| Message ID | 20230824162654.2890992-1-andriy.shevchenko@linux.intel.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [v1,1/1] amba: bus: balance firmware node reference counting | expand |
On Thu, Aug 24, 2023 at 07:26:54PM +0300, Andy Shevchenko wrote: > Currently the ACPI code doesn't bump the reference count of > the firmware node, while OF counter part does. Not that it's > a problem right now, since ACPI doesn't really use the reference > counting for firmware nodes, it still makes sense to make code > robust against any changes done there. For this, > - switch ACPI case to use device_set_node() to be unified with OF > - move reference counting to amba_device_add() > - switch to use firmware nodes instead of OF ones > > In the result we will have reference counting done in the same module > for all callers independently on the nature of firmware node behind. Any comment on this? I would like to have this applied so I can do something similar to the platform driver code.
On Tue, Sep 19, 2023 at 11:01:37PM +0300, Andy Shevchenko wrote: > On Thu, Aug 24, 2023 at 07:26:54PM +0300, Andy Shevchenko wrote: > > Currently the ACPI code doesn't bump the reference count of > > the firmware node, while OF counter part does. Not that it's > > a problem right now, since ACPI doesn't really use the reference > > counting for firmware nodes, it still makes sense to make code > > robust against any changes done there. For this, > > - switch ACPI case to use device_set_node() to be unified with OF > > - move reference counting to amba_device_add() > > - switch to use firmware nodes instead of OF ones > > > > In the result we will have reference counting done in the same module > > for all callers independently on the nature of firmware node behind. > > Any comment on this? I would like to have this applied so I can do something > similar to the platform driver code. Ah, I see, I missed LKP run on this, I'll send a v2 perhaps later on this week.
On Tue, Sep 19, 2023 at 3:02 PM Andy Shevchenko <andriy.shevchenko@linux.intel.com> wrote: > > On Tue, Sep 19, 2023 at 11:01:37PM +0300, Andy Shevchenko wrote: > > On Thu, Aug 24, 2023 at 07:26:54PM +0300, Andy Shevchenko wrote: > > > Currently the ACPI code doesn't bump the reference count of > > > the firmware node, while OF counter part does. Not that it's > > > a problem right now, since ACPI doesn't really use the reference > > > counting for firmware nodes, it still makes sense to make code > > > robust against any changes done there. For this, > > > - switch ACPI case to use device_set_node() to be unified with OF > > > - move reference counting to amba_device_add() > > > - switch to use firmware nodes instead of OF ones > > > > > > In the result we will have reference counting done in the same module > > > for all callers independently on the nature of firmware node behind. > > > > Any comment on this? I would like to have this applied so I can do something > > similar to the platform driver code. > > Ah, I see, I missed LKP run on this, I'll send a v2 perhaps later on this week. What's the relationship/dependency with this and the other patches dealing with refcounting? Did the AMBA one land? Rob
On Fri, Sep 22, 2023 at 09:18:47AM -0500, Rob Herring wrote: > On Tue, Sep 19, 2023 at 3:02 PM Andy Shevchenko > <andriy.shevchenko@linux.intel.com> wrote: > > On Tue, Sep 19, 2023 at 11:01:37PM +0300, Andy Shevchenko wrote: > > > On Thu, Aug 24, 2023 at 07:26:54PM +0300, Andy Shevchenko wrote: > > > > Currently the ACPI code doesn't bump the reference count of > > > > the firmware node, while OF counter part does. Not that it's > > > > a problem right now, since ACPI doesn't really use the reference > > > > counting for firmware nodes, it still makes sense to make code > > > > robust against any changes done there. For this, > > > > - switch ACPI case to use device_set_node() to be unified with OF > > > > - move reference counting to amba_device_add() > > > > - switch to use firmware nodes instead of OF ones > > > > > > > > In the result we will have reference counting done in the same module > > > > for all callers independently on the nature of firmware node behind. > > > > > > Any comment on this? I would like to have this applied so I can do something > > > similar to the platform driver code. > > > > Ah, I see, I missed LKP run on this, I'll send a v2 perhaps later on this week. > > What's the relationship/dependency with this and the other patches > dealing with refcounting? Did the AMBA one land? The OF refcounting is spread over platform code in unexpected places. So, we have something like of_foo() bumps it, platform ->release() drops it. It sounds to me as weird design to think of and AMBA is part of this mixture as it uses platform_bus, BUT has it's own ->release(), besides the fact that it's used in OF _and_ ACPI environments. What I would expect to have is both are really agnostic to the fwnode type and take and release reference in one place, but with OF <--> platform is kinda more complicated that with AMBA.
diff --git a/drivers/acpi/arm64/amba.c b/drivers/acpi/arm64/amba.c index 60be8ee1dbdc..171b5c2c7edd 100644 --- a/drivers/acpi/arm64/amba.c +++ b/drivers/acpi/arm64/amba.c @@ -101,7 +101,7 @@ static int amba_handler_attach(struct acpi_device *adev, if (parent) dev->dev.parent = acpi_get_first_physical_node(parent); - ACPI_COMPANION_SET(&dev->dev, adev); + device_set_node(&dev->dev, acpi_fwnode_handle(adev)); ret = amba_device_add(dev, &iomem_resource); if (ret) { diff --git a/drivers/amba/bus.c b/drivers/amba/bus.c index 09e72967b8ab..05d0209a6347 100644 --- a/drivers/amba/bus.c +++ b/drivers/amba/bus.c @@ -18,6 +18,7 @@ #include <linux/limits.h> #include <linux/clk/clk-conf.h> #include <linux/platform_device.h> +#include <linux/property.h> #include <linux/reset.h> #include <linux/of_irq.h> #include <linux/of_device.h> @@ -528,7 +529,7 @@ static void amba_device_release(struct device *dev) { struct amba_device *d = to_amba_device(dev); - of_node_put(d->dev.of_node); + fwnode_handle_put(dev_fwnode(&d->dev); if (d->res.parent) release_resource(&d->res); mutex_destroy(&d->periphid_lock); @@ -548,6 +549,8 @@ int amba_device_add(struct amba_device *dev, struct resource *parent) { int ret; + fwnode_handle_get(dev_fwnode(&dev->dev)); + ret = request_resource(parent, &dev->res); if (ret) return ret; diff --git a/drivers/of/platform.c b/drivers/of/platform.c index f235ab55b91e..126d265aa7d8 100644 --- a/drivers/of/platform.c +++ b/drivers/of/platform.c @@ -273,7 +273,7 @@ static struct amba_device *of_amba_device_create(struct device_node *node, dev->dev.dma_mask = &dev->dev.coherent_dma_mask; /* setup generic device info */ - device_set_node(&dev->dev, of_fwnode_handle(of_node_get(node))); + device_set_node(&dev->dev, of_fwnode_handle(node)); dev->dev.parent = parent ? : &platform_bus; dev->dev.platform_data = platform_data; if (bus_id)
Currently the ACPI code doesn't bump the reference count of the firmware node, while OF counter part does. Not that it's a problem right now, since ACPI doesn't really use the reference counting for firmware nodes, it still makes sense to make code robust against any changes done there. For this, - switch ACPI case to use device_set_node() to be unified with OF - move reference counting to amba_device_add() - switch to use firmware nodes instead of OF ones In the result we will have reference counting done in the same module for all callers independently on the nature of firmware node behind. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> --- drivers/acpi/arm64/amba.c | 2 +- drivers/amba/bus.c | 5 ++++- drivers/of/platform.c | 2 +- 3 files changed, 6 insertions(+), 3 deletions(-)