void
pexit(char *exitstr, int freemem)
{
Proc *up = externup();
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 && up->nfullq > 0)
iprint(" %s=%d", up->text, up->nfullq);
if(0 && up->nicc > 0)
iprint(" [%s nicc %ud tctime %ulld actime %ulld]\n",
up->text, up->nicc, up->tctime, up->actime);
if(up->syscalltrace != nil)
free(up->syscalltrace);
up->syscalltrace = nil;
up->alarm = 0;
clearwakeups(up);
if (up->Timer.tt)
timerdel(&up->Timer);
if(up->trace)
proctrace(up, SDead, 0);
/* nil out all the resources under lock (free later) */
qlock(&up->debug);
fgrp = up->fgrp;
up->fgrp = nil;
egrp = up->egrp;
up->egrp = nil;
rgrp = up->rgrp;
up->rgrp = nil;
pgrp = up->pgrp;
up->pgrp = nil;
dot = up->dot;
up->dot = nil;
qunlock(&up->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(up->kp == 0) {
p = up->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 = up->pid;
utime = up->time[TUser] + up->time[TCUser];
stime = up->time[TSys] + up->time[TCSys];
wq->w.time[TUser] = tk2ms(utime);
wq->w.time[TSys] = tk2ms(stime);
wq->w.time[TReal] = tk2ms(sys->machptr[0]->ticks - up->time[TReal]);
if(exitstr && exitstr[0])
snprint(wq->w.msg, sizeof(wq->w.msg), "%s %d: %s",
up->text, up->pid, exitstr);
else
wq->w.msg[0] = '\0';
lock(&p->exl);
/*
* Check that parent is still alive.
*/
if(p->pid == up->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(up);
qlock(&up->seglock);
es = &up->seg[NSEG];
for(s = up->seg; s < es; s++) {
if(*s) {
putseg(*s);
*s = 0;
}
}
qunlock(&up->seglock);
lock(&up->exl); /* Prevent my children from leaving waits */
psunhash(up);
up->pid = 0;
wakeup(&up->waitr);
unlock(&up->exl);
for(f = up->waitq; f; f = next) {
next = f->next;
free(f);
}
/* release debuggers */
qlock(&up->debug);
if(up->pdbg) {
wakeup(&up->pdbg->sleep);
up->pdbg = 0;
}
qunlock(&up->debug);
/* Sched must not loop for these locks */
lock(&procalloc.l);
lock(&pga.l);
stopac();
edfstop(up);
up->state = Moribund;
sched();
panic("pexit");
}
static void
releaseintr(Ureg* ureg, Timer *t)
{
Proc *up = externup();
Proc *p;
extern int panicking;
Sched *sch;
Schedq *rq;
if(panicking || active.exiting)
return;
p = t->ta;
if((edflock(p)) == nil)
return;
sch = procsched(p);
DPRINT("%lu releaseintr %d[%s]\n", now, p->pid, statename[p->state]);
switch(p->state){
default:
edfunlock();
return;
case Ready:
/* remove proc from current runq */
rq = &sch->runq[p->priority];
if(dequeueproc(sch, rq, p) != p){
DPRINT("releaseintr: can't find proc or lock race\n");
release(p); /* It'll start best effort */
edfunlock();
return;
}
p->state = Waitrelease;
/* fall through */
case Waitrelease:
release(p);
edfunlock();
if(p->state == Wakeme){
iprint("releaseintr: wakeme\n");
}
ready(p);
if(up){
up->delaysched++;
sch->delayedscheds++;
}
return;
case Running:
release(p);
edfrun(p, 1);
break;
case Wakeme:
release(p);
edfunlock();
if(p->trend)
wakeup(p->trend);
p->trend = nil;
if(up){
up->delaysched++;
sch->delayedscheds++;
}
return;
}
edfunlock();
}
/*
* 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);
}
/*
* Call user, if necessary, with note.
* Pass user the Ureg struct and the note on his stack.
*/
int
notify(Ureg* ureg)
{
Proc *up = externup();
int l;
Mpl pl;
Note note;
uintptr_t sp;
NFrame *nf;
/*
* Calls procctl splhi, see comment in procctl for the reasoning.
*/
if(up->procctl)
procctl(up);
if(up->nnote == 0)
return 0;
fpunotify(ureg);
pl = spllo();
qlock(&up->debug);
up->notepending = 0;
memmove(¬e, &up->note[0], sizeof(Note));
if(strncmp(note.msg, "sys:", 4) == 0){
l = strlen(note.msg);
if(l > ERRMAX-sizeof(" pc=0x0123456789abcdef"))
l = ERRMAX-sizeof(" pc=0x0123456789abcdef");
sprint(note.msg+l, " pc=%#p", ureg->ip);
}
if(note.flag != NUser && (up->notified || up->notify == nil)){
qunlock(&up->debug);
if(note.flag == NDebug)
pprint("suicide: %s\n", note.msg);
pexit(note.msg, note.flag != NDebug);
}
if(up->notified){
qunlock(&up->debug);
splhi();
return 0;
}
if(up->notify == nil){
qunlock(&up->debug);
pexit(note.msg, note.flag != NDebug);
}
if(!okaddr(PTR2UINT(up->notify), sizeof(ureg->ip), 0)){
qunlock(&up->debug);
pprint("suicide: bad function address %#p in notify\n",
up->notify);
pexit("Suicide", 0);
}
sp = ureg->sp - ROUNDUP(sizeof(NFrame), 16) - 128; // amd64 red zone, also wanted by go stack traces
if(!okaddr(sp, sizeof(NFrame), 1)){
qunlock(&up->debug);
pprint("suicide: bad stack address %#p in notify\n", sp);
pexit("Suicide", 0);
}
nf = UINT2PTR(sp);
memmove(&nf->ureg, ureg, sizeof(Ureg));
nf->old = up->ureg;
up->ureg = nf; /* actually the NFrame, for noted */
memmove(nf->msg, note.msg, ERRMAX);
nf->arg1 = nf->msg;
nf->arg0 = &nf->ureg;
ureg->di = (uintptr)nf->arg0;
ureg->si = (uintptr)nf->arg1;
//print("Setting di to %p and si to %p\n", ureg->di, ureg->si);
ureg->bp = PTR2UINT(nf->arg0);
nf->ip = 0;
ureg->sp = sp;
ureg->ip = PTR2UINT(up->notify);
up->notified = 1;
up->nnote--;
memmove(&up->lastnote, ¬e, sizeof(Note));
memmove(&up->note[0], &up->note[1], up->nnote*sizeof(Note));
qunlock(&up->debug);
splx(pl);
return 1;
}
/*
* Return user to state before notify()
*/
void
noted(Ureg* cur, uintptr_t arg0)
{
Proc *up = externup();
NFrame *nf;
Note note;
Ureg *nur;
qlock(&up->debug);
if(arg0 != NRSTR && !up->notified){
qunlock(&up->debug);
pprint("suicide: call to noted when not notified\n");
pexit("Suicide", 0);
}
up->notified = 0;
fpunoted();
nf = up->ureg;
/* sanity clause */
if(!okaddr(PTR2UINT(nf), sizeof(NFrame), 0)){
qunlock(&up->debug);
pprint("suicide: bad ureg %#p in noted\n", nf);
pexit("Suicide", 0);
}
/*
* Check the segment selectors are all valid.
*/
nur = &nf->ureg;
if(nur->cs != SSEL(SiUCS, SsRPL3) || nur->ss != SSEL(SiUDS, SsRPL3)) {
qunlock(&up->debug);
pprint("suicide: bad segment selector (cs %p want %p, ss %p want %p), in noted\n",
nur->cs, SSEL(SiUCS, SsRPL3),
nur->ss, SSEL(SiUDS, SsRPL3)
);
pexit("Suicide", 0);
}
/* don't let user change system flags */
nur->flags &= (Of|Df|Sf|Zf|Af|Pf|Cf);
nur->flags |= cur->flags & ~(Of|Df|Sf|Zf|Af|Pf|Cf);
memmove(cur, nur, sizeof(Ureg));
switch((int)arg0){
case NCONT:
case NRSTR:
if(!okaddr(nur->ip, BY2SE, 0) || !okaddr(nur->sp, BY2SE, 0)){
qunlock(&up->debug);
pprint("suicide: trap in noted pc=%#p sp=%#p\n",
nur->ip, nur->sp);
pexit("Suicide", 0);
}
up->ureg = nf->old;
qunlock(&up->debug);
break;
case NSAVE:
if(!okaddr(nur->ip, BY2SE, 0) || !okaddr(nur->sp, BY2SE, 0)){
qunlock(&up->debug);
pprint("suicide: trap in noted pc=%#p sp=%#p\n",
nur->ip, nur->sp);
pexit("Suicide", 0);
}
qunlock(&up->debug);
splhi();
nf->arg1 = nf->msg;
nf->arg0 = &nf->ureg;
cur->bp = PTR2UINT(nf->arg0);
nf->ip = 0;
cur->sp = PTR2UINT(nf);
break;
default:
memmove(¬e, &up->lastnote, sizeof(Note));
qunlock(&up->debug);
pprint("suicide: bad arg %#p in noted: %s\n", arg0, note.msg);
pexit(note.msg, 0);
break;
case NDFLT:
memmove(¬e, &up->lastnote, sizeof(Note));
qunlock(&up->debug);
if(note.flag == NDebug)
pprint("suicide: %s\n", note.msg);
pexit(note.msg, note.flag != NDebug);
break;
}
}
/*
* the devxxx.c that calls us handles writing data, it knows best
*/
int32_t
netifwrite(Netif *nif, Chan *c, void *a, int32_t n)
{
Proc *up = externup();
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;
}
/*
* pg->pgszi indicates the page size in machp()->pgsz[] used for the mapping.
* For the user, it can be either 2*MiB or 1*GiB pages.
* For 2*MiB pages, we use three levels, not four.
* For 1*GiB pages, we use two levels.
*/
void
mmuput(uintptr_t va, Page *pg, uint attr)
{
Proc *up = externup();
int lvl, user, x, pgsz;
PTE *pte;
Page *page, *prev;
Mpl pl;
uintmem pa, ppn;
char buf[80];
ppn = 0;
pa = pg->pa;
if(pa == 0)
panic("mmuput: zero pa");
if(DBGFLG) {
snprint(buf, sizeof buf, "cpu%d: up %#p mmuput %#p %#P %#ux\n",
machp()->machno, up, va, pa, attr);
print("%s", buf);
}
assert(pg->pgszi >= 0);
pgsz = sys->pgsz[pg->pgszi];
if(pa & (pgsz-1))
panic("mmuput: pa offset non zero: %#ullx\n", pa);
pa |= pteflags(attr);
pl = splhi();
if(DBGFLG)
mmuptpcheck(up);
user = (va < KZERO);
x = PTLX(va, 3);
pte = UINT2PTR(machp()->MMU.pml4->va);
pte += x;
prev = machp()->MMU.pml4;
for(lvl = 3; lvl >= 0; lvl--) {
if(user) {
if(pgsz == 2*MiB && lvl == 1) /* use 2M */
break;
if(pgsz == 1ull*GiB && lvl == 2) /* use 1G */
break;
}
for(page = up->MMU.mmuptp[lvl]; page != nil; page = page->next)
if(page->prev == prev && page->daddr == x) {
if(*pte == 0) {
print("mmu: jmk and nemo had fun\n");
*pte = PPN(page->pa)|PteU|PteRW|PteP;
}
break;
}
if(page == nil) {
if(up->MMU.mmuptp[0] == nil)
page = mmuptpalloc();
else {
page = up->MMU.mmuptp[0];
up->MMU.mmuptp[0] = page->next;
}
page->daddr = x;
page->next = up->MMU.mmuptp[lvl];
up->MMU.mmuptp[lvl] = page;
page->prev = prev;
*pte = PPN(page->pa)|PteU|PteRW|PteP;
if(lvl == 3 && x >= machp()->MMU.pml4->daddr)
machp()->MMU.pml4->daddr = x+1;
}
x = PTLX(va, lvl-1);
ppn = PPN(*pte);
if(ppn == 0)
panic("mmuput: ppn=0 l%d pte %#p = %#P\n", lvl, pte, *pte);
pte = UINT2PTR(KADDR(ppn));
pte += x;
prev = page;
}
if(DBGFLG)
checkpte(ppn, pte);
*pte = pa|PteU;
if(user)
switch(pgsz) {
case 2*MiB:
case 1*GiB:
*pte |= PtePS;
break;
default:
panic("mmuput: user pages must be 2M or 1G");
}
splx(pl);
if(DBGFLG) {
snprint(buf, sizeof buf, "cpu%d: up %#p new pte %#p = %#llux\n",
machp()->machno, up, pte, pte?*pte:~0);
print("%s", buf);
}
invlpg(va); /* only if old entry valid? */
}
int
ipoput6(Fs *f, Block *bp, int gating, int ttl, int tos, Conv *c)
{
Proc *up = externup();
int medialen, len, chunk, uflen, flen, seglen, lid, offset, fragoff;
int morefrags, blklen, rv = 0, tentative;
uint8_t *gate, nexthdr;
Block *xp, *nb;
Fraghdr6 fraghdr;
IP *ip;
Ip6hdr *eh;
Ipifc *ifc;
Route *r, *sr;
ip = f->ip;
/* Fill out the ip header */
eh = (Ip6hdr*)(bp->rp);
ip->stats[OutRequests]++;
/* Number of uint8_ts in data and ip header to write */
len = blocklen(bp);
tentative = iptentative(f, eh->src);
if(tentative){
netlog(f, Logip, "reject tx of packet with tentative src address %I\n",
eh->src);
goto free;
}
if(gating){
chunk = nhgets(eh->ploadlen);
if(chunk > len){
ip->stats[OutDiscards]++;
netlog(f, Logip, "short gated packet\n");
goto free;
}
if(chunk + IP6HDR < len)
len = chunk + IP6HDR;
}
if(len >= IP_MAX){
ip->stats[OutDiscards]++;
netlog(f, Logip, "exceeded ip max size %I\n", eh->dst);
goto free;
}
r = v6lookup(f, eh->dst, c);
if(r == nil){
// print("no route for %I, src %I free\n", eh->dst, eh->src);
ip->stats[OutNoRoutes]++;
netlog(f, Logip, "no interface %I\n", eh->dst);
rv = -1;
goto free;
}
ifc = r->RouteTree.ifc;
if(r->RouteTree.type & (Rifc|Runi))
gate = eh->dst;
else if(r->RouteTree.type & (Rbcast|Rmulti)) {
gate = eh->dst;
sr = v6lookup(f, eh->src, nil);
if(sr && (sr->RouteTree.type & Runi))
ifc = sr->RouteTree.ifc;
}
else
gate = r->v6.gate;
if(!gating)
eh->vcf[0] = IP_VER6;
eh->ttl = ttl;
if(!gating) {
eh->vcf[0] |= tos >> 4;
eh->vcf[1] = tos << 4;
}
static int
procgen(Chan *c, char *name, Dirtab *tab, int j, int s, Dir *dp)
{
Proc *up = externup();
Qid qid;
Proc *p;
char *ename;
int pid, sno;
uint32_t path, perm, len;
if(s == DEVDOTDOT){
mkqid(&qid, Qdir, 0, QTDIR);
devdir(c, qid, "#p", 0, eve, 0555, dp);
return 1;
}
if(c->qid.path == Qdir){
if(s == 0){
strcpy(up->genbuf, "trace");
mkqid(&qid, Qtrace, -1, QTFILE);
devdir(c, qid, up->genbuf, 0, eve, 0444, dp);
return 1;
}
if(s == 1){
strcpy(up->genbuf, "tracepids");
mkqid(&qid, Qtracepids, -1, QTFILE);
devdir(c, qid, up->genbuf, 0, eve, 0444, dp);
return 1;
}
s -= 2;
if(name != nil){
/* ignore s and use name to find pid */
pid = strtol(name, &ename, 10);
if(pid<=0 || ename[0]!='\0')
return -1;
s = psindex(pid);
if(s < 0)
return -1;
}
else if(s >= conf.nproc)
return -1;
if((p = psincref(s)) == nil || (pid = p->pid) == 0)
return 0;
snprint(up->genbuf, sizeof up->genbuf, "%u", pid);
/*
* String comparison is done in devwalk so
* name must match its formatted pid.
*/
if(name != nil && strcmp(name, up->genbuf) != 0)
return -1;
mkqid(&qid, (s+1)<<QSHIFT, pid, QTDIR);
devdir(c, qid, up->genbuf, 0, p->user, DMDIR|0555, dp);
psdecref(p);
return 1;
}
if(c->qid.path == Qtrace){
strcpy(up->genbuf, "trace");
mkqid(&qid, Qtrace, -1, QTFILE);
devdir(c, qid, up->genbuf, 0, eve, 0444, dp);
return 1;
}
if(c->qid.path == Qtracepids){
strcpy(up->genbuf, "tracepids");
mkqid(&qid, Qtrace, -1, QTFILE);
devdir(c, qid, up->genbuf, 0, eve, 0444, dp);
return 1;
}
if(s >= nelem(procdir))
return -1;
if(tab)
panic("procgen");
tab = &procdir[s];
path = c->qid.path&~(((1<<QSHIFT)-1)); /* slot component */
if((p = psincref(SLOT(c->qid))) == nil)
return -1;
perm = tab->perm;
if(perm == 0)
perm = p->procmode;
else /* just copy read bits */
perm |= p->procmode & 0444;
len = tab->length;
switch(QID(c->qid)) {
case Qwait:
len = p->nwait; /* incorrect size, but >0 means there's something to read */
break;
case Qprofile: /* TODO(aki): test this */
len = 0;
for(sno = 0; sno < NSEG; sno++){
if(p->seg[sno] != nil && (p->seg[sno]->type & SG_EXEC) != 0){
Segment *s;
s = p->seg[sno];
if(s->profile)
len += ((s->top-s->base)>>LRESPROF) * sizeof s->profile[0];
}
}
break;
}
void
mmuinit(void)
{
Proc *up = externup();
uint8_t *p;
Page *page;
uint64_t o, pa, r, sz;
archmmu();
DBG("mach%d: %#p pml4 %#p npgsz %d\n", machp()->machno, m, machp()->pml4, m->npgsz);
if(machp()->machno != 0){
/* NIX: KLUDGE: Has to go when each mach is using
* its own page table
*/
p = UINT2PTR(m->stack);
p += MACHSTKSZ;
memmove(p, UINT2PTR(mach0pml4.va), PTSZ);
machp()->pml4 = &machp()->pml4kludge;
machp()->pml4->va = PTR2UINT(p);
machp()->pml4->pa = PADDR(p);
machp()->pml4->daddr = mach0pml4.daddr; /* # of user mappings in pml4 */
r = rdmsr(Efer);
r |= Nxe;
wrmsr(Efer, r);
cr3put(machp()->pml4->pa);
DBG("m %#p pml4 %#p\n", m, machp()->pml4);
return;
}
page = &mach0pml4;
page->pa = cr3get();
page->va = PTR2UINT(KADDR(page->pa));
machp()->pml4 = page;
r = rdmsr(Efer);
r |= Nxe;
wrmsr(Efer, r);
/*
* Set up the various kernel memory allocator limits:
* pmstart/pmend bound the unused physical memory;
* vmstart/vmend bound the total possible virtual memory
* used by the kernel;
* vmunused is the highest virtual address currently mapped
* and used by the kernel;
* vmunmapped is the highest virtual address currently
* mapped by the kernel.
* Vmunused can be bumped up to vmunmapped before more
* physical memory needs to be allocated and mapped.
*
* This is set up here so meminit can map appropriately.
*/
o = sys->pmstart;
sz = ROUNDUP(o, 4*MiB) - o;
pa = asmalloc(0, sz, 1, 0);
if(pa != o)
panic("mmuinit: pa %#llux memstart %#llux\n", pa, o);
sys->pmstart += sz;
sys->vmstart = KSEG0;
sys->vmunused = sys->vmstart + ROUNDUP(o, 4*KiB);
sys->vmunmapped = sys->vmstart + o + sz;
sys->vmend = sys->vmstart + TMFM;
print("mmuinit: vmstart %#p vmunused %#p vmunmapped %#p vmend %#p\n",
sys->vmstart, sys->vmunused, sys->vmunmapped, sys->vmend);
/*
* Set up the map for PD entry access by inserting
* the relevant PDP entry into the PD. It's equivalent
* to PADDR(sys->pd)|PteRW|PteP.
*
*/
sys->pd[PDX(PDMAP)] = sys->pdp[PDPX(PDMAP)] & ~(PteD|PteA);
print("sys->pd %#p %#p\n", sys->pd[PDX(PDMAP)], sys->pdp[PDPX(PDMAP)]);
assert((pdeget(PDMAP) & ~(PteD|PteA)) == (PADDR(sys->pd)|PteRW|PteP));
dumpmmuwalk(KZERO);
mmuphysaddr(PTR2UINT(end));
}
static int32_t
pmcwrite(Chan *c, void *a, int32_t n, int64_t mm)
{
Proc *up = externup();
Cmdbuf *cb;
Cmdtab *ct;
uint32_t type;
char str[64]; /* 0x0000000000000000\0 */
AcPmcArg p;
AcCtrArg ctr;
uint64_t coreno;
Mach *mp;
if (c->qid.type == QTDIR)
error(Eperm);
if (c->qid.path == Qgctl)
error(Eperm);
if (n >= sizeof(str))
error(Ebadctl);
pmcnull(&p);
coreno = (uint64_t)c->aux;
p.coreno = coreno;
type = PMCTYPE(c->qid.path);
p.regno = PMCID(c->qid.path);
memmove(str, a, n);
str[n] = '\0';
mp = up->ac;
ctr.coreno = coreno;
ctr.regno = p.regno;
if (type == Qdata) {
/* I am a handler for a proc in the core, run an RPC*/
if (mp != nil && mp->machno == coreno) {
if (runac(mp, acpmcsetctr, 0, &ctr, sizeof(AcCtrArg)) < 0)
n = -1;
} else {
if (pmcsetctr(coreno, strtoull(str, 0, 0), p.regno) < 0)
n = -1;
}
return n;
}
/* TODO: should iterate through multiple lines */
if (strncmp(str, "set ", 4) == 0){
memmove(p.descstr, (char *)str + 4, n - 4);
p.descstr[n - 4] = '\0';
p.nodesc = 0;
} else {
cb = parsecmd(a, n);
if(waserror()){
free(cb);
nexterror();
}
ct = lookupcmd(cb, pmcctlmsg, nelem(pmcctlmsg));
switch(ct->index){
case Enable:
p.enab = 1;
break;
case Disable:
p.enab = 0;
break;
case User:
p.user = 1;
break;
case Os:
p.os = 1;
break;
case NoUser:
p.user = 0;
break;
case NoOs:
p.os = 0;
break;
case Reset:
p.reset = 1;
break;
case Debug:
pmcdebug = ~pmcdebug;
break;
default:
cmderror(cb, "invalid ctl");
break;
}
free(cb);
poperror();
}
/* I am a handler for a proc in the core, run an RPC*/
if (mp != nil && mp->machno == coreno) {
if (runac(mp, acpmcsetctl, 0, &p, sizeof(AcPmcArg)) < 0)
n = -1;
} else {
if (pmcsetctl(coreno, &p, p.regno) < 0)
n = -1;
}
return n;
}
int
anyhigher(void)
{
Proc *up = externup();
return run.runvec & ~((1<<(up->priority+1))-1);
}
void
nexterror(void)
{
Proc *up = externup();
/*debug*/gotolabel(&up->errlab[--up->nerrlab]);
}
/*
* If changing this routine, look also at sleep(). It
* contains a copy of the guts of sched().
*/
void
sched(void)
{
Proc *up = externup();
Proc *p;
if(!islo() && machp()->ilockdepth)
panic("cpu%d: ilockdepth %d, last lock %#p at %#p, sched called from %#p",
machp()->machno,
machp()->ilockdepth,
up? up->lastilock: nil,
(up && up->lastilock)? up->lastilock->_pc: 0,
getcallerpc());
kstackok();
if(up){
/*
* Delay the sched until the process gives up the locks
* it is holding. This avoids dumb lock loops.
* Don't delay if the process is Moribund.
* It called sched to die.
* But do sched eventually. This avoids a missing unlock
* from hanging the entire kernel.
* But don't reschedule procs holding palloc or procalloc.
* Those are far too important to be holding while asleep.
*
* This test is not exact. There can still be a few
* instructions in the middle of taslock when a process
* holds a lock but Lock.p has not yet been initialized.
*/
if(up->nlocks)
if(up->state != Moribund)
if(up->delaysched < 20
|| pga.l.p == up
|| procalloc.l.p == up){
up->delaysched++;
run.delayedscheds++;
return;
}
up->delaysched = 0;
splhi();
/* statistics */
if(up->nqtrap == 0 && up->nqsyscall == 0)
up->nfullq++;
machp()->cs++;
procsave(up);
mmuflushtlb(machp()->MMU.pml4->pa);
if(setlabel(&up->sched)){
procrestore(up);
spllo();
return;
}
/*debug*/gotolabel(&machp()->sched);
}
machp()->inidle = 1;
p = runproc(); /* core 0 never returns */
machp()->inidle = 0;
if(!p->edf){
updatecpu(p);
p->priority = reprioritize(p);
}
if(nosmp){
if(p != machp()->readied)
machp()->schedticks = machp()->ticks + HZ/10;
machp()->readied = 0;
}
machp()->externup = p;
up = p;
machp()->qstart = machp()->ticks;
up->nqtrap = 0;
up->nqsyscall = 0;
up->state = Running;
//up->mach = m;
up->mach = sys->machptr[machp()->machno];
machp()->proc = up;
// iprint("up->sched.sp %p * %p\n", up->sched.sp,
// *(void **) up->sched.sp);
mmuswitch(up);
assert(!up->wired || up->wired == machp());
if (0) hi("gotolabel\n");
/*debug*/gotolabel(&up->sched);
}
/*
* Zero copy I/O.
* I/O is performed using an array of Zio structures.
* Each one points to a shared buffer address indicating a length.
* Each entry indicating a length and using nil as the address
* is asking the system to allocate memory as needed (mread only).
*/
static int
ziorw(int fd, Zio *io, int nio, usize count, int64_t offset, int iswrite)
{
Proc *up = externup();
int i, n, isprw;
Kzio *kio, skio[16];
Chan *c;
usize tot;
if(nio <= 0 || nio > 512)
error("wrong io[] size");
zioinit();
kio = nil;
io = validaddr(io, sizeof io[0] * nio, 1);
DBG("ziorw %d io%#p[%d] %lu %lld\n", fd, io, nio, count, offset);
if(DBGFLG)
for(i = 0; i < nio; i++)
print("\tio%#p[%d] = %Z %s\n",
io, i, (Kzio*)&io[i], iswrite?"w":"r");
if(iswrite)
c = fdtochan(fd, OWRITE, 1, 1);
else
c = fdtochan(fd, OREAD, 1, 1);
isprw = offset != -1LL;
if(isprw)
offset = c->offset;
if(waserror()){
cclose(c);
if(kio != nil){
for(i = 0; i < nio; i++)
if(kio[i].seg != nil)
putseg(kio[i].seg);
if(kio != skio)
free(kio);
}
nexterror();
}
if(nio < nelem(skio))
kio = skio;
else
kio = smalloc(sizeof kio[0] * nio);
for(i = 0; i < nio; i++){
kio[i].Zio = io[i];
if(iswrite){
kio[i].seg = seg(up, PTR2UINT(io[i].data), 1);
if(kio[i].seg == nil)
error("invalid address in zio");
incref(&kio[i].seg->r);
qunlock(&kio[i].seg->lk);
validaddr(kio[i].Zio.data, kio[i].Zio.size, 1);
if((kio[i].seg->type&SG_ZIO) == 0){
/*
* It's not a segment where we can report
* addresses to anyone once they are free.
* So, allocate space in the kernel
* and copy the user data there.
*/
kernzio(&kio[i]);
}
assert(kio[i].seg->type&SG_ZIO);
}else{
kio[i].Zio.data = nil;
kio[i].seg = nil;
}
}
if(c->dev->zread == nil){
DBG("installing devzread for %s\n", c->dev->name);
c->dev->zread = devzread;
}
if(c->dev->zwrite == nil){
DBG("installing devzwrite for %s\n", c->dev->name);
c->dev->zwrite = devzwrite;
}
if(iswrite)
n = c->dev->zwrite(c, kio, nio, offset);
else
n = c->dev->zread(c, kio, nio, count, offset);
tot = 0;
for(i = 0; i < n; i++){
io[i] = kio[i].Zio;
tot += kio[i].Zio.size;
}
if(!isprw){
/* unlike in syswrite, we update offsets at the end */
lock(&c->r.l);
c->devoffset += tot;
c->offset += tot;
unlock(&c->r.l);
}
poperror();
cclose(c);
if(kio != skio)
free(kio);
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
vno = ureg->type;
uint64_t gsbase = rdmsr(GSbase);
//if (sce > scx) iprint("====================");
lastvno = vno;
if (gsbase < 1ULL<<63)
die("bogus gsbase");
Proc *up = externup();
char buf[ERRMAX];
Vctl *ctl, *v;
machp()->perf.intrts = perfticks();
user = userureg(ureg);
if(user && (machp()->NIX.nixtype == NIXTC)) {
up->dbgreg = ureg;
cycles(&up->kentry);
}
clockintr = 0;
//_pmcupdate(machp());
if(ctl = vctl[vno]) {
if(ctl->isintr) {
machp()->intr++;
if(vno >= VectorPIC && vno != VectorSYSCALL)
machp()->lastintr = ctl->Vkey.irq;
} else if(up)
up->nqtrap++;
if(ctl->isr) {
ctl->isr(vno);
if(islo())print("trap %d: isr %p enabled interrupts\n", vno, ctl->isr);
}
for(v = ctl; v != nil; v = v->next) {
if(v->f) {
v->f(ureg, v->a);
if(islo())print("trap %d: ctlf %p enabled interrupts\n", vno, v->f);
}
}
if(ctl->eoi) {
ctl->eoi(vno);
if(islo())print("trap %d: eoi %p enabled interrupts\n", vno, ctl->eoi);
}
intrtime(vno);
if(ctl->isintr) {
if(ctl->Vkey.irq == IrqCLOCK || ctl->Vkey.irq == IrqTIMER)
clockintr = 1;
if (ctl->Vkey.irq == IrqTIMER)
oprof_alarm_handler(ureg);
if(up && !clockintr)
preempted();
}
}
else if(vno < nelem(excname) && user) {
spllo();
snprint(buf, sizeof buf, "sys: trap: %s", excname[vno]);
postnote(up, 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",
machp()->machno, vno, machp()->lastintr);
intrtime(vno);
if(user)
kexit(ureg);
return;
}
else {
if(vno == VectorNMI) {
nmienable();
if(machp()->machno != 0) {
iprint("cpu%d: PC %#llx\n",
machp()->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(up && up->delaysched && clockintr) {
if(0)
if(user && up->ac == nil && up->nqtrap == 0 && up->nqsyscall == 0) {
if(!waserror()) {
up->ac = getac(up, -1);
poperror();
runacore();
return;
}
}
sched();
splhi();
}
if(user) {
if(up && up->procctl || up->nnote)
notify(ureg);
kexit(ureg);
}
}
int32_t
netifread(Netif *nif, Chan *c, void *a, int32_t n, int64_t off)
{
Proc *up = externup();
int i, j;
Netfile *f;
char *p;
int32_t offset;
if(c->qid.type & QTDIR)
return devdirread(c, a, n, (Dirtab*)nif, 0, netifgen);
offset = off;
switch(NETTYPE(c->qid.path)){
case Ndataqid:
f = nif->f[NETID(c->qid.path)];
return qread(f->iq, a, n);
case Nctlqid:
return readnum(offset, a, n, NETID(c->qid.path), NUMSIZE);
case Nstatqid:
p = malloc(READSTR);
if(p == nil)
error(Enomem);
j = snprint(p, READSTR, "in: %llud\n", nif->inpackets);
j += snprint(p+j, READSTR-j, "link: %d\n", nif->link);
j += snprint(p+j, READSTR-j, "out: %llud\n", nif->outpackets);
j += snprint(p+j, READSTR-j, "crc errs: %llud\n", nif->crcs);
j += snprint(p+j, READSTR-j, "overflows: %llud\n", nif->overflows);
j += snprint(p+j, READSTR-j, "soft overflows: %llud\n", nif->soverflows);
j += snprint(p+j, READSTR-j, "framing errs: %llud\n", nif->frames);
j += snprint(p+j, READSTR-j, "buffer errs: %llud\n", nif->buffs);
j += snprint(p+j, READSTR-j, "output errs: %llud\n", nif->oerrs);
j += snprint(p+j, READSTR-j, "prom: %d\n", nif->prom);
j += snprint(p+j, READSTR-j, "mbps: %d\n", nif->mbps);
j += snprint(p+j, READSTR-j, "addr: ");
for(i = 0; i < nif->alen; i++)
j += snprint(p+j, READSTR-j, "%2.2ux", nif->addr[i]);
snprint(p+j, READSTR-j, "\n");
n = readstr(offset, a, n, p);
free(p);
return n;
case Naddrqid:
p = malloc(READSTR);
j = 0;
for(i = 0; i < nif->alen; i++)
j += snprint(p+j, READSTR-j, "%2.2ux", nif->addr[i]);
n = readstr(offset, a, n, p);
free(p);
return n;
case Ntypeqid:
f = nif->f[NETID(c->qid.path)];
return readnum(offset, a, n, f->type, NUMSIZE);
case Nifstatqid:
return 0;
case Nmtuqid:
snprint(up->genbuf, sizeof up->genbuf, "%11.ud %11.ud %11.ud\n", nif->minmtu, nif->mtu, nif->maxmtu);
return readstr(offset, a, n, up->genbuf);
}
error(Ebadarg);
return -1; /* not reached */
}
void
asmmeminit(void)
{
Proc *up = externup();
int i, l;
Asm* assem;
PTE *pte, *pml4;
uintptr va;
uintmem hi, lo, mem, nextmem, pa;
#ifdef ConfCrap
int cx;
#endif /* ConfCrap */
assert(!((sys->vmunmapped|sys->vmend) & machp()->pgszmask[1]));
if((pa = mmuphysaddr(sys->vmunused)) == ~0)
panic("asmmeminit 1");
pa += sys->vmunmapped - sys->vmunused;
mem = asmalloc(pa, sys->vmend - sys->vmunmapped, 1, 0);
if(mem != pa)
panic("asmmeminit 2");
DBG("pa %#llux mem %#llux\n", pa, mem);
/* assume already 2MiB aligned*/
assert(ALIGNED(sys->vmunmapped, 2*MiB));
pml4 = UINT2PTR(machp()->pml4->va);
while(sys->vmunmapped < sys->vmend) {
l = mmuwalk(pml4, sys->vmunmapped, 1, &pte, asmwalkalloc);
DBG("%#p l %d\n", sys->vmunmapped, l);
*pte = pa|PtePS|PteRW|PteP;
sys->vmunmapped += 2*MiB;
pa += 2*MiB;
}
#ifdef ConfCrap
cx = 0;
#endif /* ConfCrap */
for(assem = asmlist; assem != nil; assem = assem->next) {
if(assem->type != AsmMEMORY)
continue;
va = KSEG2+assem->addr;
print("asm: addr %#P end %#P type %d size %P\n",
assem->addr, assem->addr+assem->size,
assem->type, assem->size);
lo = assem->addr;
hi = assem->addr+assem->size;
/* Convert a range into pages */
for(mem = lo; mem < hi; mem = nextmem) {
nextmem = (mem + PGLSZ(0)) & ~machp()->pgszmask[0];
/* Try large pages first */
for(i = m->npgsz - 1; i >= 0; i--) {
if((mem & machp()->pgszmask[i]) != 0)
continue;
if(mem + PGLSZ(i) > hi)
continue;
/* This page fits entirely within the range. */
/* Mark it a usable */
if((l = mmuwalk(pml4, va, i, &pte, asmwalkalloc)) < 0)
panic("asmmeminit 3");
*pte = mem|PteRW|PteP;
if(l > 0)
*pte |= PtePS;
nextmem = mem + PGLSZ(i);
va += PGLSZ(i);
npg[i]++;
break;
}
}
#ifdef ConfCrap
/*
* Fill in conf crap.
*/
if(cx >= nelem(conf.mem))
continue;
lo = ROUNDUP(assem->addr, PGSZ);
//if(lo >= 600ull*MiB)
// continue;
conf.mem[cx].base = lo;
hi = ROUNDDN(hi, PGSZ);
//if(hi > 600ull*MiB)
// hi = 600*MiB;
conf.mem[cx].npage = (hi - lo)/PGSZ;
conf.npage += conf.mem[cx].npage;
print("cm %d: addr %#llux npage %lud\n",
cx, conf.mem[cx].base, conf.mem[cx].npage);
cx++;
#endif /* ConfCrap */
}
print("%d %d %d\n", npg[0], npg[1], npg[2]);
#ifdef ConfCrap
/*
* Fill in more conf crap.
* This is why I hate Plan 9.
*/
conf.upages = conf.npage;
i = (sys->vmend - sys->vmstart)/PGSZ; /* close enough */
conf.ialloc = (i/2)*PGSZ;
print("npage %llud upage %lud kpage %d\n",
conf.npage, conf.upages, i);
#endif /* ConfCrap */
}
/*
* generate a 3 level directory
*/
static int
netifgen(Chan *c, char* j, Dirtab *vp, int n, int i, Dir *dp)
{
Proc *up = externup();
Qid q;
Netif *nif = (Netif*)vp;
Netfile *f;
int t;
int perm;
char *o;
q.type = QTFILE;
q.vers = 0;
/* top level directory contains the name of the network */
if(c->qid.path == 0){
switch(i){
case DEVDOTDOT:
q.path = 0;
q.type = QTDIR;
devdir(c, q, ".", 0, eve, 0555, dp);
break;
case 0:
q.path = N2ndqid;
q.type = QTDIR;
strcpy(up->genbuf, nif->name);
devdir(c, q, up->genbuf, 0, eve, 0555, dp);
break;
default:
return -1;
}
return 1;
}
/* second level contains clone plus all the conversations */
t = NETTYPE(c->qid.path);
if(t == N2ndqid || t == Ncloneqid || t == Naddrqid
|| t == Nstatqid || t == Nifstatqid || t == Nmtuqid){
switch(i) {
case DEVDOTDOT:
q.type = QTDIR;
q.path = 0;
devdir(c, q, ".", 0, eve, DMDIR|0555, dp);
break;
case 0:
q.path = Ncloneqid;
devdir(c, q, "clone", 0, eve, 0666, dp);
break;
case 1:
q.path = Naddrqid;
devdir(c, q, "addr", 0, eve, 0666, dp);
break;
case 2:
q.path = Nstatqid;
devdir(c, q, "stats", 0, eve, 0444, dp);
break;
case 3:
q.path = Nifstatqid;
devdir(c, q, "ifstats", 0, eve, 0444, dp);
break;
case 4:
q.path = Nmtuqid;
devdir(c, q, "mtu", 0, eve, 0444, dp);
break;
default:
i -= 5;
if(i >= nif->nfile)
return -1;
if(nif->f[i] == 0)
return 0;
q.type = QTDIR;
q.path = NETQID(i, N3rdqid);
snprint(up->genbuf, sizeof up->genbuf, "%d", i);
devdir(c, q, up->genbuf, 0, eve, DMDIR|0555, dp);
break;
}
return 1;
}
/* third level */
f = nif->f[NETID(c->qid.path)];
if(f == 0)
return 0;
if(*f->owner){
o = f->owner;
perm = f->mode;
} else {
o = eve;
perm = 0666;
}
switch(i){
case DEVDOTDOT:
q.type = QTDIR;
q.path = N2ndqid;
strcpy(up->genbuf, nif->name);
devdir(c, q, up->genbuf, 0, eve, DMDIR|0555, dp);
break;
case 0:
q.path = NETQID(NETID(c->qid.path), Ndataqid);
devdir(c, q, "data", 0, o, perm, dp);
break;
case 1:
q.path = NETQID(NETID(c->qid.path), Nctlqid);
devdir(c, q, "ctl", 0, o, perm, dp);
break;
case 2:
q.path = NETQID(NETID(c->qid.path), Nstatqid);
devdir(c, q, "stats", 0, eve, 0444, dp);
break;
case 3:
q.path = NETQID(NETID(c->qid.path), Ntypeqid);
devdir(c, q, "type", 0, eve, 0444, dp);
break;
case 4:
q.path = NETQID(NETID(c->qid.path), Nifstatqid);
devdir(c, q, "ifstats", 0, eve, 0444, dp);
break;
default:
return -1;
}
return 1;
}
void
rudpkick(void *x)
{
Proc *up = externup();
Conv *c = x;
Udphdr *uh;
uint16_t rport;
uint8_t laddr[IPaddrlen], raddr[IPaddrlen];
Block *bp;
Rudpcb *ucb;
Rudphdr *rh;
Reliable *r;
int dlen, ptcllen;
Rudppriv *upriv;
Fs *f;
upriv = c->p->priv;
f = c->p->f;
netlog(c->p->f, Logrudp, "rudp: kick\n");
bp = qget(c->wq);
if(bp == nil)
return;
ucb = (Rudpcb*)c->ptcl;
switch(ucb->headers) {
case 7:
/* get user specified addresses */
bp = pullupblock(bp, UDP_USEAD7);
if(bp == nil)
return;
ipmove(raddr, bp->rp);
bp->rp += IPaddrlen;
ipmove(laddr, bp->rp);
bp->rp += IPaddrlen;
/* pick interface closest to dest */
if(ipforme(f, laddr) != Runi)
findlocalip(f, laddr, raddr);
bp->rp += IPaddrlen; /* Ignore ifc address */
rport = nhgets(bp->rp);
bp->rp += 2+2; /* Ignore local port */
break;
default:
ipmove(raddr, c->raddr);
ipmove(laddr, c->laddr);
rport = c->rport;
break;
}
dlen = blocklen(bp);
/* Make space to fit rudp & ip header */
bp = padblock(bp, UDP_IPHDR+UDP_RHDRSIZE);
if(bp == nil)
return;
uh = (Udphdr *)(bp->rp);
uh->vihl = IP_VER4;
rh = (Rudphdr*)uh;
ptcllen = dlen + (UDP_RHDRSIZE-UDP_PHDRSIZE);
uh->Unused = 0;
uh->udpproto = IP_UDPPROTO;
uh->frag[0] = 0;
uh->frag[1] = 0;
hnputs(uh->udpplen, ptcllen);
switch(ucb->headers){
case 7:
v6tov4(uh->udpdst, raddr);
hnputs(uh->udpdport, rport);
v6tov4(uh->udpsrc, laddr);
break;
default:
v6tov4(uh->udpdst, c->raddr);
hnputs(uh->udpdport, c->rport);
if(ipcmp(c->laddr, IPnoaddr) == 0)
findlocalip(f, c->laddr, c->raddr);
v6tov4(uh->udpsrc, c->laddr);
break;
}
hnputs(uh->udpsport, c->lport);
hnputs(uh->udplen, ptcllen);
uh->udpcksum[0] = 0;
uh->udpcksum[1] = 0;
qlock(&ucb->ql);
r = relstate(ucb, raddr, rport, "kick");
r->sndseq = NEXTSEQ(r->sndseq);
hnputl(rh->relseq, r->sndseq);
hnputl(rh->relsgen, r->sndgen);
hnputl(rh->relack, r->rcvseq); /* ACK last rcvd packet */
hnputl(rh->relagen, r->rcvgen);
if(r->rcvseq != r->acksent)
r->acksent = r->rcvseq;
hnputs(uh->udpcksum, ptclcsum(bp, UDP_IPHDR, dlen+UDP_RHDRSIZE));
relackq(r, bp);
qunlock(&ucb->ql);
upriv->ustats.rudpOutDatagrams++;
DPRINT("sent: %lud/%lud, %lud/%lud\n",
r->sndseq, r->sndgen, r->rcvseq, r->rcvgen);
doipoput(c, f, bp, 0, c->ttl, c->tos);
if(waserror()) {
relput(r);
qunlock(&r->lock);
nexterror();
}
/* flow control of sorts */
qlock(&r->lock);
if(UNACKED(r) > Maxunacked){
r->blocked = 1;
sleep(&r->vous, flow, r);
r->blocked = 0;
}
qunlock(&r->lock);
relput(r);
poperror();
}
/*
* attach a device (or pkt driver) to the interface.
* called with c locked
*/
static char*
ipifcbind(Conv *c, char **argv, int argc)
{
Proc *up = externup();
Ipifc *ifc;
Medium *medium;
if(argc < 2)
return Ebadarg;
ifc = (Ipifc*)c->ptcl;
/* bind the device to the interface */
medium = ipfindmedium(argv[1]);
if(medium == nil)
return "unknown interface type";
wlock(ifc);
if(ifc->medium != nil){
wunlock(ifc);
return "interface already bound";
}
if(waserror()){
wunlock(ifc);
nexterror();
}
/* do medium specific binding */
(*medium->bind)(ifc, argc, argv);
/* set the bound device name */
if(argc > 2)
strncpy(ifc->dev, argv[2], sizeof(ifc->dev));
else
snprint(ifc->dev, sizeof ifc->dev, "%s%d", medium->name, c->x);
ifc->dev[sizeof(ifc->dev)-1] = 0;
/* set up parameters */
ifc->medium = medium;
ifc->mintu = ifc->medium->mintu;
ifc->maxtu = ifc->medium->maxtu;
if(ifc->medium->unbindonclose == 0)
ifc->conv->inuse++;
ifc->rp.mflag = 0; /* default not managed */
ifc->rp.oflag = 0;
ifc->rp.maxraint = 600000; /* millisecs */
ifc->rp.minraint = 200000;
ifc->rp.linkmtu = 0; /* no mtu sent */
ifc->rp.reachtime = 0;
ifc->rp.rxmitra = 0;
ifc->rp.ttl = MAXTTL;
ifc->rp.routerlt = 3 * ifc->rp.maxraint;
/* any ancillary structures (like routes) no int32_ter pertain */
ifc->ifcid++;
/* reopen all the queues closed by a previous unbind */
qreopen(c->rq);
qreopen(c->eq);
qreopen(c->sq);
wunlock(ifc);
poperror();
return nil;
}
Walkqid*
devwalk(Chan *c, Chan *nc, char **name, int nname, Dirtab *tab, int ntab,
Devgen *gen)
{
Proc *up = externup();
if (0) print_func_entry();
int i, j, alloc;
Walkqid *wq;
char *n;
Dir dir;
if(nname > 0)
isdir(c);
if (0)
{ int i; iprint("%d names:", nname); for(i = 0; i < nname; i++) iprint("%s ", name[i]); iprint("\n");}
alloc = 0;
wq = smalloc(sizeof(Walkqid)+(nname-1)*sizeof(Qid));
if(waserror()){
if(alloc && wq->clone!=nil)
cclose(wq->clone);
free(wq);
if (0) print_func_exit();
return nil;
}
if(nc == nil){
nc = devclone(c);
/*
* nc->dev remains nil for now. //XDYNX
*/
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){
/*
* Use c->dev->name in the error because
* nc->dev should be nil here.
*/
if((*gen)(nc, nil, tab, ntab, DEVDOTDOT, &dir) != 1){
print("devgen walk .. in dev%s %#llx broken\n",
c->dev->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 */
//what goes here: //XDYNX
// ->dev must be nil because can't walk an open chan, right?
// what about ref count on dev?
wq->clone->dev = c->dev;
//if(wq->clone->dev) //XDYNX
// devtabincr(wq->clone->dev);
}
if (0) print_func_exit();
return wq;
}
/* it should be unsigned. FIXME */
void
syscall(int badscallnr, Ureg *ureg)
{
// can only handle 4 args right now.
uintptr_t a0, a1, a2, a3;
uintptr_t a4, a5 = 0;
a0 = ureg->di;
a1 = ureg->si;
a2 = ureg->dx;
a3 = ureg->r10;
a4 = ureg->r8;
Proc *up = externup();
unsigned int scallnr = (unsigned int) badscallnr;
if (0) iprint("Syscall %d, %lx, %lx, %lx %lx %lx\n", scallnr, a0, a1, a2, a3, a4);
char *e;
uintptr_t sp;
int s;
int64_t startns, stopns;
Ar0 ar0;
static Ar0 zar0;
if(!userureg(ureg))
panic("syscall: cs %#llux\n", ureg->cs);
cycles(&up->kentry);
machp()->syscall++;
up->nsyscall++;
up->nqsyscall++;
up->insyscall = 1;
up->pc = ureg->ip;
up->dbgreg = ureg;
sp = ureg->sp;
startns = 0;
if (0) hi("so far syscall!\n");
if (printallsyscalls) {
syscallfmt(scallnr, a0, a1, a2, a3, a4, a5);
if(up->syscalltrace) {
print("E %s\n", up->syscalltrace);
free(up->syscalltrace);
up->syscalltrace = nil;
}
}
if(up->procctl == Proc_tracesyscall){
/*
* Redundant validaddr. Do we care?
* Tracing syscalls is not exactly a fast path...
* Beware, validaddr currently does a pexit rather
* than an error if there's a problem; that might
* change in the future.
*/
if(sp < (USTKTOP-BIGPGSZ) || sp > (USTKTOP-sizeof(up->arg)-BY2SE))
validaddr(UINT2PTR(sp), sizeof(up->arg)+BY2SE, 0);
syscallfmt(scallnr, a0, a1, a2, a3, a4, a5);
up->procctl = Proc_stopme;
procctl(up);
if(up->syscalltrace)
free(up->syscalltrace);
up->syscalltrace = nil;
startns = todget(nil);
}
if (0) hi("more syscall!\n");
up->scallnr = scallnr;
if(scallnr == RFORK)
fpusysrfork(ureg);
spllo();
sp = ureg->sp;
up->nerrlab = 0;
ar0 = zar0;
if(!waserror()){
if(scallnr >= nsyscall || systab[scallnr].f == nil){
pprint("bad sys call number %d pc %#llux\n",
scallnr, ureg->ip);
postnote(up, 1, "sys: bad sys call", NDebug);
error(Ebadarg);
}
if(sp < (USTKTOP-BIGPGSZ) || sp > (USTKTOP-sizeof(up->arg)-BY2SE))
validaddr(UINT2PTR(sp), sizeof(up->arg)+BY2SE, 0);
memmove(up->arg, UINT2PTR(sp+BY2SE), sizeof(up->arg));
up->psstate = systab[scallnr].n;
if (0) hi("call syscall!\n");
systab[scallnr].f(&ar0, a0, a1, a2, a3, a4, a5);
if (0) hi("it returned!\n");
if(scallnr == SYSR1){
/*
* BUG: must go when ron binaries go.
* NIX: Returning from execac().
* This means that the process is back to the
* time sharing core. However, the process did
* already return from the system call, when dispatching
* the user code to the AC. The only thing left is to
* return. The user registers should be ok, because
* up->dbgreg has been the user context for the process.
*/
return;
}
poperror();
}
else{
/* failure: save the error buffer for errstr */
e = up->syserrstr;
up->syserrstr = up->errstr;
up->errstr = e;
if(DBGFLG && up->pid == 1)
iprint("%s: syscall %s error %s\n",
up->text, systab[scallnr].n, up->syserrstr);
ar0 = systab[scallnr].r;
}
/*
* NIX: for the execac() syscall, what follows is done within
* the system call, because it never returns.
* See acore.c:/^retfromsyscall
*/
noerrorsleft();
/*
* Put return value in frame.
*/
ureg->ax = ar0.p;
if (printallsyscalls) {
stopns = todget(nil);
sysretfmt(scallnr, &ar0, startns, stopns, a0, a1, a2, a3, a4, a5);
if(up->syscalltrace) {
print("X %s\n", up->syscalltrace);
free(up->syscalltrace);
up->syscalltrace = nil;
}
}
if(up->procctl == Proc_tracesyscall){
stopns = todget(nil);
up->procctl = Proc_stopme;
sysretfmt(scallnr, &ar0, startns, stopns, a0, a1, a2, a3, a4, a5);
s = splhi();
procctl(up);
splx(s);
if(up->syscalltrace)
free(up->syscalltrace);
up->syscalltrace = nil;
}else if(up->procctl == Proc_totc || up->procctl == Proc_toac)
procctl(up);
if (0) hi("past sysretfmt\n");
up->insyscall = 0;
up->psstate = 0;
if(scallnr == NOTED)
noted(ureg, a0);
if (0) hi("now to splhi\n");
splhi();
if(scallnr != RFORK && (up->procctl || up->nnote))
notify(ureg);
/* if we delayed sched because we held a lock, sched now */
if(up->delaysched){
sched();
splhi();
}
kexit(ureg);
if (0) hi("done kexit\n");
}
static int32_t
vgaread(Chan* c, void* a, int32_t n, int64_t off)
{
Proc *up = externup();
int len;
char *p, *s;
VGAscr *scr;
uint32_t offset = off;
char chbuf[30];
switch((uint32_t)c->qid.path){
case Qdir:
return devdirread(c, a, n, vgadir, nelem(vgadir), devgen);
case Qvgabios:
if(offset >= 0x100000)
return 0;
if(offset+n >= 0x100000)
n = 0x100000 - offset;
memmove(a, (unsigned char*)KADDR(0)+offset, n);
return n;
case Qvgactl:
scr = &vgascreen[0];
p = malloc(READSTR);
if(p == nil)
error(Enomem);
if(waserror()){
free(p);
nexterror();
}
len = 0;
if(scr->dev)
s = scr->dev->name;
else
s = "cga";
len += snprint(p+len, READSTR-len, "type %s\n", s);
if(scr->gscreen) {
len += snprint(p+len, READSTR-len, "size %dx%dx%d %s\n",
scr->gscreen->r.max.x, scr->gscreen->r.max.y,
scr->gscreen->depth, chantostr(chbuf, scr->gscreen->chan));
if(Dx(scr->gscreen->r) != Dx(physgscreenr)
|| Dy(scr->gscreen->r) != Dy(physgscreenr))
len += snprint(p+len, READSTR-len, "actualsize %dx%d\n",
physgscreenr.max.x, physgscreenr.max.y);
}
len += snprint(p+len, READSTR-len, "blank time %lu idle %d state %s\n",
blanktime, drawidletime(), scr->isblank ? "off" : "on");
len += snprint(p+len, READSTR-len, "hwaccel %s\n", hwaccel ? "on" : "off");
len += snprint(p+len, READSTR-len, "hwblank %s\n", hwblank ? "on" : "off");
len += snprint(p+len, READSTR-len, "panning %s\n", panning ? "on" : "off");
len += snprint(p+len, READSTR-len, "addr p 0x%lx v 0x%p size 0x%x\n", scr->paddr, scr->vaddr, scr->apsize);
USED(len);
n = readstr(offset, a, n, p);
poperror();
free(p);
return n;
case Qvgaovl:
case Qvgaovlctl:
error(Ebadusefd);
break;
default:
error(Egreg);
break;
}
return 0;
}
int
screensize(int x, int y, int z, uint32_t chan)
{
Proc *up = externup();
VGAscr *scr;
void *oldsoft;
lock(&vgascreenlock);
if(waserror()){
unlock(&vgascreenlock);
nexterror();
}
memimageinit();
scr = &vgascreen[0];
oldsoft = softscreen;
if(scr->paddr == 0){
int width = (x*z)/BI2WD;
void *p;
p = malloc(width*BY2WD*y);
if(p == nil)
error("no memory for vga soft screen");
gscreendata.bdata = softscreen = p;
if(scr->dev && scr->dev->page){
scr->vaddr = KADDR(VGAMEM());
scr->apsize = 1<<16;
}
scr->useflush = 1;
}
else{
gscreendata.bdata = scr->vaddr;
scr->useflush = scr->dev && scr->dev->flush;
}
scr->gscreen = nil;
if(gscreen)
freememimage(gscreen);
gscreen = allocmemimaged(Rect(0,0,x,y), chan, &gscreendata);
if(gscreen == nil)
error("no memory for vga memimage");
vgaimageinit(chan);
scr->palettedepth = 6; /* default */
scr->gscreendata = &gscreendata;
scr->memdefont = getmemdefont();
scr->gscreen = gscreen;
physgscreenr = gscreen->r;
unlock(&vgascreenlock);
poperror();
if(oldsoft)
free(oldsoft);
memimagedraw(gscreen, gscreen->r, memblack, ZP, nil, ZP, S);
flushmemscreen(gscreen->r);
if(didswcursorinit)
swcursorinit();
drawcmap();
return 0;
}
static int32_t
consread(Chan *c, void *buf, int32_t n, int64_t off)
{
Proc *up = externup();
uint32_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->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 %uld\n", lockstats.locks);
s = seprint(s, tmp + sizeof tmp, "glare %uld\n", lockstats.glare);
s = seprint(s, tmp + sizeof tmp, "inglare %uld\n", lockstats.inglare);
s = seprint(s, tmp + sizeof tmp, "qlock %uld\n", qlockstats.qlock);
seprint(s, tmp + sizeof tmp, "qlockq %uld\n", qlockstats.qlockq);
return readstr(offset, buf, n, tmp);
break;
default:
print("consread %#llux\n", c->qid.path);
error(Egreg);
}
return -1; /* never reached */
}
char *
edfadmit(Proc *p)
{
Proc *up = externup();
char *err;
Edf *e;
int i;
Proc *r;
int32_t tns;
e = p->edf;
if (e->flags & Admitted)
return "task state"; /* should never happen */
/* simple sanity checks */
if (e->T == 0)
return "T not set";
if (e->C == 0)
return "C not set";
if (e->D > e->T)
return "D > T";
if (e->D == 0) /* if D is not set, set it to T */
e->D = e->T;
if (e->C > e->D)
return "C > D";
qlock(&edfschedlock);
if (err = testschedulability(p)){
qunlock(&edfschedlock);
return err;
}
e->flags |= Admitted;
edflock(p);
if(p->trace)
proctrace(p, SAdmit, 0);
/* Look for another proc with the same period to synchronize to */
for(i=0; (r = psincref(i)) != nil; i++) {
if(r->state == Dead || r == p){
psdecref(r);
continue;
}
if (r->edf == nil || (r->edf->flags & Admitted) == 0){
psdecref(r);
continue;
}
if (r->edf->T == e->T)
break;
}
if (r == nil){
/* Can't synchronize to another proc, release now */
e->t = now;
e->d = 0;
release(p);
if (p == up){
DPRINT("%lu edfadmit self %d[%s], release now: r=%lu d=%lu t=%lu\n",
now, p->pid, statename[p->state], e->r, e->d, e->t);
/* We're already running */
edfrun(p, 1);
}else{
/* We're releasing another proc */
DPRINT("%lu edfadmit other %d[%s], release now: r=%lu d=%lu t=%lu\n",
now, p->pid, statename[p->state], e->r, e->d, e->t);
p->Timer.ta = p;
edfunlock();
qunlock(&edfschedlock);
releaseintr(nil, &p->Timer);
return nil;
}
}else{
/* Release in synch to something else */
e->t = r->edf->t;
psdecref(r);
if (p == up){
DPRINT("%lu edfadmit self %d[%s], release at %lu\n",
now, p->pid, statename[p->state], e->t);
}else{
DPRINT("%lu edfadmit other %d[%s], release at %lu\n",
now, p->pid, statename[p->state], e->t);
if(e->Timer.tt == nil){
e->Timer.tf = releaseintr;
e->Timer.ta = p;
tns = e->t - now;
if(tns < 20)
tns = 20;
e->Timer.tns = 1000LL * tns;
e->Timer.tmode = Trelative;
timeradd(&e->Timer);
}
}
}
edfunlock();
qunlock(&edfschedlock);
return nil;
}
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:
*(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;
default:
print("conswrite: %#llux\n", c->qid.path);
error(Egreg);
}
return n;
}
/* 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 int
tfn(void *arg)
{
Proc *up = externup();
return up->trend == nil || up->tfn(arg);
}