diff options
| author | Jamie Liu <jamieliu@google.com> | 2021-10-08 18:00:36 -0700 |
|---|---|---|
| committer | gVisor bot <gvisor-bot@google.com> | 2021-10-08 18:03:23 -0700 |
| commit | 3f1642e4bc86c6e7febc6b2dd2b83ad48c5ee201 (patch) | |
| tree | d7b6428e36cda75eb7a8702d22923cc13c22f0cc /pkg/ring0/lib_amd64.s | |
| parent | 34e68b6b4ff04de5c7b5e2dc46e5bd44c6845e63 (diff) | |
Remove ring0 floating point save/load functions on amd64.
ring0.Save/LoadFloatingPoint() are only usable if the caller can ensure that Go
will not clobber floating point registers before/after calling them
respectively. Due to regabig in Go 1.17, this is no longer the case; regabig
(among other things) maintains a zeroed XMM15 during ABIInternal execution,
including by zeroing it after ABI0-to-ABIInternal transitions. In
ring0.sysenter/exception, this happens in
ring0.kernelSyscall/kernelException.abi0 respectively; in
ring0.CPU.SwitchToUser, this happens after returning from
ring0.sysret/iret.abi0. Delete these functions and do floating point save/load
in assembly.
While arm64 doesn't appear to be immediately affected (so this CL permits us to
resume usage of Go 1.17), its use of Save/LoadFloatingPoint() still seems to be
incorrect for the same fundamental reason (Go code can't sanely assume what
registers the Go compiler will or won't use) and should be fixed eventually.
PiperOrigin-RevId: 401895658
Diffstat (limited to 'pkg/ring0/lib_amd64.s')
| -rw-r--r-- | pkg/ring0/lib_amd64.s | 23 |
1 files changed, 23 insertions, 0 deletions
diff --git a/pkg/ring0/lib_amd64.s b/pkg/ring0/lib_amd64.s index 8ed98fc84..0f283aaae 100644 --- a/pkg/ring0/lib_amd64.s +++ b/pkg/ring0/lib_amd64.s @@ -128,6 +128,29 @@ TEXT ·wrfsmsr(SB),NOSPLIT,$0-8 BYTE $0x0f; BYTE $0x30; RET +// writeGS writes to the GS base. +// +// This is written in assembly because it must be callable from assembly (ABI0) +// without an intermediate transition to ABIInternal. +// +// Preconditions: must be running in the lower address space, as it accesses +// global data. +TEXT ·writeGS(SB),NOSPLIT,$8-8 + MOVQ addr+0(FP), AX + + CMPB ·hasFSGSBASE(SB), $1 + JNE msr + + PUSHQ AX + CALL ·wrgsbase(SB) + POPQ AX + RET +msr: + PUSHQ AX + CALL ·wrgsmsr(SB) + POPQ AX + RET + // wrgsbase writes to the GS base. // // The code corresponds to: |