have crashlog.py insert a stack frame with $lr when stack frame 0 is address 0

[jasonmolenda]

Summary

A common crash is calling through a null function pointer and trying to execute an instruction at address 0. lldb's backtracer special cases this, but the macOS crash tracer algorithm does not; crash reports will have a stack frame at 0 and then skip the calling stack frame, assuming a stack frame was set up. On arm64, the caller function is likely available in $lr, so let's insert a stack frame with that return pc value.

Also fix a little bug where register names without a "prefix" would get "None" prepended, so "Nonepc", "Nonelr" etc.

rdar://92631787

[JDevlieghere]

Does that mean that the interactive crashlogs (crashlog -i) do this correctly?

[jasonmolenda]

In D125042#3495475, @JDevlieghere wrote:

Does that mean that the interactive crashlogs (crashlog -i) do this correctly?

I didn't look at that when I wrote the patch; asked @mib and it sounded like he was going to do that separately.

[mib]

In D125042#3495479, @jasonmolenda wrote:

In D125042#3495475, @JDevlieghere wrote:

Does that mean that the interactive crashlogs (crashlog -i) do this correctly?

I didn't look at that when I wrote the patch; asked @mib and it sounded like he was going to do that separately.

Interactive crashlogs use the same crashlog parser to fetch the register context, so I think they should work out of the box with this patch applied.

[kastiglione]

thanks for fixing these, especially the lr technique!

lldb/examples/python/crashlog.py

520

(minor) I think a word is missing between "if" and "jump".

533

What do you think of adding a break here? Not that it would change the outcome, but it might make the intention of the loop slightly better.

[mib]

In D125042#3495488, @mib wrote:

In D125042#3495479, @jasonmolenda wrote:

In D125042#3495475, @JDevlieghere wrote:

Does that mean that the interactive crashlogs (crashlog -i) do this correctly?

I didn't look at that when I wrote the patch; asked @mib and it sounded like he was going to do that separately.

Interactive crashlogs use the same crashlog parser to fetch the register context, so I think they should work out of the box with this patch applied.

I applied @jasonmolenda's patch to test it with interactive crashlog and I can see the added stack frame as expected.

[mib]

LGTM! Thanks J!

[jasonmolenda]

lldb/examples/python/crashlog.py

520

yep good catch.

533

ok.

[jasonmolenda]

update to incorporate dave's suggestions.

lldb/examples/python/crashlog.py

def parse_frames(self, thread, json_frames):
             thread.add_ident(ident)
             if ident not in self.crashlog.idents:
                 self.crashlog.idents.append(ident)
 
             frame_offset = int(json_frame['imageOffset'])
             image_addr = self.get_used_image(image_id)['base']
             pc = image_addr + frame_offset
             thread.frames.append(self.crashlog.Frame(idx, pc, frame_offset))
+
+            # on arm64 systems, if it jump through a null function pointer,
+            # we end up at address 0 and the crash reporter unwinder 
+            # misses the frame that actually faulted.  
+            # But $lr can tell us where the last BL/BLR instruction used 
+            # was at, so insert that address as the caller stack frame.  
+            if idx == 0 and pc == 0 and "lr" in thread.registers:
+                pc = thread.registers["lr"]
+                for image in self.data['usedImages']:
+                    text_lo = image['base']
+                    text_hi = text_lo + image['size']
+                    if text_lo <= pc < text_hi:
+                      idx += 1
+                      frame_offset = pc - text_lo
+                      thread.frames.append(self.crashlog.Frame(idx, pc, frame_offset))
+                      break
+
             idx += 1
 
     def parse_threads(self, json_threads):
         idx = 0
         for json_thread in json_threads:
             thread = self.crashlog.Thread(idx, False)
             if 'name' in json_thread:
                 thread.name = json_thread['name']
def parse_thread_registers(self, json_thread_state, prefix=None):
                 registers.update(self.parse_thread_registers(state))
                 continue
             if key == "x":
                 gpr_dict = { str(idx) : reg for idx,reg in enumerate(state) }
                 registers.update(self.parse_thread_registers(gpr_dict, key))
                 continue
             try:
                 value = int(state['value'])
-                registers["{}{}".format(prefix,key)] = value
+                registers["{}{}".format(prefix or '',key)] = value
             except (KeyError, ValueError, TypeError):
                 pass
         return registers
 
     def parse_errors(self, json_data):
        if 'reportNotes' in json_data:
           self.crashlog.errors = json_data['reportNotes']