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: <1634621361-17155-3-git-send-email-quic_fenglinw@quicinc.com>
Date:   Tue, 19 Oct 2021 13:29:13 +0800
From:   Fenglin Wu <quic_fenglinw@...cinc.com>
To:     <linux-arm-msm@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <sboyd@...nel.org>
CC:     <collinsd@...eaurora.org>, <subbaram@...eaurora.org>,
        <quic_fenglinw@...cinc.com>, <tglx@...utronix.de>, <maz@...nel.org>
Subject: [PATCH v2 02/10] spmi: pmic-arb: do not ack and clear peripheral interrupts in cleanup_irq

From: Subbaraman Narayanamurthy <subbaram@...eaurora.org>

Currently, cleanup_irq() is invoked when a peripheral's interrupt
fires and there is no mapping present in the interrupt domain of
spmi interrupt controller.

The cleanup_irq clears the arbiter bit, clears the pmic interrupt
and disables it at the pmic in that order. The last disable in
cleanup_irq races with request_irq() in that it stomps over the
enable issued by request_irq. Fix this by not writing to the pmic
in cleanup_irq. The latched bit will be left set in the pmic,
which will not send us more interrupts even if the enable bit
stays enabled.

When a client wants to request an interrupt, use the activate
callback on the irq_domain to clear latched bit. This ensures
that the latched, if set due to the above changes in cleanup_irq
or when the bootloader leaves it set, gets cleaned up, paving way
for upcoming interrupts to trigger.

With this, there is a possibility of unwanted triggering of
interrupt right after the latched bit is cleared - the interrupt
may be left enabled too. To avoid that, clear the enable first
followed by clearing the latched bit in the activate callback.

Signed-off-by: Subbaraman Narayanamurthy <subbaram@...eaurora.org>
[collinsd@...eaurora.org: fix merge conflict]
Signed-off-by: David Collins <collinsd@...eaurora.org>
Signed-off-by: Fenglin Wu <quic_fenglinw@...cinc.com>
---
 drivers/spmi/spmi-pmic-arb.c | 15 +++++----------
 1 file changed, 5 insertions(+), 10 deletions(-)

diff --git a/drivers/spmi/spmi-pmic-arb.c b/drivers/spmi/spmi-pmic-arb.c
index da629cc..ce7ae99 100644
--- a/drivers/spmi/spmi-pmic-arb.c
+++ b/drivers/spmi/spmi-pmic-arb.c
@@ -492,16 +492,6 @@ static void cleanup_irq(struct spmi_pmic_arb *pmic_arb, u16 apid, int id)
 	dev_err_ratelimited(&pmic_arb->spmic->dev, "%s apid=%d sid=0x%x per=0x%x irq=%d\n",
 			__func__, apid, sid, per, id);
 	writel_relaxed(irq_mask, pmic_arb->ver_ops->irq_clear(pmic_arb, apid));
-
-	if (pmic_arb_write_cmd(pmic_arb->spmic, SPMI_CMD_EXT_WRITEL, sid,
-			(per << 8) + QPNPINT_REG_LATCHED_CLR, &irq_mask, 1))
-		dev_err_ratelimited(&pmic_arb->spmic->dev, "failed to ack irq_mask = 0x%x for ppid = %x\n",
-				irq_mask, ppid);
-
-	if (pmic_arb_write_cmd(pmic_arb->spmic, SPMI_CMD_EXT_WRITEL, sid,
-			       (per << 8) + QPNPINT_REG_EN_CLR, &irq_mask, 1))
-		dev_err_ratelimited(&pmic_arb->spmic->dev, "failed to ack irq_mask = 0x%x for ppid = %x\n",
-				irq_mask, ppid);
 }
 
 static int periph_interrupt(struct spmi_pmic_arb *pmic_arb, u16 apid)
@@ -674,6 +664,7 @@ static int qpnpint_irq_domain_activate(struct irq_domain *domain,
 	u16 apid = hwirq_to_apid(d->hwirq);
 	u16 sid = hwirq_to_sid(d->hwirq);
 	u16 irq = hwirq_to_irq(d->hwirq);
+	u8 buf;
 
 	if (pmic_arb->apid_data[apid].irq_ee != pmic_arb->ee) {
 		dev_err(&pmic_arb->spmic->dev, "failed to xlate sid = %#x, periph = %#x, irq = %u: ee=%u but owner=%u\n",
@@ -682,6 +673,10 @@ static int qpnpint_irq_domain_activate(struct irq_domain *domain,
 		return -ENODEV;
 	}
 
+	buf = BIT(irq);
+	qpnpint_spmi_write(d, QPNPINT_REG_EN_CLR, &buf, 1);
+	qpnpint_spmi_write(d, QPNPINT_REG_LATCHED_CLR, &buf, 1);
+
 	return 0;
 }
 
-- 
2.7.4

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ