Skip to contentSkip to conversation

dma-debug: don't call __dma_entry_alloc_check_leak() under free_entries_lock

PATCH New Sasha Levin 1 participant 3 messages Sep 14, 2023

Subsystems (3)
lore
Share & tools
lore

Sasha Levin

author maintainer
#0 lore raw

PATCH 1/3 dma-debug: don't call __dma_entry_alloc_check_leak() under free_entries_lock

From: Sergey Senozhatsky <[email protected]>

[ Upstream commit fb5a4315591dae307a65fc246ca80b5159d296e1 ]

__dma_entry_alloc_check_leak() calls into printk -> serial console output (qcom geni) and grabs port->lock under free_entries_lock spin lock, which is a reverse locking dependency chain as qcom_geni IRQ handler can call into dma-debug code and grab free_entries_lock under port->lock.

Move __dma_entry_alloc_check_leak() call out of free_entries_lock scope so that we don't acquire serial console's port->lock under it.

Trimmed-down lockdep splat:

The existing dependency chain (in reverse order) is:

               -> #2 (free_entries_lock){-.-.}-{2:2}:
        _raw_spin_lock_irqsave+0x60/0x80
        dma_entry_alloc+0x38/0x110
        debug_dma_map_page+0x60/0xf8
        dma_map_page_attrs+0x1e0/0x230
        dma_map_single_attrs.constprop.0+0x6c/0xc8
        geni_se_rx_dma_prep+0x40/0xcc
        qcom_geni_serial_isr+0x310/0x510
        __handle_irq_event_percpu+0x110/0x244
        handle_irq_event_percpu+0x20/0x54
        handle_irq_event+0x50/0x88
        handle_fasteoi_irq+0xa4/0xcc
        handle_irq_desc+0x28/0x40
        generic_handle_domain_irq+0x24/0x30
        gic_handle_irq+0xc4/0x148
        do_interrupt_handler+0xa4/0xb0
        el1_interrupt+0x34/0x64
        el1h_64_irq_handler+0x18/0x24
        el1h_64_irq+0x64/0x68
        arch_local_irq_enable+0x4/0x8
        ____do_softirq+0x18/0x24
        ...

               -> #1 (&port_lock_key){-.-.}-{2:2}:
        _raw_spin_lock_irqsave+0x60/0x80
        qcom_geni_serial_console_write+0x184/0x1dc
        console_flush_all+0x344/0x454
        console_unlock+0x94/0xf0
        vprintk_emit+0x238/0x24c
        vprintk_default+0x3c/0x48
        vprintk+0xb4/0xbc
        _printk+0x68/0x90
        register_console+0x230/0x38c
        uart_add_one_port+0x338/0x494
        qcom_geni_serial_probe+0x390/0x424
        platform_probe+0x70/0xc0
        really_probe+0x148/0x280
        __driver_probe_device+0xfc/0x114
        driver_probe_device+0x44/0x100
        __device_attach_driver+0x64/0xdc
        bus_for_each_drv+0xb0/0xd8
        __device_attach+0xe4/0x140
        device_initial_probe+0x1c/0x28
        bus_probe_device+0x44/0xb0
        device_add+0x538/0x668
        of_device_add+0x44/0x50
        of_platform_device_create_pdata+0x94/0xc8
        of_platform_bus_create+0x270/0x304
        of_platform_populate+0xac/0xc4
        devm_of_platform_populate+0x60/0xac
        geni_se_probe+0x154/0x160
        platform_probe+0x70/0xc0
        ...

               -> #0 (console_owner){-...}-{0:0}:
        __lock_acquire+0xdf8/0x109c
        lock_acquire+0x234/0x284
        console_flush_all+0x330/0x454
        console_unlock+0x94/0xf0
        vprintk_emit+0x238/0x24c
        vprintk_default+0x3c/0x48
        vprintk+0xb4/0xbc
        _printk+0x68/0x90
        dma_entry_alloc+0xb4/0x110
        debug_dma_map_sg+0xdc/0x2f8
        __dma_map_sg_attrs+0xac/0xe4
        dma_map_sgtable+0x30/0x4c
        get_pages+0x1d4/0x1e4 [msm]
        msm_gem_pin_pages_locked+0x38/0xac [msm]
        msm_gem_pin_vma_locked+0x58/0x88 [msm]
        msm_ioctl_gem_submit+0xde4/0x13ac [msm]
        drm_ioctl_kernel+0xe0/0x15c
        drm_ioctl+0x2e8/0x3f4
        vfs_ioctl+0x30/0x50
        ...
 Chain exists of:
   console_owner --> &port_lock_key --> free_entries_lock

  Possible unsafe locking scenario:
        CPU0                    CPU1
        ----                    ----
   lock(free_entries_lock);
                                lock(&port_lock_key);
                                lock(free_entries_lock);

lock(console_owner);

                *** DEADLOCK ***
 Call trace:
  dump_backtrace+0xb4/0xf0
  show_stack+0x20/0x30
  dump_stack_lvl+0x60/0x84
  dump_stack+0x18/0x24
  print_circular_bug+0x1cc/0x234
  check_noncircular+0x78/0xac
  __lock_acquire+0xdf8/0x109c
  lock_acquire+0x234/0x284
  console_flush_all+0x330/0x454
  console_unlock+0x94/0xf0
  vprintk_emit+0x238/0x24c
  vprintk_default+0x3c/0x48
  vprintk+0xb4/0xbc
  _printk+0x68/0x90
  dma_entry_alloc+0xb4/0x110
  debug_dma_map_sg+0xdc/0x2f8
  __dma_map_sg_attrs+0xac/0xe4
  dma_map_sgtable+0x30/0x4c
  get_pages+0x1d4/0x1e4 [msm]
  msm_gem_pin_pages_locked+0x38/0xac [msm]
  msm_gem_pin_vma_locked+0x58/0x88 [msm]
  msm_ioctl_gem_submit+0xde4/0x13ac [msm]
  drm_ioctl_kernel+0xe0/0x15c
  drm_ioctl+0x2e8/0x3f4
  vfs_ioctl+0x30/0x50
  ...
---
 kernel/dma/debug.c | 20 +++++++++++++++-----
 1 file changed, 15 insertions(+), 5 deletions(-)
kernel/dma/debug.c +15 -5
index 2caafd13f8aac..1f9a8cee42241 100644
@@ -605,15 +605,19 @@ static struct dma_debug_entry *__dma_entry_alloc(void)
 	return entry;
 }
 
-static void __dma_entry_alloc_check_leak(void)
+/*
+ * This should be called outside of free_entries_lock scope to avoid potential
+ * deadlocks with serial consoles that use DMA.
+ */
+static void __dma_entry_alloc_check_leak(u32 nr_entries)
 {
-	u32 tmp = nr_total_entries % nr_prealloc_entries;
+	u32 tmp = nr_entries % nr_prealloc_entries;
 
 	/* Shout each time we tick over some multiple of the initial pool */
 	if (tmp < DMA_DEBUG_DYNAMIC_ENTRIES) {
 		pr_info("dma_debug_entry pool grown to %u (%u00%%)\n",
-			nr_total_entries,
-			(nr_total_entries / nr_prealloc_entries));
+			nr_entries,
+			(nr_entries / nr_prealloc_entries));
 	}
 }
 
@@ -624,8 +628,10 @@ static void __dma_entry_alloc_check_leak(void)
  */
 static struct dma_debug_entry *dma_entry_alloc(void)
 {
+	bool alloc_check_leak = false;
 	struct dma_debug_entry *entry;
 	unsigned long flags;
+	u32 nr_entries;
 
 	spin_lock_irqsave(&free_entries_lock, flags);
 	if (num_free_entries == 0) {
@@ -635,13 +641,17 @@ static struct dma_debug_entry *dma_entry_alloc(void)
 			pr_err("debugging out of memory - disabling\n");
 			return NULL;
 		}
-		__dma_entry_alloc_check_leak();
+		alloc_check_leak = true;
+		nr_entries = nr_total_entries;
 	}
 
 	entry = __dma_entry_alloc();
 
 	spin_unlock_irqrestore(&free_entries_lock, flags);
 
+	if (alloc_check_leak)
+		__dma_entry_alloc_check_leak(nr_entries);
+
 #ifdef CONFIG_STACKTRACE
 	entry->stack_len = stack_trace_save(entry->stack_entries,
 					    ARRAY_SIZE(entry->stack_entries),
Signature · 1 line
2.40.1
author maintainer
#1 lore raw

PATCH 2/3 spi: sun6i: reduce DMA RX transfer width to single byte

From: Tobias Schramm <[email protected]>

[ Upstream commit 171f8a49f212e87a8b04087568e1b3d132e36a18 ]

Through empirical testing it has been determined that sometimes RX SPI transfers with DMA enabled return corrupted data. This is down to single or even multiple bytes lost during DMA transfer from SPI peripheral to memory. It seems the RX FIFO within the SPI peripheral can become confused when performing bus read accesses wider than a single byte to it during an active SPI transfer.

This patch reduces the width of individual DMA read accesses to the RX FIFO to a single byte to mitigate that issue.

---
 drivers/spi/spi-sun6i.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/spi/spi-sun6i.c +1 -1
index 23ad052528dbe..2bfe87873edb3 100644
@@ -200,7 +200,7 @@ static int sun6i_spi_prepare_dma(struct sun6i_spi *sspi,
 		struct dma_slave_config rxconf = {
 			.direction = DMA_DEV_TO_MEM,
 			.src_addr = sspi->dma_addr_rx,
-			.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES,
+			.src_addr_width = DMA_SLAVE_BUSWIDTH_1_BYTE,
 			.src_maxburst = 8,
 		};
 
Signature · 1 line
2.40.1
author maintainer
#2 lore raw

PATCH 3/3 spi: sun6i: fix race between DMA RX transfer completion and RX FIFO drain

In reply to Sasha Levin

From: Tobias Schramm <[email protected]>

[ Upstream commit 1f11f4202caf5710204d334fe63392052783876d ]

Previously the transfer complete IRQ immediately drained to RX FIFO to read any data remaining in FIFO to the RX buffer. This behaviour is correct when dealing with SPI in interrupt mode. However in DMA mode the transfer complete interrupt still fires as soon as all bytes to be transferred have been stored in the FIFO. At that point data in the FIFO still needs to be picked up by the DMA engine. Thus the drain procedure and DMA engine end up racing to read from RX FIFO, corrupting any data read. Additionally the RX buffer pointer is never adjusted according to DMA progress in DMA mode, thus calling the RX FIFO drain procedure in DMA mode is a bug.
Fix corruptions in DMA RX mode by draining RX FIFO only in interrupt mode. Also wait for completion of RX DMA when in DMA mode before returning to ensure all data has been copied to the supplied memory buffer.

---
 drivers/spi/spi-sun6i.c | 29 ++++++++++++++++++++++++++++-
 1 file changed, 28 insertions(+), 1 deletion(-)
drivers/spi/spi-sun6i.c +28 -1
index 2bfe87873edb3..d79853ba7792a 100644
@@ -95,6 +95,7 @@ struct sun6i_spi {
 	struct reset_control	*rstc;
 
 	struct completion	done;
+	struct completion	dma_rx_done;
 
 	const u8		*tx_buf;
 	u8			*rx_buf;
@@ -189,6 +190,13 @@ static size_t sun6i_spi_max_transfer_size(struct spi_device *spi)
 	return SUN6I_MAX_XFER_SIZE - 1;
 }
 
+static void sun6i_spi_dma_rx_cb(void *param)
+{
+	struct sun6i_spi *sspi = param;
+
+	complete(&sspi->dma_rx_done);
+}
+
 static int sun6i_spi_prepare_dma(struct sun6i_spi *sspi,
 				 struct spi_transfer *tfr)
 {
@@ -213,6 +221,8 @@ static int sun6i_spi_prepare_dma(struct sun6i_spi *sspi,
 						 DMA_PREP_INTERRUPT);
 		if (!rxdesc)
 			return -EINVAL;
+		rxdesc->callback_param = sspi;
+		rxdesc->callback = sun6i_spi_dma_rx_cb;
 	}
 
 	txdesc = NULL;
@@ -268,6 +278,7 @@ static int sun6i_spi_transfer_one(struct spi_master *master,
 		return -EINVAL;
 
 	reinit_completion(&sspi->done);
+	reinit_completion(&sspi->dma_rx_done);
 	sspi->tx_buf = tfr->tx_buf;
 	sspi->rx_buf = tfr->rx_buf;
 	sspi->len = tfr->len;
@@ -426,6 +437,22 @@ static int sun6i_spi_transfer_one(struct spi_master *master,
 	start = jiffies;
 	timeout = wait_for_completion_timeout(&sspi->done,
 					      msecs_to_jiffies(tx_time));
+
+	if (!use_dma) {
+		sun6i_spi_drain_fifo(sspi);
+	} else {
+		if (timeout && rx_len) {
+			/*
+			 * Even though RX on the peripheral side has finished
+			 * RX DMA might still be in flight
+			 */
+			timeout = wait_for_completion_timeout(&sspi->dma_rx_done,
+							      timeout);
+			if (!timeout)
+				dev_warn(&master->dev, "RX DMA timeout\n");
+		}
+	}
+
 	end = jiffies;
 	if (!timeout) {
 		dev_warn(&master->dev,
@@ -453,7 +480,6 @@ static irqreturn_t sun6i_spi_handler(int irq, void *dev_id)
 	/* Transfer complete */
 	if (status & SUN6I_INT_CTL_TC) {
 		sun6i_spi_write(sspi, SUN6I_INT_STA_REG, SUN6I_INT_CTL_TC);
-		sun6i_spi_drain_fifo(sspi);
 		complete(&sspi->done);
 		return IRQ_HANDLED;
 	}
@@ -611,6 +637,7 @@ static int sun6i_spi_probe(struct platform_device *pdev)
 	}
 
 	init_completion(&sspi->done);
+	init_completion(&sspi->dma_rx_done);
 
 	sspi->rstc = devm_reset_control_get_exclusive(&pdev->dev, NULL);
 	if (IS_ERR(sspi->rstc)) {
Signature · 1 line
2.40.1