Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Chengfeng Ye <dg573847474@gmail.com>
To:     sean.wang@mediatek.com, vkoul@kernel.org, matthias.bgg@gmail.com,
        angelogioacchino.delregno@collabora.com
Cc:     dmaengine@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
        linux-mediatek@lists.infradead.org, linux-kernel@vger.kernel.org,
        Chengfeng Ye <dg573847474@gmail.com>
Subject: [PATCH] dmaengine: mediatek: mtk-hsdma: fix potential deadlock on &vc->lock
Date:   Tue, 27 Jun 2023 16:43:09 +0000
Message-Id: <20230627164309.59922-1-dg573847474@gmail.com>
Precedence: bulk

As &vc->lock is acquired by the irq mtk_hsdma_irq(), other process
context code acquiring the lock should disable irq. The terminate
callback mtk_hsdma_terminate_all() acquires the same lock without
closing irq.

Possible deadlock scenario:
mtk_hsdma_free_active_desc()
    -> spin_lock(&hvc->vc.lock)
        <irq interrupt>
        -> mtk_hsdma_irq()
        -> mtk_hsdma_free_rooms_in_ring()
        -> spin_lock(&hvc->vc.lock)  (deadlock here)

This flaw was found using an experimental static analysis tool we are
developing for irq-related deadlock.

The tentative patch fix the potential deadlock by spin_lock_irqsave().

Signed-off-by: Chengfeng Ye <dg573847474@gmail.com>
---
 drivers/dma/mediatek/mtk-hsdma.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/dma/mediatek/mtk-hsdma.c b/drivers/dma/mediatek/mtk-hsdma.c
index 69cc61c0b262..6b6773575893 100644
--- a/drivers/dma/mediatek/mtk-hsdma.c
+++ b/drivers/dma/mediatek/mtk-hsdma.c
@@ -757,18 +757,19 @@ static void mtk_hsdma_free_active_desc(struct dma_chan *c)
 {
 	struct mtk_hsdma_vchan *hvc = to_hsdma_vchan(c);
 	bool sync_needed = false;
+	unsigned long flags;
 
 	/*
 	 * Once issue_synchronize is being set, which means once the hardware
 	 * consumes all descriptors for the channel in the ring, the
 	 * synchronization must be notified immediately it is completed.
 	 */
-	spin_lock(&hvc->vc.lock);
+	spin_lock_irqsave(&hvc->vc.lock, flags);
 	if (!list_empty(&hvc->desc_hw_processing)) {
 		hvc->issue_synchronize = true;
 		sync_needed = true;
 	}
-	spin_unlock(&hvc->vc.lock);
+	spin_unlock_irqrestore(&hvc->vc.lock, flags);
 
 	if (sync_needed)
 		wait_for_completion(&hvc->issue_completion);
-- 
2.17.1