failed to generate peripherals for nrf52.svd

[fanoush]

Hi, this is great stuff. However I tried it with https://github.com/posborne/cmsis-svd/blob/master/data/Nordic/nrf52.svd and it only creates memory areas but nothing in Peripherals namespace. The output is:

SVD-Loader.py> Running...
Loading SVD file...
	Done!
Generating memory regions...
	Done!
Generating peripherals...
	FICR
		Failed to generate peripheral FICR
	UICR
		Failed to generate peripheral UICR
	BPROT
		Failed to generate peripheral BPROT
	POWER
		Failed to generate peripheral POWER
	CLOCK
		Failed to generate peripheral CLOCK
	RADIO
		Failed to generate peripheral RADIO
	UARTE0
		Failed to generate peripheral UARTE0
	SPIM0
		Failed to generate peripheral SPIM0
	SPIS0
		Failed to generate peripheral SPIS0
	TWIM0
		Failed to generate peripheral TWIM0
	TWIS0
		Failed to generate peripheral TWIS0
	SPI0
		Failed to generate peripheral SPI0
	TWI0
		Failed to generate peripheral TWI0
	SPIM1
		Failed to generate peripheral SPIM1
	SPIS1
		Failed to generate peripheral SPIS1
	TWIM1
		Failed to generate peripheral TWIM1
	TWIS1
		Failed to generate peripheral TWIS1
	SPI1
		Failed to generate peripheral SPI1
	NFCT
		Failed to generate peripheral NFCT
	GPIOTE
		Failed to generate peripheral GPIOTE
	SAADC
		Failed to generate peripheral SAADC
	TIMER0
		Failed to generate peripheral TIMER0
	TIMER1
		Failed to generate peripheral TIMER1
	TIMER2
		Failed to generate peripheral TIMER2
	RTC0
		Failed to generate peripheral RTC0
	TEMP
		Failed to generate peripheral TEMP
	RNG
		Failed to generate peripheral RNG
	ECB
		Failed to generate peripheral ECB
	CCM
		Failed to generate peripheral CCM
	AAR
		Failed to generate peripheral AAR
	WDT
		Failed to generate peripheral WDT
	RTC1
		Failed to generate peripheral RTC1
	QDEC
		Failed to generate peripheral QDEC
	COMP
		Failed to generate peripheral COMP
	LPCOMP
		Failed to generate peripheral LPCOMP
	SWI0
		Failed to generate peripheral SWI0
	EGU0
		Failed to generate peripheral EGU0
	SWI1
		Failed to generate peripheral SWI1
	EGU1
		Failed to generate peripheral EGU1
	SWI2
		Failed to generate peripheral SWI2
	EGU2
		Failed to generate peripheral EGU2
	SWI3
		Failed to generate peripheral SWI3
	EGU3
		Failed to generate peripheral EGU3
	SWI4
		Failed to generate peripheral SWI4
	EGU4
		Failed to generate peripheral EGU4
	SWI5
		Failed to generate peripheral SWI5
	EGU5
		Failed to generate peripheral EGU5
	TIMER3
		Failed to generate peripheral TIMER3
	TIMER4
		Failed to generate peripheral TIMER4
	PWM0
		Failed to generate peripheral PWM0
	PDM
		Failed to generate peripheral PDM
	NVMC
		Failed to generate peripheral NVMC
	PPI
		Failed to generate peripheral PPI
	MWU
		Failed to generate peripheral MWU
	PWM1
		Failed to generate peripheral PWM1
	PWM2
		Failed to generate peripheral PWM2
	SPIM2
		Failed to generate peripheral SPIM2
	SPIS2
		Failed to generate peripheral SPIS2
	SPI2
		Failed to generate peripheral SPI2
	RTC2
		Failed to generate peripheral RTC2
	I2S
		Failed to generate peripheral I2S
	P0
		Failed to generate peripheral P0
	TWI1
		Failed to generate peripheral TWI1
	UART0
		Failed to generate peripheral UART0
SVD-Loader.py> Finished!

[fanoush]

I have rethrown the exception in the code to see it and it fails with

Traceback (most recent call last):
  File "/home/fanoush/SVD-Loader-Ghidra/SVD-Loader.py", line 135, in <module>
    length = calculate_peripheral_size(peripheral)
  File "/home/fanoush/SVD-Loader-Ghidra/SVD-Loader.py", line 55, in calculate_peripheral_size
    size = max(size, register.address_offset + register._size/8)
TypeError: unsupported operand type(s) for /: 'NoneType' and 'int'

so I guess register._size is not there

[fanoush]

OK, register._size is optional, default is in peripheral.size so I have fixed it like this

 diff --git a/SVD-Loader.py b/SVD-Loader.py
index a55b7a5..689c68d 100644
--- a/SVD-Loader.py
+++ b/SVD-Loader.py
@@ -52,7 +52,8 @@ def reduce_memory_regions(regions):
 def calculate_peripheral_size(peripheral):
        size = 0
        for register in peripheral.registers:
-               size = max(size, register.address_offset + register._size/8)
+               rs = ( peripheral._size if (register._size is None) else register._size ) / 8
+               size = max(size, register.address_offset + rs)
        return size
 
 
@@ -142,14 +143,14 @@ for peripheral in peripherals:
 
                for register in peripheral.registers:
                        r_type = UnsignedIntegerDataType()
-                       rs = register._size / 8
+                       rs = ( peripheral._size if (register._size is None) else register._size ) / 8
                        if rs == 1:
                                r_type = ByteDataType()
                        elif rs == 2:
                                r_type = UnsignedShortDataType()
                        elif rs == 8:
                                r_type = UnsignedLongLongDataType()
-                       peripheral_struct.replaceAtOffset(register.address_offset, r_type, register._size/8, register.name, register.description)
+                       peripheral_struct.replaceAtOffset(register.address_offset, r_type, rs, register.name, register.description)
 
 
                addr = space.getAddress(peripheral_start)

However then I didn't get very far. Some peripherals still fail since they are overlapping which is intentional as they really share same space (how to solve this?) but even for those not overlapping it prints no error but the namespace does not look so rich as on video and analyzed code still looks almost same as before.

[nezza]

Heya, thanks for the feedback! I'll try to look at this in the next days and see what I can figure out!

[nezza]

Yea so this is an issue - for those overlaps we might just create names instead of structures..

[Calvin69]

Hello, sorry for being late to the party. May I know if there is any solutions to this issue? Because I got the same exact error @fanoush had.

[fanoush]

@Calvin69 did you try my patch above to at least get rid of exception because of missing register._size value (which seems to be optional but the code expected it to be always present)?

Another isue is overlapping of address spaces. Just to explain how it is with nrf52 - there are 3 reasons for using same address space:

1. older deprecated and newer (DMA based) version of same HW e.g. UART vs UARTE (E=EasyDMA)
2. older deprected vs several variants of same iterface SPI vs SPIM (master with DMA) vs SPIS (slave with DMA)
3. same HW block used for two mutually exclusive interfaces SPI and TWI=I2C

Type is selected by writing specific value to ENABLE 0x500 register.
The list what is shared is here so e.g. 0x40003000 can be used for old style SPI0 or newer DMA based variants SPIS0 SPIM0 or older TWI0 or newer DMA based TWIM0 TWIS0. Selection is done by writing values 1,2,7 (SPI variants) or 5,6,9 (TWI variants) to 0x40003500.

Typically this selection for specific decompiled code is static so in theory it could be automagically detected from code writing specific value to ENABLE register however the easiest would be to simply set a preference at decompile/loading time which one is preferred and ignore the others.

[nezza]

Heya,

Yea so I have a basic solution working for overlapping address spaces, however I didn't get to fully test it yet unfortunately. In my basic solution I just generate a pspec for the processor, which then will not have structures but at least labels... However your solution with having the user select the right way sounds interesting!
Maybe I'll create an SVD-Loader option to just name the memory addresses without creating structs, that would allow overlapping without issues, at the expense of worse names...

Cheers and thanks for investigating,

Thomas

[Calvin69]

Hello guys,

I just realized that the errors on my console is different that @fanoush . Here's mine:

  File "C:\Users\Admin\ghidra_scripts\SVD-Loader-Ghidra-master\SVD-Loader.py", line 61, in <module>
    parser = SVDParser.for_xml_file(str(svd_file))
  File "C:\Users\Admin\ghidra_scripts\SVD-Loader-Ghidra-master\cmsis_svd\parser.py", line 72, in for_xml_file
    return cls(ET.parse(path), remove_reserved)
  File "C:\Users\Admin\Desktop\ghidra_9.1.2_PUBLIC\Ghidra\Features\Python\data\jython-2.7.1\Lib\xml\etree\ElementTree.py", line 1184, in parse
    tree.parse(source, parser)
  File "C:\Users\Admin\Desktop\ghidra_9.1.2_PUBLIC\Ghidra\Features\Python\data\jython-2.7.1\Lib\xml\etree\ElementTree.py", line 657, in parse
    self._root = parser.close()
  File "C:\Users\Admin\Desktop\ghidra_9.1.2_PUBLIC\Ghidra\Features\Python\data\jython-2.7.1\Lib\xml\etree\ElementTree.py", line 1667, in close
    self._raiseerror(v)
  File "C:\Users\Admin\Desktop\ghidra_9.1.2_PUBLIC\Ghidra\Features\Python\data\jython-2.7.1\Lib\xml\etree\ElementTree.py", line 1519, in _raiseerror
    raise err
  File "<string>", line None
xml.etree.ElementTree.ParseError: Attribute name "data-pjax-transient" associated with an element type "meta" must be followed by the ' = ' character.
SVD-Loader.py> Finished!

It doesn't seems to be an overlapping issue, isn't it? I have never modified any SVD Loader files. I had only modified a Ghidra file which is the ia.sinc which is a solution to this issue answered by nsadeveloper789. Could that be causing the issue?

@nezza are you GhidraNinja? If so, I'd like to say thanks for uploading videos on YT. Really helped me learn a lot.

UPDATE: Looking at the first line of error, I uploaded a .xml file and now I get the same error. Will try fanoush's method and will update again.

UPDATE 2: Just tried fanoush's patch and still get the same thing. No change.

[SourceDiver42]

Hello, I have a similar problem when generating the peripherals but this time it is on the nrf51 platform. As this is an issue which is shared between the nrf51 and nrf52 processor, may I suggest changing the title of the issue to integrate nrf51 as well? Other than that, the overlap seems to be the only issue on my side.

EDIT: I double checked for the nrf51.svd file and indeed, some are requesting the same region. The alternatePeripheral tag references other failing peripherals. In my case for instance the peripheral CLOCK has the alternatePeripheral POWER, which has its own definition somewhere else.

[nezza]

Awesome, I'll try to review on the weekend!

[Avamander]

Yeah, I just encountered this as well with the nrf51.svd file.

[bananabr]

I do have similar problems with rp2040.svd.