int
exec_sigcode_map(struct proc *p, struct emul *e)
{
vsize_t sz;
sz = (vaddr_t)e->e_esigcode - (vaddr_t)e->e_sigcode;
/*
* If we don't have a sigobject for this emulation, create one.
*
* sigobject is an anonymous memory object (just like SYSV shared
* memory) that we keep a permanent reference to and that we map
* in all processes that need this sigcode. The creation is simple,
* we create an object, add a permanent reference to it, map it in
* kernel space, copy out the sigcode to it and unmap it.
* Then we map it with PROT_READ|PROT_EXEC into the process just
* the way sys_mmap would map it.
*/
if (e->e_sigobject == NULL) {
vaddr_t va;
int r;
e->e_sigobject = uao_create(sz, 0);
uao_reference(e->e_sigobject); /* permanent reference */
va = vm_map_min(kernel_map); /* hint */
if ((r = uvm_map(kernel_map, &va, round_page(sz), e->e_sigobject,
0, 0, UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_RW,
UVM_INH_SHARE, UVM_ADV_RANDOM, 0)))) {
uao_detach(e->e_sigobject);
return (ENOMEM);
}
memcpy((void *)va, e->e_sigcode, sz);
uvm_unmap(kernel_map, va, va + round_page(sz));
}
/* Just a hint to uvm_mmap where to put it. */
p->p_sigcode = uvm_map_hint(p, VM_PROT_READ|VM_PROT_EXECUTE);
uao_reference(e->e_sigobject);
if (uvm_map(&p->p_vmspace->vm_map, &p->p_sigcode, round_page(sz),
e->e_sigobject, 0, 0, UVM_MAPFLAG(UVM_PROT_RX, UVM_PROT_RX,
UVM_INH_SHARE, UVM_ADV_RANDOM, 0))) {
uao_detach(e->e_sigobject);
return (ENOMEM);
}
return (0);
}
int
sys_shmat(struct proc *p, void *v, register_t *retval)
{
struct sys_shmat_args /* {
syscallarg(int) shmid;
syscallarg(const void *) shmaddr;
syscallarg(int) shmflg;
} */ *uap = v;
int error, i, flags;
struct ucred *cred = p->p_ucred;
struct shmid_ds *shmseg;
struct shmmap_head *shmmap_h;
struct shmmap_state *shmmap_s;
struct shm_handle *shm_handle;
vaddr_t attach_va;
vm_prot_t prot;
vsize_t size;
shmmap_h = (struct shmmap_head *)p->p_vmspace->vm_shm;
if (shmmap_h == NULL) {
size = sizeof(int) +
shminfo.shmseg * sizeof(struct shmmap_state);
shmmap_h = malloc(size, M_SHM, M_WAITOK);
shmmap_h->shmseg = shminfo.shmseg;
for (i = 0, shmmap_s = shmmap_h->state; i < shmmap_h->shmseg;
i++, shmmap_s++)
shmmap_s->shmid = -1;
p->p_vmspace->vm_shm = (caddr_t)shmmap_h;
}
shmseg = shm_find_segment_by_shmid(SCARG(uap, shmid));
if (shmseg == NULL)
return (EINVAL);
error = ipcperm(cred, &shmseg->shm_perm,
(SCARG(uap, shmflg) & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W);
if (error)
return (error);
for (i = 0, shmmap_s = shmmap_h->state; i < shmmap_h->shmseg; i++) {
if (shmmap_s->shmid == -1)
break;
shmmap_s++;
}
if (i >= shmmap_h->shmseg)
return (EMFILE);
size = round_page(shmseg->shm_segsz);
prot = VM_PROT_READ;
if ((SCARG(uap, shmflg) & SHM_RDONLY) == 0)
prot |= VM_PROT_WRITE;
flags = MAP_ANON | MAP_SHARED;
if (SCARG(uap, shmaddr)) {
flags |= MAP_FIXED;
if (SCARG(uap, shmflg) & SHM_RND)
attach_va =
(vaddr_t)SCARG(uap, shmaddr) & ~(SHMLBA-1);
else if (((vaddr_t)SCARG(uap, shmaddr) & (SHMLBA-1)) == 0)
attach_va = (vaddr_t)SCARG(uap, shmaddr);
else
return (EINVAL);
} else {
/* This is just a hint to uvm_map() about where to put it. */
attach_va = uvm_map_hint(p, prot);
}
shm_handle = shmseg->shm_internal;
uao_reference(shm_handle->shm_object);
error = uvm_map(&p->p_vmspace->vm_map, &attach_va, size,
shm_handle->shm_object, 0, 0, UVM_MAPFLAG(prot, prot,
UVM_INH_SHARE, UVM_ADV_RANDOM, 0));
if (error) {
uao_detach(shm_handle->shm_object);
return (error);
}
shmmap_s->va = attach_va;
shmmap_s->shmid = SCARG(uap, shmid);
shmseg->shm_lpid = p->p_p->ps_mainproc->p_pid;
shmseg->shm_atime = time_second;
shmseg->shm_nattch++;
*retval = attach_va;
return (0);
}
int
sys_mmap(struct proc *p, void *v, register_t *retval)
{
struct sys_mmap_args /* {
syscallarg(void *) addr;
syscallarg(size_t) len;
syscallarg(int) prot;
syscallarg(int) flags;
syscallarg(int) fd;
syscallarg(long) pad;
syscallarg(off_t) pos;
} */ *uap = v;
vaddr_t addr;
struct vattr va;
off_t pos;
vsize_t size, pageoff;
vm_prot_t prot, maxprot;
int flags, fd;
vaddr_t vm_min_address = VM_MIN_ADDRESS;
struct filedesc *fdp = p->p_fd;
struct file *fp = NULL;
struct vnode *vp;
caddr_t handle;
int error;
/*
* first, extract syscall args from the uap.
*/
addr = (vaddr_t) SCARG(uap, addr);
size = (vsize_t) SCARG(uap, len);
prot = SCARG(uap, prot);
flags = SCARG(uap, flags);
fd = SCARG(uap, fd);
pos = SCARG(uap, pos);
/*
* Fixup the old deprecated MAP_COPY into MAP_PRIVATE, and
* validate the flags.
*/
if ((prot & VM_PROT_ALL) != prot)
return (EINVAL);
if ((flags & MAP_FLAGMASK) != flags)
return (EINVAL);
if (flags & MAP_COPY)
flags = (flags & ~MAP_COPY) | MAP_PRIVATE;
if ((flags & (MAP_SHARED|MAP_PRIVATE)) == (MAP_SHARED|MAP_PRIVATE))
return (EINVAL);
if (flags & MAP_DENYWRITE)
return (EINVAL);
/*
* align file position and save offset. adjust size.
*/
ALIGN_ADDR(pos, size, pageoff);
/*
* now check (MAP_FIXED) or get (!MAP_FIXED) the "addr"
*/
if (flags & MAP_FIXED) {
/* adjust address by the same amount as we did the offset */
addr -= pageoff;
if (addr & PAGE_MASK)
return (EINVAL); /* not page aligned */
if (addr > SIZE_MAX - size)
return (EINVAL); /* no wrapping! */
if (VM_MAXUSER_ADDRESS > 0 &&
(addr + size) > VM_MAXUSER_ADDRESS)
return (EINVAL);
if (vm_min_address > 0 && addr < vm_min_address)
return (EINVAL);
} else {
/*
* not fixed: make sure we skip over the largest possible heap.
* we will refine our guess later (e.g. to account for VAC, etc)
*/
if (addr == 0)
addr = uvm_map_hint(p, prot);
else if (!(flags & MAP_TRYFIXED) &&
addr < (vaddr_t)p->p_vmspace->vm_daddr)
addr = uvm_map_hint(p, prot);
}
/*
* check for file mappings (i.e. not anonymous) and verify file.
*/
if ((flags & MAP_ANON) == 0) {
if ((fp = fd_getfile(fdp, fd)) == NULL)
return (EBADF);
FREF(fp);
if (fp->f_type != DTYPE_VNODE) {
error = ENODEV; /* only mmap vnodes! */
goto out;
}
vp = (struct vnode *)fp->f_data; /* convert to vnode */
if (vp->v_type != VREG && vp->v_type != VCHR &&
vp->v_type != VBLK) {
error = ENODEV; /* only REG/CHR/BLK support mmap */
goto out;
}
if (vp->v_type == VREG && (pos + size) < pos) {
error = EINVAL; /* no offset wrapping */
goto out;
}
/* special case: catch SunOS style /dev/zero */
if (vp->v_type == VCHR && iszerodev(vp->v_rdev)) {
flags |= MAP_ANON;
FRELE(fp);
fp = NULL;
goto is_anon;
}
/*
* Old programs may not select a specific sharing type, so
* default to an appropriate one.
*
* XXX: how does MAP_ANON fit in the picture?
*/
if ((flags & (MAP_SHARED|MAP_PRIVATE)) == 0) {
#if defined(DEBUG)
printf("WARNING: defaulted mmap() share type to "
"%s (pid %d comm %s)\n", vp->v_type == VCHR ?
"MAP_SHARED" : "MAP_PRIVATE", p->p_pid,
p->p_comm);
#endif
if (vp->v_type == VCHR)
flags |= MAP_SHARED; /* for a device */
else
flags |= MAP_PRIVATE; /* for a file */
}
/*
* MAP_PRIVATE device mappings don't make sense (and aren't
* supported anyway). However, some programs rely on this,
* so just change it to MAP_SHARED.
*/
if (vp->v_type == VCHR && (flags & MAP_PRIVATE) != 0) {
flags = (flags & ~MAP_PRIVATE) | MAP_SHARED;
}
#ifdef ANOUBIS
/* Force DENYWRITE mappings if file->denywrite is set. */
if (fp->denywrite)
flags |= MAP_DENYWRITE;
#endif
/*
* now check protection
*/
/*
* Don't allow the file to be mapped into executable memory if
* the underlying file system is marked as 'noexec'.
*/
if (prot & PROT_EXEC && vp->v_mount->mnt_flag & MNT_NOEXEC) {
error = EACCES;
goto out;
}
maxprot = VM_PROT_EXECUTE;
/* check read access */
if (fp->f_flag & FREAD)
maxprot |= VM_PROT_READ;
else if (prot & PROT_READ) {
error = EACCES;
goto out;
}
/* PROT_EXEC only makes sense if the descriptor is readable. */
if (!(fp->f_flag & FREAD) && prot & PROT_EXEC) {
error = EACCES;
goto out;
}
/* check write access, shared case first */
if (flags & MAP_SHARED) {
/*
* if the file is writable, only add PROT_WRITE to
* maxprot if the file is not immutable, append-only.
* otherwise, if we have asked for PROT_WRITE, return
* EPERM.
*/
if (fp->f_flag & FWRITE) {
if ((error =
VOP_GETATTR(vp, &va, p->p_ucred, p)))
goto out;
if ((va.va_flags & (IMMUTABLE|APPEND)) == 0)
maxprot |= VM_PROT_WRITE;
else if (prot & PROT_WRITE) {
error = EPERM;
goto out;
}
} else if (prot & PROT_WRITE) {
error = EACCES;
goto out;
}
} else {
/* MAP_PRIVATE mappings can always write to */
maxprot |= VM_PROT_WRITE;
}
#ifdef MAC
error = mac_vnode_check_mmap(p->p_ucred, vp, prot, flags);
if (error)
goto out;
#endif
vfs_mark_atime(vp, p->p_ucred);
/*
* set handle to vnode
*/
handle = (caddr_t)vp;
} else { /* MAP_ANON case */
/*
* XXX What do we do about (MAP_SHARED|MAP_PRIVATE) == 0?
*/
if (fd != -1) {
error = EINVAL;
goto out;
}
is_anon: /* label for SunOS style /dev/zero */
handle = NULL;
maxprot = VM_PROT_ALL;
pos = 0;
}
if ((flags & MAP_ANON) != 0 ||
((flags & MAP_PRIVATE) != 0 && (prot & PROT_WRITE) != 0)) {
if (size >
(p->p_rlimit[RLIMIT_DATA].rlim_cur - ptoa(p->p_vmspace->vm_dused))) {
error = ENOMEM;
goto out;
}
}
/*
* now let kernel internal function uvm_mmap do the work.
*/
error = uvm_mmap(&p->p_vmspace->vm_map, &addr, size, prot, maxprot,
flags, handle, pos, p->p_rlimit[RLIMIT_MEMLOCK].rlim_cur, p);
if (error == 0)
/* remember to add offset */
*retval = (register_t)(addr + pageoff);
out:
if (fp)
FRELE(fp);
return (error);
}
/*
* Load a file (interpreter/library) pointed to by path [stolen from
* coff_load_shlib()]. Made slightly generic so it might be used externally.
*/
int
ELFNAME(load_file)(struct proc *p, char *path, struct exec_package *epp,
struct elf_args *ap, Elf_Addr *last)
{
int error, i;
struct nameidata nd;
Elf_Ehdr eh;
Elf_Phdr *ph = NULL;
u_long phsize;
char *bp = NULL;
Elf_Addr addr;
struct vnode *vp;
u_int8_t os; /* Just a dummy in this routine */
Elf_Phdr *base_ph = NULL;
struct interp_ld_sec {
Elf_Addr vaddr;
u_long memsz;
} loadmap[ELF_MAX_VALID_PHDR];
int nload, idx = 0;
Elf_Addr pos = *last;
int file_align;
bp = path;
NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, path, p);
if ((error = namei(&nd)) != 0) {
return (error);
}
vp = nd.ni_vp;
if (vp->v_type != VREG) {
error = EACCES;
goto bad;
}
if ((error = VOP_GETATTR(vp, epp->ep_vap, p->p_ucred, p)) != 0)
goto bad;
if (vp->v_mount->mnt_flag & MNT_NOEXEC) {
error = EACCES;
goto bad;
}
if ((error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) != 0)
goto bad1;
if ((error = ELFNAME(read_from)(p, nd.ni_vp, 0,
(caddr_t)&eh, sizeof(eh))) != 0)
goto bad1;
if (ELFNAME(check_header)(&eh, ET_DYN) &&
ELFNAME(olf_check_header)(&eh, ET_DYN, &os)) {
error = ENOEXEC;
goto bad1;
}
phsize = eh.e_phnum * sizeof(Elf_Phdr);
ph = (Elf_Phdr *)malloc(phsize, M_TEMP, M_WAITOK);
if ((error = ELFNAME(read_from)(p, nd.ni_vp, eh.e_phoff, (caddr_t)ph,
phsize)) != 0)
goto bad1;
for (i = 0; i < eh.e_phnum; i++) {
if (ph[i].p_type == PT_LOAD) {
loadmap[idx].vaddr = trunc_page(ph[i].p_vaddr);
loadmap[idx].memsz = round_page (ph[i].p_vaddr +
ph[i].p_memsz - loadmap[idx].vaddr);
file_align = ph[i].p_align;
idx++;
}
}
nload = idx;
/*
* If no position to load the interpreter was set by a probe
* function, pick the same address that a non-fixed mmap(0, ..)
* would (i.e. something safely out of the way).
*/
if (pos == ELFDEFNNAME(NO_ADDR)) {
pos = uvm_map_hint(p, VM_PROT_EXECUTE);
}
pos = ELF_ROUND(pos, file_align);
*last = epp->ep_interp_pos = pos;
for (i = 0; i < nload;/**/) {
vaddr_t addr;
struct uvm_object *uobj;
off_t uoff;
size_t size;
#ifdef this_needs_fixing
if (i == 0) {
uobj = &vp->v_uvm.u_obj;
/* need to fix uoff */
} else {
#endif
uobj = NULL;
uoff = 0;
#ifdef this_needs_fixing
}
#endif
addr = trunc_page(pos + loadmap[i].vaddr);
size = round_page(addr + loadmap[i].memsz) - addr;
/* CRAP - map_findspace does not avoid daddr+MAXDSIZ */
if ((addr + size > (vaddr_t)p->p_vmspace->vm_daddr) &&
(addr < (vaddr_t)p->p_vmspace->vm_daddr + MAXDSIZ))
addr = round_page((vaddr_t)p->p_vmspace->vm_daddr +
MAXDSIZ);
if (uvm_map_findspace(&p->p_vmspace->vm_map, addr, size,
&addr, uobj, uoff, 0, UVM_FLAG_FIXED) == NULL) {
if (uvm_map_findspace(&p->p_vmspace->vm_map, addr, size,
&addr, uobj, uoff, 0, 0) == NULL) {
error = ENOMEM; /* XXX */
goto bad1;
}
}
if (addr != pos + loadmap[i].vaddr) {
/* base changed. */
pos = addr - trunc_page(loadmap[i].vaddr);
pos = ELF_ROUND(pos,file_align);
epp->ep_interp_pos = *last = pos;
i = 0;
continue;
}
i++;
}
/*
* Load all the necessary sections
*/
for (i = 0; i < eh.e_phnum; i++) {
Elf_Addr size = 0;
int prot = 0;
int flags;
switch (ph[i].p_type) {
case PT_LOAD:
if (base_ph == NULL) {
flags = VMCMD_BASE;
addr = *last;
base_ph = &ph[i];
} else {
flags = VMCMD_RELATIVE;
addr = ph[i].p_vaddr - base_ph->p_vaddr;
}
ELFNAME(load_psection)(&epp->ep_vmcmds, nd.ni_vp,
&ph[i], &addr, &size, &prot, flags);
/* If entry is within this section it must be text */
if (eh.e_entry >= ph[i].p_vaddr &&
eh.e_entry < (ph[i].p_vaddr + size)) {
epp->ep_entry = addr + eh.e_entry -
ELF_TRUNC(ph[i].p_vaddr,ph[i].p_align);
ap->arg_interp = addr;
}
addr += size;
break;
case PT_DYNAMIC:
case PT_PHDR:
case PT_NOTE:
break;
default:
break;
}
}
vn_marktext(nd.ni_vp);
bad1:
VOP_CLOSE(nd.ni_vp, FREAD, p->p_ucred, p);
bad:
if (ph != NULL)
free((char *)ph, M_TEMP);
*last = addr;
vput(nd.ni_vp);
return (error);
}
int
sys_mquery(struct proc *p, void *v, register_t *retval)
{
struct sys_mquery_args /* {
syscallarg(void *) addr;
syscallarg(size_t) len;
syscallarg(int) prot;
syscallarg(int) flags;
syscallarg(int) fd;
syscallarg(long) pad;
syscallarg(off_t) pos;
} */ *uap = v;
struct file *fp;
struct uvm_object *uobj;
voff_t uoff;
int error;
vaddr_t vaddr;
int flags = 0;
vsize_t size;
vm_prot_t prot;
int fd;
vaddr = (vaddr_t) SCARG(uap, addr);
prot = SCARG(uap, prot);
size = (vsize_t) SCARG(uap, len);
fd = SCARG(uap, fd);
if ((prot & VM_PROT_ALL) != prot)
return (EINVAL);
if (SCARG(uap, flags) & MAP_FIXED)
flags |= UVM_FLAG_FIXED;
if (fd >= 0) {
if ((error = getvnode(p->p_fd, fd, &fp)) != 0)
return (error);
uobj = &((struct vnode *)fp->f_data)->v_uvm.u_obj;
uoff = SCARG(uap, pos);
} else {
fp = NULL;
uobj = NULL;
uoff = 0;
}
if (vaddr == 0)
vaddr = uvm_map_hint(p, prot);
/* prevent a user requested address from falling in heap space */
if ((vaddr + size > (vaddr_t)p->p_vmspace->vm_daddr) &&
(vaddr < (vaddr_t)p->p_vmspace->vm_daddr + MAXDSIZ)) {
if (flags & UVM_FLAG_FIXED) {
error = EINVAL;
goto done;
}
vaddr = round_page((vaddr_t)p->p_vmspace->vm_daddr + MAXDSIZ);
}
vm_map_lock(&p->p_vmspace->vm_map);
again:
if (uvm_map_findspace(&p->p_vmspace->vm_map, vaddr, size,
&vaddr, uobj, uoff, 0, flags) == NULL) {
if (flags & UVM_FLAG_FIXED)
error = EINVAL;
else
error = ENOMEM;
} else {
/* prevent a returned address from falling in heap space */
if ((vaddr + size > (vaddr_t)p->p_vmspace->vm_daddr)
&& (vaddr < (vaddr_t)p->p_vmspace->vm_daddr + MAXDSIZ)) {
vaddr = round_page((vaddr_t)p->p_vmspace->vm_daddr +
MAXDSIZ);
goto again;
}
error = 0;
*retval = (register_t)(vaddr);
}
vm_map_unlock(&p->p_vmspace->vm_map);
done:
if (fp != NULL)
FRELE(fp);
return (error);
}