aboutsummaryrefslogtreecommitdiff
path: root/src/pkg/runtime/stack.c
diff options
context:
space:
mode:
authorRuss Cox <rsc@golang.org>2013-09-12 14:00:16 -0400
committerRuss Cox <rsc@golang.org>2013-09-12 14:00:16 -0400
commit7276c02b4193edb19bc0d2d36a786238564db03f (patch)
treee5d13c00ad0b813e8a1edcf9381a8b242780ef2f /src/pkg/runtime/stack.c
parent1ea0c480dc16a986c2c335ff2965e70d99bfa654 (diff)
downloadgo-7276c02b4193edb19bc0d2d36a786238564db03f.tar.xz
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover
Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
Diffstat (limited to 'src/pkg/runtime/stack.c')
-rw-r--r--src/pkg/runtime/stack.c13
1 files changed, 12 insertions, 1 deletions
diff --git a/src/pkg/runtime/stack.c b/src/pkg/runtime/stack.c
index 6b34f091e1..011c616bac 100644
--- a/src/pkg/runtime/stack.c
+++ b/src/pkg/runtime/stack.c
@@ -174,6 +174,7 @@ runtime·oldstack(void)
gp->stackbase = top->stackbase;
gp->stackguard = top->stackguard;
gp->stackguard0 = gp->stackguard;
+ gp->panicwrap = top->panicwrap;
if(top->free != 0) {
gp->stacksize -= top->free;
@@ -195,7 +196,7 @@ void
runtime·newstack(void)
{
int32 framesize, argsize, oldstatus;
- Stktop *top;
+ Stktop *top, *oldtop;
byte *stk;
uintptr sp;
uintptr *src, *dst, *dstend;
@@ -316,6 +317,16 @@ runtime·newstack(void)
// copy flag from panic
top->panic = gp->ispanic;
gp->ispanic = false;
+
+ // if this isn't a panic, maybe we're splitting the stack for a panic.
+ // if we're splitting in the top frame, propagate the panic flag
+ // forward so that recover will know we're in a panic.
+ oldtop = (Stktop*)top->stackbase;
+ if(oldtop != nil && oldtop->panic && top->argp == (byte*)oldtop - oldtop->argsize - gp->panicwrap)
+ top->panic = true;
+
+ top->panicwrap = gp->panicwrap;
+ gp->panicwrap = 0;
gp->stackbase = (uintptr)top;
gp->stackguard = (uintptr)stk + StackGuard;