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: <20231206-rkisp-irq-fix-v2-4-6ba4185eeb1f@ideasonboard.com>
Date:   Wed, 06 Dec 2023 12:12:31 +0200
From:   Tomi Valkeinen <tomi.valkeinen@...asonboard.com>
To:     Dafna Hirschfeld <dafna@...tmail.com>,
        Laurent Pinchart <laurent.pinchart@...asonboard.com>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        Heiko Stuebner <heiko@...ech.de>,
        Paul Elder <paul.elder@...asonboard.com>
Cc:     Alexander Stein <alexander.stein@...tq-group.com>,
        kieran.bingham@...asonboard.com, umang.jain@...asonboard.com,
        aford173@...il.com, linux-media@...r.kernel.org,
        linux-rockchip@...ts.infradead.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        Tomi Valkeinen <tomi.valkeinen@...asonboard.com>
Subject: [PATCH v2 4/4] media: rkisp1: Fix IRQ disable race issue

In rkisp1_isp_stop() and rkisp1_csi_disable() the driver masks the
interrupts and then apparently assumes that the interrupt handler won't
be running, and proceeds in the stop procedure. This is not the case, as
the interrupt handler can already be running, which would lead to the
ISP being disabled while the interrupt handler handling a captured
frame.

This brings up two issues: 1) the ISP could be powered off while the
interrupt handler is still running and accessing registers, leading to
board lockup, and 2) the interrupt handler code and the code that
disables the streaming might do things that conflict.

It is not clear to me if 2) causes a real issue, but 1) can be seen with
a suitable delay (or printk in my case) in the interrupt handler,
leading to board lockup.

Signed-off-by: Tomi Valkeinen <tomi.valkeinen@...asonboard.com>
---
 drivers/media/platform/rockchip/rkisp1/rkisp1-csi.c | 14 +++++++++++++-
 drivers/media/platform/rockchip/rkisp1/rkisp1-isp.c | 20 +++++++++++++++++---
 2 files changed, 30 insertions(+), 4 deletions(-)

diff --git a/drivers/media/platform/rockchip/rkisp1/rkisp1-csi.c b/drivers/media/platform/rockchip/rkisp1/rkisp1-csi.c
index 47f4353a1784..0bab3303f2e4 100644
--- a/drivers/media/platform/rockchip/rkisp1/rkisp1-csi.c
+++ b/drivers/media/platform/rockchip/rkisp1/rkisp1-csi.c
@@ -125,8 +125,20 @@ static void rkisp1_csi_disable(struct rkisp1_csi *csi)
 	struct rkisp1_device *rkisp1 = csi->rkisp1;
 	u32 val;
 
-	/* Mask and clear interrupts. */
+	/* Mask MIPI interrupts. */
 	rkisp1_write(rkisp1, RKISP1_CIF_MIPI_IMSC, 0);
+
+	/* Flush posted writes */
+	rkisp1_read(rkisp1, RKISP1_CIF_MIPI_IMSC);
+
+	/*
+	 * Wait until the IRQ handler has ended. The IRQ handler may get called
+	 * even after this, but it will return immediately as the MIPI
+	 * interrupts have been masked.
+	 */
+	synchronize_irq(rkisp1->irqs[RKISP1_IRQ_MIPI]);
+
+	/* Clear MIPI interrupt status */
 	rkisp1_write(rkisp1, RKISP1_CIF_MIPI_ICR, ~0);
 
 	val = rkisp1_read(rkisp1, RKISP1_CIF_MIPI_CTRL);
diff --git a/drivers/media/platform/rockchip/rkisp1/rkisp1-isp.c b/drivers/media/platform/rockchip/rkisp1/rkisp1-isp.c
index dafbfd230542..33b5a714d117 100644
--- a/drivers/media/platform/rockchip/rkisp1/rkisp1-isp.c
+++ b/drivers/media/platform/rockchip/rkisp1/rkisp1-isp.c
@@ -364,11 +364,25 @@ static void rkisp1_isp_stop(struct rkisp1_isp *isp)
 	 * ISP(mi) stop in mi frame end -> Stop ISP(mipi) ->
 	 * Stop ISP(isp) ->wait for ISP isp off
 	 */
-	/* stop and clear MI and ISP interrupts */
-	rkisp1_write(rkisp1, RKISP1_CIF_ISP_IMSC, 0);
-	rkisp1_write(rkisp1, RKISP1_CIF_ISP_ICR, ~0);
 
+	/* Mask MI and ISP interrupts */
+	rkisp1_write(rkisp1, RKISP1_CIF_ISP_IMSC, 0);
 	rkisp1_write(rkisp1, RKISP1_CIF_MI_IMSC, 0);
+
+	/* Flush posted writes */
+	rkisp1_read(rkisp1, RKISP1_CIF_MI_IMSC);
+
+	/*
+	 * Wait until the IRQ handler has ended. The IRQ handler may get called
+	 * even after this, but it will return immediately as the MI and ISP
+	 * interrupts have been masked.
+	 */
+	synchronize_irq(rkisp1->irqs[RKISP1_IRQ_ISP]);
+	if (rkisp1->irqs[RKISP1_IRQ_ISP] != rkisp1->irqs[RKISP1_IRQ_MI])
+		synchronize_irq(rkisp1->irqs[RKISP1_IRQ_MI]);
+
+	/* Clear MI and ISP interrupt status */
+	rkisp1_write(rkisp1, RKISP1_CIF_ISP_ICR, ~0);
 	rkisp1_write(rkisp1, RKISP1_CIF_MI_ICR, ~0);
 
 	/* stop ISP */

-- 
2.34.1

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ