[PATCH v2 6/6] spi: tegra210-quad: Protect curr_xfer check in IRQ handler

[Breno Leitao <leitao@debian.org>]

Now that all other accesses to curr_xfer are done under the lock,
protect the curr_xfer NULL check in tegra_qspi_isr_thread() with the
spinlock. Without this protection, the following race can occur:

  CPU0 (ISR thread)              CPU1 (timeout path)
  ----------------               -------------------
  if (!tqspi->curr_xfer)
    // sees non-NULL
                                 spin_lock()
                                 tqspi->curr_xfer = NULL
                                 spin_unlock()
  handle_*_xfer()
    spin_lock()
    t = tqspi->curr_xfer  // NULL!
    ... t->len ...        // NULL dereference!

With this patch, all curr_xfer accesses are now properly synchronized.

Although all accesses to curr_xfer are done under the lock, in
tegra_qspi_isr_thread() it checks for NULL, releases the lock and
reacquires it later in handle_cpu_based_xfer()/handle_dma_based_xfer().
There is a potential for an update in between, which could cause a NULL
pointer dereference.

To handle this, add a NULL check inside the handlers after acquiring
the lock. This ensures that if the timeout path has already cleared
curr_xfer, the handler will safely return without dereferencing the
NULL pointer.

Fixes: b4e002d8a7ce ("spi: tegra210-quad: Fix timeout handling")
Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/spi/spi-tegra210-quad.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/drivers/spi/spi-tegra210-quad.c b/drivers/spi/spi-tegra210-quad.c
index 79aeb80aa4a70..f425d62e0c276 100644
--- a/drivers/spi/spi-tegra210-quad.c
+++ b/drivers/spi/spi-tegra210-quad.c
@@ -1457,6 +1457,11 @@ static irqreturn_t handle_cpu_based_xfer(struct tegra_qspi *tqspi)
 	spin_lock_irqsave(&tqspi->lock, flags);
 	t = tqspi->curr_xfer;
 
+	if (!t) {
+		spin_unlock_irqrestore(&tqspi->lock, flags);
+		return IRQ_HANDLED;
+	}
+
 	if (tqspi->tx_status ||  tqspi->rx_status) {
 		tegra_qspi_handle_error(tqspi);
 		complete(&tqspi->xfer_completion);
@@ -1527,6 +1532,11 @@ static irqreturn_t handle_dma_based_xfer(struct tegra_qspi *tqspi)
 	spin_lock_irqsave(&tqspi->lock, flags);
 	t = tqspi->curr_xfer;
 
+	if (!t) {
+		spin_unlock_irqrestore(&tqspi->lock, flags);
+		return IRQ_HANDLED;
+	}
+
 	if (num_errors) {
 		tegra_qspi_dma_unmap_xfer(tqspi, t);
 		tegra_qspi_handle_error(tqspi);
@@ -1565,6 +1575,7 @@ static irqreturn_t handle_dma_based_xfer(struct tegra_qspi *tqspi)
 static irqreturn_t tegra_qspi_isr_thread(int irq, void *context_data)
 {
 	struct tegra_qspi *tqspi = context_data;
+	unsigned long flags;
 	u32 status;
 
 	/*
@@ -1582,7 +1593,9 @@ static irqreturn_t tegra_qspi_isr_thread(int irq, void *context_data)
 	 * If no transfer is in progress, check if this was a real interrupt
 	 * that the timeout handler already processed, or a spurious one.
 	 */
+	spin_lock_irqsave(&tqspi->lock, flags);
 	if (!tqspi->curr_xfer) {
+		spin_unlock_irqrestore(&tqspi->lock, flags);
 		/* Spurious interrupt - transfer not ready */
 		if (!(status & QSPI_RDY))
 			return IRQ_NONE;
@@ -1599,7 +1612,14 @@ static irqreturn_t tegra_qspi_isr_thread(int irq, void *context_data)
 		tqspi->rx_status = tqspi->status_reg & (QSPI_RX_FIFO_OVF | QSPI_RX_FIFO_UNF);
 
 	tegra_qspi_mask_clear_irq(tqspi);
+	spin_unlock_irqrestore(&tqspi->lock, flags);
 
+	/*
+	 * Lock is released here but handlers safely re-check curr_xfer under
+	 * lock before dereferencing.
+	 * DMA handler also needs to sleep in wait_for_completion_*(), which
+	 * cannot be done while holding spinlock.
+	 */
 	if (!tqspi->is_curr_dma_xfer)
 		return handle_cpu_based_xfer(tqspi);
 

-- 
2.47.3