9fans - fans of the OS Plan 9 from Bell Labs
 help / color / mirror / Atom feed
From: Pavel Zholkover <paulzhol@gmail.com>
To: Fans of the OS Plan 9 from Bell Labs <9fans@9fans.net>
Subject: Re: [9fans] plan9 go output faults on 9vx but ok on cpu
Date: Fri, 14 Jan 2011 19:32:39 +0200	[thread overview]
Message-ID: <AANLkTim9Vw10LF626ZSx1QLBQeAr75oUBVgpQwZp-MFC@mail.gmail.com> (raw)
In-Reply-To: <AANLkTi=Z7+g+o-KYvWci2Tf27m5qgYvcAcTEvWmBa75a@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 997 bytes --]

Hi,

Attached a patch to get cross compiled go binaries to run on 9vx and
native Plan 9.

I tried the 9vx-0.12 binaries from http://swtch.com/9vx on a x86-64
linux, but they don't contain support for the CMPXCHG instruction.
After recompiling 9vx from rminnich branch, I can get hello.go to run
correctly both on 9vx and a Plan 9 terminal inside qemu-kvm with this
patch.

Unfortunately rfork seems to be broken (on the 9vx); The PUSHAL, POPAL
instructions are not supported so I had to comment out
PUSHAL
CALL	runtime·emptyfunc(SB)
POPAL
in src/runtime/plan9/386/sys.s, and then 9vx crashes:

% goroutines
9vx panic: kernel fault: signo=11 addr=0[ffffffffbf44a000] 1 eip=0
esp=7f37b5eecd70
aborting, to dump core.

I have no idea what is causing it to die, I've tried adding a bunch of
INT $3 to see how far in the flow I can get, but all I get this:

9vx panic: unknown trap/intr: 0x203
aborting, to dump core.

for an INT $3 just before
CALL	SI	// fn()

Pavel

[-- Attachment #2: golang_plan9_9vx.diff --]
[-- Type: text/x-patch, Size: 3739 bytes --]

diff -r ad4f2cbb85ef src/cmd/8l/obj.c
--- a/src/cmd/8l/obj.c	Thu Jan 13 20:05:08 2011 -0800
+++ b/src/cmd/8l/obj.c	Fri Jan 14 19:04:02 2011 +0200
@@ -185,6 +185,7 @@
 			INITRND = 0;
 		break;
 	case 2:	/* plan 9 */
+		tlsoffset = -8;
 		HEADR = 32L;
 		if(INITTEXT == -1)
 			INITTEXT = 4096+32;
diff -r ad4f2cbb85ef src/cmd/8l/pass.c
--- a/src/cmd/8l/pass.c	Thu Jan 13 20:05:08 2011 -0800
+++ b/src/cmd/8l/pass.c	Fri Jan 14 19:04:02 2011 +0200
@@ -243,6 +243,7 @@
 	Prog *p, *q;
 	Sym *s;
 	int32 vexit;
+	Sym *plan9_tos;
 
 	if(debug['v'])
 		Bprint(&bso, "%5.2f mkfwd\n", cputime());
@@ -253,6 +254,10 @@
 	Bflush(&bso);
 	s = lookup("exit", 0);
 	vexit = s->value;
+	
+	if(HEADTYPE == 2)
+		plan9_tos = lookup("_tos", 0);
+	
 	for(cursym = textp; cursym != nil; cursym = cursym->next) {
 		for(p = cursym->text; p != P; p = p->link) {
 			if(HEADTYPE == 10) {	// Windows
@@ -296,9 +301,15 @@
 			if(HEADTYPE == 2) {	// Plan 9
 				if(p->from.type == D_INDIR+D_GS
 				&& p->to.type >= D_AX && p->to.type <= D_DI) {
+					q = appendp(p);
+					q->from = p->from;
+					q->from.type = D_INDIR + p->to.type;
+					q->to = p->to;
+					q->as = p->as;
 					p->as = AMOVL;
-					p->from.type = D_ADDR+D_STATIC;
-					p->from.offset += 0xdfffefc0;
+					p->from.type = D_EXTERN;
+					p->from.sym = plan9_tos;
+					p->from.offset = 0;
 				}
 			}
 			if(p->as == ACALL || (p->as == AJMP && p->to.type != D_BRANCH)) {
@@ -382,6 +393,7 @@
 	int a;
 	Prog *pmorestack;
 	Sym *symmorestack;
+	Sym *plan9_tos;
 
 	pmorestack = P;
 	symmorestack = lookup("runtime.morestack", 0);
@@ -392,6 +404,9 @@
 		pmorestack = symmorestack->text;
 		symmorestack->text->from.scale |= NOSPLIT;
 	}
+	
+	if(HEADTYPE == 2)	
+		plan9_tos = lookup("_tos", 0);
 
 	for(cursym = textp; cursym != nil; cursym = cursym->next) {
 		if(cursym->text == nil || cursym->text->link == nil)
@@ -436,9 +451,15 @@
 			
 			case 2:	// Plan 9
 				p->as = AMOVL;
-				p->from.type = D_ADDR+D_STATIC;
-				p->from.offset = 0xdfffefc0;
+				p->from.type = D_EXTERN;
+				p->from.sym = plan9_tos;
 				p->to.type = D_CX;
+				
+				p = appendp(p);
+				p->as = AMOVL;
+				p->from.type = D_INDIR+D_CX;
+				p->from.offset = tlsoffset + 0;
+				p->to.type = D_CX;				
 				break;
 			
 			default:
diff -r ad4f2cbb85ef src/pkg/runtime/mkasmh.sh
--- a/src/pkg/runtime/mkasmh.sh	Thu Jan 13 20:05:08 2011 -0800
+++ b/src/pkg/runtime/mkasmh.sh	Fri Jan 14 19:04:02 2011 +0200
@@ -25,9 +25,9 @@
 		echo '#define	m(r)	4(r)'
 		;;
 	plan9)
-		echo '#define	get_tls(r)'
-		echo '#define	g(r)	0xdfffefc0'
-		echo '#define	m(r)	0xdfffefc4'
+		echo '#define	get_tls(r)	MOVL _tos(SB), r '
+		echo '#define	g(r)	-8(r)'
+		echo '#define	m(r)	-4(r)'
 		;;
 	linux)
 		# On Linux systems, what we call 0(GS) and 4(GS) for g and m
diff -r ad4f2cbb85ef src/pkg/runtime/plan9/386/defs.h
--- a/src/pkg/runtime/plan9/386/defs.h	Thu Jan 13 20:05:08 2011 -0800
+++ b/src/pkg/runtime/plan9/386/defs.h	Fri Jan 14 19:04:02 2011 +0200
@@ -1,1 +1,2 @@
 // nothing to see here
+#define tos_pid 48
diff -r ad4f2cbb85ef src/pkg/runtime/plan9/386/sys.s
--- a/src/pkg/runtime/plan9/386/sys.s	Thu Jan 13 20:05:08 2011 -0800
+++ b/src/pkg/runtime/plan9/386/sys.s	Fri Jan 14 19:04:02 2011 +0200
@@ -58,18 +58,19 @@
 	MOVL	BX, m(AX)
 
 	// Initialize AX from _tos->pid
-	MOVL	0xdfffeff8, AX
+	MOVL	_tos(SB), AX
+	MOVL	tos_pid(AX), AX
 	MOVL	AX, m_procid(BX)	// save pid as m->procid
-
+	
 	CALL	runtime·stackcheck(SB)	// smashes AX, CX
 	
 	MOVL	0(DX), DX	// paranoia; check they are not nil
 	MOVL	0(BX), BX
 	
 	// more paranoia; check that stack splitting code works
-	PUSHAL
-	CALL	runtime·emptyfunc(SB)
-	POPAL
+//	PUSHAL
+//	CALL	runtime·emptyfunc(SB)
+//	POPAL
 	
 	CALL	SI	// fn()
 	CALL	runtime·exit(SB)

[-- Attachment #3: hello.go --]
[-- Type: application/octet-stream, Size: 59 bytes --]

package main

func main() {
        println("Hello Go!")
}

[-- Attachment #4: goroutines.go --]
[-- Type: application/octet-stream, Size: 225 bytes --]

package main

import "runtime"


func until(id int) {
	for i := 0; i < 10; i++ {
		println("id:", id, i)
        	runtime.Gosched()	
	}
        
}

func main() {
	runtime.GOMAXPROCS(2)
        go until(2)
        until(1)	
}

[-- Attachment #5: goroutines --]
[-- Type: application/octet-stream, Size: 122007 bytes --]

[-- Attachment #6: hello --]
[-- Type: application/octet-stream, Size: 96564 bytes --]

  reply	other threads:[~2011-01-14 17:32 UTC|newest]

Thread overview: 36+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-01-11  7:54 Skip Tavakkolian
2011-01-11 14:11 ` erik quanstrom
2011-01-11 15:06   ` Devon H. O'Dell
2011-01-11 17:03   ` Charles Forsyth
2011-01-11 17:51     ` ron minnich
2011-01-11 19:35       ` Pavel Zholkover
2011-01-11 23:58       ` Charles Forsyth
2011-01-11 16:25 ` ron minnich
2011-01-12  8:32   ` Skip Tavakkolian
2011-01-12 16:12     ` andrey mirtchovski
2011-01-12 22:03       ` ron minnich
2011-01-13  5:54     ` ron minnich
2011-01-14 17:32       ` Pavel Zholkover [this message]
2011-01-15  2:00         ` Anthony Martin
2011-01-15  9:47           ` Pavel Zholkover
2011-01-15  9:01         ` kokamoto
2011-01-15  9:19           ` Pavel Zholkover
2011-01-15  9:30           ` lucio
2011-01-15  9:58             ` Pavel Zholkover
2011-01-15 10:29               ` lucio
2011-01-15 10:40               ` lucio
2011-01-15 11:25                 ` Pavel Zholkover
2011-01-16  1:51                   ` kokamoto
2011-01-16  2:04                     ` kokamoto
2011-01-16  6:21                       ` kokamoto
2011-01-17 12:03                         ` kokamoto
2011-01-17 15:20                         ` [9fans] how to make hardware work? sergey.kish
2011-01-16  2:06                     ` [9fans] plan9 go output faults on 9vx but ok on cpu Skip Tavakkolian
2011-01-11 17:03 ` Pavel Zholkover
2011-01-11 18:55   ` Anthony Martin
2011-01-11 19:06     ` Anthony Martin
2011-01-11 22:58     ` ron minnich
2011-01-12 22:31       ` Anthony Martin
2011-01-12 22:36         ` erik quanstrom
2011-01-12 23:28           ` Pavel Zholkover
2011-01-12 23:29         ` Charles Forsyth

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=AANLkTim9Vw10LF626ZSx1QLBQeAr75oUBVgpQwZp-MFC@mail.gmail.com \
    --to=paulzhol@gmail.com \
    --cc=9fans@9fans.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).