Hi,
On 6/16/21 4:23 PM, Rafael J. Wysocki wrote:
> From: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
>
> In general, acpi_bus_attach() can only be run safely under
> acpi_scan_lock, but that lock cannot be acquired under
> acpi_dep_list_lock, so make acpi_scan_clear_dep() schedule deferred
> execution of acpi_bus_attach() under acpi_scan_lock instead of
> calling it directly.
>
> This also fixes a possible race between acpi_scan_clear_dep() and
> device removal that might cause a device object that went away to
> be accessed, because acpi_scan_clear_dep() is changed to acquire
> a reference on the consumer device object.
>
> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> ---
> drivers/acpi/scan.c | 50 +++++++++++++++++++++++++++++++++++++++++++++-----
> 1 file changed, 45 insertions(+), 5 deletions(-)
>
> Index: linux-pm/drivers/acpi/scan.c
> ===================================================================
> --- linux-pm.orig/drivers/acpi/scan.c
> +++ linux-pm/drivers/acpi/scan.c
> @@ -2115,16 +2115,56 @@ static int acpi_dev_get_first_consumer_d
> return 0;
> }
>
> -static int acpi_scan_clear_dep(struct acpi_dep_data *dep, void *data)
> -{
> +struct acpi_scan_clear_dep_work {
> + struct work_struct work;
> struct acpi_device *adev;
> +};
> +
> +static void acpi_scan_clear_dep_fn(struct work_struct *work)
> +{
> + struct acpi_scan_clear_dep_work *cdw;
> +
> + cdw = container_of(work, struct acpi_scan_clear_dep_work, work);
>
> - acpi_bus_get_device(dep->consumer, &adev);
> + acpi_scan_lock_acquire();
> + acpi_bus_attach(cdw->adev, true);
> + acpi_scan_lock_release();
> +
> + acpi_dev_put(cdw->adev);
> + kfree(cdw);
> +}
> +
> +static bool acpi_scan_clear_dep_queue(struct acpi_device *adev)
> +{
> + struct acpi_scan_clear_dep_work *cdw;
> +
> + if (adev->dep_unmet)
> + return false;
> +
> + cdw = kmalloc(sizeof(*cdw), GFP_KERNEL);
> + if (!cdw)
> + return false;
> +
> + cdw->adev = adev;
> + INIT_WORK(&cdw->work, acpi_scan_clear_dep_fn);
> + /*
> + * Since the work function may block on the lock until the entire
> + * initial enumeration of devices is complete, put it into the unbound
> + * workqueue.
> + */
> + queue_work(system_unbound_wq, &cdw->work);
Hmm, I'm a bit worried about this. Even with the system_unbound_wq
some code may expect at least some progress being made with processing
works during the initial enumeration. OTOH this does run pretty early on.
Still I wonder if it would not be better to create + use our own workqueue
for this ?
I guess we can always do this if we run into issues later...
With that said / otherwise the patch looks good to me:
Reviewed-by: Hans de Goede <hdegoede@xxxxxxxxxx>
Regards,
Hans
> +
> + return true;
> +}
> +
> +static int acpi_scan_clear_dep(struct acpi_dep_data *dep, void *data)
> +{
> + struct acpi_device *adev = acpi_bus_get_acpi_device(dep->consumer);
>
> if (adev) {
> adev->dep_unmet--;
> - if (!adev->dep_unmet)
> - acpi_bus_attach(adev, true);
> + if (!acpi_scan_clear_dep_queue(adev))
> + acpi_dev_put(adev);
> }
>
> list_del(&dep->node);
>
>
>
