void
dvma_init()
{
vm_offset_t segmap_addr;
/*
* Create phys_map covering the entire DVMA space,
* then allocate the segment pool from that. The
* remainder will be used as the DVMA page pool.
*/
phys_map = uvm_map_create(pmap_kernel(),
DVMA_MAP_BASE, DVMA_MAP_END, 1);
if (phys_map == NULL)
panic("unable to create DVMA map");
/*
* Reserve the DVMA space used for segment remapping.
* The remainder of phys_map is used for DVMA scratch
* memory pages (i.e. driver control blocks, etc.)
*/
segmap_addr = uvm_km_valloc_wait(phys_map, dvma_segmap_size);
if (segmap_addr != DVMA_MAP_BASE)
panic("dvma_init: unable to allocate DVMA segments");
/*
* Create the VM pool used for mapping whole segments
* into DVMA space for the purpose of data transfer.
*/
rminit(dvma_segmap, dvma_segmap_size, segmap_addr,
"dvma_segmap", NUM_DVMA_SEGS);
}
/*
* Map a user I/O request into kernel virtual address space.
* Note: the pages are already locked by uvm_vslock(), so we
* do not need to pass an access_type to pmap_enter().
*/
void
vmapbuf(struct buf *bp, vsize_t len)
{
vaddr_t faddr, taddr, off;
paddr_t fpa;
if ((bp->b_flags & B_PHYS) == 0)
panic("vmapbuf");
faddr = trunc_page((vaddr_t)(bp->b_saveaddr = bp->b_data));
off = (vaddr_t)bp->b_data - faddr;
len = round_page(off + len);
taddr= uvm_km_valloc_wait(phys_map, len);
bp->b_data = (caddr_t)(taddr + off);
/*
* The region is locked, so we expect that pmap_pte() will return
* non-NULL.
* XXX: unwise to expect this in a multithreaded environment.
* anything can happen to a pmap between the time we lock a
* region, release the pmap lock, and then relock it for
* the pmap_extract().
*
* no need to flush TLB since we expect nothing to be mapped
* where we we just allocated (TLB will be flushed when our
* mapping is removed).
*/
while (len) {
(void) pmap_extract(vm_map_pmap(&bp->b_proc->p_vmspace->vm_map),
faddr, &fpa);
pmap_kenter_pa(taddr, fpa, PROT_READ | PROT_WRITE);
faddr += PAGE_SIZE;
taddr += PAGE_SIZE;
len -= PAGE_SIZE;
}
}
/*
* Map an IO request into kernel virtual address space.
*/
void
vmapbuf(struct buf *bp, vsize_t len)
{
vaddr_t faddr, taddr, off;
paddr_t pa;
#ifdef DIAGNOSTIC
if (!(bp->b_flags & B_PHYS))
panic("vmapbuf");
#endif
faddr = trunc_page((vaddr_t)(bp->b_saveaddr = bp->b_data));
off = (vaddr_t)bp->b_data - faddr;
len = round_page(off + len);
taddr = uvm_km_valloc_wait(phys_map, len);
bp->b_data = (caddr_t)(taddr + off);
for (; len > 0; len -= NBPG) {
pmap_extract(vm_map_pmap(&bp->b_proc->p_vmspace->vm_map),
faddr, &pa);
pmap_enter(vm_map_pmap(phys_map), taddr, pa,
VM_PROT_READ | VM_PROT_WRITE, PMAP_WIRED);
faddr += NBPG;
taddr += NBPG;
}
pmap_update(vm_map_pmap(phys_map));
}
/*
* Machine-dependent startup code
*/
void
cpu_startup()
{
caddr_t v;
int sz;
#ifdef DEBUG
extern int pmapdebug;
int opmapdebug = pmapdebug;
#endif
vaddr_t minaddr, maxaddr;
extern struct user *proc0paddr;
#ifdef DEBUG
pmapdebug = 0;
#endif
if (CPU_ISSUN4M) {
extern int stackgap_random;
stackgap_random = STACKGAP_RANDOM_SUN4M;
}
/*
* fix message buffer mapping, note phys addr of msgbuf is 0
*/
pmap_map(MSGBUF_VA, 0, MSGBUFSIZE, VM_PROT_READ|VM_PROT_WRITE);
initmsgbuf((caddr_t)(MSGBUF_VA + (CPU_ISSUN4 ? 4096 : 0)), MSGBUFSIZE);
proc0.p_addr = proc0paddr;
/*
* Good {morning,afternoon,evening,night}.
*/
printf(version);
/*identifycpu();*/
printf("real mem = %u (%uMB)\n", ptoa(physmem),
ptoa(physmem)/1024/1024);
/*
* Find out how much space we need, allocate it,
* and then give everything true virtual addresses.
*/
sz = (int)allocsys((caddr_t)0);
if ((v = (caddr_t)uvm_km_alloc(kernel_map, round_page(sz))) == 0)
panic("startup: no room for tables");
if (allocsys(v) - v != sz)
panic("startup: table size inconsistency");
/*
* Determine how many buffers to allocate.
* We allocate bufcachepercent% of memory for buffer space.
*/
if (bufpages == 0)
bufpages = physmem * bufcachepercent / 100;
/* Restrict to at most 25% filled kvm */
if (bufpages >
(VM_MAX_KERNEL_ADDRESS-VM_MIN_KERNEL_ADDRESS) / PAGE_SIZE / 4)
bufpages = (VM_MAX_KERNEL_ADDRESS-VM_MIN_KERNEL_ADDRESS) /
PAGE_SIZE / 4;
/*
* Allocate a submap for exec arguments. This map effectively
* limits the number of processes exec'ing at any time.
*/
minaddr = vm_map_min(kernel_map);
exec_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
16*NCARGS, VM_MAP_PAGEABLE, FALSE, NULL);
/*
* Allocate a map for physio. Others use a submap of the kernel
* map, but we want one completely separate, even though it uses
* the same pmap.
*/
dvma_base = CPU_ISSUN4M ? DVMA4M_BASE : DVMA_BASE;
dvma_end = CPU_ISSUN4M ? DVMA4M_END : DVMA_END;
#if defined(SUN4M)
if (CPU_ISSUN4M) {
/*
* The DVMA space we want partially overrides kernel_map.
* Allocate it in kernel_map as well to prevent it from being
* used for other things.
*/
if (uvm_map(kernel_map, &dvma_base,
vm_map_max(kernel_map) - dvma_base,
NULL, UVM_UNKNOWN_OFFSET, 0,
UVM_MAPFLAG(UVM_PROT_NONE, UVM_PROT_NONE, UVM_INH_NONE,
UVM_ADV_NORMAL, 0)))
panic("startup: can not steal dvma map");
}
#endif
phys_map = uvm_map_create(pmap_kernel(), dvma_base, dvma_end,
VM_MAP_INTRSAFE);
if (phys_map == NULL)
panic("unable to create DVMA map");
/*
* Allocate DVMA space and dump into a privately managed
* resource map for double mappings which is usable from
* interrupt contexts.
*/
if (uvm_km_valloc_wait(phys_map, (dvma_end-dvma_base)) != dvma_base)
panic("unable to allocate from DVMA map");
dvmamap_extent = extent_create("dvmamap", dvma_base, dvma_end,
M_DEVBUF, NULL, 0, EX_NOWAIT);
if (dvmamap_extent == 0)
panic("unable to allocate extent for dvma");
#ifdef DEBUG
pmapdebug = opmapdebug;
#endif
printf("avail mem = %lu (%luMB)\n", ptoa(uvmexp.free),
ptoa(uvmexp.free)/1024/1024);
/*
* Set up buffers, so they can be used to read disk labels.
*/
bufinit();
/* Early interrupt handlers initialization */
intr_init();
}
/*
* cpu_startup: allocate memory for variable-sized tables,
* initialize cpu, and do autoconfiguration.
*
* This is called early in init_main.c:main(), after the
* kernel memory allocator is ready for use, but before
* the creation of processes 1,2, and mountroot, etc.
*/
void
cpu_startup()
{
caddr_t v;
int sz, i;
vsize_t size;
int base, residual;
vaddr_t minaddr, maxaddr;
char pbuf[9];
/*
* Initialize message buffer (for kernel printf).
* This is put in physical pages four through seven
* so it will always be in the same place after a
* reboot. (physical pages 0-3 are reserved by the PROM
* for its vector table and other stuff.)
* Its mapping was prepared in pmap_bootstrap().
* Also, offset some to avoid PROM scribbles.
*/
v = (caddr_t) (NBPG * 4);
msgbufaddr = (caddr_t)(v + MSGBUFOFF);
initmsgbuf(msgbufaddr, MSGBUFSIZE);
#ifdef DDB
{
extern int end[];
extern char *esym;
ddb_init(end[0], end + 1, (int*)esym);
}
#endif /* DDB */
/*
* Good {morning,afternoon,evening,night}.
*/
printf(version);
identifycpu();
fputype = FPU_NONE;
#ifdef FPU_EMULATE
printf("fpu: emulator\n");
#else
printf("fpu: no math support\n");
#endif
format_bytes(pbuf, sizeof(pbuf), ctob(physmem));
printf("total memory = %s\n", pbuf);
/*
* XXX fredette - we force a small number of buffers
* to help me debug this on my low-memory machine.
* this should go away at some point, allowing the
* normal automatic buffer-sizing to happen.
*/
bufpages = 37;
/*
* Get scratch page for dumpsys().
*/
if ((dumppage = uvm_km_alloc(kernel_map, NBPG)) == 0)
panic("startup: alloc dumppage");
/*
* Find out how much space we need, allocate it,
* and then give everything true virtual addresses.
*/
sz = (int)allocsys(NULL, NULL);
if ((v = (caddr_t)uvm_km_alloc(kernel_map, round_page(sz))) == 0)
panic("startup: no room for tables");
if (allocsys(v, NULL) - v != sz)
panic("startup: table size inconsistency");
/*
* Now allocate buffers proper. They are different than the above
* in that they usually occupy more virtual memory than physical.
*/
size = MAXBSIZE * nbuf;
if (uvm_map(kernel_map, (vaddr_t *) &buffers, round_page(size),
NULL, UVM_UNKNOWN_OFFSET, 0,
UVM_MAPFLAG(UVM_PROT_NONE, UVM_PROT_NONE, UVM_INH_NONE,
UVM_ADV_NORMAL, 0)) != 0)
panic("startup: cannot allocate VM for buffers");
minaddr = (vaddr_t)buffers;
if ((bufpages / nbuf) >= btoc(MAXBSIZE)) {
/* don't want to alloc more physical mem than needed */
bufpages = btoc(MAXBSIZE) * nbuf;
}
base = bufpages / nbuf;
residual = bufpages % nbuf;
for (i = 0; i < nbuf; i++) {
vsize_t curbufsize;
vaddr_t curbuf;
struct vm_page *pg;
/*
* Each buffer has MAXBSIZE bytes of VM space allocated. Of
* that MAXBSIZE space, we allocate and map (base+1) pages
* for the first "residual" buffers, and then we allocate
* "base" pages for the rest.
*/
curbuf = (vaddr_t) buffers + (i * MAXBSIZE);
curbufsize = NBPG * ((i < residual) ? (base+1) : base);
while (curbufsize) {
pg = uvm_pagealloc(NULL, 0, NULL, 0);
if (pg == NULL)
panic("cpu_startup: not enough memory for "
"buffer cache");
pmap_kenter_pa(curbuf, VM_PAGE_TO_PHYS(pg),
VM_PROT_READ|VM_PROT_WRITE);
curbuf += PAGE_SIZE;
curbufsize -= PAGE_SIZE;
}
}
pmap_update(pmap_kernel());
/*
* Allocate a submap for exec arguments. This map effectively
* limits the number of processes exec'ing at any time.
*/
exec_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
NCARGS, VM_MAP_PAGEABLE, FALSE, NULL);
/*
* We don't use a submap for physio, and use a separate map
* for DVMA allocations. Our vmapbuf just maps pages into
* the kernel map (any kernel mapping is OK) and then the
* device drivers clone the kernel mappings into DVMA space.
*/
/*
* Finally, allocate mbuf cluster submap.
*/
mb_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
nmbclusters * mclbytes, VM_MAP_INTRSAFE,
FALSE, NULL);
format_bytes(pbuf, sizeof(pbuf), ptoa(uvmexp.free));
printf("avail memory = %s\n", pbuf);
format_bytes(pbuf, sizeof(pbuf), bufpages * NBPG);
printf("using %d buffers containing %s of memory\n", nbuf, pbuf);
/*
* Allocate a virtual page (for use by /dev/mem)
* This page is handed to pmap_enter() therefore
* it has to be in the normal kernel VA range.
*/
vmmap = uvm_km_valloc_wait(kernel_map, NBPG);
/*
* Allocate dma map for devices on the bus.
*/
dvmamap = extent_create("dvmamap",
DVMA_MAP_BASE, DVMA_MAP_BASE + DVMA_MAP_AVAIL,
M_DEVBUF, 0, 0, EX_NOWAIT);
if (dvmamap == NULL)
panic("unable to allocate DVMA map");
/*
* Set up CPU-specific registers, cache, etc.
*/
initcpu();
/*
* Set up buffers, so they can be used to read disk labels.
*/
bufinit();
}
/* ARGSUSED */
int
sys_execve(struct proc *p, void *v, register_t *retval)
{
struct sys_execve_args /* {
syscallarg(const char *) path;
syscallarg(char *const *) argp;
syscallarg(char *const *) envp;
} */ *uap = v;
int error;
struct exec_package pack;
struct nameidata nid;
struct vattr attr;
struct ucred *cred = p->p_ucred;
char *argp;
char * const *cpp, *dp, *sp;
struct process *pr = p->p_p;
long argc, envc;
size_t len, sgap;
#ifdef MACHINE_STACK_GROWS_UP
size_t slen;
#endif
char *stack;
struct ps_strings arginfo;
struct vmspace *vm = p->p_vmspace;
char **tmpfap;
extern struct emul emul_native;
#if NSYSTRACE > 0
int wassugid = ISSET(pr->ps_flags, PS_SUGID | PS_SUGIDEXEC);
size_t pathbuflen;
#endif
char *pathbuf = NULL;
/*
* Cheap solution to complicated problems.
* Mark this process as "leave me alone, I'm execing".
*/
atomic_setbits_int(&p->p_flag, P_INEXEC);
#if NSYSTRACE > 0
if (ISSET(p->p_flag, P_SYSTRACE)) {
systrace_execve0(p);
pathbuf = pool_get(&namei_pool, PR_WAITOK);
error = copyinstr(SCARG(uap, path), pathbuf, MAXPATHLEN,
&pathbuflen);
if (error != 0)
goto clrflag;
}
#endif
if (pathbuf != NULL) {
NDINIT(&nid, LOOKUP, NOFOLLOW, UIO_SYSSPACE, pathbuf, p);
} else {
NDINIT(&nid, LOOKUP, NOFOLLOW, UIO_USERSPACE,
SCARG(uap, path), p);
}
/*
* initialize the fields of the exec package.
*/
if (pathbuf != NULL)
pack.ep_name = pathbuf;
else
pack.ep_name = (char *)SCARG(uap, path);
pack.ep_hdr = malloc(exec_maxhdrsz, M_EXEC, M_WAITOK);
pack.ep_hdrlen = exec_maxhdrsz;
pack.ep_hdrvalid = 0;
pack.ep_ndp = &nid;
pack.ep_interp = NULL;
pack.ep_emul_arg = NULL;
VMCMDSET_INIT(&pack.ep_vmcmds);
pack.ep_vap = &attr;
pack.ep_emul = &emul_native;
pack.ep_flags = 0;
/* see if we can run it. */
if ((error = check_exec(p, &pack)) != 0) {
goto freehdr;
}
/* XXX -- THE FOLLOWING SECTION NEEDS MAJOR CLEANUP */
/* allocate an argument buffer */
argp = (char *) uvm_km_valloc_wait(exec_map, NCARGS);
#ifdef DIAGNOSTIC
if (argp == NULL)
panic("execve: argp == NULL");
#endif
dp = argp;
argc = 0;
/* copy the fake args list, if there's one, freeing it as we go */
if (pack.ep_flags & EXEC_HASARGL) {
tmpfap = pack.ep_fa;
while (*tmpfap != NULL) {
char *cp;
cp = *tmpfap;
while (*cp)
*dp++ = *cp++;
*dp++ = '\0';
free(*tmpfap, M_EXEC);
tmpfap++; argc++;
}
free(pack.ep_fa, M_EXEC);
pack.ep_flags &= ~EXEC_HASARGL;
}
/* Now get argv & environment */
if (!(cpp = SCARG(uap, argp))) {
error = EFAULT;
goto bad;
}
if (pack.ep_flags & EXEC_SKIPARG)
cpp++;
while (1) {
len = argp + ARG_MAX - dp;
if ((error = copyin(cpp, &sp, sizeof(sp))) != 0)
goto bad;
if (!sp)
break;
if ((error = copyinstr(sp, dp, len, &len)) != 0) {
if (error == ENAMETOOLONG)
error = E2BIG;
goto bad;
}
dp += len;
cpp++;
argc++;
}
envc = 0;
/* environment does not need to be there */
if ((cpp = SCARG(uap, envp)) != NULL ) {
while (1) {
len = argp + ARG_MAX - dp;
if ((error = copyin(cpp, &sp, sizeof(sp))) != 0)
goto bad;
if (!sp)
break;
if ((error = copyinstr(sp, dp, len, &len)) != 0) {
if (error == ENAMETOOLONG)
error = E2BIG;
goto bad;
}
dp += len;
cpp++;
envc++;
}
}
dp = (char *)ALIGN(dp);
sgap = STACKGAPLEN;
if (stackgap_random != 0)
sgap += (arc4random() * ALIGNBYTES) & (stackgap_random - 1);
#ifdef MACHINE_STACK_GROWS_UP
sgap = ALIGN(sgap);
#endif
/* Now check if args & environ fit into new stack */
len = ((argc + envc + 2 + pack.ep_emul->e_arglen) * sizeof(char *) +
sizeof(long) + dp + sgap + sizeof(struct ps_strings)) - argp;
len = ALIGN(len); /* make the stack "safely" aligned */
if (len > pack.ep_ssize) { /* in effect, compare to initial limit */
error = ENOMEM;
goto bad;
}
/* adjust "active stack depth" for process VSZ */
pack.ep_ssize = len; /* maybe should go elsewhere, but... */
/*
* Prepare vmspace for remapping. Note that uvmspace_exec can replace
* p_vmspace!
*/
uvmspace_exec(p, VM_MIN_ADDRESS, VM_MAXUSER_ADDRESS);
vm = p->p_vmspace;
/* Now map address space */
vm->vm_taddr = (char *)pack.ep_taddr;
vm->vm_tsize = atop(round_page(pack.ep_tsize));
vm->vm_daddr = (char *)pack.ep_daddr;
vm->vm_dsize = atop(round_page(pack.ep_dsize));
vm->vm_dused = 0;
vm->vm_ssize = atop(round_page(pack.ep_ssize));
vm->vm_maxsaddr = (char *)pack.ep_maxsaddr;
vm->vm_minsaddr = (char *)pack.ep_minsaddr;
/* create the new process's VM space by running the vmcmds */
#ifdef DIAGNOSTIC
if (pack.ep_vmcmds.evs_used == 0)
panic("execve: no vmcmds");
#endif
error = exec_process_vmcmds(p, &pack);
/* if an error happened, deallocate and punt */
if (error)
goto exec_abort;
/* remember information about the process */
arginfo.ps_nargvstr = argc;
arginfo.ps_nenvstr = envc;
#ifdef MACHINE_STACK_GROWS_UP
stack = (char *)USRSTACK + sizeof(arginfo) + sgap;
slen = len - sizeof(arginfo) - sgap;
#else
stack = (char *)(USRSTACK - len);
#endif
/* Now copy argc, args & environ to new stack */
if (!(*pack.ep_emul->e_copyargs)(&pack, &arginfo, stack, argp))
goto exec_abort;
/* copy out the process's ps_strings structure */
if (copyout(&arginfo, (char *)PS_STRINGS, sizeof(arginfo)))
goto exec_abort;
stopprofclock(p); /* stop profiling */
fdcloseexec(p); /* handle close on exec */
execsigs(p); /* reset caught signals */
TCB_SET(p, NULL); /* reset the TCB address */
/* set command name & other accounting info */
len = min(nid.ni_cnd.cn_namelen, MAXCOMLEN);
bcopy(nid.ni_cnd.cn_nameptr, p->p_comm, len);
p->p_comm[len] = 0;
p->p_acflag &= ~AFORK;
/* record proc's vnode, for use by procfs and others */
if (p->p_textvp)
vrele(p->p_textvp);
vref(pack.ep_vp);
p->p_textvp = pack.ep_vp;
atomic_setbits_int(&pr->ps_flags, PS_EXEC);
if (pr->ps_flags & PS_PPWAIT) {
atomic_clearbits_int(&pr->ps_flags, PS_PPWAIT);
atomic_clearbits_int(&pr->ps_pptr->ps_flags, PS_ISPWAIT);
wakeup(pr->ps_pptr);
}
/*
* If process does execve() while it has a mismatched real,
* effective, or saved uid/gid, we set PS_SUGIDEXEC.
*/
if (p->p_ucred->cr_uid != p->p_cred->p_ruid ||
p->p_ucred->cr_uid != p->p_cred->p_svuid ||
p->p_ucred->cr_gid != p->p_cred->p_rgid ||
p->p_ucred->cr_gid != p->p_cred->p_svgid)
atomic_setbits_int(&pr->ps_flags, PS_SUGIDEXEC);
else
atomic_clearbits_int(&pr->ps_flags, PS_SUGIDEXEC);
/*
* deal with set[ug]id.
* MNT_NOEXEC has already been used to disable s[ug]id.
*/
if ((attr.va_mode & (VSUID | VSGID)) && proc_cansugid(p)) {
int i;
atomic_setbits_int(&pr->ps_flags, PS_SUGID|PS_SUGIDEXEC);
#ifdef KTRACE
/*
* If process is being ktraced, turn off - unless
* root set it.
*/
if (p->p_tracep && !(p->p_traceflag & KTRFAC_ROOT)) {
p->p_traceflag = 0;
ktrsettracevnode(p, NULL);
}
#endif
p->p_ucred = crcopy(cred);
if (attr.va_mode & VSUID)
p->p_ucred->cr_uid = attr.va_uid;
if (attr.va_mode & VSGID)
p->p_ucred->cr_gid = attr.va_gid;
/*
* For set[ug]id processes, a few caveats apply to
* stdin, stdout, and stderr.
*/
for (i = 0; i < 3; i++) {
struct file *fp = NULL;
/*
* NOTE - This will never return NULL because of
* immature fds. The file descriptor table is not
* shared because we're suid.
*/
fp = fd_getfile(p->p_fd, i);
#ifdef PROCFS
/*
* Close descriptors that are writing to procfs.
*/
if (fp && fp->f_type == DTYPE_VNODE &&
((struct vnode *)(fp->f_data))->v_tag == VT_PROCFS &&
(fp->f_flag & FWRITE)) {
fdrelease(p, i);
fp = NULL;
}
#endif
/*
* Ensure that stdin, stdout, and stderr are already
* allocated. We do not want userland to accidentally
* allocate descriptors in this range which has implied
* meaning to libc.
*/
if (fp == NULL) {
short flags = FREAD | (i == 0 ? 0 : FWRITE);
struct vnode *vp;
int indx;
if ((error = falloc(p, &fp, &indx)) != 0)
goto exec_abort;
#ifdef DIAGNOSTIC
if (indx != i)
panic("sys_execve: falloc indx != i");
#endif
if ((error = cdevvp(getnulldev(), &vp)) != 0) {
fdremove(p->p_fd, indx);
closef(fp, p);
goto exec_abort;
}
if ((error = VOP_OPEN(vp, flags, p->p_ucred, p)) != 0) {
fdremove(p->p_fd, indx);
closef(fp, p);
vrele(vp);
goto exec_abort;
}
if (flags & FWRITE)
vp->v_writecount++;
fp->f_flag = flags;
fp->f_type = DTYPE_VNODE;
fp->f_ops = &vnops;
fp->f_data = (caddr_t)vp;
FILE_SET_MATURE(fp);
}
}
} else
atomic_clearbits_int(&pr->ps_flags, PS_SUGID);
p->p_cred->p_svuid = p->p_ucred->cr_uid;
p->p_cred->p_svgid = p->p_ucred->cr_gid;
if (pr->ps_flags & PS_SUGIDEXEC) {
int i, s = splclock();
timeout_del(&p->p_realit_to);
timerclear(&p->p_realtimer.it_interval);
timerclear(&p->p_realtimer.it_value);
for (i = 0; i < sizeof(p->p_stats->p_timer) /
sizeof(p->p_stats->p_timer[0]); i++) {
timerclear(&p->p_stats->p_timer[i].it_interval);
timerclear(&p->p_stats->p_timer[i].it_value);
}
splx(s);
}
uvm_km_free_wakeup(exec_map, (vaddr_t) argp, NCARGS);
pool_put(&namei_pool, nid.ni_cnd.cn_pnbuf);
vn_close(pack.ep_vp, FREAD, cred, p);
/*
* notify others that we exec'd
*/
KNOTE(&pr->ps_klist, NOTE_EXEC);
/* setup new registers and do misc. setup. */
if (pack.ep_emul->e_fixup != NULL) {
if ((*pack.ep_emul->e_fixup)(p, &pack) != 0)
goto free_pack_abort;
}
#ifdef MACHINE_STACK_GROWS_UP
(*pack.ep_emul->e_setregs)(p, &pack, (u_long)stack + slen, retval);
#else
(*pack.ep_emul->e_setregs)(p, &pack, (u_long)stack, retval);
#endif
/* map the process's signal trampoline code */
if (exec_sigcode_map(p, pack.ep_emul))
goto free_pack_abort;
#ifdef __HAVE_EXEC_MD_MAP
/* perform md specific mappings that process might need */
if (exec_md_map(p, &pack))
goto free_pack_abort;
#endif
if (p->p_flag & P_TRACED)
psignal(p, SIGTRAP);
free(pack.ep_hdr, M_EXEC);
/*
* Call emulation specific exec hook. This can setup per-process
* p->p_emuldata or do any other per-process stuff an emulation needs.
*
* If we are executing process of different emulation than the
* original forked process, call e_proc_exit() of the old emulation
* first, then e_proc_exec() of new emulation. If the emulation is
* same, the exec hook code should deallocate any old emulation
* resources held previously by this process.
*/
if (p->p_emul && p->p_emul->e_proc_exit &&
p->p_emul != pack.ep_emul)
(*p->p_emul->e_proc_exit)(p);
p->p_descfd = 255;
if ((pack.ep_flags & EXEC_HASFD) && pack.ep_fd < 255)
p->p_descfd = pack.ep_fd;
/*
* Call exec hook. Emulation code may NOT store reference to anything
* from &pack.
*/
if (pack.ep_emul->e_proc_exec)
(*pack.ep_emul->e_proc_exec)(p, &pack);
/* update p_emul, the old value is no longer needed */
p->p_emul = pack.ep_emul;
#ifdef KTRACE
if (KTRPOINT(p, KTR_EMUL))
ktremul(p, p->p_emul->e_name);
#endif
atomic_clearbits_int(&p->p_flag, P_INEXEC);
#if NSYSTRACE > 0
if (ISSET(p->p_flag, P_SYSTRACE) &&
wassugid && !ISSET(pr->ps_flags, PS_SUGID | PS_SUGIDEXEC))
systrace_execve1(pathbuf, p);
#endif
if (pathbuf != NULL)
pool_put(&namei_pool, pathbuf);
return (0);
bad:
/* free the vmspace-creation commands, and release their references */
kill_vmcmds(&pack.ep_vmcmds);
/* kill any opened file descriptor, if necessary */
if (pack.ep_flags & EXEC_HASFD) {
pack.ep_flags &= ~EXEC_HASFD;
(void) fdrelease(p, pack.ep_fd);
}
if (pack.ep_interp != NULL)
pool_put(&namei_pool, pack.ep_interp);
if (pack.ep_emul_arg != NULL)
free(pack.ep_emul_arg, M_TEMP);
/* close and put the exec'd file */
vn_close(pack.ep_vp, FREAD, cred, p);
pool_put(&namei_pool, nid.ni_cnd.cn_pnbuf);
uvm_km_free_wakeup(exec_map, (vaddr_t) argp, NCARGS);
freehdr:
free(pack.ep_hdr, M_EXEC);
#if NSYSTRACE > 0
clrflag:
#endif
atomic_clearbits_int(&p->p_flag, P_INEXEC);
if (pathbuf != NULL)
pool_put(&namei_pool, pathbuf);
return (error);
exec_abort:
/*
* the old process doesn't exist anymore. exit gracefully.
* get rid of the (new) address space we have created, if any, get rid
* of our namei data and vnode, and exit noting failure
*/
uvm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
if (pack.ep_interp != NULL)
pool_put(&namei_pool, pack.ep_interp);
if (pack.ep_emul_arg != NULL)
free(pack.ep_emul_arg, M_TEMP);
pool_put(&namei_pool, nid.ni_cnd.cn_pnbuf);
vn_close(pack.ep_vp, FREAD, cred, p);
uvm_km_free_wakeup(exec_map, (vaddr_t) argp, NCARGS);
free_pack_abort:
free(pack.ep_hdr, M_EXEC);
exit1(p, W_EXITCODE(0, SIGABRT), EXIT_NORMAL);
/* NOTREACHED */
atomic_clearbits_int(&p->p_flag, P_INEXEC);
if (pathbuf != NULL)
pool_put(&namei_pool, pathbuf);
return (0);
}
/*
* Machine-dependent startup code
*/
void
cpu_startup()
{
#ifdef DEBUG
extern int pmapdebug;
int opmapdebug = pmapdebug;
#endif
vaddr_t minaddr, maxaddr;
extern struct user *proc0paddr;
#ifdef DEBUG
pmapdebug = 0;
#endif
/*
* fix message buffer mapping
*/
pmap_map(MSGBUF_VA, MSGBUF_PA, MSGBUF_PA + MSGBUFSIZE, UVM_PROT_RW);
initmsgbuf((caddr_t)(MSGBUF_VA + (CPU_ISSUN4 ? 4096 : 0)), MSGBUFSIZE);
proc0.p_addr = proc0paddr;
/* I would print this earlier, but I want it in the message buffer */
if (kap_maskcheck() == 0) {
printf("WARNING: KAP M2C3 or earlier mask detected.\n"
"THE PROCESSOR IN THIS MACHINE SUFFERS FROM SEVERE HARDWARE ISSUES.\n"
"M2C3 PROCESSORS MAY RUN RELIABLY ENOUGH, OLDER WILL DEFINITELY NOT.\n\n");
}
/*
* Good {morning,afternoon,evening,night}.
*/
printf(version);
/*identifycpu();*/
printf("real mem = %d (%dMB)\n", ptoa(physmem),
ptoa(physmem) / 1024 / 1024);
/*
* Allocate a submap for exec arguments. This map effectively
* limits the number of processes exec'ing at any time.
*/
minaddr = vm_map_min(kernel_map);
exec_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
16*NCARGS, VM_MAP_PAGEABLE, FALSE, NULL);
/*
* Allocate a map for physio. Others use a submap of the kernel
* map, but we want one completely separate, even though it uses
* the same pmap.
*/
dvma_base = CPU_ISSUN4M ? DVMA4M_BASE : DVMA_BASE;
dvma_end = CPU_ISSUN4M ? DVMA4M_END : DVMA_END;
phys_map = uvm_map_create(pmap_kernel(), dvma_base, dvma_end,
VM_MAP_INTRSAFE);
if (phys_map == NULL)
panic("unable to create DVMA map");
/*
* Allocate DVMA space and dump into a privately managed
* resource map for double mappings which is usable from
* interrupt contexts.
*/
if (uvm_km_valloc_wait(phys_map, (dvma_end-dvma_base)) != dvma_base)
panic("unable to allocate from DVMA map");
dvmamap_extent = extent_create("dvmamap", dvma_base, dvma_end,
M_DEVBUF, NULL, 0, EX_NOWAIT);
if (dvmamap_extent == 0)
panic("unable to allocate extent for dvma");
#ifdef DEBUG
pmapdebug = opmapdebug;
#endif
printf("avail mem = %lu (%luMB)\n", ptoa(uvmexp.free),
ptoa(uvmexp.free) / 1024 / 1024);
/*
* Set up buffers, so they can be used to read disk labels.
*/
bufinit();
/* Early interrupt handlers initialization */
intr_init();
}
