lib: sbi: pmu: Return SBI_EINVAL if cidx_mask is 0 in ctr_cfg_match
PATCH v2 1/1 lib: sbi: pmu: Return SBI_EINVAL if cidx_mask is 0 in ctr_cfg_match
Currently, when configuring a matching programmable HPM counter with Sscofpmf being present, cidx_base > 2, and cidx_mask == 0 to monitor either the CPU_CYCLES or INSTRUCTIONS hardware event, ctr_cfg_match will succeed but it will configure the corresponding fixed counter instead of the counter specified by the cidx_base parameter.
During counter configuration, the following issues may arise:
- If the CLEAR_VALUE flag is set, the corresponding fixed counter will be reset, which could be considered unexpected behavior.
- If the AUTO_START flag is set, ctr_start_hw will silently start the fixed counter, even though it has already started. From the supervisor's perspective, nothing has changed, which could be confusing. The supervisor will not see the SBI_ERR_ALREADY_STARTED error code since ctr_cfg_match does not return the error code of ctr_start_hw.
The only way to detect these issues is to check the counter_idx return value of ctr_cfg_match and compare it with cidx_base.
Fix these issues by returning the SBI_ERR_INVALID_PARAM error code if the cidx_mask parameter value being passed in is 0 since an invalid parameter should not lead to a successful ctr_cfg_match but with unexpected side effects. This also aligns OpenSBI's behavior with KVM's.
--- lib/sbi/sbi_pmu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
lib/sbi/sbi_pmu.c +1 -1
@@ -840,7 +840,7 @@ int sbi_pmu_ctr_cfg_match(unsigned long cidx_base, unsigned long cidx_mask,u32 event_code;/* Do a basic sanity check of counter base & mask */ -if ((cidx_base +sbi_fls (cidx_mask)) >= total_ctrs) +if (!cidx_mask || ((cidx_base +sbi_fls (cidx_mask)) >= total_ctrs))return SBI_EINVAL ; event_type =pmu_event_validate (phs, event_idx, event_data);
Signature · 7 lines
2.43.0 -- opensbi mailing list [email protected] http://lists.infradead.org/mailman/listinfo/opensbi
Quoted 34 lines · James Raphael Tiovalen
Currently, when configuring a matching programmable HPM counter with Sscofpmf being present,
cidx_base> 2, andcidx_mask== 0 to monitor either theCPU_CYCLESor INSTRUCTIONS hardware event,ctr_cfg_matchwill succeed but it will configure the corresponding fixed counter instead of the counter specified by thecidx_baseparameter.During counter configuration, the following issues may arise:
- If theCLEAR_VALUEflag is set, the corresponding fixed counter will be reset, which could be considered unexpected behavior.- If the
AUTO_STARTflag is set,ctr_start_hwwill silently start the fixed counter, even though it has already started. From the supervisor's perspective, nothing has changed, which could be confusing. The supervisor will not see theSBI_ERR_ALREADY_STARTEDerror code sincectr_cfg_matchdoes not return the error code ofctr_start_hw.The only way to detect these issues is to check the
counter_idxreturn value ofctr_cfg_matchand compare it withcidx_base.Fix these issues by returning the
SBI_ERR_INVALID_PARAMerror code if thecidx_maskparameter value being passed in is 0 since an invalid parameter should not lead to a successfulctr_cfg_matchbut with unexpected side effects. This also aligns OpenSBI's behavior with KVM's.Signed-off-by James Raphael Tiovalen <[email protected]>--- lib/sbi/sbi_pmu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)lib/sbi/sbi_pmu.c +0
-0
lib/sbi/sbi_pmu.c int sbi_pmu_ctr_cfg_match(unsigned long cidx_base, unsigned long cidx_mask, · line 846
Quoted 8 lines
lib/sbi/sbi_pmu.c +1
-1@@ -840,7 +840,7 @@ int sbi_pmu_ctr_cfg_match(unsigned long cidx_base, unsigned long cidx_mask,u32 event_code;/* Do a basic sanity check of counter base & mask */ -if ((cidx_base +sbi_fls (cidx_mask)) >= total_ctrs) +if (!cidx_mask || ((cidx_base +sbi_fls (cidx_mask)) >= total_ctrs))return SBI_EINVAL ; event_type =pmu_event_validate (phs, event_idx, event_data);
Nice catch. Thanks for the fix. I think we need the same fix in sbi_pmu_ctr_start/stop.sbi_fls behavior is undefined when the there are no set bits. So checking the mask is a correct thing do anyways.
Regards,
Atish
Signature · 3 lines
opensbi mailing list [email protected] http://lists.infradead.org/mailman/listinfo/opensbi