Skip to content

Commit

Permalink
cmd/compile: convert 386 port to use addressing modes pass (take 2)
Browse files Browse the repository at this point in the history 
Retrying CL 222782, with a fix that will hopefully stop the random crashing.

The issue with the previous CL is that it does pointer arithmetic
in a way that may briefly generate an out-of-bounds pointer. If an
interrupt happens to occur in that state, the referenced object may
be collected incorrectly.

Suppose there was code that did s[x+c].  The previous CL had a rule
to the effect of ptr + (x + c) -> c + (ptr + x).  But ptr+x is not
guaranteed to point to the same object as ptr. In contrast,
ptr+(x+c) is guaranteed to point to the same object as ptr, because
we would have already checked that x+c is in bounds.

For example, strconv.trim used to have this code:
  MOVZX -0x1(BX)(DX*1), BP
  CMPL $0x30, AL
After CL 222782, it had this code:
  LEAL 0(BX)(DX*1), BP
  CMPB $0x30, -0x1(BP)

An interrupt between those last two instructions could see BP pointing
outside the backing store of the slice involved.

It's really hard to actually demonstrate a bug. First, you need to
have an interrupt occur at exactly the right time. Then, there must
be no other pointers to the object in question. Since the interrupted
frame will be scanned conservatively, there can't even be a dead
pointer in another register or on the stack. (In the example above,
a bug can't happen because BX still holds the original pointer.)
Then, the object in question needs to be collected (or at least
scanned?) before the interrupted code continues.

This CL needs to handle load combining somewhat differently than CL 222782
because of the new restriction on arithmetic. That's the only real
difference (other than removing the bad rules) from that old CL.

This bug is also present in the amd64 rewrite rules, and we haven't
seen any crashing as a result. I will fix up that code similarly to
this one in a separate CL.

Update #37881

Change-Id: I5f0d584d9bef4696bfe89a61ef0a27c8d507329f
Reviewed-on: https://go-review.googlesource.com/c/go/+/225798
Run-TryBot: Keith Randall <[email protected]>
TryBot-Result: Gobot Gobot <[email protected]>
Reviewed-by: Cherry Zhang <[email protected]>
  • Loading branch information
@randall77
randall77 committed Mar 27, 2020
1 parent 4a8b9bd commit af7eafd
Show file tree
Hide file tree
Showing 9 changed files with 1,270 additions and 5,325 deletions.
83 changes: 77 additions & 6 deletions src/cmd/compile/internal/ssa/addressingmodes.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -11,8 +11,8 @@ func addressingModes(f *Func) {
default:
// Most architectures can't do this.
return
case "amd64":
// TODO: 386, s390x?
case "amd64", "386":
// TODO: s390x?
}

var tmp []*Value
Expand All @@ -21,7 +21,17 @@ func addressingModes(f *Func) {
if !combineFirst[v.Op] {
continue
}
p := v.Args[0]
// All matched operations have the pointer in arg[0].
// All results have the pointer in arg[0] and the index in arg[1].
// *Except* for operations which update a register,
// which are marked with resultInArg0. Those have
// the pointer in arg[1], and the corresponding result op
// has the pointer in arg[1] and the index in arg[2].
ptrIndex := 0
if opcodeTable[v.Op].resultInArg0 {
ptrIndex = 1
}
p := v.Args[ptrIndex]
c, ok := combine[[2]Op{v.Op, p.Op}]
if !ok {
continue
Expand Down Expand Up @@ -71,10 +81,11 @@ func addressingModes(f *Func) {
f.Fatalf("unknown aux combining for %s and %s\n", v.Op, p.Op)
}
// Combine the operations.
tmp = append(tmp[:0], v.Args[1:]...)
tmp = append(tmp[:0], v.Args[:ptrIndex]...)
tmp = append(tmp, p.Args...)
tmp = append(tmp, v.Args[ptrIndex+1:]...)
v.resetArgs()
v.Op = c
v.AddArgs(p.Args...)
v.AddArgs(tmp...)
}
}
Expand All @@ -97,6 +108,7 @@ func init() {
// x.Args[0].Args + x.Args[1:]
// Additionally, the Aux/AuxInt from x.Args[0] is merged into x.
var combine = map[[2]Op]Op{
// amd64
[2]Op{OpAMD64MOVBload, OpAMD64ADDQ}: OpAMD64MOVBloadidx1,
[2]Op{OpAMD64MOVWload, OpAMD64ADDQ}: OpAMD64MOVWloadidx1,
[2]Op{OpAMD64MOVLload, OpAMD64ADDQ}: OpAMD64MOVLloadidx1,
Expand Down Expand Up @@ -150,5 +162,64 @@ var combine = map[[2]Op]Op{
[2]Op{OpAMD64MOVQstoreconst, OpAMD64LEAQ1}: OpAMD64MOVQstoreconstidx1,
[2]Op{OpAMD64MOVQstoreconst, OpAMD64LEAQ8}: OpAMD64MOVQstoreconstidx8,

// TODO: 386
// 386
[2]Op{Op386MOVBload, Op386ADDL}: Op386MOVBloadidx1,
[2]Op{Op386MOVWload, Op386ADDL}: Op386MOVWloadidx1,
[2]Op{Op386MOVLload, Op386ADDL}: Op386MOVLloadidx1,
[2]Op{Op386MOVSSload, Op386ADDL}: Op386MOVSSloadidx1,
[2]Op{Op386MOVSDload, Op386ADDL}: Op386MOVSDloadidx1,

[2]Op{Op386MOVBstore, Op386ADDL}: Op386MOVBstoreidx1,
[2]Op{Op386MOVWstore, Op386ADDL}: Op386MOVWstoreidx1,
[2]Op{Op386MOVLstore, Op386ADDL}: Op386MOVLstoreidx1,
[2]Op{Op386MOVSSstore, Op386ADDL}: Op386MOVSSstoreidx1,
[2]Op{Op386MOVSDstore, Op386ADDL}: Op386MOVSDstoreidx1,

[2]Op{Op386MOVBstoreconst, Op386ADDL}: Op386MOVBstoreconstidx1,
[2]Op{Op386MOVWstoreconst, Op386ADDL}: Op386MOVWstoreconstidx1,
[2]Op{Op386MOVLstoreconst, Op386ADDL}: Op386MOVLstoreconstidx1,

[2]Op{Op386MOVBload, Op386LEAL1}: Op386MOVBloadidx1,
[2]Op{Op386MOVWload, Op386LEAL1}: Op386MOVWloadidx1,
[2]Op{Op386MOVWload, Op386LEAL2}: Op386MOVWloadidx2,
[2]Op{Op386MOVLload, Op386LEAL1}: Op386MOVLloadidx1,
[2]Op{Op386MOVLload, Op386LEAL4}: Op386MOVLloadidx4,
[2]Op{Op386MOVSSload, Op386LEAL1}: Op386MOVSSloadidx1,
[2]Op{Op386MOVSSload, Op386LEAL4}: Op386MOVSSloadidx4,
[2]Op{Op386MOVSDload, Op386LEAL1}: Op386MOVSDloadidx1,
[2]Op{Op386MOVSDload, Op386LEAL8}: Op386MOVSDloadidx8,

[2]Op{Op386MOVBstore, Op386LEAL1}: Op386MOVBstoreidx1,
[2]Op{Op386MOVWstore, Op386LEAL1}: Op386MOVWstoreidx1,
[2]Op{Op386MOVWstore, Op386LEAL2}: Op386MOVWstoreidx2,
[2]Op{Op386MOVLstore, Op386LEAL1}: Op386MOVLstoreidx1,
[2]Op{Op386MOVLstore, Op386LEAL4}: Op386MOVLstoreidx4,
[2]Op{Op386MOVSSstore, Op386LEAL1}: Op386MOVSSstoreidx1,
[2]Op{Op386MOVSSstore, Op386LEAL4}: Op386MOVSSstoreidx4,
[2]Op{Op386MOVSDstore, Op386LEAL1}: Op386MOVSDstoreidx1,
[2]Op{Op386MOVSDstore, Op386LEAL8}: Op386MOVSDstoreidx8,

[2]Op{Op386MOVBstoreconst, Op386LEAL1}: Op386MOVBstoreconstidx1,
[2]Op{Op386MOVWstoreconst, Op386LEAL1}: Op386MOVWstoreconstidx1,
[2]Op{Op386MOVWstoreconst, Op386LEAL2}: Op386MOVWstoreconstidx2,
[2]Op{Op386MOVLstoreconst, Op386LEAL1}: Op386MOVLstoreconstidx1,
[2]Op{Op386MOVLstoreconst, Op386LEAL4}: Op386MOVLstoreconstidx4,

[2]Op{Op386ADDLload, Op386LEAL4}: Op386ADDLloadidx4,
[2]Op{Op386SUBLload, Op386LEAL4}: Op386SUBLloadidx4,
[2]Op{Op386MULLload, Op386LEAL4}: Op386MULLloadidx4,
[2]Op{Op386ANDLload, Op386LEAL4}: Op386ANDLloadidx4,
[2]Op{Op386ORLload, Op386LEAL4}: Op386ORLloadidx4,
[2]Op{Op386XORLload, Op386LEAL4}: Op386XORLloadidx4,

[2]Op{Op386ADDLmodify, Op386LEAL4}: Op386ADDLmodifyidx4,
[2]Op{Op386SUBLmodify, Op386LEAL4}: Op386SUBLmodifyidx4,
[2]Op{Op386ANDLmodify, Op386LEAL4}: Op386ANDLmodifyidx4,
[2]Op{Op386ORLmodify, Op386LEAL4}: Op386ORLmodifyidx4,
[2]Op{Op386XORLmodify, Op386LEAL4}: Op386XORLmodifyidx4,

[2]Op{Op386ADDLconstmodify, Op386LEAL4}: Op386ADDLconstmodifyidx4,
[2]Op{Op386ANDLconstmodify, Op386LEAL4}: Op386ANDLconstmodifyidx4,
[2]Op{Op386ORLconstmodify, Op386LEAL4}: Op386ORLconstmodifyidx4,
[2]Op{Op386XORLconstmodify, Op386LEAL4}: Op386XORLconstmodifyidx4,
}
Loading

0 comments on commit af7eafd

Please sign in to comment.