Re: [PATCH] phy: use per-PHY lockdep keys

[Neil Armstrong <neil.armstrong@linaro.org>]

On 28/05/2025 23:29, Dmitry Baryshkov wrote:
> If the PHY driver uses another PHY internally (e.g. in case of eUSB2,
> repeaters are represented as PHYs), then it would trigger the following
> lockdep splat because all PHYs use a single static lockdep key and thus
> lockdep can not identify whether there is a dependency or not and
> reports a false positive.
> 
> Make PHY subsystem use dynamic lockdep keys, assigning each driver a
> separate key. This way lockdep can correctly identify dependency graph
> between mutexes.

Nice solution !

> 
>   ============================================
>   WARNING: possible recursive locking detected
>   6.15.0-rc7-next-20250522-12896-g3932f283970c #3455 Not tainted
>   --------------------------------------------
>   kworker/u51:0/78 is trying to acquire lock:
>   ffff0008116554f0 (&phy->mutex){+.+.}-{4:4}, at: phy_init+0x4c/0x12c
> 
>   but task is already holding lock:
>   ffff000813c10cf0 (&phy->mutex){+.+.}-{4:4}, at: phy_init+0x4c/0x12c
> 
>   other info that might help us debug this:
>    Possible unsafe locking scenario:
> 
>          CPU0
>          ----
>     lock(&phy->mutex);
>     lock(&phy->mutex);
> 
>    *** DEADLOCK ***
> 
>    May be due to missing lock nesting notation
> 
>   4 locks held by kworker/u51:0/78:
>    #0: ffff000800010948 ((wq_completion)events_unbound){+.+.}-{0:0}, at: process_one_work+0x18c/0x5ec
>    #1: ffff80008036bdb0 (deferred_probe_work){+.+.}-{0:0}, at: process_one_work+0x1b4/0x5ec
>    #2: ffff0008094ac8f8 (&dev->mutex){....}-{4:4}, at: __device_attach+0x38/0x188
>    #3: ffff000813c10cf0 (&phy->mutex){+.+.}-{4:4}, at: phy_init+0x4c/0x12c
> 
>   stack backtrace:
>   CPU: 0 UID: 0 PID: 78 Comm: kworker/u51:0 Not tainted 6.15.0-rc7-next-20250522-12896-g3932f283970c #3455 PREEMPT
>   Hardware name: Qualcomm CRD, BIOS 6.0.240904.BOOT.MXF.2.4-00528.1-HAMOA-1 09/ 4/2024
>   Workqueue: events_unbound deferred_probe_work_func
>   Call trace:
>    show_stack+0x18/0x24 (C)
>    dump_stack_lvl+0x90/0xd0
>    dump_stack+0x18/0x24
>    print_deadlock_bug+0x258/0x348
>    __lock_acquire+0x10fc/0x1f84
>    lock_acquire+0x1c8/0x338
>    __mutex_lock+0xb8/0x59c
>    mutex_lock_nested+0x24/0x30
>    phy_init+0x4c/0x12c
>    snps_eusb2_hsphy_init+0x54/0x1a0
>    phy_init+0xe0/0x12c
>    dwc3_core_init+0x450/0x10b4
>    dwc3_core_probe+0xce4/0x15fc
>    dwc3_probe+0x64/0xb0
>    platform_probe+0x68/0xc4
>    really_probe+0xbc/0x298
>    __driver_probe_device+0x78/0x12c
>    driver_probe_device+0x3c/0x160
>    __device_attach_driver+0xb8/0x138
>    bus_for_each_drv+0x84/0xe0
>    __device_attach+0x9c/0x188
>    device_initial_probe+0x14/0x20
>    bus_probe_device+0xac/0xb0
>    deferred_probe_work_func+0x8c/0xc8
>    process_one_work+0x208/0x5ec
>    worker_thread+0x1c0/0x368
>    kthread+0x14c/0x20c
>    ret_from_fork+0x10/0x20
> 
> Fixes: 3584f6392f09 ("phy: qcom: phy-qcom-snps-eusb2: Add support for eUSB2 repeater")
> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
> ---
> Note: I've used a Fixes tag pointing to the commit which actually
> started using nested PHYs. If you think that it's incorrect, I'm fine
> with dropping it.

In fact, other drivers were using sub-phys before:
./amlogic/phy-meson-axg-pcie.c
./amlogic/phy-meson-axg-mipi-dphy.c

(yeah I know, my bad !)

so it should be:
Fixes: e2463559ff1d ("phy: amlogic: Add Amlogic AXG PCIE PHY Driver")

> 
> Note2: I've tried using mutex_lock_nested, however that didn't play
> well. We can not store nest level in the struct phy (as it can be used
> by different drivers), so using mutex_lock_nested() would require us to
> change and wrap all PHY APIs which take a lock internally. Using dynamic
> lockdep keys looks like a more ellegant solution, especially granted
> that there is no extra impact if lockdep is disabled.
> ---
>   drivers/phy/phy-core.c  | 5 ++++-
>   include/linux/phy/phy.h | 2 ++
>   2 files changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/phy/phy-core.c b/drivers/phy/phy-core.c
> index 8e2daea81666bf8a76d9c936c1a16d6318105c91..04a5a34e7a950ae94fae915673c25d476fc071c1 100644
> --- a/drivers/phy/phy-core.c
> +++ b/drivers/phy/phy-core.c
> @@ -994,7 +994,8 @@ struct phy *phy_create(struct device *dev, struct device_node *node,
>   	}
>   
>   	device_initialize(&phy->dev);
> -	mutex_init(&phy->mutex);
> +	lockdep_register_key(&phy->lockdep_key);
> +	mutex_init_with_key(&phy->mutex, &phy->lockdep_key);
>   
>   	phy->dev.class = &phy_class;
>   	phy->dev.parent = dev;
> @@ -1259,6 +1260,8 @@ static void phy_release(struct device *dev)
>   	dev_vdbg(dev, "releasing '%s'\n", dev_name(dev));
>   	debugfs_remove_recursive(phy->debugfs);
>   	regulator_put(phy->pwr);
> +	mutex_destroy(&phy->mutex);
> +	lockdep_unregister_key(&phy->lockdep_key);
>   	ida_free(&phy_ida, phy->id);
>   	kfree(phy);
>   }
> diff --git a/include/linux/phy/phy.h b/include/linux/phy/phy.h
> index 437769e061b7030105c9ea4e9b0da9d32b6fa158..13add0c2c40721fe9ca3f0350d13c035cd25af45 100644
> --- a/include/linux/phy/phy.h
> +++ b/include/linux/phy/phy.h
> @@ -154,6 +154,7 @@ struct phy_attrs {
>    * @id: id of the phy device
>    * @ops: function pointers for performing phy operations
>    * @mutex: mutex to protect phy_ops
> + * @lockdep_key: lockdep information for this mutex
>    * @init_count: used to protect when the PHY is used by multiple consumers
>    * @power_count: used to protect when the PHY is used by multiple consumers
>    * @attrs: used to specify PHY specific attributes
> @@ -165,6 +166,7 @@ struct phy {
>   	int			id;
>   	const struct phy_ops	*ops;
>   	struct mutex		mutex;
> +	struct lock_class_key	lockdep_key;
>   	int			init_count;
>   	int			power_count;
>   	struct phy_attrs	attrs;
> 
> ---
> base-commit: 460178e842c7a1e48a06df684c66eb5fd630bcf7
> change-id: 20250528-phy-subinit-42c988a12b6c
> 
> Best regards,

With the Fixes changed:

Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>