diff options
| author | Keith Randall <khr@golang.org> | 2025-05-17 15:05:56 -0700 |
|---|---|---|
| committer | Keith Randall <khr@golang.org> | 2025-10-06 14:11:41 -0700 |
| commit | 719dfcf8a8478d70360bf3c34c0e920be7b32994 (patch) | |
| tree | d58aaf3289de3bb18901e34b336da46b425f8075 /src/cmd/link/internal | |
| parent | f3312124c2370c2f64a7f9ad29732ec30209647a (diff) | |
| download | go-719dfcf8a8478d70360bf3c34c0e920be7b32994.tar.xz | |
cmd/compile: redo arm64 LR/FP save and restore
Instead of storing LR (the return address) at 0(SP) and the FP
(parent's frame pointer) at -8(SP), store them at framesize-8(SP)
and framesize-16(SP), respectively.
We push and pop data onto the stack such that we're never accessing
anything below SP.
The prolog/epilog lengths are unchanged (3 insns for a typical prolog,
2 for a typical epilog).
We use 8 bytes more per frame.
Typical prologue:
STP.W (FP, LR), -16(SP)
MOVD SP, FP
SUB $C, SP
Typical epilogue:
ADD $C, SP
LDP.P 16(SP), (FP, LR)
RET
The previous word where we stored LR, at 0(SP), is now unused.
We could repurpose that slot for storing a local variable.
The new prolog and epilog instructions are recognized by libunwind,
so pc-sampling tools like perf should now be accurate. (TODO: except
maybe after the first RET instruction? Have to look into that.)
Update #73753 (fixes, for arm64)
Update #57302 (Quim thinks this will help on that issue)
Change-Id: I4800036a9a9a08aaaf35d9f99de79a36cf37ebb8
Reviewed-on: https://go-review.googlesource.com/c/go/+/674615
Reviewed-by: David Chase <drchase@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>
Diffstat (limited to 'src/cmd/link/internal')
| -rw-r--r-- | src/cmd/link/internal/amd64/obj.go | 19 | ||||
| -rw-r--r-- | src/cmd/link/internal/arm64/obj.go | 23 | ||||
| -rw-r--r-- | src/cmd/link/internal/ld/dwarf.go | 7 | ||||
| -rw-r--r-- | src/cmd/link/internal/ld/lib.go | 4 | ||||
| -rw-r--r-- | src/cmd/link/internal/ld/stackcheck.go | 5 | ||||
| -rw-r--r-- | src/cmd/link/internal/x86/obj.go | 15 |
6 files changed, 40 insertions, 33 deletions
diff --git a/src/cmd/link/internal/amd64/obj.go b/src/cmd/link/internal/amd64/obj.go index 3a6141b909..761496549f 100644 --- a/src/cmd/link/internal/amd64/obj.go +++ b/src/cmd/link/internal/amd64/obj.go @@ -51,15 +51,16 @@ func Init() (*sys.Arch, ld.Arch) { Plan9Magic: uint32(4*26*26 + 7), Plan9_64Bit: true, - Adddynrel: adddynrel, - Archinit: archinit, - Archreloc: archreloc, - Archrelocvariant: archrelocvariant, - Gentext: gentext, - Machoreloc1: machoreloc1, - MachorelocSize: 8, - PEreloc1: pereloc1, - TLSIEtoLE: tlsIEtoLE, + Adddynrel: adddynrel, + Archinit: archinit, + Archreloc: archreloc, + Archrelocvariant: archrelocvariant, + Gentext: gentext, + Machoreloc1: machoreloc1, + MachorelocSize: 8, + PEreloc1: pereloc1, + TLSIEtoLE: tlsIEtoLE, + ReturnAddressAtTopOfFrame: true, ELF: ld.ELFArch{ Linuxdynld: "/lib64/ld-linux-x86-64.so.2", diff --git a/src/cmd/link/internal/arm64/obj.go b/src/cmd/link/internal/arm64/obj.go index 3d358155ba..e1e4ade818 100644 --- a/src/cmd/link/internal/arm64/obj.go +++ b/src/cmd/link/internal/arm64/obj.go @@ -47,17 +47,18 @@ func Init() (*sys.Arch, ld.Arch) { Dwarfreglr: dwarfRegLR, TrampLimit: 0x7c00000, // 26-bit signed offset * 4, leave room for PLT etc. - Adddynrel: adddynrel, - Archinit: archinit, - Archreloc: archreloc, - Archrelocvariant: archrelocvariant, - Extreloc: extreloc, - Gentext: gentext, - GenSymsLate: gensymlate, - Machoreloc1: machoreloc1, - MachorelocSize: 8, - PEreloc1: pereloc1, - Trampoline: trampoline, + Adddynrel: adddynrel, + Archinit: archinit, + Archreloc: archreloc, + Archrelocvariant: archrelocvariant, + Extreloc: extreloc, + Gentext: gentext, + GenSymsLate: gensymlate, + Machoreloc1: machoreloc1, + MachorelocSize: 8, + PEreloc1: pereloc1, + Trampoline: trampoline, + ReturnAddressAtTopOfFrame: true, ELF: ld.ELFArch{ Androiddynld: "/system/bin/linker64", diff --git a/src/cmd/link/internal/ld/dwarf.go b/src/cmd/link/internal/ld/dwarf.go index 0003938ef2..c4d12a5488 100644 --- a/src/cmd/link/internal/ld/dwarf.go +++ b/src/cmd/link/internal/ld/dwarf.go @@ -1544,9 +1544,14 @@ func (d *dwctxt) writeframes(fs loader.Sym) dwarfSecInfo { if pcsp.Value > 0 { // The return address is preserved at (CFA-frame_size) // after a stack frame has been allocated. + off := -spdelta + if thearch.ReturnAddressAtTopOfFrame { + // Except arm64, which has it at the top of frame. + off = -int64(d.arch.PtrSize) + } deltaBuf = append(deltaBuf, dwarf.DW_CFA_offset_extended_sf) deltaBuf = dwarf.AppendUleb128(deltaBuf, uint64(thearch.Dwarfreglr)) - deltaBuf = dwarf.AppendSleb128(deltaBuf, -spdelta/dataAlignmentFactor) + deltaBuf = dwarf.AppendSleb128(deltaBuf, off/dataAlignmentFactor) } else { // The return address is restored into the link register // when a stack frame has been de-allocated. diff --git a/src/cmd/link/internal/ld/lib.go b/src/cmd/link/internal/ld/lib.go index 2c861129b5..5f5ebfc1d9 100644 --- a/src/cmd/link/internal/ld/lib.go +++ b/src/cmd/link/internal/ld/lib.go @@ -263,6 +263,10 @@ type Arch struct { // optional override for assignAddress AssignAddress func(ldr *loader.Loader, sect *sym.Section, n int, s loader.Sym, va uint64, isTramp bool) (*sym.Section, int, uint64) + // Reports whether the return address is stored at the top (highest address) + // of the stack frame. + ReturnAddressAtTopOfFrame bool + // ELF specific information. ELF ELFArch } diff --git a/src/cmd/link/internal/ld/stackcheck.go b/src/cmd/link/internal/ld/stackcheck.go index 98e7edaeb1..14cd3a2238 100644 --- a/src/cmd/link/internal/ld/stackcheck.go +++ b/src/cmd/link/internal/ld/stackcheck.go @@ -9,7 +9,6 @@ import ( "cmd/internal/objabi" "cmd/link/internal/loader" "fmt" - "internal/buildcfg" "sort" "strings" ) @@ -62,10 +61,6 @@ func (ctxt *Link) doStackCheck() { // that there are at least StackLimit bytes available below SP // when morestack returns. limit := objabi.StackNosplit(*flagRace) - sc.callSize - if buildcfg.GOARCH == "arm64" { - // Need an extra 8 bytes below SP to save FP. - limit -= 8 - } // Compute stack heights without any back-tracking information. // This will almost certainly succeed and we can simply diff --git a/src/cmd/link/internal/x86/obj.go b/src/cmd/link/internal/x86/obj.go index 4336f01ea3..a4885fde8f 100644 --- a/src/cmd/link/internal/x86/obj.go +++ b/src/cmd/link/internal/x86/obj.go @@ -50,13 +50,14 @@ func Init() (*sys.Arch, ld.Arch) { Plan9Magic: uint32(4*11*11 + 7), - Adddynrel: adddynrel, - Archinit: archinit, - Archreloc: archreloc, - Archrelocvariant: archrelocvariant, - Gentext: gentext, - Machoreloc1: machoreloc1, - PEreloc1: pereloc1, + Adddynrel: adddynrel, + Archinit: archinit, + Archreloc: archreloc, + Archrelocvariant: archrelocvariant, + Gentext: gentext, + Machoreloc1: machoreloc1, + PEreloc1: pereloc1, + ReturnAddressAtTopOfFrame: true, ELF: ld.ELFArch{ Linuxdynld: "/lib/ld-linux.so.2", |
