void
pexit(char *exitstr, int freemem)
{
Mach *m = machp();
Proc *p;
Segment **s, **es;
int32_t utime, stime;
Waitq *wq, *f, *next;
Fgrp *fgrp;
Egrp *egrp;
Rgrp *rgrp;
Pgrp *pgrp;
Chan *dot;
if(0 && m->externup->nfullq > 0)
iprint(" %s=%d", m->externup->text, m->externup->nfullq);
if(0 && m->externup->nicc > 0)
iprint(" [%s nicc %ud tctime %ulld actime %ulld]\n",
m->externup->text, m->externup->nicc, m->externup->tctime, m->externup->actime);
if(m->externup->syscalltrace != nil)
free(m->externup->syscalltrace);
m->externup->syscalltrace = nil;
m->externup->alarm = 0;
if (m->externup->tt)
timerdel(m->externup);
if(m->externup->trace)
proctrace(m->externup, SDead, 0);
/* nil out all the resources under lock (free later) */
qlock(&m->externup->debug);
fgrp = m->externup->fgrp;
m->externup->fgrp = nil;
egrp = m->externup->egrp;
m->externup->egrp = nil;
rgrp = m->externup->rgrp;
m->externup->rgrp = nil;
pgrp = m->externup->pgrp;
m->externup->pgrp = nil;
dot = m->externup->dot;
m->externup->dot = nil;
qunlock(&m->externup->debug);
if(fgrp)
closefgrp(fgrp);
if(egrp)
closeegrp(egrp);
if(rgrp)
closergrp(rgrp);
if(dot)
cclose(dot);
if(pgrp)
closepgrp(pgrp);
/*
* if not a kernel process and have a parent,
* do some housekeeping.
*/
if(m->externup->kp == 0) {
p = m->externup->parent;
if(p == 0) {
if(exitstr == 0)
exitstr = "unknown";
//die("bootprocessdeath");
panic("boot process died: %s", exitstr);
}
while(waserror())
;
wq = smalloc(sizeof(Waitq));
poperror();
wq->w.pid = m->externup->pid;
utime = m->externup->time[TUser] + m->externup->time[TCUser];
stime = m->externup->time[TSys] + m->externup->time[TCSys];
wq->w.time[TUser] = tk2ms(utime);
wq->w.time[TSys] = tk2ms(stime);
wq->w.time[TReal] = tk2ms(sys->machptr[0]->ticks - m->externup->time[TReal]);
if(exitstr && exitstr[0])
snprint(wq->w.msg, sizeof(wq->w.msg), "%s %d: %s",
m->externup->text, m->externup->pid, exitstr);
else
wq->w.msg[0] = '\0';
lock(&p->exl);
/*
* Check that parent is still alive.
*/
if(p->pid == m->externup->parentpid && p->state != Broken) {
p->nchild--;
p->time[TCUser] += utime;
p->time[TCSys] += stime;
/*
* If there would be more than 128 wait records
* processes for my parent, then don't leave a wait
* record behind. This helps prevent badly written
* daemon processes from accumulating lots of wait
* records.
*/
if(p->nwait < 128) {
wq->next = p->waitq;
p->waitq = wq;
p->nwait++;
wq = nil;
wakeup(&p->waitr);
}
}
unlock(&p->exl);
if(wq)
free(wq);
}
if(!freemem)
addbroken(m->externup);
qlock(&m->externup->seglock);
es = &m->externup->seg[NSEG];
for(s = m->externup->seg; s < es; s++) {
if(*s) {
putseg(*s);
*s = 0;
}
}
qunlock(&m->externup->seglock);
lock(&m->externup->exl); /* Prevent my children from leaving waits */
psunhash(m->externup);
m->externup->pid = 0;
wakeup(&m->externup->waitr);
unlock(&m->externup->exl);
for(f = m->externup->waitq; f; f = next) {
next = f->next;
free(f);
}
/* release debuggers */
qlock(&m->externup->debug);
if(m->externup->pdbg) {
wakeup(&m->externup->pdbg->sleep);
m->externup->pdbg = 0;
}
qunlock(&m->externup->debug);
/* Sched must not loop for these locks */
lock(&procalloc);
lock(&pga);
stopac();
edfstop(m->externup);
m->externup->state = Moribund;
sched();
panic("pexit");
}
static int32_t
conswrite(Chan *c, void *va, int32_t n, int64_t off)
{
Proc *up = externup();
char buf[256];
int32_t l, bp;
char *a;
Mach *mp;
int i;
uint32_t offset;
Cmdbuf *cb;
Cmdtab *ct;
a = va;
offset = off;
extern int printallsyscalls;
switch((uint32_t)c->qid.path){
case Qcons:
/*
* Can't page fault in putstrn, so copy the data locally.
*/
l = n;
while(l > 0){
bp = l;
if(bp > sizeof buf)
bp = sizeof buf;
memmove(buf, a, bp);
putstrn0(buf, bp, 1);
a += bp;
l -= bp;
}
break;
case Qconsctl:
print("consctl\n");
if(n >= sizeof(buf))
n = sizeof(buf)-1;
strncpy(buf, a, n);
buf[n] = 0;
for(a = buf; a;){
if(strncmp(a, "sys", 3) == 0) {
printallsyscalls = ! printallsyscalls;
print("%sracing syscalls\n", printallsyscalls ? "T" : "Not t");
}
if(a = strchr(a, ' '))
a++;
}
break;
case Qtime:
if(!iseve())
error(Eperm);
return writetime(a, n);
case Qbintime:
if(!iseve())
error(Eperm);
return writebintime(a, n);
case Qhostowner:
return hostownerwrite(a, n);
case Qhostdomain:
return hostdomainwrite(a, n);
case Quser:
return userwrite(a, n);
case Qnull:
break;
case Qreboot:
if(!iseve())
error(Eperm);
cb = parsecmd(a, n);
if(waserror()) {
free(cb);
nexterror();
}
ct = lookupcmd(cb, rebootmsg, nelem(rebootmsg));
switch(ct->index) {
case CMhalt:
reboot(nil, 0, 0);
break;
case CMreboot:
rebootcmd(cb->nf-1, cb->f+1);
break;
case CMpanic:
*(volatile uint32_t*)0=0;
panic("/dev/reboot");
}
poperror();
free(cb);
break;
case Qsysstat:
for(i = 0; i < MACHMAX; i++)
if((mp = sys->machptr[i]) != nil && mp->online){
mp = sys->machptr[i];
mp->cs = 0;
mp->intr = 0;
mp->syscall = 0;
mp->pfault = 0;
mp->tlbfault = 0; /* not updated */
mp->tlbpurge = 0; /* # mmuflushtlb */
}
break;
case Qswap:
if(n >= sizeof buf)
error(Egreg);
memmove(buf, va, n); /* so we can NUL-terminate */
buf[n] = 0;
if(!iseve())
error(Eperm);
if(buf[0]<'0' || '9'<buf[0])
error(Ebadarg);
if(strncmp(buf, "start", 5) == 0){
print("request to start pager ignored\n");
break;
}
break;
case Qsysname:
if(offset != 0)
error(Ebadarg);
if(n <= 0 || n >= sizeof buf)
error(Ebadarg);
strncpy(buf, a, n);
buf[n] = 0;
if(buf[n-1] == '\n')
buf[n-1] = 0;
kstrdup(&sysname, buf);
break;
case Qdebug:
if(n >= sizeof(buf))
n = sizeof(buf)-1;
strncpy(buf, a, n);
buf[n] = 0;
if(n > 0 && buf[n-1] == '\n')
buf[n-1] = 0;
error(Ebadctl);
break;
case Qsyscall:
if(n >= sizeof(buf))
n = sizeof(buf)-1;
strncpy(buf, a, n);
buf[n] = 0;
if(n > 0 && buf[n-1] == '\n')
buf[n-1] = 0;
// Doing strncmp right is just painful and overkill here.
if (buf[0] == 'o') {
if (buf[1] == 'n' && buf[2] == 0) {
printallsyscalls = 1;
break;
}
if (buf[1] == 'f' && buf[2] == 'f' && buf[3] == 0) {
printallsyscalls = 0;
break;
}
}
error("#c/syscall: can only write on or off");
break;
default:
print("conswrite: %#llx\n", c->qid.path);
error(Egreg);
}
return n;
}
/*
* the devxxx.c that calls us handles writing data, it knows best
*/
int32_t
netifwrite(Netif *nif, Chan *c, void *a, int32_t n)
{
Mach *m = machp();
Netfile *f;
int type, mtu;
char *p, buf[64];
uint8_t binaddr[Nmaxaddr];
if(NETTYPE(c->qid.path) != Nctlqid)
error(Eperm);
if(n >= sizeof(buf))
n = sizeof(buf)-1;
memmove(buf, a, n);
buf[n] = 0;
if(waserror()){
qunlock(nif);
nexterror();
}
qlock(nif);
f = nif->f[NETID(c->qid.path)];
if((p = matchtoken(buf, "connect")) != 0){
qclose(f->iq);
type = atoi(p);
if(typeinuse(nif, type))
error(Einuse);
f->type = type;
if(f->type < 0)
nif->all++;
qreopen(f->iq);
} else if(matchtoken(buf, "promiscuous")){
if(f->prom == 0){
if(nif->prom == 0 && nif->promiscuous != nil)
nif->promiscuous(nif->arg, 1);
f->prom = 1;
nif->prom++;
}
} else if((p = matchtoken(buf, "scanbs")) != 0){
/* scan for base stations */
if(f->scan == 0){
type = atoi(p);
if(type < 5)
type = 5;
if(nif->scanbs != nil)
nif->scanbs(nif->arg, type);
f->scan = type;
nif->_scan++;
}
} else if((p = matchtoken(buf, "mtu")) != 0){
/* poor planning. */
if(!iseve())
error(Eperm);
mtu = atoi(p);
/* zero resets default. */
if(mtu != 0)
if(mtu < nif->minmtu || mtu > nif->maxmtu)
error(Ebadarg);
if(nif->hwmtu)
nif->mtu = nif->hwmtu(nif->arg, mtu);
else
nif->mtu = mtu;
} else if(matchtoken(buf, "l2bridge")){
f->bridge |= 2;
} else if(matchtoken(buf, "bridge")){
f->bridge |= 1;
} else if(matchtoken(buf, "headersonly")){
f->headersonly = 1;
} else if((p = matchtoken(buf, "addmulti")) != 0){
if(parseaddr(binaddr, p, nif->alen) < 0)
error("bad address");
p = netmulti(nif, f, binaddr, 1);
if(p)
error(p);
} else if((p = matchtoken(buf, "remmulti")) != 0){
if(parseaddr(binaddr, p, nif->alen) < 0)
error("bad address");
p = netmulti(nif, f, binaddr, 0);
if(p)
error(p);
} else
n = -1;
qunlock(nif);
poperror();
return n;
}
/*
* the devxxx.c that calls us handles writing data, it knows best
*/
long netifwrite(struct ether *nif, struct chan *c, void *a, long n)
{
ERRSTACK(1);
struct netfile *f;
int type;
char *p, buf[64];
uint8_t binaddr[Nmaxaddr];
if (NETTYPE(c->qid.path) != Nctlqid)
error(Eperm);
if (n >= sizeof(buf))
n = sizeof(buf) - 1;
memmove(buf, a, n);
buf[n] = 0;
if (waserror()) {
qunlock(&nif->qlock);
nexterror();
}
qlock(&nif->qlock);
f = nif->f[NETID(c->qid.path)];
if ((p = matchtoken(buf, "connect")) != 0) {
type = strtol(p, 0, 0); /* allows any base, though usually hex */
if (typeinuse(nif, type))
error(Einuse);
f->type = type;
if (f->type < 0)
nif->all++;
} else if (matchtoken(buf, "promiscuous")) {
if (f->prom == 0) {
if (nif->prom == 0 && nif->promiscuous != NULL)
nif->promiscuous(nif->arg, 1);
f->prom = 1;
nif->prom++;
}
} else if ((p = matchtoken(buf, "scanbs")) != 0) {
/* scan for base stations */
if (f->scan == 0) {
type = strtol(p, 0, 0); /* allows any base, though usually hex */
if (type < 5)
type = 5;
if (nif->scanbs != NULL)
nif->scanbs(nif->arg, type);
f->scan = type;
nif->scan++;
}
} else if (matchtoken(buf, "bridge")) {
f->bridge = 1;
} else if (matchtoken(buf, "headersonly")) {
f->headersonly = 1;
} else if ((p = matchtoken(buf, "addmulti")) != 0) {
if (parseaddr(binaddr, p, nif->alen) < 0)
error("bad address");
p = netmulti(nif, f, binaddr, 1);
if (p)
error(p);
} else if ((p = matchtoken(buf, "remmulti")) != 0) {
if (parseaddr(binaddr, p, nif->alen) < 0)
error("bad address");
p = netmulti(nif, f, binaddr, 0);
if (p)
error(p);
} else
n = -1;
qunlock(&nif->qlock);
poperror();
return n;
}
static long
conswrite(Chan *c, void *va, long n, vlong offset)
{
char buf[128], *a, ch;
int x;
if(c->qid.type & QTDIR)
error(Eperm);
switch((ulong)c->qid.path) {
default:
error(Egreg);
case Qcons:
if(canrlock(&kprintq.l)){
if(kprintq.q != nil){
if(waserror()){
runlock(&kprintq.l);
nexterror();
}
qwrite(kprintq.q, va, n);
poperror();
runlock(&kprintq.l);
return n;
}
runlock(&kprintq.l);
}
return write(1, va, n);
case Qsysctl:
return sysconwrite(va, n);
case Qconsctl:
if(n >= sizeof(buf))
n = sizeof(buf)-1;
strncpy(buf, va, n);
buf[n] = 0;
for(a = buf; a;){
if(strncmp(a, "rawon", 5) == 0){
kbd.raw = 1;
/* clumsy hack - wake up reader */
ch = 0;
qwrite(kbdq, &ch, 1);
} else if(strncmp(buf, "rawoff", 6) == 0){
kbd.raw = 0;
}
if((a = strchr(a, ' ')) != nil)
a++;
}
break;
case Qkeyboard:
for(x=0; x<n; ) {
Rune r;
x += chartorune(&r, &((char*)va)[x]);
gkbdputc(gkbdq, r);
}
break;
case Qnull:
break;
case Qtime:
if(n >= sizeof(buf))
n = sizeof(buf)-1;
strncpy(buf, va, n);
buf[n] = '\0';
timeoffset = strtoll(buf, 0, 0)-osusectime();
break;
case Qhostowner:
if(!iseve())
error(Eperm);
if(offset != 0 || n >= sizeof(buf))
error(Ebadarg);
memmove(buf, va, n);
buf[n] = '\0';
if(n > 0 && buf[n-1] == '\n')
buf[--n] = '\0';
if(n == 0)
error(Ebadarg);
/* renameuser(eve, buf); */
/* renameproguser(eve, buf); */
kstrdup(&eve, buf);
kstrdup(&up->env->user, buf);
break;
case Quser:
if(!iseve())
error(Eperm);
if(offset != 0)
error(Ebadarg);
if(n <= 0 || n >= sizeof(buf))
error(Ebadarg);
strncpy(buf, va, n);
buf[n] = '\0';
if(n > 0 && buf[n-1] == '\n')
buf[--n] = '\0';
if(n == 0)
error(Ebadarg);
setid(buf, 0);
break;
case Qhoststdout:
return write(1, va, n);
case Qhoststderr:
return write(2, va, n);
case Qjit:
if(n >= sizeof(buf))
n = sizeof(buf)-1;
strncpy(buf, va, n);
buf[n] = '\0';
x = atoi(buf);
if(x < 0 || x > 9)
error(Ebadarg);
cflag = x;
break;
case Qsysname:
if(offset != 0)
error(Ebadarg);
if(n < 0 || n >= sizeof(buf))
error(Ebadarg);
strncpy(buf, va, n);
buf[n] = '\0';
if(buf[n-1] == '\n')
buf[n-1] = 0;
kstrdup(&ossysname, buf);
break;
}
return n;
}
/*
* All traps come here. It is slower to have all traps call trap()
* rather than directly vectoring the handler. However, this avoids a
* lot of code duplication and possible bugs. The only exception is
* VectorSYSCALL.
* Trap is called with interrupts disabled via interrupt-gates.
*/
void
trap(Ureg* ureg)
{
int clockintr, vno, user;
// cache the previous vno to see what might be causing
// trouble
static int lastvno;
vno = ureg->type;
uint64_t gsbase = rdmsr(GSbase);
//if (sce > scx) iprint("====================");
if (vno == 8) {
iprint("Lstar is %p\n", (void *)rdmsr(Lstar));
iprint("GSbase is %p\n", (void *)gsbase);
iprint("ire %d irx %d sce %d scx %d lastvno %d\n",
ire, irx, sce, scx, lastvno);
iprint("irxe %d \n",
irxe);
die("8");
}
lastvno = vno;
if (gsbase < 1ULL<<63)
die("bogus gsbase");
Mach *m = machp();
char buf[ERRMAX];
Vctl *ctl, *v;
if (0 && m && m->externup && m->externup->pid == 6) {
//iprint("type %x\n", ureg->type);
if (ureg->type != 0x49)
die("6\n");
}
m->perf.intrts = perfticks();
user = userureg(ureg);
if(user && (m->nixtype == NIXTC)){
m->externup->dbgreg = ureg;
cycles(&m->externup->kentry);
}
clockintr = 0;
//_pmcupdate(m);
if(ctl = vctl[vno]){
if(ctl->isintr){
m->intr++;
if(vno >= VectorPIC && vno != VectorSYSCALL)
m->lastintr = ctl->irq;
}else
if(m->externup)
m->externup->nqtrap++;
if(ctl->isr)
ctl->isr(vno);
for(v = ctl; v != nil; v = v->next){
if(v->f)
v->f(ureg, v->a);
}
if(ctl->eoi)
ctl->eoi(vno);
intrtime(vno);
if(ctl->isintr){
if(ctl->irq == IrqCLOCK || ctl->irq == IrqTIMER)
clockintr = 1;
if(m->externup && !clockintr)
preempted();
}
}
else if(vno < nelem(excname) && user){
spllo();
snprint(buf, sizeof buf, "sys: trap: %s", excname[vno]);
postnote(m->externup, 1, buf, NDebug);
}
else if(vno >= VectorPIC && vno != VectorSYSCALL){
/*
* An unknown interrupt.
* Check for a default IRQ7. This can happen when
* the IRQ input goes away before the acknowledge.
* In this case, a 'default IRQ7' is generated, but
* the corresponding bit in the ISR isn't set.
* In fact, just ignore all such interrupts.
*/
/* clear the interrupt */
i8259isr(vno);
iprint("cpu%d: spurious interrupt %d, last %d\n",
m->machno, vno, m->lastintr);
intrtime(vno);
if(user)
kexit(ureg);
return;
}
else{
if(vno == VectorNMI){
nmienable();
if(m->machno != 0){
iprint("cpu%d: PC %#llux\n",
m->machno, ureg->ip);
for(;;);
}
}
dumpregs(ureg);
if(!user){
ureg->sp = PTR2UINT(&ureg->sp);
dumpstackwithureg(ureg);
}
if(vno < nelem(excname))
panic("%s", excname[vno]);
panic("unknown trap/intr: %d\n", vno);
}
splhi();
/* delaysched set because we held a lock or because our quantum ended */
if(m->externup && m->externup->delaysched && clockintr){
if(0)
if(user && m->externup->ac == nil && m->externup->nqtrap == 0 && m->externup->nqsyscall == 0){
if(!waserror()){
m->externup->ac = getac(m->externup, -1);
poperror();
runacore();
return;
}
}
sched();
splhi();
}
if(user){
if(m->externup && m->externup->procctl || m->externup->nnote)
notify(ureg);
kexit(ureg);
}
}
static long
loopbackwrite(Chan *c, void *va, long n, vlong off)
{
Loop *lb;
Link *link;
Cmdbuf *volatile cb;
Block *volatile bp;
vlong d0, d0ns;
long dn, dnns;
switch(TYPE(c->qid.path)){
case Qdata:
bp = allocb(n);
if(waserror()){
freeb(bp);
nexterror();
}
memmove(bp->wp, va, n);
poperror();
bp->wp += n;
return loopbackbwrite(c, bp, off);
case Qctl:
lb = c->aux;
link = &lb->link[ID(c->qid.path)];
cb = parsecmd(va, n);
if(waserror()){
free(cb);
nexterror();
}
if(cb->nf < 1)
error("short control request");
if(strcmp(cb->f[0], "delay") == 0){
if(cb->nf != 3)
error("usage: delay latency bytedelay");
d0ns = strtoll(cb->f[1], nil, 10);
dnns = strtol(cb->f[2], nil, 10);
/*
* it takes about 20000 cycles on a pentium ii
* to run pushlink; perhaps this should be accounted.
*/
ilock(link);
link->delay0ns = d0ns;
link->delaynns = dnns;
iunlock(link);
}else if(strcmp(cb->f[0], "indrop") == 0){
if(cb->nf != 2)
error("usage: indrop [01]");
ilock(link);
link->indrop = strtol(cb->f[1], nil, 0) != 0;
iunlock(link);
}else if(strcmp(cb->f[0], "droprate") == 0){
if(cb->nf != 2)
error("usage: droprate ofn");
ilock(link);
link->droprate = strtol(cb->f[1], nil, 0);
iunlock(link);
}else if(strcmp(cb->f[0], "limit") == 0){
if(cb->nf != 2)
error("usage: limit maxqsize");
ilock(link);
link->limit = strtol(cb->f[1], nil, 0);
qsetlimit(link->oq, link->limit);
qsetlimit(link->iq, link->limit);
iunlock(link);
}else if(strcmp(cb->f[0], "reset") == 0){
if(cb->nf != 1)
error("usage: reset");
ilock(link);
link->packets = 0;
link->bytes = 0;
link->indrop = 0;
link->soverflows = 0;
link->drops = 0;
iunlock(link);
}else
error("unknown control request");
poperror();
free(cb);
break;
default:
error(Eperm);
}
return n;
}
long
bindmount(int ismount, int fd, int afd, char* arg0, char* arg1, ulong flag, char* spec)
{
int ret;
Chan *c0, *c1, *ac, *bc;
struct{
Chan *chan;
Chan *authchan;
char *spec;
int flags;
}bogus;
if((flag&~MMASK) || (flag&MORDER)==(MBEFORE|MAFTER))
error(Ebadarg);
if(ismount){
validaddr((ulong)spec, 1, 0);
spec = validnamedup(spec, 1);
if(waserror()){
free(spec);
nexterror();
}
if(up->pgrp->noattach)
error(Enoattach);
ac = nil;
bc = fdtochan(fd, ORDWR, 0, 1);
if(waserror()) {
if(ac)
cclose(ac);
cclose(bc);
nexterror();
}
if(afd >= 0)
ac = fdtochan(afd, ORDWR, 0, 1);
bogus.flags = flag & MCACHE;
bogus.chan = bc;
bogus.authchan = ac;
bogus.spec = spec;
ret = devno('M', 0);
c0 = devtab[ret]->attach((char*)&bogus);
poperror(); /* ac bc */
if(ac)
cclose(ac);
cclose(bc);
}else{
spec = 0;
validaddr((ulong)arg0, 1, 0);
c0 = namec(arg0, Abind, 0, 0);
}
if(waserror()){
cclose(c0);
nexterror();
}
validaddr((ulong)arg1, 1, 0);
c1 = namec(arg1, Amount, 0, 0);
if(waserror()){
cclose(c1);
nexterror();
}
ret = cmount(&c0, c1, flag, spec);
poperror();
cclose(c1);
poperror();
cclose(c0);
if(ismount){
fdclose(fd, 0);
poperror();
free(spec);
}
return ret;
}
/* if dc is non-zero, it means we're doing a mount and dc is the mount device to use. */
static int
bindmount(int dc, int fd, int afd, char* arg0, char* arg1, int flag, char* spec)
{
Proc *up = externup();
int i;
Dev *dev;
Chan *c0, *c1, *ac, *bc;
struct{
Chan *chan;
Chan *authchan;
char *spec;
int flags;
}bogus;
if((flag&~MMASK) || (flag&MORDER)==(MBEFORE|MAFTER))
error(Ebadarg);
bogus.flags = flag & MCACHE;
if(dc){
if (! checkdc(dc))
error(Ebadarg);
if(up->pgrp->noattach)
error(Enoattach);
ac = nil;
bc = fdtochan(fd, ORDWR, 0, 1);
if(waserror()) {
if(ac)
cclose(ac);
cclose(bc);
nexterror();
}
if(afd >= 0)
ac = fdtochan(afd, ORDWR, 0, 1);
bogus.chan = bc;
bogus.authchan = ac;
bogus.spec = validaddr(spec, 1, 0);
if(waserror())
error(Ebadspec);
spec = validnamedup(spec, 1);
poperror();
if(waserror()){
free(spec);
nexterror();
}
dev = devtabget(dc, 0); //XDYNX
if(waserror()){
//devtabdecr(dev);
nexterror();
}
c0 = dev->attach((char*)&bogus);
poperror();
//devtabdecr(dev);
poperror(); /* spec */
free(spec);
poperror(); /* ac bc */
if(ac)
cclose(ac);
cclose(bc);
}else{
bogus.spec = nil;
c0 = namec(validaddr(arg0, 1, 0), Abind, 0, 0);
}
if(waserror()){
cclose(c0);
nexterror();
}
c1 = namec(validaddr(arg1, 1, 0), Amount, 0, 0);
if(waserror()){
cclose(c1);
nexterror();
}
i = cmount(&c0, c1, flag, bogus.spec);
poperror();
cclose(c1);
poperror();
cclose(c0);
if(dc)
fdclose(fd, 0);
return i;
}
static long
read(ulong *arg, vlong *offp)
{
long n, nn, nnn;
uchar *p;
Chan *c;
vlong off;
n = arg[2];
validaddr(arg[1], n, 1);
p = (void*)arg[1];
c = fdtochan(arg[0], OREAD, 1, 1);
if(waserror()){
cclose(c);
nexterror();
}
/*
* The offset is passed through on directories, normally.
* Sysseek complains, but pread is used by servers like exportfs,
* that shouldn't need to worry about this issue.
*
* Notice that c->devoffset is the offset that c's dev is seeing.
* The number of bytes read on this fd (c->offset) may be different
* due to rewritings in rockfix.
*/
if(offp == nil) /* use and maintain channel's offset */
off = c->offset;
else
off = *offp;
if(off < 0)
error(Enegoff);
if(off == 0){ /* rewind to the beginning of the directory */
if(offp == nil){
c->offset = 0;
c->devoffset = 0;
}
mountrewind(c);
unionrewind(c);
}
if(c->qid.type & QTDIR){
if(mountrockread(c, p, n, &nn)){
/* do nothing: mountrockread filled buffer */
}else if(c->umh)
nn = unionread(c, p, n);
else{
if(off != c->offset)
error(Edirseek);
nn = devtab[c->type]->read(c, p, n, c->devoffset);
}
nnn = mountfix(c, p, nn, n);
}else
nnn = nn = devtab[c->type]->read(c, p, n, off);
lock(c);
c->devoffset += nn;
c->offset += nnn;
unlock(c);
poperror();
cclose(c);
return nnn;
}
static void
sseek(ulong *arg)
{
Chan *c;
uchar buf[sizeof(Dir)+100];
Dir dir;
int n;
vlong off;
union {
vlong v;
ulong u[2];
} o;
c = fdtochan(arg[1], -1, 1, 1);
if(waserror()){
cclose(c);
nexterror();
}
if(devtab[c->type]->dc == '|')
error(Eisstream);
off = 0;
o.u[0] = arg[2];
o.u[1] = arg[3];
switch(arg[4]){
case 0:
off = o.v;
if((c->qid.type & QTDIR) && off != 0)
error(Eisdir);
if(off < 0)
error(Enegoff);
c->offset = off;
break;
case 1:
if(c->qid.type & QTDIR)
error(Eisdir);
lock(c); /* lock for read/write update */
off = o.v + c->offset;
if(off < 0){
unlock(c);
error(Enegoff);
}
c->offset = off;
unlock(c);
break;
case 2:
if(c->qid.type & QTDIR)
error(Eisdir);
n = devtab[c->type]->stat(c, buf, sizeof buf);
if(convM2D(buf, n, &dir, nil) == 0)
error("internal error: stat error in seek");
off = dir.length + o.v;
if(off < 0)
error(Enegoff);
c->offset = off;
break;
default:
error(Ebadarg);
}
*(vlong*)arg[0] = off;
c->uri = 0;
c->dri = 0;
cclose(c);
poperror();
}
/*
* Rewrite the results of a directory read to reflect current
* name space bindings and mounts. Specifically, replace
* directory entries for bind and mount points with the results
* of statting what is mounted there. Except leave the old names.
*/
static long
mountfix(Chan *c, uchar *op, long n, long maxn)
{
char *name;
int nbuf, nname;
Chan *nc;
Mhead *mh;
Mount *m;
uchar *p;
int dirlen, rest;
long l;
uchar *buf, *e;
Dir d;
p = op;
buf = nil;
nbuf = 0;
for(e=&p[n]; p+BIT16SZ<e; p+=dirlen){
dirlen = dirfixed(p, e, &d);
if(dirlen < 0)
break;
nc = nil;
mh = nil;
if(findmount(&nc, &mh, d.type, d.dev, d.qid)){
/*
* If it's a union directory and the original is
* in the union, don't rewrite anything.
*/
for(m=mh->mount; m; m=m->next)
if(eqchantdqid(m->to, d.type, d.dev, d.qid, 1))
goto Norewrite;
name = dirname(p, &nname);
/*
* Do the stat but fix the name. If it fails, leave old entry.
* BUG: If it fails because there isn't room for the entry,
* what can we do? Nothing, really. Might as well skip it.
*/
if(buf == nil){
buf = smalloc(4096);
nbuf = 4096;
}
if(waserror())
goto Norewrite;
l = devtab[nc->type]->stat(nc, buf, nbuf);
l = dirsetname(name, nname, buf, l, nbuf);
if(l == BIT16SZ)
error("dirsetname");
poperror();
/*
* Shift data in buffer to accomodate new entry,
* possibly overflowing into rock.
*/
rest = e - (p+dirlen);
if(l > dirlen){
while(p+l+rest > op+maxn){
mountrock(c, p, &e);
if(e == p){
dirlen = 0;
goto Norewrite;
}
rest = e - (p+dirlen);
}
}
if(l != dirlen){
memmove(p+l, p+dirlen, rest);
dirlen = l;
e = p+dirlen+rest;
}
/*
* Rewrite directory entry.
*/
memmove(p, buf, l);
Norewrite:
cclose(nc);
putmhead(mh);
}
}
if(buf)
free(buf);
if(p != e)
error("oops in rockfix");
return e-op;
}
/*
* Move the current process to an application core.
* This is performed at the end of execac(), and
* we pretend to be returning to user-space, but instead we
* dispatch the process to another core.
* 1. We do the final bookkeeping that syscall() would do after
* a return from sysexec(), because we are not returning.
* 2. We dispatch the process to an AC using an ICC.
*
* This function won't return unless the process is reclaimed back
* to the time-sharing core, and is the handler for the process
* to deal with traps and system calls until the process dies.
*
* Remember that this function is the "line" between user and kernel
* space, it's not expected to raise|handle any error.
*
* We install a safety error label, just in case we raise errors,
* which we shouldn't. (noerrorsleft knows that for exotic processes
* there is an error label pushed by us).
*/
void
runacore(void)
{
Proc *up = externup();
Ureg *ureg;
void (*fn)(void);
int rc, flush, s;
char *n;
uint64_t t1;
if(waserror())
panic("runacore: error: %s\n", up->errstr);
ureg = up->dbgreg;
fakeretfromsyscall(ureg);
fpusysrfork(ureg);
procpriority(up, PriKproc, 1);
rc = runac(up->ac, actouser, 1, nil, 0);
procpriority(up, PriNormal, 0);
for(;;){
t1 = fastticks(nil);
flush = 0;
fn = nil;
switch(rc){
case ICCTRAP:
s = splhi();
machp()->MMU.cr2 = up->ac->MMU.cr2;
DBG("runacore: trap %llu cr2 %#llx ureg %#p\n",
ureg->type, machp()->MMU.cr2, ureg);
switch(ureg->type){
case IdtIPI:
if(up->procctl || up->nnote)
notify(up->dbgreg);
if(up->ac == nil)
goto ToTC;
kexit(up->dbgreg);
break;
case IdtNM:
case IdtMF:
case IdtXF:
/* these are handled in the AC;
* If we get here, they left in m->NIX.icc->data
* a note to be posted to the process.
* Post it, and make the vector a NOP.
*/
n = up->ac->NIX.icc->note;
if(n != nil)
postnote(up, 1, n, NDebug);
ureg->type = IdtIPI; /* NOP */
break;
default:
cr3put(machp()->MMU.pml4->pa);
if(0 && ureg->type == IdtPF){
print("before PF:\n");
print("AC:\n");
dumpptepg(4, up->ac->MMU.pml4->pa);
print("\n%s:\n", rolename[NIXTC]);
dumpptepg(4, machp()->MMU.pml4->pa);
}
trap(ureg);
}
splx(s);
flush = 1;
fn = actrapret;
break;
case ICCSYSCALL:
DBG("runacore: syscall ax %#llx ureg %#p\n",
ureg->ax, ureg);
cr3put(machp()->MMU.pml4->pa);
//syscall(ureg->ax, ureg);
flush = 1;
fn = acsysret;
if(0)
if(up->nqtrap > 2 || up->nsyscall > 1)
goto ToTC;
if(up->ac == nil)
goto ToTC;
break;
default:
panic("runacore: unexpected rc = %d", rc);
}
up->tctime += fastticks2us(fastticks(nil) - t1);
procpriority(up, PriExtra, 1);
rc = runac(up->ac, fn, flush, nil, 0);
procpriority(up, PriNormal, 0);
}
ToTC:
/*
* to procctl, then syscall, to
* be back in the TC
*/
DBG("runacore: up %#p: return\n", up);
}
void
trap(Ureg *ureg)
{
int rem, itype, t;
if(up != nil)
rem = ((char*)ureg)-up->kstack;
else rem = ((char*)ureg)-(char*)m->stack;
if(ureg->type != PsrMfiq && rem < 256) {
dumpregs(ureg);
panic("trap %d stack bytes remaining (%s), "
"up=#%8.8lux ureg=#%8.8lux pc=#%8.8ux"
,rem, up?up->text:"", up, ureg, ureg->pc);
for(;;);
}
itype = ureg->type;
/* All interrupts/exceptions should be resumed at ureg->pc-4,
except for Data Abort which resumes at ureg->pc-8. */
if(itype == PsrMabt+1)
ureg->pc -= 8;
else ureg->pc -= 4;
if(up){
up->pc = ureg->pc;
up->dbgreg = ureg;
}
switch(itype) {
case PsrMirq:
t = m->ticks; /* CPU time per proc */
up = nil; /* no process at interrupt level */
irq(ureg);
up = m->proc;
preemption(m->ticks - t);
m->intr++;
break;
case PsrMund:
if(*(ulong*)ureg->pc == BREAK && breakhandler) {
int s;
Proc *p;
p = up;
s = breakhandler(ureg, p);
if(s == BrkSched) {
p->preempted = 0;
sched();
} else if(s == BrkNoSched) {
/* stop it being preempted until next instruction */
p->preempted = 1;
if(up)
up->dbgreg = 0;
return;
}
break;
}
if(up == nil) goto faultpanic;
spllo();
if(waserror()) {
if(waslo(ureg->psr) && up->type == Interp)
disfault(ureg, up->env->errstr);
setpanic();
dumpregs(ureg);
panic("%s", up->env->errstr);
}
if(!fpiarm(ureg)) {
dumpregs(ureg);
sys_trap_error(ureg->type);
}
poperror();
break;
case PsrMsvc: /* Jump through 0 or SWI */
if(waslo(ureg->psr) && up && up->type == Interp) {
spllo();
dumpregs(ureg);
sys_trap_error(ureg->type);
}
setpanic();
dumpregs(ureg);
panic("SVC/SWI exception");
break;
case PsrMabt: /* Prefetch abort */
if(catchdbg && catchdbg(ureg, 0))
break;
/* FALL THROUGH */
case PsrMabt+1: /* Data abort */
if(waslo(ureg->psr) && up && up->type == Interp) {
spllo();
faultarm(ureg);
}
print("Data Abort\n");
/* FALL THROUGH */
default:
faultpanic:
setpanic();
dumpregs(ureg);
panic("exception %uX %s\n", ureg->type, trapname(ureg->type));
break;
}
splhi();
if(up)
up->dbgreg = 0; /* becomes invalid after return from trap */
}
/*
* Print a string on the console. Convert \n to \r\n for serial
* line consoles. Locking of the queues is left up to the screen
* or uart code. Multi-line messages to serial consoles may get
* interspersed with other messages.
*/
static void
putstrn0(char *str, int n, int usewrite)
{
int m;
char *t;
char buf[PRINTSIZE+2];
/*
* if kprint is open, put the message there, otherwise
* if there's an attached bit mapped display,
* put the message there.
*/
m = consoleprint;
if(canrlock(&kprintq)){
if(kprintq.q != nil){
if(waserror()){
runlock(&kprintq);
nexterror();
}
if(usewrite)
qwrite(kprintq.q, str, n);
else
qiwrite(kprintq.q, str, n);
poperror();
m = 0;
}
runlock(&kprintq);
}
if(m && screenputs != nil)
screenputs(str, n);
/*
* if there's a serial line being used as a console,
* put the message there.
*/
if(serwrite != nil) {
serwrite(str, n);
return;
}
if(printq == 0)
return;
while(n > 0) {
t = memchr(str, '\n', n);
if(t && !kbd.raw) {
m = t - str;
if(m > sizeof(buf)-2)
m = sizeof(buf)-2;
memmove(buf, str, m);
buf[m] = '\r';
buf[m+1] = '\n';
if(usewrite)
qwrite(printq, buf, m+2);
else
qiwrite(printq, buf, m+2);
str = t + 1;
n -= m + 1;
} else {
if(usewrite)
qwrite(printq, str, n);
else
qiwrite(printq, str, n);
break;
}
}
}
/*
* Rewrite the results of a directory read to reflect current
* name space bindings and mounts. Specifically, replace
* directory entries for bind and mount points with the results
* of statting what is mounted there. Except leave the old names.
*/
static int32_t
mountfix(Chan *c, uint8_t *op, int32_t n, int32_t maxn)
{
Proc *up = externup();
char *name;
int nbuf;
Chan *nc;
Mhead *mh;
Mount *mount;
usize dirlen, nname, r, rest;
int32_t l;
uint8_t *buf, *e, *p;
Dir d;
p = op;
buf = nil;
nbuf = 0;
for(e=&p[n]; p+BIT16SZ<e; p+=dirlen){
dirlen = dirfixed(p, e, &d);
if(dirlen == 0)
break;
nc = nil;
mh = nil;
if(findmount(&nc, &mh, d.type, d.dev, d.qid)){
/*
* If it's a union directory and the original is
* in the union, don't rewrite anything.
*/
for(mount=mh->mount; mount; mount=mount->next)
if(eqchanddq(mount->to, d.type, d.dev, d.qid, 1))
goto Norewrite;
name = dirname(p, &nname);
/*
* Do the stat but fix the name. If it fails,
* leave old entry.
* BUG: If it fails because there isn't room for
* the entry, what can we do? Nothing, really.
* Might as well skip it.
*/
if(buf == nil){
buf = smalloc(4096);
nbuf = 4096;
}
if(waserror())
goto Norewrite;
l = nc->dev->stat(nc, buf, nbuf);
r = dirsetname(name, nname, buf, l, nbuf);
if(r == BIT16SZ)
error("dirsetname");
poperror();
/*
* Shift data in buffer to accomodate new entry,
* possibly overflowing into rock.
*/
rest = e - (p+dirlen);
if(r > dirlen){
while(p+r+rest > op+maxn){
mountrock(c, p, &e);
if(e == p){
dirlen = 0;
goto Norewrite;
}
rest = e - (p+dirlen);
}
}
if(r != dirlen){
memmove(p+r, p+dirlen, rest);
dirlen = r;
e = p+dirlen+rest;
}
/*
* Rewrite directory entry.
*/
memmove(p, buf, r);
Norewrite:
cclose(nc);
putmhead(mh);
}
}
if(buf)
free(buf);
if(p != e)
error("oops in mountfix");
return e-op;
}
static long
consread(Chan *c, void *buf, long n, vlong offset)
{
int l;
Osenv *o;
int ch, eol, i;
char *p, tmp[128];
char *cbuf = buf;
if(n <= 0)
return n;
o = up->env;
switch((ulong)c->qid.path){
case Qdir:
return devdirread(c, buf, n, consdir, nelem(consdir), devgen);
case Qsysctl:
return readstr(offset, buf, n, VERSION);
case Qcons:
case Qkeyboard:
qlock(&kbd);
if(waserror()) {
qunlock(&kbd);
nexterror();
}
if(kbd.raw || kbd.kbdr) {
if(qcanread(lineq))
n = qread(lineq, buf, n);
else {
/* read as much as possible */
do {
i = qread(kbdq, cbuf, n);
cbuf += i;
n -= i;
} while(n>0 && qcanread(kbdq));
n = cbuf - (char*)buf;
}
} else {
while(!qcanread(lineq)) {
qread(kbdq, &kbd.line[kbd.x], 1);
ch = kbd.line[kbd.x];
eol = 0;
switch(ch){
case '\b':
if(kbd.x)
kbd.x--;
break;
case 0x15:
kbd.x = 0;
break;
case '\n':
case 0x04:
eol = 1;
default:
kbd.line[kbd.x++] = ch;
break;
}
if(kbd.x == sizeof(kbd.line) || eol) {
if(ch == 0x04)
kbd.x--;
qwrite(lineq, kbd.line, kbd.x);
kbd.x = 0;
}
}
n = qread(lineq, buf, n);
}
qunlock(&kbd);
poperror();
return n;
case Qscancode:
if(offset == 0)
return readstr(0, buf, n, kscanid);
else
return qread(kscanq, buf, n);
case Qtime:
snprint(tmp, sizeof(tmp), "%.lld", (vlong)mseconds()*1000);
return readstr(offset, buf, n, tmp);
case Qhostowner:
return readstr(offset, buf, n, eve);
case Quser:
return readstr(offset, buf, n, o->user);
case Qjit:
snprint(tmp, sizeof(tmp), "%d", cflag);
return readstr(offset, buf, n, tmp);
case Qnull:
return 0;
case Qmsec:
return readnum(offset, buf, n, TK2MS(MACHP(0)->ticks), NUMSIZE);
case Qsysname:
if(sysname == nil)
return 0;
return readstr(offset, buf, n, sysname);
case Qnotquiterandom:
genrandom(buf, n);
return n;
case Qrandom:
return randomread(buf, n);
case Qmemory:
return poolread(buf, n, offset);
case Qdrivers:
p = malloc(READSTR);
if(p == nil)
error(Enomem);
l = 0;
for(i = 0; devtab[i] != nil; i++)
l += snprint(p+l, READSTR-l, "#%C %s\n", devtab[i]->dc, devtab[i]->name);
if(waserror()){
free(p);
nexterror();
}
n = readstr(offset, buf, n, p);
free(p);
poperror();
return n;
case Qklog:
return qread(klogq, buf, n);
case Qkprint:
rlock(&kprintq);
if(waserror()){
runlock(&kprintq);
nexterror();
}
n = qread(kprintq.q, buf, n);
poperror();
runlock(&kprintq);
return n;
default:
print("consread %llud\n", c->qid.path);
error(Egreg);
}
return -1; /* never reached */
}
static int32_t
read(int ispread, int fd, void *p, int32_t n, int64_t off)
{
Proc *up = externup();
int32_t nn, nnn;
Chan *c;
p = validaddr(p, n, 1);
c = fdtochan(fd, OREAD, 1, 1);
if(waserror()){
cclose(c);
nexterror();
}
/*
* The offset is passed through on directories, normally.
* Sysseek complains, but pread is used by servers like exportfs,
* that shouldn't need to worry about this issue.
*
* Notice that c->devoffset is the offset that c's dev is seeing.
* The number of bytes read on this fd (c->offset) may be different
* due to rewritings in mountfix.
*/
if(ispread){
if(off == ~0LL){ /* use and maintain channel's offset */
off = c->offset;
ispread = 0;
}
}
else
off = c->offset;
if(c->qid.type & QTDIR){
/*
* Directory read:
* rewind to the beginning of the file if necessary;
* try to fill the buffer via mountrockread;
* clear ispread to always maintain the Chan offset.
*/
if(off == 0LL){
if(!ispread){
c->offset = 0;
c->devoffset = 0;
}
mountrewind(c);
unionrewind(c);
}
if(!mountrockread(c, p, n, &nn)){
if(c->umh)
nn = unionread(c, p, n);
else{
if(off != c->offset)
error(Edirseek);
nn = c->dev->read(c, p, n, c->devoffset);
}
}
nnn = mountfix(c, p, nn, n);
ispread = 0;
}
else
nnn = nn = c->dev->read(c, p, n, off);
if(!ispread){
lock(&c->r.l);
c->devoffset += nn;
c->offset += nnn;
unlock(&c->r.l);
}
poperror();
cclose(c);
return nnn;
}
static Chan*
loopbackattach(char *spec)
{
Loop *volatile lb;
Queue *q;
Chan *c;
int chan;
int dev;
dev = 0;
if(spec != nil){
dev = atoi(spec);
if(dev >= Nloopbacks)
error(Ebadspec);
}
c = devattach('X', spec);
if(waserror()){
chanfree(c);
nexterror();
}
lb = &loopbacks[dev];
qlock(lb);
if(waserror()){
lb->ref--;
qunlock(lb);
nexterror();
}
lb->ref++;
if(lb->ref == 1){
for(chan = 0; chan < 2; chan++){
lb->link[chan].ci.mode = Trelative;
lb->link[chan].ci.a = &lb->link[chan];
lb->link[chan].ci.f = linkintr;
lb->link[chan].limit = Loopqlim;
q = qopen(lb->link[chan].limit, 0, 0, 0);
lb->link[chan].iq = q;
if(q == nil){
freelb(lb);
exhausted("memory");
}
q = qopen(lb->link[chan].limit, 0, 0, 0);
lb->link[chan].oq = q;
if(q == nil){
freelb(lb);
exhausted("memory");
}
lb->link[chan].indrop = 1;
lb->link[chan].delaynns = Delayn;
lb->link[chan].delay0ns = Delay0;
}
}
poperror();
qunlock(lb);
poperror();
mkqid(&c->qid, QID(0, Qtopdir), 0, QTDIR);
c->aux = lb;
c->dev = dev;
return c;
}
static char*
qcmd(Ctlr *ctlr, uint qid, uint code, uchar *data, int size, Block *block)
{
uchar *d, *c;
int pad;
TXQ *q;
assert(qid < nelem(ctlr->tx));
assert(size <= Tcmdsize-4);
ilock(ctlr);
q = &ctlr->tx[qid];
while(q->n >= Ntx && !ctlr->broken){
iunlock(ctlr);
qlock(q);
if(!waserror()){
tsleep(q, txqready, q, 10);
poperror();
}
qunlock(q);
ilock(ctlr);
}
if(ctlr->broken){
iunlock(ctlr);
return "qcmd: broken";
}
q->n++;
q->lastcmd = code;
q->b[q->i] = block;
c = q->c + q->i * Tcmdsize;
d = q->d + q->i * Tdscsize;
/* build command */
c[0] = code;
c[1] = 0; /* flags */
c[2] = q->i;
c[3] = qid;
if(size > 0)
memmove(c+4, data, size);
size += 4;
memset(d, 0, Tdscsize);
pad = size - 4;
if(block != nil)
pad += BLEN(block);
pad = ((pad + 3) & ~3) - pad;
put32(d, (pad << 28) | ((1 + (block != nil)) << 24)), d += 4;
put32(d, PCIWADDR(c)), d += 4;
put32(d, size), d += 4;
if(block != nil){
size = BLEN(block);
put32(d, PCIWADDR(block->rp)), d += 4;
put32(d, size), d += 4;
}
USED(d);
coherence();
q->i = (q->i+1) % Ntx;
csr32w(ctlr, HbusTargWptr, (qid<<8) | q->i);
iunlock(ctlr);
return nil;
}
static long
powerwrite(Chan* c, void *a, long n, vlong)
{
Cmdbuf *cmd;
Puser *p;
if(c->qid.type & QTDIR)
error(Ebadusefd);
cmd = parsecmd(a, n);
if(waserror()){
free(cmd);
nexterror();
}
switch(c->qid.path & ~CHDIR){
case Qdata:
p = c->aux;
if(cmd->nf < 2)
error(Ebadarg);
if(strcmp(cmd->f[0], "ack") == 0){
if(strcmp(cmd->f[1], "power") == 0){
lock(p);
if(p->state == Pwrack){
lock(&pwrusers.l);
if(pwrusers.shutdown && pwrusers.nwaiting > 0)
pwrusers.nwaiting--;
unlock(&pwrusers.l);
wakeup(&pwrusers.ackr);
p->state = Pwridle;
}
unlock(p);
}else
error(Ebadarg);
}else if(strcmp(cmd->f[0], "alarm") == 0){
/* set alarm */
}else
error(Ebadarg);
break;
case Qctl:
if(cmd->nf < 1)
error(Ebadarg);
if(strcmp(cmd->f[0], "suspend") == 0){
/* start the suspend action */
qlock(&pwrusers);
//powersuspend(0); /* calls poweringdown, then archsuspend() */
qunlock(&pwrusers);
}else if(strcmp(cmd->f[0], "shutdown") == 0){
/* go to it */
qlock(&pwrusers);
if(waserror()){
lock(&pwrusers.l);
pwrusers.shutdown = 0; /* hard luck for those already notified */
unlock(&pwrusers.l);
qunlock(&pwrusers);
nexterror();
}
lock(&pwrusers.l);
pwrusers.shutdown = 1;
pwrusers.nwaiting = 0;
unlock(&pwrusers.l);
for(p = pwrusers.list; p != nil; p = p->next){
lock(p);
if(p->state == Pwridle){
p->state = Pwroff;
lock(&pwrusers.l);
pwrusers.nwaiting++;
unlock(&pwrusers.l);
}
unlock(p);
wakeup(&p->r);
/* putting the tsleep here does each in turn; move out of loop to multicast */
tsleep(&pwrusers.ackr, alldown, nil, 1000);
}
poperror();
qunlock(&pwrusers);
//powersuspend(1);
}else
error(Ebadarg);
free(cmd);
break;
default:
error(Ebadusefd);
}
poperror();
return n;
}
/*
* Check all frames on device and resend any frames that have been
* outstanding for 200% of the device round trip time average.
*/
static void
aoesweepproc(void*)
{
ulong i, tx, timeout, nbc;
vlong starttick;
enum { Nms = 100, Nbcms = 30*1000, }; /* magic */
uchar *ea;
Aoeata *a;
Aoedev *d;
Devlink *l;
Frame *f, *e;
nbc = Nbcms/Nms;
loop:
if(nbc-- == 0){
if(rediscover && !waserror()){
discover(0xffff, 0xff);
poperror();
}
nbc = Nbcms/Nms;
}
starttick = MACHP(0)->ticks;
rlock(&devs);
for(d = devs.d; d; d = d->next){
if(!canqlock(d))
continue;
if(!UP(d)){
qunlock(d);
continue;
}
tx = 0;
f = d->frames;
e = f + d->nframes;
for (; f < e; f++){
if(f->tag == Tfree)
continue;
l = f->dl;
timeout = l->rttavg << 1;
i = tsince(f->tag);
if(i < timeout)
continue;
if(d->nout == d->maxout){
if(d->maxout > 1)
d->maxout--;
d->lastwadj = MACHP(0)->ticks;
}
a = (Aoeata*)f->hdr;
if(a->scnt > Dbcnt / Aoesectsz &&
++f->nl->lostjumbo > (d->nframes << 1)){
ea = f->dl->eatab[f->eaidx];
eventlog("%æ: jumbo failure on %s:%E; lba%lld\n",
d, f->nl->path, ea, f->lba);
d->maxbcnt = Dbcnt;
d->flag &= ~Djumbo;
}
resend(d, f);
if(tx++ == 0){
if((l->rttavg <<= 1) > Rtmax)
l->rttavg = Rtmax;
eventlog("%æ: rtt %ldms\n", d, TK2MS(l->rttavg));
}
}
if(d->nout == d->maxout && d->maxout < d->nframes &&
TK2MS(MACHP(0)->ticks - d->lastwadj) > 10*1000){ /* more magic */
d->maxout++;
d->lastwadj = MACHP(0)->ticks;
}
qunlock(d);
}
runlock(&devs);
i = Nms - TK2MS(MACHP(0)->ticks - starttick);
if(i > 0)
tsleep(&up->sleep, return0, 0, i);
goto loop;
}
static long
consread(Chan *c, void *va, long n, vlong offset)
{
int send;
char buf[64], ch;
if(c->qid.type & QTDIR)
return devdirread(c, va, n, contab, nelem(contab), devgen);
switch((ulong)c->qid.path) {
default:
error(Egreg);
case Qsysctl:
return readstr(offset, va, n, VERSION);
case Qsysname:
if(ossysname == nil)
return 0;
return readstr(offset, va, n, ossysname);
case Qrandom:
return randomread(va, n);
case Qnotquiterandom:
genrandom(va, n);
return n;
case Qhostowner:
return readstr(offset, va, n, eve);
case Qhoststdin:
return read(0, va, n); /* should be pread */
case Quser:
return readstr(offset, va, n, up->env->user);
case Qjit:
snprint(buf, sizeof(buf), "%d", cflag);
return readstr(offset, va, n, buf);
case Qtime:
snprint(buf, sizeof(buf), "%.lld", timeoffset + osusectime());
return readstr(offset, va, n, buf);
case Qdrivers:
return devtabread(c, va, n, offset);
case Qmemory:
return poolread(va, n, offset);
case Qnull:
return 0;
case Qmsec:
return readnum(offset, va, n, osmillisec(), NUMSIZE);
case Qcons:
qlock(&kbd.q);
if(waserror()){
qunlock(&kbd.q);
nexterror();
}
if(dflag)
error(Enonexist);
while(!qcanread(lineq)) {
if(qread(kbdq, &ch, 1) == 0)
continue;
send = 0;
if(ch == 0){
/* flush output on rawoff -> rawon */
if(kbd.x > 0)
send = !qcanread(kbdq);
}else if(kbd.raw){
kbd.line[kbd.x++] = ch;
send = !qcanread(kbdq);
}else{
switch(ch){
case '\b':
if(kbd.x)
kbd.x--;
break;
case 0x15:
kbd.x = 0;
break;
case 0x04:
send = 1;
break;
case '\n':
send = 1;
default:
kbd.line[kbd.x++] = ch;
break;
}
}
if(send || kbd.x == sizeof kbd.line){
qwrite(lineq, kbd.line, kbd.x);
kbd.x = 0;
}
}
n = qread(lineq, va, n);
qunlock(&kbd.q);
poperror();
return n;
case Qscancode:
if(offset == 0)
return readstr(0, va, n, gkscanid);
return qread(gkscanq, va, n);
case Qkeyboard:
return qread(gkbdq, va, n);
case Qkprint:
rlock(&kprintq.l);
if(waserror()){
runlock(&kprintq.l);
nexterror();
}
n = qread(kprintq.q, va, n);
poperror();
runlock(&kprintq.l);
return n;
}
}
static void
atarw(Aoedev *d, Frame *f)
{
ulong bcnt;
char extbit, writebit;
Aoeata *ah;
Srb *srb;
extbit = 0x4;
writebit = 0x10;
srb = d->inprocess;
bcnt = d->maxbcnt;
if(bcnt > srb->len)
bcnt = srb->len;
f->nhdr = AOEATASZ;
memset(f->hdr, 0, f->nhdr);
ah = (Aoeata*)f->hdr;
if(hset(d, f, ah, ACata) == -1) {
d->inprocess = nil;
return;
}
f->dp = srb->dp;
f->bcnt = bcnt;
f->lba = srb->sector;
f->srb = srb;
ah->scnt = bcnt / Aoesectsz;
putlba(ah, f->lba);
if(d->flag & Dllba)
ah->aflag |= AAFext;
else {
extbit = 0;
ah->lba[3] &= 0x0f;
ah->lba[3] |= 0xe0; /* LBA bit+obsolete 0xa0 */
}
if(srb->write){
ah->aflag |= AAFwrite;
f->dlen = bcnt;
}else{
writebit = 0;
f->dlen = 0;
}
ah->cmdstat = 0x20 | writebit | extbit;
/* mark tracking fields and load out */
srb->nout++;
srb->dp = (uchar*)srb->dp + bcnt;
srb->len -= bcnt;
srb->sector += bcnt / Aoesectsz;
if(srb->len == 0)
d->inprocess = nil;
d->nout++;
f->dl->npkt++;
if(waserror()){
f->tag = Tfree;
d->inprocess = nil;
nexterror();
}
devtab[f->nl->dc->type]->bwrite(f->nl->dc, allocfb(f), 0);
poperror();
}
static int32_t
consread(Chan *c, void *buf, int32_t n, int64_t off)
{
Proc *up = externup();
uint64_t l;
Mach *mp;
char *b, *bp, *s, *e;
char tmp[512]; /* Qswap is 381 bytes at clu */
int i, k, id;
int32_t offset;
if(n <= 0)
return n;
offset = off;
switch((uint32_t)c->qid.path){
case Qdir:
return devdirread(c, buf, n, consdir, nelem(consdir), devgen);
case Qcons:
error(Egreg);
case Qcputime:
k = offset;
if(k >= 6*NUMSIZE)
return 0;
if(k+n > 6*NUMSIZE)
n = 6*NUMSIZE - k;
/* easiest to format in a separate buffer and copy out */
for(i=0; i<6 && NUMSIZE*i<k+n; i++){
l = up->time[i];
if(i == TReal)
l = sys->ticks - l;
l = TK2MS(l);
readnum(0, tmp+NUMSIZE*i, NUMSIZE, l, NUMSIZE);
}
memmove(buf, tmp+k, n);
return n;
case Qkmesg:
/*
* This is unlocked to avoid tying up a process
* that's writing to the buffer. kmesg.n never
* gets smaller, so worst case the reader will
* see a slurred buffer.
*/
if(off >= kmesg.n)
n = 0;
else{
if(off+n > kmesg.n)
n = kmesg.n - off;
memmove(buf, kmesg.buf+off, n);
}
return n;
case Qkprint:
error(Egreg);
case Qpgrpid:
return readnum(offset, buf, n, up->pgrp->pgrpid, NUMSIZE);
case Qpid:
return readnum(offset, buf, n, up->pid, NUMSIZE);
case Qppid:
return readnum(offset, buf, n, up->parentpid, NUMSIZE);
case Qtime:
return readtime(offset, buf, n);
case Qbintime:
return readbintime(buf, n);
case Qhostowner:
return readstr(offset, buf, n, eve);
case Qhostdomain:
return readstr(offset, buf, n, hostdomain);
case Quser:
return readstr(offset, buf, n, up->user);
case Qnull:
return 0;
case Qsysstat:
n = MACHMAX*(NUMSIZE*11+2+1);
b = smalloc(n + 1); /* +1 for NUL */
bp = b;
e = bp + n;
for(id = 0; id < MACHMAX; id++)
if((mp = sys->machptr[id]) != nil && mp->online){
readnum(0, bp, NUMSIZE, mp->machno, NUMSIZE);
bp += NUMSIZE;
readnum(0, bp, NUMSIZE, mp->cs, NUMSIZE);
bp += NUMSIZE;
readnum(0, bp, NUMSIZE, mp->intr, NUMSIZE);
bp += NUMSIZE;
readnum(0, bp, NUMSIZE, mp->syscall, NUMSIZE);
bp += NUMSIZE;
readnum(0, bp, NUMSIZE, mp->pfault, NUMSIZE);
bp += NUMSIZE;
readnum(0, bp, NUMSIZE, mp->tlbfault, NUMSIZE);
bp += NUMSIZE;
readnum(0, bp, NUMSIZE, mp->tlbpurge, NUMSIZE);
bp += NUMSIZE;
readnum(0, bp, NUMSIZE, sys->load, NUMSIZE);
bp += NUMSIZE;
readnum(0, bp, NUMSIZE,
(mp->perf.avg_inidle*100)/mp->perf.period,
NUMSIZE);
bp += NUMSIZE;
readnum(0, bp, NUMSIZE,
(mp->perf.avg_inintr*100)/mp->perf.period,
NUMSIZE);
bp += NUMSIZE;
readnum(0, bp, NUMSIZE, 0, NUMSIZE); /* sched # */
bp += NUMSIZE;
bp = strecpy(bp, e, rolename[mp->NIX.nixtype]);
*bp++ = '\n';
}
if(waserror()){
free(b);
nexterror();
}
n = readstr(offset, buf, n, b);
free(b);
poperror();
return n;
case Qswap:
tmp[0] = 0;
s = seprintpagestats(tmp, tmp + sizeof tmp);
s = seprintphysstats(s, tmp + sizeof tmp);
b = buf;
l = s - tmp;
i = readstr(offset, b, l, tmp);
b += i;
n -= i;
if(offset > l)
offset -= l;
else
offset = 0;
return i + mallocreadsummary(c, b, n, offset);
case Qsysname:
if(sysname == nil)
return 0;
return readstr(offset, buf, n, sysname);
case Qrandom:
return randomread(buf, n);
case Qurandom:
return urandomread(buf, n);
case Qdrivers:
return devtabread(c, buf, n, off);
case Qzero:
memset(buf, 0, n);
return n;
case Qosversion:
snprint(tmp, sizeof tmp, "2000");
n = readstr(offset, buf, n, tmp);
return n;
case Qdebug:
s = seprint(tmp, tmp + sizeof tmp, "locks %lu\n", lockstats.locks);
s = seprint(s, tmp + sizeof tmp, "glare %lu\n", lockstats.glare);
s = seprint(s, tmp + sizeof tmp, "inglare %lu\n", lockstats.inglare);
s = seprint(s, tmp + sizeof tmp, "qlock %lu\n", qlockstats.qlock);
seprint(s, tmp + sizeof tmp, "qlockq %lu\n", qlockstats.qlockq);
return readstr(offset, buf, n, tmp);
break;
case Qsyscall:
snprint(tmp, sizeof tmp, "%s", printallsyscalls ? "on" : "off");
return readstr(offset, buf, n, tmp);
break;
default:
print("consread %#llx\n", c->qid.path);
error(Egreg);
}
return -1; /* never reached */
}
Walkqid*
devwalk(Chan *c, Chan *nc, char **name, int nname, Dirtab *tab, int ntab, Devgen *gen)
{
int i, j, alloc;
Walkqid *wq;
char *n;
Dir dir;
if(nname > 0)
isdir(c);
alloc = 0;
wq = smalloc(sizeof(Walkqid)+(nname-1)*sizeof(Qid));
if(waserror()){
if(alloc && wq->clone!=nil)
cclose(wq->clone);
free(wq);
return nil;
}
if(nc == nil){
nc = devclone(c);
nc->type = 0; /* device doesn't know about this channel yet */
alloc = 1;
}
wq->clone = nc;
for(j=0; j<nname; j++){
if(!(nc->qid.type&QTDIR)){
if(j==0)
error(Enotdir);
goto Done;
}
n = name[j];
if(strcmp(n, ".") == 0){
Accept:
wq->qid[wq->nqid++] = nc->qid;
continue;
}
if(strcmp(n, "..") == 0){
if((*gen)(nc, nil, tab, ntab, DEVDOTDOT, &dir) != 1){
print("devgen walk .. in dev%s %llux broken\n",
devtab[nc->type]->name, nc->qid.path);
error("broken devgen");
}
nc->qid = dir.qid;
goto Accept;
}
/*
* Ugly problem: If we're using devgen, make sure we're
* walking the directory itself, represented by the first
* entry in the table, and not trying to step into a sub-
* directory of the table, e.g. /net/net. Devgen itself
* should take care of the problem, but it doesn't have
* the necessary information (that we're doing a walk).
*/
if(gen==devgen && nc->qid.path!=tab[0].qid.path)
goto Notfound;
for(i=0;; i++) {
switch((*gen)(nc, n, tab, ntab, i, &dir)){
case -1:
Notfound:
if(j == 0)
error(Enonexist);
kstrcpy(up->errstr, Enonexist, ERRMAX);
goto Done;
case 0:
continue;
case 1:
if(strcmp(n, dir.name) == 0){
nc->qid = dir.qid;
goto Accept;
}
continue;
}
}
}
/*
* We processed at least one name, so will return some data.
* If we didn't process all nname entries succesfully, we drop
* the cloned channel and return just the Qids of the walks.
*/
Done:
poperror();
if(wq->nqid < nname){
if(alloc)
cclose(wq->clone);
wq->clone = nil;
}else if(wq->clone){
/* attach cloned channel to same device */
wq->clone->type = c->type;
}
return wq;
}
static long
capwrite(Chan *c, void *va, long n, vlong vl)
{
Caphash *p;
char *cp;
uchar hash[Hashlen];
char *key, *from, *to;
char err[256];
switch((ulong)c->qid.path){
case Qhash:
if(!iseve())
error(Eperm);
if(n < Hashlen)
error(Eshort);
memmove(hash, va, Hashlen);
addcap(hash);
break;
case Quse:
/* copy key to avoid a fault in hmac_xx */
cp = nil;
if(waserror()){
free(cp);
nexterror();
}
cp = smalloc(n+1);
memmove(cp, va, n);
cp[n] = 0;
from = cp;
key = strrchr(cp, '@');
if(key == nil)
error(Eshort);
*key++ = 0;
hmac_sha1((uchar*)from, strlen(from), (uchar*)key, strlen(key), hash, nil);
p = remcap(hash);
if(p == nil){
snprint(err, sizeof err, "invalid capability %s@%s", from, key);
error(err);
}
/* if a from user is supplied, make sure it matches */
to = strchr(from, '@');
if(to == nil){
to = from;
} else {
*to++ = 0;
if(strcmp(from, up->user) != 0)
error("capability must match user");
}
/* set user id */
kstrdup(&up->user, to);
up->basepri = PriNormal;
free(p);
free(cp);
poperror();
break;
default:
error(Eperm);
break;
}
return n;
}
static long
conswrite(Chan *c, void *va, long n, vlong offset)
{
vlong t;
long l, bp;
char *a = va;
Cmdbuf *cb;
Cmdtab *ct;
char buf[256];
int x;
switch((ulong)c->qid.path){
case Qcons:
/*
* Can't page fault in putstrn, so copy the data locally.
*/
l = n;
while(l > 0){
bp = l;
if(bp > sizeof buf)
bp = sizeof buf;
memmove(buf, a, bp);
putstrn0(a, bp, 1);
a += bp;
l -= bp;
}
break;
case Qconsctl:
if(n >= sizeof(buf))
n = sizeof(buf)-1;
strncpy(buf, a, n);
buf[n] = 0;
for(a = buf; a;){
if(strncmp(a, "rawon", 5) == 0){
qlock(&kbd);
flushkbdline(kbdq);
kbd.raw = 1;
qunlock(&kbd);
} else if(strncmp(a, "rawoff", 6) == 0){
qlock(&kbd);
kbd.raw = 0;
kbd.x = 0;
qunlock(&kbd);
}
if(a = strchr(a, ' '))
a++;
}
break;
case Qkeyboard:
for(x=0; x<n; ) {
Rune r;
x += chartorune(&r, &a[x]);
kbdputc(kbdq, r);
}
break;
case Qtime:
if(n >= sizeof(buf))
n = sizeof(buf)-1;
strncpy(buf, a, n);
buf[n] = 0;
t = strtoll(buf, 0, 0)/1000000;
boottime = t - TK2SEC(MACHP(0)->ticks);
break;
case Qhostowner:
if(!iseve())
error(Eperm);
if(offset != 0 || n >= sizeof(buf))
error(Ebadarg);
memmove(buf, a, n);
buf[n] = '\0';
if(n > 0 && buf[n-1] == '\n')
buf[--n] = 0;
if(n <= 0)
error(Ebadarg);
renameuser(eve, buf);
renameproguser(eve, buf);
kstrdup(&eve, buf);
kstrdup(&up->env->user, buf);
break;
case Quser:
if(!iseve())
error(Eperm);
if(offset != 0)
error(Ebadarg);
if(n <= 0 || n >= sizeof(buf))
error(Ebadarg);
strncpy(buf, a, n);
buf[n] = 0;
if(buf[n-1] == '\n')
buf[n-1] = 0;
kstrdup(&up->env->user, buf);
break;
case Qjit:
if(n >= sizeof(buf))
n = sizeof(buf)-1;
strncpy(buf, va, n);
buf[n] = '\0';
x = atoi(buf);
if(x < 0 || x > 9)
error(Ebadarg);
cflag = x;
return n;
case Qnull:
break;
case Qsysname:
if(offset != 0)
error(Ebadarg);
if(n <= 0 || n >= sizeof(buf))
error(Ebadarg);
strncpy(buf, a, n);
buf[n] = 0;
if(buf[n-1] == '\n')
buf[n-1] = 0;
kstrdup(&sysname, buf);
break;
case Qsysctl:
if(!iseve())
error(Eperm);
cb = parsecmd(a, n);
if(waserror()){
free(cb);
nexterror();
}
ct = lookupcmd(cb, sysctlcmd, nelem(sysctlcmd));
switch(ct->index){
case CMreboot:
reboot();
break;
case CMhalt:
halt();
break;
case CMpanic:
panic("sysctl");
case CMconsole:
consoleprint = strcmp(cb->f[1], "off") != 0;
break;
case CMbroken:
keepbroken = 1;
break;
case CMnobroken:
keepbroken = 0;
break;
}
poperror();
free(cb);
break;
default:
print("conswrite: %llud\n", c->qid.path);
error(Egreg);
}
return n;
}
long
gpiowrite(Chan *c, void *va, long n, vlong)
{
int type, i, scheme;
uint pin;
char *arg;
Cmdbuf *cb;
Cmdtab *ct;
if(c->qid.type & QTDIR)
{
error(Eisdir);
}
type = FILE_TYPE(c->qid);
scheme = SCHEME_TYPE(c->qid);
if(scheme != Qgeneric && scheme != pinscheme)
{
error(nil);
}
cb = parsecmd(va, n);
if(waserror())
{
free(cb);
nexterror();
}
ct = lookupcmd(cb, gpiocmd, nelem(gpiocmd));
if(ct == nil)
{
error(Ebadctl);
}
switch(type)
{
case Qdata:
pin = PIN_NUMBER(c->qid);
switch(ct->index)
{
case CMzero:
gpioout(pin, 0);
break;
case CMone:
gpioout(pin, 1);
break;
default:
error(Ebadctl);
}
break;
case Qctl:
switch(ct->index)
{
case CMscheme:
arg = cb->f[1];
for(i = 0; i < nelem(schemename); i++)
{
if(strncmp(schemename[i], arg, strlen(schemename[i])) == 0)
{
pinscheme = i;
break;
}
}
break;
case CMfunc:
pin = getpin(cb->f[2]);
arg = cb->f[1];
if(pin == -1) {
error(Ebadctl);
}
for(i = 0; i < nelem(funcname); i++)
{
if(strncmp(funcname[i], arg, strlen(funcname[i])) == 0)
{
gpiofuncset(pin, i);
break;
}
}
break;
case CMpull:
pin = getpin(cb->f[2]);
if(pin == -1) {
error(Ebadctl);
}
arg = cb->f[1];
for(i = 0; i < nelem(pudname); i++)
{
if(strncmp(pudname[i], arg, strlen(pudname[i])) == 0)
{
gpiopullset(pin, i);
break;
}
}
break;
case CMevent:
pin = getpin(cb->f[3]);
if(pin == -1) {
error(Ebadctl);
}
arg = cb->f[1];
for(i = 0; i < nelem(evtypename); i++)
{
if(strncmp(evtypename[i], arg, strlen(evtypename[i])) == 0)
{
gpioevent(pin, i, (cb->f[2][0] == 'e'));
break;
}
}
break;
default:
error(Ebadctl);
}
break;
}
free(cb);
poperror();
return n;
}
static void
w_timer(void* arg)
{
Ether* ether = (Ether*) arg;
Ctlr* ctlr = (Ctlr*)ether->ctlr;
ctlr->timerproc = up;
while(waserror())
;
for(;;){
tsleep(&up->sleep, return0, 0, MSperTick);
ctlr = (Ctlr*)ether->ctlr;
if(ctlr == 0)
break;
if((ctlr->state & (Attached|Power)) != (Attached|Power))
continue;
ctlr->ticks++;
ilock(ctlr);
// Seems that the card gets frames BUT does
// not send the interrupt; this is a problem because
// I suspect it runs out of receive buffers and
// stops receiving until a transmit watchdog
// reenables the card.
// The problem is serious because it leads to
// poor rtts.
// This can be seen clearly by commenting out
// the next if and doing a ping: it will stop
// receiving (although the icmp replies are being
// issued from the remote) after a few seconds.
// Of course this `bug' could be because I'm reading
// the card frames in the wrong way; due to the
// lack of documentation I cannot know.
if(csr_ins(ctlr, WR_EvSts)&WEvs){
ctlr->tickintr++;
w_intr(ether);
}
if((ctlr->ticks % 10) == 0) {
if(ctlr->txtmout && --ctlr->txtmout == 0){
ctlr->nwatchdogs++;
w_txdone(ctlr, WTxErrEv);
if(w_enable(ether)){
DEBUG("wavelan: wdog enable failed\n");
}
w_txstart(ether);
}
if((ctlr->ticks % 120) == 0)
if(ctlr->txbusy == 0)
w_cmd(ctlr, WCmdEnquire, WTyp_Stats);
if(ctlr->scanticks > 0)
if((ctlr->ticks % ctlr->scanticks) == 0)
if(ctlr->txbusy == 0)
w_cmd(ctlr, WCmdEnquire, WTyp_Scan);
}
iunlock(ctlr);
}
pexit("terminated", 1);
}