qbit/go
1
0
Fork 0
mirror of https://github.com/golang/go synced 2024-10-04 02:21:21 -06:00
Code Releases Activity
a1c9e10371
go/src/liblink/obj8.c

874 lines
18 KiB
C
Raw Normal View History

liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
// Inferno utils/8l/pass.c
// http://code.google.com/p/inferno-os/source/browse/utils/8l/pass.c
//
// Copyright © 1994-1999 Lucent Technologies Inc. All rights reserved.
// Portions Copyright © 1995-1997 C H Forsyth (forsyth@terzarima.net)
// Portions Copyright © 1997-1999 Vita Nuova Limited
// Portions Copyright © 2000-2007 Vita Nuova Holdings Limited (www.vitanuova.com)
// Portions Copyright © 2004,2006 Bruce Ellis
// Portions Copyright © 2005-2007 C H Forsyth (forsyth@terzarima.net)
// Revisions Copyright © 2000-2007 Lucent Technologies Inc. and others
// Portions Copyright © 2009 The Go Authors. All rights reserved.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.
#include <u.h>
#include <libc.h>
#include <bio.h>
#include <link.h>
#include "../cmd/8l/8.out.h"
build: adjustments for move from src/pkg to src This CL adjusts code referring to src/pkg to refer to src. Immediately after submitting this CL, I will submit a change doing 'hg mv src/pkg/* src'. That change will be too large to review with Rietveld but will contain only the 'hg mv'. This CL will break the build. The followup 'hg mv' will fix it. For more about the move, see golang.org/s/go14nopkg. LGTM=r R=r CC=golang-codereviews https://golang.org/cl/134570043
2014-09-07 22:06:45 -06:00
#include "../runtime/stack.h"
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
static Prog zprg = {
.back = 2,
.as = AGOK,
.from = {
.type = D_NONE,
.index = D_NONE,
.scale = 1,
},
.to = {
.type = D_NONE,
.index = D_NONE,
.scale = 1,
},
};
static int
symtype(Addr *a)
{
int t;
t = a->type;
if(t == D_ADDR)
t = a->index;
return t;
}
static int
isdata(Prog *p)
{
return p->as == ADATA || p->as == AGLOBL;
}
static int
iscall(Prog *p)
{
return p->as == ACALL;
}
static int
datasize(Prog *p)
{
return p->from.scale;
}
static int
textflag(Prog *p)
{
return p->from.scale;
}
static void
settextflag(Prog *p, int f)
{
p->from.scale = f;
}
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
static int
canuselocaltls(Link *ctxt)
{
switch(ctxt->headtype) {
case Hlinux:
case Hnacl:
case Hplan9:
case Hwindows:
return 0;
}
return 1;
}
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
static void
progedit(Link *ctxt, Prog *p)
{
cmd/ld: move instruction selection + layout into compilers, assemblers - new object file reader/writer (liblink/objfile.c) - remove old object file writing routines - add pcdata iterator - remove all trace of "line number stack" and "path fragments" from object files, linker (!!!) - dwarf now writes a single "compilation unit" instead of one per package This CL disables the check for chains of no-split functions that could overflow the stack red zone. A future CL will attack the problem of reenabling that check (issue 6931). This CL is just the liblink and cmd/ld changes. There are minor associated adjustments in CL 37030045. Each depends on the other. R=golang-dev, dave, iant CC=golang-dev https://golang.org/cl/39680043
2013-12-16 10:51:58 -07:00
char literal[64];
LSym *s;
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
Prog *q;
// See obj6.c for discussion of TLS.
if(canuselocaltls(ctxt)) {
// Reduce TLS initial exec model to TLS local exec model.
// Sequences like
// MOVL TLS, BX
// ... off(BX)(TLS*1) ...
// become
// NOP
// ... off(TLS) ...
if(p->as == AMOVL && p->from.type == D_TLS && D_AX <= p->to.type && p->to.type <= D_DI) {
p->as = ANOP;
p->from.type = D_NONE;
p->to.type = D_NONE;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
}
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
if(p->from.index == D_TLS && D_INDIR+D_AX <= p->from.type && p->from.type <= D_INDIR+D_DI) {
p->from.type = D_INDIR+D_TLS;
p->from.scale = 0;
p->from.index = D_NONE;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
}
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
if(p->to.index == D_TLS && D_INDIR+D_AX <= p->to.type && p->to.type <= D_INDIR+D_DI) {
p->to.type = D_INDIR+D_TLS;
p->to.scale = 0;
p->to.index = D_NONE;
}
} else {
// As a courtesy to the C compilers, rewrite TLS local exec load as TLS initial exec load.
// The instruction
// MOVL off(TLS), BX
// becomes the sequence
// MOVL TLS, BX
// MOVL off(BX)(TLS*1), BX
// This allows the C compilers to emit references to m and g using the direct off(TLS) form.
if(p->as == AMOVL && p->from.type == D_INDIR+D_TLS && D_AX <= p->to.type && p->to.type <= D_DI) {
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
q = appendp(ctxt, p);
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
q->as = p->as;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
q->from = p->from;
q->from.type = D_INDIR + p->to.type;
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
q->from.index = D_TLS;
q->from.scale = 2; // TODO: use 1
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
q->to = p->to;
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
p->from.type = D_TLS;
p->from.index = D_NONE;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
p->from.offset = 0;
}
}
cmd/ld: move instruction selection + layout into compilers, assemblers - new object file reader/writer (liblink/objfile.c) - remove old object file writing routines - add pcdata iterator - remove all trace of "line number stack" and "path fragments" from object files, linker (!!!) - dwarf now writes a single "compilation unit" instead of one per package This CL disables the check for chains of no-split functions that could overflow the stack red zone. A future CL will attack the problem of reenabling that check (issue 6931). This CL is just the liblink and cmd/ld changes. There are minor associated adjustments in CL 37030045. Each depends on the other. R=golang-dev, dave, iant CC=golang-dev https://golang.org/cl/39680043
2013-12-16 10:51:58 -07:00
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
// TODO: Remove.
if(ctxt->headtype == Hplan9) {
if(p->from.scale == 1 && p->from.index == D_TLS)
p->from.scale = 2;
if(p->to.scale == 1 && p->to.index == D_TLS)
p->to.scale = 2;
}
cmd/ld: move instruction selection + layout into compilers, assemblers - new object file reader/writer (liblink/objfile.c) - remove old object file writing routines - add pcdata iterator - remove all trace of "line number stack" and "path fragments" from object files, linker (!!!) - dwarf now writes a single "compilation unit" instead of one per package This CL disables the check for chains of no-split functions that could overflow the stack red zone. A future CL will attack the problem of reenabling that check (issue 6931). This CL is just the liblink and cmd/ld changes. There are minor associated adjustments in CL 37030045. Each depends on the other. R=golang-dev, dave, iant CC=golang-dev https://golang.org/cl/39680043
2013-12-16 10:51:58 -07:00
// Rewrite CALL/JMP/RET to symbol as D_BRANCH.
switch(p->as) {
case ACALL:
case AJMP:
case ARET:
if((p->to.type == D_EXTERN || p->to.type == D_STATIC) && p->to.sym != nil)
p->to.type = D_BRANCH;
break;
}
// Rewrite float constants to values stored in memory.
switch(p->as) {
case AFMOVF:
case AFADDF:
case AFSUBF:
case AFSUBRF:
case AFMULF:
case AFDIVF:
case AFDIVRF:
case AFCOMF:
case AFCOMFP:
case AMOVSS:
case AADDSS:
case ASUBSS:
case AMULSS:
case ADIVSS:
case ACOMISS:
case AUCOMISS:
if(p->from.type == D_FCONST) {
int32 i32;
float32 f32;
f32 = p->from.u.dval;
memmove(&i32, &f32, 4);
sprint(literal, "$f32.%08ux", (uint32)i32);
s = linklookup(ctxt, literal, 0);
if(s->type == 0) {
s->type = SRODATA;
adduint32(ctxt, s, i32);
s->reachable = 0;
}
p->from.type = D_EXTERN;
p->from.sym = s;
p->from.offset = 0;
}
break;
case AFMOVD:
case AFADDD:
case AFSUBD:
case AFSUBRD:
case AFMULD:
case AFDIVD:
case AFDIVRD:
case AFCOMD:
case AFCOMDP:
case AMOVSD:
case AADDSD:
case ASUBSD:
case AMULSD:
case ADIVSD:
case ACOMISD:
case AUCOMISD:
if(p->from.type == D_FCONST) {
int64 i64;
memmove(&i64, &p->from.u.dval, 8);
sprint(literal, "$f64.%016llux", (uvlong)i64);
s = linklookup(ctxt, literal, 0);
if(s->type == 0) {
s->type = SRODATA;
adduint64(ctxt, s, i64);
s->reachable = 0;
}
p->from.type = D_EXTERN;
p->from.sym = s;
p->from.offset = 0;
}
break;
}
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
}
static Prog*
prg(void)
{
Prog *p;
p = emallocz(sizeof(*p));
*p = zprg;
return p;
}
static Prog* load_g_cx(Link*, Prog*);
cmd/ld: clear unused ctxt before morestack For non-closure functions, the context register is uninitialized on entry and will not be used, but morestack saves it and then the garbage collector treats it as live. This can be a source of memory leaks if the context register points at otherwise dead memory. Avoid this by introducing a parallel set of morestack functions that clear the context register, and use those for the non-closure functions. I hope this will help with some of the finalizer flakiness, but it probably won't. Fixes #7244. LGTM=dvyukov R=khr, dvyukov CC=golang-codereviews https://golang.org/cl/71030044
2014-03-04 11:53:08 -07:00
static Prog* stacksplit(Link*, Prog*, int32, int, Prog**);
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
static void
addstacksplit(Link *ctxt, LSym *cursym)
{
runtime: fix panic/wrapper/recover math The gp->panicwrap adjustment is just fatally flawed. Now that there is a Panic.argp field, update that instead. That can be done on entry only, so that unwinding doesn't need to worry about undoing anything. The wrappers emit a few more instructions in the prologue but everything else in the system gets much simpler. It also fixes (without trying) a broken test I never checked in. Fixes #7491. LGTM=khr R=khr CC=dvyukov, golang-codereviews, iant, r https://golang.org/cl/135490044
2014-09-06 11:19:08 -06:00
Prog *p, *q, *p1, *p2;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
int32 autoffset, deltasp;
int a;
cmd/ld: clear unused ctxt before morestack For non-closure functions, the context register is uninitialized on entry and will not be used, but morestack saves it and then the garbage collector treats it as live. This can be a source of memory leaks if the context register points at otherwise dead memory. Avoid this by introducing a parallel set of morestack functions that clear the context register, and use those for the non-closure functions. I hope this will help with some of the finalizer flakiness, but it probably won't. Fixes #7244. LGTM=dvyukov R=khr, dvyukov CC=golang-codereviews https://golang.org/cl/71030044
2014-03-04 11:53:08 -07:00
if(ctxt->symmorestack[0] == nil) {
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
ctxt->symmorestack[0] = linklookup(ctxt, "runtime.morestack", 0);
cmd/ld: clear unused ctxt before morestack For non-closure functions, the context register is uninitialized on entry and will not be used, but morestack saves it and then the garbage collector treats it as live. This can be a source of memory leaks if the context register points at otherwise dead memory. Avoid this by introducing a parallel set of morestack functions that clear the context register, and use those for the non-closure functions. I hope this will help with some of the finalizer flakiness, but it probably won't. Fixes #7244. LGTM=dvyukov R=khr, dvyukov CC=golang-codereviews https://golang.org/cl/71030044
2014-03-04 11:53:08 -07:00
ctxt->symmorestack[1] = linklookup(ctxt, "runtime.morestack_noctxt", 0);
}
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
liblink, runtime: preliminary support for plan9/amd64 A TLS slot is reserved by _rt0_.*_plan9 as an automatic and its address (which is static on Plan 9) is saved in the global _privates symbol. The startup linkage now is exactly like that from Plan 9 libc, and the way we access g is exactly as if we'd have used privalloc(2). Aside from making the code more standard, this change drastically simplifies it, both for 386 and for amd64, and makes the Plan 9 code in liblink common for both 386 and amd64. The amd64 runtime code was cleared of nxm assumptions, and now runs on the standard Plan 9 kernel. Note handling fixes will follow in a separate CL. LGTM=rsc R=golang-codereviews, rsc, bradfitz, dave CC=0intro, ality, golang-codereviews, jas, minux.ma, mischief https://golang.org/cl/101510049
2014-07-02 05:04:10 -06:00
if(ctxt->headtype == Hplan9 && ctxt->plan9privates == nil)
ctxt->plan9privates = linklookup(ctxt, "_privates", 0);
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
ctxt->cursym = cursym;
if(cursym->text == nil || cursym->text->link == nil)
return;
p = cursym->text;
autoffset = p->to.offset;
if(autoffset < 0)
autoffset = 0;
cmd/ld: move instruction selection + layout into compilers, assemblers - new object file reader/writer (liblink/objfile.c) - remove old object file writing routines - add pcdata iterator - remove all trace of "line number stack" and "path fragments" from object files, linker (!!!) - dwarf now writes a single "compilation unit" instead of one per package This CL disables the check for chains of no-split functions that could overflow the stack red zone. A future CL will attack the problem of reenabling that check (issue 6931). This CL is just the liblink and cmd/ld changes. There are minor associated adjustments in CL 37030045. Each depends on the other. R=golang-dev, dave, iant CC=golang-dev https://golang.org/cl/39680043
2013-12-16 10:51:58 -07:00
cursym->locals = autoffset;
cursym->args = p->to.offset2;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
q = nil;
if(!(p->from.scale & NOSPLIT) || (p->from.scale & WRAPPER)) {
p = appendp(ctxt, p);
p = load_g_cx(ctxt, p); // load g into CX
}
if(!(cursym->text->from.scale & NOSPLIT))
cmd/ld: clear unused ctxt before morestack For non-closure functions, the context register is uninitialized on entry and will not be used, but morestack saves it and then the garbage collector treats it as live. This can be a source of memory leaks if the context register points at otherwise dead memory. Avoid this by introducing a parallel set of morestack functions that clear the context register, and use those for the non-closure functions. I hope this will help with some of the finalizer flakiness, but it probably won't. Fixes #7244. LGTM=dvyukov R=khr, dvyukov CC=golang-codereviews https://golang.org/cl/71030044
2014-03-04 11:53:08 -07:00
p = stacksplit(ctxt, p, autoffset, !(cursym->text->from.scale&NEEDCTXT), &q); // emit split check
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
if(autoffset) {
p = appendp(ctxt, p);
p->as = AADJSP;
p->from.type = D_CONST;
p->from.offset = autoffset;
p->spadj = autoffset;
} else {
// zero-byte stack adjustment.
// Insert a fake non-zero adjustment so that stkcheck can
// recognize the end of the stack-splitting prolog.
p = appendp(ctxt, p);
p->as = ANOP;
p->spadj = -ctxt->arch->ptrsize;
p = appendp(ctxt, p);
p->as = ANOP;
p->spadj = ctxt->arch->ptrsize;
}
if(q != nil)
q->pcond = p;
deltasp = autoffset;
if(cursym->text->from.scale & WRAPPER) {
runtime: fix panic/wrapper/recover math The gp->panicwrap adjustment is just fatally flawed. Now that there is a Panic.argp field, update that instead. That can be done on entry only, so that unwinding doesn't need to worry about undoing anything. The wrappers emit a few more instructions in the prologue but everything else in the system gets much simpler. It also fixes (without trying) a broken test I never checked in. Fixes #7491. LGTM=khr R=khr CC=dvyukov, golang-codereviews, iant, r https://golang.org/cl/135490044
2014-09-06 11:19:08 -06:00
// if(g->panic != nil && g->panic->argp == FP) g->panic->argp = bottom-of-frame
//
// MOVL g_panic(CX), BX
// TESTL BX, BX
// JEQ end
// LEAL (autoffset+4)(SP), DI
// CMPL panic_argp(BX), DI
// JNE end
// MOVL SP, panic_argp(BX)
// end:
// NOP
//
// The NOP is needed to give the jumps somewhere to land.
// It is a liblink NOP, not an x86 NOP: it encodes to 0 instruction bytes.
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
p = appendp(ctxt, p);
runtime: fix panic/wrapper/recover math The gp->panicwrap adjustment is just fatally flawed. Now that there is a Panic.argp field, update that instead. That can be done on entry only, so that unwinding doesn't need to worry about undoing anything. The wrappers emit a few more instructions in the prologue but everything else in the system gets much simpler. It also fixes (without trying) a broken test I never checked in. Fixes #7491. LGTM=khr R=khr CC=dvyukov, golang-codereviews, iant, r https://golang.org/cl/135490044
2014-09-06 11:19:08 -06:00
p->as = AMOVL;
p->from.type = D_INDIR+D_CX;
liblink, cmd/ld, runtime: remove stackguard1 Now that we've removed all the C code in runtime and the C compilers, there is no need to have a separate stackguard field to check for C code on Go stack. Remove field g.stackguard1 and rename g.stackguard0 to g.stackguard. Adjust liblink and cmd/ld as necessary. Change-Id: I54e75db5a93d783e86af5ff1a6cd497d669d8d33 Reviewed-on: https://go-review.googlesource.com/2144 Reviewed-by: Keith Randall <khr@golang.org>
2014-12-28 23:08:40 -07:00
p->from.offset = 3*ctxt->arch->ptrsize; // G.panic
runtime: fix panic/wrapper/recover math The gp->panicwrap adjustment is just fatally flawed. Now that there is a Panic.argp field, update that instead. That can be done on entry only, so that unwinding doesn't need to worry about undoing anything. The wrappers emit a few more instructions in the prologue but everything else in the system gets much simpler. It also fixes (without trying) a broken test I never checked in. Fixes #7491. LGTM=khr R=khr CC=dvyukov, golang-codereviews, iant, r https://golang.org/cl/135490044
2014-09-06 11:19:08 -06:00
p->to.type = D_BX;
p = appendp(ctxt, p);
p->as = ATESTL;
p->from.type = D_BX;
p->to.type = D_BX;
p = appendp(ctxt, p);
p->as = AJEQ;
p->to.type = D_BRANCH;
p1 = p;
p = appendp(ctxt, p);
p->as = ALEAL;
p->from.type = D_INDIR+D_SP;
p->from.offset = autoffset+4;
p->to.type = D_DI;
p = appendp(ctxt, p);
p->as = ACMPL;
p->from.type = D_INDIR+D_BX;
p->from.offset = 0; // Panic.argp
p->to.type = D_DI;
p = appendp(ctxt, p);
p->as = AJNE;
p->to.type = D_BRANCH;
p2 = p;
p = appendp(ctxt, p);
p->as = AMOVL;
p->from.type = D_SP;
p->to.type = D_INDIR+D_BX;
p->to.offset = 0; // Panic.argp
p = appendp(ctxt, p);
p->as = ANOP;
p1->pcond = p;
p2->pcond = p;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
}
if(ctxt->debugzerostack && autoffset && !(cursym->text->from.scale&NOSPLIT)) {
// 8l -Z means zero the stack frame on entry.
// This slows down function calls but can help avoid
// false positives in garbage collection.
p = appendp(ctxt, p);
p->as = AMOVL;
p->from.type = D_SP;
p->to.type = D_DI;
p = appendp(ctxt, p);
p->as = AMOVL;
p->from.type = D_CONST;
p->from.offset = autoffset/4;
p->to.type = D_CX;
p = appendp(ctxt, p);
p->as = AMOVL;
p->from.type = D_CONST;
p->from.offset = 0;
p->to.type = D_AX;
p = appendp(ctxt, p);
p->as = AREP;
p = appendp(ctxt, p);
p->as = ASTOSL;
}
for(; p != nil; p = p->link) {
a = p->from.type;
if(a == D_AUTO)
p->from.offset += deltasp;
if(a == D_PARAM)
p->from.offset += deltasp + 4;
a = p->to.type;
if(a == D_AUTO)
p->to.offset += deltasp;
if(a == D_PARAM)
p->to.offset += deltasp + 4;
switch(p->as) {
default:
continue;
case APUSHL:
case APUSHFL:
deltasp += 4;
p->spadj = 4;
continue;
case APUSHW:
case APUSHFW:
deltasp += 2;
p->spadj = 2;
continue;
case APOPL:
case APOPFL:
deltasp -= 4;
p->spadj = -4;
continue;
case APOPW:
case APOPFW:
deltasp -= 2;
p->spadj = -2;
continue;
case ARET:
break;
}
if(autoffset != deltasp)
ctxt->diag("unbalanced PUSH/POP");
if(autoffset) {
p->as = AADJSP;
p->from.type = D_CONST;
p->from.offset = -autoffset;
p->spadj = -autoffset;
p = appendp(ctxt, p);
p->as = ARET;
// If there are instructions following
// this ARET, they come from a branch
// with the same stackframe, so undo
// the cleanup.
p->spadj = +autoffset;
}
if(p->to.sym) // retjmp
p->as = AJMP;
}
}
// Append code to p to load g into cx.
// Overwrites p with the first instruction (no first appendp).
// Overwriting p is unusual but it lets use this in both the
// prologue (caller must call appendp first) and in the epilogue.
// Returns last new instruction.
static Prog*
load_g_cx(Link *ctxt, Prog *p)
{
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
Prog *next;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
p->as = AMOVL;
p->from.type = D_INDIR+D_TLS;
p->from.offset = 0;
p->to.type = D_CX;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
next = p->link;
progedit(ctxt, p);
while(p->link != next)
p = p->link;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
liblink: introduce TLS register on 386 and amd64 When I did the original 386 ports on Linux and OS X, I chose to define GS-relative expressions like 4(GS) as relative to the actual thread-local storage base, which was usually GS but might not be (it might be FS, or it might be a different constant offset from GS or FS). The original scope was limited but since then the rewrites have gotten out of control. Sometimes GS is rewritten, sometimes FS. Some ports do other rewrites to enable shared libraries and other linking. At no point in the code is it clear whether you are looking at the real GS/FS or some synthesized thing that will be rewritten. The code manipulating all these is duplicated in many places. The first step to fixing issue 7719 is to make the code intelligible again. This CL adds an explicit TLS pseudo-register to the 386 and amd64. As a register, TLS refers to the thread-local storage base, and it can only be loaded into another register: MOVQ TLS, AX An offset from the thread-local storage base is written off(reg)(TLS*1). Semantically it is off(reg), but the (TLS*1) annotation marks this as indexing from the loaded TLS base. This emits a relocation so that if the linker needs to adjust the offset, it can. For example: MOVQ TLS, AX MOVQ 8(AX)(TLS*1), CX // load m into CX On systems that support direct access to the TLS memory, this pair of instructions can be reduced to a direct TLS memory reference: MOVQ 8(TLS), CX // load m into CX The 2-instruction and 1-instruction forms correspond roughly to ELF TLS initial exec mode and ELF TLS local exec mode, respectively. Liblink applies this rewrite on systems that support the 1-instruction form. The decision is made using only the operating system (and probably the -shared flag, eventually), not the link mode. If some link modes on a particular operating system require the 2-instruction form, then all builds for that operating system will use the 2-instruction form, so that the link mode decision can be delayed to link time. Obviously it is late to be making changes like this, but I despair of correcting issue 7719 and issue 7164 without it. To make sure I am not changing existing behavior, I built a "hello world" program for every GOOS/GOARCH combination we have and then worked to make sure that the rewrite generates exactly the same binaries, byte for byte. There are a handful of TODOs in the code marking kludges to get the byte-for-byte property, but at least now I can explain exactly how each binary is handled. The targets I tested this way are: darwin-386 darwin-amd64 dragonfly-386 dragonfly-amd64 freebsd-386 freebsd-amd64 freebsd-arm linux-386 linux-amd64 linux-arm nacl-386 nacl-amd64p32 netbsd-386 netbsd-amd64 openbsd-386 openbsd-amd64 plan9-386 plan9-amd64 solaris-amd64 windows-386 windows-amd64 There were four exceptions to the byte-for-byte goal: windows-386 and windows-amd64 have a time stamp at bytes 137 and 138 of the header. darwin-386 and plan9-386 have five or six modified bytes in the middle of the Go symbol table, caused by editing comments in runtime/sys_{darwin,plan9}_386.s. Fixes #7164. LGTM=iant R=iant, aram, minux.ma, dave CC=golang-codereviews https://golang.org/cl/87920043
2014-04-15 11:45:39 -06:00
if(p->from.index == D_TLS)
p->from.scale = 2;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
return p;
}
// Append code to p to check for stack split.
// Appends to (does not overwrite) p.
// Assumes g is in CX.
// Returns last new instruction.
// On return, *jmpok is the instruction that should jump
// to the stack frame allocation if no split is needed.
static Prog*
cmd/ld: clear unused ctxt before morestack For non-closure functions, the context register is uninitialized on entry and will not be used, but morestack saves it and then the garbage collector treats it as live. This can be a source of memory leaks if the context register points at otherwise dead memory. Avoid this by introducing a parallel set of morestack functions that clear the context register, and use those for the non-closure functions. I hope this will help with some of the finalizer flakiness, but it probably won't. Fixes #7244. LGTM=dvyukov R=khr, dvyukov CC=golang-codereviews https://golang.org/cl/71030044
2014-03-04 11:53:08 -07:00
stacksplit(Link *ctxt, Prog *p, int32 framesize, int noctxt, Prog **jmpok)
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
{
Prog *q, *q1;
if(ctxt->debugstack) {
// 8l -K means check not only for stack
// overflow but stack underflow.
// On underflow, INT 3 (breakpoint).
// Underflow itself is rare but this also
// catches out-of-sync stack guard info.
p = appendp(ctxt, p);
p->as = ACMPL;
p->from.type = D_INDIR+D_CX;
p->from.offset = 4;
p->to.type = D_SP;
p = appendp(ctxt, p);
p->as = AJCC;
p->to.type = D_BRANCH;
p->to.offset = 4;
q1 = p;
p = appendp(ctxt, p);
p->as = AINT;
p->from.type = D_CONST;
p->from.offset = 3;
p = appendp(ctxt, p);
p->as = ANOP;
q1->pcond = p;
}
q1 = nil;
if(framesize <= StackSmall) {
// small stack: SP <= stackguard
// CMPL SP, stackguard
p = appendp(ctxt, p);
p->as = ACMPL;
p->from.type = D_SP;
p->to.type = D_INDIR+D_CX;
liblink, cmd/ld, runtime: remove stackguard1 Now that we've removed all the C code in runtime and the C compilers, there is no need to have a separate stackguard field to check for C code on Go stack. Remove field g.stackguard1 and rename g.stackguard0 to g.stackguard. Adjust liblink and cmd/ld as necessary. Change-Id: I54e75db5a93d783e86af5ff1a6cd497d669d8d33 Reviewed-on: https://go-review.googlesource.com/2144 Reviewed-by: Keith Randall <khr@golang.org>
2014-12-28 23:08:40 -07:00
p->to.offset = 2*ctxt->arch->ptrsize; // G.stackguard
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
} else if(framesize <= StackBig) {
// large stack: SP-framesize <= stackguard-StackSmall
// LEAL -(framesize-StackSmall)(SP), AX
// CMPL AX, stackguard
p = appendp(ctxt, p);
p->as = ALEAL;
p->from.type = D_INDIR+D_SP;
p->from.offset = -(framesize-StackSmall);
p->to.type = D_AX;
p = appendp(ctxt, p);
p->as = ACMPL;
p->from.type = D_AX;
p->to.type = D_INDIR+D_CX;
liblink, cmd/ld, runtime: remove stackguard1 Now that we've removed all the C code in runtime and the C compilers, there is no need to have a separate stackguard field to check for C code on Go stack. Remove field g.stackguard1 and rename g.stackguard0 to g.stackguard. Adjust liblink and cmd/ld as necessary. Change-Id: I54e75db5a93d783e86af5ff1a6cd497d669d8d33 Reviewed-on: https://go-review.googlesource.com/2144 Reviewed-by: Keith Randall <khr@golang.org>
2014-12-28 23:08:40 -07:00
p->to.offset = 2*ctxt->arch->ptrsize; // G.stackguard
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
} else {
// Such a large stack we need to protect against wraparound
// if SP is close to zero.
// SP-stackguard+StackGuard <= framesize + (StackGuard-StackSmall)
// The +StackGuard on both sides is required to keep the left side positive:
// SP is allowed to be slightly below stackguard. See stack.h.
//
// Preemption sets stackguard to StackPreempt, a very large value.
// That breaks the math above, so we have to check for that explicitly.
// MOVL stackguard, CX
// CMPL CX, $StackPreempt
// JEQ label-of-call-to-morestack
// LEAL StackGuard(SP), AX
// SUBL stackguard, AX
// CMPL AX, $(framesize+(StackGuard-StackSmall))
p = appendp(ctxt, p);
p->as = AMOVL;
p->from.type = D_INDIR+D_CX;
p->from.offset = 0;
liblink, cmd/ld, runtime: remove stackguard1 Now that we've removed all the C code in runtime and the C compilers, there is no need to have a separate stackguard field to check for C code on Go stack. Remove field g.stackguard1 and rename g.stackguard0 to g.stackguard. Adjust liblink and cmd/ld as necessary. Change-Id: I54e75db5a93d783e86af5ff1a6cd497d669d8d33 Reviewed-on: https://go-review.googlesource.com/2144 Reviewed-by: Keith Randall <khr@golang.org>
2014-12-28 23:08:40 -07:00
p->from.offset = 2*ctxt->arch->ptrsize; // G.stackguard
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
p->to.type = D_SI;
p = appendp(ctxt, p);
p->as = ACMPL;
p->from.type = D_SI;
p->to.type = D_CONST;
p->to.offset = (uint32)StackPreempt;
p = appendp(ctxt, p);
p->as = AJEQ;
p->to.type = D_BRANCH;
q1 = p;
p = appendp(ctxt, p);
p->as = ALEAL;
p->from.type = D_INDIR+D_SP;
p->from.offset = StackGuard;
p->to.type = D_AX;
p = appendp(ctxt, p);
p->as = ASUBL;
p->from.type = D_SI;
p->from.offset = 0;
p->to.type = D_AX;
p = appendp(ctxt, p);
p->as = ACMPL;
p->from.type = D_AX;
p->to.type = D_CONST;
p->to.offset = framesize+(StackGuard-StackSmall);
}
// common
p = appendp(ctxt, p);
p->as = AJHI;
p->to.type = D_BRANCH;
p->to.offset = 4;
q = p;
p = appendp(ctxt, p);
p->as = ACALL;
p->to.type = D_BRANCH;
liblink, cmd/ld, runtime: remove stackguard1 Now that we've removed all the C code in runtime and the C compilers, there is no need to have a separate stackguard field to check for C code on Go stack. Remove field g.stackguard1 and rename g.stackguard0 to g.stackguard. Adjust liblink and cmd/ld as necessary. Change-Id: I54e75db5a93d783e86af5ff1a6cd497d669d8d33 Reviewed-on: https://go-review.googlesource.com/2144 Reviewed-by: Keith Randall <khr@golang.org>
2014-12-28 23:08:40 -07:00
p->to.sym = ctxt->symmorestack[noctxt];
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
p = appendp(ctxt, p);
p->as = AJMP;
p->to.type = D_BRANCH;
p->pcond = ctxt->cursym->text->link;
if(q != nil)
q->pcond = p->link;
if(q1 != nil)
q1->pcond = q->link;
*jmpok = q;
return p;
}
static void xfol(Link*, Prog*, Prog**);
static void
follow(Link *ctxt, LSym *s)
{
Prog *firstp, *lastp;
ctxt->cursym = s;
firstp = ctxt->arch->prg();
lastp = firstp;
xfol(ctxt, s->text, &lastp);
lastp->link = nil;
s->text = firstp->link;
}
static int
nofollow(int a)
{
switch(a) {
case AJMP:
case ARET:
case AIRETL:
case AIRETW:
case AUNDEF:
return 1;
}
return 0;
}
static int
pushpop(int a)
{
switch(a) {
case APUSHL:
case APUSHFL:
case APUSHW:
case APUSHFW:
case APOPL:
case APOPFL:
case APOPW:
case APOPFW:
return 1;
}
return 0;
}
static int
relinv(int a)
{
switch(a) {
case AJEQ: return AJNE;
case AJNE: return AJEQ;
case AJLE: return AJGT;
case AJLS: return AJHI;
case AJLT: return AJGE;
case AJMI: return AJPL;
case AJGE: return AJLT;
case AJPL: return AJMI;
case AJGT: return AJLE;
case AJHI: return AJLS;
case AJCS: return AJCC;
case AJCC: return AJCS;
case AJPS: return AJPC;
case AJPC: return AJPS;
case AJOS: return AJOC;
case AJOC: return AJOS;
}
sysfatal("unknown relation: %s", anames8[a]);
return 0;
}
static void
xfol(Link *ctxt, Prog *p, Prog **last)
{
Prog *q;
int i;
build: annotations and modifications for c2go The main changes fall into a few patterns: 1. Replace #define with enum. 2. Add /*c2go */ comment giving effect of #define. This is necessary for function-like #defines and non-enum-able #defined constants. (Not all compilers handle negative or large enums.) 3. Add extra braces in struct initializer. (c2go does not implement the full rules.) This is enough to let c2go typecheck the source tree. There may be more changes once it is doing other semantic analyses. LGTM=minux, iant R=minux, dave, iant CC=golang-codereviews https://golang.org/cl/106860045
2014-07-02 13:41:29 -06:00
int a;
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
loop:
if(p == nil)
return;
if(p->as == AJMP)
if((q = p->pcond) != nil && q->as != ATEXT) {
/* mark instruction as done and continue layout at target of jump */
p->mark = 1;
p = q;
if(p->mark == 0)
goto loop;
}
if(p->mark) {
/*
* p goes here, but already used it elsewhere.
* copy up to 4 instructions or else branch to other copy.
*/
for(i=0,q=p; i<4; i++,q=q->link) {
if(q == nil)
break;
if(q == *last)
break;
a = q->as;
if(a == ANOP) {
i--;
continue;
}
if(nofollow(a) || pushpop(a))
break; // NOTE(rsc): arm does goto copy
if(q->pcond == nil || q->pcond->mark)
continue;
if(a == ACALL || a == ALOOP)
continue;
for(;;) {
if(p->as == ANOP) {
p = p->link;
continue;
}
q = copyp(ctxt, p);
p = p->link;
q->mark = 1;
(*last)->link = q;
*last = q;
if(q->as != a || q->pcond == nil || q->pcond->mark)
continue;
q->as = relinv(q->as);
p = q->pcond;
q->pcond = q->link;
q->link = p;
xfol(ctxt, q->link, last);
p = q->link;
if(p->mark)
return;
goto loop;
}
} /* */
q = ctxt->arch->prg();
q->as = AJMP;
q->lineno = p->lineno;
q->to.type = D_BRANCH;
q->to.offset = p->pc;
q->pcond = p;
p = q;
}
/* emit p */
p->mark = 1;
(*last)->link = p;
*last = p;
a = p->as;
/* continue loop with what comes after p */
if(nofollow(a))
return;
if(p->pcond != nil && a != ACALL) {
/*
* some kind of conditional branch.
* recurse to follow one path.
* continue loop on the other.
*/
if((q = brchain(ctxt, p->pcond)) != nil)
p->pcond = q;
if((q = brchain(ctxt, p->link)) != nil)
p->link = q;
if(p->from.type == D_CONST) {
if(p->from.offset == 1) {
/*
* expect conditional jump to be taken.
* rewrite so that's the fall-through case.
*/
p->as = relinv(a);
q = p->link;
p->link = p->pcond;
p->pcond = q;
}
} else {
q = p->link;
if(q->mark)
if(a != ALOOP) {
p->as = relinv(a);
p->link = p->pcond;
p->pcond = q;
}
}
xfol(ctxt, p->link, last);
if(p->pcond->mark)
return;
p = p->pcond;
goto loop;
}
p = p->link;
goto loop;
}
LinkArch link386 = {
.name = "386",
cmd/ld: move instruction selection + layout into compilers, assemblers - new object file reader/writer (liblink/objfile.c) - remove old object file writing routines - add pcdata iterator - remove all trace of "line number stack" and "path fragments" from object files, linker (!!!) - dwarf now writes a single "compilation unit" instead of one per package This CL disables the check for chains of no-split functions that could overflow the stack red zone. A future CL will attack the problem of reenabling that check (issue 6931). This CL is just the liblink and cmd/ld changes. There are minor associated adjustments in CL 37030045. Each depends on the other. R=golang-dev, dave, iant CC=golang-dev https://golang.org/cl/39680043
2013-12-16 10:51:58 -07:00
.thechar = '8',
liblink: support big-endian properly LGTM=rsc R=rsc, iant CC=golang-codereviews https://golang.org/cl/115300044
2014-08-05 22:25:41 -06:00
.endian = LittleEndian,
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
.addstacksplit = addstacksplit,
.assemble = span8,
.datasize = datasize,
.follow = follow,
.iscall = iscall,
.isdata = isdata,
.prg = prg,
.progedit = progedit,
.settextflag = settextflag,
.symtype = symtype,
.textflag = textflag,
.minlc = 1,
.ptrsize = 4,
all: final merge of NaCl tree This CL replays the following one CL from the rsc-go13nacl repo. This is the last replay CL: after this CL the main repo will have everything the rsc-go13nacl repo did. Changes made to the main repo after the rsc-go13nacl repo branched off probably mean that NaCl doesn't actually work after this CL, but all the code is now moved over and just needs to be redebugged. --- cmd/6l, cmd/8l, cmd/ld: support for Native Client See golang.org/s/go13nacl for design overview. This CL is publicly visible but not CC'ed to golang-dev, to avoid distracting from the preparation of the Go 1.2 release. This CL and the others will be checked into my rsc-go13nacl clone repo for now, and I will send CLs against the main repo early in the Go 1.3 development. R≡khr https://golang.org/cl/15750044 --- LGTM=bradfitz, dave, iant R=dave, bradfitz, iant CC=golang-codereviews https://golang.org/cl/69040044
2014-02-27 18:37:00 -07:00
.regsize = 4,
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
.D_ADDR = D_ADDR,
liblink: remove arch-specific constants from file format The relocation and automatic variable types were using arch-specific numbers. Introduce portable enumerations instead. To the best of my knowledge, these are the only arch-specific bits left in the new object file format. Remove now, before Go 1.3, because file formats are forever. LGTM=iant R=iant CC=golang-codereviews https://golang.org/cl/87670044
2014-04-14 13:54:20 -06:00
.D_AUTO = D_AUTO,
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
.D_BRANCH = D_BRANCH,
.D_CONST = D_CONST,
.D_EXTERN = D_EXTERN,
.D_FCONST = D_FCONST,
.D_NONE = D_NONE,
liblink: remove arch-specific constants from file format The relocation and automatic variable types were using arch-specific numbers. Introduce portable enumerations instead. To the best of my knowledge, these are the only arch-specific bits left in the new object file format. Remove now, before Go 1.3, because file formats are forever. LGTM=iant R=iant CC=golang-codereviews https://golang.org/cl/87670044
2014-04-14 13:54:20 -06:00
.D_PARAM = D_PARAM,
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
.D_SCONST = D_SCONST,
cmd/ld: move instruction selection + layout into compilers, assemblers - new object file reader/writer (liblink/objfile.c) - remove old object file writing routines - add pcdata iterator - remove all trace of "line number stack" and "path fragments" from object files, linker (!!!) - dwarf now writes a single "compilation unit" instead of one per package This CL disables the check for chains of no-split functions that could overflow the stack red zone. A future CL will attack the problem of reenabling that check (issue 6931). This CL is just the liblink and cmd/ld changes. There are minor associated adjustments in CL 37030045. Each depends on the other. R=golang-dev, dave, iant CC=golang-dev https://golang.org/cl/39680043
2013-12-16 10:51:58 -07:00
.D_STATIC = D_STATIC,
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
.ACALL = ACALL,
cmd/ld: move instruction selection + layout into compilers, assemblers - new object file reader/writer (liblink/objfile.c) - remove old object file writing routines - add pcdata iterator - remove all trace of "line number stack" and "path fragments" from object files, linker (!!!) - dwarf now writes a single "compilation unit" instead of one per package This CL disables the check for chains of no-split functions that could overflow the stack red zone. A future CL will attack the problem of reenabling that check (issue 6931). This CL is just the liblink and cmd/ld changes. There are minor associated adjustments in CL 37030045. Each depends on the other. R=golang-dev, dave, iant CC=golang-dev https://golang.org/cl/39680043
2013-12-16 10:51:58 -07:00
.ADATA = ADATA,
.AEND = AEND,
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
.AFUNCDATA = AFUNCDATA,
cmd/ld: move instruction selection + layout into compilers, assemblers - new object file reader/writer (liblink/objfile.c) - remove old object file writing routines - add pcdata iterator - remove all trace of "line number stack" and "path fragments" from object files, linker (!!!) - dwarf now writes a single "compilation unit" instead of one per package This CL disables the check for chains of no-split functions that could overflow the stack red zone. A future CL will attack the problem of reenabling that check (issue 6931). This CL is just the liblink and cmd/ld changes. There are minor associated adjustments in CL 37030045. Each depends on the other. R=golang-dev, dave, iant CC=golang-dev https://golang.org/cl/39680043
2013-12-16 10:51:58 -07:00
.AGLOBL = AGLOBL,
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
.AJMP = AJMP,
.ANOP = ANOP,
.APCDATA = APCDATA,
.ARET = ARET,
.ATEXT = ATEXT,
cmd/ld: move instruction selection + layout into compilers, assemblers - new object file reader/writer (liblink/objfile.c) - remove old object file writing routines - add pcdata iterator - remove all trace of "line number stack" and "path fragments" from object files, linker (!!!) - dwarf now writes a single "compilation unit" instead of one per package This CL disables the check for chains of no-split functions that could overflow the stack red zone. A future CL will attack the problem of reenabling that check (issue 6931). This CL is just the liblink and cmd/ld changes. There are minor associated adjustments in CL 37030045. Each depends on the other. R=golang-dev, dave, iant CC=golang-dev https://golang.org/cl/39680043
2013-12-16 10:51:58 -07:00
.ATYPE = ATYPE,
liblink: create new library based on linker code There is an enormous amount of code moving around in this CL, but the code is the same, and it is invoked in the same ways. This CL is preparation for the new linker structure, not the new structure itself. The new library's definition is in include/link.h. The main change is the use of a Link structure to hold all the linker-relevant state, replacing the smattering of global variables. The Link structure should both make it clearer which state must be carried around and make it possible to parallelize more easily later. The main body of the linker has moved into the architecture-independent cmd/ld directory. That includes the list of known header types, so the distinction between Hplan9x32 and Hplan9x64 is removed (no other header type distinguished 32- and 64-bit formats), and code for unused formats such as ipaq kernels has been deleted. The code being deleted from 5l, 6l, and 8l reappears in liblink or in ld. Because multiple files are being merged in the liblink directory, it is not possible to show the diffs nicely in hg. The Prog and Addr structures have been unified into an architecture-independent form and moved to link.h, where they will be shared by all tools: the assemblers, the compilers, and the linkers. The unification makes it possible to write architecture-independent traversal of Prog lists, among other benefits. The Sym structures cannot be unified: they are too fundamentally different between the linker and the compilers. Instead, liblink defines an LSym - a linker Sym - to be used in the Prog and Addr structures, and the linker now refers exclusively to LSyms. The compilers will keep using their own syms but will fill out the corresponding LSyms in the Prog and Addr structures. Although code from 5l, 6l, and 8l is now in a single library, the code has been arranged so that only one architecture needs to be linked into a particular program: 5l will not contain the code needed for x86 instruction layout, for example. The object file writing code in liblink/obj.c is from cmd/gc/obj.c. Preparation for golang.org/s/go13linker work. This CL does not build by itself. It depends on 35740044 and will be submitted at the same time. R=iant CC=golang-dev https://golang.org/cl/35790044
2013-12-08 20:49:37 -07:00
.AUSEFIELD = AUSEFIELD,
};
Copy Permalink