Re: [PATCH 4/5] ACPI / property: Allow holes in reference properties
From: Mika Westerberg
Date: Thu Sep 29 2016 - 07:57:31 EST
On Thu, Sep 29, 2016 at 01:53:50PM +0200, Rafael J. Wysocki wrote:
> On Thursday, September 29, 2016 09:48:29 AM Mika Westerberg wrote:
> > On Wed, Sep 28, 2016 at 11:36:10PM +0200, Rafael J. Wysocki wrote:
> > > On Fri, Sep 23, 2016 at 4:57 PM, Mika Westerberg
> > > <mika.westerberg@xxxxxxxxxxxxxxx> wrote:
> > > > DT allows holes or empty phandles for references. This is used for example
> > > > in SPI subsystem where some chip selects are native and others are regular
> > > > GPIOs. In ACPI _DSD we currently do not support this but instead the
> > > > preceding reference "consumes" all following integer arguments.
> > > >
> > > > For example we would like to support something like the below ASL fragment
> > > > for SPI:
> > > >
> > > > Package () {
> > > > "cs-gpios",
> > > > Package () {
> > > > ^GPIO, 19, 0, 0, // GPIO CS0
> > > > 0, // Native CS
>
> One thing just occurred to me: what if this value is not 0?
It can be any integer value actually. We allow it but 0 is the preferred
value. I did not find anything in DT saying what it should be either.
Alternatively I can add a check that it must be 0.
> > > > ^GPIO, 20, 0, 0, // GPIO CS1
> > > > }
> > > > }
> > > >
> > > > The zero in the middle means "no entry" or NULL reference. To support this
> > > > we add a new function acpi_node_get_property_reference_fixed_args() that
> > > > takes number of expected arguments as parameter. Function returns -ENOENT
> > > > if the corresponding index resolves to the "no entry" reference.
> > > >
> > > > Signed-off-by: Mika Westerberg <mika.westerberg@xxxxxxxxxxxxxxx>
> > > > ---
> > > > drivers/acpi/property.c | 116 +++++++++++++++++++++++++++++++++++-------------
> > > > include/linux/acpi.h | 3 ++
> > > > 2 files changed, 89 insertions(+), 30 deletions(-)
> > > >
> > > > diff --git a/drivers/acpi/property.c b/drivers/acpi/property.c
> > > > index f2fd3fee588a..b255c4442a05 100644
> > > > --- a/drivers/acpi/property.c
> > > > +++ b/drivers/acpi/property.c
> > > > @@ -472,6 +472,7 @@ static int acpi_data_get_property_array(struct acpi_device_data *data,
> > > > * @data: ACPI device data object containing the property
> > > > * @propname: Name of the property
> > > > * @index: Index of the reference to return
> > > > + * @num_args: Maximum number of arguments after each reference
> > > > * @args: Location to store the returned reference with optional arguments
> > > > *
> > > > * Find property with @name, verifify that it is a package containing at least
> > > > @@ -485,8 +486,8 @@ static int acpi_data_get_property_array(struct acpi_device_data *data,
> > > > * Return: %0 on success, negative error code on failure.
> > > > */
> > > > static int acpi_data_get_property_reference(struct acpi_device_data *data,
> > > > - const char *propname, size_t index,
> > > > - struct acpi_reference_args *args)
> > > > + const char *propname, size_t index, size_t num_args,
> > > > + struct acpi_reference_args *args)
> > > > {
> > > > const union acpi_object *element, *end;
> > > > const union acpi_object *obj;
> > > > @@ -532,41 +533,52 @@ static int acpi_data_get_property_reference(struct acpi_device_data *data,
> > > > while (element < end) {
> > > > u32 nargs, i;
> > > >
> > > > - if (element->type != ACPI_TYPE_LOCAL_REFERENCE)
> > > > - return -EPROTO;
> > > > -
> > > > - ret = acpi_bus_get_device(element->reference.handle, &device);
> > > > - if (ret)
> > > > - return -ENODEV;
> > > > -
> > > > - element++;
> > > > - nargs = 0;
> > > > -
> > > > - /* assume following integer elements are all args */
> > > > - for (i = 0; element + i < end; i++) {
> > > > - int type = element[i].type;
> > > > + if (element->type == ACPI_TYPE_LOCAL_REFERENCE) {
> > > > + ret = acpi_bus_get_device(element->reference.handle,
> > > > + &device);
> > > > + if (ret)
> > > > + return -ENODEV;
> > > > +
> > > > + nargs = 0;
> > > > + element++;
> > > > +
> > > > + /* assume following integer elements are all args */
> > > > + for (i = 0; element + i < end && i < num_args; i++) {
> > > > + int type = element[i].type;
> > > > +
> > > > + if (type == ACPI_TYPE_INTEGER)
> > > > + nargs++;
> > > > + else if (type == ACPI_TYPE_LOCAL_REFERENCE)
> > > > + break;
> > > > + else
> > > > + return -EPROTO;
> > > > + }
> > > >
> > > > - if (type == ACPI_TYPE_INTEGER)
> > > > - nargs++;
> > > > - else if (type == ACPI_TYPE_LOCAL_REFERENCE)
> > > > - break;
> > > > - else
> > > > + if (nargs > MAX_ACPI_REFERENCE_ARGS)
> > > > return -EPROTO;
> > > > - }
> > > >
> > > > - if (idx++ == index) {
> > > > - args->adev = device;
> > > > - args->nargs = nargs;
> > > > - for (i = 0; i < nargs; i++)
> > > > - args->args[i] = element[i].integer.value;
> > > > + if (idx == index) {
> > > > + args->adev = device;
> > > > + args->nargs = nargs;
> > > > + for (i = 0; i < nargs; i++)
> > > > + args->args[i] = element[i].integer.value;
> > > >
> > > > - return 0;
> > > > + return 0;
> > > > + }
> > > > +
> > > > + element += nargs;
> > > > + } else if (element->type == ACPI_TYPE_INTEGER) {
> > > > + if (idx == index)
> > > > + return -ENOENT;
> > > > + element++;
> > > > + } else {
> > > > + return -EPROTO;
> > > > }
> > > >
> > > > - element += nargs;
> > > > + idx++;
> > > > }
> > > >
> > > > - return -EPROTO;
> > > > + return -ENODATA;
> > > > }
> > > >
> > > > /**
> > > > @@ -582,10 +594,54 @@ int acpi_node_get_property_reference(struct fwnode_handle *fwnode,
> > > > {
> > > > struct acpi_device_data *data = acpi_device_data_of_node(fwnode);
> > > >
> > > > - return data ? acpi_data_get_property_reference(data, name, index, args) : -EINVAL;
> > > > + if (!data)
> > > > + return -EINVAL;
> > > > +
> > > > + return acpi_data_get_property_reference(data, name, index,
> > > > + MAX_ACPI_REFERENCE_ARGS, args);
> > > > }
> > > > EXPORT_SYMBOL_GPL(acpi_node_get_property_reference);
> > >
> > > gpiolib-acpi is the only user of this AFAICS, so why don't we change
> > > the definition of acpi_node_get_property_reference() to take one extra
> > > argument and update the caller to pass MAX_ACPI_REFERENCE_ARGS?
> >
> > Unfortunately there are these (which I think should not call that
> > function directly anyway):
>
> Right.
>
> It looks like exactly the same usage pattern that should better be
> implemented as a library function, ie. workarounds for something missing.
>
> > drivers/net/ethernet/apm/xgene/xgene_enet_hw.c: status = acpi_node_get_property_reference(fw_node, "phy-handle", 0,
> > drivers/net/ethernet/hisilicon/hns/hns_dsaf_mac.c: rc = acpi_node_get_property_reference(
> > drivers/net/ethernet/hisilicon/hns/hns_enet.c: ret = acpi_node_get_property_reference(dev->fwnode,
> >
> > So we would need to update those as well. Not a big deal, though.
>
> Or rename it to __acpi_node_get_property_reference() and define
> acpi_node_get_property_reference() as a static inline wrapper passing
> MAX_ACPI_REFERENCE_ARGS to it.
OK, I'll do this.