linux-kernel - [PATCH 3/3] perf/arm-cmn: Enable per-DTC counter allocation

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <849f65566582cb102c6d0843d0f26e231180f8ac.1697824215.git.robin.murphy@arm.com>
Date:   Fri, 20 Oct 2023 18:51:27 +0100
From:   Robin Murphy <robin.murphy@....com>
To:     will@...nel.org
Cc:     mark.rutland@....com, linux-arm-kernel@...ts.infradead.org,
        linux-kernel@...r.kernel.org, jeremy.linton@....com,
        ilkka@...amperecomputing.com, renyu.zj@...ux.alibaba.com
Subject: [PATCH 3/3] perf/arm-cmn: Enable per-DTC counter allocation

Finally enable independent per-DTC-domain counter allocation, except on
CMN-600 where we still need to cope with not knowing the domain topology
and thus keep counter indices sychronised across domains. This allows
users to simultaneously count up to 8 targeted events per domain, rather
than 8 globally, for up to 4x wider coverage on maximum configurations.

Even though this now looks deceptively simple, I stand by my previous
assertion that it was a flippin' nightmare to implement; all the real
head-scratchers are hidden in the foundations in the previous patch...

Signed-off-by: Robin Murphy <robin.murphy@....com>
---
 drivers/perf/arm-cmn.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c
index 675f1638013e..9479e919c063 100644
--- a/drivers/perf/arm-cmn.c
+++ b/drivers/perf/arm-cmn.c
@@ -1570,7 +1570,7 @@ struct arm_cmn_val {
 	u8 dtm_count[CMN_MAX_DTMS];
 	u8 occupid[CMN_MAX_DTMS][SEL_MAX];
 	u8 wp[CMN_MAX_DTMS][4];
-	int dtc_count;
+	int dtc_count[CMN_MAX_DTCS];
 	bool cycles;
 };
 
@@ -1591,7 +1591,8 @@ static void arm_cmn_val_add_event(struct arm_cmn *cmn, struct arm_cmn_val *val,
 		return;
 	}
 
-	val->dtc_count++;
+	for_each_hw_dtc_idx(hw, dtc, idx)
+		val->dtc_count[dtc]++;
 
 	for_each_hw_dn(hw, dn, i) {
 		int wp_idx, dtm = dn->dtm, sel = hw->filter_sel;
@@ -1638,8 +1639,9 @@ static int arm_cmn_validate_group(struct arm_cmn *cmn, struct perf_event *event)
 		goto done;
 	}
 
-	if (val->dtc_count == CMN_DT_NUM_COUNTERS)
-		goto done;
+	for (i = 0; i < CMN_MAX_DTCS; i++)
+		if (val->dtc_count[i] == CMN_DT_NUM_COUNTERS)
+			goto done;
 
 	for_each_hw_dn(hw, dn, i) {
 		int wp_idx, wp_cmb, dtm = dn->dtm, sel = hw->filter_sel;
@@ -1806,9 +1808,9 @@ static int arm_cmn_event_add(struct perf_event *event, int flags)
 		return 0;
 	}
 
-	/* Grab a free global counter first... */
+	/* Grab the global counters first... */
 	for_each_hw_dtc_idx(hw, j, idx) {
-		if (j > 0) {
+		if (cmn->part == PART_CMN600 && j > 0) {
 			idx = hw->dtc_idx[0];
 		} else {
 			idx = 0;
@@ -1819,10 +1821,10 @@ static int arm_cmn_event_add(struct perf_event *event, int flags)
 		hw->dtc_idx[j] = idx;
 	}
 
-	/* ...then the local counters to feed it. */
+	/* ...then the local counters to feed them */
 	for_each_hw_dn(hw, dn, i) {
 		struct arm_cmn_dtm *dtm = &cmn->dtms[dn->dtm] + hw->dtm_offset;
-		unsigned int dtm_idx, shift, d = 0;
+		unsigned int dtm_idx, shift, d = max_t(int, dn->dtc, 0);
 		u64 reg;
 
 		dtm_idx = 0;
-- 
2.39.2.101.g768bb238c484.dirty