linux-kernel - [PATCH] Partially revert "perf/arm-cmn: Optimise DTC counter accesses"

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-next>] [day] [month] [year] [list]

Message-Id: <b41bb4ed7283c3d8400ce5cf5e6ec94915e6750f.1674498637.git.robin.murphy@arm.com>
Date:   Mon, 23 Jan 2023 18:30:38 +0000
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, ilkka@...amperecomputing.com
Subject: [PATCH] Partially revert "perf/arm-cmn: Optimise DTC counter accesses"

It turns out the optimisation implemented by commit 4f2c3872dde5 is
totally broken, since all the places that consume hw->dtcs_used for
events other than cycle count are still not expecting it to be sparsely
populated, and fail to read all the relevant DTC counters correctly if
so.

If implemented correctly, the optimisation potentially saves up to 3
register reads per event update, which is reasonably significant for
events targeting a single node, but still not worth a massive amount of
additional code complexity overall. Getting it right within the current
design looks a fair bit more involved than it was ever intended to be,
so let's just make a functional revert which restores the old behaviour
while still backporting easily.

Fixes: 4f2c3872dde5 ("perf/arm-cmn: Optimise DTC counter accesses")
Reported-by: Ilkka Koskinen <ilkka@...amperecomputing.com>
Signed-off-by: Robin Murphy <robin.murphy@....com>
---
 drivers/perf/arm-cmn.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c
index b80a9b74662b..1deb61b22bc7 100644
--- a/drivers/perf/arm-cmn.c
+++ b/drivers/perf/arm-cmn.c
@@ -1576,7 +1576,6 @@ static int arm_cmn_event_init(struct perf_event *event)
 			hw->dn++;
 			continue;
 		}
-		hw->dtcs_used |= arm_cmn_node_to_xp(cmn, dn)->dtc;
 		hw->num_dns++;
 		if (bynodeid)
 			break;
@@ -1589,6 +1588,12 @@ static int arm_cmn_event_init(struct perf_event *event)
 			nodeid, nid.x, nid.y, nid.port, nid.dev, type);
 		return -EINVAL;
 	}
+	/*
+	 * Keep assuming non-cycles events count in all DTC domains; turns out
+	 * it's hard to make a worthwhile optimisation around this, short of
+	 * going all-in with domain-local counter allocation as well.
+	 */
+	hw->dtcs_used = (1U << cmn->num_dtcs) - 1;

 	return arm_cmn_validate_group(cmn, event);
 }
-- 
2.36.1.dirty