zephyr/scripts/dts/gen_defines.py

828 lines
26 KiB
Python
Executable File

#!/usr/bin/env python3
# Copyright (c) 2019 Nordic Semiconductor ASA
# Copyright (c) 2019 Linaro Limited
# SPDX-License-Identifier: BSD-3-Clause
# This script uses edtlib to generate a header file and a .conf file (both
# containing the same values) from a devicetree (.dts) file. Information from
# binding files in YAML format is used as well.
#
# Bindings are files that describe devicetree nodes. Devicetree nodes are
# usually mapped to bindings via their 'compatible = "..."' property.
#
# See the docstring/comments at the top of edtlib.py for more information.
#
# Note: Do not access private (_-prefixed) identifiers from edtlib here (and
# also note that edtlib is not meant to expose the dtlib API directly).
# Instead, think of what API you need, and add it as a public documented API in
# edtlib. This will keep this script simple.
import argparse
import os
import pathlib
import sys
import edtlib
def main():
global conf_file
global header_file
args = parse_args()
try:
edt = edtlib.EDT(args.dts, args.bindings_dirs)
except edtlib.EDTError as e:
sys.exit("devicetree error: " + str(e))
conf_file = open(args.conf_out, "w", encoding="utf-8")
header_file = open(args.header_out, "w", encoding="utf-8")
write_top_comment(edt)
active_compats = set()
for node in edt.nodes:
if node.enabled and node.matching_compat:
# Skip 'fixed-partitions' devices since they are handled by
# write_flash() and would generate extra spurious #defines
if node.matching_compat == "fixed-partitions":
continue
write_node_comment(node)
write_regs(node)
write_irqs(node)
write_props(node)
write_clocks(node)
write_spi_dev(node)
write_bus(node)
write_existence_flags(node)
active_compats.update(node.compats)
out_comment("Active compatibles (mentioned in DTS + binding found)")
for compat in sorted(active_compats):
#define DT_COMPAT_<COMPAT> 1
out("COMPAT_{}".format(str2ident(compat)), 1)
# Derived from /chosen
write_addr_size(edt, "zephyr,sram", "SRAM")
write_addr_size(edt, "zephyr,ccm", "CCM")
write_addr_size(edt, "zephyr,dtcm", "DTCM")
write_addr_size(edt, "zephyr,ipc_shm", "IPC_SHM")
write_flash(edt)
print("Devicetree configuration written to " + args.conf_out)
def parse_args():
# Returns parsed command-line arguments
parser = argparse.ArgumentParser()
parser.add_argument("--dts", required=True, help="DTS file")
parser.add_argument("--bindings-dirs", nargs='+', required=True,
help="directory with bindings in YAML format, "
"we allow multiple")
parser.add_argument("--header-out", required=True,
help="path to write header to")
parser.add_argument("--conf-out", required=True,
help="path to write configuration file to")
return parser.parse_args()
def write_top_comment(edt):
# Writes an overview comment with misc. info at the top of the header and
# configuration file
s = """\
Generated by gen_defines.py
DTS input file:
{}
Directories with bindings:
{}
Nodes in dependency order (ordinal and path):
""".format(edt.dts_path, ", ".join(map(relativize, edt.bindings_dirs)))
for scc in edt.scc_order():
if len(scc) > 1:
err("cycle in devicetree involving "
+ ", ".join(node.path for node in scc))
s += " {0.dep_ordinal:<3} {0.path}\n".format(scc[0])
out_comment(s, blank_before=False)
def write_node_comment(node):
# Writes a comment describing 'node' to the header and configuration file
s = """\
Devicetree node:
{}
Binding (compatible = {}):
{}
Dependency Ordinal: {}
""".format(node.path, node.matching_compat, relativize(node.binding_path),
node.dep_ordinal)
if node.depends_on:
s += "\nRequires:\n"
for dep in node.depends_on:
s += " {:<3} {}\n".format(dep.dep_ordinal, dep.path)
if node.required_by:
s += "\nSupports:\n"
for req in node.required_by:
s += " {:<3} {}\n".format(req.dep_ordinal, req.path)
# Indent description by two spaces
s += "\nDescription:\n" + \
"\n".join(" " + line for line in node.description.splitlines())
out_comment(s)
def relativize(path):
# If 'path' is within $ZEPHYR_BASE, returns it relative to $ZEPHYR_BASE,
# with a "$ZEPHYR_BASE/..." hint at the start of the string. Otherwise,
# returns 'path' unchanged.
zbase = os.getenv("ZEPHYR_BASE")
if zbase is None:
return path
try:
return str("$ZEPHYR_BASE" / pathlib.Path(path).relative_to(zbase))
except ValueError:
# Not within ZEPHYR_BASE
return path
def write_regs(node):
# Writes address/size output for the registers in the node's 'reg' property
def reg_addr_name_alias(reg):
return str2ident(reg.name) + "_BASE_ADDRESS" if reg.name else None
def reg_size_name_alias(reg):
return str2ident(reg.name) + "_SIZE" if reg.name else None
for reg in node.regs:
out_dev(node, reg_addr_ident(reg), hex(reg.addr),
name_alias=reg_addr_name_alias(reg))
if reg.size:
out_dev(node, reg_size_ident(reg), reg.size,
name_alias=reg_size_name_alias(reg))
def write_props(node):
# Writes any properties defined in the "properties" section of the binding
# for the node
for prop in node.props.values():
# Skip #size-cell and other property starting with #. Also skip mapping
# properties like 'gpio-map'.
if prop.name[0] == "#" or prop.name.endswith("-map"):
continue
# See write_clocks()
if prop.name == "clocks":
continue
# edtlib provides these as well (Property.val becomes an edtlib.Node
# and a list of edtlib.Nodes, respectively). Nothing is generated for
# them currently though.
if prop.type in {"phandle", "phandles"}:
continue
# Skip properties that we handle elsewhere
if prop.name in {
"reg", "compatible", "status", "interrupts",
"interrupt-controller", "gpio-controller"
}:
continue
if prop.description is not None:
out_comment(prop.description, blank_before=False)
ident = str2ident(prop.name)
if prop.type == "boolean":
out_dev(node, ident, 1 if prop.val else 0)
elif prop.type == "string":
out_dev_s(node, ident, prop.val)
elif prop.type == "int":
out_dev(node, ident, prop.val)
elif prop.type == "array":
for i, val in enumerate(prop.val):
out_dev(node, "{}_{}".format(ident, i), val)
out_dev(node, ident,
"{" + ", ".join(map(str, prop.val)) + "}")
elif prop.type == "string-array":
for i, val in enumerate(prop.val):
out_dev_s(node, "{}_{}".format(ident, i), val)
elif prop.type == "uint8-array":
out_dev(node, ident,
"{ " + ", ".join("0x{:02x}".format(b) for b in prop.val) + " }")
elif prop.type == "phandle-array":
write_phandle_val_list(prop)
# Generate DT_..._ENUM if there's an 'enum:' key in the binding
if prop.enum_index is not None:
out_dev(node, ident + "_ENUM", prop.enum_index)
def write_bus(node):
# Generate bus-related #defines
if not node.bus:
return
if node.parent.label is None:
err("missing 'label' property on {!r}".format(node.parent))
# #define DT_<DEV-IDENT>_BUS_NAME <BUS-LABEL>
out_dev_s(node, "BUS_NAME", str2ident(node.parent.label))
for compat in node.compats:
# #define DT_<COMPAT>_BUS_<BUS-TYPE> 1
out("{}_BUS_{}".format(str2ident(compat), str2ident(node.bus)), 1)
def write_existence_flags(node):
# Generate #defines of the form
#
# #define DT_INST_<INSTANCE>_<COMPAT> 1
#
# These are flags for which devices exist.
for compat in node.compats:
out("INST_{}_{}".format(node.instance_no[compat],
str2ident(compat)), 1)
def reg_addr_ident(reg):
# Returns the identifier (e.g., macro name) to be used for the address of
# 'reg' in the output
node = reg.node
# NOTE: to maintain compat wit the old script we special case if there's
# only a single register (we drop the '_0').
if len(node.regs) > 1:
return "BASE_ADDRESS_{}".format(node.regs.index(reg))
else:
return "BASE_ADDRESS"
def reg_size_ident(reg):
# Returns the identifier (e.g., macro name) to be used for the size of
# 'reg' in the output
node = reg.node
# NOTE: to maintain compat wit the old script we special case if there's
# only a single register (we drop the '_0').
if len(node.regs) > 1:
return "SIZE_{}".format(node.regs.index(reg))
else:
return "SIZE"
def dev_ident(node):
# Returns an identifier for the device given by 'node'. Used when building
# e.g. macro names.
# TODO: Handle PWM on STM
# TODO: Better document the rules of how we generate things
ident = ""
if node.bus:
ident += "{}_{:X}_".format(
str2ident(node.parent.matching_compat), node.parent.unit_addr)
ident += "{}_".format(str2ident(node.matching_compat))
if node.unit_addr is not None:
ident += "{:X}".format(node.unit_addr)
elif node.parent.unit_addr is not None:
ident += "{:X}_{}".format(node.parent.unit_addr, str2ident(node.name))
else:
# This is a bit of a hack
ident += "{}".format(str2ident(node.name))
return ident
def dev_aliases(node):
# Returns a list of aliases for the device given by 'node', used e.g. when
# building macro names
return dev_path_aliases(node) + dev_instance_aliases(node)
def dev_path_aliases(node):
# Returns a list of aliases for the device given by 'node', based on the
# aliases registered for it, in the /aliases node. Used when building e.g.
# macro names.
if node.matching_compat is None:
return []
compat_s = str2ident(node.matching_compat)
aliases = []
for alias in node.aliases:
aliases.append("ALIAS_{}".format(str2ident(alias)))
# TODO: See if we can remove or deprecate this form
aliases.append("{}_{}".format(compat_s, str2ident(alias)))
return aliases
def dev_instance_aliases(node):
# Returns a list of aliases for the device given by 'node', based on the
# instance number of the device (based on how many instances of that
# particular device there are).
#
# This is a list since a device can have multiple 'compatible' strings,
# each with their own instance number.
return ["INST_{}_{}".format(node.instance_no[compat], str2ident(compat))
for compat in node.compats]
def write_addr_size(edt, prop_name, prefix):
# Writes <prefix>_BASE_ADDRESS and <prefix>_SIZE for the node pointed at by
# the /chosen property named 'prop_name', if it exists
node = edt.chosen_node(prop_name)
if not node:
return
if not node.regs:
err("missing 'reg' property in node pointed at by /chosen/{} ({!r})"
.format(prop_name, node))
out_comment("/chosen/{} ({})".format(prop_name, node.path))
out("{}_BASE_ADDRESS".format(prefix), hex(node.regs[0].addr))
out("{}_SIZE".format(prefix), node.regs[0].size//1024)
def write_flash(edt):
# Writes flash-related output
write_flash_node(edt)
write_code_partition(edt)
flash_index = 0
for node in edt.nodes:
if node.name.startswith("partition@"):
write_flash_partition(node, flash_index)
flash_index += 1
if flash_index != 0:
out_comment("Number of flash partitions")
out("FLASH_AREA_NUM", flash_index)
def write_flash_node(edt):
# Writes output for the top-level flash node pointed at by
# zephyr,flash in /chosen
node = edt.chosen_node("zephyr,flash")
out_comment("/chosen/zephyr,flash ({})"
.format(node.path if node else "missing"))
if not node:
# No flash node. Write dummy values.
out("FLASH_BASE_ADDRESS", 0)
out("FLASH_SIZE", 0)
return
if len(node.regs) != 1:
err("expected zephyr,flash to have a single register, has {}"
.format(len(node.regs)))
if node.bus == "spi" and len(node.parent.regs) == 2:
reg = node.parent.regs[1] # QSPI flash
else:
reg = node.regs[0]
out("FLASH_BASE_ADDRESS", hex(reg.addr))
if reg.size:
out("FLASH_SIZE", reg.size//1024)
if "erase-block-size" in node.props:
out("FLASH_ERASE_BLOCK_SIZE", node.props["erase-block-size"].val)
if "write-block-size" in node.props:
out("FLASH_WRITE_BLOCK_SIZE", node.props["write-block-size"].val)
def write_code_partition(edt):
# Writes output for the node pointed at by zephyr,code-partition in /chosen
node = edt.chosen_node("zephyr,code-partition")
out_comment("/chosen/zephyr,code-partition ({})"
.format(node.path if node else "missing"))
if not node:
# No code partition. Write dummy values.
out("CODE_PARTITION_OFFSET", 0)
out("CODE_PARTITION_SIZE", 0)
return
if not node.regs:
err("missing 'regs' property on {!r}".format(node))
out("CODE_PARTITION_OFFSET", node.regs[0].addr)
out("CODE_PARTITION_SIZE", node.regs[0].size)
def write_flash_partition(partition_node, index):
out_comment("Flash partition at " + partition_node.path)
if partition_node.label is None:
err("missing 'label' property on {!r}".format(partition_node))
# Generate label-based identifiers
write_flash_partition_prefix(
"FLASH_AREA_" + str2ident(partition_node.label), partition_node, index)
# Generate index-based identifiers
write_flash_partition_prefix(
"FLASH_AREA_{}".format(index), partition_node, index)
def write_flash_partition_prefix(prefix, partition_node, index):
# write_flash_partition() helper. Generates identifiers starting with
# 'prefix'.
out("{}_ID".format(prefix), index)
out("{}_READ_ONLY".format(prefix), 1 if partition_node.read_only else 0)
for i, reg in enumerate(partition_node.regs):
# Also add aliases that point to the first sector (TODO: get rid of the
# aliases?)
out("{}_OFFSET_{}".format(prefix, i), reg.addr,
aliases=["{}_OFFSET".format(prefix)] if i == 0 else [])
out("{}_SIZE_{}".format(prefix, i), reg.size,
aliases=["{}_SIZE".format(prefix)] if i == 0 else [])
controller = partition_node.flash_controller
if controller.label is not None:
out_s("{}_DEV".format(prefix), controller.label)
def write_irqs(node):
# Writes IRQ num and data for the interrupts in the node's 'interrupt'
# property
def irq_name_alias(irq, cell_name):
if not irq.name:
return None
alias = "IRQ_{}".format(str2ident(irq.name))
if cell_name != "irq":
alias += "_" + str2ident(cell_name)
return alias
def map_arm_gic_irq_type(irq, irq_num):
# Maps ARM GIC IRQ (type)+(index) combo to linear IRQ number
if "type" not in irq.data:
err("Expected binding for {!r} to have 'type' in interrupt-cells"
.format(irq.controller))
irq_type = irq.data["type"]
if irq_type == 0: # GIC_SPI
return irq_num + 32
if irq_type == 1: # GIC_PPI
return irq_num + 16
err("Invalid interrupt type specified for {!r}"
.format(irq))
def encode_zephyr_multi_level_irq(irq, irq_num):
# See doc/reference/kernel/other/interrupts.rst for details
# on how this encoding works
irq_ctrl = irq.controller
# Look for interrupt controller parent until we have none
while irq_ctrl.interrupts:
irq_num = (irq_num + 1) << 8
if "irq" not in irq_ctrl.interrupts[0].data:
err("Expected binding for {!r} to have 'irq' in interrupt-cells"
.format(irq_ctrl))
irq_num |= irq_ctrl.interrupts[0].data["irq"]
irq_ctrl = irq_ctrl.interrupts[0].controller
return irq_num
for irq_i, irq in enumerate(node.interrupts):
for cell_name, cell_value in irq.data.items():
ident = "IRQ_{}".format(irq_i)
if cell_name == "irq":
if "arm,gic" in irq.controller.compats:
cell_value = map_arm_gic_irq_type(irq, cell_value)
cell_value = encode_zephyr_multi_level_irq(irq, cell_value)
else:
ident += "_" + str2ident(cell_name)
out_dev(node, ident, cell_value,
name_alias=irq_name_alias(irq, cell_name))
def write_spi_dev(node):
# Writes SPI device GPIO chip select data if there is any
cs_gpio = edtlib.spi_dev_cs_gpio(node)
if cs_gpio is not None:
write_phandle_val_list_entry(node, cs_gpio, None, "CS_GPIOS")
def write_phandle_val_list(prop):
# Writes output for a phandle/value list, e.g.
#
# pwms = <&pwm-ctrl-1 10 20
# &pwm-ctrl-2 30 40>;
#
# prop:
# phandle/value Property instance.
#
# If only one entry appears in 'prop' (the example above has two), the
# generated identifier won't get a '_0' suffix, and the '_COUNT' and
# group initializer are skipped too.
#
# The base identifier is derived from the property name. For example, 'pwms = ...'
# generates output like this:
#
# #define <device prefix>_PWMS_CONTROLLER_0 "PWM_0" (name taken from 'label = ...')
# #define <device prefix>_PWMS_CHANNEL_0 123 (name taken from *-cells in binding)
# #define <device prefix>_PWMS_0 {"PWM_0", 123}
# #define <device prefix>_PWMS_CONTROLLER_1 "PWM_1"
# #define <device prefix>_PWMS_CHANNEL_1 456
# #define <device prefix>_PWMS_1 {"PWM_1", 456}
# #define <device prefix>_PWMS_COUNT 2
# #define <device prefix>_PWMS {<device prefix>_PWMS_0, <device prefix>_PWMS_1}
# ...
# pwms -> PWMS
# foo-gpios -> FOO_GPIOS
ident = str2ident(prop.name)
initializer_vals = []
for i, entry in enumerate(prop.val):
initializer_vals.append(write_phandle_val_list_entry(
prop.node, entry, i if len(prop.val) > 1 else None, ident))
if len(prop.val) > 1:
out_dev(prop.node, ident + "_COUNT", len(initializer_vals))
out_dev(prop.node, ident, "{" + ", ".join(initializer_vals) + "}")
def write_phandle_val_list_entry(node, entry, i, ident):
# write_phandle_val_list() helper. We could get rid of it if it wasn't for
# write_spi_dev(). Adds 'i' as an index to identifiers unless it's None.
#
# 'entry' is an edtlib.ControllerAndData instance.
#
# Returns the identifier for the macro that provides the
# initializer for the entire entry.
initializer_vals = []
if entry.controller.label is not None:
ctrl_ident = ident + "_CONTROLLER" # e.g. PWMS_CONTROLLER
if entry.name:
name_alias = str2ident(entry.name) + "_" + ctrl_ident
else:
name_alias = None
# Ugly backwards compatibility hack. Only add the index if there's
# more than one entry.
if i is not None:
ctrl_ident += "_{}".format(i)
initializer_vals.append(quote_str(entry.controller.label))
out_dev_s(node, ctrl_ident, entry.controller.label, name_alias)
for cell, val in entry.data.items():
cell_ident = ident + "_" + str2ident(cell) # e.g. PWMS_CHANNEL
if entry.name:
# From e.g. 'pwm-names = ...'
name_alias = str2ident(entry.name) + "_" + cell_ident
else:
name_alias = None
# Backwards compatibility (see above)
if i is not None:
cell_ident += "_{}".format(i)
out_dev(node, cell_ident, val, name_alias)
initializer_vals += entry.data.values()
initializer_ident = ident
if entry.name:
name_alias = initializer_ident + "_" + str2ident(entry.name)
else:
name_alias = None
if i is not None:
initializer_ident += "_{}".format(i)
return out_dev(node, initializer_ident,
"{" + ", ".join(map(str, initializer_vals)) + "}",
name_alias)
def write_clocks(node):
# Writes clock information.
#
# Most of this ought to be handled in write_props(), but the identifiers
# that get generated for 'clocks' are inconsistent with the with other
# 'phandle-array' properties.
#
# See https://github.com/zephyrproject-rtos/zephyr/pull/19327#issuecomment-534081845.
if "clocks" not in node.props:
return
for clock_i, clock in enumerate(node.props["clocks"].val):
controller = clock.controller
if controller.label is not None:
out_dev_s(node, "CLOCK_CONTROLLER", controller.label)
for name, val in clock.data.items():
if clock_i == 0:
clk_name_alias = "CLOCK_" + str2ident(name)
else:
clk_name_alias = None
out_dev(node, "CLOCK_{}_{}".format(str2ident(name), clock_i), val,
name_alias=clk_name_alias)
if "fixed-clock" not in controller.compats:
continue
if "clock-frequency" not in controller.props:
err("{!r} is a 'fixed-clock' but lacks a 'clock-frequency' "
"property".format(controller))
out_dev(node, "CLOCKS_CLOCK_FREQUENCY",
controller.props["clock-frequency"].val)
def str2ident(s):
# Converts 's' to a form suitable for (part of) an identifier
return s.replace("-", "_") \
.replace(",", "_") \
.replace("@", "_") \
.replace("/", "_") \
.replace(".", "_") \
.replace("+", "PLUS") \
.upper()
def out_dev(node, ident, val, name_alias=None):
# Writes an
#
# <device prefix>_<ident> = <val>
#
# assignment, along with a set of
#
# <device alias>_<ident>
#
# aliases, for each device alias. If 'name_alias' (a string) is passed,
# then these additional aliases are generated:
#
# <device prefix>_<name alias>
# <device alias>_<name alias> (for each device alias)
#
# 'name_alias' is used for reg-names and the like.
#
# Returns the identifier used for the macro that provides the value
# for 'ident' within 'node', e.g. DT_MFG_MODEL_CTL_GPIOS_PIN.
dev_prefix = dev_ident(node)
aliases = [alias + "_" + ident for alias in dev_aliases(node)]
if name_alias is not None:
aliases.append(dev_prefix + "_" + name_alias)
aliases += [alias + "_" + name_alias for alias in dev_aliases(node)]
return out(dev_prefix + "_" + ident, val, aliases)
def out_dev_s(node, ident, s, name_alias=None):
# Like out_dev(), but emits 's' as a string literal
#
# Returns the generated macro name for 'ident'.
return out_dev(node, ident, quote_str(s), name_alias)
def out_s(ident, val):
# Like out(), but puts quotes around 'val' and escapes any double
# quotes and backslashes within it
#
# Returns the generated macro name for 'ident'.
return out(ident, quote_str(val))
def out(ident, val, aliases=()):
# Writes '#define <ident> <val>' to the header and '<ident>=<val>' to the
# the configuration file.
#
# Also writes any aliases listed in 'aliases' (an iterable). For the
# header, these look like '#define <alias> <ident>'. For the configuration
# file, the value is just repeated as '<alias>=<val>' for each alias.
#
# Returns the generated macro name for 'ident'.
print("#define DT_{:40} {}".format(ident, val), file=header_file)
primary_ident = "DT_{}".format(ident)
# Exclude things that aren't single token values from .conf. At
# the moment the only such items are unquoted string
# representations of initializer lists, which begin with a curly
# brace.
output_to_conf = not (isinstance(val, str) and val.startswith("{"))
if output_to_conf:
print("{}={}".format(primary_ident, val), file=conf_file)
for alias in aliases:
if alias != ident:
print("#define DT_{:40} DT_{}".format(alias, ident),
file=header_file)
if output_to_conf:
# For the configuration file, the value is just repeated for all
# the aliases
print("DT_{}={}".format(alias, val), file=conf_file)
return primary_ident
def out_comment(s, blank_before=True):
# Writes 's' as a comment to the header and configuration file. 's' is
# allowed to have multiple lines. blank_before=True adds a blank line
# before the comment.
if blank_before:
print(file=header_file)
print(file=conf_file)
if "\n" in s:
# Format multi-line comments like
#
# /*
# * first line
# * second line
# *
# * empty line before this line
# */
res = ["/*"]
for line in s.splitlines():
# Avoid an extra space after '*' for empty lines. They turn red in
# Vim if space error checking is on, which is annoying.
res.append(" *" if not line.strip() else " * " + line)
res.append(" */")
print("\n".join(res), file=header_file)
else:
# Format single-line comments like
#
# /* foo bar */
print("/* " + s + " */", file=header_file)
print("\n".join("# " + line for line in s.splitlines()), file=conf_file)
def escape(s):
# Backslash-escapes any double quotes and backslashes in 's'
# \ must be escaped before " to avoid double escaping
return s.replace("\\", "\\\\").replace('"', '\\"')
def quote_str(s):
# Puts quotes around 's' and escapes any double quotes and
# backslashes within it
return '"{}"'.format(escape(s))
def err(s):
raise Exception(s)
if __name__ == "__main__":
main()