From nobody Tue Dec 16 03:22:38 2025 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id AD5A2C0015E for ; Fri, 7 Jul 2023 16:38:32 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S232919AbjGGQib (ORCPT ); Fri, 7 Jul 2023 12:38:31 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:55008 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S232908AbjGGQiZ (ORCPT ); Fri, 7 Jul 2023 12:38:25 -0400 Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by lindbergh.monkeyblade.net (Postfix) with ESMTP id 07F58211E for ; Fri, 7 Jul 2023 09:38:22 -0700 (PDT) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.121.207.14]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id BF7BC1516; Fri, 7 Jul 2023 09:39:04 -0700 (PDT) Received: from e121345-lin.cambridge.arm.com (e121345-lin.cambridge.arm.com [10.1.196.40]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPA id C851F3F740; Fri, 7 Jul 2023 09:38:21 -0700 (PDT) From: Robin Murphy To: will@kernel.org Cc: mark.rutland@arm.com, linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, ilkka@os.amperecomputing.com, blakgeof@amazon.com Subject: [PATCH 3/3] perf/arm-cmn: Add CMN-700 r3 support Date: Fri, 7 Jul 2023 17:38:13 +0100 Message-Id: <49153b72253f6af0e625cb55b9e1b825b110c49c.1688746690.git.robin.murphy@arm.com> X-Mailer: git-send-email 2.39.2.101.g768bb238c484.dirty In-Reply-To: References: MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Type: text/plain; charset="utf-8" CMN-700 r3 has a special configuration option for a so-called "Super Home Node", which is a superset of the standard HN-F that also manages remote-chip coherency for multi-chip setups. As such it has a similar but expanded set of PMU events compared to HN-F, with some additional filtering options to boot. Signed-off-by: Robin Murphy --- drivers/perf/arm-cmn.c | 94 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 94 insertions(+) diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c index 1efe9b72c0e6..913dc04b3a40 100644 --- a/drivers/perf/arm-cmn.c +++ b/drivers/perf/arm-cmn.c @@ -72,6 +72,8 @@ /* For most nodes, this is all there is */ #define CMN_PMU_EVENT_SEL 0x000 #define CMN__PMU_CBUSY_SNTHROTTLE_SEL GENMASK_ULL(44, 42) +#define CMN__PMU_SN_HOME_SEL GENMASK_ULL(40, 39) +#define CMN__PMU_HBT_LBT_SEL GENMASK_ULL(38, 37) #define CMN__PMU_CLASS_OCCUP_ID GENMASK_ULL(36, 35) /* Technically this is 4 bits wide on DNs, but we only use 2 there anyway = */ #define CMN__PMU_OCCUP1_ID GENMASK_ULL(34, 32) @@ -226,6 +228,7 @@ enum cmn_revision { REV_CMN700_R0P0 =3D 0, REV_CMN700_R1P0, REV_CMN700_R2P0, + REV_CMN700_R3P0, REV_CI700_R0P0 =3D 0, REV_CI700_R1P0, REV_CI700_R2P0, @@ -254,6 +257,9 @@ enum cmn_node_type { CMN_TYPE_CCHA, CMN_TYPE_CCLA, CMN_TYPE_CCLA_RNI, + CMN_TYPE_HNS =3D 0x200, + CMN_TYPE_HNS_MPAM_S, + CMN_TYPE_HNS_MPAM_NS, /* Not a real node type */ CMN_TYPE_WP =3D 0x7770 }; @@ -263,6 +269,8 @@ enum cmn_filter_select { SEL_OCCUP1ID, SEL_CLASS_OCCUP_ID, SEL_CBUSY_SNTHROTTLE_SEL, + SEL_HBT_LBT_SEL, + SEL_SN_HOME_SEL, SEL_MAX }; =20 @@ -768,6 +776,8 @@ static umode_t arm_cmn_event_attr_is_visible(struct kob= ject *kobj, CMN_EVENT_ATTR(CMN_ANY, ccla_##_name, CMN_TYPE_CCLA, _event) #define CMN_EVENT_CCLA_RNI(_name, _event) \ CMN_EVENT_ATTR(CMN_ANY, ccla_rni_##_name, CMN_TYPE_CCLA_RNI, _event) +#define CMN_EVENT_HNS(_name, _event) \ + CMN_EVENT_ATTR(CMN_ANY, hns_##_name, CMN_TYPE_HNS, _event) =20 #define CMN_EVENT_DVM(_model, _name, _event) \ _CMN_EVENT_DVM(_model, _name, _event, 0, SEL_NONE) @@ -803,6 +813,23 @@ static umode_t arm_cmn_event_attr_is_visible(struct ko= bject *kobj, #define CMN_EVENT_HNF_SNT(_model, _name, _event) \ CMN_EVENT_HN_SNT(_model, hnf_##_name, CMN_TYPE_HNF, _event) =20 +#define CMN_EVENT_HNS_OCC(_name, _event) \ + CMN_EVENT_HN_OCC(CMN_ANY, hns_##_name, CMN_TYPE_HNS, _event), \ + _CMN_EVENT_ATTR(CMN_ANY, hns_##_name##_rxsnp, CMN_TYPE_HNS, _event, 5, SE= L_OCCUP1ID), \ + _CMN_EVENT_ATTR(CMN_ANY, hns_##_name##_lbt, CMN_TYPE_HNS, _event, 6, SEL_= OCCUP1ID), \ + _CMN_EVENT_ATTR(CMN_ANY, hns_##_name##_hbt, CMN_TYPE_HNS, _event, 7, SEL_= OCCUP1ID) +#define CMN_EVENT_HNS_CLS( _name, _event) \ + CMN_EVENT_HN_CLS(CMN_ANY, hns_##_name, CMN_TYPE_HNS, _event) +#define CMN_EVENT_HNS_SNT(_name, _event) \ + CMN_EVENT_HN_SNT(CMN_ANY, hns_##_name, CMN_TYPE_HNS, _event) +#define CMN_EVENT_HNS_HBT(_name, _event) \ + _CMN_EVENT_ATTR(CMN_ANY, hns_##_name##_all, CMN_TYPE_HNS, _event, 0, SEL_= HBT_LBT_SEL), \ + _CMN_EVENT_ATTR(CMN_ANY, hns_##_name##_hbt, CMN_TYPE_HNS, _event, 1, SEL_= HBT_LBT_SEL), \ + _CMN_EVENT_ATTR(CMN_ANY, hns_##_name##_lbt, CMN_TYPE_HNS, _event, 2, SEL_= HBT_LBT_SEL) +#define CMN_EVENT_HNS_SNH(_name, _event) \ + _CMN_EVENT_ATTR(CMN_ANY, hns_##_name##_all, CMN_TYPE_HNS, _event, 0, SEL_= SN_HOME_SEL), \ + _CMN_EVENT_ATTR(CMN_ANY, hns_##_name##_sn, CMN_TYPE_HNS, _event, 1, SEL_S= N_HOME_SEL), \ + _CMN_EVENT_ATTR(CMN_ANY, hns_##_name##_home, CMN_TYPE_HNS, _event, 2, SEL= _SN_HOME_SEL) =20 #define _CMN_EVENT_XP_MESH(_name, _event) \ __CMN_EVENT_XP(e_##_name, (_event) | (0 << 2)), \ @@ -1151,6 +1178,66 @@ static struct attribute *arm_cmn_event_attrs[] =3D { CMN_EVENT_CCLA(pfwd_sndr_stalls_static_crd, 0x2a), CMN_EVENT_CCLA(pfwd_sndr_stalls_dynmaic_crd, 0x2b), =20 + CMN_EVENT_HNS_HBT(cache_miss, 0x01), + CMN_EVENT_HNS_HBT(slc_sf_cache_access, 0x02), + CMN_EVENT_HNS_HBT(cache_fill, 0x03), + CMN_EVENT_HNS_HBT(pocq_retry, 0x04), + CMN_EVENT_HNS_HBT(pocq_reqs_recvd, 0x05), + CMN_EVENT_HNS_HBT(sf_hit, 0x06), + CMN_EVENT_HNS_HBT(sf_evictions, 0x07), + CMN_EVENT_HNS(dir_snoops_sent, 0x08), + CMN_EVENT_HNS(brd_snoops_sent, 0x09), + CMN_EVENT_HNS_HBT(slc_eviction, 0x0a), + CMN_EVENT_HNS_HBT(slc_fill_invalid_way, 0x0b), + CMN_EVENT_HNS(mc_retries_local, 0x0c), + CMN_EVENT_HNS_SNH(mc_reqs_local, 0x0d), + CMN_EVENT_HNS(qos_hh_retry, 0x0e), + CMN_EVENT_HNS_OCC(qos_pocq_occupancy, 0x0f), + CMN_EVENT_HNS(pocq_addrhaz, 0x10), + CMN_EVENT_HNS(pocq_atomic_addrhaz, 0x11), + CMN_EVENT_HNS(ld_st_swp_adq_full, 0x12), + CMN_EVENT_HNS(cmp_adq_full, 0x13), + CMN_EVENT_HNS(txdat_stall, 0x14), + CMN_EVENT_HNS(txrsp_stall, 0x15), + CMN_EVENT_HNS(seq_full, 0x16), + CMN_EVENT_HNS(seq_hit, 0x17), + CMN_EVENT_HNS(snp_sent, 0x18), + CMN_EVENT_HNS(sfbi_dir_snp_sent, 0x19), + CMN_EVENT_HNS(sfbi_brd_snp_sent, 0x1a), + CMN_EVENT_HNS(intv_dirty, 0x1c), + CMN_EVENT_HNS(stash_snp_sent, 0x1d), + CMN_EVENT_HNS(stash_data_pull, 0x1e), + CMN_EVENT_HNS(snp_fwded, 0x1f), + CMN_EVENT_HNS(atomic_fwd, 0x20), + CMN_EVENT_HNS(mpam_hardlim, 0x21), + CMN_EVENT_HNS(mpam_softlim, 0x22), + CMN_EVENT_HNS(snp_sent_cluster, 0x23), + CMN_EVENT_HNS(sf_imprecise_evict, 0x24), + CMN_EVENT_HNS(sf_evict_shared_line, 0x25), + CMN_EVENT_HNS_CLS(pocq_class_occup, 0x26), + CMN_EVENT_HNS_CLS(pocq_class_retry, 0x27), + CMN_EVENT_HNS_CLS(class_mc_reqs_local, 0x28), + CMN_EVENT_HNS_CLS(class_cgnt_cmin, 0x29), + CMN_EVENT_HNS_SNT(sn_throttle, 0x2a), + CMN_EVENT_HNS_SNT(sn_throttle_min, 0x2b), + CMN_EVENT_HNS(sf_precise_to_imprecise, 0x2c), + CMN_EVENT_HNS(snp_intv_cln, 0x2d), + CMN_EVENT_HNS(nc_excl, 0x2e), + CMN_EVENT_HNS(excl_mon_ovfl, 0x2f), + CMN_EVENT_HNS(snp_req_recvd, 0x30), + CMN_EVENT_HNS(snp_req_byp_pocq, 0x31), + CMN_EVENT_HNS(dir_ccgha_snp_sent, 0x32), + CMN_EVENT_HNS(brd_ccgha_snp_sent, 0x33), + CMN_EVENT_HNS(ccgha_snp_stall, 0x34), + CMN_EVENT_HNS(lbt_req_hardlim, 0x35), + CMN_EVENT_HNS(hbt_req_hardlim, 0x36), + CMN_EVENT_HNS(sf_reupdate, 0x37), + CMN_EVENT_HNS(excl_sf_imprecise, 0x38), + CMN_EVENT_HNS(snp_pocq_addrhaz, 0x39), + CMN_EVENT_HNS(mc_retries_remote, 0x3a), + CMN_EVENT_HNS_SNH(mc_reqs_remote, 0x3b), + CMN_EVENT_HNS_CLS(class_mc_reqs_remote, 0x3c), + NULL }; =20 @@ -1392,6 +1479,10 @@ static int arm_cmn_set_event_sel_hi(struct arm_cmn_n= ode *dn, dn->occupid[fsel].val =3D occupid; reg =3D FIELD_PREP(CMN__PMU_CBUSY_SNTHROTTLE_SEL, dn->occupid[SEL_CBUSY_SNTHROTTLE_SEL].val) | + FIELD_PREP(CMN__PMU_SN_HOME_SEL, + dn->occupid[SEL_SN_HOME_SEL].val) | + FIELD_PREP(CMN__PMU_HBT_LBT_SEL, + dn->occupid[SEL_HBT_LBT_SEL].val) | FIELD_PREP(CMN__PMU_CLASS_OCCUP_ID, dn->occupid[SEL_CLASS_OCCUP_ID].val) | FIELD_PREP(CMN__PMU_OCCUP1_ID, @@ -2219,6 +2310,7 @@ static int arm_cmn_discover(struct arm_cmn *cmn, unsi= gned int rgn_offset) case CMN_TYPE_CCRA: case CMN_TYPE_CCHA: case CMN_TYPE_CCLA: + case CMN_TYPE_HNS: dn++; break; /* Nothing to see here */ @@ -2226,6 +2318,8 @@ static int arm_cmn_discover(struct arm_cmn *cmn, unsi= gned int rgn_offset) case CMN_TYPE_MPAM_NS: case CMN_TYPE_RNSAM: case CMN_TYPE_CXLA: + case CMN_TYPE_HNS_MPAM_S: + case CMN_TYPE_HNS_MPAM_NS: break; /* * Split "optimised" combination nodes into separate --=20 2.39.2.101.g768bb238c484.dirty