[lldb-dev] Disassembling issue

Ted Woodward via lldb-dev lldb-dev at lists.llvm.org
Thu Jun 13 09:38:50 PDT 2019


Hi Romaric,

The call to Disassembler::FindPlugin should have arch set to your triple. If m_options.arch is not set on the disassembly command line, it will get it from the target:

  if (!m_options.arch.IsValid())
    m_options.arch = target->GetArchitecture();


That leads me to wonder if your triple is set correctly after you do the run. What is the output from these commands, when you’re stopped?:

target list
platform status

My guess is you’re connected to a remote gdb-server on a simulator for dpu, and it’s telling lldb that the triple is x86_64.x.x.

If target list says the arch is dpu before the run, and x86_64 after the run, we should look at the RSP log. Before the run, type
log enable -f <file> gdb-remote packets

Attach the file to your next message.

Ted

From: Romaric Jodin <rjodin at upmem.com>
Sent: Thursday, June 13, 2019 1:16 AM
To: Ted Woodward <tedwood at quicinc.com>
Cc: LLDB <lldb-dev at lists.llvm.org>
Subject: [EXT] Re: [lldb-dev] Disassembling issue

Hi Ted,

Thank you for the answer.
I'm not sure I understand you well. It seems to me that my program call "Disassembler::FindPlugin" with the "m_options.arch" set to "x86_64". But my toolchain does not contain a LLVM supporting x86_64, so it feels normal to me that it is not able to create this disassembler. What is weird is that the first time I'm executing it, it managed to do so.

I also have kind of the same issue with breakpoint. It's like if I'm missing the debug information after I do the "run" command:

(lldb) file test
Current executable set to 'test' (dpu).
(lldb) b main
Breakpoint 1: where = test`main + 24 at test.c:8, address = 0x80000078
(lldb) r
Process 21312 launched: '/home/rjodin/work/dpu_tools/llvm/lldb/test/test' (dpu)
Process 21312 stopped
* thread #1, name = 'DPUthread0', stop reason = Thread finished
    frame #0: 0x00000000800000c0
(lldb) b main
Breakpoint 2: no locations (pending).
WARNING:  Unable to resolve breakpoint to any actual locations.

When I'm doing the disassembly, I can manage to disassemble if I add the option with my architecture to the command line, but as you can see below, I also do not have the debug information that I had at the first disassembly:

(lldb) file test
Current executable set to 'test' (dpu).
(lldb) dis -s 0x80000000
test`__bootstrap:
test[0x80000000] <+0>:  jnz    id, 0x80000020
test[0x80000008] <+8>:  move   r0, 0xff
test[0x80000010] <+16>: release r0, 0x0, nz, 0x80000018
test[0x80000018] <+24>: sub    r0, r0, 0x1, pl, 0x80000010
(lldb) r
Process 21312 launched: '/home/rjodin/work/dpu_tools/llvm/lldb/test/test' (dpu)
Process 21312 stopped
* thread #1, name = 'DPUthread0', stop reason = Thread finished
    frame #0: 0x00000000800000a0
(lldb) dis -s 0x80000000
error: Unable to find Disassembler plug-in for the 'x86_64' architecture.
(lldb) dis -s 0x80000000 -A dpu
    0x80000000: jnz    id, 0x80000020
    0x80000008: move   r0, 0xff
    0x80000010: release r0, 0x0, nz, 0x80000018
    0x80000018: sub    r0, r0, 0x1, pl, 0x80000010

Any ideas about it?
Thanks,
Romaric

On Wed, Jun 12, 2019 at 6:55 PM Ted Woodward <tedwood at quicinc.com<mailto:tedwood at quicinc.com>> wrote:
The error is coming from source/Commands/CommandObjectDisassemble.cpp . It’s not able to create the disassembler for your architecture. The call to Disassembler::FindPlugin fails.
My guess is something in the call to MCDisasmInstance::Create in source/Plugins/Disassembler/llvm/DisassemblerLLVMC.cpp is failing. Step through that and make sure everything there is working.

Ted

From: lldb-dev <lldb-dev-bounces at lists.llvm.org<mailto:lldb-dev-bounces at lists.llvm.org>> On Behalf Of Romaric Jodin via lldb-dev
Sent: Wednesday, June 12, 2019 8:23 AM
To: lldb-dev at lists.llvm.org<mailto:lldb-dev at lists.llvm.org>
Subject: [EXT] [lldb-dev] Disassembling issue

Hi everyone,

I'm trying to have lldb being able to debug my architecture. It does it well after loading my elf. But then I run it, stop it in middle, try to disassemble and at that point it seems that I lost some information about my architecture. lldb tries to disassemble it for x86, which in not possible with my toolchain.

Here is a log of what I am doing:

(lldb) file test
Current executable set to 'test' (dpu).
(lldb) dis -s 0x80000000
test`__bootstrap:
test[0x80000000] <+0>:  jnz    id, 0x80000020
test[0x80000008] <+8>:  move   r0, 0xff
test[0x80000010] <+16>: release r0, 0x0, nz, 0x80000018
test[0x80000018] <+24>: sub    r0, r0, 0x1, pl, 0x80000010
(lldb) r
Process 21312 launched: '/home/rjodin/work/dpu_tools/llvm/lldb/test/test' (dpu)
Process 21312 stopped
* thread #1, name = 'DPUthread0', stop reason = Thread stopped
    frame #0: 0x0000000080000090
(lldb) dis -s 0x80000000
error: Unable to find Disassembler plug-in for the 'x86_64' architecture.
(lldb)

Any ideas of what I am missing in my implementation?
Thanks,
--
Romaric JODIN
UPMEM
Software Engineer

[cid:image001.png at 01D521DB.DDFD0A00]


--
Romaric JODIN
UPMEM
Software Engineer

[cid:image001.png at 01D521DB.DDFD0A00]
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/lldb-dev/attachments/20190613/fb2b2c68/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 1939 bytes
Desc: image001.png
URL: <http://lists.llvm.org/pipermail/lldb-dev/attachments/20190613/fb2b2c68/attachment-0001.png>


More information about the lldb-dev mailing list