All of lore.kernel.org
 help / color / mirror / Atom feed
From: Sasha Levin <sashal@kernel.org>
To: patches@lists.linux.dev, stable@vger.kernel.org
Cc: Fabian Godehardt <fg@emlix.com>, Mark Brown <broonie@kernel.org>,
	Sasha Levin <sashal@kernel.org>,
	linux-spi@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: [PATCH AUTOSEL 6.19-6.12] spi: spidev: fix lock inversion between spi_lock and buf_lock
Date: Mon, 23 Feb 2026 07:37:20 -0500	[thread overview]
Message-ID: <20260223123738.1532940-15-sashal@kernel.org> (raw)
In-Reply-To: <20260223123738.1532940-1-sashal@kernel.org>

From: Fabian Godehardt <fg@emlix.com>

[ Upstream commit 40534d19ed2afb880ecf202dab26a8e7a5808d16 ]

The spidev driver previously used two mutexes, spi_lock and buf_lock,
but acquired them in different orders depending on the code path:

  write()/read(): buf_lock -> spi_lock
  ioctl():       spi_lock -> buf_lock

This AB-BA locking pattern triggers lockdep warnings and can
cause real deadlocks:

  WARNING: possible circular locking dependency detected
  spidev_ioctl() -> mutex_lock(&spidev->buf_lock)
  spidev_sync_write() -> mutex_lock(&spidev->spi_lock)
  *** DEADLOCK ***

The issue is reproducible with a simple userspace program that
performs write() and SPI_IOC_WR_MAX_SPEED_HZ ioctl() calls from
separate threads on the same spidev file descriptor.

Fix this by simplifying the locking model and removing the lock
inversion entirely. spidev_sync() no longer performs any locking,
and all callers serialize access using spi_lock.

buf_lock is removed since its functionality is fully covered by
spi_lock, eliminating the possibility of lock ordering issues.

This removes the lock inversion and prevents deadlocks without
changing userspace ABI or behaviour.

Signed-off-by: Fabian Godehardt <fg@emlix.com>
Link: https://patch.msgid.link/20260211072616.489522-1-fg@emlix.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---

LLM Generated explanations, may be completely bogus:

### 3. Fix Approach

The fix is clean and straightforward:

1. **Removes `buf_lock` entirely** from the `spidev_data` structure
2. **Removes the `spidev_sync()` wrapper** that was taking `spi_lock`
   internally
3. **Changes `spidev_read()`/`spidev_write()`** to use `spi_lock`
   directly (instead of `buf_lock`), adding a NULL check for
   `spidev->spi` before proceeding
4. **Removes `buf_lock` acquisition from `spidev_ioctl()`** and
   `spidev_compat_ioc_message()` since `spi_lock` is already held

This eliminates the lock inversion by using a single lock (`spi_lock`)
for all serialization, which is correct since `spi_lock` already
protected the SPI device pointer and was needed for all SPI operations.

### 4. Stable Kernel Criteria Assessment

| Criterion | Assessment |
|-----------|-----------|
| **Fixes a real bug** | YES - deadlock reproducible from userspace |
| **Obviously correct** | YES - consolidating to single lock eliminates
AB-BA inversion |
| **Small and contained** | YES - single file, single driver, ~30 lines
net removal |
| **No new features** | YES - pure bug fix, no ABI/behavior changes |
| **Tested** | YES - accepted by SPI maintainer Mark Brown |

### 5. Risk Assessment

**Low risk:**
- The change is within a single, self-contained userspace-facing driver
  (`spidev`)
- It simplifies locking (removes one mutex entirely), reducing
  complexity
- `spi_lock` was already held in the ioctl path that also held
  `buf_lock` — so `spi_lock` already provided the serialization
- The old `buf_lock` comment in ioctl stated it was for "triple duty"
  but all three duties are covered by `spi_lock`
- No changes to userspace ABI

**Potential concern:** Slightly coarser locking (one lock instead of
two), but since `spi_lock` was always taken in both paths anyway, the
practical concurrency difference is negligible.

### 6. User Impact

- **spidev is widely used** - it's the primary way userspace programs
  interact with SPI devices (used in embedded Linux, Raspberry Pi, IoT
  devices, etc.)
- The deadlock is **trivially reproducible** from userspace with
  concurrent read/write and ioctl calls on the same fd
- Any multithreaded application using spidev is at risk of hitting this
  deadlock
- Lockdep warnings are triggered, which is disruptive even when the
  deadlock doesn't manifest

### Verification

- **Lock ordering confirmed:** The subagent verified that
  `spidev_read()`/`spidev_write()` took `buf_lock` first then `spi_lock`
  (via `spidev_sync()`), while `spidev_ioctl()` took `spi_lock` first
  then `buf_lock` — classic AB-BA
- **Bug age verified:** `spi_lock` was introduced in commit `25d5cb4b`
  (May 2008); the AB-BA pattern has existed since then, becoming more
  prominent after the spinlock-to-mutex conversion in Jan 2023
- **Commit accepted by SPI maintainer Mark Brown** (Signed-off-by),
  confirming the fix is correct
- **The fix removes `spidev_sync()` entirely** and replaces calls with
  `spidev_sync_unlocked()` — verified in the diff that `spidev_sync()`
  only added `spi_lock` around `spidev_sync_unlocked()`, so callers that
  now hold `spi_lock` themselves can call `spidev_sync_unlocked()`
  directly
- **NULL check for `spidev->spi` added in read/write paths** —
  previously this check was inside `spidev_sync()`, now it's done by the
  callers before calling `spidev_sync_unlocked()`, maintaining the same
  safety against device removal
- **Could not independently verify** the specific lockdep warning output
  in the commit message, but the AB-BA pattern is self-evident from code
  inspection

### Conclusion

This is a textbook stable backport candidate. It fixes a real,
reproducible deadlock in a widely-used driver. The fix is small,
obviously correct (consolidates two locks into one, eliminating the
inversion), contained to a single file, and introduces no new features
or ABI changes. The bug affects any multithreaded userspace application
using spidev, which is extremely common in embedded Linux.

**YES**

 drivers/spi/spidev.c | 63 ++++++++++++++++----------------------------
 1 file changed, 22 insertions(+), 41 deletions(-)

diff --git a/drivers/spi/spidev.c b/drivers/spi/spidev.c
index 9a0160f6dc3dc..f28528ed1c24e 100644
--- a/drivers/spi/spidev.c
+++ b/drivers/spi/spidev.c
@@ -74,7 +74,6 @@ struct spidev_data {
 	struct list_head	device_entry;
 
 	/* TX/RX buffers are NULL unless this device is open (users > 0) */
-	struct mutex		buf_lock;
 	unsigned		users;
 	u8			*tx_buffer;
 	u8			*rx_buffer;
@@ -102,24 +101,6 @@ spidev_sync_unlocked(struct spi_device *spi, struct spi_message *message)
 	return status;
 }
 
-static ssize_t
-spidev_sync(struct spidev_data *spidev, struct spi_message *message)
-{
-	ssize_t status;
-	struct spi_device *spi;
-
-	mutex_lock(&spidev->spi_lock);
-	spi = spidev->spi;
-
-	if (spi == NULL)
-		status = -ESHUTDOWN;
-	else
-		status = spidev_sync_unlocked(spi, message);
-
-	mutex_unlock(&spidev->spi_lock);
-	return status;
-}
-
 static inline ssize_t
 spidev_sync_write(struct spidev_data *spidev, size_t len)
 {
@@ -132,7 +113,8 @@ spidev_sync_write(struct spidev_data *spidev, size_t len)
 
 	spi_message_init(&m);
 	spi_message_add_tail(&t, &m);
-	return spidev_sync(spidev, &m);
+
+	return spidev_sync_unlocked(spidev->spi, &m);
 }
 
 static inline ssize_t
@@ -147,7 +129,8 @@ spidev_sync_read(struct spidev_data *spidev, size_t len)
 
 	spi_message_init(&m);
 	spi_message_add_tail(&t, &m);
-	return spidev_sync(spidev, &m);
+
+	return spidev_sync_unlocked(spidev->spi, &m);
 }
 
 /*-------------------------------------------------------------------------*/
@@ -157,7 +140,7 @@ static ssize_t
 spidev_read(struct file *filp, char __user *buf, size_t count, loff_t *f_pos)
 {
 	struct spidev_data	*spidev;
-	ssize_t			status;
+	ssize_t			status = -ESHUTDOWN;
 
 	/* chipselect only toggles at start or end of operation */
 	if (count > bufsiz)
@@ -165,7 +148,11 @@ spidev_read(struct file *filp, char __user *buf, size_t count, loff_t *f_pos)
 
 	spidev = filp->private_data;
 
-	mutex_lock(&spidev->buf_lock);
+	mutex_lock(&spidev->spi_lock);
+
+	if (spidev->spi == NULL)
+		goto err_spi_removed;
+
 	status = spidev_sync_read(spidev, count);
 	if (status > 0) {
 		unsigned long	missing;
@@ -176,7 +163,9 @@ spidev_read(struct file *filp, char __user *buf, size_t count, loff_t *f_pos)
 		else
 			status = status - missing;
 	}
-	mutex_unlock(&spidev->buf_lock);
+
+err_spi_removed:
+	mutex_unlock(&spidev->spi_lock);
 
 	return status;
 }
@@ -187,7 +176,7 @@ spidev_write(struct file *filp, const char __user *buf,
 		size_t count, loff_t *f_pos)
 {
 	struct spidev_data	*spidev;
-	ssize_t			status;
+	ssize_t			status = -ESHUTDOWN;
 	unsigned long		missing;
 
 	/* chipselect only toggles at start or end of operation */
@@ -196,13 +185,19 @@ spidev_write(struct file *filp, const char __user *buf,
 
 	spidev = filp->private_data;
 
-	mutex_lock(&spidev->buf_lock);
+	mutex_lock(&spidev->spi_lock);
+
+	if (spidev->spi == NULL)
+		goto err_spi_removed;
+
 	missing = copy_from_user(spidev->tx_buffer, buf, count);
 	if (missing == 0)
 		status = spidev_sync_write(spidev, count);
 	else
 		status = -EFAULT;
-	mutex_unlock(&spidev->buf_lock);
+
+err_spi_removed:
+	mutex_unlock(&spidev->spi_lock);
 
 	return status;
 }
@@ -379,14 +374,6 @@ spidev_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 
 	ctlr = spi->controller;
 
-	/* use the buffer lock here for triple duty:
-	 *  - prevent I/O (from us) so calling spi_setup() is safe;
-	 *  - prevent concurrent SPI_IOC_WR_* from morphing
-	 *    data fields while SPI_IOC_RD_* reads them;
-	 *  - SPI_IOC_MESSAGE needs the buffer locked "normally".
-	 */
-	mutex_lock(&spidev->buf_lock);
-
 	switch (cmd) {
 	/* read requests */
 	case SPI_IOC_RD_MODE:
@@ -510,7 +497,6 @@ spidev_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 		break;
 	}
 
-	mutex_unlock(&spidev->buf_lock);
 	spi_dev_put(spi);
 	mutex_unlock(&spidev->spi_lock);
 	return retval;
@@ -541,9 +527,6 @@ spidev_compat_ioc_message(struct file *filp, unsigned int cmd,
 		return -ESHUTDOWN;
 	}
 
-	/* SPI_IOC_MESSAGE needs the buffer locked "normally" */
-	mutex_lock(&spidev->buf_lock);
-
 	/* Check message and copy into scratch area */
 	ioc = spidev_get_ioc_message(cmd, u_ioc, &n_ioc);
 	if (IS_ERR(ioc)) {
@@ -564,7 +547,6 @@ spidev_compat_ioc_message(struct file *filp, unsigned int cmd,
 	kfree(ioc);
 
 done:
-	mutex_unlock(&spidev->buf_lock);
 	spi_dev_put(spi);
 	mutex_unlock(&spidev->spi_lock);
 	return retval;
@@ -802,7 +784,6 @@ static int spidev_probe(struct spi_device *spi)
 	/* Initialize the driver data */
 	spidev->spi = spi;
 	mutex_init(&spidev->spi_lock);
-	mutex_init(&spidev->buf_lock);
 
 	INIT_LIST_HEAD(&spidev->device_entry);
 
-- 
2.51.0


  parent reply	other threads:[~2026-02-23 12:38 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-02-23 12:37 [PATCH AUTOSEL 6.19-6.1] drm/amd/display: Remove conditional for shaper 3DLUT power-on Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.18] ASoC: rt721-sdca: Fix issue of fail to detect OMTP jack type Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.18] ALSA: hda/tas2781: Ignore reset check for SPI device Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-5.15] btrfs: replace BUG() with error handling in __btrfs_balance() Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-5.15] ALSA: usb-audio: Add sanity check for OOB writes at silencing Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.12] drm/amd/display: Fix system resume lag issue Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.12] arm64: hugetlbpage: avoid unused-but-set-parameter warning (gcc-16) Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.12] drm/amd/display: Fix writeback on DCN 3.2+ Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.18] drm/amdgpu: Skip vcn poison irq release on VF Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.18] drm/amdgpu: return when ras table checksum is error Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.18] regulator: core: Remove regulator supply_name length limit Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-5.10] ARM: 9467/1: mm: Don't use %pK through printk Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-5.10] drm/radeon: Add HAINAN clock adjustment Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.18] drm/amdgpu: avoid sdma ring reset in sriov Sasha Levin
2026-02-23 12:37 ` Sasha Levin [this message]
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-5.15] drm/amdgpu: Adjust usleep_range in fence wait Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19] mshv: Ignore second stats page map result failure Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19] btrfs: do not ASSERT() when the fs flips RO inside btrfs_repair_io_failure() Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.18] ALSA: hda/hdmi: Add quirk for TUXEDO IBS14G6 Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19] drm/amd/display: set enable_legacy_fast_update to false for DCN36 Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19] x86/hyperv: Move hv crash init after hypercall pg setup Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.18] mshv: clear eventfd counter on irqfd shutdown Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-5.10] drm/amd/display: Avoid updating surface with the same surface under MPO Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-5.15] ALSA: usb-audio: Update the number of packets properly at receiving Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.12] drm/amd/display: bypass post csc for additional color spaces in dal Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.18] ASoC: amd: amd_sdw: add machine driver quirk for Lenovo models Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-6.18] ALSA: hda/realtek: Fix headset mic on ASUS Zenbook 14 UX3405MA Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19] Drivers: hv: vmbus: Use kthread for vmbus interrupts on PREEMPT_RT Sasha Levin
2026-02-23 12:37 ` [PATCH AUTOSEL 6.19-5.10] drm/amdgpu: Add HAINAN clock adjustment Sasha Levin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260223123738.1532940-15-sashal@kernel.org \
    --to=sashal@kernel.org \
    --cc=broonie@kernel.org \
    --cc=fg@emlix.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-spi@vger.kernel.org \
    --cc=patches@lists.linux.dev \
    --cc=stable@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.