This is used to enable the guard page when an elf binary is built with
BTI instructions.
Sponsored by: Arm Ltd
Sponsored by: The FreeBSD Foundation
Details
Details
- Reviewers
kib markj manu - Group Reviewers
arm64 - Commits
- rGc4e4a7596beb: Add a BTI sysarch
Diff Detail
Diff Detail
- Repository
- rG FreeBSD src repository
- Lint
Lint Passed - Unit
No Test Coverage - Build Status
Buildable 50798 Build 47689: arc lint + arc unit
Event Timeline
Comment Actions
Where will BTI be documented? Perhaps bti.3 and mitigations.7?
| sys/arm64/arm64/sys_machdep.c | ||
|---|---|---|
| 63 | ||
| 89 | IMO it is clearer to write this in terms of a range: sva = trunc_page(gp_args.addr);
eva = round_page(gp_args.addr + gp_args.len);
if (sva > eva || eva > VM_MAX_USER_ADDRESS)
return (EINVAL);
error = pmap_pti_set(pmap, sva, eva); | |
| 91 | Shouldn't the pmap should be derived from td, not using pcpu? This looks strange, I don't know why PKRU does it this way. | |
| 92 | pmap_bti_set() returns 0 if the executable didn't have BTI configured. Don't we want to return an error here in that case? | |
| sys/arm64/include/sysarch.h | ||
| 39 | What is this constant for? | |
Comment Actions
Updates based on feedback from @markj.
Only page align the start and end address when calling pmap_bti_set to ensure we catch address space overflows
| sys/arm64/arm64/sys_machdep.c | ||
|---|---|---|
| 68 | All of the other error cases below are handled by returning directly. | |