[PATCH OLK-6.6 00/33] arm_mpam: Introduce Narrow-PARTID feature
Background ========== On x86, the resctrl allows creating up to num_rmids monitoring groups under parent control group. However, ARM64 MPAM is currently limited by the PMG (Performance Monitoring Group) count, which is typically much smaller than the theoretical RMID limit. This creates a significant scalability gap: users expecting fine-grained per-process or per-thread monitoring quickly exhaust the PMG space, even when plenty of reqPARTIDs remain available. The Narrow-PARTID feature, defined in the ARM MPAM architecture, addresses this by associating reqPARTIDs with intPARTIDs through a programmable many-to-one mapping. This allows the kernel to present more logical monitoring contexts. Design Overview =============== The implementation extends the RMID encoding to carry reqPARTID information: RMID = reqPARTID * NUM_PMG + PMG In this patchset, a monitoring group is uniquely identified by the combination of reqPARTID and PMG. The closid is represented by intPARTID, which is exactly the original PARTID. For systems with homogeneous MSCs (all supporting Narrow-PARTID), the driver exposes the full reqPARTID range directly. For heterogeneous systems where some MSCs lack Narrow-PARTID support, the driver utilizes PARTIDs beyond the intPARTID range as reqPARTIDs to expand monitoring capacity. The sole exception is when MBA MSCs lack Narrow-PARTID support, their percentage-based control mechanism prevents the use of PARTIDs as reqPARTIDs. Capacity Improvements ===================== -------------------------------------------------------------------------- The maximum | Sub-monitoring groups | System-wide number of | under a control group | monitoring groups -------------------------------------------------------------------------- Without | | reqPARTID | PMG | intPARTID * PMG -------------------------------------------------------------------------- reqPARTID | | static allocation | (reqPARTID // intPARTID) * PMG | reqPARTID * PMG -------------------------------------------------------------------------- reqPARTID | | dynamic allocation | (reqPARTID − intPARTID + 1) * PMG | reqPARTID * PMG -------------------------------------------------------------------------- Under MPAM, the number of reqPARTID is always greater than or equal to intPARTID. Dave Martin (1): arm_mpam: Set INTERNAL as needed when setting MSC controls James Morse (1): debugfs: Add helpers for creating cpumask entries in debugfs Yushan Wang (1): cputype: add cputype for HIP13 Zeng Heng (30): arm_mpam: Add intPARTID and reqPARTID support for narrow PARTID feature arm_mpam: Refactor rmid to reqPARTID/PMG mapping arm_mpam: Propagate control group config to sub-monitoring groups fs/resctrl: Add rmid_entry state helpers arm_mpam: Implement dynamic reqPARTID allocation for monitoring groups fs/resctrl: Wire up rmid expansion and reclaim functions arm64/mpam: Add mpam_sync_config() for dynamic rmid expansion fs/resctrl: Prevent rmid parsing errors by flushing limbo on umount arm_mpam: Disable reqPARTID expansion when Narrow-PARTID is unavailable arm_mpam: Add MBWU counter scaling support fs/resctrl: Fix incorrect closid propagation during monitor group reparenting arm64/mpam: Fix wd boundary check and GENMASK range in percent-mbw conversion arm64/mpam: Add HISI_HIP13 support fs/resctrl: Introduce per-resource MBM enable check fs/resctrl: Add arch-specific MBM overflow setup helpers arm64/mpam: Add selective L2 MBM overflow checking arm64/mpam: Add MBM overflow handling for HIP13 arm64/mpam: Allow zero-bit CPBM for L2 cache arm64/mpam: Adapt SMMU IOMMU interface for Narrow PARTID support arm64/mpam: Add MBOPT (Memory Bandwidth Optimal) feature resctrl: Remove unused mbm_core field from rdt_domain fs/resctrl: Introduce resctrl_arch_mbm_update() for arch-specific MBM event handling arm64/mpam: Add MBM_TOTAL_EVENT configurability check arm64/mpam: Add debugfs entries to show the MSC/RIS the driver discovered iommu/arm-smmu-v3: Add debug logging for MPAM STE configuration Revert "fs/resctrl: Re-allocate rmid for the monitor when migrating across control groups" fs/resctrl: Fix iommu_group and kobject reference leaks arm_mpam: Fix NULL pointer dereference after freeing comp->cfg arm_mpam: Wait for SRCU grace period before destroying RIS arm_mpam: Fix percpu memory leak in __setup_ppi() error path arch/arm64/include/asm/cputype.h | 2 + arch/arm64/include/asm/mpam.h | 17 +- drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 3 + drivers/platform/mpam/mpam_devices.c | 271 ++++++++- drivers/platform/mpam/mpam_internal.h | 21 +- drivers/platform/mpam/mpam_resctrl.c | 617 ++++++++++++++++---- fs/debugfs/file.c | 64 ++ fs/resctrl/internal.h | 1 - fs/resctrl/monitor.c | 123 ++-- fs/resctrl/rdtgroup.c | 132 +++-- include/linux/arm_mpam.h | 20 +- include/linux/debugfs.h | 6 + include/linux/resctrl.h | 35 +- include/linux/resctrl_types.h | 2 + tools/arch/arm64/include/asm/cputype.h | 2 + 15 files changed, 1035 insertions(+), 281 deletions(-) -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Introduce narrow PARTID (partid_nrw) feature support, which enables many-to-one mapping of request PARTIDs (reqPARTID) to internal PARTIDs (intPARTID). This expands monitoring capability by allowing a single control group to track more task types through multiple reqPARTIDs per intPARTID, bypassing the PMG limit in some extent. intPARTID: Internal PARTID used for control group configuration. Configurations are synchronized to all reqPARTIDs mapped to the same intPARTID. Count is indicated by MPAMF_PARTID_NRW_IDR.INTPARTID_MAX, or defaults to PARTID count if narrow PARTID is unsupported. reqPARTID: Request PARTID used to expand monitoring groups. Enables a single control group to monitor more task types by multiple reqPARTIDs within one intPARTID, overcoming the PMG count limitation. Control group and monitoring limits are calculated as: n = min(intPARTID, PARTID) /* control groups (closids) */ l = min(reqPARTID, PARTID) /* total reqPARTIDs */ m = l // n /* reqPARTIDs per control group */ Where: intPARTID: intPARTIDs on narrow-PARTID-capable MSCs reqPARTID: reqPARTIDs on narrow-PARTID-capable MSCs PARTID: PARTIDs on non-narrow-PARTID-capable MSCs Example: L3 cache (256 PARTIDs, without narrow PARTID feature) + MATA (32 intPARTIDs, 256 reqPARTIDs): n = min( 32, 256) = 32 intPARTIDs l = min(256, 256) = 256 reqPARTIDs m = 256 / 32 = 8 reqPARTIDs per intPARTID Implementation notes: * Handle mixed MSC systems (some support narrow PARTID, some don't) by taking minimum values across all MSCs. * Add get_num_reqpartid() for reqPARTID count used in RMID calculation. * resctrl_arch_get_num_closid() now returns intPARTID count (was PARTID). Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 12 +++++++++++- drivers/platform/mpam/mpam_internal.h | 2 ++ drivers/platform/mpam/mpam_resctrl.c | 9 +++++++-- 3 files changed, 20 insertions(+), 3 deletions(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 6c1c82f79ec04..42d1e5bbe810b 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -56,6 +56,7 @@ static DEFINE_MUTEX(mpam_cpuhp_state_lock); * Generating traffic outside this range will result in screaming interrupts. */ u16 mpam_partid_max; +u16 mpam_intpartid_max; u8 mpam_pmg_max; static bool partid_max_init, partid_max_published; static DEFINE_SPINLOCK(partid_max_lock); @@ -201,10 +202,16 @@ int mpam_register_requestor(u16 partid_max, u8 pmg_max) spin_lock(&partid_max_lock); if (!partid_max_init) { mpam_partid_max = partid_max; + /* + * Update mpam_intpartid_max here, in case the + * system doesn't have narrow-partid feature. + */ + mpam_intpartid_max = partid_max; mpam_pmg_max = pmg_max; partid_max_init = true; } else if (!partid_max_published) { mpam_partid_max = min(mpam_partid_max, partid_max); + mpam_intpartid_max = min(mpam_intpartid_max, partid_max); mpam_pmg_max = min(mpam_pmg_max, pmg_max); } else { /* New requestors can't lower the values */ @@ -741,7 +748,9 @@ static void mpam_ris_hw_probe(struct mpam_msc_ris *ris) u16 partid_max = FIELD_GET(MPAMF_PARTID_NRW_IDR_INTPARTID_MAX, nrwidr); mpam_set_feature(mpam_feat_partid_nrw, props); - msc->partid_max = min(msc->partid_max, partid_max); + msc->intpartid_max = min(msc->partid_max, partid_max); + } else { + msc->intpartid_max = msc->partid_max; } } @@ -806,6 +815,7 @@ static int mpam_msc_hw_probe(struct mpam_msc *msc) spin_lock(&partid_max_lock); mpam_partid_max = min(mpam_partid_max, msc->partid_max); + mpam_intpartid_max = min(mpam_intpartid_max, msc->intpartid_max); mpam_pmg_max = min(mpam_pmg_max, msc->pmg_max); spin_unlock(&partid_max_lock); diff --git a/drivers/platform/mpam/mpam_internal.h b/drivers/platform/mpam/mpam_internal.h index 5f60e0cc4eeb1..e4809f2242ec0 100644 --- a/drivers/platform/mpam/mpam_internal.h +++ b/drivers/platform/mpam/mpam_internal.h @@ -58,6 +58,7 @@ struct mpam_msc bool error_irq_requested; bool error_irq_hw_enabled; u16 partid_max; + u16 intpartid_max; u8 pmg_max; unsigned long ris_idxs[128 / BITS_PER_LONG]; u32 ris_max; @@ -303,6 +304,7 @@ extern struct srcu_struct mpam_srcu; /* System wide partid/pmg values */ extern u16 mpam_partid_max; +extern u16 mpam_intpartid_max; extern u8 mpam_pmg_max; /* Scheduled work callback to enable mpam once all MSC have been probed */ diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index de0f0136abe3c..6d6f328a7f716 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -147,6 +147,11 @@ static bool mpam_resctrl_hide_cdp(enum resctrl_res_level rid) * only the system wide safe value is safe to use. */ u32 resctrl_arch_get_num_closid(struct rdt_resource *ignored) +{ + return mpam_intpartid_max + 1; +} + +static u32 get_num_reqpartid(void) { return mpam_partid_max + 1; } @@ -154,9 +159,9 @@ u32 resctrl_arch_get_num_closid(struct rdt_resource *ignored) u32 resctrl_arch_system_num_rmid_idx(void) { u8 closid_shift = fls(mpam_pmg_max); - u32 num_partid = resctrl_arch_get_num_closid(NULL); + u32 num_reqpartid = get_num_reqpartid(); - return num_partid << closid_shift; + return num_reqpartid << closid_shift; } u32 resctrl_arch_rmid_idx_encode(u32 closid, u32 rmid) -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ The Narrow PARTID feature allows the MPAM driver to statically or dynamically allocate request PARTIDs (reqPARTIDs) to internal PARTIDs (intPARTIDs). This enables expanding the number of monitoring groups beyond the hardware PMG limit. For systems with mixed MSCs (Memory System Components), MSCs that do not support narrow PARTID use PARTIDs exceeding the minimum number of intPARTIDs as reqPARTIDs to expand monitoring groups. Expand RMID to include reqPARTID information: rmid = (reqPARTID << shift | PMG). To maintain compatibility with the existing resctrl layer, reqPARTIDs are allocated statically with a linear mapping to intPARTIDs via req2intpartid(). Mapping relationships (n = intPARTID count, m = reqPARTIDs per intPARTID): P - Partition group M - Monitoring group Group closid rmid.reqPARTID MSCs w/ narrow-PARTID MSCs w/o narrow-PARTID P1 0 intPARTID_1 PARTID_1 M1_1 0 0 ├── reqPARTID_1_1 ├── PARTID_1 M1_2 0 0+n ├── reqPARTID_1_2 ├── PARTID_1_2 M1_3 0 0+n*2 ├── reqPARTID_1_3 ├── PARTID_1_3 ... ├── ... ├── ... M1_m 0 0+n*(m-1) └── reqPARTID_1_m └── PARTID_1_m P2 1 intPARTID_2 PARTID_2 M2_1 1 1 ├── reqPARTID_2_1 ├── PARTID_2 M2_2 1 1+n ├── reqPARTID_2_2 ├── PARTID_2_2 M2_3 1 1+n*2 ├── reqPARTID_2_3 ├── PARTID_2_3 ... ├── ... ├── ... M2_m 1 1+n*(m-1) └── reqPARTID_2_m └── PARTID_2_m Pn n-1 intPARTID_n PARTID_n Mn_1 n-1 n-1 ├── reqPARTID_n_1 ├── PARTID_n Mn_2 n-1 n-1+n ├── reqPARTID_n_2 ├── PARTID_n_2 Mn_3 n-1 n-1+n*2 ├── reqPARTID_n_3 ├── PARTID_n_3 ... ├── ... ├── ... Mn_m n-1 n*m-1 └── reqPARTID_n_m └── PARTID_n_m Refactor the glue layer between resctrl abstractions (rmid) and MPAM hardware registers (reqPARTID/PMG) to support narrow PARTID. The resctrl layer uses rmid2reqpartid() and rmid2pmg() to extract components from rmid. The closid-to-intPARTID translation remains unchanged via resctrl_get_config_index(). Since narrow PARTID is a monitoring enhancement, reqPARTID is only used in monitoring paths while configuration paths maintain original semantics of closid. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- arch/arm64/include/asm/mpam.h | 17 +-- drivers/platform/mpam/mpam_resctrl.c | 160 ++++++++++++++++----------- include/linux/arm_mpam.h | 2 +- 3 files changed, 105 insertions(+), 74 deletions(-) diff --git a/arch/arm64/include/asm/mpam.h b/arch/arm64/include/asm/mpam.h index 5f1ac30ea4702..4d0a66013e3ff 100644 --- a/arch/arm64/include/asm/mpam.h +++ b/arch/arm64/include/asm/mpam.h @@ -129,18 +129,13 @@ static inline u64 mpam_get_regval(struct task_struct *tsk) #endif } -static inline void resctrl_arch_set_rmid(struct task_struct *tsk, u32 rmid) -{ -#ifdef CONFIG_ARM64_MPAM - u64 regval = mpam_get_regval(tsk); +u32 req2intpartid(u32 reqpartid); - regval &= ~MPAM_SYSREG_PMG_D; - regval &= ~MPAM_SYSREG_PMG_I; - regval |= FIELD_PREP(MPAM_SYSREG_PMG_D, rmid); - regval |= FIELD_PREP(MPAM_SYSREG_PMG_I, rmid); +static inline u32 mpam_get_regval_partid(u64 regval) +{ + u32 reqpartid = (regval & MPAM_SYSREG_PARTID_D) >> 16; - WRITE_ONCE(task_thread_info(tsk)->mpam_partid_pmg, regval); -#endif + return req2intpartid(reqpartid); } static inline void mpam_thread_switch(struct task_struct *tsk) @@ -153,7 +148,7 @@ static inline void mpam_thread_switch(struct task_struct *tsk) !static_branch_likely(&mpam_enabled)) return; - if (!regval) + if (!regval || !mpam_get_regval_partid(regval)) regval = READ_ONCE(per_cpu(arm64_mpam_default, cpu)); oldregval = READ_ONCE(per_cpu(arm64_mpam_current, cpu)); diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index 6d6f328a7f716..a2b0267502ef9 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -164,24 +164,73 @@ u32 resctrl_arch_system_num_rmid_idx(void) return num_reqpartid << closid_shift; } +static u32 rmid2reqpartid(u32 rmid) +{ + u8 pmg_shift = fls(mpam_pmg_max); + u32 reqpartid; + + WARN_ON_ONCE(pmg_shift > 8); + + rmid >>= pmg_shift; + + if (cdp_enabled) + reqpartid = resctrl_get_config_index(rmid, CDP_DATA); + else + reqpartid = resctrl_get_config_index(rmid, CDP_NONE); + + return reqpartid; +} + +static u8 rmid2pmg(u32 rmid) +{ + u8 pmg_shift = fls(mpam_pmg_max); + u32 pmg_mask = ~(~0 << pmg_shift); + + return rmid & pmg_mask; +} + +u32 req2intpartid(u32 reqpartid) +{ + u8 intpartid_shift = fls(mpam_intpartid_max); + u32 intpartid_mask = ~(~0 << intpartid_shift); + + return reqpartid & intpartid_mask; +} + +/* + * To avoid the reuse of rmid across multiple control groups, check + * the incoming closid to prevent rmid from being reallocated by + * resctrl_find_free_rmid(). + * + * If the closid and rmid do not match upon inspection, immediately + * returns an invalid rmid. A valid rmid must not exceed 24 bits. + */ u32 resctrl_arch_rmid_idx_encode(u32 closid, u32 rmid) { - u8 closid_shift = fls(mpam_pmg_max); + u32 reqpartid = rmid2reqpartid(rmid); + u32 intpartid = req2intpartid(reqpartid); + + if (cdp_enabled) + intpartid >>= 1; - BUG_ON(closid_shift > 8); + if (closid != intpartid) + return U32_MAX; - return (closid << closid_shift) | rmid; + return rmid; } void resctrl_arch_rmid_idx_decode(u32 idx, u32 *closid, u32 *rmid) { - u8 closid_shift = fls(mpam_pmg_max); - u32 pmg_mask = ~(~0 << closid_shift); - - BUG_ON(closid_shift > 8); + u32 reqpartid = rmid2reqpartid(idx); + u32 intpartid = req2intpartid(reqpartid); - *closid = idx >> closid_shift; - *rmid = idx & pmg_mask; + if (rmid) + *rmid = idx; + if (closid) { + if (cdp_enabled) + intpartid >>= 1; + *closid = intpartid; + } } void resctrl_sched_in(struct task_struct *tsk) @@ -191,23 +240,22 @@ void resctrl_sched_in(struct task_struct *tsk) mpam_thread_switch(tsk); } -void resctrl_arch_set_cpu_default_closid_rmid(int cpu, u32 closid, u32 pmg) +void resctrl_arch_set_cpu_default_closid_rmid(int cpu, u32 closid, u32 rmid) { - BUG_ON(closid > U16_MAX); - BUG_ON(pmg > U8_MAX); + u32 reqpartid = rmid2reqpartid(rmid); + u8 pmg = rmid2pmg(rmid); - if (!cdp_enabled) { - mpam_set_cpu_defaults(cpu, closid, closid, pmg, pmg); - } else { + WARN_ON_ONCE(reqpartid > U16_MAX); + WARN_ON_ONCE(pmg > U8_MAX); + + if (!cdp_enabled) + mpam_set_cpu_defaults(cpu, reqpartid, reqpartid, pmg, pmg); + else /* * When CDP is enabled, resctrl halves the closid range and we * use odd/even partid for one closid. */ - u32 partid_d = resctrl_get_config_index(closid, CDP_DATA); - u32 partid_i = resctrl_get_config_index(closid, CDP_CODE); - - mpam_set_cpu_defaults(cpu, partid_d, partid_i, pmg, pmg); - } + mpam_set_cpu_defaults(cpu, reqpartid, reqpartid + 1, pmg, pmg); } void resctrl_arch_sync_cpu_defaults(void *info) @@ -226,43 +274,40 @@ void resctrl_arch_sync_cpu_defaults(void *info) void resctrl_arch_set_closid_rmid(struct task_struct *tsk, u32 closid, u32 rmid) { + u32 reqpartid = rmid2reqpartid(rmid); + u8 pmg = rmid2pmg(rmid); + WARN_ON_ONCE(reqpartid > U16_MAX); + WARN_ON_ONCE(pmg > U8_MAX); - BUG_ON(closid > U16_MAX); - BUG_ON(rmid > U8_MAX); - - if (!cdp_enabled) { - mpam_set_task_partid_pmg(tsk, closid, closid, rmid, rmid); - } else { - u32 partid_d = resctrl_get_config_index(closid, CDP_DATA); - u32 partid_i = resctrl_get_config_index(closid, CDP_CODE); - - mpam_set_task_partid_pmg(tsk, partid_d, partid_i, rmid, rmid); - } + if (!cdp_enabled) + mpam_set_task_partid_pmg(tsk, reqpartid, reqpartid, pmg, pmg); + else + mpam_set_task_partid_pmg(tsk, reqpartid, reqpartid + 1, pmg, pmg); } bool resctrl_arch_match_closid(struct task_struct *tsk, u32 closid) { u64 regval = mpam_get_regval(tsk); - u32 tsk_closid = FIELD_GET(MPAM_SYSREG_PARTID_D, regval); + u32 tsk_partid = FIELD_GET(MPAM1_EL1_PARTID_D, regval); + + tsk_partid = req2intpartid(tsk_partid); if (cdp_enabled) - tsk_closid >>= 1; + tsk_partid >>= 1; - return tsk_closid == closid; + return tsk_partid == closid; } /* The task's pmg is not unique, the partid must be considered too */ bool resctrl_arch_match_rmid(struct task_struct *tsk, u32 closid, u32 rmid) { u64 regval = mpam_get_regval(tsk); - u32 tsk_closid = FIELD_GET(MPAM_SYSREG_PARTID_D, regval); - u32 tsk_rmid = FIELD_GET(MPAM_SYSREG_PMG_D, regval); - - if (cdp_enabled) - tsk_closid >>= 1; + u32 tsk_partid = FIELD_GET(MPAM1_EL1_PARTID_D, regval); + u32 tsk_pmg = FIELD_GET(MPAM1_EL1_PMG_D, regval); - return (tsk_closid == closid) && (tsk_rmid == rmid); + return (tsk_partid == rmid2reqpartid(rmid)) && + (tsk_pmg == rmid2pmg(rmid)); } #ifdef CONFIG_RESCTRL_IOMMU @@ -389,29 +434,24 @@ int resctrl_arch_rmid_read(struct rdt_resource *r, struct rdt_domain *d, num_mon = res->class->props.num_csu_mon; cfg.match_pmg = true; - cfg.pmg = rmid; + cfg.pmg = rmid2pmg(rmid); cfg.opts = resctrl_evt_config_to_mpam(dom->mbm_local_evt_cfg); + cfg.partid = rmid2reqpartid(rmid); - if (cdp_enabled) { - cfg.partid = resctrl_get_config_index(closid, CDP_DATA); - cfg.mon = cfg.partid % num_mon; - err = mpam_msmon_read(dom->comp, &cfg, type, val); - if (err) - return err; + cfg.mon = cfg.partid % num_mon; + err = mpam_msmon_read(dom->comp, &cfg, type, val); + if (err) + return err; - cfg.partid = resctrl_get_config_index(closid, CDP_CODE); + if (cdp_enabled) { + cfg.partid += 1; cfg.mon = cfg.partid % num_mon; err = mpam_msmon_read(dom->comp, &cfg, type, &cdp_val); if (!err) { - pr_debug("read monitor rmid %u %s:%u CODE/DATA: %lld/%lld\n", - resctrl_arch_rmid_idx_encode(closid, rmid), - r->name, dom->comp->comp_id, cdp_val, *val); + pr_debug("read monitor closid %u rmid %u %s:%u CODE/DATA: %lld/%lld\n", + closid, rmid, r->name, dom->comp->comp_id, cdp_val, *val); *val += cdp_val; } - } else { - cfg.partid = closid; - cfg.mon = cfg.partid % num_mon; - err = mpam_msmon_read(dom->comp, &cfg, type, val); } return err; @@ -433,19 +473,15 @@ void resctrl_arch_reset_rmid(struct rdt_resource *r, struct rdt_domain *d, num_mbwu_mon = res->class->props.num_mbwu_mon; cfg.mon = resctrl_arch_rmid_idx_encode(closid, rmid) % num_mbwu_mon; cfg.match_pmg = true; - cfg.pmg = rmid; + cfg.pmg = rmid2pmg(rmid); + cfg.partid = rmid2reqpartid(rmid); dom = container_of(d, struct mpam_resctrl_dom, resctrl_dom); + mpam_msmon_reset_mbwu(dom->comp, &cfg); if (cdp_enabled) { - cfg.partid = closid << 1; - mpam_msmon_reset_mbwu(dom->comp, &cfg); - cfg.partid += 1; mpam_msmon_reset_mbwu(dom->comp, &cfg); - } else { - cfg.partid = closid; - mpam_msmon_reset_mbwu(dom->comp, &cfg); } } diff --git a/include/linux/arm_mpam.h b/include/linux/arm_mpam.h index 5d60e1c6ca6c5..541c2f1bf6f4c 100644 --- a/include/linux/arm_mpam.h +++ b/include/linux/arm_mpam.h @@ -79,7 +79,7 @@ bool resctrl_arch_match_closid(struct task_struct *tsk, u32 closid); bool resctrl_arch_match_rmid(struct task_struct *tsk, u32 closid, u32 rmid); void resctrl_arch_set_cpu_default_closid(int cpu, u32 closid); void resctrl_arch_set_closid_rmid(struct task_struct *tsk, u32 closid, u32 rmid); -void resctrl_arch_set_cpu_default_closid_rmid(int cpu, u32 closid, u32 pmg); +void resctrl_arch_set_cpu_default_closid_rmid(int cpu, u32 closid, u32 rmid); void resctrl_sched_in(struct task_struct *tsk); u32 resctrl_arch_rmid_idx_encode(u32 closid, u32 rmid); void resctrl_arch_rmid_idx_decode(u32 idx, u32 *closid, u32 *rmid); -- 2.43.0
From: Dave Martin <Dave.Martin@arm.com> hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Currently, when an MSC implements PARTID narrowing, MPAMCFG_PART_SEL is left set to the reqPARTID while programming resource controls in an MSC. The MPAM architecture does not guarantee that any particular resource controls will be updated correctly in this scenario. Instead, MPAMCFG_PART_SEL must be written with the corresponding intPARTID and with the INTERNAL bit set before attempting to program resource controls. Only the PARTID->intPARTID mappings can be written without the INTERNAL bit set in MPAMCFG_PART_SEL. Fix it, by rewriting MPAMCFG_PART_SEL appropriately after setting the intPARTID mapping. The MPAMCFG_INTPARTID_INTERNAL flag is not currently accepted as input to the __mpam_part_sel() helper. In the interest of keeping the code clean, break this helper up. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: James Morse <james.morse@arm.com> Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 26 ++++++++++++++++++++------ 1 file changed, 20 insertions(+), 6 deletions(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 42d1e5bbe810b..a190be18fff8f 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -184,15 +184,27 @@ static u64 mpam_msc_read_esr(struct mpam_msc *msc) return (esr_high << 32) | esr_low; } +static void __mpam_part_sel_raw(u32 partsel, struct mpam_msc *msc) +{ + lockdep_assert_held(&msc->part_sel_lock); + mpam_write_partsel_reg(msc, PART_SEL, partsel); +} + static void __mpam_part_sel(u8 ris_idx, u16 partid, struct mpam_msc *msc) { - u32 partsel; + u32 partsel = FIELD_PREP(MPAMCFG_PART_SEL_RIS, ris_idx) | + FIELD_PREP(MPAMCFG_PART_SEL_PARTID_SEL, partid); - lockdep_assert_held(&msc->part_sel_lock); + __mpam_part_sel_raw(partsel, msc); +} - partsel = FIELD_PREP(MPAMCFG_PART_SEL_RIS, ris_idx) | - FIELD_PREP(MPAMCFG_PART_SEL_PARTID_SEL, partid); - mpam_write_partsel_reg(msc, PART_SEL, partsel); +static void __mpam_intpart_sel(u8 ris_idx, u16 intpartid, struct mpam_msc *msc) +{ + u32 partsel = FIELD_PREP(MPAMCFG_PART_SEL_RIS, ris_idx) | + FIELD_PREP(MPAMCFG_PART_SEL_PARTID_SEL, intpartid) | + MPAMCFG_PART_SEL_INTERNAL; + + __mpam_part_sel_raw(partsel, msc); } int mpam_register_requestor(u16 partid_max, u8 pmg_max) @@ -1364,9 +1376,11 @@ static void mpam_reprogram_ris_partid(struct mpam_msc_ris *ris, u16 partid, spin_lock(&msc->part_sel_lock); __mpam_part_sel(ris->ris_idx, partid, msc); - if(mpam_has_feature(mpam_feat_partid_nrw, rprops)) + if (mpam_has_feature(mpam_feat_partid_nrw, rprops)) { mpam_write_partsel_reg(msc, INTPARTID, (MPAMCFG_PART_SEL_INTERNAL | partid)); + __mpam_intpart_sel(ris->ris_idx, partid, msc); + } if (mpam_has_feature(mpam_feat_cpor_part, rprops)) { if (mpam_has_feature(mpam_feat_cpor_part, cfg)) -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ With the narrow PARTID feature, each control group is assigned multiple (req)PARTIDs to expand monitoring capacity. When a control group's configuration is updated, all associated sub-monitoring groups (each identified by a unique reqPARTID) should be synchronized. In __write_config(), iterate over all reqPARTIDs belonging to the control group and propagate the configuration to each sub-monitoring group: 1. For MSCs supporting narrow PARTID, establish the reqPARTID to intPARTID mapping. 2. For MSCs without narrow PARTID support, synchronize the configuration to new PARTIDs directly. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 29 ++++++++++++++++++++++++--- drivers/platform/mpam/mpam_internal.h | 2 ++ drivers/platform/mpam/mpam_resctrl.c | 5 +++++ 3 files changed, 33 insertions(+), 3 deletions(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index a190be18fff8f..7ebb9e5ebf578 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -1371,6 +1371,7 @@ static void mpam_reprogram_ris_partid(struct mpam_msc_ris *ris, u16 partid, u16 cmax = MPAMCFG_CMAX_CMAX; struct mpam_msc *msc = ris->msc; u16 bwa_fract = MPAMCFG_MBW_MAX_MAX; + u16 intpartid = req2intpartid(partid); struct mpam_props *rprops = &ris->props; spin_lock(&msc->part_sel_lock); @@ -1378,8 +1379,17 @@ static void mpam_reprogram_ris_partid(struct mpam_msc_ris *ris, u16 partid, if (mpam_has_feature(mpam_feat_partid_nrw, rprops)) { mpam_write_partsel_reg(msc, INTPARTID, - (MPAMCFG_PART_SEL_INTERNAL | partid)); - __mpam_intpart_sel(ris->ris_idx, partid, msc); + MPAMCFG_INTPARTID_INTERNAL | + intpartid); + + /* + * Mapping from reqpartid to intpartid already established. + * Sub-monitoring groups share the parent's configuration. + */ + if (partid != intpartid) + goto out; + + __mpam_intpart_sel(ris->ris_idx, intpartid, msc); } if (mpam_has_feature(mpam_feat_cpor_part, rprops)) { @@ -1463,6 +1473,7 @@ static void mpam_reprogram_ris_partid(struct mpam_msc_ris *ris, u16 partid, mpam_has_feature(mpam_feat_dspri_part, rprops)) mpam_write_partsel_reg(msc, PRI, pri_val); +out: spin_unlock(&msc->part_sel_lock); } @@ -2570,9 +2581,21 @@ struct mpam_write_config_arg { static int __write_config(void *arg) { + int closid_num = resctrl_arch_get_num_closid(NULL); struct mpam_write_config_arg *c = arg; + u32 reqpartid, req_idx; - mpam_reprogram_ris_partid(c->ris, c->partid, &c->comp->cfg[c->partid]); + /* c->partid should be within the range of intPARTIDs */ + WARN_ON_ONCE(c->partid >= closid_num); + + /* Synchronize the configuration to each sub-monitoring group. */ + for (req_idx = 0; req_idx < get_num_reqpartid_per_closid(); + req_idx++) { + reqpartid = req_idx * closid_num + c->partid; + + mpam_reprogram_ris_partid(c->ris, reqpartid, + &c->comp->cfg[c->partid]); + } return 0; } diff --git a/drivers/platform/mpam/mpam_internal.h b/drivers/platform/mpam/mpam_internal.h index e4809f2242ec0..adef0aae7cff0 100644 --- a/drivers/platform/mpam/mpam_internal.h +++ b/drivers/platform/mpam/mpam_internal.h @@ -594,4 +594,6 @@ bool mpam_cpbm_hisi_check_invalid(struct rdt_resource *r, unsigned long val); int mpam_resctrl_prepare_offline(void); +u32 get_num_reqpartid_per_closid(void); + #endif /* MPAM_INTERNAL_H */ diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index a2b0267502ef9..7786dc1c087af 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -156,6 +156,11 @@ static u32 get_num_reqpartid(void) return mpam_partid_max + 1; } +u32 get_num_reqpartid_per_closid(void) +{ + return get_num_reqpartid() / resctrl_arch_get_num_closid(NULL); +} + u32 resctrl_arch_system_num_rmid_idx(void) { u8 closid_shift = fls(mpam_pmg_max); -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Introduce helper functions for rmid_entry management, in preparation for upcoming patches supporting dynamic monitoring group allocation: - rmid_is_occupied(): Query whether a rmid_entry is currently allocated by checking if its list node has been removed from the free list. - rmid_entry_reassign_closid(): Update the closid associated with a rmid entry. Fix list node initialization in alloc_rmid() and dom_data_init() by using list_del_init() instead of list_del(). This ensures list_empty() checks in rmid_is_occupied() work correctly without encountering LIST_POISON values. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- fs/resctrl/monitor.c | 18 ++++++++++++++++-- include/linux/resctrl.h | 21 +++++++++++++++++++++ 2 files changed, 37 insertions(+), 2 deletions(-) diff --git a/fs/resctrl/monitor.c b/fs/resctrl/monitor.c index b651b37e7e657..df6839ebb2d18 100644 --- a/fs/resctrl/monitor.c +++ b/fs/resctrl/monitor.c @@ -265,7 +265,7 @@ int alloc_rmid(u32 closid) if (IS_ERR(entry)) return PTR_ERR(entry); - list_del(&entry->list); + list_del_init(&entry->list); return entry->rmid; } @@ -324,6 +324,13 @@ void free_rmid(u32 closid, u32 rmid) list_add_tail(&entry->list, &rmid_free_lru); } +bool rmid_is_occupied(u32 closid, u32 rmid) +{ + u32 idx = resctrl_arch_rmid_idx_encode(closid, rmid); + + return list_empty(&rmid_ptrs[idx].list); +} + static struct mbm_state *get_mbm_state(struct rdt_domain *d, u32 closid, u32 rmid, enum resctrl_event_id evtid) { @@ -733,6 +740,13 @@ void mbm_setup_overflow_handler(struct rdt_domain *dom, unsigned long delay_ms, schedule_delayed_work_on(cpu, &dom->mbm_over, delay); } +void rmid_entry_reassign_closid(u32 closid, u32 rmid) +{ + u32 idx = resctrl_arch_rmid_idx_encode(closid, rmid); + + rmid_ptrs[idx].closid = closid; +} + static int dom_data_init(struct rdt_resource *r) { u32 idx_limit = resctrl_arch_system_num_rmid_idx(); @@ -780,7 +794,7 @@ static int dom_data_init(struct rdt_resource *r) idx = resctrl_arch_rmid_idx_encode(RESCTRL_RESERVED_CLOSID, RESCTRL_RESERVED_RMID); entry = __rmid_entry(idx); - list_del(&entry->list); + list_del_init(&entry->list); out_unlock: mutex_unlock(&rdtgroup_mutex); diff --git a/include/linux/resctrl.h b/include/linux/resctrl.h index 74fc71d9a6bae..dd3c2b660b393 100644 --- a/include/linux/resctrl.h +++ b/include/linux/resctrl.h @@ -444,6 +444,27 @@ void resctrl_arch_reset_rmid_all(struct rdt_resource *r, struct rdt_domain *d); extern unsigned int resctrl_rmid_realloc_threshold; extern unsigned int resctrl_rmid_realloc_limit; +/** + * rmid_is_occupied() - Check whether the specified rmid has been + * allocated. + * @closid: Specify the closid that matches the rmid. + * @rmid: Specify the rmid entry to check status. + * + * This function checks if the rmid_entry is currently allocated by testing + * whether its list node is empty (removed from the free list). + * + * Return: + * True if the specified rmid is still in use. + */ +bool rmid_is_occupied(u32 closid, u32 rmid); + +/** + * rmid_entry_reassign_closid() - Update the closid field of a rmid_entry. + * @closid: Specify the reassigned closid. + * @rmid: Specify the rmid entry to update closid. + */ +void rmid_entry_reassign_closid(u32 closid, u32 rmid); + extern bool resctrl_mounted; int resctrl_init(void); -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Replace static reqPARTID allocation with dynamic binding to maximize monitoring group utilization. Static allocation wastes resources when control groups create fewer sub-groups than the pre-allocated limit. Add a lookup table (reqpartid_map) to dynamically bind reqPARTIDs to control groups needing extended monitoring capacity: - resctrl_arch_rmid_expand(): Find and bind a free reqPARTID to the specified closid when creating monitoring groups. - resctrl_arch_rmid_reclaim(): Unbind reqPARTID when all monitoring groups associated with pmg are freed, making it available for reuse. Update conversion helpers for dynamic mapping: - req2intpartid(): Use lookup table for dynamic entries - Add partid2closid() and req_pmg2rmid() helpers Refactor __write_config() to iterate over all reqPARTIDs that match by intPARTID, removing fixed per-closid slot assumption. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 16 ++- drivers/platform/mpam/mpam_internal.h | 2 +- drivers/platform/mpam/mpam_resctrl.c | 177 +++++++++++++++++++++----- 3 files changed, 154 insertions(+), 41 deletions(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 7ebb9e5ebf578..6ded9769ae5d2 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -2583,18 +2583,20 @@ static int __write_config(void *arg) { int closid_num = resctrl_arch_get_num_closid(NULL); struct mpam_write_config_arg *c = arg; - u32 reqpartid, req_idx; + u32 reqpartid; /* c->partid should be within the range of intPARTIDs */ WARN_ON_ONCE(c->partid >= closid_num); - /* Synchronize the configuration to each sub-monitoring group. */ - for (req_idx = 0; req_idx < get_num_reqpartid_per_closid(); - req_idx++) { - reqpartid = req_idx * closid_num + c->partid; + mpam_reprogram_ris_partid(c->ris, c->partid, + &c->comp->cfg[c->partid]); - mpam_reprogram_ris_partid(c->ris, reqpartid, - &c->comp->cfg[c->partid]); + /* Synchronize the configuration to each sub-monitoring group. */ + for (reqpartid = closid_num; + reqpartid < get_num_reqpartid(); reqpartid++) { + if (req2intpartid(reqpartid) == c->partid) + mpam_reprogram_ris_partid(c->ris, reqpartid, + &c->comp->cfg[c->partid]); } return 0; diff --git a/drivers/platform/mpam/mpam_internal.h b/drivers/platform/mpam/mpam_internal.h index adef0aae7cff0..74abba9497ceb 100644 --- a/drivers/platform/mpam/mpam_internal.h +++ b/drivers/platform/mpam/mpam_internal.h @@ -594,6 +594,6 @@ bool mpam_cpbm_hisi_check_invalid(struct rdt_resource *r, unsigned long val); int mpam_resctrl_prepare_offline(void); -u32 get_num_reqpartid_per_closid(void); +u32 get_num_reqpartid(void); #endif /* MPAM_INTERNAL_H */ diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index 7786dc1c087af..8ccf5a84f954d 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -151,16 +151,11 @@ u32 resctrl_arch_get_num_closid(struct rdt_resource *ignored) return mpam_intpartid_max + 1; } -static u32 get_num_reqpartid(void) +u32 get_num_reqpartid(void) { return mpam_partid_max + 1; } -u32 get_num_reqpartid_per_closid(void) -{ - return get_num_reqpartid() / resctrl_arch_get_num_closid(NULL); -} - u32 resctrl_arch_system_num_rmid_idx(void) { u8 closid_shift = fls(mpam_pmg_max); @@ -194,12 +189,37 @@ static u8 rmid2pmg(u32 rmid) return rmid & pmg_mask; } +static u32 req_pmg2rmid(u32 reqpartid, u8 pmg) +{ + u8 pmg_shift = fls(mpam_pmg_max); + u32 pmg_mask = ~(~0 << pmg_shift); + + if (cdp_enabled) + reqpartid >>= 1; + + return (reqpartid << pmg_shift) | (pmg & pmg_mask); +} + +static u32 *reqpartid_map; + u32 req2intpartid(u32 reqpartid) { - u8 intpartid_shift = fls(mpam_intpartid_max); - u32 intpartid_mask = ~(~0 << intpartid_shift); + /* + * Directly return intPartid in case that mpam_reset_ris() access + * NULL pointer. + */ + if (reqpartid < resctrl_arch_get_num_closid(NULL)) + return reqpartid; + + return reqpartid_map[reqpartid]; +} + +static u32 partid2closid(u32 partid) +{ + if (cdp_enabled) + partid >>= 1; - return reqpartid & intpartid_mask; + return partid; } /* @@ -213,12 +233,12 @@ u32 req2intpartid(u32 reqpartid) u32 resctrl_arch_rmid_idx_encode(u32 closid, u32 rmid) { u32 reqpartid = rmid2reqpartid(rmid); - u32 intpartid = req2intpartid(reqpartid); - if (cdp_enabled) - intpartid >>= 1; + /* When enable CDP mode, needs to filter invalid rmid entry out */ + if (reqpartid >= get_num_reqpartid()) + return U32_MAX; - if (closid != intpartid) + if (closid != partid2closid(req2intpartid(reqpartid))) return U32_MAX; return rmid; @@ -231,11 +251,9 @@ void resctrl_arch_rmid_idx_decode(u32 idx, u32 *closid, u32 *rmid) if (rmid) *rmid = idx; - if (closid) { - if (cdp_enabled) - intpartid >>= 1; - *closid = intpartid; - } + + if (closid) + *closid = partid2closid(intpartid); } void resctrl_sched_in(struct task_struct *tsk) @@ -1191,6 +1209,87 @@ static int mpam_resctrl_resource_init(struct mpam_resctrl_res *res) return 0; } +static int reqpartid_init(void) +{ + int req_num, idx; + + req_num = get_num_reqpartid(); + reqpartid_map = kcalloc(req_num, sizeof(u32), GFP_KERNEL); + if (!reqpartid_map) + return -ENOMEM; + + for (idx = 0; idx < req_num; idx++) + reqpartid_map[idx] = idx; + + return 0; +} + +void reqpartid_exit(void) +{ + kfree(reqpartid_map); +} + +void update_rmid_entries_for_reqpartid(u32 reqpartid) +{ + int pmg; + u32 intpartid = reqpartid_map[reqpartid]; + u32 closid = partid2closid(intpartid); + + for (pmg = 0; pmg <= mpam_pmg_max; pmg++) + rmid_entry_reassign_closid(closid, req_pmg2rmid(reqpartid, pmg)); +} + +int resctrl_arch_rmid_expand(u32 closid) +{ + int i; + + for (i = resctrl_arch_get_num_closid(NULL); + i < get_num_reqpartid(); i++) { + if (reqpartid_map[i] >= resctrl_arch_get_num_closid(NULL)) { + if (cdp_enabled) { + reqpartid_map[i] = resctrl_get_config_index(closid, CDP_DATA); + reqpartid_map[i + 1] = resctrl_get_config_index(closid, CDP_CODE); + } else { + reqpartid_map[i] = resctrl_get_config_index(closid, CDP_NONE); + } + update_rmid_entries_for_reqpartid(i); + return i; + } + } + + return -ENOSPC; +} + +void resctrl_arch_rmid_reclaim(u32 closid, u32 rmid) +{ + int pmg; + u32 intpartid; + int reqpartid = rmid2reqpartid(rmid); + + if (reqpartid < resctrl_arch_get_num_closid(NULL)) + return; + + if (cdp_enabled) + intpartid = resctrl_get_config_index(closid, CDP_DATA); + else + intpartid = resctrl_get_config_index(closid, CDP_NONE); + + WARN_ON_ONCE(intpartid != req2intpartid(reqpartid)); + + for (pmg = 0; pmg <= mpam_pmg_max; pmg++) { + if (rmid_is_occupied(closid, req_pmg2rmid(reqpartid, pmg))) + break; + } + + if (pmg > mpam_pmg_max) { + reqpartid_map[reqpartid] = reqpartid; + if (cdp_enabled) + reqpartid_map[reqpartid + 1] = reqpartid + 1; + + update_rmid_entries_for_reqpartid(reqpartid); + } +} + int mpam_resctrl_setup(void) { int err = 0; @@ -1222,24 +1321,35 @@ int mpam_resctrl_setup(void) } cpus_read_unlock(); - if (!err && !exposed_alloc_capable && !exposed_mon_capable) - err = -EOPNOTSUPP; - - if (!err) { - if (!is_power_of_2(mpam_pmg_max + 1)) { - /* - * If not all the partid*pmg values are valid indexes, - * resctrl may allocate pmg that don't exist. This - * should cause an error interrupt. - */ - pr_warn("Number of PMG is not a power of 2! resctrl may misbehave"); - } + if (err) + return err; + + if (!exposed_alloc_capable && !exposed_mon_capable) + return -EOPNOTSUPP; - err = resctrl_init(); - if (!err) - WRITE_ONCE(resctrl_enabled, true); + err = reqpartid_init(); + if (err) + return err; + + if (!is_power_of_2(mpam_pmg_max + 1)) { + /* + * If not all the partid*pmg values are valid indexes, + * resctrl may allocate pmg that don't exist. This + * should cause an error interrupt. + */ + pr_warn("Number of PMG is not a power of 2! resctrl may misbehave"); } + err = resctrl_init(); + if (err) + goto out; + + WRITE_ONCE(resctrl_enabled, true); + + return 0; + +out: + reqpartid_exit(); return err; } @@ -1250,6 +1360,7 @@ void mpam_resctrl_exit(void) WRITE_ONCE(resctrl_enabled, false); resctrl_exit(); + reqpartid_exit(); } u32 resctrl_arch_get_config(struct rdt_resource *r, struct rdt_domain *d, -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ The previous patch implemented resctrl_arch_rmid_expand() and resctrl_arch_rmid_reclaim() for ARM MPAM. This patch integrates these architecture-specific functions into the generic resctrl layer. Refactor resctrl_find_free_rmid() to support dynamic rmid expansion. If no free rmid is available for the current closid, attempt to expand via resctrl_arch_expand_rmid(). On success, retry the rmid allocation. As this capability is architecture-specific, x86 maintains existing behavior by returning -ENOSPC when rmid resources are exhausted. Additionally, invoke resctrl_arch_rmid_reclaim() when rmids are released to enable architecture-specific resource cleanup. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- fs/resctrl/monitor.c | 32 ++++++++++++++++++++++++++++++-- include/linux/arm_mpam.h | 17 +++++++++++++++++ 2 files changed, 47 insertions(+), 2 deletions(-) diff --git a/fs/resctrl/monitor.c b/fs/resctrl/monitor.c index df6839ebb2d18..e5eb971ee39da 100644 --- a/fs/resctrl/monitor.c +++ b/fs/resctrl/monitor.c @@ -115,6 +115,8 @@ static void limbo_release_entry(struct rmid_entry *entry) if (IS_ENABLED(CONFIG_RESCTRL_RMID_DEPENDS_ON_CLOSID)) closid_num_dirty_rmid[entry->closid]--; + + resctrl_arch_rmid_reclaim(entry->closid, entry->rmid); } /* @@ -178,7 +180,7 @@ bool has_busy_rmid(struct rdt_domain *d) return find_first_bit(d->rmid_busy_llc, idx_limit) != idx_limit; } -struct rmid_entry *resctrl_find_free_rmid(u32 closid) +static struct rmid_entry *__resctrl_find_free_rmid(u32 closid) { struct rmid_entry *itr; u32 itr_idx, cmp_idx; @@ -195,7 +197,12 @@ struct rmid_entry *resctrl_find_free_rmid(u32 closid) * very first entry will be returned. */ itr_idx = resctrl_arch_rmid_idx_encode(itr->closid, itr->rmid); + if (itr_idx == U32_MAX) + continue; + cmp_idx = resctrl_arch_rmid_idx_encode(closid, itr->rmid); + if (cmp_idx == U32_MAX) + continue; if (itr_idx == cmp_idx) return itr; @@ -204,6 +211,25 @@ struct rmid_entry *resctrl_find_free_rmid(u32 closid) return ERR_PTR(-ENOSPC); } +struct rmid_entry *resctrl_find_free_rmid(u32 closid) +{ + struct rmid_entry *err; + int ret; + + err = __resctrl_find_free_rmid(closid); + if (err == ERR_PTR(-ENOSPC)) { + ret = resctrl_arch_rmid_expand(closid); + if (ret < 0) + /* Out of rmid */ + goto out; + + /* Try it again */ + return __resctrl_find_free_rmid(closid); + } +out: + return err; +} + /** * resctrl_find_cleanest_closid() - Find a CLOSID where all the associated * RMID are clean, or the CLOSID that has @@ -320,8 +346,10 @@ void free_rmid(u32 closid, u32 rmid) if (resctrl_arch_is_llc_occupancy_enabled()) add_rmid_to_limbo(entry); - else + else { list_add_tail(&entry->list, &rmid_free_lru); + resctrl_arch_rmid_reclaim(closid, rmid); + } } bool rmid_is_occupied(u32 closid, u32 rmid) diff --git a/include/linux/arm_mpam.h b/include/linux/arm_mpam.h index 541c2f1bf6f4c..a1a511d07eec1 100644 --- a/include/linux/arm_mpam.h +++ b/include/linux/arm_mpam.h @@ -77,6 +77,23 @@ bool resctrl_arch_get_cdp_enabled(enum resctrl_res_level ignored); int resctrl_arch_set_cdp_enabled(enum resctrl_res_level ignored, bool enable); bool resctrl_arch_match_closid(struct task_struct *tsk, u32 closid); bool resctrl_arch_match_rmid(struct task_struct *tsk, u32 closid, u32 rmid); + +/** + * resctrl_arch_rmid_expand() - Expand the RMID resources for the specified closid. + * @closid: closid that matches the rmid. + * + * Return: + * 0 on success, or -ENOSPC etc on error. + */ +int resctrl_arch_rmid_expand(u32 closid); + +/** + * resctrl_arch_rmid_reclaim() - Reclaim the rmid resources for the specified closid. + * @closid: closid that matches the rmid. + * @rmid: Reclaim the rmid specified. + */ +void resctrl_arch_rmid_reclaim(u32 closid, u32 rmid); + void resctrl_arch_set_cpu_default_closid(int cpu, u32 closid); void resctrl_arch_set_closid_rmid(struct task_struct *tsk, u32 closid, u32 rmid); void resctrl_arch_set_cpu_default_closid_rmid(int cpu, u32 closid, u32 rmid); -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Add mpam_sync_config() to synchronize configuration when dynamically expanding rmid resources. When binding a new reqpartid to a control group, the driver maps the reqpartid to the corresponding intpartid or applies the control group's existing configuration to new partid if without Narrow partid feature. Extend mpam_apply_config() with a sync mode: - Sync mode: mpam_sync_config() calls this to apply existing configuration without updating config. - Non-sync mode: resctrl_arch_update_one() calls this to compare, update, and apply configuration. This mode retains the original behavior. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 25 +++++++++++++++++++---- drivers/platform/mpam/mpam_internal.h | 2 +- drivers/platform/mpam/mpam_resctrl.c | 29 ++++++++++++++++++++++++--- 3 files changed, 48 insertions(+), 8 deletions(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 6ded9769ae5d2..62b7ae418a882 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -2577,6 +2577,7 @@ struct mpam_write_config_arg { struct mpam_msc_ris *ris; struct mpam_component *comp; u16 partid; + bool sync; }; static int __write_config(void *arg) @@ -2585,6 +2586,15 @@ static int __write_config(void *arg) struct mpam_write_config_arg *c = arg; u32 reqpartid; + if (c->sync) { + /* c->partid should be within the range of reqPARTIDs */ + WARN_ON_ONCE(c->partid < closid_num); + + mpam_reprogram_ris_partid(c->ris, c->partid, + &c->comp->cfg[req2intpartid(c->partid)]); + return 0; + } + /* c->partid should be within the range of intPARTIDs */ WARN_ON_ONCE(c->partid >= closid_num); @@ -2605,7 +2615,7 @@ static int __write_config(void *arg) /* TODO: split into write_config/sync_config */ /* TODO: add config_dirty bitmap to drive sync_config */ int mpam_apply_config(struct mpam_component *comp, u16 partid, - struct mpam_config *cfg) + struct mpam_config *cfg, bool sync) { struct mpam_write_config_arg arg; struct mpam_msc_ris *ris; @@ -2613,12 +2623,19 @@ int mpam_apply_config(struct mpam_component *comp, u16 partid, lockdep_assert_cpus_held(); - if (!memcmp(&comp->cfg[partid], cfg, sizeof(*cfg))) - return 0; + if (!sync) { + /* The partid is within the range of intPARTIDs */ + WARN_ON_ONCE(partid >= resctrl_arch_get_num_closid(NULL)); + + if (!memcmp(&comp->cfg[partid], cfg, sizeof(*cfg))) + return 0; + + comp->cfg[partid] = *cfg; + } - comp->cfg[partid] = *cfg; arg.comp = comp; arg.partid = partid; + arg.sync = sync; idx = srcu_read_lock(&mpam_srcu); list_for_each_entry_rcu(ris, &comp->ris, comp_list) { diff --git a/drivers/platform/mpam/mpam_internal.h b/drivers/platform/mpam/mpam_internal.h index 74abba9497ceb..ebb6d7357a980 100644 --- a/drivers/platform/mpam/mpam_internal.h +++ b/drivers/platform/mpam/mpam_internal.h @@ -314,7 +314,7 @@ void mpam_disable(struct work_struct *work); void mpam_reset_class(struct mpam_class *class); int mpam_apply_config(struct mpam_component *comp, u16 partid, - struct mpam_config *cfg); + struct mpam_config *cfg, bool sync); int mpam_msmon_read(struct mpam_component *comp, struct mon_cfg *ctx, enum mpam_device_features, u64 *val); diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index 8ccf5a84f954d..ad8ed64af4153 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -1239,6 +1239,23 @@ void update_rmid_entries_for_reqpartid(u32 reqpartid) rmid_entry_reassign_closid(closid, req_pmg2rmid(reqpartid, pmg)); } +static int mpam_sync_config(u32 reqpartid) +{ + struct mpam_component *comp; + struct mpam_class *class; + int err; + + list_for_each_entry(class, &mpam_classes, classes_list) { + list_for_each_entry(comp, &class->components, class_list) { + err = mpam_apply_config(comp, reqpartid, NULL, true); + if (err) + return err; + } + } + + return 0; +} + int resctrl_arch_rmid_expand(u32 closid) { int i; @@ -1248,10 +1265,16 @@ int resctrl_arch_rmid_expand(u32 closid) if (reqpartid_map[i] >= resctrl_arch_get_num_closid(NULL)) { if (cdp_enabled) { reqpartid_map[i] = resctrl_get_config_index(closid, CDP_DATA); + mpam_sync_config(i); + reqpartid_map[i + 1] = resctrl_get_config_index(closid, CDP_CODE); + mpam_sync_config(i + 1); + } else { reqpartid_map[i] = resctrl_get_config_index(closid, CDP_NONE); + mpam_sync_config(i); } + update_rmid_entries_for_reqpartid(i); return i; } @@ -1544,15 +1567,15 @@ int resctrl_arch_update_one(struct rdt_resource *r, struct rdt_domain *d, */ if (mpam_resctrl_hide_cdp(r->rid)) { partid = resctrl_get_config_index(closid, CDP_CODE); - err = mpam_apply_config(dom->comp, partid, &cfg); + err = mpam_apply_config(dom->comp, partid, &cfg, false); if (err) return err; partid = resctrl_get_config_index(closid, CDP_DATA); - return mpam_apply_config(dom->comp, partid, &cfg); + return mpam_apply_config(dom->comp, partid, &cfg, false); } else { - return mpam_apply_config(dom->comp, partid, &cfg); + return mpam_apply_config(dom->comp, partid, &cfg, false); } } -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Reorder rdt_kill_sb() to prevent rmid parsing errors caused by premature cdp_enabled reset. The rmid to reqpartid conversion depends on cdp_enabled state, but rdt_disable_ctx() clears this flag. Delay rdt_disable_ctx() until after rmdir_all_sub() completes, ensuring free_rmid() operates with correct CDP state. Additionally, flush all pending limbo work before umount completes. This prevents rmid release errors on subsequent mounts if CDP state changes between mount sessions. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- fs/resctrl/rdtgroup.c | 24 ++++++++++++++++++++++-- 1 file changed, 22 insertions(+), 2 deletions(-) diff --git a/fs/resctrl/rdtgroup.c b/fs/resctrl/rdtgroup.c index 642f91435d810..428e6538072e0 100644 --- a/fs/resctrl/rdtgroup.c +++ b/fs/resctrl/rdtgroup.c @@ -3000,17 +3000,37 @@ static void rmdir_all_sub(void) kernfs_remove(kn_mondata); } +static void rdt_flush_limbo(void) +{ + struct rdt_resource *r = resctrl_arch_get_resource(RDT_RESOURCE_L3); + struct rdt_domain *d; + + if (!IS_ENABLED(CONFIG_RESCTRL_RMID_DEPENDS_ON_CLOSID)) + return; + + if (!resctrl_arch_is_llc_occupancy_enabled()) + return; + + list_for_each_entry(d, &r->domains, list) { + if (has_busy_rmid(d)) { + __check_limbo(d, true); + cancel_delayed_work(&d->cqm_limbo); + } + } +} + static void rdt_kill_sb(struct super_block *sb) { cpus_read_lock(); mutex_lock(&rdtgroup_mutex); - rdt_disable_ctx(); - /* Put everything back to default values. */ resctrl_arch_reset_resources(); rmdir_all_sub(); + rdt_flush_limbo(); + rdt_disable_ctx(); + if (IS_ENABLED(CONFIG_RESCTRL_FS_PSEUDO_LOCK)) rdt_pseudo_lock_release(); rdtgroup_default.mode = RDT_MODE_SHAREABLE; -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ MPAM supports heterogeneous systems where some type of MSCs may implement Narrow-PARTID while others do not. However, when an MSC uses percentage-based throttling (non-bitmap partition control) and lacks Narrow-PARTID support, resctrl cannot correctly apply control group configurations across multiple PARTIDs. To enable free assignment of multiple reqPARTIDs to resource control groups, all MSCs used by resctrl must either: Implement Narrow-PARTID, allowing explicit PARTID remapping, or only have stateless resource controls (non-percentage-based), such that splitting a control group across multiple PARTIDs does not affect behavior. The detection occurs at initialization time on the first call to get_num_reqpartid() from update_rmid_limits(). This call is guaranteed to occur after mpam_resctrl_pick_{mba,caches}() have set up the resource classes, ensuring the necessary properties are available for the Narrow-PARTID capability check. When an MSC with percentage-based control lacks Narrow-PARTID support, get_num_reqpartid() falls back to returning the number of intPARTIDs, effectively disabling the reqPARTID expansion for monitoring groups. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_resctrl.c | 75 +++++++++++++++++++++------- 1 file changed, 56 insertions(+), 19 deletions(-) diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index ad8ed64af4153..e3fd8361c22d9 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -151,17 +151,46 @@ u32 resctrl_arch_get_num_closid(struct rdt_resource *ignored) return mpam_intpartid_max + 1; } +/* + * Determine the effective number of PARTIDs available for resctrl. + * + * This function performs a one-time check to determine if Narrow-PARTID + * can be used. It must be called after mpam_resctrl_pick_{mba,caches}() + * have initialized the resource classes, as class properties are used + * to detect Narrow-PARTID support. + * + * The first call occurs in update_rmid_limits(), ensuring the + * prerequisite initialization is complete. + */ u32 get_num_reqpartid(void) { + struct mpam_props *cprops; + struct mpam_class *class; + static bool first = true; + + if (first) { + list_for_each_entry_rcu(class, &mpam_classes, classes_list) { + cprops = &class->props; + if (mpam_has_feature(mpam_feat_partid_nrw, cprops)) + continue; + + if (mpam_has_feature(mpam_feat_mbw_max, cprops) || + mpam_has_feature(mpam_feat_mbw_min, cprops) || + mpam_has_feature(mpam_feat_ccap_part, cprops) || + mpam_has_feature(mpam_feat_cmin, cprops)) { + mpam_partid_max = mpam_intpartid_max; + break; + } + } + } + + first = false; return mpam_partid_max + 1; } u32 resctrl_arch_system_num_rmid_idx(void) { - u8 closid_shift = fls(mpam_pmg_max); - u32 num_reqpartid = get_num_reqpartid(); - - return num_reqpartid << closid_shift; + return (mpam_pmg_max + 1) * get_num_reqpartid(); } static u32 rmid2reqpartid(u32 rmid) @@ -512,9 +541,13 @@ void resctrl_arch_reset_rmid(struct rdt_resource *r, struct rdt_domain *d, * The rmid realloc threshold should be for the smallest cache exposed to * resctrl. */ -static void update_rmid_limits(unsigned int size) +static void update_rmid_limits(struct mpam_class *class) { u32 num_unique_pmg = resctrl_arch_system_num_rmid_idx(); + unsigned int size; + + /* Assume cache levels are the same size for all CPUs... */ + size = get_cpu_cacheinfo_size(smp_processor_id(), class->level); if (WARN_ON_ONCE(!size)) return; @@ -773,7 +806,6 @@ static u16 ca_max_to_percent(u16 ca_max, u8 wd) static void mpam_resctrl_pick_caches(void) { int idx; - unsigned int cache_size; struct mpam_class *class; struct mpam_resctrl_res *res; bool has_cpor, has_cmax, has_cmin, has_intpri; @@ -813,18 +845,6 @@ static void mpam_resctrl_pick_caches(void) continue; } - /* Assume cache levels are the same size for all CPUs... */ - cache_size = get_cpu_cacheinfo_size(smp_processor_id(), class->level); - if (!cache_size) { - pr_debug("pick_caches: Could not read cache size\n"); - continue; - } - - if (mpam_has_feature(mpam_feat_msmon_csu, cprops)) { - if (class->level == 3) - update_rmid_limits(cache_size); - } - if (has_cpor) { if (class->level == 2) { res = &mpam_resctrl_exports[RDT_RESOURCE_L2]; @@ -923,6 +943,23 @@ static void mpam_resctrl_pick_mba(void) srcu_read_unlock(&mpam_srcu, idx); } +static void mpam_resctrl_pick_counters(void) +{ + struct mpam_class *class; + int idx; + + idx = srcu_read_lock(&mpam_srcu); + + list_for_each_entry_rcu(class, &mpam_classes, classes_list) { + if (mpam_has_feature(mpam_feat_msmon_csu, &class->props)) { + if (class->level == 3) + update_rmid_limits(class); + } + } + + srcu_read_unlock(&mpam_srcu, idx); +} + bool resctrl_arch_is_evt_configurable(enum resctrl_event_id evt) { struct mpam_props *cprops; @@ -1331,7 +1368,7 @@ int mpam_resctrl_setup(void) mpam_resctrl_pick_caches(); mpam_resctrl_pick_mba(); - /* TODO: mpam_resctrl_pick_counters(); */ + mpam_resctrl_pick_counters(); for (i = 0; i < RDT_NUM_RESOURCES; i++) { res = &mpam_resctrl_exports[i]; -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Add support for the MBWU (Memory Bandwidth Usage) counter scaling feature as defined in the ARM MPAM specification. This allows hardware to report bandwidth usage with granularity by applying a scaling factor to raw counter values. The scaling factor is extracted from MPAMF_MBWUMON_IDR.SCALE field during hardware probing. When scale bit is enabled, extend the measurable range by 2^scale while reducing granularity accordingly. The scaling is applied to: - Overflow threshold calculations - Raw counter value readings This feature is mutually exclusive with the "long" counter format, as both provide alternative mechanisms for extending the effective range of bandwidth measurements. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 21 ++++++++++++++++----- drivers/platform/mpam/mpam_internal.h | 3 +++ 2 files changed, 19 insertions(+), 5 deletions(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 62b7ae418a882..3e875f29bd03b 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -745,6 +745,14 @@ static void mpam_ris_hw_probe(struct mpam_msc_ris *ris) else mpam_set_feature(mpam_feat_msmon_mbwu_44counter, props); } + + if (!has_long) { + props->mbwu_scale = FIELD_GET(MPAMF_MBWUMON_IDR_SCALE, mbwumonidr); + if (props->mbwu_scale) + mpam_set_feature(mpam_feat_msmon_mbwu_scale, props); + } else { + props->mbwu_scale = 0; + } } } @@ -909,6 +917,9 @@ static void gen_msmon_ctl_flt_vals(struct mon_read *m, u32 *ctl_val, */ *ctl_val |= MSMON_CFG_x_CTL_MATCH_PARTID; + if (mpam_has_feature(mpam_feat_msmon_mbwu_scale, &m->ris->props)) + *ctl_val |= MSMON_CFG_x_CTL_SCLEN; + *flt_val = FIELD_PREP(MSMON_CFG_MBWU_FLT_PARTID, ctx->partid); *flt_val |= FIELD_PREP(MSMON_CFG_MBWU_FLT_RWBW, ctx->opts); if (m->ctx->match_pmg) { @@ -995,7 +1006,8 @@ static u64 mpam_msmon_overflow_val(struct mpam_msc_ris *ris) else if (mpam_has_feature(mpam_feat_msmon_mbwu_44counter, &ris->props)) return GENMASK_ULL(43, 0); else - return GENMASK_ULL(30, 0); + /* Only non-long MBWU counter enables scale */ + return GENMASK_ULL(30, 0) << ris->props.mbwu_scale; } bool resctrl_arch_would_mbm_overflow(void) @@ -1044,7 +1056,7 @@ static void __ris_msmon_read(void *arg) unsigned long flags; bool config_mismatch; struct mon_read *m = arg; - u64 now, overflow_val = 0; + u64 now; bool mbwu_overflow = false; struct mon_cfg *ctx = m->ctx; bool reset_on_next_read = false; @@ -1120,6 +1132,7 @@ static void __ris_msmon_read(void *arg) now = mpam_read_monsel_reg(msc, MBWU); nrdy = now & MSMON___NRDY; now = FIELD_GET(MSMON___VALUE, now); + now <<= ris->props.mbwu_scale; } if (config_mismatch && !mpam_ris_has_nrdy_bit(ris)) @@ -1145,9 +1158,7 @@ static void __ris_msmon_read(void *arg) /* Add any pre-overflow value to the mbwu_state->val */ if (mbwu_overflow) - overflow_val = mpam_msmon_overflow_val(ris); - - mbwu_state->correction += overflow_val; + mbwu_state->correction += mpam_msmon_overflow_val(ris); /* Include bandwidth consumed before the last hardware reset */ now += mbwu_state->correction; diff --git a/drivers/platform/mpam/mpam_internal.h b/drivers/platform/mpam/mpam_internal.h index ebb6d7357a980..70d4abb93fd19 100644 --- a/drivers/platform/mpam/mpam_internal.h +++ b/drivers/platform/mpam/mpam_internal.h @@ -110,6 +110,7 @@ enum mpam_device_features { mpam_feat_msmon_mbwu_63counter, mpam_feat_msmon_mbwu_capture, mpam_feat_msmon_mbwu_rwbw, + mpam_feat_msmon_mbwu_scale, mpam_feat_msmon_capt, mpam_feat_partid_nrw, MPAM_FEATURE_LAST, @@ -128,6 +129,7 @@ struct mpam_props u16 dspri_wd; u16 num_csu_mon; u16 num_mbwu_mon; + u8 mbwu_scale; }; #define mpam_has_feature(_feat, x) ((1<<_feat) & (x)->features) @@ -434,6 +436,7 @@ bool mpam_cpbm_hisi_check_invalid(struct rdt_resource *r, unsigned long val); /* MPAMF_MBWUMON_IDR - MPAM memory bandwidth usage monitor ID register */ #define MPAMF_MBWUMON_IDR_NUM_MON GENMASK(15, 0) +#define MPAMF_MBWUMON_IDR_SCALE GENMASK(20, 16) #define MPAMF_MBWUMON_IDR_HAS_RWBW BIT(28) #define MPAMF_MBWUMON_IDR_LWD BIT(29) #define MPAMF_MBWUMON_IDR_HAS_LONG BIT(30) -- 2.43.0
hulk inclusion category: bugfix bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ When reparenting a monitor group (RDTMON_GROUP) to a new control group (RDTCTRL_GROUP), mongrp_reparent() updates rdtgrp->closid to the new parent's closid before calling rdt_move_group_tasks(). However, rdt_move_group_tasks() expects the first argument to be the *original* group (to identify tasks by their current closid) and the second argument to be the *destination* group (to update tasks to the new closid). By passing rdtgrp for both arguments after modifying its closid, the function fails to find any tasks to migrate because it searches for the new closid instead of the old one. Fix this by saving a copy of the original rdtgrp state before modifying it, then passing the copy as the source and the updated rdtgrp as the destination to rdt_move_group_tasks(). Fixes: 43132396266d ("fs/resctrl: Re-allocate rmid for the monitor when migrating across control groups") Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- fs/resctrl/rdtgroup.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/resctrl/rdtgroup.c b/fs/resctrl/rdtgroup.c index 428e6538072e0..16c4f9311111d 100644 --- a/fs/resctrl/rdtgroup.c +++ b/fs/resctrl/rdtgroup.c @@ -3852,6 +3852,7 @@ static void mongrp_reparent(struct rdtgroup *rdtgrp, cpumask_var_t cpus) { struct rdtgroup *prdtgrp = rdtgrp->mon.parent; + struct rdtgroup old_rdtgrp = *rdtgrp; WARN_ON(rdtgrp->type != RDTMON_GROUP); WARN_ON(new_prdtgrp->type != RDTCTRL_GROUP); @@ -3870,7 +3871,7 @@ static void mongrp_reparent(struct rdtgroup *rdtgrp, rdtgrp->closid = new_prdtgrp->closid; /* Propagate updated closid to all tasks in this group. */ - rdt_move_group_tasks(rdtgrp, rdtgrp, cpus); + rdt_move_group_tasks(&old_rdtgrp, rdtgrp, cpus); update_closid_rmid(cpus, NULL); } -- 2.43.0
hulk inclusion category: bugfix bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ The wd (weight divisor) field in MPAM MSC registers supports values 1-16 per the MPAM specification. However, the sanity check "wd > 15" incorrectly rejects the valid wd=16 case. Fix this by checking "wd > 16" instead. Fix both percent_to_mbw_max() and mbw_max_to_percent() to use the correct boundary check and GENMASK range. Return GENMASK(15, 0) for the error case in percent_to_mbw_max() to indicate "all bits valid" rather than the misleading MAX_MBA_BW. Fixes: 58db5c68e84a ("untested: arm_mpam: resctrl: Add support for MB resource") Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_resctrl.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index e3fd8361c22d9..a16f1318799c1 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -711,7 +711,7 @@ static u32 mbw_max_to_percent(u16 mbw_max, u8 wd) u8 bit; u32 divisor = 2, value = 0, precision = get_wd_precision(wd); - if (mbw_max == GENMASK(15, 15 - wd + 1)) + if (mbw_max == GENMASK(15, 16 - wd)) return MAX_MBA_BW; for (bit = 15; bit; bit--) { @@ -740,11 +740,12 @@ static u16 percent_to_mbw_max(u32 pc, u8 wd) u8 bit; u32 divisor = 2, value = 0, precision = get_wd_precision(wd); - if (WARN_ON_ONCE(wd > 15)) - return MAX_MBA_BW; + if (WARN_ON_ONCE(wd > 16)) + /* All bits valid as fallback */ + return GENMASK(15, 0); if (pc == MAX_MBA_BW) - return GENMASK(15, 15 - wd + 1); + return GENMASK(15, 16 - wd); pc *= precision; @@ -759,7 +760,7 @@ static u16 percent_to_mbw_max(u32 pc, u8 wd) break; } - value &= GENMASK(15, 15 - wd + 1); + value &= GENMASK(15, 16 - wd); return value; } -- 2.43.0
From: Yushan Wang <wangyushan12@huawei.com> hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Add new HiSilicon CPU type HIP13. Signed-off-by: Yushan Wang <wangyushan12@huawei.com> Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- arch/arm64/include/asm/cputype.h | 2 ++ tools/arch/arm64/include/asm/cputype.h | 2 ++ 2 files changed, 4 insertions(+) diff --git a/arch/arm64/include/asm/cputype.h b/arch/arm64/include/asm/cputype.h index 9cd4532daff3c..32e610cb281a0 100644 --- a/arch/arm64/include/asm/cputype.h +++ b/arch/arm64/include/asm/cputype.h @@ -147,6 +147,7 @@ #define HISI_CPU_PART_LINXICORE9100 0xD02 #define HISI_CPU_PART_HIP11 0xD22 #define HISI_CPU_PART_HIP12 0xD06 +#define HISI_CPU_PART_HIP13 0xD08 #define APPLE_CPU_PART_M1_ICESTORM 0x022 #define APPLE_CPU_PART_M1_FIRESTORM 0x023 @@ -239,6 +240,7 @@ #define MIDR_HISI_LINXICORE9100 MIDR_CPU_MODEL(ARM_CPU_IMP_HISI, HISI_CPU_PART_LINXICORE9100) #define MIDR_HISI_HIP11 MIDR_CPU_MODEL(ARM_CPU_IMP_HISI, HISI_CPU_PART_HIP11) #define MIDR_HISI_HIP12 MIDR_CPU_MODEL(ARM_CPU_IMP_HISI, HISI_CPU_PART_HIP12) +#define MIDR_HISI_HIP13 MIDR_CPU_MODEL(ARM_CPU_IMP_HISI, HISI_CPU_PART_HIP13) #define MIDR_APPLE_M1_ICESTORM MIDR_CPU_MODEL(ARM_CPU_IMP_APPLE, APPLE_CPU_PART_M1_ICESTORM) #define MIDR_APPLE_M1_FIRESTORM MIDR_CPU_MODEL(ARM_CPU_IMP_APPLE, APPLE_CPU_PART_M1_FIRESTORM) #define MIDR_APPLE_M1_ICESTORM_PRO MIDR_CPU_MODEL(ARM_CPU_IMP_APPLE, APPLE_CPU_PART_M1_ICESTORM_PRO) diff --git a/tools/arch/arm64/include/asm/cputype.h b/tools/arch/arm64/include/asm/cputype.h index c269f3bc87b3f..bacb39bd65a92 100644 --- a/tools/arch/arm64/include/asm/cputype.h +++ b/tools/arch/arm64/include/asm/cputype.h @@ -129,6 +129,7 @@ #define HISI_CPU_PART_TSV110 0xD01 #define HISI_CPU_PART_HIP11 0xD22 #define HISI_CPU_PART_HIP12 0xD06 +#define HISI_CPU_PART_HIP13 0xD08 #define APPLE_CPU_PART_M1_ICESTORM 0x022 #define APPLE_CPU_PART_M1_FIRESTORM 0x023 @@ -203,6 +204,7 @@ #define MIDR_HISI_TSV110 MIDR_CPU_MODEL(ARM_CPU_IMP_HISI, HISI_CPU_PART_TSV110) #define MIDR_HISI_HIP11 MIDR_CPU_MODEL(ARM_CPU_IMP_HISI, HISI_CPU_PART_HIP11) #define MIDR_HISI_HIP12 MIDR_CPU_MODEL(ARM_CPU_IMP_HISI, HISI_CPU_PART_HIP12) +#define MIDR_HISI_HIP13 MIDR_CPU_MODEL(ARM_CPU_IMP_HISI, HISI_CPU_PART_HIP13) #define MIDR_APPLE_M1_ICESTORM MIDR_CPU_MODEL(ARM_CPU_IMP_APPLE, APPLE_CPU_PART_M1_ICESTORM) #define MIDR_APPLE_M1_FIRESTORM MIDR_CPU_MODEL(ARM_CPU_IMP_APPLE, APPLE_CPU_PART_M1_FIRESTORM) #define MIDR_APPLE_M1_ICESTORM_PRO MIDR_CPU_MODEL(ARM_CPU_IMP_APPLE, APPLE_CPU_PART_M1_ICESTORM_PRO) -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Add support for the Hisilicon HIP13 processor by extending the existing HIP12-specific quirks to cover HIP13 as well. HIP13 inherits the following quirks from HIP12: * PMG_MAX workaround: Force PMG_MAX to 0 * CSU retrigger: Re-trigger cache monitoring reads for accurate values * CSU halved: Halve cache utilization values to match actual capacity Create a unified hisi_cpus[] list containing both HIP12 and HIP13 to simplify the quirk checks. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 3e875f29bd03b..60ab22fed7fde 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -97,6 +97,12 @@ static const struct midr_range hip12_cpus[] = { { /* sentinel */ } }; +static const struct midr_range hisi_cpus[] = { + MIDR_ALL_VERSIONS(MIDR_HISI_HIP12), + MIDR_ALL_VERSIONS(MIDR_HISI_HIP13), + { /* sentinel */ } +}; + static u32 __mpam_read_reg(struct mpam_msc *msc, u16 reg) { WARN_ON_ONCE(reg + sizeof(u32) > msc->mapped_hwpage_sz); @@ -776,7 +782,7 @@ static void mpam_ris_hw_probe(struct mpam_msc_ris *ris) static int mpam_pmg_max_workaround(u64 idr) { - if (is_midr_in_range_list(hip12_cpus)) + if (is_midr_in_range_list(hisi_cpus)) return 0; return FIELD_GET(MPAMF_IDR_PMG_MAX, idr); @@ -957,7 +963,7 @@ static bool mpam_csu_hisi_need_retrigger(struct mpam_msc_ris *ris, ris->comp->class->level != 3) return false; - if (!is_midr_in_range_list(hip12_cpus)) + if (!is_midr_in_range_list(hisi_cpus)) return false; if (read_again) @@ -1025,7 +1031,7 @@ static bool mpam_ris_has_nrdy_bit(struct mpam_msc_ris *ris) static u64 mpam_csu_hisi_need_halved(struct mpam_msc_ris *ris, u64 now) { - if (!is_midr_in_range_list(hip12_cpus)) + if (!is_midr_in_range_list(hisi_cpus)) return now; if (ris->comp->class->type != MPAM_CLASS_CACHE || -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Replace the global resctrl_is_mbm_enabled() with a resource-specific resctrl_arch_is_mbm_enabled(rid) that allows architectures to determine MBM support based on the resource level (L2/L3/MBA). This change is necessary because: * x86 supports MBM only on L3 (total and local) * ARM MPAM supports MBM on L2 (core), L3 (local), and MBA (total) * The previous global check can't aggregate all MBM types, leading to false positives when only specific resources support MBM Update all call sites to pass the appropriate rid. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_resctrl.c | 14 ++++++++++++++ fs/resctrl/rdtgroup.c | 15 ++++----------- include/linux/arm_mpam.h | 1 + 3 files changed, 19 insertions(+), 11 deletions(-) diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index a16f1318799c1..9c1d98791e225 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -85,6 +85,20 @@ bool resctrl_arch_is_mbm_core_enabled(void) return mbm_core_class; } +bool resctrl_arch_is_mbm_enabled(enum resctrl_res_level rid) +{ + switch (rid) { + case RDT_RESOURCE_L2: + return resctrl_arch_is_mbm_core_enabled(); + case RDT_RESOURCE_L3: + return resctrl_arch_is_mbm_local_enabled(); + case RDT_RESOURCE_MBA: + return resctrl_arch_is_mbm_total_enabled(); + default: + return false; + } +} + bool resctrl_arch_get_cdp_enabled(enum resctrl_res_level rid) { switch (rid) { diff --git a/fs/resctrl/rdtgroup.c b/fs/resctrl/rdtgroup.c index 16c4f9311111d..139ea6bc446cd 100644 --- a/fs/resctrl/rdtgroup.c +++ b/fs/resctrl/rdtgroup.c @@ -111,13 +111,6 @@ void rdt_staged_configs_clear(void) } } -static bool resctrl_is_mbm_enabled(void) -{ - return (resctrl_arch_is_mbm_total_enabled() || - resctrl_arch_is_mbm_local_enabled() || - resctrl_arch_is_mbm_core_enabled()); -} - static bool resctrl_is_mbm_event(int e) { return (e == QOS_L3_MBM_TOTAL_EVENT_ID || @@ -2765,7 +2758,7 @@ static int rdt_get_tree(struct fs_context *fc) if (resctrl_arch_alloc_capable() || resctrl_arch_mon_capable()) resctrl_mounted = true; - if (resctrl_is_mbm_enabled() && resctrl_arch_would_mbm_overflow()) { + if (resctrl_arch_is_mbm_enabled(l3->rid) && resctrl_arch_would_mbm_overflow()) { list_for_each_entry(dom, &l3->domains, list) mbm_setup_overflow_handler(dom, MBM_OVERFLOW_INTERVAL, RESCTRL_PICK_ANY_CPU); @@ -4072,7 +4065,7 @@ void resctrl_offline_domain(struct rdt_resource *r, struct rdt_domain *d) if (resctrl_mounted && resctrl_arch_mon_capable()) rmdir_mondata_subdir_allrdtgrp(r, d->id); - if (resctrl_is_mbm_enabled() && resctrl_arch_would_mbm_overflow()) + if (resctrl_arch_is_mbm_enabled(r->rid) && resctrl_arch_would_mbm_overflow()) cancel_delayed_work(&d->mbm_over); if (resctrl_arch_is_llc_occupancy_enabled() && has_busy_rmid(d)) { /* @@ -4153,7 +4146,7 @@ int resctrl_online_domain(struct rdt_resource *r, struct rdt_domain *d) if (err) goto out_unlock; - if (resctrl_is_mbm_enabled() && resctrl_arch_would_mbm_overflow()) { + if (resctrl_arch_is_mbm_enabled(r->rid) && resctrl_arch_would_mbm_overflow()) { INIT_DELAYED_WORK(&d->mbm_over, mbm_handle_overflow); mbm_setup_overflow_handler(d, MBM_OVERFLOW_INTERVAL, RESCTRL_PICK_ANY_CPU); @@ -4214,7 +4207,7 @@ void resctrl_offline_cpu(unsigned int cpu) d = resctrl_get_domain_from_cpu(cpu, l3); if (d) { - if (resctrl_is_mbm_enabled() && cpu == d->mbm_work_cpu && + if (resctrl_arch_is_mbm_enabled(l3->rid) && cpu == d->mbm_work_cpu && resctrl_arch_would_mbm_overflow()) { cancel_delayed_work(&d->mbm_over); mbm_setup_overflow_handler(d, 0, cpu); diff --git a/include/linux/arm_mpam.h b/include/linux/arm_mpam.h index a1a511d07eec1..25c62a42bd2c2 100644 --- a/include/linux/arm_mpam.h +++ b/include/linux/arm_mpam.h @@ -68,6 +68,7 @@ bool resctrl_arch_is_l2c_occupancy_enabled(void); bool resctrl_arch_is_mbm_core_enabled(void); bool resctrl_arch_is_mbm_local_enabled(void); bool resctrl_arch_is_mbm_total_enabled(void); +bool resctrl_arch_is_mbm_enabled(enum resctrl_res_level rid); bool resctrl_arch_would_mbm_overflow(void); /* reset cached configurations, then all devices */ -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Introduce resctrl_arch_setup_res_mbm_over() and resctrl_arch_setup_res_mbm_over_exclude_cpu() to abstract MBM overflow handling setup for different architectures. This abstraction is required because: * x86 only supports MBM on L3 cache, requiring single-resource setup. * ARM MPAM supports MBM on multiple resources (L2, L3, MBA), requiring iteration over all enabled resources. Add generic helper: * resctrl_setup_dom_overflow(): Set up overflow handlers for all domains in a resource * resctrl_setup_dom_overflow_exclude_cpu(): Re-setup handlers excluding a specific CPU (used during CPU hotplug) Refactor rdt_get_tree() and resctrl_offline_cpu() to use these helpers, eliminating hardcoded L3 dependencies. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_resctrl.c | 34 ++++++++++++++++++++++ fs/resctrl/rdtgroup.c | 42 ++++++++++++++++++++-------- include/linux/resctrl.h | 7 +++++ 3 files changed, 71 insertions(+), 12 deletions(-) diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index 9c1d98791e225..ad9e6b1ac7167 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -99,6 +99,40 @@ bool resctrl_arch_is_mbm_enabled(enum resctrl_res_level rid) } } +void resctrl_arch_setup_res_mbm_over(void) +{ + struct mpam_resctrl_res *res; + int i; + + for (i = 0; i < RDT_NUM_RESOURCES; i++) { + res = &mpam_resctrl_exports[i]; + + if (!res->class) + continue; + + resctrl_setup_dom_overflow(&res->resctrl_res); + } +} + +void resctrl_arch_setup_res_mbm_over_exclude_cpu(unsigned int exclude_cpu) +{ + struct mpam_resctrl_res *res; + struct rdt_domain *d; + int i; + + for (i = 0; i < RDT_NUM_RESOURCES; i++) { + res = &mpam_resctrl_exports[i]; + + if (!res->class) + continue; + + d = resctrl_get_domain_from_cpu(exclude_cpu, &res->resctrl_res); + if (d) + resctrl_setup_dom_overflow_exclude_cpu(&res->resctrl_res, + d, exclude_cpu); + } +} + bool resctrl_arch_get_cdp_enabled(enum resctrl_res_level rid) { switch (rid) { diff --git a/fs/resctrl/rdtgroup.c b/fs/resctrl/rdtgroup.c index 139ea6bc446cd..b719eded4839b 100644 --- a/fs/resctrl/rdtgroup.c +++ b/fs/resctrl/rdtgroup.c @@ -2679,12 +2679,22 @@ static void schemata_list_destroy(void) } } +void resctrl_setup_dom_overflow(struct rdt_resource *r) +{ + struct rdt_domain *d; + + if (resctrl_arch_is_mbm_enabled(r->rid) && + resctrl_arch_would_mbm_overflow()) { + list_for_each_entry(d, &r->domains, list) + mbm_setup_overflow_handler(d, MBM_OVERFLOW_INTERVAL, + RESCTRL_PICK_ANY_CPU); + } +} + static int rdt_get_tree(struct fs_context *fc) { - struct rdt_resource *l3 = resctrl_arch_get_resource(RDT_RESOURCE_L3); struct rdt_fs_context *ctx = rdt_fc2context(fc); unsigned long flags = RFTYPE_CTRL_BASE; - struct rdt_domain *dom; int ret; cpus_read_lock(); @@ -2758,11 +2768,7 @@ static int rdt_get_tree(struct fs_context *fc) if (resctrl_arch_alloc_capable() || resctrl_arch_mon_capable()) resctrl_mounted = true; - if (resctrl_arch_is_mbm_enabled(l3->rid) && resctrl_arch_would_mbm_overflow()) { - list_for_each_entry(dom, &l3->domains, list) - mbm_setup_overflow_handler(dom, MBM_OVERFLOW_INTERVAL, - RESCTRL_PICK_ANY_CPU); - } + resctrl_arch_setup_res_mbm_over(); goto out; @@ -4188,6 +4194,21 @@ static void clear_childcpus(struct rdtgroup *r, unsigned int cpu) } } +void resctrl_setup_dom_overflow_exclude_cpu(struct rdt_resource *r, + struct rdt_domain *d, + unsigned int exclude_cpu) +{ + if (!d) + return; + + if (resctrl_arch_is_mbm_enabled(r->rid) && + exclude_cpu == d->mbm_work_cpu && + resctrl_arch_would_mbm_overflow()) { + cancel_delayed_work(&d->mbm_over); + mbm_setup_overflow_handler(d, 0, exclude_cpu); + } +} + void resctrl_offline_cpu(unsigned int cpu) { struct rdt_resource *l3 = resctrl_arch_get_resource(RDT_RESOURCE_L3); @@ -4207,11 +4228,6 @@ void resctrl_offline_cpu(unsigned int cpu) d = resctrl_get_domain_from_cpu(cpu, l3); if (d) { - if (resctrl_arch_is_mbm_enabled(l3->rid) && cpu == d->mbm_work_cpu && - resctrl_arch_would_mbm_overflow()) { - cancel_delayed_work(&d->mbm_over); - mbm_setup_overflow_handler(d, 0, cpu); - } if (resctrl_arch_is_llc_occupancy_enabled() && cpu == d->cqm_work_cpu && has_busy_rmid(d)) { cancel_delayed_work(&d->cqm_limbo); @@ -4219,6 +4235,8 @@ void resctrl_offline_cpu(unsigned int cpu) } } + resctrl_arch_setup_res_mbm_over_exclude_cpu(cpu); + out_unlock: mutex_unlock(&rdtgroup_mutex); } diff --git a/include/linux/resctrl.h b/include/linux/resctrl.h index dd3c2b660b393..6fb5e7ca29a26 100644 --- a/include/linux/resctrl.h +++ b/include/linux/resctrl.h @@ -473,6 +473,13 @@ void resctrl_exit(void); int resctrl_arch_mon_resource_init(void); void mbm_config_rftype_init(const char *config); +void resctrl_setup_dom_overflow(struct rdt_resource *r); +void resctrl_setup_dom_overflow_exclude_cpu(struct rdt_resource *r, + struct rdt_domain *d, + unsigned int exclude_cpu); +void resctrl_arch_setup_res_mbm_over(void); +void resctrl_arch_setup_res_mbm_over_exclude_cpu(unsigned int exclude_cpu); + /* When supported, the architecture must implement these */ #ifdef CONFIG_RESCTRL_IOMMU int resctrl_arch_set_iommu_closid_rmid(struct iommu_group *group, u32 closid, -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ L2 monitors are typically fewer than available RMIDs. To avoid false overflow detections, only perform MBM overflow checking for RMIDs that are currently being monitored by a physical monitor. Introduce mpam_skip_check_l2_overflow() to verify if the current monitoring configuration (partid-pmg pair) matches the saved state in mbwu_state. Skip overflow handling when no match is found, indicating the RMID is not actively monitored on this component. Add a new event type QOS_L2_MBM_CORE_OVERFLOW_EVENT_ID to distinguish overflow checking from regular MBM reads, allowing the skip logic to be applied only where needed. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 3 ++ drivers/platform/mpam/mpam_resctrl.c | 42 ++++++++++++++++++++++++++++ fs/resctrl/monitor.c | 2 +- include/linux/resctrl_types.h | 1 + 4 files changed, 47 insertions(+), 1 deletion(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 60ab22fed7fde..d6e29f4fd710d 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -1086,6 +1086,9 @@ static void __ris_msmon_read(void *arg) if (mbwu_state) { reset_on_next_read = mbwu_state->reset_on_next_read; mbwu_state->reset_on_next_read = false; + + mbwu_state->cfg.partid = ctx->partid; + mbwu_state->cfg.pmg = ctx->pmg; } mbwu_overflow = read_msmon_mbwu_is_overflow(msc); diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index ad9e6b1ac7167..93d7d14753237 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -493,6 +493,42 @@ static enum mon_filter_options resctrl_evt_config_to_mpam(u32 local_evt_cfg) } } +/* + * Check whether to skip L2 MBM overflow checking for a given component. + * + * The number of L2 monitors is less than the number of RMIDs, so we only + * check MBM overflow for RMIDs currently being monitored by the monitor. + * When handling QOS_L2_MBM_CORE_OVERFLOW_EVENT_ID, we verify if the current + * monitoring configuration (partid/pmg) matches the previously saved one in + * mbwu_state. If they match, it means this RMID is still being monitored + * and we should proceed with overflow check. Otherwise, skip it. + * + * Returns: + * false - Don't skip, proceed with overflow check (partid/pmg match) + * true - Skip overflow check (no matching configuration found) + */ +static bool mpam_skip_check_l2_overflow(struct mpam_component *comp, + struct mon_cfg *cfg) +{ + bool ret; + unsigned long flags; + struct mpam_msc_ris *ris; + struct msmon_mbwu_state *mbwu_state; + + ris = list_first_or_null_rcu(&comp->ris, struct mpam_msc_ris, comp_list); + if (!ris) + return true; + + mbwu_state = &ris->mbwu_state[cfg->mon]; + + spin_lock_irqsave(&ris->msc->mon_sel_lock, flags); + ret = (mbwu_state->cfg.partid != cfg->partid || + mbwu_state->cfg.pmg != cfg->pmg); + spin_unlock_irqrestore(&ris->msc->mon_sel_lock, flags); + + return ret; +} + int resctrl_arch_rmid_read(struct rdt_resource *r, struct rdt_domain *d, u32 closid, u32 rmid, enum resctrl_event_id eventid, u64 *val, void *arch_mon_ctx) @@ -539,6 +575,12 @@ int resctrl_arch_rmid_read(struct rdt_resource *r, struct rdt_domain *d, cfg.partid = rmid2reqpartid(rmid); cfg.mon = cfg.partid % num_mon; + + if (eventid == QOS_L2_MBM_CORE_OVERFLOW_EVENT_ID) { + if (mpam_skip_check_l2_overflow(dom->comp, &cfg)) + return 0; + } + err = mpam_msmon_read(dom->comp, &cfg, type, val); if (err) return err; diff --git a/fs/resctrl/monitor.c b/fs/resctrl/monitor.c index e5eb971ee39da..516a94d833d4d 100644 --- a/fs/resctrl/monitor.c +++ b/fs/resctrl/monitor.c @@ -633,7 +633,7 @@ static void mbm_update(struct rdt_resource *r, struct rdt_domain *d, resctrl_arch_mon_ctx_free(rr.r, rr.evtid, rr.arch_mon_ctx); } if (resctrl_arch_is_mbm_core_enabled()) { - rr.evtid = QOS_L2_MBM_CORE_EVENT_ID; + rr.evtid = QOS_L2_MBM_CORE_OVERFLOW_EVENT_ID; rr.val = 0; rr.arch_mon_ctx = resctrl_arch_mon_ctx_alloc(rr.r, rr.evtid); if (IS_ERR(rr.arch_mon_ctx)) { diff --git a/include/linux/resctrl_types.h b/include/linux/resctrl_types.h index eaea801a85523..debba6674578c 100644 --- a/include/linux/resctrl_types.h +++ b/include/linux/resctrl_types.h @@ -120,6 +120,7 @@ enum resctrl_event_id { QOS_L2_OCCUP_EVENT_ID, QOS_L2_MBM_CORE_EVENT_ID, + QOS_L2_MBM_CORE_OVERFLOW_EVENT_ID, }; #endif /* CONFIG_X86_CPU_RESCTRL */ -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ The current resctrl_arch_would_mbm_overflow() assumes all HiSilicon CPUs do not require MBM overflow handling, returning false for any ARM_CPU_IMP_HISI implementor. However, HIP13 follows the ARM spec and requires proper overflow detection. This leads to incorrect memory bandwidth measurements on HIP13 systems when the MBM counter wraps around. Fix this by explicitly checking for HIP13 in the HiSilicon branch and returning true to enable overflow handling. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index d6e29f4fd710d..7180b253cf978 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -97,6 +97,11 @@ static const struct midr_range hip12_cpus[] = { { /* sentinel */ } }; +static const struct midr_range hip13_cpus[] = { + MIDR_ALL_VERSIONS(MIDR_HISI_HIP13), + { /* sentinel */ } +}; + static const struct midr_range hisi_cpus[] = { MIDR_ALL_VERSIONS(MIDR_HISI_HIP12), MIDR_ALL_VERSIONS(MIDR_HISI_HIP13), @@ -1018,7 +1023,12 @@ static u64 mpam_msmon_overflow_val(struct mpam_msc_ris *ris) bool resctrl_arch_would_mbm_overflow(void) { - return read_cpuid_implementor() != ARM_CPU_IMP_HISI; + /* Non-HiSilicon CPUs and HIP13 need overflow handling */ + if (read_cpuid_implementor() != ARM_CPU_IMP_HISI) + return true; + + /* HIP13 is the exception among HiSilicon CPUs */ + return is_midr_in_range_list(hip13_cpus); } static bool mpam_ris_has_nrdy_bit(struct mpam_msc_ris *ris) -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ HIP13 supports disabling L2 cache entirely by setting CPBM (Cache Portion Bitmask) to zero. This capability is useful for L2 cache isolation scenarios where a task or group should not use any L2 cache. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 10 ++++++++++ drivers/platform/mpam/mpam_internal.h | 1 + drivers/platform/mpam/mpam_resctrl.c | 2 +- 3 files changed, 12 insertions(+), 1 deletion(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 7180b253cf978..3c93dc6b2b1ca 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -1051,6 +1051,16 @@ static u64 mpam_csu_hisi_need_halved(struct mpam_msc_ris *ris, u64 now) return now >> 1; } +u32 mpam_min_cbm_bits(enum resctrl_res_level rid) +{ + if (rid == RDT_RESOURCE_L2) { + if (is_midr_in_range_list(hip13_cpus)) + return 0; + } + + return 1; +} + static bool read_msmon_mbwu_is_overflow(struct mpam_msc *msc) { u32 ctl; diff --git a/drivers/platform/mpam/mpam_internal.h b/drivers/platform/mpam/mpam_internal.h index 70d4abb93fd19..ce5af77e3d300 100644 --- a/drivers/platform/mpam/mpam_internal.h +++ b/drivers/platform/mpam/mpam_internal.h @@ -331,6 +331,7 @@ void mpam_resctrl_exit(void); u16 mpam_cpbm_wd_hisi_workaround(u16 cpbm_wd, enum mpam_device_features feat, u8 cache_level); bool mpam_cpbm_hisi_check_invalid(struct rdt_resource *r, unsigned long val); +u32 mpam_min_cbm_bits(enum resctrl_res_level rid); /* * MPAM MSCs have the following register layout. See: diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index 93d7d14753237..ce5a894b962ba 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -1130,7 +1130,7 @@ static int mpam_resctrl_resource_init(struct mpam_resctrl_res *res) r->cache.arch_has_sparse_bitmasks = true; /* mpam_devices will reject empty bitmaps */ - r->cache.min_cbm_bits = 1; + r->cache.min_cbm_bits = mpam_min_cbm_bits(res->resctrl_res.rid); /* TODO: kill these properties off as they are derivatives */ r->format_str = "%d=%0*x"; -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Update the SMMU IOMMU group QoS parameter functions to use the Narrow PARTID abstraction helpers instead of manual partid calculations: This enables SMMU devices to participate in the Narrow PARTID scheme where multiple reqPARTIDs can be mapped to a single internal PARTID, expanding the number of available monitoring contexts for IOMMU groups. This patch builds on the Narrow PARTID feature introduced in earlier commits, extending support to SMMU/IOMMU devices. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_resctrl.c | 27 +++++++++------------------ 1 file changed, 9 insertions(+), 18 deletions(-) diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index ce5a894b962ba..569a5254d65e0 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -414,44 +414,35 @@ bool resctrl_arch_match_rmid(struct task_struct *tsk, u32 closid, u32 rmid) int resctrl_arch_set_iommu_closid_rmid(struct iommu_group *group, u32 closid, u32 rmid) { - u16 partid; - - if (cdp_enabled) - partid = closid << 1; - else - partid = closid; - - return iommu_group_set_qos_params(group, partid, rmid); + return iommu_group_set_qos_params(group, rmid2reqpartid(rmid), + rmid2pmg(rmid)); } bool resctrl_arch_match_iommu_closid(struct iommu_group *group, u32 closid) { - u16 partid; - int err = iommu_group_get_qos_params(group, &partid, NULL); + u16 reqpartid; + int err = iommu_group_get_qos_params(group, &reqpartid, NULL); if (err) return false; if (cdp_enabled) - partid >>= 1; + closid <<= 1; - return (partid == closid); + return req2intpartid(reqpartid) == closid; } bool resctrl_arch_match_iommu_closid_rmid(struct iommu_group *group, u32 closid, u32 rmid) { u8 pmg; - u16 partid; - int err = iommu_group_get_qos_params(group, &partid, &pmg); + u16 reqpartid; + int err = iommu_group_get_qos_params(group, &reqpartid, &pmg); if (err) return false; - if (cdp_enabled) - partid >>= 1; - - return (partid == closid) && (rmid == pmg); + return req_pmg2rmid(reqpartid, pmg) == rmid; } #endif -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Add support for the MBOPT memory bandwidth resource, which provides an alternative bandwidth allocation mechanism compared to the standard MBA (Memory Bandwidth Allocation) resource. MBOPT uses a different scaling and value interpretation for bandwidth limits, allowing finer-grained control. Unlike MBA which converts values to percentages, MBOPT uses raw register values shifted by the hardware standard width. The MBOPT resource is exposed as "MBOPT" in the resctrl schemata, alongside the existing "MB" resource. MBOPT defaults to maximum bandwidth, allowing all traffic by default until explicitly restricted by the user. Example: # cat schemata MBOPT:0=0001023;1=0001023 MB:0=0000100;1=0000100 # echo MBOPT:0=000102 > schemata # cat schemata MBOPT:0=0000102;1=0001023 MB:0=0000010;1=0000100 Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_resctrl.c | 39 ++++++++++++++++++++++++++-- include/linux/resctrl_types.h | 1 + 2 files changed, 38 insertions(+), 2 deletions(-) diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index 569a5254d65e0..a8f8f1a802eae 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -1002,6 +1002,10 @@ static void mpam_resctrl_pick_mba(void) res = &mpam_resctrl_exports[RDT_RESOURCE_MBA]; res->class = class; res->resctrl_res.name = "MB"; + + res = &mpam_resctrl_exports[RDT_RESOURCE_MB_OPT]; + res->class = class; + res->resctrl_res.name = "MBOPT"; } if (has_mbw_min) { @@ -1307,6 +1311,23 @@ static int mpam_resctrl_resource_init(struct mpam_resctrl_res *res) r->alloc_capable = true; break; + case RDT_RESOURCE_MB_OPT: + r->format_str = "%d=%0*u"; + r->schema_fmt = RESCTRL_SCHEMA_RANGE; + r->fflags = RFTYPE_RES_MB; + r->default_ctrl = GENMASK(cprops->bwa_wd - 1, 0); + r->membw.max_bw = GENMASK(cprops->bwa_wd - 1, 0); + r->data_width = 5; + + r->membw.delay_linear = true; + r->membw.throttle_mode = THREAD_THROTTLE_UNDEFINED; + r->membw.min_bw = 1; + r->membw.bw_gran = 1; + + if (class_has_usable_mba(cprops)) + r->alloc_capable = true; + break; + default: break; } @@ -1571,6 +1592,10 @@ u32 resctrl_arch_get_config(struct rdt_resource *r, struct rdt_domain *d, configured_by = mpam_feat_max_limit; break; + case RDT_RESOURCE_MB_OPT: + configured_by = mpam_feat_mbw_max; + break; + default: return -EINVAL; } @@ -1593,14 +1618,20 @@ u32 resctrl_arch_get_config(struct rdt_resource *r, struct rdt_domain *d, /* TODO: Scaling is not yet supported */ return mbw_pbm_to_percent(cfg->mbw_pbm, cprops); case mpam_feat_mbw_max: - return mbw_max_to_percent(cfg->mbw_max, cprops->bwa_wd); + if (r->rid == RDT_RESOURCE_MBA) + return mbw_max_to_percent(cfg->mbw_max, cprops->bwa_wd); + else if (r->rid == RDT_RESOURCE_MB_OPT) + return cfg->mbw_max >> (16 - cprops->bwa_wd); + break; case mpam_feat_mbw_min: return mbw_max_to_percent(cfg->mbw_min, cprops->bwa_wd); case mpam_feat_max_limit: return cfg->max_limit; default: - return -EINVAL; + break; } + + return -EINVAL; } int resctrl_arch_update_one(struct rdt_resource *r, struct rdt_domain *d, @@ -1676,6 +1707,10 @@ int resctrl_arch_update_one(struct rdt_resource *r, struct rdt_domain *d, cfg.max_limit = cfg_val; mpam_set_feature(mpam_feat_max_limit, &cfg); break; + case RDT_RESOURCE_MB_OPT: + cfg.mbw_max = cfg_val << (16 - cprops->bwa_wd); + mpam_set_feature(mpam_feat_mbw_max, &cfg); + break; default: return -EINVAL; } diff --git a/include/linux/resctrl_types.h b/include/linux/resctrl_types.h index debba6674578c..8cdbaf2eb6074 100644 --- a/include/linux/resctrl_types.h +++ b/include/linux/resctrl_types.h @@ -94,6 +94,7 @@ enum resctrl_res_level { RDT_RESOURCE_L2_PRI, RDT_RESOURCE_MB_PRI, RDT_RESOURCE_MB_HDL, + RDT_RESOURCE_MB_OPT, #endif /* Must be the last */ -- 2.43.0
hulk inclusion category: bugfix bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Remove the mbm_core field and its associated allocation from struct rdt_domain, as it is no referenced anywhere in the codebase. This is a code cleanup with no functional change. Fixes: 556688623b2b ("fs/resctrl: Create l2 cache monitors") Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- fs/resctrl/rdtgroup.c | 10 ---------- include/linux/resctrl.h | 1 - 2 files changed, 11 deletions(-) diff --git a/fs/resctrl/rdtgroup.c b/fs/resctrl/rdtgroup.c index b719eded4839b..601c74e18315b 100644 --- a/fs/resctrl/rdtgroup.c +++ b/fs/resctrl/rdtgroup.c @@ -4119,16 +4119,6 @@ static int domain_setup_mon_state(struct rdt_resource *r, struct rdt_domain *d) return -ENOMEM; } } - if (resctrl_arch_is_mbm_core_enabled()) { - tsize = sizeof(*d->mbm_core); - d->mbm_core = kcalloc(idx_limit, tsize, GFP_KERNEL); - if (!d->mbm_core) { - bitmap_free(d->rmid_busy_llc); - kfree(d->mbm_total); - kfree(d->mbm_local); - return -ENOMEM; - } - } return 0; } diff --git a/include/linux/resctrl.h b/include/linux/resctrl.h index 6fb5e7ca29a26..a9ef4cfce2a62 100644 --- a/include/linux/resctrl.h +++ b/include/linux/resctrl.h @@ -119,7 +119,6 @@ struct rdt_domain { unsigned long *rmid_busy_llc; struct mbm_state *mbm_total; struct mbm_state *mbm_local; - struct mbm_state *mbm_core; struct delayed_work mbm_over; struct delayed_work cqm_limbo; int mbm_work_cpu; -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Refactor the MBM (Memory Bandwidth Monitoring) update path to allow arch-specific selection of which events to sample. Previously, mbm_update() hardcoded checks for QOS_L3_MBM_TOTAL_EVENT_ID, QOS_L3_MBM_LOCAL_EVENT_ID, and QOS_L2_MBM_CORE_OVERFLOW_EVENT_ID, coupling the generic resctrl code to x86-specific event IDs. This patch introduces resctrl_arch_mbm_update() as an arch-specific hook that decides which MBM events to update: - x86/rdt: Samples L3 MBM total and local bandwidth events, same as before. - ARM/MPAM: Samples MBA total/local or L2 core overflow events, depending on the resource type. Extracts the per-event sampling logic into a new generic helper resctrl_mbm_update_one(), which handles the allocation of monitor context, event counting, and context freeing for a single event type. Adds a 'res' pointer to struct rdt_domain so that the MPAM implementation can determine the resource type from the domain. This enables MPAM platforms to handle MBM overflow detection correctly without polluting the generic code with platform-specific event logic. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_resctrl.c | 31 ++++++++++++ fs/resctrl/monitor.c | 73 ++++++++-------------------- include/linux/resctrl.h | 6 +++ 3 files changed, 58 insertions(+), 52 deletions(-) diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index a8f8f1a802eae..c279ba142ca3d 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -133,6 +133,36 @@ void resctrl_arch_setup_res_mbm_over_exclude_cpu(unsigned int exclude_cpu) } } +void resctrl_arch_mbm_update(struct rdt_domain *d, + u32 closid, u32 rmid) +{ + switch (d->res->rid) { + case RDT_RESOURCE_MBA: + if (resctrl_arch_is_mbm_total_enabled()) + resctrl_mbm_update_one(d->res, d, + QOS_L3_MBM_TOTAL_EVENT_ID, + closid, rmid); + break; + + case RDT_RESOURCE_L3: + if (resctrl_arch_is_mbm_local_enabled()) + resctrl_mbm_update_one(d->res, d, + QOS_L3_MBM_LOCAL_EVENT_ID, + closid, rmid); + break; + + case RDT_RESOURCE_L2: + if (resctrl_arch_is_mbm_core_enabled() && !d->res->invisible) + resctrl_mbm_update_one(d->res, d, + QOS_L2_MBM_CORE_OVERFLOW_EVENT_ID, + closid, rmid); + break; + + default: + break; + } +} + bool resctrl_arch_get_cdp_enabled(enum resctrl_res_level rid) { switch (rid) { @@ -1818,6 +1848,7 @@ mpam_resctrl_alloc_domain(unsigned int cpu, struct mpam_resctrl_res *res) /* TODO: this list should be sorted */ list_add_tail(&dom->resctrl_dom.list, &res->resctrl_res.domains); + dom->resctrl_dom.res = &res->resctrl_res; return dom; } diff --git a/fs/resctrl/monitor.c b/fs/resctrl/monitor.c index 516a94d833d4d..b3eb1a02ac607 100644 --- a/fs/resctrl/monitor.c +++ b/fs/resctrl/monitor.c @@ -583,8 +583,9 @@ static void update_mba_bw(struct rdtgroup *rgrp, struct rdt_domain *dom_mbm) } } -static void mbm_update(struct rdt_resource *r, struct rdt_domain *d, - u32 closid, u32 rmid) +void resctrl_mbm_update_one(struct rdt_resource *r, struct rdt_domain *d, + enum resctrl_event_id evtid, + u32 closid, u32 rmid) { struct rmid_read rr; @@ -596,56 +597,26 @@ static void mbm_update(struct rdt_resource *r, struct rdt_domain *d, * This is protected from concurrent reads from user * as both the user and we hold the global mutex. */ - if (resctrl_arch_is_mbm_total_enabled()) { - rr.evtid = QOS_L3_MBM_TOTAL_EVENT_ID; - rr.val = 0; - rr.arch_mon_ctx = resctrl_arch_mon_ctx_alloc(rr.r, rr.evtid); - if (IS_ERR(rr.arch_mon_ctx)) { - pr_warn_ratelimited("Failed to allocate monitor context: %ld", - PTR_ERR(rr.arch_mon_ctx)); - return; - } - - __mon_event_count(closid, rmid, &rr); - - resctrl_arch_mon_ctx_free(rr.r, rr.evtid, rr.arch_mon_ctx); + rr.evtid = evtid; + rr.val = 0; + rr.arch_mon_ctx = resctrl_arch_mon_ctx_alloc(rr.r, rr.evtid); + if (IS_ERR(rr.arch_mon_ctx)) { + pr_warn_ratelimited("Failed to allocate monitor context: %ld", + PTR_ERR(rr.arch_mon_ctx)); + return; } - if (resctrl_arch_is_mbm_local_enabled()) { - rr.evtid = QOS_L3_MBM_LOCAL_EVENT_ID; - rr.val = 0; - rr.arch_mon_ctx = resctrl_arch_mon_ctx_alloc(rr.r, rr.evtid); - if (IS_ERR(rr.arch_mon_ctx)) { - pr_warn_ratelimited("Failed to allocate monitor context: %ld", - PTR_ERR(rr.arch_mon_ctx)); - return; - } - __mon_event_count(closid, rmid, &rr); + __mon_event_count(closid, rmid, &rr); - /* - * Call the MBA software controller only for the - * control groups and when user has enabled - * the software controller explicitly. - */ - if (is_mba_sc(NULL)) - mbm_bw_count(closid, rmid, &rr); - - resctrl_arch_mon_ctx_free(rr.r, rr.evtid, rr.arch_mon_ctx); - } - if (resctrl_arch_is_mbm_core_enabled()) { - rr.evtid = QOS_L2_MBM_CORE_OVERFLOW_EVENT_ID; - rr.val = 0; - rr.arch_mon_ctx = resctrl_arch_mon_ctx_alloc(rr.r, rr.evtid); - if (IS_ERR(rr.arch_mon_ctx)) { - pr_warn_ratelimited("Failed to allocate monitor context: %ld", - PTR_ERR(rr.arch_mon_ctx)); - return; - } - - __mon_event_count(closid, rmid, &rr); + /* + * Call the MBA software controller only for the + * control groups and when user has enabled + * the software controller explicitly. + */ + if ((evtid == QOS_L3_MBM_LOCAL_EVENT_ID) && is_mba_sc(NULL)) + mbm_bw_count(closid, rmid, &rr); - resctrl_arch_mon_ctx_free(rr.r, rr.evtid, rr.arch_mon_ctx); - } + resctrl_arch_mon_ctx_free(rr.r, rr.evtid, rr.arch_mon_ctx); } /* @@ -701,7 +672,6 @@ void mbm_handle_overflow(struct work_struct *work) unsigned long delay = msecs_to_jiffies(MBM_OVERFLOW_INTERVAL); struct rdtgroup *prgrp, *crgrp; struct list_head *head; - struct rdt_resource *r; struct rdt_domain *d; cpus_read_lock(); @@ -714,15 +684,14 @@ void mbm_handle_overflow(struct work_struct *work) if (!resctrl_mounted || !resctrl_arch_mon_capable()) goto out_unlock; - r = resctrl_arch_get_resource(RDT_RESOURCE_L3); d = container_of(work, struct rdt_domain, mbm_over.work); list_for_each_entry(prgrp, &rdt_all_groups, rdtgroup_list) { - mbm_update(r, d, prgrp->closid, prgrp->mon.rmid); + resctrl_arch_mbm_update(d, prgrp->closid, prgrp->mon.rmid); head = &prgrp->mon.crdtgrp_list; list_for_each_entry(crgrp, head, mon.crdtgrp_list) - mbm_update(r, d, crgrp->closid, crgrp->mon.rmid); + resctrl_arch_mbm_update(d, crgrp->closid, crgrp->mon.rmid); if (is_mba_sc(NULL)) update_mba_bw(prgrp, d); diff --git a/include/linux/resctrl.h b/include/linux/resctrl.h index a9ef4cfce2a62..1b8f89a5830e0 100644 --- a/include/linux/resctrl.h +++ b/include/linux/resctrl.h @@ -126,6 +126,7 @@ struct rdt_domain { struct pseudo_lock_region *plr; struct resctrl_staged_config staged_config[CDP_NUM_TYPES]; u32 *mbps_val; + struct rdt_resource *res; /* Just for MPAM */ }; /** @@ -479,6 +480,11 @@ void resctrl_setup_dom_overflow_exclude_cpu(struct rdt_resource *r, void resctrl_arch_setup_res_mbm_over(void); void resctrl_arch_setup_res_mbm_over_exclude_cpu(unsigned int exclude_cpu); +void resctrl_mbm_update_one(struct rdt_resource *r, struct rdt_domain *d, + enum resctrl_event_id evtid, + u32 closid, u32 rmid); +void resctrl_arch_mbm_update(struct rdt_domain *d, u32 closid, u32 rmid); + /* When supported, the architecture must implement these */ #ifdef CONFIG_RESCTRL_IOMMU int resctrl_arch_set_iommu_closid_rmid(struct iommu_group *group, u32 closid, -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Add support for checking whether the MBM (Memory Bandwidth Monitoring) total bandwidth event is configurable on MPAM-enabled platforms. This mirrors the existing handling for the MBM local bandwidth event by checking the msmon_mbwu_rwbw feature flag on the mbm_total_class's properties. The QOS_L3_MBM_TOTAL_EVENT_ID case was missing from resctrl_arch_is_evt_configurable(), causing the total bandwidth monitoring event to be incorrectly reported as non-configurable even when the underlying hardware supports it. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_resctrl.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/drivers/platform/mpam/mpam_resctrl.c b/drivers/platform/mpam/mpam_resctrl.c index c279ba142ca3d..5f470d07d1f04 100644 --- a/drivers/platform/mpam/mpam_resctrl.c +++ b/drivers/platform/mpam/mpam_resctrl.c @@ -1086,6 +1086,12 @@ bool resctrl_arch_is_evt_configurable(enum resctrl_event_id evt) return false; cprops = &mbm_local_class->props; + return mpam_has_feature(mpam_feat_msmon_mbwu_rwbw, cprops); + case QOS_L3_MBM_TOTAL_EVENT_ID: + if (!mbm_total_class) + return false; + + cprops = &mbm_total_class->props; return mpam_has_feature(mpam_feat_msmon_mbwu_rwbw, cprops); default: return false; -- 2.43.0
From: James Morse <james.morse@arm.com> maillist inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 Reference: https://git.kernel.org/pub/scm/linux/kernel/git/morse/linux.git/commit/?h=mp... ------------------ debugfs has handy helpers to make a bool, integer or string available through debugfs. Add helpers to do the same for cpumasks. These are read only. CC: Ben Horgan <ben.horgan@arm.com> Signed-off-by: James Morse <james.morse@arm.com> Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- fs/debugfs/file.c | 64 +++++++++++++++++++++++++++++++++++++++++ include/linux/debugfs.h | 6 ++++ 2 files changed, 70 insertions(+) diff --git a/fs/debugfs/file.c b/fs/debugfs/file.c index 33676b8cf56c9..9bf45fe95aa37 100644 --- a/fs/debugfs/file.c +++ b/fs/debugfs/file.c @@ -1003,6 +1003,70 @@ void debugfs_create_str(const char *name, umode_t mode, } EXPORT_SYMBOL_GPL(debugfs_create_str); +static ssize_t debugfs_read_file_cpumask(struct file *file, + char __user *user_buf, + size_t count, loff_t *ppos) +{ + struct dentry *dentry = F_DENTRY(file); + struct cpumask *cpumask; + char *kernel_buf; + ssize_t ret; + int len; + + ret = debugfs_file_get(dentry); + if (unlikely(ret)) + return ret; + + /* How long is a piece of string? */ + kernel_buf = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (!kernel_buf) { + debugfs_file_put(dentry); + return -ENOMEM; + } + + cpumask = (struct cpumask *)file->private_data; + len = scnprintf(kernel_buf, PAGE_SIZE, + "%*pb\n", cpumask_pr_args(cpumask)); + debugfs_file_put(dentry); + if (len + 1 >= PAGE_SIZE) { + kfree(kernel_buf); + return -EIO; + } + + ret = simple_read_from_buffer(user_buf, count, ppos, kernel_buf, len); + kfree(kernel_buf); + + return ret; +} + +static const struct file_operations fops_cpumask_ro = { + .read = debugfs_read_file_cpumask, + .open = simple_open, + .llseek = default_llseek, +}; + +/** + * debugfs_create_cpumask - create a read-only debugfs file that is used to read a cpumask + * @name: a pointer to a string containing the name of the file to create. + * @mode: the permission that the file should have + * @parent: a pointer to the parent dentry for this file. This should be a + * directory dentry if set. If this parameter is %NULL, then the + * file will be created in the root of the debugfs filesystem. + * @value: a pointer to the variable that the file should read from. + * + * This function creates a file in debugfs with the given name that + * contains the value of the variable @value. + */ +void debugfs_create_cpumask(const char *name, umode_t mode, + struct dentry *parent, struct cpumask *value) +{ + /* Only read-only is supported */ + WARN_ON_ONCE(mode & S_IWUGO); + + debugfs_create_mode_unsafe(name, mode, parent, value, &fops_cpumask_ro, + &fops_cpumask_ro, &fops_cpumask_ro); +} + static ssize_t read_file_blob(struct file *file, char __user *user_buf, size_t count, loff_t *ppos) { diff --git a/include/linux/debugfs.h b/include/linux/debugfs.h index ea2d919fd9c79..4dfc67da0bb00 100644 --- a/include/linux/debugfs.h +++ b/include/linux/debugfs.h @@ -140,6 +140,8 @@ void debugfs_create_bool(const char *name, umode_t mode, struct dentry *parent, bool *value); void debugfs_create_str(const char *name, umode_t mode, struct dentry *parent, char **value); +void debugfs_create_cpumask(const char *name, umode_t mode, + struct dentry *parent, struct cpumask *value); struct dentry *debugfs_create_blob(const char *name, umode_t mode, struct dentry *parent, @@ -320,6 +322,10 @@ static inline void debugfs_create_str(const char *name, umode_t mode, char **value) { } +static inline void debugfs_create_cpumask(const char *name, umode_t mode, + struct dentry *parent, struct cpumask *value) +{ } + static inline struct dentry *debugfs_create_blob(const char *name, umode_t mode, struct dentry *parent, struct debugfs_blob_wrapper *blob) -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Not all of MPAM is visible through the resctrl user-space interface. To make it easy to debug why certain devices were not exposed through resctrl, allow the properties of the devices to be read through debugfs. This adds an mpam directory to debugfs, and exposes the devices as well as the hierarchy that was built. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 117 ++++++++++++++++++++++++-- drivers/platform/mpam/mpam_internal.h | 11 +++ 2 files changed, 122 insertions(+), 6 deletions(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 3c93dc6b2b1ca..418a341f3821a 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -394,6 +394,9 @@ static void mpam_msc_destroy(struct mpam_msc *msc) list_for_each_entry_safe(ris, tmp, &msc->ris, msc_list) mpam_ris_destroy(ris); + + debugfs_remove_recursive(msc->debugfs); + msc->debugfs = NULL; } /* @@ -625,6 +628,7 @@ u16 mpam_cpbm_wd_hisi_workaround(u16 cpbm_wd, enum mpam_device_features feat, return cpbm_wd; } +static struct dentry *mpam_debugfs; static void mpam_ris_hw_probe(struct mpam_msc_ris *ris) { int err; @@ -637,25 +641,25 @@ static void mpam_ris_hw_probe(struct mpam_msc_ris *ris) /* Cache Capacity Partitioning */ if (FIELD_GET(MPAMF_IDR_HAS_CCAP_PART, ris->idr)) { - u32 ccap_features = mpam_read_partsel_reg(msc, CCAP_IDR); + ris->ccap_idr = mpam_read_partsel_reg(msc, CCAP_IDR); - props->cmax_wd = FIELD_GET(MPAMF_CCAP_IDR_CMAX_WD, ccap_features); + props->cmax_wd = FIELD_GET(MPAMF_CCAP_IDR_CMAX_WD, ris->ccap_idr); if (props->cmax_wd) { - if (!FIELD_GET(MPAMF_CCAP_IDR_NO_CMAX, ccap_features)) + if (!FIELD_GET(MPAMF_CCAP_IDR_NO_CMAX, ris->ccap_idr)) mpam_set_feature(mpam_feat_ccap_part, props); - if (FIELD_GET(MPAMF_CCAP_IDR_HAS_CMIN, ccap_features)) + if (FIELD_GET(MPAMF_CCAP_IDR_HAS_CMIN, ris->ccap_idr)) mpam_set_feature(mpam_feat_cmin, props); } } /* Cache Portion partitioning */ if (FIELD_GET(MPAMF_IDR_HAS_CPOR_PART, ris->idr)) { - u32 cpor_features = mpam_read_partsel_reg(msc, CPOR_IDR); + ris->cpor_idr = mpam_read_partsel_reg(msc, CPOR_IDR); props->cpbm_wd = mpam_cpbm_wd_hisi_workaround( - FIELD_GET(MPAMF_CPOR_IDR_CPBM_WD, cpor_features), + FIELD_GET(MPAMF_CPOR_IDR_CPBM_WD, ris->cpor_idr), mpam_feat_cpor_part, class->level); if (props->cpbm_wd) mpam_set_feature(mpam_feat_cpor_part, props); @@ -1992,6 +1996,7 @@ static int mpam_msc_drv_probe(struct platform_device *pdev) { int err; pgprot_t prot; + char name[20]; void * __iomem io; struct mpam_msc *msc; struct resource *msc_res; @@ -2090,6 +2095,11 @@ static int mpam_msc_drv_probe(struct platform_device *pdev) list_add_rcu(&msc->glbl_list, &mpam_all_msc); platform_set_drvdata(pdev, msc); + + snprintf(name, sizeof(name), "msc.%u", msc->id); + msc->debugfs = debugfs_create_dir(name, mpam_debugfs); + debugfs_create_x32("max_nrdy_usec", 0400, msc->debugfs, &msc->nrdy_usec); + } while (0); mutex_unlock(&mpam_list_lock); @@ -2448,6 +2458,97 @@ static int mpam_allocate_config(void) return 0; } +static void mpam_debugfs_setup_ris(struct mpam_msc_ris *ris) +{ + char name[40]; + struct dentry *d; + struct mpam_props *rprops = &ris->props; + + snprintf(name, sizeof(name), "ris.%u", ris->ris_idx); + d = debugfs_create_dir(name, ris->msc->debugfs); + debugfs_create_x64("mpamf_idr", 0400, d, &ris->idr); + debugfs_create_x32("mpamf_cpor_idr", 0400, d, &ris->cpor_idr); + debugfs_create_x32("mpamf_ccap_idr", 0400, d, &ris->ccap_idr); + debugfs_create_x32("features", 0400, d, &rprops->features); + debugfs_create_x16("cpbm_wd", 0400, d, &rprops->cpbm_wd); + debugfs_create_x16("cmax_wd", 0400, d, &rprops->cmax_wd); + debugfs_create_x16("mbw_pbm_bits", 0400, d, &rprops->mbw_pbm_bits); + debugfs_create_x16("intpri_wd", 0400, d, &rprops->intpri_wd); + debugfs_create_x8("bwa_wd", 0400, d, &rprops->bwa_wd); + debugfs_create_x8("mbwu_scale", 0400, d, &rprops->mbwu_scale); + debugfs_create_x16("num_csu_mon", 0400, d, &rprops->num_csu_mon); + debugfs_create_x16("num_mbwu_mon", 0400, d, &rprops->num_mbwu_mon); + debugfs_create_cpumask("affinity", 0400, d, &ris->affinity); + ris->debugfs = d; +} + +static void mpam_debugfs_setup_comp_ris(struct mpam_component *comp, + struct mpam_msc_ris *ris) +{ + char name[40]; + char path[40]; + u8 ris_idx = ris->ris_idx; + int msc_id = ris->msc->id; + struct dentry *d = comp->debugfs; + + snprintf(name, sizeof(name), "msc.%u_ris.%u", + msc_id, ris_idx); + snprintf(path, sizeof(path), "../../msc.%u/ris.%u", + msc_id, ris_idx); + debugfs_create_symlink(name, d, path); +} + +static void mpam_debugfs_setup_comp(struct mpam_class *class, + struct mpam_component *comp) +{ + char name[40]; + struct dentry *d; + struct mpam_msc_ris *ris; + + snprintf(name, sizeof(name), "comp.%u", comp->comp_id); + d = debugfs_create_dir(name, class->debugfs); + comp->debugfs = d; + + list_for_each_entry_rcu(ris, &comp->ris, comp_list) + mpam_debugfs_setup_comp_ris(comp, ris); +} + +static void mpam_debugfs_setup(void) +{ + char name[40]; + struct dentry *d; + struct mpam_msc *msc; + struct mpam_class *class; + struct mpam_msc_ris *ris; + struct mpam_component *comp; + + lockdep_assert_held(&mpam_list_lock); + + list_for_each_entry(msc, &mpam_all_msc, glbl_list) { + d = msc->debugfs; + debugfs_create_u32("fw_id", 0400, d, &msc->pdev->id); + debugfs_create_x32("iface", 0400, d, &msc->iface); + debugfs_create_u16("partid_max", 0400, d, &msc->partid_max); + debugfs_create_u16("intpartid_max", 0400, d, &msc->intpartid_max); + debugfs_create_u8("pmg_max", 0400, d, &msc->pmg_max); + list_for_each_entry(ris, &msc->ris, msc_list) + mpam_debugfs_setup_ris(ris); + } + + list_for_each_entry_rcu(class, &mpam_classes, classes_list) { + snprintf(name, sizeof(name), "class.%u", class->level); + d = debugfs_create_dir(name, mpam_debugfs); + debugfs_create_x32("features", 0400, d, &class->props.features); + debugfs_create_x32("nrdy_usec", 0400, d, &class->nrdy_usec); + debugfs_create_x8("level", 0400, d, &class->level); + debugfs_create_cpumask("affinity", 0400, d, &class->affinity); + class->debugfs = d; + + list_for_each_entry_rcu(comp, &class->components, class_list) + mpam_debugfs_setup_comp(class, comp); + } +} + static void mpam_enable_once(void) { int err; @@ -2473,6 +2574,8 @@ static void mpam_enable_once(void) pr_warn("Failed to register irqs: %d\n", err); break; } + + mpam_debugfs_setup(); } while (0); mutex_unlock(&mpam_list_lock); cpus_read_unlock(); @@ -2757,6 +2860,8 @@ static int __init mpam_msc_driver_init(void) if (acpi_disabled) mpam_dt_create_foundling_msc(); + mpam_debugfs = debugfs_create_dir("mpam", NULL); + return platform_driver_register(&mpam_msc_driver); } /* Must occur after arm64_mpam_register_cpus() from arch_initcall() */ diff --git a/drivers/platform/mpam/mpam_internal.h b/drivers/platform/mpam/mpam_internal.h index ce5af77e3d300..c7087ca35e200 100644 --- a/drivers/platform/mpam/mpam_internal.h +++ b/drivers/platform/mpam/mpam_internal.h @@ -7,6 +7,7 @@ #include <linux/arm_mpam.h> #include <linux/atomic.h> #include <linux/cpumask.h> +#include <linux/debugfs.h> #include <linux/io.h> #include <linux/jump_label.h> #include <linux/mailbox_client.h> @@ -75,6 +76,8 @@ struct mpam_msc spinlock_t mon_sel_lock; void __iomem * mapped_hwpage; size_t mapped_hwpage_sz; + + struct dentry *debugfs; }; /* @@ -158,6 +161,8 @@ struct mpam_class struct ida ida_csu_mon; struct ida ida_mbwu_mon; + + struct dentry *debugfs; }; struct mpam_config { @@ -200,6 +205,8 @@ struct mpam_component /* parent: */ struct mpam_class *class; + + struct dentry *debugfs; }; /* The values for MSMON_CFG_MBWU_FLT.RWBW */ @@ -237,6 +244,8 @@ struct msmon_mbwu_state { struct mpam_msc_ris { u8 ris_idx; u64 idr; + u32 cpor_idr; + u32 ccap_idr; struct mpam_props props; bool in_reset_state; @@ -254,6 +263,8 @@ struct mpam_msc_ris { /* msmon mbwu configuration is preserved over reset */ struct msmon_mbwu_state *mbwu_state; + + struct dentry *debugfs; }; struct mpam_resctrl_dom { -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ Add a pr_debug() statement to arm_smmu_group_set_mpam() to log the SMMU features, PARTID, PMG, and Stream ID when binding MPAM parameters to a Stream Table Entry (STE). This aids debugging of MPAM (Memory System Resource Partitioning and Monitoring) configuration issues, especially when verifying that the correct partition ID and performance monitoring group are being programmed for each stream. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c index fa2bb54a49e4e..f63fab2a8d11d 100644 --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c @@ -4155,6 +4155,9 @@ static int arm_smmu_group_set_mpam(struct iommu_group *group, u16 partid, sid = master->streams[i].id; step = arm_smmu_get_step_for_sid(smmu, sid); + pr_debug("smmu feat 0x%x bind partid %d pmg %d to sid %d\n", + smmu->features, partid, pmg, sid); + /* These need locking if the VMSPtr is ever used */ step->data[4] = FIELD_PREP(STRTAB_STE_4_PARTID, partid); step->data[5] = FIELD_PREP(STRTAB_STE_5_PMG, pmg); -- 2.43.0
hulk inclusion category: feature bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ This reverts commit 43132396266db21803149f07afff112bc56f19a6. The rdtgroup_rename() operation for monitor group migration involves RMID free and reallocation under MPAM, which alters the RDT operational flow. Additionally, on kernfs_rename() failure, resctrl_find_free_rmid() is not properly rolled back. Since upstream MPAM does not yet fully support this functionality, the conditions for monitor group 'mv' operations are not currently met. Return -EPERM directly to disallow the operation. Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- fs/resctrl/internal.h | 1 - fs/resctrl/monitor.c | 2 +- fs/resctrl/rdtgroup.c | 20 +++----------------- 3 files changed, 4 insertions(+), 19 deletions(-) diff --git a/fs/resctrl/internal.h b/fs/resctrl/internal.h index 4c5ddd03ce08a..a5f35e72ebc69 100644 --- a/fs/resctrl/internal.h +++ b/fs/resctrl/internal.h @@ -299,6 +299,5 @@ bool has_busy_rmid(struct rdt_domain *d); void __check_limbo(struct rdt_domain *d, bool force_free); void rdt_staged_configs_clear(void); int resctrl_find_cleanest_closid(void); -struct rmid_entry *resctrl_find_free_rmid(u32 closid); #endif /* _FS_RESCTRL_INTERNAL_H */ diff --git a/fs/resctrl/monitor.c b/fs/resctrl/monitor.c index b3eb1a02ac607..e152c646f35b2 100644 --- a/fs/resctrl/monitor.c +++ b/fs/resctrl/monitor.c @@ -211,7 +211,7 @@ static struct rmid_entry *__resctrl_find_free_rmid(u32 closid) return ERR_PTR(-ENOSPC); } -struct rmid_entry *resctrl_find_free_rmid(u32 closid) +static struct rmid_entry *resctrl_find_free_rmid(u32 closid) { struct rmid_entry *err; int ret; diff --git a/fs/resctrl/rdtgroup.c b/fs/resctrl/rdtgroup.c index 601c74e18315b..2fab48af47000 100644 --- a/fs/resctrl/rdtgroup.c +++ b/fs/resctrl/rdtgroup.c @@ -3864,8 +3864,6 @@ static void mongrp_reparent(struct rdtgroup *rdtgrp, list_move_tail(&rdtgrp->mon.crdtgrp_list, &new_prdtgrp->mon.crdtgrp_list); - free_rmid(rdtgrp->closid, rdtgrp->mon.rmid); - rdtgrp->mon.rmid = alloc_rmid(new_prdtgrp->closid); rdtgrp->mon.parent = new_prdtgrp; rdtgrp->closid = new_prdtgrp->closid; @@ -3880,11 +3878,13 @@ static int rdtgroup_rename(struct kernfs_node *kn, { struct kernfs_node *kn_parent; struct rdtgroup *new_prdtgrp; - struct rmid_entry *entry; struct rdtgroup *rdtgrp; cpumask_var_t tmpmask; int ret; + if (IS_ENABLED(CONFIG_ARM64_MPAM)) + return -EPERM; + rdtgrp = kernfs_to_rdtgroup(kn); new_prdtgrp = kernfs_to_rdtgroup(new_parent); if (!rdtgrp || !new_prdtgrp) @@ -3940,20 +3940,6 @@ static int rdtgroup_rename(struct kernfs_node *kn, goto out; } - /* - * Unlike RDT, the rmid and closid in MPAM have a hierarchical - * relationship. Therefore, first check whether there are still - * free rmids available under the target closid. - */ - if (IS_ENABLED(CONFIG_ARM64_MPAM)) { - entry = resctrl_find_free_rmid(new_prdtgrp->closid); - if (IS_ERR(entry)) { - rdt_last_cmd_puts("Destination has been out of RMIDs\n"); - ret = PTR_ERR(entry); - goto out; - } - } - /* * Allocate the cpumask for use in mongrp_reparent() to avoid the * possibility of failing to allocate it after kernfs_rename() has -- 2.43.0
hulk inclusion category: bugfix bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ iommu_group_get_from_kobj() returns holding a reference on group->devices_kobj, which iommu_group_put() releases. The kobject_get() call only compensates for the kobject_put(&group->kobj) done internally by iommu_group_get_from_kobj(). Since iommu_group_put() was never called, this leaks one iommu_group reference per group on every call (e.g. on rmdir and every tasks file read). Since device_kobj pins the group's kobject, the leaked groups can never be freed. kset_get_next_obj() only drops its reference on the current kobject when it is passed back as 'prev' on the next iteration. Breaking out of the loop leaks the group_kobj reference taken by the iterator. The same patterns exist in both rdt_move_group_iommus() and show_rdt_iommu(). Fixes: 7e3f95a1a861 ("fs/resctrl: Add support for assigning iommu_groups to resctrl groups") Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- fs/resctrl/rdtgroup.c | 22 +++++++++++++++------- 1 file changed, 15 insertions(+), 7 deletions(-) diff --git a/fs/resctrl/rdtgroup.c b/fs/resctrl/rdtgroup.c index 2fab48af47000..17100c37b635f 100644 --- a/fs/resctrl/rdtgroup.c +++ b/fs/resctrl/rdtgroup.c @@ -815,17 +815,21 @@ static int rdt_move_group_iommus(struct rdtgroup *from, struct rdtgroup *to) kobject_get(group_kobj); group = iommu_group_get_from_kobj(group_kobj); - if (!group) + if (!group) { + kobject_put(group_kobj); continue; + } if (!from || iommu_matches_rdtgroup(group, from)) { err = kstrtoint(group_kobj->name, 0, &iommu_group_id); - if (err) - break; + if (!err) + err = rdtgroup_move_iommu(iommu_group_id, to); + } - err = rdtgroup_move_iommu(iommu_group_id, to); - if (err) - break; + iommu_group_put(group); + if (err) { + kobject_put(group_kobj); + break; } } @@ -926,11 +930,15 @@ static void show_rdt_iommu(struct rdtgroup *r, struct seq_file *s) kobject_get(group_kobj); group = iommu_group_get_from_kobj(group_kobj); - if (!group) + if (!group) { + kobject_put(group_kobj); continue; + } if (iommu_matches_rdtgroup(group, r)) seq_printf(s, "iommu_group:%s\n", group_kobj->name); + + iommu_group_put(group); } kset_put(iommu_groups); -- 2.43.0
hulk inclusion category: bugfix bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ In __destroy_component_cfg() comp->cfg is freed with kfree() but not set to NULL, leaving a dangling pointer. If the function is subsequently called again (e.g. from the error path of __allocate_component_cfg()), the check "if (comp->cfg)" in __allocate_component_cfg() would see the stale non-NULL pointer and skip re-allocation, leading to a use-after-free when comp->cfg[partid] is later accessed. NULL the pointer after freeing to avoid the dangling reference. Fixes: 21ac9fc8f275 ("arm_mpam: Track bandwidth counter state for overflow and power management") Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 418a341f3821a..73f105072ff43 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -2392,6 +2392,7 @@ static void __destroy_component_cfg(struct mpam_component *comp) struct msmon_mbwu_state *mbwu_state; kfree(comp->cfg); + comp->cfg = NULL; list_for_each_entry(ris, &comp->ris, comp_list) { mutex_lock(&ris->msc->lock); spin_lock_irqsave(&ris->msc->mon_sel_lock, flags); -- 2.43.0
hulk inclusion category: bugfix bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ mpam_msc_drv_remove() calls mpam_msc_destroy() before synchronize_srcu(). mpam_msc_destroy() frees the per-RIS state (via mpam_ris_destroy()), but readers that entered a read-side critical section by locking mpam_srcu may still hold pointers into that memory, and the SRCU grace period is only waited for afterwards. This leaves a window for an RCU-side use-after-free after the MSC is unbound. Call mpam_msc_destroy() only after synchronize_srcu() so all read-side critical sections have completed before the RIS state is freed. Fixes: 848cefee5a21 ("arm_mpam: Add the class and component structures for ris firmware described") Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 73f105072ff43..2f44e1cfbd1a8 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -2710,8 +2710,8 @@ static int mpam_msc_drv_remove(struct platform_device *pdev) mpam_num_msc--; platform_set_drvdata(pdev, NULL); list_del_rcu(&msc->glbl_list); - mpam_msc_destroy(msc); synchronize_srcu(&mpam_srcu); + mpam_msc_destroy(msc); mutex_unlock(&mpam_list_lock); return 0; -- 2.43.0
hulk inclusion category: bugfix bugzilla: https://atomgit.com/openeuler/kernel/issues/9840 ------------------ error_dev_id is allocated with alloc_percpu_gfp() in __setup_ppi(). Since it is a plain percpu allocation (not devm-managed like msc), devm_kfree() in the mpam_msc_setup() does not release it. When a PPI sharing conflict is detected and the function returns -EBUSY, error_dev_id is left allocated, leaking percpu memory. Free it before returning the error. Fixes: 79a4fc6c372c ("arm_mpam: Register and enable IRQs") Signed-off-by: Zeng Heng <zengheng4@huawei.com> --- drivers/platform/mpam/mpam_devices.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/platform/mpam/mpam_devices.c b/drivers/platform/mpam/mpam_devices.c index 2f44e1cfbd1a8..f8b6fea70784c 100644 --- a/drivers/platform/mpam/mpam_devices.c +++ b/drivers/platform/mpam/mpam_devices.c @@ -1835,6 +1835,7 @@ static int __setup_ppi(struct mpam_msc *msc) pr_err_once("%s shares PPI with %s!\n", dev_name(&msc->pdev->dev), dev_name(&empty->pdev->dev)); + free_percpu(msc->error_dev_id); return -EBUSY; } *per_cpu_ptr(msc->error_dev_id, cpu) = msc; -- 2.43.0
反馈： 您发送到kernel@openeuler.org的补丁/补丁集，已成功转换为PR！ PR链接地址： https://atomgit.com/openeuler/kernel/merge_requests/27127 邮件列表地址：https://mailweb.openeuler.org/archives/list/kernel@openeuler.org/message/MJV... FeedBack: The patch(es) which you have sent to kernel@openeuler.org mailing list has been converted to a pull request successfully! Pull request link: https://atomgit.com/openeuler/kernel/merge_requests/27127 Mailing list address: https://mailweb.openeuler.org/archives/list/kernel@openeuler.org/message/MJV...
participants (2)
-
patchwork bot -
Zeng Heng