Hi Walter,

On Fri, 29 May 2020 at 12:15, Walter Lozano walter.lozano@collabora.com wrote:

Currently dtoc scans dtbs to convert them to struct platdata and to generate U_BOOT_DEVICE entries. These entries need to be filled with the driver name, but at this moment the information used is the compatible name present in the dtb. This causes that only nodes with a compatible name that matches a driver name generate a working entry.

In order to improve this behaviour, this patch adds to dtoc the capability of scan drivers source code to generate a list of valid driver names. This allows to rise a warning in the case that an U_BOOT_DEVICE entry will try to use a name not valid.

Additionally, in order to add more flexibility to the solution, adds the U_BOOT_DRIVER_ALIAS macro, which generates no code at all, but allows an easy way to declare driver name aliases. Thanks to this, dtoc can look for the driver name based on its alias when it populates the U_BOOT_DEVICE entry.

Signed-off-by: Walter Lozano walter.lozano@collabora.com

include/dm/device.h | 7 ++++ tools/dtoc/dtb_platdata.py | 83 ++++++++++++++++++++++++++++++++++++-- 2 files changed, 86 insertions(+), 4 deletions(-)

diff --git a/include/dm/device.h b/include/dm/device.h index 975eec5d0e..2cfe10766f 100644 --- a/include/dm/device.h +++ b/include/dm/device.h @@ -282,6 +282,13 @@ struct driver { #define DM_GET_DRIVER(__name) \ ll_entry_get(struct driver, __name, driver)

+/**

• • Declare a macro to state a alias for a driver name. This macro will
• • produce no code but its information will be parsed by tools like
• • dtoc
• */

+#define U_BOOT_DRIVER_ALIAS(__name, __alias)

/**

• dev_get_platdata() - Get the platform data for a device

diff --git a/tools/dtoc/dtb_platdata.py b/tools/dtoc/dtb_platdata.py index ecfe0624d1..23cfda2f88 100644 --- a/tools/dtoc/dtb_platdata.py +++ b/tools/dtoc/dtb_platdata.py @@ -13,6 +13,8 @@ static data.

import collections import copy +import os +import re import sys

from dtoc import fdt @@ -140,6 +142,9 @@ class DtbPlatdata(object): _include_disabled: true to include nodes marked status = "disabled" _outfile: The current output file (sys.stdout or a real file) _lines: Stashed list of output lines for outputting in the future

•    _aliases: Dict that hold aliases for compatible strings
  

key: The driver name, i.e. the part between brackets in U_BOOT_DRIVER(xx) ?? value: ...

•    _drivers: List of valid driver names found in drivers/
  

•    _driver_aliases: Dict that holds aliases for driver names
  

key: vaue:

 """
 def __init__(self, dtb_fname, include_disabled):
     self._fdt = None

@@ -149,6 +154,35 @@ class DtbPlatdata(object): self._outfile = None self._lines = [] self._aliases = {}

•    self._drivers = []
  

•    self._driver_aliases = {}
  

• def get_normalized_compat_name(self, node):

•    """Get a node's normalized compat name
  

•    Returns a valid driver name by retrieving node's first compatible
  

•    string as a C identifier and perfomrming a check against _drivers
  

performing

•    and a lookup in driver_aliases rising a warning in case of failure.
  

s/ rising/, printing/

•    Args:
  

•        node: Node object to check
  

•    Return:
  

•        Tuple:
  

•            Driver name associated with the first compatible string
  

•            List of C identifiers for all the other compatible strings
  

•                (possibly empty)
  

Can you update this comment to explain what is returned when it is not found?

•    """
  

•    compat_c, aliases_c = get_compat_name(node)
  

•    if compat_c not in self._drivers:
  

•        compat_c_old = compat_c
  

•        compat_c = self._driver_aliases.get(compat_c)
  

•        if not compat_c:
  

•            print('WARNING: the driver %s was not found in the driver list' % (compat_c_old))
  

This creates lots of warnings at present. Either we need a patch to clean up the differences in the source code, or we need to disable the warning.

Also the warning is not really actionable. It needs to add mention of U_BOOT_DEVICE and ...ALIAS.

In future we can scan the compatible strings and tell the user what changes to make, I suppose.

•            compat_c = compat_c_old
  

•        else: # pragma: no cover
  

Need to fix the coverage here

•            aliases_c = [compat_c_old] + aliases_c
  

•    return compat_c, aliases_c
  
  

  def setup_output(self, fname): """Set up the output destination

@@ -243,6 +277,46 @@ class DtbPlatdata(object): return PhandleInfo(max_args, args) return None

• def scan_driver(self, fn):

•    """Scan a driver file to build a list of driver names and aliases
  

•    This procedure will populate self._drivers and self._driver_aliases
  

•    Args
  

•        fn: Driver filename to scan
  

•    """
  

•    with open(fn) as fd:
  

drop blank line

•        buff = fd.read()
  

•        # The following re will search for driver names declared as
  

•        # U_BOOT_DRIVER(driver_name)
  

•        drivers = re.findall('U_BOOT_DRIVER\((.*)\)', buff)
  

•        for driver in drivers:
  

•            self._drivers.append(driver)
  

•        # The following re will search for driver aliases declared as
  

•        # U_BOOT_DRIVER_ALIAS(alias, driver_name)
  

•        driver_aliases = re.findall('U_BOOT_DRIVER_ALIAS\(\s*(\w+)\s*,\s*(\w+)\s*\)', buff)
  

•        for alias in driver_aliases: # pragma: no cover
  

•            if len(alias) != 2:
  

•                continue
  

•            self._driver_aliases[alias[1]] = alias[0]
  

• def scan_drivers(self):

•    """Scan the driver folders to build a list of driver names and aliases
  

•    This procedure will populate self._drivers and self._driver_aliases
  

•    """
  

•    for (dirpath, dirnames, filenames) in os.walk('./'):
  

This doesn't work for out-of-tree cases (make O=xxx). You may need to pass $(srctree) to dtc as an argument.

•        for fn in filenames:
  

•            if not fn.endswith('.c'):
  

•                continue
  

•            self.scan_driver(dirpath + '/' + fn)
  

• def scan_dtb(self): """Scan the device tree to obtain a tree of nodes and properties

@@ -353,7 +427,7 @@ class DtbPlatdata(object): """ structs = {} for node in self._valid_nodes:

•        node_name, _ = get_compat_name(node)
  

•        node_name, _ = self.get_normalized_compat_name(node)
         fields = {}
  
         # Get a list of all the valid properties in this node.
  

@@ -377,14 +451,14 @@ class DtbPlatdata(object):

     upto = 0
     for node in self._valid_nodes:

•        node_name, _ = get_compat_name(node)
  

•        node_name, _ = self.get_normalized_compat_name(node)
         struct = structs[node_name]
         for name, prop in node.props.items():
             if name not in PROP_IGNORE_LIST and name[0] != '#':
                 prop.Widen(struct[name])
         upto += 1
  
  

•        struct_name, aliases = get_compat_name(node)
  

•        struct_name, aliases = self.get_normalized_compat_name(node)
         for alias in aliases:
             self._aliases[alias] = struct_name
  
  

@@ -461,7 +535,7 @@ class DtbPlatdata(object): Args: node: node to output """

•    struct_name, _ = get_compat_name(node)
  

•    struct_name, _ = self.get_normalized_compat_name(node)
     var_name = conv_name_to_c(node.name)
     self.buf('static const struct %s%s %s%s = {\n' %
              (STRUCT_PREFIX, struct_name, VAL_PREFIX, var_name))
  

@@ -562,6 +636,7 @@ def run_steps(args, dtb_file, include_disabled, output): raise ValueError('Please specify a command: struct, platdata')

 plat = DtbPlatdata(dtb_file, include_disabled)

• plat.scan_drivers() plat.scan_dtb() plat.scan_tree() plat.scan_reg_sizes()

-- 2.20.1

Regards, Simon

Hi Simon,

On 8/6/20 12:49, Walter Lozano wrote:

Hi Simon,

On 4/6/20 12:59, Simon Glass wrote:

...

Hi Walter,

On Fri, 29 May 2020 at 12:15, Walter Lozano walter.lozano@collabora.com wrote:

...

Currently dtoc scans dtbs to convert them to struct platdata and to generate U_BOOT_DEVICE entries. These entries need to be filled with the driver name, but at this moment the information used is the compatible name present in the dtb. This causes that only nodes with a compatible name that matches a driver name generate a working entry.

In order to improve this behaviour, this patch adds to dtoc the capability of scan drivers source code to generate a list of valid driver names. This allows to rise a warning in the case that an U_BOOT_DEVICE entry will try to use a name not valid.

Additionally, in order to add more flexibility to the solution, adds the U_BOOT_DRIVER_ALIAS macro, which generates no code at all, but allows an easy way to declare driver name aliases. Thanks to this, dtoc can look for the driver name based on its alias when it populates the U_BOOT_DEVICE entry.

Signed-off-by: Walter Lozano walter.lozano@collabora.com

include/dm/device.h        |  7 ++++   tools/dtoc/dtb_platdata.py | 83 ++++++++++++++++++++++++++++++++++++--   2 files changed, 86 insertions(+), 4 deletions(-)

diff --git a/include/dm/device.h b/include/dm/device.h index 975eec5d0e..2cfe10766f 100644 --- a/include/dm/device.h +++ b/include/dm/device.h @@ -282,6 +282,13 @@ struct driver {   #define DM_GET_DRIVER(__name) \          ll_entry_get(struct driver, __name, driver)

+/**

• • Declare a macro to state a alias for a driver name. This macro will
• • produce no code but its information will be parsed by tools like
• • dtoc
• */

+#define U_BOOT_DRIVER_ALIAS(__name, __alias)

/**    * dev_get_platdata() - Get the platform data for a device    * diff --git a/tools/dtoc/dtb_platdata.py b/tools/dtoc/dtb_platdata.py index ecfe0624d1..23cfda2f88 100644 --- a/tools/dtoc/dtb_platdata.py +++ b/tools/dtoc/dtb_platdata.py @@ -13,6 +13,8 @@ static data.

import collections   import copy +import os +import re   import sys

from dtoc import fdt @@ -140,6 +142,9 @@ class DtbPlatdata(object):           _include_disabled: true to include nodes marked status = "disabled"           _outfile: The current output file (sys.stdout or a real file)           _lines: Stashed list of output lines for outputting in the future +        _aliases: Dict that hold aliases for compatible strings

key: The driver name, i.e. the part between brackets in U_BOOT_DRIVER(xx)  ?? value: ...

Noted.

...

...

+        _drivers: List of valid driver names found in drivers/ +        _driver_aliases: Dict that holds aliases for driver names

key: vaue:

OK.

...

...

"""       def __init__(self, dtb_fname, include_disabled):           self._fdt = None @@ -149,6 +154,35 @@ class DtbPlatdata(object):           self._outfile = None           self._lines = []           self._aliases = {} +        self._drivers = [] +        self._driver_aliases = {}

+    def get_normalized_compat_name(self, node): +        """Get a node's normalized compat name

+        Returns a valid driver name by retrieving node's first compatible +        string as a C identifier and perfomrming a check against _drivers

performing

Noted.

...

...

+        and a lookup in driver_aliases rising a warning in case of failure.

s/ rising/, printing/

OK.

...

...

+        Args: +            node: Node object to check +        Return: +            Tuple: +                Driver name associated with the first compatible string +                List of C identifiers for all the other compatible strings +                    (possibly empty)

Can you update this comment to explain what is returned when it is not found?

Sure.

...

...

+        """ +        compat_c, aliases_c = get_compat_name(node) +        if compat_c not in self._drivers: +            compat_c_old = compat_c +            compat_c = self._driver_aliases.get(compat_c) +            if not compat_c: +                print('WARNING: the driver %s was not found in the driver list' % (compat_c_old))

This creates lots of warnings at present. Either we need a patch to clean up the differences in the source code, or we need to disable the warning.

Regarding this, maybe we should have a list of driver names we don't expect to support, like simple_bus. For this to work probably the best approach is to have a config option similar to CONFIG_OF_REMOVE_PROPS, so each config can add their owns.

I've been thinking about this issue, I think it would be better to only print the warning if build is issue with

make V=1

What do you think?

...

Also the warning is not really actionable. It needs to add mention of U_BOOT_DEVICE and ...ALIAS.

I agree with you. Thanks.

...

In future we can scan the compatible strings and tell the user what changes to make, I suppose.

Yes, it will be a great improvement!

...

...

+                compat_c = compat_c_old +            else: # pragma: no cover

Need to fix the coverage here

Noted. I will a few more tests.

...

...

+                aliases_c = [compat_c_old] + aliases_c

+        return compat_c, aliases_c

def setup_output(self, fname):           """Set up the output destination @@ -243,6 +277,46 @@ class DtbPlatdata(object):               return PhandleInfo(max_args, args)           return None

+    def scan_driver(self, fn): +        """Scan a driver file to build a list of driver names and aliases

+        This procedure will populate self._drivers and self._driver_aliases

+        Args +            fn: Driver filename to scan +        """ +        with open(fn) as fd:

drop blank line

OK.

...

...

+            buff = fd.read()

+            # The following re will search for driver names declared as +            # U_BOOT_DRIVER(driver_name) +            drivers = re.findall('U_BOOT_DRIVER((.*))', buff)

+            for driver in drivers: +                self._drivers.append(driver)

+            # The following re will search for driver aliases declared as +            # U_BOOT_DRIVER_ALIAS(alias, driver_name) +            driver_aliases = re.findall('U_BOOT_DRIVER_ALIAS(\s*(\w+)\s*,\s*(\w+)\s*)', buff)

+            for alias in driver_aliases: # pragma: no cover +                if len(alias) != 2: +                    continue +                self._driver_aliases[alias[1]] = alias[0]

+    def scan_drivers(self): +        """Scan the driver folders to build a list of driver names and aliases

+        This procedure will populate self._drivers and self._driver_aliases

+        """ +        for (dirpath, dirnames, filenames) in os.walk('./'):

This doesn't work for out-of-tree cases (make O=xxx). You may need to pass $(srctree) to dtc as an argument.

Thanks for pointing at this.

...

...

+            for fn in filenames: +                if not fn.endswith('.c'): +                    continue +                self.scan_driver(dirpath + '/' + fn)

def scan_dtb(self):           """Scan the device tree to obtain a tree of nodes and properties

@@ -353,7 +427,7 @@ class DtbPlatdata(object):           """           structs = {}           for node in self._valid_nodes: -            node_name, _ = get_compat_name(node) +            node_name, _ = self.get_normalized_compat_name(node)               fields = {}

# Get a list of all the valid properties in this node. @@ -377,14 +451,14 @@ class DtbPlatdata(object):

upto = 0           for node in self._valid_nodes: -            node_name, _ = get_compat_name(node) +            node_name, _ = self.get_normalized_compat_name(node)               struct = structs[node_name]               for name, prop in node.props.items():                   if name not in PROP_IGNORE_LIST and name[0] != '#':                       prop.Widen(struct[name])               upto += 1

-            struct_name, aliases = get_compat_name(node) +            struct_name, aliases = self.get_normalized_compat_name(node)               for alias in aliases:                   self._aliases[alias] = struct_name

@@ -461,7 +535,7 @@ class DtbPlatdata(object):           Args:               node: node to output           """ -        struct_name, _ = get_compat_name(node) +        struct_name, _ = self.get_normalized_compat_name(node)           var_name = conv_name_to_c(node.name)           self.buf('static const struct %s%s %s%s = {\n' %                    (STRUCT_PREFIX, struct_name, VAL_PREFIX, var_name)) @@ -562,6 +636,7 @@ def run_steps(args, dtb_file, include_disabled, output):           raise ValueError('Please specify a command: struct, platdata')

plat = DtbPlatdata(dtb_file, include_disabled) +    plat.scan_drivers()       plat.scan_dtb()       plat.scan_tree()       plat.scan_reg_sizes() -- 2.20.1

Thanks once again for your review

Regards,

Walter

Hi Walter,

On Mon, 8 Jun 2020 at 09:49, Walter Lozano walter.lozano@collabora.com wrote:

Hi Simon,

On 4/6/20 12:59, Simon Glass wrote:

...

Hi Walter,

On Fri, 29 May 2020 at 12:15, Walter Lozano walter.lozano@collabora.com wrote:

...

Currently dtoc scans dtbs to convert them to struct platdata and to generate U_BOOT_DEVICE entries. These entries need to be filled with the driver name, but at this moment the information used is the compatible name present in the dtb. This causes that only nodes with a compatible name that matches a driver name generate a working entry.

In order to improve this behaviour, this patch adds to dtoc the capability of scan drivers source code to generate a list of valid driver names. This allows to rise a warning in the case that an U_BOOT_DEVICE entry will try to use a name not valid.

Additionally, in order to add more flexibility to the solution, adds the U_BOOT_DRIVER_ALIAS macro, which generates no code at all, but allows an easy way to declare driver name aliases. Thanks to this, dtoc can look for the driver name based on its alias when it populates the U_BOOT_DEVICE entry.

Signed-off-by: Walter Lozano walter.lozano@collabora.com

include/dm/device.h | 7 ++++ tools/dtoc/dtb_platdata.py | 83 ++++++++++++++++++++++++++++++++++++-- 2 files changed, 86 insertions(+), 4 deletions(-)

diff --git a/include/dm/device.h b/include/dm/device.h index 975eec5d0e..2cfe10766f 100644 --- a/include/dm/device.h +++ b/include/dm/device.h @@ -282,6 +282,13 @@ struct driver { #define DM_GET_DRIVER(__name) \ ll_entry_get(struct driver, __name, driver)

+/**

• • Declare a macro to state a alias for a driver name. This macro will
• • produce no code but its information will be parsed by tools like
• • dtoc
• */

+#define U_BOOT_DRIVER_ALIAS(__name, __alias)

• /**
  • dev_get_platdata() - Get the platform data for a device

diff --git a/tools/dtoc/dtb_platdata.py b/tools/dtoc/dtb_platdata.py index ecfe0624d1..23cfda2f88 100644 --- a/tools/dtoc/dtb_platdata.py +++ b/tools/dtoc/dtb_platdata.py @@ -13,6 +13,8 @@ static data.

import collections import copy +import os +import re import sys

from dtoc import fdt @@ -140,6 +142,9 @@ class DtbPlatdata(object): _include_disabled: true to include nodes marked status = "disabled" _outfile: The current output file (sys.stdout or a real file) _lines: Stashed list of output lines for outputting in the future

•    _aliases: Dict that hold aliases for compatible strings
  

key: The driver name, i.e. the part between brackets in U_BOOT_DRIVER(xx) ?? value: ...

Noted.

...

...

•    _drivers: List of valid driver names found in drivers/
  

•    _driver_aliases: Dict that holds aliases for driver names
  

key: vaue:

OK.

...

...

  """
  def __init__(self, dtb_fname, include_disabled):
      self._fdt = None

@@ -149,6 +154,35 @@ class DtbPlatdata(object): self._outfile = None self._lines = [] self._aliases = {}

•    self._drivers = []
  

•    self._driver_aliases = {}
  

• def get_normalized_compat_name(self, node):

•    """Get a node's normalized compat name
  

•    Returns a valid driver name by retrieving node's first compatible
  

•    string as a C identifier and perfomrming a check against _drivers
  

performing

Noted.

...

...

•    and a lookup in driver_aliases rising a warning in case of failure.
  

s/ rising/, printing/

OK.

...

...

•    Args:
  

•        node: Node object to check
  

•    Return:
  

•        Tuple:
  

•            Driver name associated with the first compatible string
  

•            List of C identifiers for all the other compatible strings
  

•                (possibly empty)
  

Can you update this comment to explain what is returned when it is not found?

Sure.

...

...

•    """
  

•    compat_c, aliases_c = get_compat_name(node)
  

•    if compat_c not in self._drivers:
  

•        compat_c_old = compat_c
  

•        compat_c = self._driver_aliases.get(compat_c)
  

•        if not compat_c:
  

•            print('WARNING: the driver %s was not found in the driver list' % (compat_c_old))
  

This creates lots of warnings at present. Either we need a patch to clean up the differences in the source code, or we need to disable the warning.

Regarding this, maybe we should have a list of driver names we don't expect to support, like simple_bus. For this to work probably the best approach is to have a config option similar to CONFIG_OF_REMOVE_PROPS, so each config can add their owns.

Or perhaps have another macro in the source code that indicates that the driver cannot be used with of-platdata and should be ignored?

Regards, Simon

Hi Walter,

On Thu, 11 Jun 2020 at 11:11, Walter Lozano walter.lozano@collabora.com wrote:

Hi Simon

On 11/6/20 13:45, Simon Glass wrote:

...

Hi Walter,

On Mon, 8 Jun 2020 at 09:49, Walter Lozano walter.lozano@collabora.com wrote:

...

Hi Simon,

On 4/6/20 12:59, Simon Glass wrote:

...

Hi Walter,

On Fri, 29 May 2020 at 12:15, Walter Lozano walter.lozano@collabora.com wrote:

...

Currently dtoc scans dtbs to convert them to struct platdata and to generate U_BOOT_DEVICE entries. These entries need to be filled with the driver name, but at this moment the information used is the compatible name present in the dtb. This causes that only nodes with a compatible name that matches a driver name generate a working entry.

In order to improve this behaviour, this patch adds to dtoc the capability of scan drivers source code to generate a list of valid driver names. This allows to rise a warning in the case that an U_BOOT_DEVICE entry will try to use a name not valid.

Additionally, in order to add more flexibility to the solution, adds the U_BOOT_DRIVER_ALIAS macro, which generates no code at all, but allows an easy way to declare driver name aliases. Thanks to this, dtoc can look for the driver name based on its alias when it populates the U_BOOT_DEVICE entry.

Signed-off-by: Walter Lozano walter.lozano@collabora.com

include/dm/device.h | 7 ++++ tools/dtoc/dtb_platdata.py | 83 ++++++++++++++++++++++++++++++++++++-- 2 files changed, 86 insertions(+), 4 deletions(-)

diff --git a/include/dm/device.h b/include/dm/device.h index 975eec5d0e..2cfe10766f 100644 --- a/include/dm/device.h +++ b/include/dm/device.h @@ -282,6 +282,13 @@ struct driver { #define DM_GET_DRIVER(__name) \ ll_entry_get(struct driver, __name, driver)

+/**

• • Declare a macro to state a alias for a driver name. This macro will
• • produce no code but its information will be parsed by tools like
• • dtoc
• */

+#define U_BOOT_DRIVER_ALIAS(__name, __alias)

• /**
  • dev_get_platdata() - Get the platform data for a device

diff --git a/tools/dtoc/dtb_platdata.py b/tools/dtoc/dtb_platdata.py index ecfe0624d1..23cfda2f88 100644 --- a/tools/dtoc/dtb_platdata.py +++ b/tools/dtoc/dtb_platdata.py @@ -13,6 +13,8 @@ static data.

import collections import copy +import os +import re import sys

from dtoc import fdt @@ -140,6 +142,9 @@ class DtbPlatdata(object): _include_disabled: true to include nodes marked status = "disabled" _outfile: The current output file (sys.stdout or a real file) _lines: Stashed list of output lines for outputting in the future

•    _aliases: Dict that hold aliases for compatible strings
  

key: The driver name, i.e. the part between brackets in U_BOOT_DRIVER(xx) ?? value: ...

Noted.

...

...

•    _drivers: List of valid driver names found in drivers/
  

•    _driver_aliases: Dict that holds aliases for driver names
  

key: vaue:

OK.

...

...

   """
   def __init__(self, dtb_fname, include_disabled):
       self._fdt = None

@@ -149,6 +154,35 @@ class DtbPlatdata(object): self._outfile = None self._lines = [] self._aliases = {}

•    self._drivers = []
  

•    self._driver_aliases = {}
  

• def get_normalized_compat_name(self, node):

•    """Get a node's normalized compat name
  

•    Returns a valid driver name by retrieving node's first compatible
  

•    string as a C identifier and perfomrming a check against _drivers
  

performing

Noted.

...

...

•    and a lookup in driver_aliases rising a warning in case of failure.
  

s/ rising/, printing/

OK.

...

...

•    Args:
  

•        node: Node object to check
  

•    Return:
  

•        Tuple:
  

•            Driver name associated with the first compatible string
  

•            List of C identifiers for all the other compatible strings
  

•                (possibly empty)
  

Can you update this comment to explain what is returned when it is not found?

Sure.

...

...

•    """
  

•    compat_c, aliases_c = get_compat_name(node)
  

•    if compat_c not in self._drivers:
  

•        compat_c_old = compat_c
  

•        compat_c = self._driver_aliases.get(compat_c)
  

•        if not compat_c:
  

•            print('WARNING: the driver %s was not found in the driver list' % (compat_c_old))
  

This creates lots of warnings at present. Either we need a patch to clean up the differences in the source code, or we need to disable the warning.

Regarding this, maybe we should have a list of driver names we don't expect to support, like simple_bus. For this to work probably the best approach is to have a config option similar to CONFIG_OF_REMOVE_PROPS, so each config can add their owns.

Or perhaps have another macro in the source code that indicates that the driver cannot be used with of-platdata and should be ignored?

I don't fully understand your idea. As I see, the warning should help to spot that you will be trying to create a U_BOOT_DEVICE without a proper driver name, which means that compatible string does not match a driver name. The most probably reason for this is that driver doesn't fully support of-platdata, or at least an U_BOOT_DRIVER_ALIAS is missing.

From my understanding by adding a another macro to indicate that a driver cannot be used, or even better to add a macro which tells that a driver supports of-platdata, will give us a cleaner dt-struct, which will be nice, however my first sentence still makes sense.

Could you clarify?

I just mean that you should fix all the warnings, so that none are printed in the normal case. Then people can see the problems they create. Perhaps then it could even be an error rather than a warning?

Also I have mentioned

...

...

I've been thinking about this issue, I think it would be better to only print the warning if build is issue with

make V=1

What do you think?

...

Can we not fix the warnings?

So it is not clear to me which idea is better from your point if view.

I don't think V=1 helps. Then you suppress the warning for most people. If it isn't important, we shouldn't print it. But see above - hoping these warnings can be fixed by renaming drivers.

Regards, Simon

Hi Walter,

On Thu, 11 Jun 2020 at 13:07, Walter Lozano walter.lozano@collabora.com wrote:

Hi Simon,

On 11/6/20 14:22, Simon Glass wrote:

...

Hi Walter,

On Thu, 11 Jun 2020 at 11:11, Walter Lozano walter.lozano@collabora.com wrote:

...

Hi Simon

On 11/6/20 13:45, Simon Glass wrote:

...

Hi Walter,

On Mon, 8 Jun 2020 at 09:49, Walter Lozano walter.lozano@collabora.com wrote:

...

Hi Simon,

On 4/6/20 12:59, Simon Glass wrote:

...

Hi Walter,

On Fri, 29 May 2020 at 12:15, Walter Lozano walter.lozano@collabora.com wrote: > Currently dtoc scans dtbs to convert them to struct platdata and > to generate U_BOOT_DEVICE entries. These entries need to be filled > with the driver name, but at this moment the information used is the > compatible name present in the dtb. This causes that only nodes with > a compatible name that matches a driver name generate a working > entry. > > In order to improve this behaviour, this patch adds to dtoc the > capability of scan drivers source code to generate a list of valid driver > names. This allows to rise a warning in the case that an U_BOOT_DEVICE > entry will try to use a name not valid. > > Additionally, in order to add more flexibility to the solution, adds the > U_BOOT_DRIVER_ALIAS macro, which generates no code at all, but allows an > easy way to declare driver name aliases. Thanks to this, dtoc can look > for the driver name based on its alias when it populates the U_BOOT_DEVICE > entry. > > Signed-off-by: Walter Lozano walter.lozano@collabora.com > --- > include/dm/device.h | 7 ++++ > tools/dtoc/dtb_platdata.py | 83 ++++++++++++++++++++++++++++++++++++-- > 2 files changed, 86 insertions(+), 4 deletions(-) > > diff --git a/include/dm/device.h b/include/dm/device.h > index 975eec5d0e..2cfe10766f 100644 > --- a/include/dm/device.h > +++ b/include/dm/device.h > @@ -282,6 +282,13 @@ struct driver { > #define DM_GET_DRIVER(__name) \ > ll_entry_get(struct driver, __name, driver) > > +/** > + * Declare a macro to state a alias for a driver name. This macro will > + * produce no code but its information will be parsed by tools like > + * dtoc > + */ > +#define U_BOOT_DRIVER_ALIAS(__name, __alias) > + > /** > * dev_get_platdata() - Get the platform data for a device > * > diff --git a/tools/dtoc/dtb_platdata.py b/tools/dtoc/dtb_platdata.py > index ecfe0624d1..23cfda2f88 100644 > --- a/tools/dtoc/dtb_platdata.py > +++ b/tools/dtoc/dtb_platdata.py > @@ -13,6 +13,8 @@ static data. > > import collections > import copy > +import os > +import re > import sys > > from dtoc import fdt > @@ -140,6 +142,9 @@ class DtbPlatdata(object): > _include_disabled: true to include nodes marked status = "disabled" > _outfile: The current output file (sys.stdout or a real file) > _lines: Stashed list of output lines for outputting in the future > + _aliases: Dict that hold aliases for compatible strings key: The driver name, i.e. the part between brackets in U_BOOT_DRIVER(xx) ?? value: ...

Noted.

...

> + _drivers: List of valid driver names found in drivers/ > + _driver_aliases: Dict that holds aliases for driver names key: vaue:

OK.

...

> """ > def __init__(self, dtb_fname, include_disabled): > self._fdt = None > @@ -149,6 +154,35 @@ class DtbPlatdata(object): > self._outfile = None > self._lines = [] > self._aliases = {} > + self._drivers = [] > + self._driver_aliases = {} > + > + def get_normalized_compat_name(self, node): > + """Get a node's normalized compat name > + > + Returns a valid driver name by retrieving node's first compatible > + string as a C identifier and perfomrming a check against _drivers performing

Noted.

...

> + and a lookup in driver_aliases rising a warning in case of failure. s/ rising/, printing/

OK.

...

> + > + Args: > + node: Node object to check > + Return: > + Tuple: > + Driver name associated with the first compatible string > + List of C identifiers for all the other compatible strings > + (possibly empty) Can you update this comment to explain what is returned when it is not found?

Sure.

...

> + """ > + compat_c, aliases_c = get_compat_name(node) > + if compat_c not in self._drivers: > + compat_c_old = compat_c > + compat_c = self._driver_aliases.get(compat_c) > + if not compat_c: > + print('WARNING: the driver %s was not found in the driver list' % (compat_c_old)) This creates lots of warnings at present. Either we need a patch to clean up the differences in the source code, or we need to disable the warning.

Regarding this, maybe we should have a list of driver names we don't expect to support, like simple_bus. For this to work probably the best approach is to have a config option similar to CONFIG_OF_REMOVE_PROPS, so each config can add their owns.

Or perhaps have another macro in the source code that indicates that the driver cannot be used with of-platdata and should be ignored?

I don't fully understand your idea. As I see, the warning should help to spot that you will be trying to create a U_BOOT_DEVICE without a proper driver name, which means that compatible string does not match a driver name. The most probably reason for this is that driver doesn't fully support of-platdata, or at least an U_BOOT_DRIVER_ALIAS is missing.

From my understanding by adding a another macro to indicate that a driver cannot be used, or even better to add a macro which tells that a driver supports of-platdata, will give us a cleaner dt-struct, which will be nice, however my first sentence still makes sense.

Could you clarify?

I just mean that you should fix all the warnings, so that none are printed in the normal case. Then people can see the problems they create. Perhaps then it could even be an error rather than a warning?

Thanks for taking the time to explain your point. Let me put an example in order to check if we agree.

Currently, using sandbox_spl_defconfig several warnings arise, for instance

WARNING: the driver sandbox_serial was not found in the driver list

the driver is driver/serial/sandbox.c

The reason for this warning is that in sandbox_serial is not declared neither as a driver nor as an alias. In this case, this device won't work with of-platdata as it could not be bound. Am I correct?

To disable the warning is to rename the driver or to add an alias as

U_BOOT_DRIVER_ALIAS(serial_sandbox, sandbox_serial)

Would you like me to add U_BOOT_DRIVER_ALIAS for these kind of cases?

I think it would be better to rename the driver. The names are a bit arbitrary anyway at present.

However removing the warning without properly testing the driver with of-platdata might hide runtime issues, don't you think so?

Well you can only make it better, I suspect, since you are correcting the name.

Also, if you feel that this discussion will take time, I have no problem in moving the warning to a different patchset, to avoid delay your work. I totally open to your suggestions.

Sure I suppose we could start with what you have, with the warnings, and then submit a fixup afterwards.

Regards, Simon

Hi Walter,

On Fri, 12 Jun 2020 at 11:38, Walter Lozano walter.lozano@collabora.com wrote:

On 11/6/20 23:22, Simon Glass wrote:

...

Hi Walter,

On Thu, 11 Jun 2020 at 13:07, Walter Lozano walter.lozano@collabora.com wrote:

...

Hi Simon,

On 11/6/20 14:22, Simon Glass wrote:

...

Hi Walter,

On Thu, 11 Jun 2020 at 11:11, Walter Lozano walter.lozano@collabora.com wrote:

...

Hi Simon

On 11/6/20 13:45, Simon Glass wrote:

...

Hi Walter,

On Mon, 8 Jun 2020 at 09:49, Walter Lozano walter.lozano@collabora.com wrote: > Hi Simon, > > On 4/6/20 12:59, Simon Glass wrote: >> Hi Walter, >> >> On Fri, 29 May 2020 at 12:15, Walter Lozano walter.lozano@collabora.com wrote: >>> Currently dtoc scans dtbs to convert them to struct platdata and >>> to generate U_BOOT_DEVICE entries. These entries need to be filled >>> with the driver name, but at this moment the information used is the >>> compatible name present in the dtb. This causes that only nodes with >>> a compatible name that matches a driver name generate a working >>> entry. >>> >>> In order to improve this behaviour, this patch adds to dtoc the >>> capability of scan drivers source code to generate a list of valid driver >>> names. This allows to rise a warning in the case that an U_BOOT_DEVICE >>> entry will try to use a name not valid. >>> >>> Additionally, in order to add more flexibility to the solution, adds the >>> U_BOOT_DRIVER_ALIAS macro, which generates no code at all, but allows an >>> easy way to declare driver name aliases. Thanks to this, dtoc can look >>> for the driver name based on its alias when it populates the U_BOOT_DEVICE >>> entry. >>> >>> Signed-off-by: Walter Lozano walter.lozano@collabora.com >>> --- >>> include/dm/device.h | 7 ++++ >>> tools/dtoc/dtb_platdata.py | 83 ++++++++++++++++++++++++++++++++++++-- >>> 2 files changed, 86 insertions(+), 4 deletions(-) >>> >>> diff --git a/include/dm/device.h b/include/dm/device.h >>> index 975eec5d0e..2cfe10766f 100644 >>> --- a/include/dm/device.h >>> +++ b/include/dm/device.h >>> @@ -282,6 +282,13 @@ struct driver { >>> #define DM_GET_DRIVER(__name) \ >>> ll_entry_get(struct driver, __name, driver) >>> >>> +/** >>> + * Declare a macro to state a alias for a driver name. This macro will >>> + * produce no code but its information will be parsed by tools like >>> + * dtoc >>> + */ >>> +#define U_BOOT_DRIVER_ALIAS(__name, __alias) >>> + >>> /** >>> * dev_get_platdata() - Get the platform data for a device >>> * >>> diff --git a/tools/dtoc/dtb_platdata.py b/tools/dtoc/dtb_platdata.py >>> index ecfe0624d1..23cfda2f88 100644 >>> --- a/tools/dtoc/dtb_platdata.py >>> +++ b/tools/dtoc/dtb_platdata.py >>> @@ -13,6 +13,8 @@ static data. >>> >>> import collections >>> import copy >>> +import os >>> +import re >>> import sys >>> >>> from dtoc import fdt >>> @@ -140,6 +142,9 @@ class DtbPlatdata(object): >>> _include_disabled: true to include nodes marked status = "disabled" >>> _outfile: The current output file (sys.stdout or a real file) >>> _lines: Stashed list of output lines for outputting in the future >>> + _aliases: Dict that hold aliases for compatible strings >> key: The driver name, i.e. the part between brackets in U_BOOT_DRIVER(xx) ?? >> value: ... > Noted. >>> + _drivers: List of valid driver names found in drivers/ >>> + _driver_aliases: Dict that holds aliases for driver names >> key: >> vaue: > OK. >>> """ >>> def __init__(self, dtb_fname, include_disabled): >>> self._fdt = None >>> @@ -149,6 +154,35 @@ class DtbPlatdata(object): >>> self._outfile = None >>> self._lines = [] >>> self._aliases = {} >>> + self._drivers = [] >>> + self._driver_aliases = {} >>> + >>> + def get_normalized_compat_name(self, node): >>> + """Get a node's normalized compat name >>> + >>> + Returns a valid driver name by retrieving node's first compatible >>> + string as a C identifier and perfomrming a check against _drivers >> performing > Noted. >>> + and a lookup in driver_aliases rising a warning in case of failure. >> s/ rising/, printing/ > OK. >>> + >>> + Args: >>> + node: Node object to check >>> + Return: >>> + Tuple: >>> + Driver name associated with the first compatible string >>> + List of C identifiers for all the other compatible strings >>> + (possibly empty) >> Can you update this comment to explain what is returned when it is not found? > Sure. >>> + """ >>> + compat_c, aliases_c = get_compat_name(node) >>> + if compat_c not in self._drivers: >>> + compat_c_old = compat_c >>> + compat_c = self._driver_aliases.get(compat_c) >>> + if not compat_c: >>> + print('WARNING: the driver %s was not found in the driver list' % (compat_c_old)) >> This creates lots of warnings at present. Either we need a patch to >> clean up the differences in the source code, or we need to disable the >> warning. > Regarding this, maybe we should have a list of driver names we don't > expect to support, like simple_bus. For this to work probably the best > approach is to have a config option similar to CONFIG_OF_REMOVE_PROPS, > so each config can add their owns. Or perhaps have another macro in the source code that indicates that the driver cannot be used with of-platdata and should be ignored?

I don't fully understand your idea. As I see, the warning should help to spot that you will be trying to create a U_BOOT_DEVICE without a proper driver name, which means that compatible string does not match a driver name. The most probably reason for this is that driver doesn't fully support of-platdata, or at least an U_BOOT_DRIVER_ALIAS is missing.

From my understanding by adding a another macro to indicate that a driver cannot be used, or even better to add a macro which tells that a driver supports of-platdata, will give us a cleaner dt-struct, which will be nice, however my first sentence still makes sense.

Could you clarify?

I just mean that you should fix all the warnings, so that none are printed in the normal case. Then people can see the problems they create. Perhaps then it could even be an error rather than a warning?

Thanks for taking the time to explain your point. Let me put an example in order to check if we agree.

Currently, using sandbox_spl_defconfig several warnings arise, for instance

WARNING: the driver sandbox_serial was not found in the driver list

the driver is driver/serial/sandbox.c

The reason for this warning is that in sandbox_serial is not declared neither as a driver nor as an alias. In this case, this device won't work with of-platdata as it could not be bound. Am I correct?

To disable the warning is to rename the driver or to add an alias as

U_BOOT_DRIVER_ALIAS(serial_sandbox, sandbox_serial)

Would you like me to add U_BOOT_DRIVER_ALIAS for these kind of cases?

I think it would be better to rename the driver. The names are a bit arbitrary anyway at present.

Yes, I agree. Actually I rename some drivers for iMX6 for that reason. Let me share some examples in order to check if we agree

diff --git a/drivers/gpio/rk_gpio.c b/drivers/gpio/rk_gpio.c index 3d96678a45..8cc288581c 100644 --- a/drivers/gpio/rk_gpio.c +++ b/drivers/gpio/rk_gpio.c @@ -172,8 +172,8 @@ static const struct udevice_id rockchip_gpio_ids[] = { { } };

-U_BOOT_DRIVER(gpio_rockchip) = {

•   .name   = "gpio_rockchip",
  

+U_BOOT_DRIVER(rockchip_gpio_bank) = {

•   .name   = "rockchip_gpio_bank",
    .id     = UCLASS_GPIO,
    .of_match = rockchip_gpio_ids,
    .ops    = &gpio_rockchip_ops,
  

diff --git a/drivers/gpio/sandbox.c b/drivers/gpio/sandbox.c index 9549c74c2b..ff46d3c8d1 100644 --- a/drivers/gpio/sandbox.c +++ b/drivers/gpio/sandbox.c @@ -243,8 +243,8 @@ static const struct udevice_id sandbox_gpio_ids[] = { { } };

-U_BOOT_DRIVER(gpio_sandbox) = {

•   .name   = "gpio_sandbox",
  

+U_BOOT_DRIVER(sandbox_gpio) = {

•   .name   = "sandbox_gpio",
    .id     = UCLASS_GPIO,
    .of_match = sandbox_gpio_ids,
    .ofdata_to_platdata = sandbox_gpio_ofdata_to_platdata,
  

diff --git a/drivers/pinctrl/rockchip/pinctrl-rk3288.c b/drivers/pinctrl/rockchip/pinctrl-rk3288.c index 7ae147f304..c617d21b7a 100644 --- a/drivers/pinctrl/rockchip/pinctrl-rk3288.c +++ b/drivers/pinctrl/rockchip/pinctrl-rk3288.c @@ -240,7 +240,7 @@ static const struct udevice_id rk3288_pinctrl_ids[] = { { } };

-U_BOOT_DRIVER(pinctrl_rk3288) = { +U_BOOT_DRIVER(rockchip_rk3288_pinctrl) = { .name = "rockchip_rk3288_pinctrl", .id = UCLASS_PINCTRL, .of_match = rk3288_pinctrl_ids, diff --git a/drivers/power/pmic/rk8xx.c b/drivers/power/pmic/rk8xx.c index 52e6d9d8c0..d870ed7113 100644 --- a/drivers/power/pmic/rk8xx.c +++ b/drivers/power/pmic/rk8xx.c @@ -182,8 +182,8 @@ static const struct udevice_id rk8xx_ids[] = { { } };

-U_BOOT_DRIVER(pmic_rk8xx) = {

•   .name = "rk8xx pmic",
  

+U_BOOT_DRIVER(rockchip_rk805) = {

•   .name = "rockchip_rk805",
    .id = UCLASS_PMIC,
    .of_match = rk8xx_ids,
  

  #if CONFIG_IS_ENABLED(PMIC_CHILDREN)

Yes that seems OK.

...

...

However removing the warning without properly testing the driver with of-platdata might hide runtime issues, don't you think so?

Well you can only make it better, I suspect, since you are correcting the name.

...

Also, if you feel that this discussion will take time, I have no problem in moving the warning to a different patchset, to avoid delay your work. I totally open to your suggestions.

Sure I suppose we could start with what you have, with the warnings, and then submit a fixup afterwards.

I have no problem, let's see if we can agree in this patchset in order to keep improving things.

OK.

Regards, Simon

Hi Walter,

On Fri, 29 May 2020 at 12:15, Walter Lozano walter.lozano@collabora.com wrote:

As dtoc now performs checks for valid driver names, when running dtoc tests several warnings arise as these tests don't use valid driver names.

This patch adds an option to disable those warning, which is only intended for running tests.

Signed-off-by: Walter Lozano walter.lozano@collabora.com

tools/dtoc/dtb_platdata.py | 11 +++++--- tools/dtoc/test_dtoc.py | 54 +++++++++++++++++++------------------- 2 files changed, 34 insertions(+), 31 deletions(-)

diff --git a/tools/dtoc/dtb_platdata.py b/tools/dtoc/dtb_platdata.py index 23cfda2f88..0a54188348 100644 --- a/tools/dtoc/dtb_platdata.py +++ b/tools/dtoc/dtb_platdata.py @@ -141,17 +141,19 @@ class DtbPlatdata(object): _valid_nodes: A list of Node object with compatible strings _include_disabled: true to include nodes marked status = "disabled" _outfile: The current output file (sys.stdout or a real file)

•    _warning_disabled: true to disable warnings about driver names not found
     _lines: Stashed list of output lines for outputting in the future
     _aliases: Dict that hold aliases for compatible strings
     _drivers: List of valid driver names found in drivers/
     _driver_aliases: Dict that holds aliases for driver names
  

  """

• def __init__(self, dtb_fname, include_disabled):

• def __init__(self, dtb_fname, include_disabled, warning_disable): self._fdt = None self._dtb_fname = dtb_fname self._valid_nodes = None self._include_disabled = include_disabled self._outfile = None

•    self._warning_disable = warning_disable
     self._lines = []
     self._aliases = {}
     self._drivers = []
  

@@ -177,7 +179,8 @@ class DtbPlatdata(object): compat_c_old = compat_c compat_c = self._driver_aliases.get(compat_c) if not compat_c:

•            print('WARNING: the driver %s was not found in the driver list' % (compat_c_old))
  

•            if not self._warning_disable: # pragma: no cover
  

Need coverage for this.

•                print('WARNING: the driver %s was not found in the driver list' % (compat_c_old))
             compat_c = compat_c_old
         else: # pragma: no cover
             aliases_c = [compat_c_old] + aliases_c
  

@@ -623,7 +626,7 @@ class DtbPlatdata(object): nodes_to_output.remove(node)

-def run_steps(args, dtb_file, include_disabled, output): +def run_steps(args, dtb_file, include_disabled, output, warning_disable = False):

no spaces around =

 """Run all the steps of the dtoc tool

 Args:

@@ -635,7 +638,7 @@ def run_steps(args, dtb_file, include_disabled, output): if not args: raise ValueError('Please specify a command: struct, platdata')

• plat = DtbPlatdata(dtb_file, include_disabled)

• plat = DtbPlatdata(dtb_file, include_disabled, warning_disable) plat.scan_drivers() plat.scan_dtb() plat.scan_tree()

diff --git a/tools/dtoc/test_dtoc.py b/tools/dtoc/test_dtoc.py index 8498e8303c..a9b605cac8 100755 --- a/tools/dtoc/test_dtoc.py +++ b/tools/dtoc/test_dtoc.py @@ -154,12 +154,12 @@ class TestDtoc(unittest.TestCase): """Test output from a device tree file with no nodes""" dtb_file = get_dtb_file('dtoc_test_empty.dts') output = tools.GetOutputFilename('output')

•    dtb_platdata.run_steps(['struct'], dtb_file, False, output)
  

•    dtb_platdata.run_steps(['struct'], dtb_file, False, output, True)
     with open(output) as infile:
         lines = infile.read().splitlines()
     self.assertEqual(HEADER.splitlines(), lines)
  
  

•    dtb_platdata.run_steps(['platdata'], dtb_file, False, output)
  

•    dtb_platdata.run_steps(['platdata'], dtb_file, False, output, True)
  

Can you create run_test which calls run_steps with that set that to True, to avoid adding the param everywhere in this file?

Regards, Simon

Hi Walter,

On Fri, 29 May 2020 at 12:56, Walter Lozano walter.lozano@collabora.com wrote:

On 29/5/20 15:15, Walter Lozano wrote:

...

Currently when creating an U_BOOT_DEVICE entry a struct driver_info is declared, which contains the data needed to instantiate the device. However, the actual device is created at runtime and there is no proper way to get the device based on its struct driver_info.

This patch extends struct driver_info adding a pointer to udevice which is populated during the bind process, allowing to generate a set of functions to get the device based on its struct driver_info.

Signed-off-by: Walter Lozano walter.lozano@collabora.com

drivers/core/device.c | 26 +++++++++++++++++++++++--- drivers/core/root.c | 4 ++++ include/dm/device.h | 14 ++++++++++++++ include/dm/platdata.h | 14 ++++++++++++++ 4 files changed, 55 insertions(+), 3 deletions(-)

diff --git a/drivers/core/device.c b/drivers/core/device.c index a0ad080aaf..5adbc30849 100644 --- a/drivers/core/device.c +++ b/drivers/core/device.c @@ -250,6 +250,7 @@ int device_bind_by_name(struct udevice *parent, bool pre_reloc_only, { struct driver *drv; uint platdata_size = 0;

• int ret = 0;
  
  drv = lists_driver_lookup_name(info->name);
  if (!drv)
  

@@ -260,9 +261,16 @@ int device_bind_by_name(struct udevice *parent, bool pre_reloc_only, #if CONFIG_IS_ENABLED(OF_PLATDATA) platdata_size = info->platdata_size; #endif

• return device_bind_common(parent, drv, info->name,
  

•                 (void *)info->platdata, 0, ofnode_null(), platdata_size,
  

•                 devp);
  

• ret = device_bind_common(parent, drv, info->name,
  

•                          (void *)info->platdata, 0, ofnode_null(),
  

•                          platdata_size, devp);
  

• if (ret)
  

•         return ret;
  

+#if CONFIG_IS_ENABLED(OF_PLATDATA)

• info->dev = *devp;
  

+#endif

I have tried to test this using sandbox_spl_defconfig but I've received a segmentation fault when trying to update info->dev, however this code works on iMX6.

Could it be some kind of protection? Any thoughts?

Yes, see u-boot-dm/dtoc-working - arch/sandbox/cpu/u-boot-spl.lds has an attempt to move some of the list stuff into the data region.

[..]

Regards, Simon

Hi Walter,

On Fri, 29 May 2020 at 13:21, Walter Lozano walter.lozano@collabora.com wrote:

Hi Simon,

On 29/5/20 16:00, Simon Glass wrote:

...

Hi Walter,

On Fri, 29 May 2020 at 12:56, Walter Lozano walter.lozano@collabora.com wrote:

...

On 29/5/20 15:15, Walter Lozano wrote:

...

Currently when creating an U_BOOT_DEVICE entry a struct driver_info is declared, which contains the data needed to instantiate the device. However, the actual device is created at runtime and there is no proper way to get the device based on its struct driver_info.

This patch extends struct driver_info adding a pointer to udevice which is populated during the bind process, allowing to generate a set of functions to get the device based on its struct driver_info.

Signed-off-by: Walter Lozano walter.lozano@collabora.com

drivers/core/device.c | 26 +++++++++++++++++++++++--- drivers/core/root.c | 4 ++++ include/dm/device.h | 14 ++++++++++++++ include/dm/platdata.h | 14 ++++++++++++++ 4 files changed, 55 insertions(+), 3 deletions(-)

diff --git a/drivers/core/device.c b/drivers/core/device.c index a0ad080aaf..5adbc30849 100644 --- a/drivers/core/device.c +++ b/drivers/core/device.c @@ -250,6 +250,7 @@ int device_bind_by_name(struct udevice *parent, bool pre_reloc_only, { struct driver *drv; uint platdata_size = 0;

• int ret = 0;
  
   drv = lists_driver_lookup_name(info->name);
   if (!drv)
  

@@ -260,9 +261,16 @@ int device_bind_by_name(struct udevice *parent, bool pre_reloc_only, #if CONFIG_IS_ENABLED(OF_PLATDATA) platdata_size = info->platdata_size; #endif

• return device_bind_common(parent, drv, info->name,
  

•                 (void *)info->platdata, 0, ofnode_null(), platdata_size,
  

•                 devp);
  

• ret = device_bind_common(parent, drv, info->name,
  

•                          (void *)info->platdata, 0, ofnode_null(),
  

•                          platdata_size, devp);
  

• if (ret)
  

•         return ret;
  

+#if CONFIG_IS_ENABLED(OF_PLATDATA)

• info->dev = *devp;
  

+#endif

I have tried to test this using sandbox_spl_defconfig but I've received a segmentation fault when trying to update info->dev, however this code works on iMX6.

Could it be some kind of protection? Any thoughts?

Yes, see u-boot-dm/dtoc-working - arch/sandbox/cpu/u-boot-spl.lds has an attempt to move some of the list stuff into the data region.

Thanks for confirmed it and also for the quick response. I'm about to start a deeper review to your work about tiny-dm now.

OK, but don't take too much notice of it as it is very rough.

One other thing that might be interesting is that I found a way to put links in data structures:

#define DM_DECL_TINY_DRIVER(__name) \ ll_entry_decl(struct tiny_drv, __name, tiny_drv)

/* * Get a pointer to a given tiny driver, for use in data structures. This * requires that the symbol be declared with DM_DECL_TINY_DRIVER() first */ #define DM_REF_TINY_DRIVER(__name) \ ll_entry_ref(struct tiny_drv, __name, tiny_drv)

Regards, Simon

On 4/6/20 12:59, Simon Glass wrote:

Hi Walter,

On Fri, 29 May 2020 at 12:15, Walter Lozano walter.lozano@collabora.com wrote:

...

Currently when creating an U_BOOT_DEVICE entry a struct driver_info is declared, which contains the data needed to instantiate the device. However, the actual device is created at runtime and there is no proper way to get the device based on its struct driver_info.

This patch extends struct driver_info adding a pointer to udevice which is populated during the bind process, allowing to generate a set of functions to get the device based on its struct driver_info.

Signed-off-by: Walter Lozano walter.lozano@collabora.com

drivers/core/device.c | 26 +++++++++++++++++++++++--- drivers/core/root.c | 4 ++++ include/dm/device.h | 14 ++++++++++++++ include/dm/platdata.h | 14 ++++++++++++++ 4 files changed, 55 insertions(+), 3 deletions(-)

Reviewed-by: Simon Glass sjg@chromium.org

Nits below

Thanks for taking the time!

...

diff --git a/drivers/core/device.c b/drivers/core/device.c index a0ad080aaf..5adbc30849 100644 --- a/drivers/core/device.c +++ b/drivers/core/device.c @@ -250,6 +250,7 @@ int device_bind_by_name(struct udevice *parent, bool pre_reloc_only, { struct driver *drv; uint platdata_size = 0;

•   int ret = 0;
  

Can you drop = 0 ?

Noted!

...

     drv = lists_driver_lookup_name(info->name);
     if (!drv)

@@ -260,9 +261,16 @@ int device_bind_by_name(struct udevice *parent, bool pre_reloc_only, #if CONFIG_IS_ENABLED(OF_PLATDATA) platdata_size = info->platdata_size; #endif

•   return device_bind_common(parent, drv, info->name,
  

•                   (void *)info->platdata, 0, ofnode_null(), platdata_size,
  

•                   devp);
  

•   ret = device_bind_common(parent, drv, info->name,
  

•                            (void *)info->platdata, 0, ofnode_null(),
  

•                            platdata_size, devp);
  

•   if (ret)
  

•           return ret;
  

+#if CONFIG_IS_ENABLED(OF_PLATDATA)

•   info->dev = *devp;
  

+#endif

•   return ret;
  

  }

  static void *alloc_priv(int size, uint flags)

@@ -727,6 +735,18 @@ int device_get_global_by_ofnode(ofnode ofnode, struct udevice **devp) return device_get_device_tail(dev, dev ? 0 : -ENOENT, devp); }

+#if CONFIG_IS_ENABLED(OF_PLATDATA) +int device_get_by_driver_info(const struct driver_info *info,

•                         struct udevice **devp)
  

+{

•   struct udevice *dev;
  

•   dev = info->dev;
  

•   return device_get_device_tail(dev, dev ? 0 : -ENOENT, devp);
  

+} +#endif

• int device_find_first_child(const struct udevice *parent, struct udevice **devp) { if (list_empty(&parent->child_head)) {

diff --git a/drivers/core/root.c b/drivers/core/root.c index c9ee56478a..8f47a6b356 100644 --- a/drivers/core/root.c +++ b/drivers/core/root.c @@ -346,6 +346,10 @@ int dm_init_and_scan(bool pre_reloc_only) { int ret;

+#if CONFIG_IS_ENABLED(OF_PLATDATA)

•   populate_phandle_data();
  

+#endif

•    ret = dm_init(IS_ENABLED(CONFIG_OF_LIVE));
     if (ret) {
             debug("dm_init() failed: %d\n", ret);
  

diff --git a/include/dm/device.h b/include/dm/device.h index 2cfe10766f..a3b3e5bc46 100644 --- a/include/dm/device.h +++ b/include/dm/device.h @@ -538,6 +538,20 @@ int device_find_global_by_ofnode(ofnode node, struct udevice **devp); */ int device_get_global_by_ofnode(ofnode node, struct udevice **devp);

+/**

• • device_get_by_driver_info() - Get a device based on driver_info
• • Locates a device by its struct driver_info.
• • The device is probed to activate it ready for use.
• • @info: Struct driver_info

Here you should mention using DM_GET_DRIVER() to find it.

OK, I'll add a reference.

...

• • @devp: Returns pointer to device if found, otherwise this is set to NULL
• • @return 0 if OK, -ve on error
• */

+int device_get_by_driver_info(const struct driver_info *info,

•                         struct udevice **devp);
  

• /**
  • device_find_first_child() - Find the first child of a device

diff --git a/include/dm/platdata.h b/include/dm/platdata.h index c972fa6936..238379b0e4 100644 --- a/include/dm/platdata.h +++ b/include/dm/platdata.h @@ -22,12 +22,14 @@

• @name: Driver name
• @platdata: Driver-specific platform data
• @platdata_size: Size of platform data structure

• • @dev: Device created from this structure data
  */ struct driver_info { const char *name; const void *platdata; #if CONFIG_IS_ENABLED(OF_PLATDATA) uint platdata_size;

•   struct udevice *dev;
  

  #endif };

@@ -43,4 +45,16 @@ struct driver_info { #define U_BOOT_DEVICES(__name) \ ll_entry_declare_list(struct driver_info, __name, driver_info)

+/* Get a pointer to a given driver */ +#define DM_GET_DEVICE(__name) \

•   ll_entry_get(struct driver_info, __name, driver_info)
  

+/**

• • populate_phandle_data() - Populates phandle data in platda
• • This populates phandle data with an U_BOOT_DEVICE entry get by
• • DM_GET_DEVICE. The implementation of this function will be done
• • by dtoc when parsing dtb.
• */

+void populate_phandle_data(void);

dm_populate_phandle_data()

Sure!

...

#endif

2.20.1

Regards,

Walter