info(uint32_t cidx, unsigned long *ctr_info)return SBI_EINVAL ;/* We have 31 HW counters with 31 being the last index(MHPMCOUNTER31) */ -if (cidx <= num_hw_ctrs) { +if (cidx < num_hw_ctrs) { cinfo.type =SBI_PMU_CTR_TYPE_HW ; cinfo.csr =CSR_CYCLE + cidx;/* mcycle & minstret are always 64 bit */lib/sbi/sbi_pmu.c +1
-1@@ -749,7 +749,7 @@ int sbi_pmu_init(struct sbi_scratch *scratch, bool cold_boot)sbi_platform_pmu_init (plat );/* mcycle & minstret is available always */ - num_hw_ctrs =sbi_hart_mhpm_count (scratch) +2 ; + num_hw_ctrs =sbi_hart_mhpm_count (scratch) +3 ; total_ctrs = num_hw_ctrs +SBI_PMU_FW_CTR_MAX ; }
Quoted 4 lines
As the counters are zero indexed, the firmware counter should start from
num_hw_ctrs
E.g. If a hardware has 8 programmable counters,num_hw_ctrswill be set to 11. with counter value (0-10). Firmware counters should range from 11-26.
Can one of you quickly send a v2 (preferably today) so that I can include it for OpenSBI v1.1 release ?
Regards,
Anup
Quoted 91 lines
Quoted 76 lines
Regards,
AnupQuoted 72 lines
lib/sbi/sbi_pmu.c +8
-9return SBI_EINVAL ; event_idx_type =get_cidx_type (event_idx_val);@@ -347,13 +347,13 @@ int sbi_pmu_ctr_start(unsigned long cbase, unsigned long cmask,int ret =SBI_EINVAL ;bool bUpdate = FALSE; -if (sbi_fls (ctr_mask) >= total_ctrs) +if (sbi_fls (ctr_mask) > total_ctrs)return ret;if (flags &SBI_PMU_START_FLAG_SET_INIT_VALUE ) bUpdate = TRUE; -for_each_set_bit_from (cbase, &ctr_mask, total_ctrs) { +for_each_set_bit_from (cbase, &ctr_mask, total_ctrs +1 ) { event_idx_type =pmu_ctr_validate (cbase, &event_code);if (event_idx_type <0 )/* Continue the start operation for other counters */ @@ -423,10 +423,10 @@ int sbi_pmu_ctr_stop(unsigned long cbase, unsigned long cmask,uint32_t event_code;unsigned long ctr_mask = cmask << cbase; -if (sbi_fls (ctr_mask) >= total_ctrs) +if (sbi_fls (ctr_mask) > total_ctrs)return SBI_EINVAL ; -for_each_set_bit_from (cbase, &ctr_mask, total_ctrs) { +for_each_set_bit_from (cbase, &ctr_mask, total_ctrs +1 ) { event_idx_type =pmu_ctr_validate (cbase, &event_code);if (event_idx_type <0 )/* Continue the stop operation for other counters */ @@ -598,7 +598,7 @@ static int pmu_ctr_find_fw(unsigned long cbase, unsigned long cmask, u32 hartid)else fw_base = cbase; -for (i = fw_base; i < total_ctrs; i++) +for (i = fw_base; i <= total_ctrs; i++)if ((active_events[hartid][i] ==SBI_PMU_EVENT_IDX_INVALID ) && ((1UL << i) & ctr_mask))return i;@@ -618,7 +618,7 @@ int sbi_pmu_ctr_cfg_match(unsigned long cidx_base, unsigned long cidx_mask,unsigned long tmp = cidx_mask << cidx_base;/* Do a basic sanity check of counter base & mask */ -if (sbi_fls (tmp) >= total_ctrs || event_type >=SBI_PMU_EVENT_TYPE_MAX ) +if (sbi_fls (tmp) > total_ctrs || event_type >=SBI_PMU_EVENT_TYPE_MAX )return SBI_EINVAL ;if (flags &SBI_PMU_CFG_FLAG_SKIP_MATCH ) {@@ -690,7 +690,7 @@ int sbi_pmu_ctr_get_info(uint32_t cidx, unsigned long *ctr_info)struct sbi_scratch *scratch =sbi_scratch_thishart_ptr ();/* Sanity check. Counter1 is not mapped at all */ -if (cidx >= total_ctrs || cidx ==1 ) +if (cidx > total_ctrs || cidx ==1 )return SBI_EINVAL ;/* We have 31 HW counters with 31 being the last index(MHPMCOUNTER31) */ @@ -719,7 +719,7 @@ static void pmu_reset_event_map(u32 hartid)int j;/* Initialize the counter to event mapping table */ -for (j =3 ; j < total_ctrs; j++) +for (j =3 ; j <= total_ctrs; j++) active_events[hartid][j] =SBI_PMU_EVENT_IDX_INVALID ;for (j =0 ; j <SBI_PMU_FW_CTR_MAX ; j++) sbi_memset(&fw_event_map[hartid][j], 0, --2.36.1
Quoted 9 lines
Quoted 4 lines
--
opensbi mailing list
opensbi at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi--
opensbi mailing list
opensbi at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi--
Regards,
Atish
PATCH 1/1 lib: pmu: allow to use the highest available counter
In reply to Atish PatraHello,
Quoted 46 lines
Quoted 37 lines
Quoted 37 lines
Both OpenSBI and OS explicitly assume that there is no hardware counter with index 1: hardware uses that bit for TM control. So OpenSBI filters out that index in sanity checks. However the range sanity checks do not treat that index in a special way. As a result, OpenSBI does not allow to use the firmware counter with the highest index. This change modifies range checks to allow access to the highest index firmware counter.
The simple test is to make sure that there is no counter multiplexing in the following command:
$ perf stat -e \ r8000000000000000,r8000000000000001,r8000000000000002,r8000000000000003, \ r8000000000000004,r8000000000000005,r8000000000000006,r8000000000000007, \ r8000000000000008,r8000000000000009,r800000000000000a,r800000000000000b, \ r800000000000000c,r800000000000000d,r800000000000000e,r800000000000000f \ lsSigned-off-by Sergey Matyukevich <[email protected]>---
Note that perf RISC-V SBI Linux driver is adapted for the current OpenSBI
behavior: it passes counters mask with exluded highest valid index. So the accompanying fix is also required for Linux, see the patches posted to the risc-v mailing list:https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3c
lib/sbi/sbi_pmu.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)lib/sbi/sbi_pmu.c +0
-0@@ -115,7 +115,7 @@ static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code)Quoted 8 lines
lib/sbi/sbi_pmu.c static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) · line 118
Quoted 4 lines
lib/sbi/sbi_pmu.c +1
-1event_idx_val = active_events[hartid][cidx]; -if (cidx >= total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID )) +if (cidx > total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID ))Instead of changing all comparisons of
total_ctrs, why not set "num_hw_ctrs=sbi_hart_mhpm_count(scratch)+ 3;" insbi_pmu_init()?
Quoted 2 lines
Yes. That is simpler. We need a few other fixes along with that as well. Here is the diff
Indeed, this is simpler. But as I mentioned in my reply to v2, this way you report incorrect number of hw counters to kernel just to get the correct mask of counters in response. And that behavior will have to be followed by all the other SBI implementations.
Current OpenSBI implementation implies the only gap in counters: index 1.
So my approach in v1 version of this patch was as follows:
- send the accurate number of counters to the kernel (hw + fw)
- fix range checks to take into account gap for index 1
This way OpenSBI just sends correct information to the kernel and
expects (and verifies) correct data in response. Accompanying kernel
fixes are the first two patches in the kernel series, see:
https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3c
The 3rd kernel patch switches from array to IDR. With that change kernel driver will support any gaps in counters, not only index 1. So OpenSBI or any other SBI implementation will be able to exclude some hw or fw counters if needed without changing kernel driver. So far I can see two possible examples where it could be useful:
- platform specific quirks in OpenSBI to exclude non-functional counters
- keep some hw counters for internal use in OpenSBI or any other SBI implementation
Regards,
Sergey
PATCH 1/1 lib: pmu: allow to use the highest available counter
Quoted 49 lines · Sergey Matyukevich
Hello,
Quoted 46 lines
Quoted 45 lines
Quoted 36 lines
Both OpenSBI and OS explicitly assume that there is no hardware counter with index 1: hardware uses that bit for TM control. So OpenSBI filters out that index in sanity checks. However the range sanity checks do not treat that index in a special way. As a result, OpenSBI does not allow to use the firmware counter with the highest index. This change modifies range checks to allow access to the highest index firmware counter.
The simple test is to make sure that there is no counter multiplexing in the following command:
$ perf stat -e \ r8000000000000000,r8000000000000001,r8000000000000002,r8000000000000003, \ r8000000000000004,r8000000000000005,r8000000000000006,r8000000000000007, \ r8000000000000008,r8000000000000009,r800000000000000a,r800000000000000b, \ r800000000000000c,r800000000000000d,r800000000000000e,r800000000000000f \ lsSigned-off-by Sergey Matyukevich <[email protected]>---
Note that perf RISC-V SBI Linux driver is adapted for the current OpenSBI
behavior: it passes counters mask with exluded highest valid index. So the accompanying fix is also required for Linux, see the patches posted to the risc-v mailing list:https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3c
lib/sbi/sbi_pmu.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)lib/sbi/sbi_pmu.c +0
-0lib/sbi/sbi_pmu.c static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) · line 118
Quoted 5 lines
lib/sbi/sbi_pmu.c +1
-1@@ -115,7 +115,7 @@ static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) event_idx_val = active_events[hartid][cidx]; -if (cidx >= total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID )) +if (cidx > total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID ))Instead of changing all comparisons of
total_ctrs, why not set "num_hw_ctrs=sbi_hart_mhpm_count(scratch)+ 3;" insbi_pmu_init()?
Quoted 7 lines
Quoted 2 lines
Yes. That is simpler. We need a few other fixes along with that as well. Here is the diff
Indeed, this is simpler. But as I mentioned in my reply to v2, this way you report incorrect number of hw counters to kernel just to get the correct mask of counters in response. And that behavior will have to be followed by all the other SBI implementations.
I don't think this patch reports an incorrect number of counters because the time CSR is indeed a counter which counts timer ticks.
Just like cycle and instret CSRs, the time CSR is also a fixed counter which cannot count any other HW event.
The only special thing about time CSR is that as-per RISC-V privilege specification the time CSR is always counting and there is no way to halt/resume it because it is an alias of platform-level free-running mtime counter.
It's totally up to the SBI implementation whether it wants to expose fixed HW counters (cycle, time, and insret) as SBI PMU counters or not. In fact, the SBI PMU specification defines counter_idx as logical counter with no relation to HW counter numbering.
A perfectly valid SBI implementation is allowed to do various things:
1) Intermix HW counters and FW counters (e.g. counter_idx = 0 is FW counter but counter_idx = 1 is HW counter)
2) Make a particular HW counter unusable through SBI PMU due to some errata (e.g. If mhpmcounter3 is broken then platform will not set bit[3] in any of the counter-mask provided in the "riscv,pmu" DT node which is parsed by OpenSBI). The sbi_pmu_config_matching() will never select a counter_idx which is not part of any counter-mask provided by the platform.
3) Create on-demand mapping of counter_idx to HW counters because SBI implementation is doing trap-n-emulate of counters. This is particularly true for hypervisors like KVM when counters are shared between host and guest so the hypervisor will on-demand allocate a HW counter and map it to a logical counter_idx seen by the guest. (Note: same approach is taken by KVM ARM)
Quoted 10 lines
Current OpenSBI implementation implies the only gap in counters: index 1.
So my approach in v1 version of this patch was as follows:
- send the accurate number of counters to the kernel (hw + fw)
- fix range checks to take into account gap for index 1
This way OpenSBI just sends correct information to the kernel and
expects (and verifies) correct data in response. Accompanying kernel
fixes are the first two patches in the kernel series, see:
https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3c
Quoted 7 lines
The 3rd kernel patch switches from array to IDR. With that change kernel driver will support any gaps in counters, not only index 1. So OpenSBI or any other SBI implementation will be able to exclude some hw or fw counters if needed without changing kernel driver. So far I can see two possible examples where it could be useful:
- platform specific quirks in OpenSBI to exclude non-functional counters
- keep some hw counters for internal use in OpenSBI or any other SBI implementation
I think there is some confusion with regards to counter_idx.
As-per SBI spec, the counter_idx is a logical number so if there are N counters then 0 <= counter_idx < N. There are no gaps in counter_idx numbering. Of course, an SBI implementation might intentionally never configure a particular counter_idx because it is special or broken.
Regards,
Anup
Quoted 8 lines
Regards,
Sergey--
opensbi mailing list
opensbi at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi
PATCH 1/1 lib: pmu: allow to use the highest available counter
Hi Anup,
Quoted 105 lines
Quoted 46 lines
Quoted 46 lines
Quoted 45 lines
Quoted 36 lines
Both OpenSBI and OS explicitly assume that there is no hardware counter with index 1: hardware uses that bit for TM control. So OpenSBI filters out that index in sanity checks. However the range sanity checks do not treat that index in a special way. As a result, OpenSBI does not allow to use the firmware counter with the highest index. This change modifies range checks to allow access to the highest index firmware counter.
The simple test is to make sure that there is no counter multiplexing in the following command:
$ perf stat -e \ r8000000000000000,r8000000000000001,r8000000000000002,r8000000000000003, \ r8000000000000004,r8000000000000005,r8000000000000006,r8000000000000007, \ r8000000000000008,r8000000000000009,r800000000000000a,r800000000000000b, \ r800000000000000c,r800000000000000d,r800000000000000e,r800000000000000f \ lsSigned-off-by Sergey Matyukevich <[email protected]>---
Note that perf RISC-V SBI Linux driver is adapted for the current OpenSBI
behavior: it passes counters mask with exluded highest valid index. So the accompanying fix is also required for Linux, see the patches posted to the risc-v mailing list:https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3c
lib/sbi/sbi_pmu.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)lib/sbi/sbi_pmu.c +0
-0lib/sbi/sbi_pmu.c static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) · line 118
Quoted 5 lines
lib/sbi/sbi_pmu.c +1
-1@@ -115,7 +115,7 @@ static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) event_idx_val = active_events[hartid][cidx]; -if (cidx >= total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID )) +if (cidx > total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID ))Instead of changing all comparisons of
total_ctrs, why not set "num_hw_ctrs=sbi_hart_mhpm_count(scratch)+ 3;" insbi_pmu_init()?Quoted 7 lines
Quoted 2 lines
Yes. That is simpler. We need a few other fixes along with that as well. Here is the diff
Indeed, this is simpler. But as I mentioned in my reply to v2, this way you report incorrect number of hw counters to kernel just to get the correct mask of counters in response. And that behavior will have to be followed by all the other SBI implementations.
I don't think this patch reports an incorrect number of counters because the time CSR is indeed a counter which counts timer ticks.
Just like cycle and instret CSRs, the time CSR is also a fixed counter which cannot count any other HW event.
The only special thing about time CSR is that as-per RISC-V privilege specification the time CSR is always counting and there is no way to halt/resume it because it is an alias of platform-level free-running mtime counter.
It's totally up to the SBI implementation whether it wants to expose fixed HW counters (cycle, time, and insret) as SBI PMU counters or not. In fact, the SBI PMU specification defines
counter_idxas logical counter with no relation to HW counter numbering.A perfectly valid SBI implementation is allowed to do various things:
1) Intermix HW counters and FW counters (e.g.counter_idx= 0 is FW counter butcounter_idx= 1 is HW counter)
2) Make a particular HW counter unusable through SBI PMU due to some errata (e.g. If mhpmcounter3 is broken then platform will not set bit[3] in any of the counter-mask provided in the "riscv,pmu" DT node which is parsed by OpenSBI). Thesbi_pmu_config_matching()will never select acounter_idxwhich is not part of any counter-mask provided by the platform.
3) Create on-demand mapping ofcounter_idxto HW counters because SBI implementation is doing trap-n-emulate of counters. This is particularly true for hypervisors like KVM when counters are shared between host and guest so the hypervisor will on-demand allocate a HW counter and map it to a logicalcounter_idxseen by the guest. (Note: same approach is taken by KVM ARM)Quoted 10 lines
Current OpenSBI implementation implies the only gap in counters: index 1.
So my approach in v1 version of this patch was as follows:
- send the accurate number of counters to the kernel (hw + fw)
- fix range checks to take into account gap for index 1
This way OpenSBI just sends correct information to the kernel and
expects (and verifies) correct data in response. Accompanying kernel
fixes are the first two patches in the kernel series, see:
https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3cQuoted 7 lines
The 3rd kernel patch switches from array to IDR. With that change kernel driver will support any gaps in counters, not only index 1. So OpenSBI or any other SBI implementation will be able to exclude some hw or fw counters if needed without changing kernel driver. So far I can see two possible examples where it could be useful:
- platform specific quirks in OpenSBI to exclude non-functional counters
- keep some hw counters for internal use in OpenSBI or any other SBI implementation
Quoted 6 lines
I think there is some confusion with regards to
counter_idx.As-per SBI spec, the
counter_idxis a logical number so if there are N counters then 0 <=counter_idx< N. There are no gaps incounter_idxnumbering. Of course, an SBI implementation might intentionally never configure a particularcounter_idxbecause it is special or broken.
Thanks for clarifications! So, in brief, it is OK for valid SBI implementation to do the following:
- to report _total_ number of counters (hw + fw), including special ones (like timer) or broken ones or reserved ones
- sbi_pmu_ctr_cfg_match should never send special/reserved/broken counters to OS
That makes sense to me. So v2 changes suggested by Atish look reasonable. But I guess commit message needs to be clarified.
I will shortly send you v3 version of the OpenSBI patch with updated commit message. I will also send v2 of the kernel patches.
Regards,
Sergey
PATCH 1/1 lib: pmu: allow to use the highest available counter
Quoted 115 lines · Sergey Matyukevich
Hi Anup,
Quoted 105 lines
Quoted 46 lines
Quoted 46 lines
Quoted 45 lines
Quoted 36 lines
Both OpenSBI and OS explicitly assume that there is no hardware counter with index 1: hardware uses that bit for TM control. So OpenSBI filters out that index in sanity checks. However the range sanity checks do not treat that index in a special way. As a result, OpenSBI does not allow to use the firmware counter with the highest index. This change modifies range checks to allow access to the highest index firmware counter.
The simple test is to make sure that there is no counter multiplexing in the following command:
$ perf stat -e \ r8000000000000000,r8000000000000001,r8000000000000002,r8000000000000003, \ r8000000000000004,r8000000000000005,r8000000000000006,r8000000000000007, \ r8000000000000008,r8000000000000009,r800000000000000a,r800000000000000b, \ r800000000000000c,r800000000000000d,r800000000000000e,r800000000000000f \ lsSigned-off-by Sergey Matyukevich <[email protected]>---
Note that perf RISC-V SBI Linux driver is adapted for the current OpenSBI
behavior: it passes counters mask with exluded highest valid index. So the accompanying fix is also required for Linux, see the patches posted to the risc-v mailing list:https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3c
lib/sbi/sbi_pmu.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)lib/sbi/sbi_pmu.c +0
-0lib/sbi/sbi_pmu.c static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) · line 118
Quoted 5 lines
lib/sbi/sbi_pmu.c +1
-1@@ -115,7 +115,7 @@ static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code) event_idx_val = active_events[hartid][cidx]; -if (cidx >= total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID )) +if (cidx > total_ctrs || (event_idx_val ==SBI_PMU_EVENT_IDX_INVALID ))Instead of changing all comparisons of
total_ctrs, why not set "num_hw_ctrs=sbi_hart_mhpm_count(scratch)+ 3;" insbi_pmu_init()?Quoted 7 lines
Quoted 2 lines
Yes. That is simpler. We need a few other fixes along with that as well. Here is the diff
Indeed, this is simpler. But as I mentioned in my reply to v2, this way you report incorrect number of hw counters to kernel just to get the correct mask of counters in response. And that behavior will have to be followed by all the other SBI implementations.
I don't think this patch reports an incorrect number of counters because the time CSR is indeed a counter which counts timer ticks.
Just like cycle and instret CSRs, the time CSR is also a fixed counter which cannot count any other HW event.
The only special thing about time CSR is that as-per RISC-V privilege specification the time CSR is always counting and there is no way to halt/resume it because it is an alias of platform-level free-running mtime counter.
It's totally up to the SBI implementation whether it wants to expose fixed HW counters (cycle, time, and insret) as SBI PMU counters or not. In fact, the SBI PMU specification defines
counter_idxas logical counter with no relation to HW counter numbering.A perfectly valid SBI implementation is allowed to do various things:
1) Intermix HW counters and FW counters (e.g.counter_idx= 0 is FW counter butcounter_idx= 1 is HW counter)
2) Make a particular HW counter unusable through SBI PMU due to some errata (e.g. If mhpmcounter3 is broken then platform will not set bit[3] in any of the counter-mask provided in the "riscv,pmu" DT node which is parsed by OpenSBI). Thesbi_pmu_config_matching()will never select acounter_idxwhich is not part of any counter-mask provided by the platform.
3) Create on-demand mapping ofcounter_idxto HW counters because SBI implementation is doing trap-n-emulate of counters. This is particularly true for hypervisors like KVM when counters are shared between host and guest so the hypervisor will on-demand allocate a HW counter and map it to a logicalcounter_idxseen by the guest. (Note: same approach is taken by KVM ARM)Quoted 10 lines
Current OpenSBI implementation implies the only gap in counters: index 1.
So my approach in v1 version of this patch was as follows:
- send the accurate number of counters to the kernel (hw + fw)
- fix range checks to take into account gap for index 1
This way OpenSBI just sends correct information to the kernel and
expects (and verifies) correct data in response. Accompanying kernel
fixes are the first two patches in the kernel series, see:
https://lore.kernel.org/linux-riscv/20220623112735.357093-1-geomatsi at gmail.com/T/#m4811c83cf767cdaaeb31e2530d4a51034fcf6c3cQuoted 7 lines
The 3rd kernel patch switches from array to IDR. With that change kernel driver will support any gaps in counters, not only index 1. So OpenSBI or any other SBI implementation will be able to exclude some hw or fw counters if needed without changing kernel driver. So far I can see two possible examples where it could be useful:
- platform specific quirks in OpenSBI to exclude non-functional counters
- keep some hw counters for internal use in OpenSBI or any other SBI implementationQuoted 6 lines
I think there is some confusion with regards to
counter_idx.As-per SBI spec, the
counter_idxis a logical number so if there are N counters then 0 <=counter_idx< N. There are no gaps incounter_idxnumbering. Of course, an SBI implementation might intentionally never configure a particularcounter_idxbecause it is special or broken.
Quoted 5 lines
Thanks for clarifications! So, in brief, it is OK for valid SBI implementation to do the following:
- to report _total_ number of counters (hw + fw), including special ones (like timer) or broken ones or reserved ones
Yes, as long as SBI implementation controls what counters can be configured successfully.
Quoted 2 lines
-
sbi_pmu_ctr_cfg_matchshould never send special/reserved/broken counters to OS
Yes, sbi_pmu_counter_config_matching() should:
1) Never select broken or reserved or special counters
2) Select fixed counters (such as cycle or insret) only for specific PMU event
Quoted 6 lines
That makes sense to me. So v2 changes suggested by Atish look reasonable. But I guess commit message needs to be clarified.
I will shortly send you v3 version of the OpenSBI patch with updated commit message. I will also send v2 of the kernel patches.
Sure, thanks.
Also, thanks for catching this issue.
Quoted 3 lines
Regards,
Sergey
Regards,
Anup