xv6-65oo2/user/usertests.c

2627 lines
51 KiB
C
Raw Normal View History

#include "kernel/param.h"
#include "kernel/types.h"
#include "kernel/stat.h"
#include "user/user.h"
#include "kernel/fs.h"
#include "kernel/fcntl.h"
#include "kernel/syscall.h"
#include "kernel/memlayout.h"
2019-07-22 22:08:52 +00:00
#include "kernel/riscv.h"
//
// Tests xv6 system calls. usertests without arguments runs them all
// and usertests <name> runs <name> test. The test runner creates for
// each test a process and based on the exit status of the process,
// the test runner reports "OK" or "FAILED". Some tests result in
// kernel printing usertrap messages, which can be ignored if test
// prints "OK".
//
#define BUFSZ (MAXOPBLOCKS+2)*BSIZE
char buf[BUFSZ];
2006-09-07 13:23:41 +00:00
char name[3];
// what if you pass ridiculous pointers to system calls
// that read user memory with copyin?
void
copyin(char *s)
{
uint64 addrs[] = { 0x80000000LL, 0xffffffffffffffff };
for(int ai = 0; ai < 2; ai++){
uint64 addr = addrs[ai];
int fd = open("copyin1", O_CREATE|O_WRONLY);
if(fd < 0){
printf("open(copyin1) failed\n");
exit(1);
}
int n = write(fd, (void*)addr, 8192);
if(n >= 0){
printf("write(fd, %p, 8192) returned %d, not -1\n", addr, n);
exit(1);
}
close(fd);
unlink("copyin1");
n = write(1, (char*)addr, 8192);
if(n > 0){
printf("write(1, %p, 8192) returned %d, not -1 or 0\n", addr, n);
exit(1);
}
int fds[2];
if(pipe(fds) < 0){
printf("pipe() failed\n");
exit(1);
}
n = write(fds[1], (char*)addr, 8192);
if(n > 0){
printf("write(pipe, %p, 8192) returned %d, not -1 or 0\n", addr, n);
exit(1);
}
close(fds[0]);
close(fds[1]);
}
}
// what if you pass ridiculous pointers to system calls
// that write user memory with copyout?
void
copyout(char *s)
{
uint64 addrs[] = { 0x80000000LL, 0xffffffffffffffff };
for(int ai = 0; ai < 2; ai++){
uint64 addr = addrs[ai];
int fd = open("README", 0);
if(fd < 0){
printf("open(README) failed\n");
exit(1);
}
int n = read(fd, (void*)addr, 8192);
if(n > 0){
printf("read(fd, %p, 8192) returned %d, not -1 or 0\n", addr, n);
exit(1);
}
close(fd);
int fds[2];
if(pipe(fds) < 0){
printf("pipe() failed\n");
exit(1);
}
n = write(fds[1], "x", 1);
if(n != 1){
printf("pipe write failed\n");
exit(1);
}
n = read(fds[0], (void*)addr, 8192);
if(n > 0){
printf("read(pipe, %p, 8192) returned %d, not -1 or 0\n", addr, n);
exit(1);
}
close(fds[0]);
close(fds[1]);
}
}
// what if you pass ridiculous string pointers to system calls?
void
copyinstr1(char *s)
{
uint64 addrs[] = { 0x80000000LL, 0xffffffffffffffff };
for(int ai = 0; ai < 2; ai++){
uint64 addr = addrs[ai];
int fd = open((char *)addr, O_CREATE|O_WRONLY);
if(fd >= 0){
printf("open(%p) returned %d, not -1\n", addr, fd);
exit(1);
}
}
}
// what if a string system call argument is exactly the size
// of the kernel buffer it is copied into, so that the null
// would fall just beyond the end of the kernel buffer?
void
copyinstr2(char *s)
{
char b[MAXPATH+1];
for(int i = 0; i < MAXPATH; i++)
b[i] = 'x';
b[MAXPATH] = '\0';
int ret = unlink(b);
if(ret != -1){
printf("unlink(%s) returned %d, not -1\n", b, ret);
exit(1);
}
int fd = open(b, O_CREATE | O_WRONLY);
if(fd != -1){
printf("open(%s) returned %d, not -1\n", b, fd);
exit(1);
}
ret = link(b, b);
if(ret != -1){
printf("link(%s, %s) returned %d, not -1\n", b, b, ret);
exit(1);
}
char *args[] = { "xx", 0 };
ret = exec(b, args);
if(ret != -1){
printf("exec(%s) returned %d, not -1\n", b, fd);
exit(1);
}
int pid = fork();
if(pid < 0){
printf("fork failed\n");
exit(1);
}
if(pid == 0){
static char big[PGSIZE+1];
for(int i = 0; i < PGSIZE; i++)
big[i] = 'x';
big[PGSIZE] = '\0';
char *args2[] = { big, big, big, 0 };
ret = exec("echo", args2);
if(ret != -1){
printf("exec(echo, BIG) returned %d, not -1\n", fd);
exit(1);
}
exit(747); // OK
}
int st = 0;
wait(&st);
if(st != 747){
printf("exec(echo, BIG) succeeded, should have failed\n");
exit(1);
}
}
// what if a string argument crosses over the end of last user page?
void
copyinstr3(char *s)
{
sbrk(8192);
uint64 top = (uint64) sbrk(0);
if((top % PGSIZE) != 0){
sbrk(PGSIZE - (top % PGSIZE));
}
top = (uint64) sbrk(0);
if(top % PGSIZE){
printf("oops\n");
exit(1);
}
char *b = (char *) (top - 1);
*b = 'x';
int ret = unlink(b);
if(ret != -1){
printf("unlink(%s) returned %d, not -1\n", b, ret);
exit(1);
}
int fd = open(b, O_CREATE | O_WRONLY);
if(fd != -1){
printf("open(%s) returned %d, not -1\n", b, fd);
exit(1);
}
ret = link(b, b);
if(ret != -1){
printf("link(%s, %s) returned %d, not -1\n", b, b, ret);
exit(1);
}
char *args[] = { "xx", 0 };
ret = exec(b, args);
if(ret != -1){
printf("exec(%s) returned %d, not -1\n", b, fd);
exit(1);
}
}
2020-07-16 15:38:08 +00:00
// test O_TRUNC.
void
truncate1(char *s)
{
char buf[32];
unlink("truncfile");
int fd1 = open("truncfile", O_CREATE|O_WRONLY|O_TRUNC);
write(fd1, "abcd", 4);
close(fd1);
int fd2 = open("truncfile", O_RDONLY);
int n = read(fd2, buf, sizeof(buf));
if(n != 4){
printf("%s: read %d bytes, wanted 4\n", s, n);
exit(1);
}
fd1 = open("truncfile", O_WRONLY|O_TRUNC);
int fd3 = open("truncfile", O_RDONLY);
n = read(fd3, buf, sizeof(buf));
if(n != 0){
printf("aaa fd3=%d\n", fd3);
printf("%s: read %d bytes, wanted 0\n", s, n);
exit(1);
}
n = read(fd2, buf, sizeof(buf));
if(n != 0){
printf("bbb fd2=%d\n", fd2);
printf("%s: read %d bytes, wanted 0\n", s, n);
exit(1);
}
write(fd1, "abcdef", 6);
n = read(fd3, buf, sizeof(buf));
if(n != 6){
printf("%s: read %d bytes, wanted 6\n", s, n);
exit(1);
}
n = read(fd2, buf, sizeof(buf));
if(n != 2){
printf("%s: read %d bytes, wanted 2\n", s, n);
exit(1);
}
unlink("truncfile");
close(fd1);
close(fd2);
close(fd3);
}
// write to an open FD whose file has just been truncated.
// this causes a write at an offset beyond the end of the file.
// such writes fail on xv6 (unlike POSIX) but at least
// they don't crash.
void
truncate2(char *s)
{
unlink("truncfile");
int fd1 = open("truncfile", O_CREATE|O_TRUNC|O_WRONLY);
write(fd1, "abcd", 4);
int fd2 = open("truncfile", O_TRUNC|O_WRONLY);
int n = write(fd1, "x", 1);
if(n != -1){
printf("%s: write returned %d, expected -1\n", s, n);
exit(1);
}
unlink("truncfile");
close(fd1);
close(fd2);
}
void
truncate3(char *s)
{
int pid, xstatus;
close(open("truncfile", O_CREATE|O_TRUNC|O_WRONLY));
pid = fork();
if(pid < 0){
printf("%s: fork failed\n", s);
exit(1);
}
if(pid == 0){
for(int i = 0; i < 100; i++){
char buf[32];
int fd = open("truncfile", O_WRONLY);
if(fd < 0){
printf("%s: open failed\n", s);
exit(1);
}
int n = write(fd, "1234567890", 10);
if(n != 10){
printf("%s: write got %d, expected 10\n", s, n);
exit(1);
}
close(fd);
fd = open("truncfile", O_RDONLY);
read(fd, buf, sizeof(buf));
close(fd);
}
exit(0);
}
for(int i = 0; i < 150; i++){
int fd = open("truncfile", O_CREATE|O_WRONLY|O_TRUNC);
if(fd < 0){
printf("%s: open failed\n", s);
exit(1);
}
int n = write(fd, "xxx", 3);
if(n != 3){
printf("%s: write got %d, expected 3\n", s, n);
exit(1);
}
close(fd);
}
wait(&xstatus);
unlink("truncfile");
exit(xstatus);
}
// does chdir() call iput(p->cwd) in a transaction?
void
iputtest(char *s)
{
if(mkdir("iputdir") < 0){
printf("%s: mkdir failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(chdir("iputdir") < 0){
printf("%s: chdir iputdir failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(unlink("../iputdir") < 0){
printf("%s: unlink ../iputdir failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(chdir("/") < 0){
printf("%s: chdir / failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
}
// does exit() call iput(p->cwd) in a transaction?
void
exitiputtest(char *s)
{
int pid, xstatus;
pid = fork();
if(pid < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid == 0){
if(mkdir("iputdir") < 0){
printf("%s: mkdir failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(chdir("iputdir") < 0){
printf("%s: child chdir failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(unlink("../iputdir") < 0){
printf("%s: unlink ../iputdir failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
exit(0);
}
wait(&xstatus);
exit(xstatus);
}
// does the error path in open() for attempt to write a
// directory call iput() in a transaction?
// needs a hacked kernel that pauses just after the namei()
// call in sys_open():
// if((ip = namei(path)) == 0)
// return -1;
// {
// int i;
// for(i = 0; i < 10000; i++)
// yield();
// }
void
openiputtest(char *s)
{
int pid, xstatus;
if(mkdir("oidir") < 0){
printf("%s: mkdir oidir failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
pid = fork();
if(pid < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid == 0){
int fd = open("oidir", O_RDWR);
if(fd >= 0){
printf("%s: open directory for write succeeded\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
exit(0);
}
sleep(1);
if(unlink("oidir") != 0){
printf("%s: unlink failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
wait(&xstatus);
exit(xstatus);
}
2006-09-07 13:23:41 +00:00
// simple file system tests
void
opentest(char *s)
2006-09-07 13:23:41 +00:00
{
int fd;
fd = open("echo", 0);
if(fd < 0){
printf("%s: open echo failed!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
fd = open("doesnotexist", 0);
if(fd >= 0){
printf("%s: open doesnotexist succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
}
void
writetest(char *s)
2006-09-07 13:23:41 +00:00
{
int fd;
int i;
enum { N=100, SZ=10 };
2006-09-07 13:23:41 +00:00
fd = open("small", O_CREATE|O_RDWR);
if(fd < 0){
printf("%s: error: creat small failed!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
for(i = 0; i < N; i++){
if(write(fd, "aaaaaaaaaa", SZ) != SZ){
printf("%s: error: write aa %d new file failed\n", i);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(write(fd, "bbbbbbbbbb", SZ) != SZ){
printf("%s: error: write bb %d new file failed\n", i);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
}
close(fd);
fd = open("small", O_RDONLY);
if(fd < 0){
printf("%s: error: open small failed!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
i = read(fd, buf, N*SZ*2);
if(i != N*SZ*2){
printf("%s: read failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
if(unlink("small") < 0){
printf("%s: unlink small failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
}
void
writebig(char *s)
2006-09-07 13:23:41 +00:00
{
int i, fd, n;
fd = open("big", O_CREATE|O_RDWR);
if(fd < 0){
printf("%s: error: creat big failed!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
for(i = 0; i < MAXFILE; i++){
((int*)buf)[0] = i;
if(write(fd, buf, BSIZE) != BSIZE){
printf("%s: error: write big file failed\n", i);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
}
close(fd);
fd = open("big", O_RDONLY);
if(fd < 0){
printf("%s: error: open big failed!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
n = 0;
for(;;){
i = read(fd, buf, BSIZE);
if(i == 0){
if(n == MAXFILE - 1){
printf("%s: read only %d blocks from big", n);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
break;
} else if(i != BSIZE){
printf("%s: read failed %d\n", i);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(((int*)buf)[0] != n){
printf("%s: read content of block %d is %d\n",
2006-09-07 13:23:41 +00:00
n, ((int*)buf)[0]);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
n++;
}
close(fd);
if(unlink("big") < 0){
printf("%s: unlink big failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
}
// many creates, followed by unlink test
2006-09-07 13:23:41 +00:00
void
createtest(char *s)
2006-09-07 13:23:41 +00:00
{
int i, fd;
enum { N=52 };
2006-09-07 13:23:41 +00:00
name[0] = 'a';
name[2] = '\0';
for(i = 0; i < N; i++){
2006-09-07 13:23:41 +00:00
name[1] = '0' + i;
fd = open(name, O_CREATE|O_RDWR);
close(fd);
}
name[0] = 'a';
name[2] = '\0';
for(i = 0; i < N; i++){
2006-09-07 13:23:41 +00:00
name[1] = '0' + i;
unlink(name);
}
}
void dirtest(char *s)
2006-09-07 13:23:41 +00:00
{
2019-08-27 17:13:03 +00:00
printf("mkdir test\n");
2006-09-07 13:23:41 +00:00
if(mkdir("dir0") < 0){
printf("%s: mkdir failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(chdir("dir0") < 0){
printf("%s: chdir dir0 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(chdir("..") < 0){
printf("%s: chdir .. failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(unlink("dir0") < 0){
printf("%s: unlink dir0 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
printf("%s: mkdir test ok\n");
2006-09-07 13:23:41 +00:00
}
void
exectest(char *s)
2006-09-07 13:23:41 +00:00
{
2019-09-19 19:14:52 +00:00
int fd, xstatus, pid;
char *echoargv[] = { "echo", "OK", 0 };
char buf[3];
unlink("echo-ok");
pid = fork();
if(pid < 0) {
printf("%s: fork failed\n", s);
exit(1);
}
if(pid == 0) {
close(1);
fd = open("echo-ok", O_CREATE|O_WRONLY);
if(fd < 0) {
printf("%s: create failed\n", s);
exit(1);
}
if(fd != 1) {
printf("%s: wrong fd\n", s);
exit(1);
}
if(exec("echo", echoargv) < 0){
printf("%s: exec echo failed\n", s);
exit(1);
}
// won't get to here
}
if (wait(&xstatus) != pid) {
printf("%s: wait failed!\n", s);
}
if(xstatus != 0)
exit(xstatus);
fd = open("echo-ok", O_RDONLY);
if(fd < 0) {
printf("%s: open failed\n", s);
exit(1);
}
if (read(fd, buf, 2) != 2) {
printf("%s: read failed\n", s);
exit(1);
}
unlink("echo-ok");
if(buf[0] == 'O' && buf[1] == 'K')
exit(0);
else {
printf("%s: wrong output\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
2019-09-19 19:14:52 +00:00
2006-09-07 13:23:41 +00:00
}
2006-06-27 14:35:53 +00:00
// simple fork and pipe read/write
2006-06-27 14:35:53 +00:00
void
pipe1(char *s)
2006-06-27 14:35:53 +00:00
{
int fds[2], pid, xstatus;
2007-08-10 17:17:42 +00:00
int seq, i, n, cc, total;
enum { N=5, SZ=1033 };
2006-08-29 19:06:37 +00:00
if(pipe(fds) != 0){
printf("%s: pipe() failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-08-29 19:06:37 +00:00
}
pid = fork();
2007-08-10 17:17:42 +00:00
seq = 0;
2006-06-27 14:35:53 +00:00
if(pid == 0){
close(fds[0]);
for(n = 0; n < N; n++){
for(i = 0; i < SZ; i++)
buf[i] = seq++;
if(write(fds[1], buf, SZ) != SZ){
printf("%s: pipe1 oops 1\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
}
exit(0);
2006-08-29 19:06:37 +00:00
} else if(pid > 0){
close(fds[1]);
total = 0;
cc = 1;
2006-07-17 01:25:22 +00:00
while((n = read(fds[0], buf, cc)) > 0){
for(i = 0; i < n; i++){
if((buf[i] & 0xff) != (seq++ & 0xff)){
printf("%s: pipe1 oops 2\n", s);
return;
}
}
total += n;
cc = cc * 2;
if(cc > sizeof(buf))
cc = sizeof(buf);
2006-06-27 14:35:53 +00:00
}
if(total != N * SZ){
printf("%s: pipe1 oops 3 total %d\n", total);
2019-09-11 14:04:40 +00:00
exit(1);
2011-08-31 00:50:19 +00:00
}
close(fds[0]);
wait(&xstatus);
exit(xstatus);
2006-08-29 19:06:37 +00:00
} else {
printf("%s: fork() failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-06-27 14:35:53 +00:00
}
}
// meant to be run w/ at most two CPUs
void
preempt(char *s)
{
int pid1, pid2, pid3;
int pfds[2];
pid1 = fork();
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
if(pid1 < 0) {
printf("%s: fork failed");
2019-09-11 14:04:40 +00:00
exit(1);
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
}
if(pid1 == 0)
2006-07-17 01:25:22 +00:00
for(;;)
;
2006-09-06 17:27:19 +00:00
pid2 = fork();
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
if(pid2 < 0) {
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
}
if(pid2 == 0)
2006-07-17 01:25:22 +00:00
for(;;)
;
pipe(pfds);
pid3 = fork();
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
if(pid3 < 0) {
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
}
if(pid3 == 0){
close(pfds[0]);
if(write(pfds[1], "x", 1) != 1)
printf("%s: preempt write error");
close(pfds[1]);
2006-07-17 01:25:22 +00:00
for(;;)
;
}
close(pfds[1]);
if(read(pfds[0], buf, sizeof(buf)) != 1){
printf("%s: preempt read error");
return;
}
close(pfds[0]);
2019-08-27 17:13:03 +00:00
printf("kill... ");
kill(pid1);
kill(pid2);
kill(pid3);
2019-08-27 17:13:03 +00:00
printf("wait... ");
wait(0);
wait(0);
wait(0);
}
// try to find any races between exit and wait
void
exitwait(char *s)
{
int i, pid;
for(i = 0; i < 100; i++){
pid = fork();
if(pid < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid){
int xstate;
if(wait(&xstate) != pid){
printf("%s: wait wrong pid\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(i != xstate) {
printf("%s: wait wrong exit status\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
} else {
exit(i);
}
}
}
// try to find races in the reparenting
// code that handles a parent exiting
// when it still has live children.
void
reparent(char *s)
{
int master_pid = getpid();
for(int i = 0; i < 200; i++){
int pid = fork();
if(pid < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid){
if(wait(0) != pid){
printf("%s: wait wrong pid\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
} else {
int pid2 = fork();
if(pid2 < 0){
kill(master_pid);
2019-09-11 14:04:40 +00:00
exit(1);
}
exit(0);
}
}
exit(0);
}
// what if two children exit() at the same time?
void
twochildren(char *s)
{
for(int i = 0; i < 1000; i++){
int pid1 = fork();
if(pid1 < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid1 == 0){
exit(0);
} else {
int pid2 = fork();
if(pid2 < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid2 == 0){
exit(0);
} else {
wait(0);
wait(0);
}
}
}
}
// concurrent forks to try to expose locking bugs.
void
forkfork(char *s)
{
enum { N=2 };
for(int i = 0; i < N; i++){
int pid = fork();
if(pid < 0){
printf("%s: fork failed", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid == 0){
for(int j = 0; j < 200; j++){
int pid1 = fork();
if(pid1 < 0){
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid1 == 0){
exit(0);
}
wait(0);
}
exit(0);
}
}
int xstatus;
for(int i = 0; i < N; i++){
wait(&xstatus);
if(xstatus != 0) {
printf("%s: fork in child failed", s);
exit(1);
}
}
}
void
forkforkfork(char *s)
{
unlink("stopforking");
int pid = fork();
if(pid < 0){
printf("%s: fork failed", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid == 0){
while(1){
int fd = open("stopforking", 0);
if(fd >= 0){
exit(0);
}
if(fork() < 0){
close(open("stopforking", O_CREATE|O_RDWR));
}
}
exit(0);
}
sleep(20); // two seconds
close(open("stopforking", O_CREATE|O_RDWR));
wait(0);
sleep(10); // one second
}
// regression test. does reparent() violate the parent-then-child
// locking order when giving away a child to init, so that exit()
// deadlocks against init's wait()? also used to trigger a "panic:
// release" due to exit() releasing a different p->parent->lock than
// it acquired.
void
reparent2(char *s)
{
for(int i = 0; i < 800; i++){
int pid1 = fork();
if(pid1 < 0){
printf("fork failed\n");
exit(1);
}
if(pid1 == 0){
fork();
fork();
exit(0);
}
wait(0);
}
exit(0);
}
2019-09-19 19:22:45 +00:00
// allocate all mem, free it, and allocate again
2006-08-24 19:24:36 +00:00
void
mem(char *s)
2006-08-24 19:24:36 +00:00
{
void *m1, *m2;
int pid;
2006-08-24 19:24:36 +00:00
2006-08-29 19:06:37 +00:00
if((pid = fork()) == 0){
m1 = 0;
while((m2 = malloc(10001)) != 0){
*(char**)m2 = m1;
2006-08-29 19:06:37 +00:00
m1 = m2;
}
while(m1){
2006-09-06 17:27:19 +00:00
m2 = *(char**)m1;
2006-08-29 19:06:37 +00:00
free(m1);
m1 = m2;
}
m1 = malloc(1024*20);
if(m1 == 0){
printf("couldn't allocate mem?!!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-08-29 19:06:37 +00:00
}
2006-08-24 19:24:36 +00:00
free(m1);
exit(0);
2006-08-29 19:06:37 +00:00
} else {
int xstatus;
wait(&xstatus);
exit(xstatus);
2006-08-24 19:24:36 +00:00
}
}
2006-09-07 13:23:41 +00:00
// More file system tests
// two processes write to the same file descriptor
// is the offset shared? does inode locking work?
void
sharedfd(char *s)
2006-09-07 13:23:41 +00:00
{
int fd, pid, i, n, nc, np;
enum { N = 1000, SZ=10};
char buf[SZ];
2006-09-07 13:23:41 +00:00
unlink("sharedfd");
fd = open("sharedfd", O_CREATE|O_RDWR);
if(fd < 0){
printf("%s: cannot open sharedfd for writing", s);
exit(1);
2006-09-07 13:23:41 +00:00
}
pid = fork();
memset(buf, pid==0?'c':'p', sizeof(buf));
for(i = 0; i < N; i++){
2006-09-07 13:23:41 +00:00
if(write(fd, buf, sizeof(buf)) != sizeof(buf)){
printf("%s: write sharedfd failed\n", s);
exit(1);
2006-09-07 13:23:41 +00:00
}
}
if(pid == 0) {
exit(0);
} else {
int xstatus;
wait(&xstatus);
if(xstatus != 0)
exit(xstatus);
}
2006-09-07 13:23:41 +00:00
close(fd);
fd = open("sharedfd", 0);
if(fd < 0){
printf("%s: cannot open sharedfd for reading\n", s);
exit(1);
2006-09-07 13:23:41 +00:00
}
nc = np = 0;
while((n = read(fd, buf, sizeof(buf))) > 0){
for(i = 0; i < sizeof(buf); i++){
if(buf[i] == 'c')
nc++;
if(buf[i] == 'p')
np++;
}
}
close(fd);
unlink("sharedfd");
if(nc == N*SZ && np == N*SZ){
exit(0);
2011-08-31 00:50:19 +00:00
} else {
printf("%s: nc/np test fails\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2011-08-31 00:50:19 +00:00
}
2006-09-07 13:23:41 +00:00
}
// four processes write different files at the same
2006-09-07 13:23:41 +00:00
// time, to test block allocation.
void
fourfiles(char *s)
2006-09-07 13:23:41 +00:00
{
int fd, pid, i, j, n, total, pi;
char *names[] = { "f0", "f1", "f2", "f3" };
2006-09-07 13:23:41 +00:00
char *fname;
enum { N=12, NCHILD=4, SZ=500 };
for(pi = 0; pi < NCHILD; pi++){
fname = names[pi];
unlink(fname);
2006-09-07 13:23:41 +00:00
pid = fork();
if(pid < 0){
printf("fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
2006-09-07 13:23:41 +00:00
if(pid == 0){
fd = open(fname, O_CREATE | O_RDWR);
if(fd < 0){
printf("create failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
memset(buf, '0'+pi, SZ);
for(i = 0; i < N; i++){
if((n = write(fd, buf, SZ)) != SZ){
2019-08-27 17:13:03 +00:00
printf("write failed %d\n", n);
2019-09-11 14:04:40 +00:00
exit(1);
}
}
exit(0);
2006-09-07 13:23:41 +00:00
}
}
int xstatus;
for(pi = 0; pi < NCHILD; pi++){
wait(&xstatus);
if(xstatus != 0)
exit(xstatus);
}
2006-09-07 13:23:41 +00:00
for(i = 0; i < NCHILD; i++){
fname = names[i];
fd = open(fname, 0);
2006-09-07 13:23:41 +00:00
total = 0;
while((n = read(fd, buf, sizeof(buf))) > 0){
for(j = 0; j < n; j++){
if(buf[j] != '0'+i){
printf("wrong char\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
}
total += n;
}
close(fd);
if(total != N*SZ){
2019-08-27 17:13:03 +00:00
printf("wrong length %d\n", total);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
unlink(fname);
2006-09-07 13:23:41 +00:00
}
}
// four processes create and delete different files in same directory
2006-09-07 13:23:41 +00:00
void
createdelete(char *s)
2006-09-07 13:23:41 +00:00
{
2018-09-29 12:29:34 +00:00
enum { N = 20, NCHILD=4 };
int pid, i, fd, pi;
2006-09-07 13:23:41 +00:00
char name[32];
2018-09-29 12:29:34 +00:00
for(pi = 0; pi < NCHILD; pi++){
pid = fork();
if(pid < 0){
printf("fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(pid == 0){
name[0] = 'p' + pi;
name[2] = '\0';
for(i = 0; i < N; i++){
name[1] = '0' + i;
fd = open(name, O_CREATE | O_RDWR);
if(fd < 0){
printf("%s: create failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
close(fd);
if(i > 0 && (i % 2 ) == 0){
name[1] = '0' + (i / 2);
if(unlink(name) < 0){
printf("%s: unlink failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
}
2006-09-07 13:23:41 +00:00
}
exit(0);
2006-09-07 13:23:41 +00:00
}
}
int xstatus;
2018-09-29 12:29:34 +00:00
for(pi = 0; pi < NCHILD; pi++){
wait(&xstatus);
if(xstatus != 0)
exit(1);
}
2006-09-07 13:23:41 +00:00
name[0] = name[1] = name[2] = 0;
2007-08-10 17:17:42 +00:00
for(i = 0; i < N; i++){
2018-09-29 12:29:34 +00:00
for(pi = 0; pi < NCHILD; pi++){
name[0] = 'p' + pi;
name[1] = '0' + i;
fd = open(name, 0);
if((i == 0 || i >= N/2) && fd < 0){
printf("%s: oops createdelete %s didn't exist\n", s, name);
2019-09-11 14:04:40 +00:00
exit(1);
} else if((i >= 1 && i < N/2) && fd >= 0){
printf("%s: oops createdelete %s did exist\n", s, name);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(fd >= 0)
close(fd);
2006-09-07 13:23:41 +00:00
}
}
2007-08-10 17:17:42 +00:00
for(i = 0; i < N; i++){
2018-09-29 12:29:34 +00:00
for(pi = 0; pi < NCHILD; pi++){
name[0] = 'p' + i;
name[1] = '0' + i;
unlink(name);
}
2006-09-07 13:23:41 +00:00
}
}
// can I unlink a file and still read it?
void
unlinkread(char *s)
2006-09-07 13:23:41 +00:00
{
enum { SZ = 5 };
2006-09-07 13:23:41 +00:00
int fd, fd1;
fd = open("unlinkread", O_CREATE | O_RDWR);
if(fd < 0){
printf("%s: create unlinkread failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
write(fd, "hello", SZ);
2006-09-07 13:23:41 +00:00
close(fd);
fd = open("unlinkread", O_RDWR);
if(fd < 0){
printf("%s: open unlinkread failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(unlink("unlinkread") != 0){
printf("%s: unlink unlinkread failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
fd1 = open("unlinkread", O_CREATE | O_RDWR);
write(fd1, "yyy", 3);
close(fd1);
if(read(fd, buf, sizeof(buf)) != SZ){
printf("%s: unlinkread read failed", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(buf[0] != 'h'){
printf("%s: unlinkread wrong data\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(write(fd, buf, 10) != 10){
printf("%s: unlinkread write failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
unlink("unlinkread");
}
void
linktest(char *s)
2006-09-07 13:23:41 +00:00
{
enum { SZ = 5 };
2006-09-07 13:23:41 +00:00
int fd;
unlink("lf1");
unlink("lf2");
fd = open("lf1", O_CREATE|O_RDWR);
if(fd < 0){
printf("%s: create lf1 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(write(fd, "hello", SZ) != SZ){
printf("%s: write lf1 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
if(link("lf1", "lf2") < 0){
printf("%s: link lf1 lf2 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
unlink("lf1");
if(open("lf1", 0) >= 0){
printf("%s: unlinked lf1 but it is still there!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
fd = open("lf2", 0);
if(fd < 0){
printf("%s: open lf2 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(read(fd, buf, sizeof(buf)) != SZ){
printf("%s: read lf2 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
if(link("lf2", "lf2") >= 0){
printf("%s: link lf2 lf2 succeeded! oops\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
unlink("lf2");
if(link("lf2", "lf1") >= 0){
printf("%s: link non-existant succeeded! oops\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(link(".", "lf1") >= 0){
printf("%s: link . lf1 succeeded! oops\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
}
// test concurrent create/link/unlink of the same file
2006-09-07 13:23:41 +00:00
void
concreate(char *s)
2006-09-07 13:23:41 +00:00
{
enum { N = 40 };
2006-09-07 13:23:41 +00:00
char file[3];
int i, pid, n, fd;
char fa[N];
2006-09-07 13:23:41 +00:00
struct {
ushort inum;
char name[DIRSIZ];
2006-09-07 13:23:41 +00:00
} de;
file[0] = 'C';
file[2] = '\0';
for(i = 0; i < N; i++){
2006-09-07 13:23:41 +00:00
file[1] = '0' + i;
unlink(file);
pid = fork();
if(pid && (i % 3) == 1){
link("C0", file);
} else if(pid == 0 && (i % 5) == 1){
link("C0", file);
} else {
fd = open(file, O_CREATE | O_RDWR);
if(fd < 0){
2019-08-27 17:13:03 +00:00
printf("concreate create %s failed\n", file);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
}
if(pid == 0) {
exit(0);
} else {
int xstatus;
wait(&xstatus);
if(xstatus != 0)
exit(1);
}
2006-09-07 13:23:41 +00:00
}
memset(fa, 0, sizeof(fa));
fd = open(".", 0);
n = 0;
while(read(fd, &de, sizeof(de)) > 0){
if(de.inum == 0)
continue;
if(de.name[0] == 'C' && de.name[2] == '\0'){
i = de.name[1] - '0';
if(i < 0 || i >= sizeof(fa)){
printf("%s: concreate weird file %s\n", s, de.name);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(fa[i]){
printf("%s: concreate duplicate file %s\n", s, de.name);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
fa[i] = 1;
n++;
}
}
close(fd);
if(n != N){
printf("%s: concreate not enough files in directory listing\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
for(i = 0; i < N; i++){
2006-09-07 13:23:41 +00:00
file[1] = '0' + i;
pid = fork();
if(pid < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(((i % 3) == 0 && pid == 0) ||
((i % 3) == 1 && pid != 0)){
close(open(file, 0));
close(open(file, 0));
close(open(file, 0));
close(open(file, 0));
close(open(file, 0));
close(open(file, 0));
} else {
unlink(file);
unlink(file);
unlink(file);
unlink(file);
unlink(file);
unlink(file);
}
if(pid == 0)
exit(0);
else
wait(0);
2006-09-07 13:23:41 +00:00
}
}
// another concurrent link/unlink/create test,
// to look for deadlocks.
void
linkunlink(char *s)
{
int pid, i;
unlink("x");
pid = fork();
if(pid < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
unsigned int x = (pid ? 1 : 97);
for(i = 0; i < 100; i++){
x = x * 1103515245 + 12345;
if((x % 3) == 0){
close(open("x", O_RDWR | O_CREATE));
} else if((x % 3) == 1){
link("cat", "x");
} else {
unlink("x");
}
}
if(pid)
wait(0);
else
exit(0);
}
2006-09-07 13:23:41 +00:00
// directory that uses indirect blocks
void
bigdir(char *s)
2006-09-07 13:23:41 +00:00
{
enum { N = 500 };
2006-09-07 13:23:41 +00:00
int i, fd;
char name[10];
unlink("bd");
fd = open("bd", O_CREATE);
if(fd < 0){
printf("%s: bigdir create failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
for(i = 0; i < N; i++){
2006-09-07 13:23:41 +00:00
name[0] = 'x';
name[1] = '0' + (i / 64);
name[2] = '0' + (i % 64);
name[3] = '\0';
if(link("bd", name) != 0){
printf("%s: bigdir link(bd, %s) failed\n", s, name);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
}
unlink("bd");
for(i = 0; i < N; i++){
2006-09-07 13:23:41 +00:00
name[0] = 'x';
name[1] = '0' + (i / 64);
name[2] = '0' + (i % 64);
name[3] = '\0';
if(unlink(name) != 0){
printf("%s: bigdir unlink failed", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
}
}
void
subdir(char *s)
2006-09-07 13:23:41 +00:00
{
int fd, cc;
unlink("ff");
if(mkdir("dd") != 0){
printf("%s: mkdir dd failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
fd = open("dd/ff", O_CREATE | O_RDWR);
if(fd < 0){
printf("%s: create dd/ff failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
write(fd, "ff", 2);
close(fd);
2007-08-22 02:21:22 +00:00
if(unlink("dd") >= 0){
printf("%s: unlink dd (non-empty dir) succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-22 02:21:22 +00:00
}
2006-09-07 13:23:41 +00:00
if(mkdir("/dd/dd") != 0){
printf("subdir mkdir dd/dd failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
fd = open("dd/dd/ff", O_CREATE | O_RDWR);
if(fd < 0){
printf("%s: create dd/dd/ff failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
write(fd, "FF", 2);
close(fd);
fd = open("dd/dd/../ff", 0);
if(fd < 0){
printf("%s: open dd/dd/../ff failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
cc = read(fd, buf, sizeof(buf));
if(cc != 2 || buf[0] != 'f'){
printf("%s: dd/dd/../ff wrong content\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
if(link("dd/dd/ff", "dd/dd/ffff") != 0){
printf("link dd/dd/ff dd/dd/ffff failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(unlink("dd/dd/ff") != 0){
printf("%s: unlink dd/dd/ff failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
2007-08-22 02:21:22 +00:00
if(open("dd/dd/ff", O_RDONLY) >= 0){
printf("%s: open (unlinked) dd/dd/ff succeeded\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-22 02:21:22 +00:00
}
2006-09-07 13:23:41 +00:00
if(chdir("dd") != 0){
printf("%s: chdir dd failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(chdir("dd/../../dd") != 0){
printf("%s: chdir dd/../../dd failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
2007-08-22 02:21:22 +00:00
if(chdir("dd/../../../dd") != 0){
printf("chdir dd/../../dd failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-22 02:21:22 +00:00
}
2006-09-07 13:23:41 +00:00
if(chdir("./..") != 0){
printf("%s: chdir ./.. failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
fd = open("dd/dd/ffff", 0);
if(fd < 0){
printf("%s: open dd/dd/ffff failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(read(fd, buf, sizeof(buf)) != 2){
printf("%s: read dd/dd/ffff wrong len\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
2007-08-22 02:21:22 +00:00
if(open("dd/dd/ff", O_RDONLY) >= 0){
printf("%s: open (unlinked) dd/dd/ff succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(open("dd/ff/ff", O_CREATE|O_RDWR) >= 0){
printf("%s: create dd/ff/ff succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(open("dd/xx/ff", O_CREATE|O_RDWR) >= 0){
printf("%s: create dd/xx/ff succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(open("dd", O_CREATE) >= 0){
printf("%s: create dd succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(open("dd", O_RDWR) >= 0){
printf("%s: open dd rdwr succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(open("dd", O_WRONLY) >= 0){
printf("%s: open dd wronly succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(link("dd/ff/ff", "dd/dd/xx") == 0){
printf("%s: link dd/ff/ff dd/dd/xx succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(link("dd/xx/ff", "dd/dd/xx") == 0){
printf("%s: link dd/xx/ff dd/dd/xx succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(link("dd/ff", "dd/dd/ffff") == 0){
printf("%s: link dd/ff dd/dd/ffff succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(mkdir("dd/ff/ff") == 0){
printf("%s: mkdir dd/ff/ff succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(mkdir("dd/xx/ff") == 0){
printf("%s: mkdir dd/xx/ff succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(mkdir("dd/dd/ffff") == 0){
printf("%s: mkdir dd/dd/ffff succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(unlink("dd/xx/ff") == 0){
printf("%s: unlink dd/xx/ff succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(unlink("dd/ff/ff") == 0){
printf("%s: unlink dd/ff/ff succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(chdir("dd/ff") == 0){
printf("%s: chdir dd/ff succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(chdir("dd/xx") == 0){
printf("%s: chdir dd/xx succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(unlink("dd/dd/ffff") != 0){
printf("%s: unlink dd/dd/ff failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(unlink("dd/ff") != 0){
printf("%s: unlink dd/ff failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
2007-08-22 02:21:22 +00:00
if(unlink("dd") == 0){
printf("%s: unlink non-empty dd succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-22 02:21:22 +00:00
}
if(unlink("dd/dd") < 0){
printf("%s: unlink dd/dd failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-22 02:21:22 +00:00
}
if(unlink("dd") < 0){
printf("%s: unlink dd failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-22 02:21:22 +00:00
}
2006-09-07 13:23:41 +00:00
}
// test writes that are larger than the log.
void
bigwrite(char *s)
{
int fd, sz;
unlink("bigwrite");
for(sz = 499; sz < (MAXOPBLOCKS+2)*BSIZE; sz += 471){
fd = open("bigwrite", O_CREATE | O_RDWR);
if(fd < 0){
printf("%s: cannot create bigwrite\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
int i;
for(i = 0; i < 2; i++){
int cc = write(fd, buf, sz);
if(cc != sz){
printf("%s: write(%d) ret %d\n", s, sz, cc);
2019-09-11 14:04:40 +00:00
exit(1);
}
}
close(fd);
unlink("bigwrite");
}
}
2006-09-07 13:23:41 +00:00
void
bigfile(char *s)
2006-09-07 13:23:41 +00:00
{
enum { N = 20, SZ=600 };
2006-09-07 13:23:41 +00:00
int fd, i, total, cc;
unlink("bigfile.dat");
fd = open("bigfile.dat", O_CREATE | O_RDWR);
2006-09-07 13:23:41 +00:00
if(fd < 0){
printf("%s: cannot create bigfile", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
for(i = 0; i < N; i++){
memset(buf, i, SZ);
if(write(fd, buf, SZ) != SZ){
printf("%s: write bigfile failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
}
close(fd);
fd = open("bigfile.dat", 0);
2006-09-07 13:23:41 +00:00
if(fd < 0){
printf("%s: cannot open bigfile\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
total = 0;
for(i = 0; ; i++){
cc = read(fd, buf, SZ/2);
2006-09-07 13:23:41 +00:00
if(cc < 0){
printf("%s: read bigfile failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(cc == 0)
break;
if(cc != SZ/2){
printf("%s: short read bigfile\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(buf[0] != i/2 || buf[SZ/2-1] != i/2){
printf("%s: read bigfile wrong data\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
total += cc;
}
close(fd);
if(total != N*SZ){
printf("%s: read bigfile wrong total\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
unlink("bigfile.dat");
2006-09-07 13:23:41 +00:00
}
void
fourteen(char *s)
2006-09-07 13:23:41 +00:00
{
int fd;
2007-08-10 17:53:09 +00:00
// DIRSIZ is 14.
2006-09-07 13:23:41 +00:00
if(mkdir("12345678901234") != 0){
printf("%s: mkdir 12345678901234 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(mkdir("12345678901234/123456789012345") != 0){
printf("%s: mkdir 12345678901234/123456789012345 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
fd = open("123456789012345/123456789012345/123456789012345", O_CREATE);
if(fd < 0){
printf("%s: create 123456789012345/123456789012345/123456789012345 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
fd = open("12345678901234/12345678901234/12345678901234", 0);
if(fd < 0){
printf("%s: open 12345678901234/12345678901234/12345678901234 failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
close(fd);
if(mkdir("12345678901234/12345678901234") == 0){
printf("%s: mkdir 12345678901234/12345678901234 succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
if(mkdir("123456789012345/12345678901234") == 0){
printf("%s: mkdir 12345678901234/123456789012345 succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2006-09-07 13:23:41 +00:00
}
// clean up
unlink("123456789012345/12345678901234");
unlink("12345678901234/12345678901234");
unlink("12345678901234/12345678901234/12345678901234");
unlink("123456789012345/123456789012345/123456789012345");
unlink("12345678901234/123456789012345");
unlink("12345678901234");
2006-09-07 13:23:41 +00:00
}
2007-08-10 17:53:09 +00:00
void
rmdot(char *s)
2007-08-10 17:53:09 +00:00
{
if(mkdir("dots") != 0){
printf("%s: mkdir dots failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-10 17:53:09 +00:00
}
if(chdir("dots") != 0){
printf("%s: chdir dots failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-10 17:53:09 +00:00
}
if(unlink(".") == 0){
printf("%s: rm . worked!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-10 17:53:09 +00:00
}
if(unlink("..") == 0){
printf("%s: rm .. worked!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-10 17:53:09 +00:00
}
if(chdir("/") != 0){
printf("%s: chdir / failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-10 17:53:09 +00:00
}
if(unlink("dots/.") == 0){
printf("%s: unlink dots/. worked!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-10 17:53:09 +00:00
}
if(unlink("dots/..") == 0){
printf("%s: unlink dots/.. worked!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-10 17:53:09 +00:00
}
if(unlink("dots") != 0){
printf("%s: unlink dots failed!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-10 17:53:09 +00:00
}
}
2007-08-24 14:56:17 +00:00
void
dirfile(char *s)
2007-08-24 14:56:17 +00:00
{
int fd;
fd = open("dirfile", O_CREATE);
if(fd < 0){
printf("%s: create dirfile failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
close(fd);
if(chdir("dirfile") == 0){
printf("%s: chdir dirfile succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
fd = open("dirfile/xx", 0);
if(fd >= 0){
printf("%s: create dirfile/xx succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
fd = open("dirfile/xx", O_CREATE);
if(fd >= 0){
printf("%s: create dirfile/xx succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
if(mkdir("dirfile/xx") == 0){
printf("%s: mkdir dirfile/xx succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
if(unlink("dirfile/xx") == 0){
printf("%s: unlink dirfile/xx succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
if(link("README", "dirfile/xx") == 0){
printf("%s: link to dirfile/xx succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
if(unlink("dirfile") != 0){
printf("%s: unlink dirfile failed!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
fd = open(".", O_RDWR);
if(fd >= 0){
printf("%s: open . for writing succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
fd = open(".", 0);
if(write(fd, "x", 1) > 0){
printf("%s: write . succeeded!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
close(fd);
}
// test that iput() is called at the end of _namei().
// also tests empty file names.
2007-08-24 14:56:17 +00:00
void
iref(char *s)
2007-08-24 14:56:17 +00:00
{
int i, fd;
for(i = 0; i < NINODE + 1; i++){
2007-08-24 14:56:17 +00:00
if(mkdir("irefd") != 0){
printf("%s: mkdir irefd failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
if(chdir("irefd") != 0){
printf("%s: chdir irefd failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 14:56:17 +00:00
}
mkdir("");
link("README", "");
fd = open("", O_CREATE);
if(fd >= 0)
close(fd);
fd = open("xx", O_CREATE);
if(fd >= 0)
close(fd);
unlink("xx");
}
// clean up
for(i = 0; i < NINODE + 1; i++){
chdir("..");
unlink("irefd");
}
2007-08-24 14:56:17 +00:00
chdir("/");
}
2007-08-24 20:20:23 +00:00
// test that fork fails gracefully
// the forktest binary also does this, but it runs out of proc entries first.
// inside the bigger usertests binary, we run out of memory first.
void
forktest(char *s)
2007-08-24 20:20:23 +00:00
{
enum{ N = 1000 };
2007-08-24 20:20:23 +00:00
int n, pid;
for(n=0; n<N; n++){
2007-08-24 20:20:23 +00:00
pid = fork();
if(pid < 0)
break;
if(pid == 0)
exit(0);
2007-08-24 20:20:23 +00:00
}
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
if (n == 0) {
printf("%s: no fork at all!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
}
if(n == N){
printf("%s: fork claimed to work 1000 times!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 20:20:23 +00:00
}
2007-08-24 20:20:23 +00:00
for(; n > 0; n--){
if(wait(0) < 0){
printf("%s: wait stopped early\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 20:20:23 +00:00
}
}
if(wait(0) != -1){
printf("%s: wait got too many\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2007-08-24 20:20:23 +00:00
}
}
void
sbrkbasic(char *s)
{
enum { TOOMUCH=1024*1024*1024};
int i, pid, xstatus;
char *c, *a, *b;
// does sbrk() return the expected failure value?
a = sbrk(TOOMUCH);
if(a != (char*)0xffffffffffffffffL){
printf("%s: sbrk(<toomuch>) returned %p\n", a);
2019-09-11 14:04:40 +00:00
exit(1);
}
// can one sbrk() less than a page?
a = sbrk(0);
for(i = 0; i < 5000; i++){
b = sbrk(1);
if(b != a){
printf("%s: sbrk test failed %d %x %x\n", i, a, b);
2019-09-11 14:04:40 +00:00
exit(1);
}
*b = 1;
a = b + 1;
}
pid = fork();
if(pid < 0){
printf("%s: sbrk test fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
c = sbrk(1);
c = sbrk(1);
if(c != a + 1){
printf("%s: sbrk test failed post-fork\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid == 0)
exit(0);
wait(&xstatus);
exit(xstatus);
}
void
sbrkmuch(char *s)
{
enum { BIG=100*1024*1024 };
char *c, *oldbrk, *a, *lastaddr, *p;
uint64 amt;
oldbrk = sbrk(0);
// can one grow address space to something big?
a = sbrk(0);
amt = BIG - (uint64)a;
p = sbrk(amt);
if (p != a) {
printf("%s: sbrk test failed to grow big address space; enough phys mem?\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
lastaddr = (char*) (BIG-1);
*lastaddr = 99;
// can one de-allocate?
a = sbrk(0);
c = sbrk(-PGSIZE);
if(c == (char*)0xffffffffffffffffL){
printf("%s: sbrk could not deallocate\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
c = sbrk(0);
if(c != a - PGSIZE){
printf("%s: sbrk deallocation produced wrong address, a %x c %x\n", a, c);
2019-09-11 14:04:40 +00:00
exit(1);
}
// can one re-allocate that page?
a = sbrk(0);
c = sbrk(PGSIZE);
if(c != a || sbrk(0) != a + PGSIZE){
printf("%s: sbrk re-allocation failed, a %x c %x\n", a, c);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(*lastaddr == 99){
// should be zero
printf("%s: sbrk de-allocation didn't really deallocate\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
a = sbrk(0);
c = sbrk(-(sbrk(0) - oldbrk));
if(c != a){
printf("%s: sbrk downsize failed, a %x c %x\n", a, c);
2019-09-11 14:04:40 +00:00
exit(1);
}
}
// can we read the kernel's memory?
void
kernmem(char *s)
{
char *a;
int pid;
for(a = (char*)(KERNBASE); a < (char*) (KERNBASE+2000000); a += 50000){
pid = fork();
if(pid < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid == 0){
printf("%s: oops could read %x = %x\n", a, *a);
2019-09-11 14:04:40 +00:00
exit(1);
}
int xstatus;
wait(&xstatus);
if(xstatus != -1) // did kernel kill child?
exit(1);
}
}
// if we run the system out of memory, does it clean up the last
// failed allocation?
void
sbrkfail(char *s)
{
enum { BIG=100*1024*1024 };
int i, xstatus;
int fds[2];
char scratch;
char *c, *a;
int pids[10];
int pid;
if(pipe(fds) != 0){
printf("%s: pipe() failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
2010-09-01 04:41:25 +00:00
for(i = 0; i < sizeof(pids)/sizeof(pids[0]); i++){
if((pids[i] = fork()) == 0){
// allocate a lot of memory
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
sbrk(BIG - (uint64)sbrk(0));
write(fds[1], "x", 1);
// sit around until killed
2010-09-01 04:41:25 +00:00
for(;;) sleep(1000);
}
2010-09-01 04:41:25 +00:00
if(pids[i] != -1)
read(fds[0], &scratch, 1);
}
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
// if those failed allocations freed up the pages they did allocate,
// we'll be able to allocate here
c = sbrk(PGSIZE);
2010-09-01 04:41:25 +00:00
for(i = 0; i < sizeof(pids)/sizeof(pids[0]); i++){
if(pids[i] == -1)
continue;
kill(pids[i]);
wait(0);
}
if(c == (char*)0xffffffffffffffffL){
printf("%s: failed sbrk leaked memory\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
// test running fork with the above allocated page
pid = fork();
if(pid < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if(pid == 0){
// allocate a lot of memory
a = sbrk(0);
sbrk(10*BIG);
int n = 0;
for (i = 0; i < 10*BIG; i += PGSIZE) {
n += *(a+i);
}
printf("%s: allocate a lot of memory succeeded %d\n", n);
2019-09-11 14:04:40 +00:00
exit(1);
}
wait(&xstatus);
if(xstatus != -1)
exit(1);
}
// test reads/writes from/to allocated memory
void
sbrkarg(char *s)
{
char *a;
int fd, n;
a = sbrk(PGSIZE);
fd = open("sbrk", O_CREATE|O_WRONLY);
unlink("sbrk");
if(fd < 0) {
printf("%s: open sbrk failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
if ((n = write(fd, a, PGSIZE)) < 0) {
printf("%s: write sbrk failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
close(fd);
// test writes to allocated memory
a = sbrk(PGSIZE);
if(pipe((int *) a) != 0){
printf("%s: pipe() failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
}
void
validatetest(char *s)
{
int hi;
Checkpoint port of xv6 to x86-64. Passed usertests on 2 processors a few times. The x86-64 doesn't just add two levels to page tables to support 64 bit addresses, but is a different processor. For example, calling conventions, system calls, and segmentation are different from 32-bit x86. Segmentation is basically gone, but gs/fs in combination with MSRs can be used to hold a per-core pointer. In general, x86-64 is more straightforward than 32-bit x86. The port uses code from sv6 and the xv6 "rsc-amd64" branch. A summary of the changes is as follows: - Booting: switch to grub instead of xv6's bootloader (pass -kernel to qemu), because xv6's boot loader doesn't understand 64bit ELF files. And, we don't care anymore about booting. - Makefile: use -m64 instead of -m32 flag for gcc, delete boot loader, xv6.img, bochs, and memfs. For now dont' use -O2, since usertests with -O2 is bigger than MAXFILE! - Update gdb.tmpl to be for i386 or x86-64 - Console/printf: use stdarg.h and treat 64-bit addresses different from ints (32-bit) - Update elfhdr to be 64 bit - entry.S/entryother.S: add code to switch to 64-bit mode: build a simple page table in 32-bit mode before switching to 64-bit mode, share code for entering boot processor and APs, and tweak boot gdt. The boot gdt is the gdt that the kernel proper also uses. (In 64-bit mode, the gdt/segmentation and task state mostly disappear.) - exec.c: fix passing argv (64-bit now instead of 32-bit). - initcode.c: use syscall instead of int. - kernel.ld: load kernel very high, in top terabyte. 64 bits is a lot of address space! - proc.c: initial return is through new syscall path instead of trapret. - proc.h: update struct cpu to have some scratch space since syscall saves less state than int, update struct context to reflect x86-64 calling conventions. - swtch: simplify for x86-64 calling conventions. - syscall: add fetcharg to handle x86-64 calling convetions (6 arguments are passed through registers), and fetchaddr to read a 64-bit value from user space. - sysfile: update to handle pointers from user space (e.g., sys_exec), which are 64 bits. - trap.c: no special trap vector for sys calls, because x86-64 has a different plan for system calls. - trapasm: one plan for syscalls and one plan for traps (interrupt and exceptions). On x86-64, the kernel is responsible for switching user/kernel stacks. To do, xv6 keeps some scratch space in the cpu structure, and uses MSR GS_KERN_BASE to point to the core's cpu structure (using swapgs). - types.h: add uint64, and change pde_t to uint64 - usertests: exit() when fork fails, which helped in tracking down one of the bugs in the switch from 32-bit to 64-bit - vectors: update to make them 64 bits - vm.c: use bootgdt in kernel too, program MSRs for syscalls and core-local state (for swapgs), walk 4 levels in walkpgdir, add DEVSPACETOP, use task segment to set kernel stack for interrupts (but simpler than in 32-bit mode), add an extra argument to freevm (size of user part of address space) to avoid checking all entries till KERNBASE (there are MANY TB before the top 1TB). - x86: update trapframe to have 64-bit entries, which is what the processor pushes on syscalls and traps. simplify lgdt and lidt, using struct desctr, which needs the gcc directives packed and aligned. TODO: - use int32 instead of int? - simplify curproc(). xv6 has per-cpu state again, but this time it must have it. - avoid repetition in walkpgdir - fix validateint() in usertests.c - fix bugs (e.g., observed one a case of entering kernel with invalid gs or proc
2018-09-23 12:24:42 +00:00
uint64 p;
hi = 1100*1024;
for(p = 0; p <= (uint)hi; p += PGSIZE){
// try to crash the kernel by passing in a bad string pointer
if(link("nosuchfile", (char*)p) != -1){
printf("%s: link should not succeed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
}
}
2010-09-20 10:00:22 +00:00
// does unintialized data start out zero?
char uninit[10000];
void
bsstest(char *s)
{
int i;
for(i = 0; i < sizeof(uninit); i++){
if(uninit[i] != '\0'){
printf("%s: bss test failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
}
}
// does exec return an error if the arguments
// are larger than a page? or does it write
// below the stack and wreck the instructions/data?
2010-09-20 10:00:22 +00:00
void
bigargtest(char *s)
2010-09-20 10:00:22 +00:00
{
int pid, fd, xstatus;
2010-09-20 10:00:22 +00:00
unlink("bigarg-ok");
2010-09-20 10:00:22 +00:00
pid = fork();
if(pid == 0){
static char *args[MAXARG];
2010-09-20 10:00:22 +00:00
int i;
for(i = 0; i < MAXARG-1; i++)
args[i] = "bigargs test: failed\n ";
args[MAXARG-1] = 0;
2010-09-20 10:00:22 +00:00
exec("echo", args);
fd = open("bigarg-ok", O_CREATE);
close(fd);
exit(0);
2010-09-20 10:00:22 +00:00
} else if(pid < 0){
printf("%s: bigargtest: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2010-09-20 10:00:22 +00:00
}
wait(&xstatus);
if(xstatus != 0)
exit(xstatus);
fd = open("bigarg-ok", 0);
if(fd < 0){
printf("%s: bigarg test failed!\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
}
close(fd);
2010-09-20 10:00:22 +00:00
}
// what happens when the file system runs out of blocks?
// answer: balloc panics, so this test is not useful.
void
fsfull()
{
int nfiles;
int fsblocks = 0;
2019-08-27 17:13:03 +00:00
printf("fsfull test\n");
for(nfiles = 0; ; nfiles++){
char name[64];
name[0] = 'f';
name[1] = '0' + nfiles / 1000;
name[2] = '0' + (nfiles % 1000) / 100;
name[3] = '0' + (nfiles % 100) / 10;
name[4] = '0' + (nfiles % 10);
name[5] = '\0';
printf("%s: writing %s\n", name);
int fd = open(name, O_CREATE|O_RDWR);
if(fd < 0){
printf("%s: open %s failed\n", name);
break;
}
int total = 0;
while(1){
int cc = write(fd, buf, BSIZE);
if(cc < BSIZE)
break;
total += cc;
fsblocks++;
}
printf("%s: wrote %d bytes\n", total);
close(fd);
if(total == 0)
break;
}
while(nfiles >= 0){
char name[64];
name[0] = 'f';
name[1] = '0' + nfiles / 1000;
name[2] = '0' + (nfiles % 1000) / 100;
name[3] = '0' + (nfiles % 100) / 10;
name[4] = '0' + (nfiles % 10);
name[5] = '\0';
unlink(name);
nfiles--;
}
2019-08-27 17:13:03 +00:00
printf("fsfull test finished\n");
}
void argptest(char *s)
2016-09-26 11:54:02 +00:00
{
int fd;
fd = open("init", O_RDONLY);
if (fd < 0) {
printf("%s: open failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2016-09-26 11:54:02 +00:00
}
read(fd, sbrk(0) - 1, -1);
close(fd);
}
2011-08-31 00:50:19 +00:00
unsigned long randstate = 1;
unsigned int
rand()
{
randstate = randstate * 1664525 + 1013904223;
return randstate;
}
2019-07-23 16:17:17 +00:00
// check that there's an invalid page beneath
// the user stack, to catch stack overflow.
2019-07-22 22:08:52 +00:00
void
stacktest(char *s)
2019-07-22 22:08:52 +00:00
{
int pid;
int xstatus;
2019-07-22 22:08:52 +00:00
pid = fork();
if(pid == 0) {
char *sp = (char *) r_sp();
sp -= PGSIZE;
2019-07-23 16:17:17 +00:00
// the *sp should cause a trap.
printf("%s: stacktest: read below stack %p\n", *sp);
2019-09-11 14:04:40 +00:00
exit(1);
2019-07-22 22:08:52 +00:00
} else if(pid < 0){
printf("%s: fork failed\n", s);
2019-09-11 14:04:40 +00:00
exit(1);
2019-07-22 22:08:52 +00:00
}
wait(&xstatus);
if(xstatus == -1) // kernel killed child?
exit(0);
else
exit(xstatus);
}
// regression test. copyin(), copyout(), and copyinstr() used to cast
// the virtual page address to uint, which (with certain wild system
// call arguments) resulted in a kernel page faults.
void
pgbug(char *s)
{
char *argv[1];
argv[0] = 0;
exec((char*)0xeaeb0b5b00002f5e, argv);
2019-09-20 14:27:03 +00:00
pipe((int*)0xeaeb0b5b00002f5e);
exit(0);
}
// regression test. does the kernel panic if a process sbrk()s its
// size to be less than a page, or zero, or reduces the break by an
// amount too small to cause a page to be freed?
void
sbrkbugs(char *s)
{
int pid = fork();
if(pid < 0){
printf("fork failed\n");
exit(1);
}
if(pid == 0){
int sz = (uint64) sbrk(0);
// free all user memory; there used to be a bug that
// would not adjust p->sz correctly in this case,
// causing exit() to panic.
sbrk(-sz);
// user page fault here.
exit(0);
}
wait(0);
pid = fork();
if(pid < 0){
printf("fork failed\n");
exit(1);
}
if(pid == 0){
int sz = (uint64) sbrk(0);
// set the break to somewhere in the very first
// page; there used to be a bug that would incorrectly
// free the first page.
sbrk(-(sz - 3500));
exit(0);
}
wait(0);
pid = fork();
if(pid < 0){
printf("fork failed\n");
exit(1);
}
if(pid == 0){
// set the break in the middle of a page.
sbrk((10*4096 + 2048) - (uint64)sbrk(0));
// reduce the break a bit, but not enough to
// cause a page to be freed. this used to cause
// a panic.
sbrk(-10);
exit(0);
}
wait(0);
exit(0);
}
// regression test. does write() with an invalid buffer pointer cause
// a block to be allocated for a file that is then not freed when the
// file is deleted? if the kernel has this bug, it will panic: balloc:
// out of blocks. assumed_free may need to be raised to be more than
// the number of free blocks. this test takes a long time.
void
badwrite(char *s)
{
int assumed_free = 600;
unlink("junk");
for(int i = 0; i < assumed_free; i++){
int fd = open("junk", O_CREATE|O_WRONLY);
if(fd < 0){
printf("open junk failed\n");
exit(1);
}
write(fd, (char*)0xffffffffffL, 1);
close(fd);
unlink("junk");
}
int fd = open("junk", O_CREATE|O_WRONLY);
if(fd < 0){
printf("open junk failed\n");
exit(1);
}
if(write(fd, "x", 1) != 1){
printf("write failed\n");
exit(1);
}
close(fd);
unlink("junk");
exit(0);
}
// regression test. test whether exec() leaks memory if one of the
// arguments is invalid. the test passes if the kernel doesn't panic.
void
badarg(char *s)
{
for(int i = 0; i < 50000; i++){
char *argv[2];
argv[0] = (char*)0xffffffff;
argv[1] = 0;
exec("echo", argv);
}
exit(0);
}
//
// use sbrk() to count how many free physical memory pages there are.
//
int
countfree()
{
uint64 sz0 = (uint64)sbrk(0);
int n = 0;
while(1){
uint64 a = (uint64) sbrk(4096);
if(a == 0xffffffffffffffff){
break;
}
// modify the memory to make sure it's really allocated.
*(char *)(a - 1) = 1;
n += 1;
}
sbrk(-((uint64)sbrk(0) - sz0));
return n;
}
2019-09-19 19:22:45 +00:00
// run each test in its own process. run returns 1 if child's exit()
// indicates success.
2019-09-19 19:14:52 +00:00
int
run(void f(char *), char *s) {
int pid;
int xstatus;
2019-09-19 19:39:05 +00:00
printf("test %s: ", s);
if((pid = fork()) < 0) {
printf("runtest: fork error\n");
exit(1);
}
if(pid == 0) {
f(s);
exit(0);
} else {
wait(&xstatus);
2019-09-19 19:14:52 +00:00
if(xstatus != 0)
2019-10-27 12:03:19 +00:00
printf("FAILED\n");
else
2019-10-27 12:03:19 +00:00
printf("OK\n");
2019-09-19 19:14:52 +00:00
return xstatus == 0;
}
2019-07-22 22:08:52 +00:00
}
int
2006-07-28 22:33:07 +00:00
main(int argc, char *argv[])
2006-06-27 14:35:53 +00:00
{
int continuous = 0;
char *justone = 0;
if(argc == 2 && strcmp(argv[1], "-c") == 0){
continuous = 1;
} else if(argc == 2 && strcmp(argv[1], "-C") == 0){
continuous = 2;
} else if(argc == 2 && argv[1][0] != '-'){
justone = argv[1];
} else if(argc > 1){
printf("Usage: usertests [-c] [testname]\n");
exit(1);
}
struct test {
void (*f)(char *);
char *s;
} tests[] = {
{copyin, "copyin"},
{copyout, "copyout"},
{copyinstr1, "copyinstr1"},
{copyinstr2, "copyinstr2"},
{copyinstr3, "copyinstr3"},
2020-07-16 15:38:08 +00:00
{truncate1, "truncate1"},
{truncate2, "truncate2"},
{truncate3, "truncate3"},
{reparent2, "reparent2"},
{pgbug, "pgbug" },
{sbrkbugs, "sbrkbugs" },
// {badwrite, "badwrite" },
{badarg, "badarg" },
{reparent, "reparent" },
{twochildren, "twochildren"},
{forkfork, "forkfork"},
{forkforkfork, "forkforkfork"},
{argptest, "argptest"},
{createdelete, "createdelete"},
{linkunlink, "linkunlink"},
{linktest, "linktest"},
{unlinkread, "unlinkread"},
{concreate, "concreate"},
{subdir, "subdir"},
{fourfiles, "fourfiles"},
{sharedfd, "sharedfd"},
2019-09-19 19:14:52 +00:00
{exectest, "exectest"},
{bigargtest, "bigargtest"},
{bigwrite, "bigwrite"},
{bsstest, "bsstest"},
{sbrkbasic, "sbrkbasic"},
{sbrkmuch, "sbrkmuch"},
{kernmem, "kernmem"},
{sbrkfail, "sbrkfail"},
{sbrkarg, "sbrkarg"},
{validatetest, "validatetest"},
{stacktest, "stacktest"},
{opentest, "opentest"},
{writetest, "writetest"},
{writebig, "writebig"},
{createtest, "createtest"},
{openiputtest, "openiput"},
{exitiputtest, "exitiput"},
{iputtest, "iput"},
{mem, "mem"},
{pipe1, "pipe1"},
{preempt, "preempt"},
{exitwait, "exitwait"},
{rmdot, "rmdot"},
{fourteen, "fourteen"},
{bigfile, "bigfile"},
{dirfile, "dirfile"},
{iref, "iref"},
{forktest, "forktest"},
{bigdir, "bigdir"}, // slow
{ 0, 0},
};
if(continuous){
printf("continuous usertests starting\n");
while(1){
int fail = 0;
int free0 = countfree();
for (struct test *t = tests; t->s != 0; t++) {
if(!run(t->f, t->s)){
fail = 1;
break;
}
}
if(fail){
printf("SOME TESTS FAILED\n");
if(continuous != 2)
exit(1);
}
int free1 = countfree();
if(free1 < free0){
printf("FAILED -- lost some free pages\n");
if(continuous != 2)
exit(1);
}
}
2006-09-07 20:06:08 +00:00
}
2006-09-07 13:23:41 +00:00
printf("usertests starting\n");
int free0 = countfree();
2019-09-19 19:14:52 +00:00
int fail = 0;
for (struct test *t = tests; t->s != 0; t++) {
if((justone == 0) || strcmp(t->s, justone) == 0) {
2019-09-19 19:14:52 +00:00
if(!run(t->f, t->s))
fail = 1;
}
}
if(fail){
2019-09-19 19:14:52 +00:00
printf("SOME TESTS FAILED\n");
exit(1);
} else if(countfree() < free0){
printf("FAILED -- lost some free pages\n");
exit(1);
} else {
printf("ALL TESTS PASSED\n");
exit(0);
}
2006-06-27 14:35:53 +00:00
}