[PATCH v2] bus: mhi: host: Fix race between channel preparation and M0 event

From: Qiang Yu <quic_qianyu@quicinc.com>
To: mani@kernel.org, loic.poulain@linaro.org
Cc: mhi@lists.linux.dev, linux-arm-msm@vger.kernel.org,
	linux-kernel@vger.kernel.org, quic_cang@quicinc.com,
	mrana@quicinc.com, Qiang Yu <quic_qianyu@quicinc.com>
Subject: [PATCH v2] bus: mhi: host: Fix race between channel preparation and M0 event
Date: Sun, 16 Oct 2022 11:05:32 +0800	[thread overview]
Message-ID: <1665889532-13634-1-git-send-email-quic_qianyu@quicinc.com> (raw)

There is a race condition where mhi_prepare_channel() updates the
read and write pointers as the base address and in parallel, if
an M0 transition occurs, the tasklet goes ahead and rings
doorbells for all channels with a delta in TRE rings assuming
they are already enabled. This causes a null pointer access. Fix
it by adding a channel enabled check before ringing channel
doorbells.

Fixes: a6e2e3522f29 "bus: mhi: core: Add support for PM state transitions"
Signed-off-by: Qiang Yu <quic_qianyu@quicinc.com>
---
v1->v2: add Fixes tags

 drivers/bus/mhi/host/pm.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/bus/mhi/host/pm.c b/drivers/bus/mhi/host/pm.c
index 4a42186..0834590 100644
--- a/drivers/bus/mhi/host/pm.c
+++ b/drivers/bus/mhi/host/pm.c
@@ -301,7 +301,8 @@ int mhi_pm_m0_transition(struct mhi_controller *mhi_cntrl)
 		read_lock_irq(&mhi_chan->lock);
 
 		/* Only ring DB if ring is not empty */
-		if (tre_ring->base && tre_ring->wp  != tre_ring->rp)
+		if (tre_ring->base && tre_ring->wp  != tre_ring->rp &&
+		    mhi_chan->ch_state == MHI_CH_STATE_ENABLED)
 			mhi_ring_chan_db(mhi_cntrl, mhi_chan);
 		read_unlock_irq(&mhi_chan->lock);
 	}
-- 
2.7.4

