Page MenuHomeFreeBSD

arm64: log vm_fault error for data_abort
ClosedPublic

Authored by bz on May 20 2021, 4:58 PM.
Tags
None
Referenced Files
F106105962: D30362.diff
Wed, Dec 25, 12:11 PM
Unknown Object (File)
Thu, Dec 5, 9:42 PM
Unknown Object (File)
Oct 11 2024, 4:01 AM
Unknown Object (File)
Oct 6 2024, 8:43 AM
Unknown Object (File)
Oct 5 2024, 4:55 AM
Unknown Object (File)
Oct 2 2024, 2:39 PM
Unknown Object (File)
Oct 2 2024, 2:07 PM
Unknown Object (File)
Oct 1 2024, 1:22 AM
Subscribers

Details

Summary

Log the vm_fault() error in the data_abort panic so it is easier to
find the reason vm_fault() failed (e.g., invalid address).

Reviewed by: ..
MFC After: 2 weeks
Differential Revision: ...

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

bz requested review of this revision.May 20 2021, 4:58 PM

We could drop printing frame->tf_elr as we already do so in print_registers

This revision is now accepted and ready to land.May 24 2021, 8:31 AM

We could drip printing frame->tf_elr as we already do so in print_registers

That brings me to a different question: print_registers and a "show reg" in ddb right after the panic result in two different outputs. Is that intentional?

They are printing different trap frames. print_registers will print the fault frame while show reg will print a frame from a breakpoint within panic. The registers seen when debugger_on_trap is true should be identical to the version printed by print_registers.

They are printing different trap frames. print_registers will print the fault frame while show reg will print a frame from a breakpoint within panic. The registers seen when debugger_on_trap is true should be identical to the version printed by print_registers.

Thank you; Let's keep the ELR in the panicstr then.

This revision was automatically updated to reflect the committed changes.