Navigation silently jumps to one of the basic blocks when there are multiple basic blocks containing the requested address

[bb33bb]

Version and Platform (required):

• Binary Ninja Version: 4.2.6444-dev Personal (f171d6c4)
• OS: > cat /proc/version
  Linux version 6.10.13-3-MANJARO (builduser@fv-az1246-770) (gcc (GCC) 14.2.1 20240910, GNU ld (GNU Binutils) 2.43.0) use Caches folder on OS X for download cache to prevent backups #1 SMP PREEMPT_DYNAMIC Tue Oct 8 03:24:49 UTC 2024
• CPU Architecture: x64

Bug Description:
use the short key of g, we jump to the address of 0xFFFFFFC008626658 in the mode of c source code, we can not jmp to the right address sometimes.
Steps To Reproduce:

simplescreenrecorder-2024-11-16_19.24.40.mp4

Binary:
maybe you can see the right address in assembly code, but if we are in c code mode, its not the correct address.

[bb33bb]

i will upload these files, plz give me a second

[bb33bb]

https://1drv.ms/f/c/01e018a652fc6a6e/Eh4HRs_pU99Gpx1tG4jrXYcB3DM8KHbGY_vvxJLB8VQi_w?e=bzYvhP

[xusheng6]

If you check the disassembly code at 0xffffffc008626658, you will see it belongs to two different functions:

Function do_coredump and sub_ffffffc008626b10 both contain the requested address, and the latter one happens to be the most recently used one, so the linear view decides to jump to it rather than do_coredump. That also explains why you think it "sometimes" fails -- because if do_coredump is the most recent function, the linear view will navigate to that, and you will think it works as expected.

I checked the original binary you provided, and our analysis did not create sub_ffffffc008626b10. Is it created by you manually?

While the above explanation gives a reason for what you are seeing, I am keeping this issue open for internal discussion to see if we can have a better way to let people know there are overlapping basic blocks and have them confirm their intention

[xusheng6]

Similar to #5978

[bb33bb]

If you check the disassembly code at 0xffffffc008626658, you will see it belongs to two different functions:

Function do_coredump and sub_ffffffc008626b10 both contain the requested address, and the latter one happens to be the most recently used one, so the linear view decides to jump to it rather than do_coredump. That also explains why you think it "sometimes" fails -- because if do_coredump is the most recent function, the linear view will navigate to that, and you will think it works as expected.

I checked the original binary you provided, and our analysis did not create sub_ffffffc008626b10. Is it created by you manually?

While the above explanation gives a reason for what you are seeing, I am keeping this issue open for internal discussion to see if we can have a better way to let people know there are overlapping basic blocks and have them confirm their intention

yes and it's true that this function is really huge.
And there are really some difference between ida and ninja.
I have to use both of these 2 software.

[plafosse]

Proposed solution. Before navigating to code we check if that address exists in multiple basic blocks. If it does we do a follow up prompt which asks them which function they want to navigate to. The second popup should contain a searchable combo box with all the possibilities. Perhaps the one that is selected is the most recently visited function.