LinuxLists.cc - [PATCH v4 0/4] OF phandle nexus support + GPIO nexus

2017-08-11 15:42:42

Subject: [PATCH v4 0/4] OF phandle nexus support + GPIO nexus

This is a continutation of my phandle remapping/nexus node series
from a while ago. I finally got around to writing the documentation
in the spec for this, but it's really rough around the edges and
could use some review/suggestions to get it into shape. Please
review!

This is one small chunk of work related to DT overlays for expansion
boards. It would be good to have a way to expose #<list>-cells types of
providers through a connector in a standard way. So we introduce a way
to make "nexus" nodes for these types of properties to remap the consumer
number space to the other side of the connector's number space. It's
basically a copy of the interrupt nexus implementation, but without
the address space matching design and interrupt-parent walking.

Changes since v3:
* Added devicetree spec patch
* Rebased to v4.13-rc1

Changes since v2:
* Picked up Linus' ack on the gpio core patch
* Made const stack arrays static (Rob Herring)
* Used of_property_read_u32() instead of open coding (Rob Herring)

Changes since v1:
* Addition of optional pass-thru parameter
* Reduction of strings passed to 1 instead of 3 (or 4 now)
* Updated gpio patch for the new function signature
* Updated test-case for new funciton signature and pass-thru parameter
* Updated kernel-doc to be correct and highlighted difference between this
API and of_parse_phandle_with_args()

Stephen Boyd (3):
Document nexus nodes/specifier remapping
of: Support parsing phandle argument lists through a nexus node
of: unittest: Add phandle remapping test
gpio: Support gpio nexus dt bindings

Kernel:
drivers/gpio/gpiolib-of.c | 4 +-
drivers/of/base.c | 184 ++++++++++++++++++++++++++++
drivers/of/unittest-data/testcases.dts | 11 ++
drivers/of/unittest-data/tests-phandle.dtsi | 25 ++++
drivers/of/unittest.c | 120 ++++++++++++++++++
include/linux/of.h | 12 ++
6 files changed, 354 insertions(+), 2 deletions(-)

Spec:
source/devicetree-basics.rst | 208 +++++++++++++++++++++++++++++++++++++++++++
1 file changed, 208 insertions(+)

--
2.7.4

2017-08-11 15:42:45

by Stephen Boyd

[permalink] [raw]

Subject: [PATCH v4 3/4] of: unittest: Add phandle remapping test

Test the functionality of of_parse_phandle_with_args_map().

Cc: Pantelis Antoniou <[email protected]>
Cc: Linus Walleij <[email protected]>
Cc: Mark Brown <[email protected]>
Signed-off-by: Stephen Boyd <[email protected]>
---
drivers/of/unittest-data/testcases.dts | 11 +++
drivers/of/unittest-data/tests-phandle.dtsi | 25 ++++++
drivers/of/unittest.c | 120 ++++++++++++++++++++++++++++
3 files changed, 156 insertions(+)

diff --git a/drivers/of/unittest-data/testcases.dts b/drivers/of/unittest-data/testcases.dts
index 12f7c3d649c8..173b96f39cd0 100644
--- a/drivers/of/unittest-data/testcases.dts
+++ b/drivers/of/unittest-data/testcases.dts
@@ -26,12 +26,23 @@
/ { __local_fixups__ {
testcase-data {
phandle-tests {
+ provider4 {
+ phandle-map = <0x00000008 0x00000018
+ 0x00000024 0x0000003c
+ 0x00000050 0x00000064>;
+ };
consumer-a {
phandle-list = <0x00000000 0x00000008
0x00000018 0x00000028
0x00000034 0x00000038>;
phandle-list-bad-args = <0x00000000 0x0000000c>;
};
+ consumer-b {
+ phandle-list = <0x00000000 0x00000008
+ 0x00000018 0x00000024
+ 0x00000030 0x00000034>;
+ phandle-list-bad-args = <0x00000000 0x0000000c>;
+ };
};
interrupts {
intmap0 {
diff --git a/drivers/of/unittest-data/tests-phandle.dtsi b/drivers/of/unittest-data/tests-phandle.dtsi
index 5b1527e8a7fb..59e297f497d1 100644
--- a/drivers/of/unittest-data/tests-phandle.dtsi
+++ b/drivers/of/unittest-data/tests-phandle.dtsi
@@ -25,6 +25,18 @@
#phandle-cells = <3>;
};

+ provider4: provider4 {
+ #phandle-cells = <2>;
+ phandle-map = <0 1 &provider1 3>,
+ <4 0 &provider0>,
+ <16 5 &provider3 3 5 0>,
+ <200 8 &provider2 23 6>,
+ <19 0 &provider2 15 0>,
+ <2 3 &provider3 2 5 3>;
+ phandle-map-mask = <0xff 0xf>;
+ phandle-map-pass-thru = <0x0 0xf0>;
+ };
+
consumer-a {
phandle-list = <&provider1 1>,
<&provider2 2 0>,
@@ -43,6 +55,19 @@
unterminated-string = [40 41 42 43];
unterminated-string-list = "first", "second", [40 41 42 43];
};
+
+ consumer-b {
+ phandle-list = <&provider1 1>,
+ <&provider4 2 3>,
+ <0>,
+ <&provider4 4 0x100>,
+ <&provider4 0 0x61>,
+ <&provider0>,
+ <&provider4 19 0x20>;
+ phandle-list-bad-phandle = <12345678 0 0>;
+ phandle-list-bad-args = <&provider2 1 0>,
+ <&provider4 0>;
+ };
};
};
};
diff --git a/drivers/of/unittest.c b/drivers/of/unittest.c
index 0107fc680335..c6091d1e55de 100644
--- a/drivers/of/unittest.c
+++ b/drivers/of/unittest.c
@@ -443,6 +443,125 @@ static void __init of_unittest_parse_phandle_with_args(void)
unittest(rc == -EINVAL, "expected:%i got:%i\n", -EINVAL, rc);
}

+static void __init of_unittest_parse_phandle_with_args_map(void)
+{
+ struct device_node *np, *p0, *p1, *p2, *p3;
+ struct of_phandle_args args;
+ int i, rc;
+
+ np = of_find_node_by_path("/testcase-data/phandle-tests/consumer-b");
+ if (!np) {
+ pr_err("missing testcase data\n");
+ return;
+ }
+
+ p0 = of_find_node_by_path("/testcase-data/phandle-tests/provider0");
+ if (!p0) {
+ pr_err("missing testcase data\n");
+ return;
+ }
+
+ p1 = of_find_node_by_path("/testcase-data/phandle-tests/provider1");
+ if (!p1) {
+ pr_err("missing testcase data\n");
+ return;
+ }
+
+ p2 = of_find_node_by_path("/testcase-data/phandle-tests/provider2");
+ if (!p2) {
+ pr_err("missing testcase data\n");
+ return;
+ }
+
+ p3 = of_find_node_by_path("/testcase-data/phandle-tests/provider3");
+ if (!p3) {
+ pr_err("missing testcase data\n");
+ return;
+ }
+
+ rc = of_count_phandle_with_args(np, "phandle-list", "#phandle-cells");
+ unittest(rc == 7, "of_count_phandle_with_args() returned %i, expected 7\n", rc);
+
+ for (i = 0; i < 8; i++) {
+ bool passed = true;
+
+ rc = of_parse_phandle_with_args_map(np, "phandle-list",
+ "phandle", i, &args);
+
+ /* Test the values from tests-phandle.dtsi */
+ switch (i) {
+ case 0:
+ passed &= !rc;
+ passed &= (args.np == p1);
+ passed &= (args.args_count == 1);
+ passed &= (args.args[0] == 1);
+ break;
+ case 1:
+ passed &= !rc;
+ passed &= (args.np == p3);
+ passed &= (args.args_count == 3);
+ passed &= (args.args[0] == 2);
+ passed &= (args.args[1] == 5);
+ passed &= (args.args[2] == 3);
+ break;
+ case 2:
+ passed &= (rc == -ENOENT);
+ break;
+ case 3:
+ passed &= !rc;
+ passed &= (args.np == p0);
+ passed &= (args.args_count == 0);
+ break;
+ case 4:
+ passed &= !rc;
+ passed &= (args.np == p1);
+ passed &= (args.args_count == 1);
+ passed &= (args.args[0] == 3);
+ break;
+ case 5:
+ passed &= !rc;
+ passed &= (args.np == p0);
+ passed &= (args.args_count == 0);
+ break;
+ case 6:
+ passed &= !rc;
+ passed &= (args.np == p2);
+ passed &= (args.args_count == 2);
+ passed &= (args.args[0] == 15);
+ passed &= (args.args[1] == 0x20);
+ break;
+ case 7:
+ passed &= (rc == -ENOENT);
+ break;
+ default:
+ passed = false;
+ }
+
+ unittest(passed, "index %i - data error on node %s rc=%i\n",
+ i, args.np->full_name, rc);
+ }
+
+ /* Check for missing list property */
+ rc = of_parse_phandle_with_args_map(np, "phandle-list-missing",
+ "phandle", 0, &args);
+ unittest(rc == -ENOENT, "expected:%i got:%i\n", -ENOENT, rc);
+
+ /* Check for missing cells,map,mask property */
+ rc = of_parse_phandle_with_args_map(np, "phandle-list",
+ "phandle-missing", 0, &args);
+ unittest(rc == -EINVAL, "expected:%i got:%i\n", -EINVAL, rc);
+
+ /* Check for bad phandle in list */
+ rc = of_parse_phandle_with_args_map(np, "phandle-list-bad-phandle",
+ "phandle", 0, &args);
+ unittest(rc == -EINVAL, "expected:%i got:%i\n", -EINVAL, rc);
+
+ /* Check for incorrectly formed argument list */
+ rc = of_parse_phandle_with_args_map(np, "phandle-list-bad-args",
+ "phandle", 1, &args);
+ unittest(rc == -EINVAL, "expected:%i got:%i\n", -EINVAL, rc);
+}
+
static void __init of_unittest_property_string(void)
{
const char *strings[4];
@@ -2326,6 +2445,7 @@ static int __init of_unittest(void)
of_unittest_find_node_by_name();
of_unittest_dynamic();
of_unittest_parse_phandle_with_args();
+ of_unittest_parse_phandle_with_args_map();
of_unittest_printf();
of_unittest_property_string();
of_unittest_property_copy();
--
2.7.4

2017-08-11 15:43:05

by Stephen Boyd

[permalink] [raw]

Subject: [PATCH v4 4/4] gpio: Support gpio nexus dt bindings

2017-08-11 15:43:08

by Stephen Boyd

[permalink] [raw]

Subject: [PATCH v4 2/4] of: Support parsing phandle argument lists through a nexus node

Platforms like 96boards have a standardized connector/expansion
slot that exposes signals like GPIOs to expansion boards in an
SoC agnostic way. We'd like the DT overlays for the expansion
boards to be written once without knowledge of the SoC on the
other side of the connector. This avoids the unscalable
combinatorial explosion of a different DT overlay for each
expansion board and SoC pair.

We need a way to describe the GPIOs routed through the connector
in an SoC agnostic way. Let's introduce nexus property parsing
into the OF core to do this. This is largely based on the
interrupt nexus support we already have. This allows us to remap
a phandle list in a consumer node (e.g. reset-gpios) through a
connector in a generic way (e.g. via gpio-map). Do this in a
generic routine so that we can remap any sort of variable length
phandle list.

Taking GPIOs as an example, the connector would be a GPIO nexus,
supporting the remapping of a GPIO specifier space to multiple
GPIO providers on the SoC. DT would look as shown below, where
'soc_gpio1' and 'soc_gpio2' are inside the SoC, 'connector' is an
expansion port where boards can be plugged in, and
'expansion_device' is a device on the expansion board.

soc {
soc_gpio1: gpio-controller1 {
#gpio-cells = <2>;
};

soc_gpio2: gpio-controller2 {
#gpio-cells = <2>;
};
};

connector: connector {
#gpio-cells = <2>;
gpio-map = <0 0 &soc_gpio1 1 0>,
<1 0 &soc_gpio2 4 0>,
<2 0 &soc_gpio1 3 0>,
<3 0 &soc_gpio2 2 0>;
gpio-map-mask = <0xf 0x0>;
gpio-map-pass-thru = <0x0 0x1>
};

expansion_device {
reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>;
};

The GPIO core would use of_parse_phandle_with_args_map() instead
of of_parse_phandle_with_args() and arrive at the same type of
result, a phandle and argument list. The difference is that the
phandle and arguments will be remapped through the nexus node to
the underlying SoC GPIO controller node. In the example above,
we would remap 'reset-gpios' from <&connector 2 GPIO_ACTIVE_LOW>
to <&soc_gpio1 3 GPIO_ACTIVE_LOW>.

Cc: Pantelis Antoniou <[email protected]>
Cc: Linus Walleij <[email protected]>
Cc: Mark Brown <[email protected]>
Signed-off-by: Stephen Boyd <[email protected]>
---
drivers/of/base.c | 184 +++++++++++++++++++++++++++++++++++++++++++++++++++++
include/linux/of.h | 12 ++++
2 files changed, 196 insertions(+)

diff --git a/drivers/of/base.c b/drivers/of/base.c
index 686628d1dfa6..60f15bf56534 100644
--- a/drivers/of/base.c
+++ b/drivers/of/base.c
@@ -1366,6 +1366,190 @@ int of_parse_phandle_with_args(const struct device_node *np, const char *list_na
EXPORT_SYMBOL(of_parse_phandle_with_args);

/**
+ * of_parse_phandle_with_args_map() - Find a node pointed by phandle in a list and remap it
+ * @np: pointer to a device tree node containing a list
+ * @list_name: property name that contains a list
+ * @stem_name: stem of property names that specify phandles' arguments count
+ * @index: index of a phandle to parse out
+ * @out_args: optional pointer to output arguments structure (will be filled)
+ *
+ * This function is useful to parse lists of phandles and their arguments.
+ * Returns 0 on success and fills out_args, on error returns appropriate errno
+ * value. The difference between this function and of_parse_phandle_with_args()
+ * is that this API remaps a phandle if the node the phandle points to has
+ * a <@stem_name>-map property.
+ *
+ * Caller is responsible to call of_node_put() on the returned out_args->np
+ * pointer.
+ *
+ * Example:
+ *
+ * phandle1: node1 {
+ * #list-cells = <2>;
+ * }
+ *
+ * phandle2: node2 {
+ * #list-cells = <1>;
+ * }
+ *
+ * phandle3: node3 {
+ * #list-cells = <1>;
+ * list-map = <0 &phandle2 3>,
+ * <1 &phandle2 2>,
+ * <2 &phandle1 5 1>;
+ * list-map-mask = <0x3>;
+ * };
+ *
+ * node4 {
+ * list = <&phandle1 1 2 &phandle3 0>;
+ * }
+ *
+ * To get a device_node of the `node2' node you may call this:
+ * of_parse_phandle_with_args(node4, "list", "list", 1, &args);
+ */
+int of_parse_phandle_with_args_map(const struct device_node *np,
+ const char *list_name,
+ const char *stem_name,
+ int index, struct of_phandle_args *out_args)
+{
+ char *cells_name, *map_name = NULL, *mask_name = NULL;
+ char *pass_name = NULL;
+ struct device_node *cur, *new = NULL;
+ const __be32 *map, *mask, *pass;
+ static const __be32 dummy_mask[] = { [0 ... MAX_PHANDLE_ARGS] = ~0 };
+ static const __be32 dummy_pass[] = { [0 ... MAX_PHANDLE_ARGS] = 0 };
+ __be32 initial_match_array[MAX_PHANDLE_ARGS];
+ const __be32 *match_array = initial_match_array;
+ int i, ret, map_len, match;
+ u32 list_size, new_size;
+
+ if (index < 0)
+ return -EINVAL;
+
+ cells_name = kasprintf(GFP_KERNEL, "#%s-cells", stem_name);
+ if (!cells_name)
+ return -ENOMEM;
+
+ ret = -ENOMEM;
+ map_name = kasprintf(GFP_KERNEL, "%s-map", stem_name);
+ if (!map_name)
+ goto free;
+
+ mask_name = kasprintf(GFP_KERNEL, "%s-map-mask", stem_name);
+ if (!mask_name)
+ goto free;
+
+ pass_name = kasprintf(GFP_KERNEL, "%s-map-pass-thru", stem_name);
+ if (!pass_name)
+ goto free;
+
+ ret = __of_parse_phandle_with_args(np, list_name, cells_name, 0, index,
+ out_args);
+ if (ret)
+ goto free;
+
+ /* Get the #<list>-cells property */
+ cur = out_args->np;
+ ret = of_property_read_u32(cur, cells_name, &list_size);
+ if (ret < 0)
+ goto put;
+
+ /* Precalculate the match array - this simplifies match loop */
+ for (i = 0; i < list_size; i++)
+ initial_match_array[i] = cpu_to_be32(out_args->args[i]);
+
+ ret = -EINVAL;
+ while (cur) {
+ /* Get the <list>-map property */
+ map = of_get_property(cur, map_name, &map_len);
+ if (!map) {
+ ret = 0;
+ goto free;
+ }
+ map_len /= sizeof(u32);
+
+ /* Get the <list>-map-mask property (optional) */
+ mask = of_get_property(cur, mask_name, NULL);
+ if (!mask)
+ mask = dummy_mask;
+ /* Iterate through <list>-map property */
+ match = 0;
+ while (map_len > (list_size + 1) && !match) {
+ /* Compare specifiers */
+ match = 1;
+ for (i = 0; i < list_size; i++, map_len--)
+ match &= !((match_array[i] ^ *map++) & mask[i]);
+
+ of_node_put(new);
+ new = of_find_node_by_phandle(be32_to_cpup(map));
+ map++;
+ map_len--;
+
+ /* Check if not found */
+ if (!new)
+ goto put;
+
+ if (!of_device_is_available(new))
+ match = 0;
+
+ ret = of_property_read_u32(new, cells_name, &new_size);
+ if (ret)
+ goto put;
+
+ /* Check for malformed properties */
+ if (WARN_ON(new_size > MAX_PHANDLE_ARGS))
+ goto put;
+ if (map_len < new_size)
+ goto put;
+
+ /* Move forward by new node's #<list>-cells amount */
+ map += new_size;
+ map_len -= new_size;
+ }
+ if (!match)
+ goto put;
+
+ /* Get the <list>-map-pass-thru property (optional) */
+ pass = of_get_property(cur, pass_name, NULL);
+ if (!pass)
+ pass = dummy_pass;
+
+ /*
+ * Successfully parsed a <list>-map translation; copy new
+ * specifier into the out_args structure, keeping the
+ * bits specified in <list>-map-pass-thru.
+ */
+ match_array = map - new_size;
+ for (i = 0; i < new_size; i++) {
+ __be32 val = *(map - new_size + i);
+
+ if (i < list_size) {
+ val &= ~pass[i];
+ val |= cpu_to_be32(out_args->args[i]) & pass[i];
+ }
+
+ out_args->args[i] = be32_to_cpu(val);
+ }
+ out_args->args_count = list_size = new_size;
+ /* Iterate again with new provider */
+ out_args->np = new;
+ of_node_put(cur);
+ cur = new;
+ }
+put:
+ of_node_put(cur);
+ of_node_put(new);
+free:
+ kfree(mask_name);
+ kfree(map_name);
+ kfree(cells_name);
+ kfree(pass_name);
+
+ return ret;
+}
+EXPORT_SYMBOL(of_parse_phandle_with_args_map);
+
+/**
* of_parse_phandle_with_fixed_args() - Find a node pointed by phandle in a list
* @np: pointer to a device tree node containing a list
* @list_name: property name that contains a list
diff --git a/include/linux/of.h b/include/linux/of.h
index 23bf46ee8686..92fd032b76c7 100644
--- a/include/linux/of.h
+++ b/include/linux/of.h
@@ -362,6 +362,9 @@ extern struct device_node *of_parse_phandle(const struct device_node *np,
extern int of_parse_phandle_with_args(const struct device_node *np,
const char *list_name, const char *cells_name, int index,
struct of_phandle_args *out_args);
+extern int of_parse_phandle_with_args_map(const struct device_node *np,
+ const char *list_name, const char *stem_name, int index,
+ struct of_phandle_args *out_args);
extern int of_parse_phandle_with_fixed_args(const struct device_node *np,
const char *list_name, int cells_count, int index,
struct of_phandle_args *out_args);
@@ -764,6 +767,15 @@ static inline int of_parse_phandle_with_args(const struct device_node *np,
return -ENOSYS;
}

+static inline int of_parse_phandle_with_args_map(const struct device_node *np,
+ const char *list_name,
+ const char *stem_name,
+ int index,
+ struct of_phandle_args *out_args)
+{
+ return -ENOSYS;
+}
+
static inline int of_parse_phandle_with_fixed_args(const struct device_node *np,
const char *list_name, int cells_count, int index,
struct of_phandle_args *out_args)
--
2.7.4

2017-08-11 15:43:52

by Stephen Boyd

[permalink] [raw]

Subject: [RFC/PATCH v4 1/4] Document nexus nodes/specifier remapping

Document the generic nexus node properties. This can be used by
any specifier that conforms to #<specifier>-cells where they
want to support remapping phandle lists through nexus nodes. This
is similar to interrupt remapping, but slightly different because
we don't consider unit addresses when doing mappings. This is
mostly a copy/paste of the interrupt specification, with the unit
address parts removed and generalized to any specifier. There's
also the addition of a pass through mechanism to make things more
compact if desired in the mapping table.

Signed-off-by: Stephen Boyd <[email protected]>
---

I still need to write the blurb about what this is all about, but
I wanted to send this out now to get early feedback. Some starting
points:

1) Replace child/parent with incoming/outgoing everywhere?

2) Make a pretty picture to describe remapping phandle+specifiers
similar to the interrupt hierarchy diagram?

3) Come up with some better name than <specifier>? Kernel-doc uses <list>
but I'm not sure that's any better.

source/devicetree-basics.rst | 208 +++++++++++++++++++++++++++++++++++++++++++
1 file changed, 208 insertions(+)

diff --git a/source/devicetree-basics.rst b/source/devicetree-basics.rst
index 02696bac15c7..b0571145ee40 100644
--- a/source/devicetree-basics.rst
+++ b/source/devicetree-basics.rst
@@ -1280,3 +1280,211 @@ performed:
* That result is looked up in the *interrupt-map* table, which maps to
the parent interrupt specifier ``<4 1>``.

+.. _sect-nexus:
+
+Nexus Nodes and Specifier Mapping
+---------------------------------
+
+TODO: Write blurb here about nexus nodes and remapping them
+
+Nexus Node Properties
+~~~~~~~~~~~~~~~~~~~~~
+
+A nexus node shall have a *#<specifier>-cells* property, where <specifier> is
+some specifier space like 'gpio', 'clock', 'reset', etc.
+
+<specifier>-map
+^^^^^^^^^^^^^^^
+
+Property: ``<specifier>-map``
+
+Value type: ``<prop-encoded-array>`` encoded as an arbitrary number of
+specifier mapping entries.
+
+Description:
+
+ A *<specifier>-map* is a property in a nexus node that bridges one
+ specifier domain with a set of parent specifier domains and describes
+ how specifiers in the child domain are mapped to their respective parent
+ domains.
+
+ The map is a table where each row is a mapping entry
+ consisting of three components: *child specifier*, *specifier parent*, and
+ *parent specifier*.
+
+ child specifier
+ The specifier of the child node being mapped. The number
+ of 32-bit cells required to specify this component is described by
+ the *#<specifier>-cells* property of this node—the nexus node
+ containing the *<specifier>-map* property.
+
+ specifier parent
+ A single *<phandle>* value that points to the specifier parent to
+ which the child domain is being mapped.
+
+ parent specifier
+ The specifier in the parent domain. The number of 32-bit
+ cells required to specify this component is described by the
+ *#<specifier>-cells* property of the specifier parent node.
+
+ Lookups are performed on the mapping table by matching a specifier against
+ the child specifier in the map. Because some fields in the specifier may
+ not be relevant or need to be modified, a mask is applied before the lookup
+ is done. This mask is defined in the *<specifier>-map-mask* property (see
+ section :ref:`sect-specifier-map-mask`).
+
+ Similarly, when the specifier is mapped some fields in the unit specifier
+ may need to be kept unmodified and passed through from the child node to the
+ parent node. In this case, a *<specifier>-map-pass-thru* property (see
+ section :ref:`sect-specifier-map-pass-thru`) may be specified to apply
+ a mask to the incoming specifier and copy any bits that match to the outgoing
+ unit specifier.
+
+.. _sect-specifier-map-mask:
+
+<specifier>-map-mask
+^^^^^^^^^^^^^^^^^^^^
+
+Property: ``<specifier>-map-mask``
+
+Value type: ``<prop-encoded-array>`` encoded as a bit mask
+
+Description:
+
+ A *<specifier>-map-mask* property may be specified for a nexus node.
+ This property specifies a mask that is applied to the incoming unit
+ specifier being looked up in the table specified in the *<specifier>-map*
+ property. If this property is not specified, the mask is assumed to be
+ a mask with all bits set.
+
+.. _sect-specifier-map-pass-thru:
+
+<specifier>-map-pass-thru
+^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Property: ``<specifier>-map-pass-thru``
+
+Value type: ``<prop-encoded-array>`` encoded as a bit mask
+
+Description:
+
+ A *<specifier>-map-pass-thru* property may be specified for a nexus node.
+ This property specifies a mask that is applied to the incoming unit
+ specifier being looked up in the table specified in the *<specifier>-map*
+ property. Any matching bits in the incoming unit specifier are copied over
+ to the outgoing specifier. If this property is not specified, the mask is
+ assumed to be a mask with no bits set.
+
+#<specifier>-cells
+^^^^^^^^^^^^^^^^^^
+
+Property: ``#<specifier>-cells``
+
+Value type: ``<u32>``
+
+Description:
+
+ The *#<specifier>-cells* property defines the number of cells required to
+ encode a specifier for a domain.
+
+Specifier Mapping Example
+~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The following shows the representation of a fragment of a devicetree with
+two GPIO controllers and a sample specifier map for describing the
+GPIO routing of a few gpios on both of the controllers through a connector
+on a board to a device. The expansion device node is one one side of the
+connector node and the SoC with the two GPIO controllers is on the other
+side of the connector.
+
+.. _example-specifier-mapping:
+
+.. code-block:: dts
+
+ soc {
+ soc_gpio1: gpio-controller1 {
+ #gpio-cells = <2>;
+ };
+
+ soc_gpio2: gpio-controller2 {
+ #gpio-cells = <2>;
+ };
+ };
+
+ connector: connector {
+ #gpio-cells = <2>;
+ gpio-map = <0 0 &soc_gpio1 1 0>,
+ <1 0 &soc_gpio2 4 0>,
+ <2 0 &soc_gpio1 3 0>,
+ <3 0 &soc_gpio2 2 0>;
+ gpio-map-mask = <0xf 0x0>;
+ gpio-map-pass-thru = <0x0 0x1>;
+ };
+
+ expansion_device {
+ reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>;
+ };
+
+
+Each row in the gpio-map table consists of three parts: a child unit
+specifier, which is mapped to a *gpio-controller*
+node with a parent specifier.
+
+* For example, the first row of the specifier-map table specifies the
+ mapping for GPIO 0 of the connector. The components of that row are shown
+ here
+
+ | child specifier: ``0 0``
+ | specifier parent: ``&soc_gpio1``
+ | parent specifier: ``1 0``
+
+ * The child specifier is ``<0 0>``, which specifies GPIO 0 in the connector
+ with a *flags* field of ``0``. This takes two 32-bit cells as specified
+ by the *#gpio-cells* property of the connector node, which is the
+ child specifier domian.
+
+ * The specifier parent is specified by a phandle which points to the
+ specifier parent of the connector, the first GPIO controller in the SoC.
+
+ * The parent specifier is ``<1 0>``. The number of cells to
+ represent the gpio specifier (two cells) is determined by the
+ *#gpio-cells* property on the specifier parent, the soc_gpio1
+ node.
+
+ * The value ``<1 0>`` is a value specified by the device binding for
+ the GPIO controller. The value ``<1>`` specifies the
+ GPIO pin number on the GPIO controller to which GPIO 0 on the connector
+ is wired. The value ``<0>`` specifies the flags (active low,
+ active high, etc.).
+
+In this example, the *gpio-map-mask* property has a value of ``<0xf 0>``.
+This mask is applied to a child unit specifier before performing a lookup in
+the *gpio-map* table. Similarly, the *gpio-map-pass-thru* property has a value
+of ``<0x0 0x1>``. This mask is applied to a child unit specifier when mapping
+it to the parent unit specifier. Any bits set in the mask are cleared out of
+the parent unit specifier and copied over from the child unit specifier
+to the parent unit specifier.
+
+To perform a lookup of the connector's specifier source number for GPIO 2
+from the expansion device's reset-gpios property, the following steps would be
+performed:
+
+* The child specifier forms the value
+ ``<2 GPIO_ACTIVE_LOW>``.
+
+ * The specifier is encoding GPIO 2 with active low flags per the GPIO
+ binding.
+
+* The *gpio-map-mask* value ``<0xf 0x0>`` is applied, giving a
+ result of ``<0x2 0>``.
+
+* That result is looked up in the *gpio-map* table, which maps to
+ the parent specifier ``<3 0>`` and &soc_gpio1 *phandle*.
+
+* That *gpio-map-pass-thru* value ``<0x0 0x1>`` is applied to child specifier,
+ forming ``<0 GPIO_ACTIVE_LOW>`` which is then ORed with the parent
+ specifier ``<3 0>`` ANDed with the inverse of the pass-thru mask
+ ``<0xffffffff 0xffffffe>`` resulting in ``<3 GPIO_ACTIVE_LOW>``.
+
+* That specifier is combined with the mapped *phandle* &soc_gpio1 resulting
+ in ``<&soc_gpio1 3 GPIO_ACTIVE_LOW>``.
--
2.7.4

2017-08-15 20:37:22

by Rob Herring

[permalink] [raw]

Subject: Re: [PATCH v4 0/4] OF phandle nexus support + GPIO nexus

On Fri, Aug 11, 2017 at 10:42 AM, Stephen Boyd <[email protected]> wrote:
> This is a continutation of my phandle remapping/nexus node series
> from a while ago. I finally got around to writing the documentation
> in the spec for this, but it's really rough around the edges and
> could use some review/suggestions to get it into shape. Please
> review!
>
> This is one small chunk of work related to DT overlays for expansion
> boards. It would be good to have a way to expose #<list>-cells types of
> providers through a connector in a standard way. So we introduce a way
> to make "nexus" nodes for these types of properties to remap the consumer
> number space to the other side of the connector's number space. It's
> basically a copy of the interrupt nexus implementation, but without
> the address space matching design and interrupt-parent walking.

This generally looks good to me. I'm inclined to merge the kernel
parts now and sort the spec addition out afterwards. With no users
yet, we can still tweak things if needed.

I would like to see more comments on this though. Perhaps "OF phandle
nexus" is too obscure. Maybe "Add DT overlay connector bindings" would
trick some people in looking at it...

Rob

2017-09-18 17:45:02

by Rob Herring

[permalink] [raw]

Subject: Re: [RFC/PATCH v4 1/4] Document nexus nodes/specifier remapping

On Fri, Aug 11, 2017 at 10:42 AM, Stephen Boyd <[email protected]> wrote:
> Document the generic nexus node properties. This can be used by
> any specifier that conforms to #<specifier>-cells where they
> want to support remapping phandle lists through nexus nodes. This
> is similar to interrupt remapping, but slightly different because
> we don't consider unit addresses when doing mappings. This is
> mostly a copy/paste of the interrupt specification, with the unit
> address parts removed and generalized to any specifier. There's
> also the addition of a pass through mechanism to make things more
> compact if desired in the mapping table.

Sorry for the slow response.

I'm still wondering how/if we can merge interrupts as part of this
(both the spec and parsing implementation). Could we simply require
that #address-cells is 0 or do we even need this distinction? If the
usecase is connectors, then we should typically be able to set
#address-cells to 0. Perhaps you could have a custom PCI connector
with additional signals and the slot/connector node would have the PCI
address. In this case, we would have #address-cells, but having them
and ignoring the address cells via the mask would still work.

Also, I don't see any issue if we allow the -map-pass-thru property
for interrupts.

>
> Signed-off-by: Stephen Boyd <[email protected]>
> ---
>
> I still need to write the blurb about what this is all about, but
> I wanted to send this out now to get early feedback. Some starting
> points:
>
> 1) Replace child/parent with incoming/outgoing everywhere?

Parent/child is more common, so I think that's fine.

>
> 2) Make a pretty picture to describe remapping phandle+specifiers
> similar to the interrupt hierarchy diagram?

Pictures are always nice, but I don't think required.

>
> 3) Come up with some better name than <specifier>? Kernel-doc uses <list>
> but I'm not sure that's any better.

specifier seems fine to me.

Rob