From: Tan En De ende.tan@starfivetech.com
[ Upstream commit 2fe2b969d911a09abcd6a47401a3c66c38a310e6 ]
Replaced pm_runtime_put() with pm_runtime_put_sync_suspend() to ensure the runtime suspend is invoked immediately when unregistering a slave. This prevents a race condition where suspend was skipped when unregistering and registering slave in quick succession.
For example, consider the rapid sequence of `delete_device -> new_device -> delete_device -> new_device`. In this sequence, it is observed that the dw_i2c_plat_runtime_suspend() might not be invoked after `delete_device` operation.
This is because after `delete_device` operation, when the pm_runtime_put() is about to trigger suspend, the following `new_device` operation might race and cancel the suspend.
If that happens, during the `new_device` operation, dw_i2c_plat_runtime_resume() is skipped (since there was no suspend), which means `i_dev->init()`, i.e. i2c_dw_init_slave(), is skipped. Since i2c_dw_init_slave() is skipped, i2c_dw_configure_fifo_slave() is skipped too, which leaves `DW_IC_INTR_MASK` unconfigured. If we inspect the interrupt mask register using devmem, it will show as zero.
Example shell script to reproduce the issue: ``` #!/bin/sh
SLAVE_LADDR=0x1010 SLAVE_BUS=13 NEW_DEVICE=/sys/bus/i2c/devices/i2c-$SLAVE_BUS/new_device DELETE_DEVICE=/sys/bus/i2c/devices/i2c-$SLAVE_BUS/delete_device
# Create initial device echo slave-24c02 $SLAVE_LADDR > $NEW_DEVICE sleep 2
# Rapid sequence of # delete_device -> new_device -> delete_device -> new_device echo $SLAVE_LADDR > $DELETE_DEVICE echo slave-24c02 $SLAVE_LADDR > $NEW_DEVICE echo $SLAVE_LADDR > $DELETE_DEVICE echo slave-24c02 $SLAVE_LADDR > $NEW_DEVICE
# Using devmem to inspect IC_INTR_MASK will show as zero ```
Signed-off-by: Tan En De ende.tan@starfivetech.com Acked-by: Jarkko Nikula jarkko.nikula@linux.intel.com Link: https://lore.kernel.org/r/20250412023303.378600-1-ende.tan@starfivetech.com Signed-off-by: Andi Shyti andi.shyti@kernel.org Signed-off-by: Sasha Levin sashal@kernel.org ---
**YES** This commit should be backported to stable kernel trees. Here's my extensive analysis: ## Critical Bug Fix Analysis ### 1. **Nature of the Bug** This commit fixes a **race condition** that causes I2C slave functionality to completely break. The problem occurs during rapid slave device registration/unregistration sequences where: - `pm_runtime_put()` is asynchronous and may not complete suspend before the next `pm_runtime_get_sync()` - If suspend doesn't happen, runtime resume is skipped - Runtime resume is where `i2c_dw_init_slave()` and `i2c_dw_configure_fifo_slave()` restore critical hardware configuration - When these initialization functions are skipped, the interrupt mask register (`DW_IC_INTR_MASK`) remains zero, disabling all interrupts ### 2. **Impact on Users** This is a **high-impact functional bug** that: - Completely breaks I2C slave functionality in affected scenarios - Affects real-world use cases (rapid device creation/deletion as shown in the shell script) - Leaves hardware in a non-functional state requiring manual intervention - Results in silent failures where the slave appears registered but cannot handle transactions ### 3. **Code Change Analysis** The fix is **minimal and surgical**: ```c - pm_runtime_put(dev->dev); + pm_runtime_put_sync_suspend(dev->dev); ``` This single-line change: - Replaces asynchronous `pm_runtime_put()` with synchronous `pm_runtime_put_sync_suspend()` - Ensures runtime suspend happens immediately before unregistration completes - Eliminates the race window where suspend could be skipped - Has **no functional side effects** beyond the intended timing change ### 4. **Risk Assessment** **Very low risk** for several reasons: - **Single line change** with clear, well-understood semantics - **Conservative fix** - makes behavior more predictable, not fundamentally different - **Same end result** - both functions eventually suspend the device, just with different timing - **No new code paths** - uses existing, well-tested runtime PM infrastructure - **Maintains all existing error handling and power management policies** ### 5. **Stable Tree Criteria Compliance** ✅ **Important bugfix** - Fixes complete loss of I2C slave functionality ✅ **Minimal risk** - Single line change with well-understood behavior ✅ **No new features** - Pure bug fix with no feature additions ✅ **No architectural changes** - Uses existing runtime PM APIs ✅ **Confined to subsystem** - Only affects i2c-designware slave mode ✅ **Clear reproduction case** - Includes shell script demonstrating the problem ### 6. **Comparison with Similar Commits** Looking at the provided examples: - **Similar Commit #3 (Status: YES)**: Also fixed a race condition in slave mode with `synchronize_irq()` - this shows precedent for backporting race condition fixes in this driver - **Similar Commits #1,2,4,5 (Status: NO)**: Were cleanup, optimization, or architectural changes rather than critical bug fixes ### 7. **Technical Validation** The commit demonstrates understanding of the underlying issue: - **Root cause identified**: Async suspend being canceled by subsequent operations - **Mechanism explained**: Missing initialization during runtime resume - **Hardware impact documented**: Interrupt mask register left unconfigured - **Reproduction provided**: Clear test case showing the failure ### 8. **Maintainer Confidence** The commit has strong technical validation: - **Acked-by: Jarkko Nikula** - A key i2c-designware maintainer - **Detailed commit message** with technical explanation and reproduction steps - **Clean implementation** using standard kernel APIs ## Conclusion This commit fixes a critical race condition that completely breaks I2C slave functionality with minimal risk and clear precedent for backporting similar fixes. It meets all stable tree criteria for an important, low-risk bugfix that should be available to users running stable kernels.
drivers/i2c/busses/i2c-designware-slave.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/i2c/busses/i2c-designware-slave.c b/drivers/i2c/busses/i2c-designware-slave.c index f5f001738df5e..57e67962a602f 100644 --- a/drivers/i2c/busses/i2c-designware-slave.c +++ b/drivers/i2c/busses/i2c-designware-slave.c @@ -96,7 +96,7 @@ static int i2c_dw_unreg_slave(struct i2c_client *slave) dev->disable(dev); synchronize_irq(dev->irq); dev->slave = NULL; - pm_runtime_put(dev->dev); + pm_runtime_put_sync_suspend(dev->dev);
return 0; }