arm64: don't pass user trapframe to kdb_trap()
Needs ReviewPublic
Actions

Authored by mhorne on Dec 22 2020, 2:30 PM.

Details

Reviewers

andrew
jhb
markj
manu

Commits

rGe9bb4ce3d0e7: arm64: don't pass user trapframe to kdb_trap()

Summary

This effectively undoes the changes made in r321571. While useful, it is
inconsistent with how other architectures handle trapframes. This change
is required to get a working gdb(4) stub on arm64, as otherwise the
backtrace will begin one frame too early.

With D27705, this information can still be obtained via show registers/u.

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 35957
Build 32846: arc lint + arc unit

Event Timeline

mhorne created this revision.Dec 22 2020, 2:30 PM

Herald added a reviewer: manu. · View Herald TranscriptDec 22 2020, 2:30 PM

Herald added subscribers: emaste, imp. · View Herald Transcript

Harbormaster completed remote builds in B35608: Diff 81048.Dec 22 2020, 2:30 PM

mhorne requested review of this revision.Dec 22 2020, 2:30 PM

mhorne added a child revision: D27727: arm64: gdb(4) stub.Dec 22 2020, 6:07 PM

jhb accepted this revision.Dec 30 2020, 9:23 PM

This revision is now accepted and ready to land.Dec 30 2020, 9:23 PM

andrew added inline comments.Jan 5 2021, 3:13 PM

sys/arm64/arm64/trap.c
422–423	You need to remove this so the next instruction is executed

Remove the tf_elr manipulation.

This revision now requires review to proceed.Jan 5 2021, 9:39 PM

Harbormaster completed remote builds in B35957: Diff 81698.Jan 5 2021, 9:39 PM

mhorne mentioned this in rGe9bb4ce3d0e7: arm64: don't pass user trapframe to kdb_trap().Jan 8 2021, 6:54 PM

jrtc27 added a subscriber: jrtc27.Jan 8 2021, 7:01 PM

jrtc27 added inline comments.

sys/arm64/arm64/trap.c
422–423	Is it not now going to just return and immediately re-trap on the same breakpoint instruction?

mhorne added inline comments.Jan 8 2021, 7:16 PM

sys/arm64/arm64/trap.c
422–423	For ddb, `tf_elr` is advanced by `BKPT_SKIP` in `db_stop_at_pc`. For gdb, the client advances the PC. Previously this would have been applied to `td->td_frame`, not `frame`, hence the need for the manual increment here.