linux-kernel - [PATCH] powerpc/perf: prevent mixed EBB and non-EBB events

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]

Date:   Wed, 24 Feb 2021 09:21:16 -0300
From:   Thadeu Lima de Souza Cascardo <cascardo@...onical.com>
To:     linuxppc-dev@...ts.ozlabs.org
Cc:     mpe@...erman.id.au, peterz@...radead.org,
        linux-kernel@...r.kernel.org, cascardo@...onical.com
Subject: [PATCH] powerpc/perf: prevent mixed EBB and non-EBB events

EBB events must be under exclusive groups, so there is no mix of EBB and
non-EBB events on the same PMU. This requirement worked fine as perf core
would not allow other pinned events to be scheduled together with exclusive
events.

This assumption was broken by commit 1908dc911792 ("perf: Tweak
perf_event_attr::exclusive semantics").

After that, the test cpu_event_pinned_vs_ebb_test started succeeding after
read_events, but worse, the task would not have given access to PMC1, so
when it tried to write to it, it was killed with "illegal instruction".

Preventing mixed EBB and non-EBB events from being add to the same PMU will
just revert to the previous behavior and the test will succeed.

Fixes: 1908dc911792 (perf: Tweak perf_event_attr::exclusive semantics)
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@...onical.com>
---
 arch/powerpc/perf/core-book3s.c | 20 ++++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/perf/core-book3s.c b/arch/powerpc/perf/core-book3s.c
index 43599e671d38..d767f7944f85 100644
--- a/arch/powerpc/perf/core-book3s.c
+++ b/arch/powerpc/perf/core-book3s.c
@@ -1010,9 +1010,25 @@ static int check_excludes(struct perf_event **ctrs, unsigned int cflags[],
 			  int n_prev, int n_new)
 {
 	int eu = 0, ek = 0, eh = 0;
+	bool ebb = false;
 	int i, n, first;
 	struct perf_event *event;
 
+	n = n_prev + n_new;
+	if (n <= 1)
+		return 0;
+
+	first = 1;
+	for (i = 0; i < n; ++i) {
+		event = ctrs[i];
+		if (first) {
+			ebb = is_ebb_event(event);
+			first = 0;
+		} else if (is_ebb_event(event) != ebb) {
+			return -EAGAIN;
+		}
+	}
+
 	/*
 	 * If the PMU we're on supports per event exclude settings then we
 	 * don't need to do any of this logic. NB. This assumes no PMU has both
@@ -1021,10 +1037,6 @@ static int check_excludes(struct perf_event **ctrs, unsigned int cflags[],
 	if (ppmu->flags & PPMU_ARCH_207S)
 		return 0;
 
-	n = n_prev + n_new;
-	if (n <= 1)
-		return 0;
-
 	first = 1;
 	for (i = 0; i < n; ++i) {
 		if (cflags[i] & PPMU_LIMITED_PMC_OK) {
-- 
2.27.0