Use frame pointer unwinding for v8-based applications #1836
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
We can't unwind nodejs and other v8-base applications as we can't continue unwinding after the first AoT frame: ``` [...] current pc: 7fb0cf84904a current sp: 7fff93701498 current bp: 7fff937014d8 [debug] Unwinding JITed stacks [debug] Within unwinding machinery loop current pc: 157a329 current sp: 7fff937014e8 current bp: 7fff93701548 ~about to check shards found=1 ~checking shards now [info] found chunk le offset: 0 ========== left 0 right 249997 .done => table_index: 206094 => adjusted pc: 157a329 cfa type: 4, offset: 0 (row pc: 156f656) [info] PC 157a329 not contained in the unwind info, found marker [warn] mapping not added yet rbp 7fff93701548 pid 227936 tgid 227936 ``` Turns out this AoT function doesn't have the corresponding DWARF unwind information: ``` $ dist/eh-frame --executable /home/javierhonduco/.nvm/versions/node/v16.13.0/bin/node | grep 157a3 $ ``` After cross-validating the output with `bpftrace`'s and checking v8's source code it was clear why this is happening: ``` 157a32a Builtins_InterpreterEntryTrampoline+202 (/home/javierhonduco/.nvm/versions/node/v16.13.0/bin/node) 157a32a Builtins_InterpreterEntryTrampoline+202 (/home/javierhonduco/.nvm/versions/node/v16.13.0/bin/node) 157a32a Builtins_InterpreterEntryTrampoline+202 (/home/javierhonduco/.nvm/versions/node/v16.13.0/bin/node) 1578538 Builtins_JSEntryTrampoline+88 (/home/javierhonduco/.nvm/versions/node/v16.13.0/bin/node) 15782c3 Builtins_JSEntry+131 (/home/javierhonduco/.nvm/versions/node/v16.13.0/bin/node) [...] 7f65c5e29510 __libc_start_call_main+128 (/usr/lib64/libc.so.6) ``` The `Builtins_` functions above seem to be emmited via a custom code generator[0]. No special linker `.cfi` directives are used, so this is effectively like writting raw assembly with frame pointers set up, but without telling the rest of the tooling how to produce unwidn information it. This commit includes other unrelated changes that have been made while debugging this issue which hopefully improve the clarity and reliability of the code. Test Plan ========= - node: https://pprof.me/95f85d0 - basic-cpp-no-fp: https://pprof.me/a6099f2 - ruby: https://pprof.me/4ecf047 - [0]: https://github.com/v8/v8/blob/e554f1e97e30f6e4cf6eef2c5049c17ca708e103/src/builtins/x64/builtins-x64.cc Signed-off-by: Francisco Javier Honduvilla Coto <javierhonduco@gmail.com>
Sylfrena
reviewed
Jul 6, 2023
| return 0; | ||
| } | ||
|
|
||
| // 3. Request unwind information. | ||
| request_unwind_information(ctx, user_pid); |
There was a problem hiding this comment.
Thanks for all the numbered comments here!
There was a problem hiding this comment.
:D hopefully things will be clearer and this will also prevent accidental reorders!
javierhonduco
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We can't unwind nodejs and other v8-base applications as we can't continue unwinding after the first AoT frame:
Turns out this AoT function doesn't have the corresponding DWARF unwind information:
After cross-validating the output with
bpftrace's and checking v8's source code it was clear why this is happening:The
Builtins_functions above seem to be emmited via a custom code generator0. No special linker.cfidirectives are used, so this is effectively like writting raw assembly with frame pointers set up, but without telling the rest of the tooling how to produce unwidn information it.Test Plan
node: https://pprof.me/95f85d0
basic-cpp-no-fp: https://pprof.me/a6099f2
ruby: https://pprof.me/4ecf047