linux-kernel - [PATCH 1/2] perf/arm-cmn: Cope with spurious IRQs better

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: <bc8183fcd7f6ca07b0d56aa8061d816485b84341.1669822447.git.robin.murphy@arm.com>
Date:   Wed, 30 Nov 2022 15:36:38 +0000
From:   Robin Murphy <robin.murphy@....com>
To:     will@...nel.org, mark.rutland@....com
Cc:     linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        Geoff Blake <blakgeof@...zon.com>
Subject: [PATCH 1/2] perf/arm-cmn: Cope with spurious IRQs better

Although we reset the CMN state during probe before requesting our
interrupt(s), a pending IRQ could already have been latched at the
interrupt controller, and thus be delivered spuriously as soon as the
IRQ is enabled. Not handling that can then lead to the IRQ layer
disabling it again, and things subseuqently going wonky.

Since we can't support shared IRQs anyway for affinity-management
reasons, the only time we should concievably return IRQ_NONE is for a
spurious interrupt which did somehow originate from our CMN, so there
should be no harm in removing the problem by simply claiming to have
handled those as well.

Reported-by: Geoff Blake <blakgeof@...zon.com>
Signed-off-by: Robin Murphy <robin.murphy@....com>
---
 drivers/perf/arm-cmn.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/perf/arm-cmn.c b/drivers/perf/arm-cmn.c
index b80a9b74662b..9e8be5586423 100644
--- a/drivers/perf/arm-cmn.c
+++ b/drivers/perf/arm-cmn.c
@@ -1798,7 +1798,6 @@ static int arm_cmn_pmu_offline_cpu(unsigned int cpu, struct hlist_node *cpuhp_no
 static irqreturn_t arm_cmn_handle_irq(int irq, void *dev_id)
 {
 	struct arm_cmn_dtc *dtc = dev_id;
-	irqreturn_t ret = IRQ_NONE;
 
 	for (;;) {
 		u32 status = readl_relaxed(dtc->base + CMN_DT_PMOVSR);
@@ -1807,7 +1806,6 @@ static irqreturn_t arm_cmn_handle_irq(int irq, void *dev_id)
 
 		for (i = 0; i < CMN_DTM_NUM_COUNTERS; i++) {
 			if (status & (1U << i)) {
-				ret = IRQ_HANDLED;
 				if (WARN_ON(!dtc->counters[i]))
 					continue;
 				delta = (u64)arm_cmn_read_counter(dtc, i) << 16;
@@ -1816,7 +1814,6 @@ static irqreturn_t arm_cmn_handle_irq(int irq, void *dev_id)
 		}
 
 		if (status & (1U << CMN_DT_NUM_COUNTERS)) {
-			ret = IRQ_HANDLED;
 			if (dtc->cc_active && !WARN_ON(!dtc->cycles)) {
 				delta = arm_cmn_read_cc(dtc);
 				local64_add(delta, &dtc->cycles->count);
@@ -1826,7 +1823,7 @@ static irqreturn_t arm_cmn_handle_irq(int irq, void *dev_id)
 		writel_relaxed(status, dtc->base + CMN_DT_PMOVSR_CLR);
 
 		if (!dtc->irq_friend)
-			return ret;
+			return IRQ_HANDLED;
 		dtc += dtc->irq_friend;
 	}
 }
-- 
2.36.1.dirty