Further tuning of DTrace stack unwinding on ARMv[67]
AcceptedPublic
Actions

Authored by rwatson on Oct 4 2015, 11:21 AM.

Details

Reviewers

andrew
gnn
bz
br

Summary

A variety of minor tweaks to DTrace stack unwinding on ARMv[67], including:

Be more robust to invalid stack state, such as potentially invalid PC values
Better adjust number of frames skipped in various situations (e.g., SDT, FBT, etc)

Test Plan

Deployed in Cambridge L41 2014-2015.

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Skipped

Unit

Tests Skipped

Event Timeline

rwatson updated this revision to Diff 9125.Oct 4 2015, 11:21 AM

rwatson retitled this revision from to Further tuning of DTrace stack unwinding on ARMv[67].

rwatson updated this object.

rwatson edited the test plan for this revision. (Show Details)

rwatson added reviewers: andrew, br, bz.

rwatson set the repository for this revision to rS FreeBSD src repository - subversion.

rwatson added a project: ARM.

Herald added a reviewer: gnn. · View Herald TranscriptOct 4 2015, 11:21 AM

Herald added a subscriber: imp. · View Herald Transcript

Updated diff from svn contains further context; no functional change.

Herald edited edge metadata. · View Herald TranscriptOct 4 2015, 11:48 AM

gnn accepted this revision.Oct 5 2015, 2:40 PM

gnn edited edge metadata.

This revision is now accepted and ready to land.Oct 5 2015, 2:40 PM

Is the change to the number of frames to unwind correct? The change to move the call to dtrace_invop_jump_addr to asm doesn't seem to have made it into HEAD.

Should be fine for SDT, but I've not tried FBT, and as you point out, stack layout may differ a bit. I will try to give that a test run in isolation from the unupstreamed dtrace_invop_jump_addr change. We really do need to upstream the latter -- is that in your court?