Add return address stack location to context backtrace #1288 #1740

dmur1 · 2023-05-24T13:17:35Z

Add pwndbg.gdblib.find_addr_on_stack
Change coloration of context backtrace to use pwnlib.color.memory
Remove now defunct address color config from context
Update test_context_backtrace_show_proper_symbol_names regex

* Add pwndbg.gdblib.find_addr_on_stack * Change coloration of context backtrace to use pwnlib.color.memory * Remove now defunct address color config from context * Update test_context_backtrace_show_proper_symbol_names regex

pwndbg/commands/context.py

dmur1 · 2023-05-24T13:28:47Z

pwndbg/gdblib/stack.py

@@ -136,3 +136,27 @@ def is_executable() -> bool:
 nx = True

 return not nx
+
+
+def find_addr_on_stack(addr):


this method is basically copied from here https://github.com/pwndbg/pwndbg/blob/dev/pwndbg/commands/stack.py#L26

i wonder if both should use this centralised approach?

Yeah I think it makes sense to centralize it. Also I believe there's a small edge case bug, it should be:

-while addresses and start < sp < stop: +while addresses and start <= sp < stop:

i'll make another issue to centralise that in another change i think ( and i'll fix that edge case and note it in the centralisation issue )

dmur1 · 2023-05-24T13:29:29Z

tests/gdb-tests/tests/test_context_commands.py

@@ -201,21 +201,19 @@ def test_context_backtrace_show_proper_symbol_names(start_binary):
 == "─────────────────────────────────[ BACKTRACE ]──────────────────────────────────"
 )

- assert re.match(r".*0 0x[0-9a-f]+ A::foo\(int, int\)", backtrace[2])


how important is it to tightly fit this regex to the entire output in a test that is mainly of proper symbol names?

The more accurate we are the less regressions we will have :). I am fine with more relaxed assertions: its better to have some tests than none and its hard to test for everything.

codecov-commenter · 2023-06-17T09:39:27Z

Codecov Report

Merging #1740 (5f7190a) into dev (9e85d0d) will decrease coverage by 0.12%.
Report is 57 commits behind head on dev.
The diff coverage is 69.81%.

❗ Your organization is not using the GitHub App Integration. As a result you may experience degraded service beginning May 15th. Please install the Github App Integration for your organization. Read more.

@@            Coverage Diff             @@
##              dev    #1740      +/-   ##
==========================================
- Coverage   60.61%   60.50%   -0.12%     
==========================================
  Files         175      179       +4     
  Lines       21300    22024     +724     
  Branches     1909     2029     +120     
==========================================
+ Hits        12912    13326     +414     
- Misses       7747     8025     +278     
- Partials      641      673      +32

Files Changed	Coverage Δ
pwndbg/ui.py	`56.06% <23.07%> (-18.94%)`	⬇️
pwndbg/gdblib/kernel/__init__.py	`79.14% <68.57%> (-1.47%)`	⬇️
pwndbg/commands/context.py	`62.83% <100.00%> (-3.91%)`	⬇️
pwndbg/commands/stack.py	`100.00% <100.00%> (ø)`
pwndbg/gdblib/stack.py	`80.48% <100.00%> (+8.02%)`	⬆️

... and 173 files with indirect coverage changes

gsingh93

Nice work overall, just needs a few minor changes.

pwndbg/commands/context.py

pwndbg/commands/stack.py

pwndbg/gdblib/stack.py

pwndbg/commands/context.py

disconnect3d · 2023-06-26T22:17:35Z

@dmur1 can u show a screenshot before and after?

dmur1

take a look at the newest stuff!

dmur1 · 2023-07-27T09:20:42Z

@dmur1 can u show a screenshot before and after?

before:

after:

disconnect3d · 2023-08-02T14:29:21Z

pwndbg/gdblib/stack.py

@@ -136,3 +136,27 @@ def is_executable() -> bool:
 nx = True

 return not nx
+
+
+def find_addr_on_stack(addr) -> int:


I am a bit worried that this may be (way) too slow.

We are reading the whole stack memory: what if stack is super big?

What if SP does not point to mapped memory? This will crash

This can be optimized out in two ways:

We could leverage the existing search functionality we have which will do a find using GDB's API which is written in C and so should be faster

We could read the whole stack memory and use simple .find(...) to find the offset of the bytes of the addr value

The 2) will rather be slower than 1) but if we read the stack memory just once, then we can find the fetched bytes multiple times to look for different return addresses --- which would probably be faster overall?

Given this I am not sure if we really want to look for those values on the stack... we are making context slower this way. We can however make this configurable.

@gsingh93 thoughts?

disconnect3d · 2023-08-02T14:30:47Z

pwndbg/commands/context.py

- symbol = c.symbol(pwndbg.gdblib.symbol.get(int(frame.pc())))
+ frame_pc = frame.pc()
+ frame_pc_on_stack = pwndbg.gdblib.stack.find_addr_on_stack(frame_pc)
+ padded_text = "0x" + hex(frame_pc_on_stack)[2:].zfill(pwndbg.gdblib.arch.ptrsize * 2)


hex(...) returns a 0x... string, so we don't need to [2:] it and prepend a 0x string

Also nitpick, but we fetch and compute pwndbg.gdblib.arch.ptrsize * 2 in each loop iteration :<

hex(1234).zfill(8)
'0000x4d2'

i was trying to avoid this - perhaps theres a way to zfill an already 0x prefixed string?

@disconnect3d

will hoist the computation out of the loop

I'm not sure about this zfill. Like, do we really want to display those zeroes? :P They are redundant

disconnect3d · 2023-08-02T14:30:59Z

pwndbg/commands/context.py

+ frame_pc_on_stack = pwndbg.gdblib.stack.find_addr_on_stack(frame_pc)
+ padded_text = "0x" + hex(frame_pc_on_stack)[2:].zfill(pwndbg.gdblib.arch.ptrsize * 2)
+ stackaddr = M.get(frame_pc_on_stack, padded_text)
+ addrsz = "(" + stackaddr + ") " + M.get(frame_pc)


Can we use f-string literals here and wherever we can?

disconnect3d · 2023-08-02T14:32:53Z

@dmur1 can u show a screenshot before and after?

before:

after:

Random thoughts about the display:

I don't like the 0x<zeros> prepended
Maybe we should display <sp register>+0x<OFFSET> instead of whole addresses?

...or maybe this should be configurable, ugh. cc: @gsingh93

dmur1 · 2023-08-03T13:04:40Z

@dmur1 can u show a screenshot before and after?

before:
after:

Random thoughts about the display:

I don't like the 0x<zeros> prepended

Maybe we should display <sp register>+0x<OFFSET> instead of whole addresses?

...or maybe this should be configurable, ugh. cc: @gsingh93

i have to admit i'm also not a fan especially as it contrasts with the other aligned addresses in the other context sections

dmur1 · 2024-02-22T18:10:10Z

bumping this - i'm still not sure i like it what do you folks think?

disconnect3d · 2024-02-22T18:21:53Z

I think prefixing with additional zeroes is not great. What else do u ask about? :P

dmur1 · 2024-02-22T18:30:23Z

I think prefixing with additional zeroes is not great. What else do u ask about? :P

so maybe if just padded as much as was needed to keep everything aligned?

dmur1 marked this pull request as draft May 24, 2023 13:18

dmur1 commented May 24, 2023

View reviewed changes

dan murray and others added 2 commits June 17, 2023 10:25

fix edge case bug comparising sp to extent of stack

2ef3ac9

Merge branch 'dev' into stack-location-context-backtrace-1288

dcfc0b1

dmur1 marked this pull request as ready for review June 17, 2023 09:29

dmur1 mentioned this pull request Jun 17, 2023

use find_addr_on_stack for retaddr #1762

Open

fix edge case bug comparing sp to extent of stack

e5fc361

gsingh93 requested changes Jun 26, 2023

View reviewed changes

gsingh93 modified the milestone: 2023.10 Jul 17, 2023

smiley added 3 commits July 27, 2023 10:00

leave a format string as it was

84ba972

add some type hints to find_addr_on_stack

0a7355a

use an already padded hex string to get aligned coloured addresses

14a3373

dmur1 commented Jul 27, 2023

View reviewed changes

dmur1 requested review from gsingh93 and disconnect3d July 27, 2023 20:01

disconnect3d reviewed Aug 2, 2023

View reviewed changes

smiley added 2 commits August 3, 2023 14:06

use an f string to compute the value of addrsz

aa7c121

hoist computation of zero fill amount out of the loop

5f7190a

dmur1 requested a review from disconnect3d August 3, 2023 13:09

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add return address stack location to context backtrace #1288 #1740

Add return address stack location to context backtrace #1288 #1740

dmur1 commented May 24, 2023

dmur1 May 24, 2023

disconnect3d May 25, 2023 •

edited

dmur1 Jun 17, 2023

dmur1 May 24, 2023

disconnect3d May 25, 2023

codecov-commenter commented Jun 17, 2023 •

edited

gsingh93 left a comment

disconnect3d commented Jun 26, 2023

dmur1 left a comment

dmur1 commented Jul 27, 2023

disconnect3d Aug 2, 2023

disconnect3d Aug 2, 2023

disconnect3d Aug 2, 2023

dmur1 Aug 3, 2023

dmur1 Aug 3, 2023

disconnect3d Aug 6, 2023

disconnect3d Aug 2, 2023

disconnect3d commented Aug 2, 2023

dmur1 commented Aug 3, 2023

dmur1 commented Feb 22, 2024

disconnect3d commented Feb 22, 2024

dmur1 commented Feb 22, 2024

Add return address stack location to context backtrace #1288 #1740

Are you sure you want to change the base?

Add return address stack location to context backtrace #1288 #1740

Conversation

dmur1 commented May 24, 2023

Choose a reason for hiding this comment

disconnect3d May 25, 2023 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

codecov-commenter commented Jun 17, 2023 • edited

Codecov Report

gsingh93 left a comment

Choose a reason for hiding this comment

disconnect3d commented Jun 26, 2023

dmur1 left a comment

Choose a reason for hiding this comment

dmur1 commented Jul 27, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

disconnect3d commented Aug 2, 2023

dmur1 commented Aug 3, 2023

dmur1 commented Feb 22, 2024

disconnect3d commented Feb 22, 2024

dmur1 commented Feb 22, 2024

disconnect3d May 25, 2023 •

edited

codecov-commenter commented Jun 17, 2023 •

edited