/*
* fp_open:
*
* Open a file as specified. Use O_* flags for flags.
*
* vn_open() asserts that the cred must match the process's cred.
*
* NOTE! when fp_open() is called from a pure thread, root creds are
* used.
*/
int
fp_open(const char *path, int flags, int mode, file_t *fpp)
{
struct nlookupdata nd;
struct thread *td;
struct file *fp;
int error;
if ((error = falloc(NULL, fpp, NULL)) != 0)
return (error);
fp = *fpp;
td = curthread;
if (td->td_proc)
fsetcred(fp, td->td_proc->p_ucred);
error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_LOCKVP);
flags = FFLAGS(flags);
if (error == 0)
error = vn_open(&nd, fp, flags, mode);
nlookup_done(&nd);
if (error) {
fdrop(fp);
*fpp = NULL;
}
return(error);
}
/*
* MPALMOSTSAFE
*/
int
sys_otruncate(struct otruncate_args *uap)
{
struct nlookupdata nd;
int error;
get_mplock();
error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
if (error == 0)
error = kern_truncate(&nd, uap->length);
nlookup_done(&nd);
rel_mplock();
return (error);
}
/*
* MPALMOSTSAFE
*/
int
sys_linux_execve(struct linux_execve_args *args)
{
struct nlookupdata nd;
struct image_args exec_args;
char *path;
int error;
error = linux_copyin_path(args->path, &path, LINUX_PATH_EXISTS);
if (error)
return (error);
#ifdef DEBUG
if (ldebug(execve))
kprintf(ARGS(execve, "%s"), path);
#endif
get_mplock();
error = nlookup_init(&nd, path, UIO_SYSSPACE, NLC_FOLLOW);
bzero(&exec_args, sizeof(exec_args));
if (error == 0) {
error = exec_copyin_args(&exec_args, path, PATH_SYSSPACE,
args->argp, args->envp);
}
if (error == 0)
error = kern_execve(&nd, &exec_args);
nlookup_done(&nd);
/*
* The syscall result is returned in registers to the new program.
* Linux will register %edx as an atexit function and we must be
* sure to set it to 0. XXX
*/
if (error == 0) {
args->sysmsg_result64 = 0;
if (curproc->p_sysent == &elf_linux_sysvec)
error = emuldata_init(curproc, NULL, 0);
}
exec_free_args(&exec_args);
linux_free_path(&path);
if (error < 0) {
/* We hit a lethal error condition. Let's die now. */
exit1(W_EXITCODE(0, SIGABRT));
/* NOTREACHED */
}
rel_mplock();
return(error);
}
/*
* MPALMOSTSAFE
*/
int
sys_ocreat(struct ocreat_args *uap)
{
struct nlookupdata nd;
int error;
get_mplock();
error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
if (error == 0) {
error = kern_open(&nd, O_WRONLY | O_CREAT | O_TRUNC,
uap->mode, &uap->sysmsg_iresult);
}
rel_mplock();
return (error);
}
/*
* MPALMOSTSAFE
*/
int
sys_olstat(struct olstat_args *uap)
{
struct nlookupdata nd;
struct stat st;
int error;
get_mplock();
error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0);
if (error == 0) {
error = kern_stat(&nd, &st);
if (error == 0)
error = compat_43_copyout_stat(&st, uap->ub);
nlookup_done(&nd);
}
rel_mplock();
return (error);
}
static int
hammer_setup_device(struct vnode **devvpp, const char *dev_path, int ronly)
{
int error;
struct nlookupdata nd;
/*
* Get the device vnode
*/
if (*devvpp == NULL) {
error = nlookup_init(&nd, dev_path, UIO_SYSSPACE, NLC_FOLLOW);
if (error == 0)
error = nlookup(&nd);
if (error == 0)
error = cache_vref(&nd.nl_nch, nd.nl_cred, devvpp);
nlookup_done(&nd);
} else {
error = 0;
}
if (error == 0) {
if (vn_isdisk(*devvpp, &error)) {
error = vfs_mountedon(*devvpp);
}
}
if (error == 0 && vcount(*devvpp) > 0)
error = EBUSY;
if (error == 0) {
vn_lock(*devvpp, LK_EXCLUSIVE | LK_RETRY);
error = vinvalbuf(*devvpp, V_SAVE, 0, 0);
if (error == 0) {
error = VOP_OPEN(*devvpp,
(ronly ? FREAD : FREAD|FWRITE),
FSCRED, NULL);
}
vn_unlock(*devvpp);
}
if (error && *devvpp) {
vrele(*devvpp);
*devvpp = NULL;
}
return (error);
}
/*
* nlookup_init() for "at" family of syscalls.
*
* Works similarly to nlookup_init() but if path is relative and fd is not
* AT_FDCWD, path is interpreted relative to the directory pointed to by fd.
* In this case, the file entry pointed to by fd is ref'ed and returned in
* *fpp.
*
* If the call succeeds, nlookup_done_at() must be called to clean-up the nd
* and release the ref to the file entry.
*/
int
nlookup_init_at(struct nlookupdata *nd, struct file **fpp, int fd,
const char *path, enum uio_seg seg, int flags)
{
struct thread *td = curthread;
struct file* fp;
struct vnode *vp;
int error;
*fpp = NULL;
if ((error = nlookup_init(nd, path, seg, flags)) != 0) {
return (error);
}
if (nd->nl_path[0] != '/' && fd != AT_FDCWD) {
if ((error = holdvnode(td, fd, &fp)) != 0)
goto done;
vp = (struct vnode*)fp->f_data;
if (vp->v_type != VDIR || fp->f_nchandle.ncp == NULL) {
fdrop(fp);
fp = NULL;
error = ENOTDIR;
goto done;
}
if (nd->nl_flags & NLC_NCDIR) {
cache_drop_ncdir(&nd->nl_nch);
nd->nl_flags &= ~NLC_NCDIR;
} else {
cache_drop(&nd->nl_nch);
}
cache_copy(&fp->f_nchandle, &nd->nl_nch);
*fpp = fp;
}
done:
if (error)
nlookup_done(nd);
return (error);
}
/*
* Simple all-in-one nlookup. Returns a locked namecache structure or NULL
* if an error occured.
*
* Note that the returned ncp is not checked for permissions, though VEXEC
* is checked on the directory path leading up to the result. The caller
* must call naccess() to check the permissions of the returned leaf.
*/
struct nchandle
nlookup_simple(const char *str, enum uio_seg seg,
int niflags, int *error)
{
struct nlookupdata nd;
struct nchandle nch;
*error = nlookup_init(&nd, str, seg, niflags);
if (*error == 0) {
if ((*error = nlookup(&nd)) == 0) {
nch = nd.nl_nch; /* keep hold ref from structure */
cache_zero(&nd.nl_nch); /* and NULL out */
} else {
cache_zero(&nch);
}
nlookup_done(&nd);
} else {
cache_zero(&nch);
}
return(nch);
}
/*
* stat_args(char *path, struct dfbsd12_stat *ub)
*
* Get file status; this version follows links.
*
* MPALMOSTSAFE
*/
int
sys_dfbsd12_stat(struct dfbsd12_stat_args *uap)
{
struct nlookupdata nd;
struct dfbsd12_stat ost;
struct stat st;
int error;
get_mplock();
error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW);
if (error == 0) {
error = kern_stat(&nd, &st);
if (error == 0) {
cvtstat(&ost, &st);
error = copyout(&ost, uap->ub, sizeof(ost));
}
}
nlookup_done(&nd);
rel_mplock();
return (error);
}
if (pmp->pm_flags & MSDOSFSMNT_NOWIN95)
pmp->pm_flags |= MSDOSFSMNT_SHORTNAME;
}
#endif
/*
* Process export requests.
*/
return (vfs_export(mp, &pmp->pm_export, &args.export));
}
}
/*
* Not an update, or updating the name: look up the name
* and verify that it refers to a sensible block device.
*/
devvp = NULL;
error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
if (error == 0)
error = nlookup(&nd);
if (error == 0)
error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
nlookup_done(&nd);
if (error)
return (error);
if (!vn_isdisk(devvp, &error)) {
vrele(devvp);
return (error);
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
static int
link_elf_load_file(const char* filename, linker_file_t* result)
{
struct nlookupdata nd;
struct thread *td = curthread; /* XXX */
struct proc *p = td->td_proc;
struct vnode *vp;
Elf_Ehdr *hdr;
caddr_t firstpage;
int nbytes, i;
Elf_Phdr *phdr;
Elf_Phdr *phlimit;
Elf_Phdr *segs[2];
int nsegs;
Elf_Phdr *phdyn;
caddr_t mapbase;
size_t mapsize;
Elf_Addr base_vaddr;
Elf_Addr base_vlimit;
int error = 0;
int resid;
elf_file_t ef;
linker_file_t lf;
char *pathname;
Elf_Shdr *shdr;
int symtabindex;
int symstrindex;
int symcnt;
int strcnt;
/* XXX Hack for firmware loading where p == NULL */
if (p == NULL) {
p = &proc0;
}
KKASSERT(p != NULL);
if (p->p_ucred == NULL) {
kprintf("link_elf_load_file: cannot load '%s' from filesystem"
" this early\n", filename);
return ENOENT;
}
shdr = NULL;
lf = NULL;
pathname = linker_search_path(filename);
if (pathname == NULL)
return ENOENT;
error = nlookup_init(&nd, pathname, UIO_SYSSPACE, NLC_FOLLOW|NLC_LOCKVP);
if (error == 0)
error = vn_open(&nd, NULL, FREAD, 0);
kfree(pathname, M_LINKER);
if (error) {
nlookup_done(&nd);
return error;
}
vp = nd.nl_open_vp;
nd.nl_open_vp = NULL;
nlookup_done(&nd);
/*
* Read the elf header from the file.
*/
firstpage = kmalloc(PAGE_SIZE, M_LINKER, M_WAITOK);
hdr = (Elf_Ehdr *)firstpage;
error = vn_rdwr(UIO_READ, vp, firstpage, PAGE_SIZE, 0,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
nbytes = PAGE_SIZE - resid;
if (error)
goto out;
if (!IS_ELF(*hdr)) {
error = ENOEXEC;
goto out;
}
if (hdr->e_ident[EI_CLASS] != ELF_TARG_CLASS
|| hdr->e_ident[EI_DATA] != ELF_TARG_DATA) {
link_elf_error("Unsupported file layout");
error = ENOEXEC;
goto out;
}
if (hdr->e_ident[EI_VERSION] != EV_CURRENT
|| hdr->e_version != EV_CURRENT) {
link_elf_error("Unsupported file version");
error = ENOEXEC;
goto out;
}
if (hdr->e_type != ET_EXEC && hdr->e_type != ET_DYN) {
error = ENOSYS;
goto out;
}
if (hdr->e_machine != ELF_TARG_MACH) {
link_elf_error("Unsupported machine");
error = ENOEXEC;
goto out;
}
/*
* We rely on the program header being in the first page. This is
* not strictly required by the ABI specification, but it seems to
* always true in practice. And, it simplifies things considerably.
*/
if (!((hdr->e_phentsize == sizeof(Elf_Phdr)) &&
(hdr->e_phoff + hdr->e_phnum*sizeof(Elf_Phdr) <= PAGE_SIZE) &&
(hdr->e_phoff + hdr->e_phnum*sizeof(Elf_Phdr) <= nbytes)))
link_elf_error("Unreadable program headers");
/*
* Scan the program header entries, and save key information.
*
* We rely on there being exactly two load segments, text and data,
* in that order.
*/
phdr = (Elf_Phdr *) (firstpage + hdr->e_phoff);
phlimit = phdr + hdr->e_phnum;
nsegs = 0;
phdyn = NULL;
while (phdr < phlimit) {
switch (phdr->p_type) {
case PT_LOAD:
if (nsegs == 2) {
link_elf_error("Too many sections");
error = ENOEXEC;
goto out;
}
segs[nsegs] = phdr;
++nsegs;
break;
case PT_PHDR:
break;
case PT_DYNAMIC:
phdyn = phdr;
break;
case PT_INTERP:
error = ENOSYS;
goto out;
}
++phdr;
}
if (phdyn == NULL) {
link_elf_error("Object is not dynamically-linked");
error = ENOEXEC;
goto out;
}
/*
* Allocate the entire address space of the object, to stake out our
* contiguous region, and to establish the base address for relocation.
*/
base_vaddr = trunc_page(segs[0]->p_vaddr);
base_vlimit = round_page(segs[1]->p_vaddr + segs[1]->p_memsz);
mapsize = base_vlimit - base_vaddr;
ef = kmalloc(sizeof(struct elf_file), M_LINKER, M_WAITOK | M_ZERO);
ef->address = kmalloc(mapsize, M_LINKER, M_WAITOK);
mapbase = ef->address;
/*
* Read the text and data sections and zero the bss.
*/
for (i = 0; i < 2; i++) {
caddr_t segbase = mapbase + segs[i]->p_vaddr - base_vaddr;
error = vn_rdwr(UIO_READ, vp,
segbase, segs[i]->p_filesz, segs[i]->p_offset,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error) {
kfree(ef->address, M_LINKER);
kfree(ef, M_LINKER);
goto out;
}
bzero(segbase + segs[i]->p_filesz,
segs[i]->p_memsz - segs[i]->p_filesz);
}
ef->dynamic = (const Elf_Dyn *) (mapbase + phdyn->p_vaddr - base_vaddr);
lf = linker_make_file(filename, ef, &link_elf_file_ops);
if (lf == NULL) {
kfree(ef->address, M_LINKER);
kfree(ef, M_LINKER);
error = ENOMEM;
goto out;
}
lf->address = ef->address;
lf->size = mapsize;
error = parse_dynamic(lf);
if (error)
goto out;
link_elf_reloc_local(lf);
error = linker_load_dependencies(lf);
if (error)
goto out;
error = relocate_file(lf);
if (error)
goto out;
/* Try and load the symbol table if it's present. (you can strip it!) */
nbytes = hdr->e_shnum * hdr->e_shentsize;
if (nbytes == 0 || hdr->e_shoff == 0)
goto nosyms;
shdr = kmalloc(nbytes, M_LINKER, M_WAITOK | M_ZERO);
error = vn_rdwr(UIO_READ, vp,
(caddr_t)shdr, nbytes, hdr->e_shoff,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error)
goto out;
symtabindex = -1;
symstrindex = -1;
for (i = 0; i < hdr->e_shnum; i++) {
if (shdr[i].sh_type == SHT_SYMTAB) {
symtabindex = i;
symstrindex = shdr[i].sh_link;
}
}
if (symtabindex < 0 || symstrindex < 0)
goto nosyms;
symcnt = shdr[symtabindex].sh_size;
ef->symbase = kmalloc(symcnt, M_LINKER, M_WAITOK);
strcnt = shdr[symstrindex].sh_size;
ef->strbase = kmalloc(strcnt, M_LINKER, M_WAITOK);
error = vn_rdwr(UIO_READ, vp,
ef->symbase, symcnt, shdr[symtabindex].sh_offset,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error)
goto out;
error = vn_rdwr(UIO_READ, vp,
ef->strbase, strcnt, shdr[symstrindex].sh_offset,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error)
goto out;
ef->ddbsymcnt = symcnt / sizeof(Elf_Sym);
ef->ddbsymtab = (const Elf_Sym *)ef->symbase;
ef->ddbstrcnt = strcnt;
ef->ddbstrtab = ef->strbase;
nosyms:
*result = lf;
out:
if (error && lf)
linker_file_unload(lf);
if (shdr)
kfree(shdr, M_LINKER);
if (firstpage)
kfree(firstpage, M_LINKER);
vn_unlock(vp);
vn_close(vp, FREAD, NULL);
return error;
}
static int
link_elf_obj_load_file(const char *filename, linker_file_t * result)
{
struct nlookupdata nd;
struct thread *td = curthread; /* XXX */
struct proc *p = td->td_proc;
char *pathname;
struct vnode *vp;
Elf_Ehdr *hdr;
Elf_Shdr *shdr;
Elf_Sym *es;
int nbytes, i, j;
vm_offset_t mapbase;
size_t mapsize;
int error = 0;
int resid;
elf_file_t ef;
linker_file_t lf;
int symtabindex;
int symstrindex;
int shstrindex;
int nsym;
int pb, rl, ra;
int alignmask;
/* XXX Hack for firmware loading where p == NULL */
if (p == NULL) {
p = &proc0;
}
KKASSERT(p != NULL);
if (p->p_ucred == NULL) {
kprintf("link_elf_obj_load_file: cannot load '%s' from filesystem"
" this early\n", filename);
return ENOENT;
}
shdr = NULL;
lf = NULL;
mapsize = 0;
hdr = NULL;
pathname = linker_search_path(filename);
if (pathname == NULL)
return ENOENT;
error = nlookup_init(&nd, pathname, UIO_SYSSPACE, NLC_FOLLOW | NLC_LOCKVP);
if (error == 0)
error = vn_open(&nd, NULL, FREAD, 0);
kfree(pathname, M_LINKER);
if (error) {
nlookup_done(&nd);
return error;
}
vp = nd.nl_open_vp;
nd.nl_open_vp = NULL;
nlookup_done(&nd);
/*
* Read the elf header from the file.
*/
hdr = kmalloc(sizeof(*hdr), M_LINKER, M_WAITOK);
error = vn_rdwr(UIO_READ, vp, (void *)hdr, sizeof(*hdr), 0,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error)
goto out;
if (resid != 0) {
error = ENOEXEC;
goto out;
}
if (!IS_ELF(*hdr)) {
error = ENOEXEC;
goto out;
}
if (hdr->e_ident[EI_CLASS] != ELF_TARG_CLASS
|| hdr->e_ident[EI_DATA] != ELF_TARG_DATA) {
link_elf_obj_error(filename, "Unsupported file layout");
error = ENOEXEC;
goto out;
}
if (hdr->e_ident[EI_VERSION] != EV_CURRENT
|| hdr->e_version != EV_CURRENT) {
link_elf_obj_error(filename, "Unsupported file version");
error = ENOEXEC;
goto out;
}
if (hdr->e_type != ET_REL) {
error = ENOSYS;
goto out;
}
if (hdr->e_machine != ELF_TARG_MACH) {
link_elf_obj_error(filename, "Unsupported machine");
error = ENOEXEC;
goto out;
}
ef = kmalloc(sizeof(struct elf_file), M_LINKER, M_WAITOK | M_ZERO);
lf = linker_make_file(filename, ef, &link_elf_obj_file_ops);
if (lf == NULL) {
kfree(ef, M_LINKER);
error = ENOMEM;
goto out;
}
ef->nprogtab = 0;
ef->e_shdr = NULL;
ef->nreltab = 0;
ef->nrelatab = 0;
/* Allocate and read in the section header */
nbytes = hdr->e_shnum * hdr->e_shentsize;
if (nbytes == 0 || hdr->e_shoff == 0 ||
hdr->e_shentsize != sizeof(Elf_Shdr)) {
error = ENOEXEC;
goto out;
}
shdr = kmalloc(nbytes, M_LINKER, M_WAITOK);
ef->e_shdr = shdr;
error = vn_rdwr(UIO_READ, vp, (caddr_t) shdr, nbytes, hdr->e_shoff,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error)
goto out;
if (resid) {
error = ENOEXEC;
goto out;
}
/* Scan the section header for information and table sizing. */
nsym = 0;
symtabindex = -1;
symstrindex = -1;
for (i = 0; i < hdr->e_shnum; i++) {
if (shdr[i].sh_size == 0)
continue;
switch (shdr[i].sh_type) {
case SHT_PROGBITS:
case SHT_NOBITS:
ef->nprogtab++;
break;
case SHT_SYMTAB:
nsym++;
symtabindex = i;
symstrindex = shdr[i].sh_link;
break;
case SHT_REL:
ef->nreltab++;
break;
case SHT_RELA:
ef->nrelatab++;
break;
case SHT_STRTAB:
break;
}
}
if (ef->nprogtab == 0) {
link_elf_obj_error(filename, "file has no contents");
error = ENOEXEC;
goto out;
}
if (nsym != 1) {
/* Only allow one symbol table for now */
link_elf_obj_error(filename, "file has no valid symbol table");
error = ENOEXEC;
goto out;
}
if (symstrindex < 0 || symstrindex > hdr->e_shnum ||
shdr[symstrindex].sh_type != SHT_STRTAB) {
link_elf_obj_error(filename, "file has invalid symbol strings");
error = ENOEXEC;
goto out;
}
/* Allocate space for tracking the load chunks */
if (ef->nprogtab != 0)
ef->progtab = kmalloc(ef->nprogtab * sizeof(*ef->progtab),
M_LINKER, M_WAITOK | M_ZERO);
if (ef->nreltab != 0)
ef->reltab = kmalloc(ef->nreltab * sizeof(*ef->reltab),
M_LINKER, M_WAITOK | M_ZERO);
if (ef->nrelatab != 0)
ef->relatab = kmalloc(ef->nrelatab * sizeof(*ef->relatab),
M_LINKER, M_WAITOK | M_ZERO);
if ((ef->nprogtab != 0 && ef->progtab == NULL) ||
(ef->nreltab != 0 && ef->reltab == NULL) ||
(ef->nrelatab != 0 && ef->relatab == NULL)) {
error = ENOMEM;
goto out;
}
if (symtabindex == -1)
panic("lost symbol table index");
/* Allocate space for and load the symbol table */
ef->ddbsymcnt = shdr[symtabindex].sh_size / sizeof(Elf_Sym);
ef->ddbsymtab = kmalloc(shdr[symtabindex].sh_size, M_LINKER, M_WAITOK);
error = vn_rdwr(UIO_READ, vp, (void *)ef->ddbsymtab,
shdr[symtabindex].sh_size, shdr[symtabindex].sh_offset,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error)
goto out;
if (resid != 0) {
error = EINVAL;
goto out;
}
if (symstrindex == -1)
panic("lost symbol string index");
/* Allocate space for and load the symbol strings */
ef->ddbstrcnt = shdr[symstrindex].sh_size;
ef->ddbstrtab = kmalloc(shdr[symstrindex].sh_size, M_LINKER, M_WAITOK);
error = vn_rdwr(UIO_READ, vp, ef->ddbstrtab,
shdr[symstrindex].sh_size, shdr[symstrindex].sh_offset,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error)
goto out;
if (resid != 0) {
error = EINVAL;
goto out;
}
/* Do we have a string table for the section names? */
shstrindex = -1;
if (hdr->e_shstrndx != 0 &&
shdr[hdr->e_shstrndx].sh_type == SHT_STRTAB) {
shstrindex = hdr->e_shstrndx;
ef->shstrcnt = shdr[shstrindex].sh_size;
ef->shstrtab = kmalloc(shdr[shstrindex].sh_size, M_LINKER,
M_WAITOK);
error = vn_rdwr(UIO_READ, vp, ef->shstrtab,
shdr[shstrindex].sh_size, shdr[shstrindex].sh_offset,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error)
goto out;
if (resid != 0) {
error = EINVAL;
goto out;
}
}
/* Size up code/data(progbits) and bss(nobits). */
alignmask = 0;
for (i = 0; i < hdr->e_shnum; i++) {
if (shdr[i].sh_size == 0)
continue;
switch (shdr[i].sh_type) {
case SHT_PROGBITS:
case SHT_NOBITS:
alignmask = shdr[i].sh_addralign - 1;
mapsize += alignmask;
mapsize &= ~alignmask;
mapsize += shdr[i].sh_size;
break;
}
}
/*
* We know how much space we need for the text/data/bss/etc. This
* stuff needs to be in a single chunk so that profiling etc can get
* the bounds and gdb can associate offsets with modules
*/
ef->object = vm_object_allocate(OBJT_DEFAULT,
round_page(mapsize) >> PAGE_SHIFT);
if (ef->object == NULL) {
error = ENOMEM;
goto out;
}
vm_object_hold(ef->object);
vm_object_reference_locked(ef->object);
ef->address = (caddr_t) vm_map_min(&kernel_map);
ef->bytes = 0;
/*
* In order to satisfy x86_64's architectural requirements on the
* location of code and data in the kernel's address space, request a
* mapping that is above the kernel.
*
* vkernel64's text+data is outside the managed VM space entirely.
*/
#if defined(__x86_64__) && defined(_KERNEL_VIRTUAL)
error = vkernel_module_memory_alloc(&mapbase, round_page(mapsize));
vm_object_drop(ef->object);
#else
mapbase = KERNBASE;
error = vm_map_find(&kernel_map, ef->object, NULL,
0, &mapbase, round_page(mapsize),
PAGE_SIZE,
TRUE, VM_MAPTYPE_NORMAL,
VM_PROT_ALL, VM_PROT_ALL, FALSE);
vm_object_drop(ef->object);
if (error) {
vm_object_deallocate(ef->object);
ef->object = NULL;
goto out;
}
/* Wire the pages */
error = vm_map_wire(&kernel_map, mapbase,
mapbase + round_page(mapsize), 0);
#endif
if (error != KERN_SUCCESS) {
error = ENOMEM;
goto out;
}
/* Inform the kld system about the situation */
lf->address = ef->address = (caddr_t) mapbase;
lf->size = round_page(mapsize);
ef->bytes = mapsize;
/*
* Now load code/data(progbits), zero bss(nobits), allocate space for
* and load relocs
*/
pb = 0;
rl = 0;
ra = 0;
alignmask = 0;
for (i = 0; i < hdr->e_shnum; i++) {
if (shdr[i].sh_size == 0)
continue;
switch (shdr[i].sh_type) {
case SHT_PROGBITS:
case SHT_NOBITS:
alignmask = shdr[i].sh_addralign - 1;
mapbase += alignmask;
mapbase &= ~alignmask;
if (ef->shstrtab && shdr[i].sh_name != 0)
ef->progtab[pb].name =
ef->shstrtab + shdr[i].sh_name;
else if (shdr[i].sh_type == SHT_PROGBITS)
ef->progtab[pb].name = "<<PROGBITS>>";
else
ef->progtab[pb].name = "<<NOBITS>>";
#if 0
if (ef->progtab[pb].name != NULL &&
!strcmp(ef->progtab[pb].name, "set_pcpu"))
ef->progtab[pb].addr =
dpcpu_alloc(shdr[i].sh_size);
#ifdef VIMAGE
else if (ef->progtab[pb].name != NULL &&
!strcmp(ef->progtab[pb].name, VNET_SETNAME))
ef->progtab[pb].addr =
vnet_data_alloc(shdr[i].sh_size);
#endif
else
#endif
ef->progtab[pb].addr =
(void *)(uintptr_t) mapbase;
if (ef->progtab[pb].addr == NULL) {
error = ENOSPC;
goto out;
}
ef->progtab[pb].size = shdr[i].sh_size;
ef->progtab[pb].sec = i;
if (shdr[i].sh_type == SHT_PROGBITS) {
error = vn_rdwr(UIO_READ, vp,
ef->progtab[pb].addr,
shdr[i].sh_size, shdr[i].sh_offset,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred,
&resid);
if (error)
goto out;
if (resid != 0) {
error = EINVAL;
goto out;
}
#if 0
/* Initialize the per-cpu or vnet area. */
if (ef->progtab[pb].addr != (void *)mapbase &&
!strcmp(ef->progtab[pb].name, "set_pcpu"))
dpcpu_copy(ef->progtab[pb].addr,
shdr[i].sh_size);
#ifdef VIMAGE
else if (ef->progtab[pb].addr !=
(void *)mapbase &&
!strcmp(ef->progtab[pb].name, VNET_SETNAME))
vnet_data_copy(ef->progtab[pb].addr,
shdr[i].sh_size);
#endif
#endif
} else
bzero(ef->progtab[pb].addr, shdr[i].sh_size);
/* Update all symbol values with the offset. */
for (j = 0; j < ef->ddbsymcnt; j++) {
es = &ef->ddbsymtab[j];
if (es->st_shndx != i)
continue;
es->st_value += (Elf_Addr) ef->progtab[pb].addr;
}
mapbase += shdr[i].sh_size;
pb++;
break;
case SHT_REL:
ef->reltab[rl].rel = kmalloc(shdr[i].sh_size, M_LINKER, M_WAITOK);
ef->reltab[rl].nrel = shdr[i].sh_size / sizeof(Elf_Rel);
ef->reltab[rl].sec = shdr[i].sh_info;
error = vn_rdwr(UIO_READ, vp,
(void *)ef->reltab[rl].rel,
shdr[i].sh_size, shdr[i].sh_offset,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error)
goto out;
if (resid != 0) {
error = EINVAL;
goto out;
}
rl++;
break;
case SHT_RELA:
ef->relatab[ra].rela = kmalloc(shdr[i].sh_size, M_LINKER, M_WAITOK);
ef->relatab[ra].nrela = shdr[i].sh_size / sizeof(Elf_Rela);
ef->relatab[ra].sec = shdr[i].sh_info;
error = vn_rdwr(UIO_READ, vp,
(void *)ef->relatab[ra].rela,
shdr[i].sh_size, shdr[i].sh_offset,
UIO_SYSSPACE, IO_NODELOCKED, p->p_ucred, &resid);
if (error)
goto out;
if (resid != 0) {
error = EINVAL;
goto out;
}
ra++;
break;
}
}
if (pb != ef->nprogtab)
panic("lost progbits");
if (rl != ef->nreltab)
panic("lost reltab");
if (ra != ef->nrelatab)
panic("lost relatab");
if (mapbase != (vm_offset_t) ef->address + mapsize)
panic("mapbase 0x%lx != address %p + mapsize 0x%lx (0x%lx)",
mapbase, ef->address, mapsize,
(vm_offset_t) ef->address + mapsize);
/* Local intra-module relocations */
link_elf_obj_reloc_local(lf);
/* Pull in dependencies */
error = linker_load_dependencies(lf);
if (error)
goto out;
/* External relocations */
error = relocate_file(lf);
if (error)
goto out;
*result = lf;
out:
if (error && lf)
linker_file_unload(lf /*, LINKER_UNLOAD_FORCE */);
if (hdr)
kfree(hdr, M_LINKER);
vn_unlock(vp);
vn_close(vp, FREAD, NULL);
return error;
}
static int
ntfs_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
{
size_t size;
int error;
struct vnode *devvp;
struct ntfs_args args;
struct nlookupdata nd;
struct vnode *rootvp;
error = 0;
/*
* Use NULL path to flag a root mount
*/
if( path == NULL) {
/*
***
* Mounting root file system
***
*/
/* Get vnode for root device*/
if( bdevvp( rootdev, &rootvp))
panic("ffs_mountroot: can't setup bdevvp for root");
/*
* FS specific handling
*/
mp->mnt_flag |= MNT_RDONLY; /* XXX globally applicable?*/
/*
* Attempt mount
*/
if( ( error = ntfs_mountfs(rootvp, mp, &args, cred)) != 0) {
/* fs specific cleanup (if any)*/
goto error_1;
}
goto dostatfs; /* success*/
}
/*
***
* Mounting non-root file system or updating a file system
***
*/
/* copy in user arguments*/
error = copyin(data, (caddr_t)&args, sizeof (struct ntfs_args));
if (error)
goto error_1; /* can't get arguments*/
/*
* If updating, check whether changing from read-only to
* read/write; if there is no device name, that's all we do.
*/
if (mp->mnt_flag & MNT_UPDATE) {
/* if not updating name...*/
if (args.fspec == NULL) {
/*
* Process export requests. Jumping to "success"
* will return the vfs_export() error code.
*/
struct ntfsmount *ntm = VFSTONTFS(mp);
error = vfs_export(mp, &ntm->ntm_export, &args.export);
goto success;
}
kprintf("ntfs_mount(): MNT_UPDATE not supported\n");
error = EINVAL;
goto error_1;
}
/*
* Not an update, or updating the name: look up the name
* and verify that it refers to a sensible block device.
*/
devvp = NULL;
error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
if (error == 0)
error = nlookup(&nd);
if (error == 0)
error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
nlookup_done(&nd);
if (error)
goto error_1;
if (!vn_isdisk(devvp, &error))
goto error_2;
if (mp->mnt_flag & MNT_UPDATE) {
#if 0
/*
********************
* UPDATE
********************
*/
if (devvp != ntmp->um_devvp)
error = EINVAL; /* needs translation */
else
vrele(devvp);
/*
* Update device name only on success
*/
if( !error) {
/* Save "mounted from" info for mount point (NULL pad)*/
copyinstr( args.fspec,
mp->mnt_stat.f_mntfromname,
MNAMELEN - 1,
&size);
bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
}
#endif
} else {
/*
********************
* NEW MOUNT
********************
*/
/* Save "mounted from" info for mount point (NULL pad)*/
copyinstr( args.fspec, /* device name*/
mp->mnt_stat.f_mntfromname, /* save area*/
MNAMELEN - 1, /* max size*/
&size); /* real size*/
bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
error = ntfs_mountfs(devvp, mp, &args, cred);
}
if (error) {
goto error_2;
}
dostatfs:
/*
* Initialize FS stat information in mount struct; uses
* mp->mnt_stat.f_mntfromname.
*
* This code is common to root and non-root mounts
*/
(void)VFS_STATFS(mp, &mp->mnt_stat, cred);
goto success;
error_2: /* error with devvp held*/
/* release devvp before failing*/
vrele(devvp);
error_1: /* no state to back out*/
success:
return(error);
}
int
ufs_quotaon(struct ucred *cred, struct mount *mp, int type, caddr_t fname)
{
struct ufsmount *ump = VFSTOUFS(mp);
struct vnode *vp, **vpp;
struct ufs_dquot *dq;
int error;
struct nlookupdata nd;
struct scaninfo scaninfo;
vpp = &ump->um_quotas[type];
error = nlookup_init(&nd, fname, UIO_USERSPACE, NLC_FOLLOW|NLC_LOCKVP);
if (error == 0)
error = vn_open(&nd, NULL, FREAD|FWRITE, 0);
if (error == 0 && nd.nl_open_vp->v_type != VREG)
error = EACCES;
if (error) {
nlookup_done(&nd);
return (error);
}
vp = nd.nl_open_vp;
nd.nl_open_vp = NULL;
nlookup_done(&nd);
vn_unlock(vp);
if (*vpp != vp)
ufs_quotaoff(mp, type);
ump->um_qflags[type] |= QTF_OPENING;
mp->mnt_flag |= MNT_QUOTA;
vsetflags(vp, VSYSTEM);
*vpp = vp;
/* XXX release duplicate vp if *vpp == vp? */
/*
* Save the credential of the process that turned on quotas.
* Set up the time limits for this quota.
*/
ump->um_cred[type] = crhold(cred);
ump->um_btime[type] = MAX_DQ_TIME;
ump->um_itime[type] = MAX_IQ_TIME;
if (ufs_dqget(NULLVP, 0, ump, type, &dq) == 0) {
if (dq->dq_btime > 0)
ump->um_btime[type] = dq->dq_btime;
if (dq->dq_itime > 0)
ump->um_itime[type] = dq->dq_itime;
ufs_dqrele(NULLVP, dq);
}
/*
* Search vnodes associated with this mount point,
* adding references to quota file being opened.
* NB: only need to add dquot's for inodes being modified.
*/
scaninfo.rescan = 1;
while (scaninfo.rescan) {
scaninfo.rescan = 0;
error = vmntvnodescan(mp, VMSC_GETVP,
NULL, ufs_quotaon_scan, &scaninfo);
if (error)
break;
}
ump->um_qflags[type] &= ~QTF_OPENING;
if (error)
ufs_quotaoff(mp, type);
return (error);
}
/************************************************************************
* VOLUMES *
************************************************************************
*
* Load a HAMMER volume by name. Returns 0 on success or a positive error
* code on failure. Volumes must be loaded at mount time, get_volume() will
* not load a new volume.
*
* The passed devvp is vref()'d but not locked. This function consumes the
* ref (typically by associating it with the volume structure).
*
* Calls made to hammer_load_volume() or single-threaded
*/
int
hammer_install_volume(struct hammer_mount *hmp, const char *volname,
struct vnode *devvp)
{
struct mount *mp;
hammer_volume_t volume;
struct hammer_volume_ondisk *ondisk;
struct nlookupdata nd;
struct buf *bp = NULL;
int error;
int ronly;
int setmp = 0;
mp = hmp->mp;
ronly = ((mp->mnt_flag & MNT_RDONLY) ? 1 : 0);
/*
* Allocate a volume structure
*/
++hammer_count_volumes;
volume = kmalloc(sizeof(*volume), hmp->m_misc, M_WAITOK|M_ZERO);
volume->vol_name = kstrdup(volname, hmp->m_misc);
volume->io.hmp = hmp; /* bootstrap */
hammer_io_init(&volume->io, volume, HAMMER_STRUCTURE_VOLUME);
volume->io.offset = 0LL;
volume->io.bytes = HAMMER_BUFSIZE;
/*
* Get the device vnode
*/
if (devvp == NULL) {
error = nlookup_init(&nd, volume->vol_name, UIO_SYSSPACE, NLC_FOLLOW);
if (error == 0)
error = nlookup(&nd);
if (error == 0)
error = cache_vref(&nd.nl_nch, nd.nl_cred, &volume->devvp);
nlookup_done(&nd);
} else {
error = 0;
volume->devvp = devvp;
}
if (error == 0) {
if (vn_isdisk(volume->devvp, &error)) {
error = vfs_mountedon(volume->devvp);
}
}
if (error == 0 && vcount(volume->devvp) > 0)
error = EBUSY;
if (error == 0) {
vn_lock(volume->devvp, LK_EXCLUSIVE | LK_RETRY);
error = vinvalbuf(volume->devvp, V_SAVE, 0, 0);
if (error == 0) {
error = VOP_OPEN(volume->devvp,
(ronly ? FREAD : FREAD|FWRITE),
FSCRED, NULL);
}
vn_unlock(volume->devvp);
}
if (error) {
hammer_free_volume(volume);
return(error);
}
volume->devvp->v_rdev->si_mountpoint = mp;
setmp = 1;
/*
* Extract the volume number from the volume header and do various
* sanity checks.
*/
error = bread(volume->devvp, 0LL, HAMMER_BUFSIZE, &bp);
if (error)
goto late_failure;
ondisk = (void *)bp->b_data;
if (ondisk->vol_signature != HAMMER_FSBUF_VOLUME) {
kprintf("hammer_mount: volume %s has an invalid header\n",
volume->vol_name);
error = EFTYPE;
goto late_failure;
}
volume->vol_no = ondisk->vol_no;
volume->buffer_base = ondisk->vol_buf_beg;
volume->vol_flags = ondisk->vol_flags;
volume->nblocks = ondisk->vol_nblocks;
volume->maxbuf_off = HAMMER_ENCODE_RAW_BUFFER(volume->vol_no,
ondisk->vol_buf_end - ondisk->vol_buf_beg);
volume->maxraw_off = ondisk->vol_buf_end;
if (RB_EMPTY(&hmp->rb_vols_root)) {
hmp->fsid = ondisk->vol_fsid;
} else if (bcmp(&hmp->fsid, &ondisk->vol_fsid, sizeof(uuid_t))) {
kprintf("hammer_mount: volume %s's fsid does not match "
"other volumes\n", volume->vol_name);
error = EFTYPE;
goto late_failure;
}
/*
* Insert the volume structure into the red-black tree.
*/
if (RB_INSERT(hammer_vol_rb_tree, &hmp->rb_vols_root, volume)) {
kprintf("hammer_mount: volume %s has a duplicate vol_no %d\n",
volume->vol_name, volume->vol_no);
error = EEXIST;
}
/*
* Set the root volume . HAMMER special cases rootvol the structure.
* We do not hold a ref because this would prevent related I/O
* from being flushed.
*/
if (error == 0 && ondisk->vol_rootvol == ondisk->vol_no) {
hmp->rootvol = volume;
hmp->nvolumes = ondisk->vol_count;
if (bp) {
brelse(bp);
bp = NULL;
}
hmp->mp->mnt_stat.f_blocks += ondisk->vol0_stat_bigblocks *
(HAMMER_LARGEBLOCK_SIZE / HAMMER_BUFSIZE);
hmp->mp->mnt_vstat.f_blocks += ondisk->vol0_stat_bigblocks *
(HAMMER_LARGEBLOCK_SIZE / HAMMER_BUFSIZE);
}
late_failure:
if (bp)
brelse(bp);
if (error) {
/*vinvalbuf(volume->devvp, V_SAVE, 0, 0);*/
if (setmp)
volume->devvp->v_rdev->si_mountpoint = NULL;
vn_lock(volume->devvp, LK_EXCLUSIVE | LK_RETRY);
VOP_CLOSE(volume->devvp, ronly ? FREAD : FREAD|FWRITE);
vn_unlock(volume->devvp);
hammer_free_volume(volume);
}
return (error);
}
/*
* XXX: Should we process mount export info ?
* If not, returning zero here is enough as the actual ro/rw update is
* being done in sys_mount().
*/
if (mp->mnt_flag & MNT_UPDATE) {
xmp = MOUNTTONULLMOUNT(mp);
error = vfs_export(mp, &xmp->export, &args.export);
return (error);
}
/*
* Find lower node
*/
rootvp = NULL;
error = nlookup_init(&nd, args.target, UIO_USERSPACE, NLC_FOLLOW);
if (error)
goto fail1;
error = nlookup(&nd);
if (error)
goto fail2;
error = cache_vget(&nd.nl_nch, nd.nl_cred, LK_EXCLUSIVE, &rootvp);
if (error)
goto fail2;
xmp = (struct null_mount *) kmalloc(sizeof(struct null_mount),
M_NULLFSMNT, M_WAITOK | M_ZERO);
/*
* Save reference to underlying FS
*