Start to clean up arm64 address space selection
ClosedPublic
Actions

Authored by andrew on Jul 14 2021, 3:20 PM.

Details

Reviewers

alc
kib
markj
manu

Group Reviewers

arm64

Commits

rGe20f0db4a94a: Start to clean up arm64 address space selection
rGb7a78d573a56: Start to clean up arm64 address space selection

Summary

On arm64 we should use bit 55 of the address to decide if aan address
is a user or kernel address. Add a new macro with this check and a
second to ensure the address is in teh canonical form, i.e.
the top bits are all zero or all one.

This will help with supporting future cpu features, including Top
Byte Ignore, Pointer Authentication, and Memory Tagging.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

andrew created this revision.Jul 14 2021, 3:20 PM

Herald added a reviewer: manu. · View Herald TranscriptJul 14 2021, 3:20 PM

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

andrew requested review of this revision.Jul 14 2021, 3:20 PM

Harbormaster completed remote builds in B40472: Diff 92189.Jul 14 2021, 3:20 PM

Typo, "Tob Byte Ignore"

andrew edited the summary of this revision. (Show Details)Jul 14 2021, 4:07 PM

Set an error in data_abort as we print it in the kernel case

Harbormaster completed remote builds in B40473: Diff 92191.Jul 14 2021, 4:18 PM

alc added inline comments.Jul 14 2021, 11:16 PM

sys/arm64/arm64/pmap.c
475–476	I would suggest moving this KASSERT() before the above comment. Otherwise, this KASSERT() sits between the comment and the KASSERT() immediately below that the comment explains.
509–510	Ditto.
sys/arm64/include/vmparam.h
161	This "it's" should be "its".

Update based on comments from alc

Harbormaster completed remote builds in B40482: Diff 92214.Jul 15 2021, 8:56 AM

kib accepted this revision.Jul 15 2021, 10:24 AM

This revision is now accepted and ready to land.Jul 15 2021, 10:24 AM

On arm64 we should use bit 55 of the address to decide if aan address is a user or kernel address.

Could you explain a bit further? Where does that number come from?

In current use he hardware will use bits 63:48 in the virtual address to decide which ttbr register to use, however I'm adding pointer authentication support. In this case the hardware will change bits 63:56 and 54:48. As it keeps bit 55 we will need to use this to decide if an address is a user or kernel address when faulting.

It is similar with top byte ignore, however bits 63:56 of the virtual address are ignored when performing the translation.

mhorne mentioned this in D31208: riscv: handle page faults in the unmappable region.Jul 17 2021, 8:42 PM

Closed by commit rGb7a78d573a56: Start to clean up arm64 address space selection (authored by andrew). · Explain WhyJul 20 2021, 11:21 AM

This revision was automatically updated to reflect the committed changes.

andrew added a commit: rGb7a78d573a56: Start to clean up arm64 address space selection.

andrew added a commit: rGe20f0db4a94a: Start to clean up arm64 address space selection.Aug 30 2021, 12:44 PM