Improve handling of newly created threads in initial seize code #255
Conversation
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #255 +/- ##
==========================================
- Coverage 66.99% 66.81% -0.19%
==========================================
Files 99 99
Lines 10321 10348 +27
Branches 1709 1718 +9
==========================================
- Hits 6915 6914 -1
- Misses 2523 2549 +26
- Partials 883 885 +2 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
| if (ptrace(PTRACE_SINGLESTEP, pid, 0, 0) == -1) { | ||
| LOG(ERROR) << "Error in singlestep!"; | ||
| LOG(ERROR) << "singleStepInst: ptrace single error: " << strerror(errno); |
There was a problem hiding this comment.
nit: this can be replaced with PLOG(ERROR) << "singleStepInst: ptrace single error: "; similarly with the other cases that append strerror.
| @@ -64,7 +64,7 @@ class OIDebugger { | |||
| OIDebugger::processTrapRet processTrap(pid_t, bool = true, bool = true); | |||
| bool contTargetThread(bool detach = true) const; | |||
| bool isGlobalDataProbeEnabled(void) const; | |||
| static uint64_t singlestepInst(pid_t, struct user_regs_struct&); | |||
| static uint64_t singleStepInst(pid_t, struct user_regs_struct&); | |||
There was a problem hiding this comment.
nit: should this return uintptr_t? it's casted to that at one of the call sites and seems to make more sense. obviously they're the same type though.
| @@ -968,7 +980,7 @@ OIDebugger::processTrapRet OIDebugger::processTrap(pid_t pid, | |||
| siginfo_t info; | |||
|
|
|||
| auto stopsig = WSTOPSIG(status); | |||
| VLOG(4) << "Stop sig: " << std::dec << stopsig; | |||
| VLOG(3) << "Stop sig: " << std::dec << stopsig; | |||
There was a problem hiding this comment.
not sure if this was for debugging, but shouldn't this remain a 4?
|
Hi @tyroguru! Thank you for your pull request. We require contributors to sign our Contributor License Agreement, and yours needs attention. You currently have a record in our system, but the CLA is no longer valid, and will need to be resubmitted. ProcessIn order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA. Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with If you have received this in error or have any questions, please contact us at [email protected]. Thanks! |
Summary
Our initial thread seize code is very racy and it's super easy to end up with threads being ignored or hung for long periods waiting for stop states to be reaped. This change aims to handle the case where a thread that has been seized executes a sequence such that a new thread is created (e.g., a clone(2) syscall and the newborn thread shows up in the /proc namespace so the directory_iterator picks it up. This new thread will be traced by us already as the tracing sate is inherited from its parent thread so we can't seize it. It will however be in a signal-stop state so that should be reaped and the thread continued.
Test plan
make 'test' returns the same failures as before the change. However, it is unlikely to show any change as this code is handling an
EPERMreturn fromptrace(2). Test in prod really. I have engineered highly controlled test cases to work through the failure scenarios and at a minimum I don't think we are any more exposed to failure with this code than without.