[U-Boot] [PATCH 20/37] rockchip: Add a script to parse datasheets

Simon Glass sjg at chromium.org
Thu Jan 14 16:58:03 CET 2016


This script has proved useful for parsing datasheets and creating register
shift/mask values for use in header files. Include it in case it is useful
for others.

Signed-off-by: Simon Glass <sjg at chromium.org>
---

 doc/README.rockchip |   6 ++
 tools/rkmux.py      | 218 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 224 insertions(+)
 create mode 100755 tools/rkmux.py

diff --git a/doc/README.rockchip b/doc/README.rockchip
index 9a2ebca..7be5a13 100644
--- a/doc/README.rockchip
+++ b/doc/README.rockchip
@@ -243,6 +243,12 @@ SPI flash.
 
 See above for instructions on how to write a SPI image.
 
+rkmux.py
+--------
+
+You can use this script to create #defines for SoC register access. See the
+script for usage.
+
 
 Device tree and driver model
 ----------------------------
diff --git a/tools/rkmux.py b/tools/rkmux.py
new file mode 100755
index 0000000..3917335
--- /dev/null
+++ b/tools/rkmux.py
@@ -0,0 +1,218 @@
+#!/usr/bin/python
+
+# Script to create enums from datasheet register tables
+#
+# Usage:
+#
+# First, create a text file from the datasheet:
+#    pdftotext -layout /path/to/rockchip-3288-trm.pdf /tmp/asc
+#
+# Then use this script to output the #defines for a particular register:
+#    ./tools/rkmux.py GRF_GPIO4C_IOMUX
+#
+# It will create output suitable for putting in a header file, with SHIFT and
+# MASK values for each bitfield in the register.
+#
+# Note: this tool is not perfect and you may need to edit the resulting code.
+# But it should speed up the process.
+
+import csv
+import re
+import sys
+
+tab_to_col = 3
+
+class RegField:
+    def __init__(self, cols=None):
+        if cols:
+            self.bits, self.attr, self.reset_val, self.desc = (
+                [x.strip() for x in cols])
+            self.desc = [self.desc]
+        else:
+            self.bits = ''
+            self.attr = ''
+            self.reset_val = ''
+            self.desc = []
+
+    def Setup(self, cols):
+        self.bits, self.attr, self.reset_val = cols[0:3]
+        if len(cols) > 3:
+            self.desc.append(cols[3])
+
+    def AddDesc(self, desc):
+        self.desc.append(desc)
+
+    def Show(self):
+        print self
+        print
+        self.__init__()
+
+    def __str__(self):
+        return '%s,%s,%s,%s' % (self.bits, self.attr, self.reset_val,
+                                '\n'.join(self.desc))
+
+class Printer:
+    def __init__(self, name):
+        self.first = True
+        self.name = name
+        self.re_sel = re.compile("[1-9]'b([01]+): (.*)")
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, type, value, traceback):
+        if not self.first:
+            self.output_footer()
+
+    def output_header(self):
+        print '/* %s */' % self.name
+        print 'enum {'
+
+    def output_footer(self):
+        print '};';
+
+    def output_regfield(self, regfield):
+        lines = regfield.desc
+        field = lines[0]
+        #print 'field:', field
+        if field in ['reserved', 'reserve', 'write_enable', 'write_mask']:
+            return
+        if field.endswith('_sel') or field.endswith('_con'):
+            field = field[:-4]
+        elif field.endswith(' iomux'):
+            field = field[:-6]
+        elif field.endswith('_mode') or field.endswith('_mask'):
+            field = field[:-5]
+        #else:
+            #print 'bad field %s' % field
+            #return
+        field = field.upper()
+        if ':' in regfield.bits:
+            bit_high, bit_low = [int(x) for x in regfield.bits.split(':')]
+        else:
+            bit_high = bit_low = int(regfield.bits)
+        bit_width = bit_high - bit_low + 1
+        mask = (1 << bit_width) - 1
+        if self.first:
+            self.first = False
+            self.output_header()
+        else:
+            print
+        out_enum(field, 'shift', bit_low)
+        out_enum(field, 'mask', mask)
+        next_val = -1
+        #print 'lines: %s', lines
+        for line in lines:
+            m = self.re_sel.match(line)
+            if m:
+                val, enum = int(m.group(1), 2), m.group(2)
+                if enum not in ['reserved', 'reserve']:
+                    out_enum(field, enum, val, val == next_val)
+                    next_val = val + 1
+
+
+def process_file(name, fd):
+    field = RegField()
+    reg = ''
+
+    fields = []
+
+    def add_it(field):
+        if field.bits:
+            if reg == name:
+                fields.append(field)
+            field = RegField()
+        return field
+
+    def is_field_start(line):
+       if '=' in line or '+' in line:
+           return False
+       if (line.startswith('gpio') or line.startswith('peri_') or
+                line.endswith('_sel') or line.endswith('_con')):
+           return True
+       if not ' ' in line: # and '_' in line:
+           return True
+       return False
+
+    for line in fd:
+        line = line.rstrip()
+        if line[:4] in ['GRF_', 'PMU_', 'CRU_']:
+            field = add_it(field)
+            reg = line
+            do_this = name == reg
+        elif not line or not line.startswith(' '):
+            continue
+        line = line.replace('\xe2\x80\x99', "'")
+        leading = len(line) - len(line.lstrip())
+        line = line.lstrip()
+        cols = re.split(' *', line, 3)
+        if leading > 15 or (len(cols) > 3 and is_field_start(cols[3])):
+            if is_field_start(line):
+                field = add_it(field)
+            field.AddDesc(line)
+        else:
+            if cols[0] == 'Bit' or len(cols) < 3:
+                continue
+            #print
+            #print field
+            field = add_it(field)
+            field.Setup(cols)
+    field = add_it(field)
+
+    with Printer(name) as printer:
+        for field in fields:
+            #print field
+            printer.output_regfield(field)
+            #print
+
+def out_enum(field, suffix, value, skip_val=False):
+    str = '%s_%s' % (field.upper(), suffix.upper())
+    if not skip_val:
+        tabs = tab_to_col - len(str) / 8
+        if value > 9:
+            val_str = '%#x' % value
+        else:
+            val_str = '%d' % value
+
+        str += '%s= %s' % ('\t' * tabs, val_str)
+    print '\t%s,' % str
+
+# Process a CSV file, e.g. from tabula
+def process_csv(name, fd):
+    reader = csv.reader(fd)
+
+    rows = []
+
+    field = RegField()
+    for row in reader:
+        #print field.desc
+        if not row[0]:
+            field.desc.append(row[3])
+            continue
+        if field.bits:
+            if field.bits != 'Bit':
+                rows.append(field)
+        #print row
+        field = RegField(row)
+
+    with Printer(name) as printer:
+        for row in rows:
+            #print field
+            printer.output_regfield(row)
+            #print
+
+fname = sys.argv[1]
+name = sys.argv[2]
+
+# Read output from pdftotext -layout
+if 1:
+    with open(fname, 'r') as fd:
+        process_file(name, fd)
+
+# Use tabula
+# It seems to be better at outputting text for an entire cell in one cell.
+# But it does not always work. E.g. GRF_GPIO7CH_IOMUX.
+# So there is no point in using it.
+if 0:
+    with open(fname, 'r') as fd:
+        process_csv(name, fd)
-- 
2.6.0.rc2.230.g3dd15c0



More information about the U-Boot mailing list