/*
* util_file_map_whole -- maps the entire file into memory
*/
void *
util_file_map_whole(const char *path)
{
LOG(3, "path \"%s\"", path);
int fd;
int olderrno;
void *addr = NULL;
if ((fd = os_open(path, O_RDWR)) < 0) {
ERR("!open \"%s\"", path);
return NULL;
}
ssize_t size = util_file_get_size(path);
if (size < 0) {
LOG(2, "cannot determine file length \"%s\"", path);
goto out;
}
addr = util_map(fd, (size_t)size, MAP_SHARED, 0, 0, NULL);
if (addr == NULL) {
LOG(2, "failed to map entire file \"%s\"", path);
goto out;
}
out:
olderrno = errno;
(void) os_close(fd);
errno = olderrno;
return addr;
}
/*
* util_file_open -- open a memory pool file
*/
int
util_file_open(const char *path, size_t *size, size_t minsize, int flags)
{
LOG(3, "path \"%s\" size %p minsize %zu flags %d", path, size, minsize,
flags);
int oerrno;
int fd;
#ifdef _WIN32
flags |= O_BINARY;
#endif
if ((fd = os_open(path, flags)) < 0) {
ERR("!open \"%s\"", path);
return -1;
}
if (os_flock(fd, OS_LOCK_EX | OS_LOCK_NB) < 0) {
ERR("!flock \"%s\"", path);
(void) os_close(fd);
return -1;
}
if (size || minsize) {
if (size)
ASSERTeq(*size, 0);
ssize_t actual_size = util_file_get_size(path);
if (actual_size < 0) {
ERR("stat \"%s\": negative size", path);
errno = EINVAL;
goto err;
}
if ((size_t)actual_size < minsize) {
ERR("size %zu smaller than %zu",
(size_t)actual_size, minsize);
errno = EINVAL;
goto err;
}
if (size) {
*size = (size_t)actual_size;
LOG(4, "actual file size %zu", *size);
}
}
return fd;
err:
oerrno = errno;
if (os_flock(fd, OS_LOCK_UN))
ERR("!flock unlock");
(void) os_close(fd);
errno = oerrno;
return -1;
}
/*
* util_file_zero -- zeroes the specified region of the file
*/
int
util_file_zero(const char *path, os_off_t off, size_t len)
{
LOG(3, "path \"%s\" off %ju len %zu", path, off, len);
int fd;
int olderrno;
int ret = 0;
if ((fd = os_open(path, O_RDWR)) < 0) {
ERR("!open \"%s\"", path);
return -1;
}
ssize_t size = util_file_get_size(path);
if (size < 0) {
LOG(2, "cannot determine file length \"%s\"", path);
ret = -1;
goto out;
}
if (off > size) {
LOG(2, "offset beyond file length, %ju > %ju", off, size);
ret = -1;
goto out;
}
if ((size_t)off + len > (size_t)size) {
LOG(2, "requested size of write goes beyond the file length, "
"%zu > %zu", (size_t)off + len, size);
LOG(4, "adjusting len to %zu", size - off);
len = (size_t)(size - off);
}
void *addr = util_map(fd, (size_t)size, MAP_SHARED, 0, 0, NULL);
if (addr == NULL) {
LOG(2, "failed to map entire file \"%s\"", path);
ret = -1;
goto out;
}
/* zero initialize the specified region */
memset((char *)addr + off, 0, len);
util_unmap(addr, (size_t)size);
out:
olderrno = errno;
(void) os_close(fd);
errno = olderrno;
return ret;
}
/*
* util_file_pread -- reads from a file with an offset
*/
ssize_t
util_file_pread(const char *path, void *buffer, size_t size,
os_off_t offset)
{
LOG(3, "path \"%s\" buffer %p size %zu offset %ju",
path, buffer, size, offset);
enum file_type type = util_file_get_type(path);
if (type < 0)
return -1;
if (type == TYPE_NORMAL) {
int fd = util_file_open(path, NULL, 0, O_RDONLY);
if (fd < 0) {
LOG(2, "failed to open file \"%s\"", path);
return -1;
}
ssize_t read_len = pread(fd, buffer, size, offset);
int olderrno = errno;
(void) os_close(fd);
errno = olderrno;
return read_len;
}
ssize_t file_size = util_file_get_size(path);
if (file_size < 0) {
LOG(2, "cannot determine file length \"%s\"", path);
return -1;
}
size_t max_size = (size_t)(file_size - offset);
if (size > max_size) {
LOG(2, "requested size of read goes beyond the file length, "
"%zu > %zu", size, max_size);
LOG(4, "adjusting size to %zu", max_size);
size = max_size;
}
void *addr = util_file_map_whole(path);
if (addr == NULL) {
LOG(2, "failed to map entire file \"%s\"", path);
return -1;
}
memcpy(buffer, ADDR_SUM(addr, offset), size);
util_unmap(addr, (size_t)file_size);
return (ssize_t)size;
}
/*
* pool_parse_params -- parse pool type, file size and block size
*/
static int
pool_params_parse(const PMEMpoolcheck *ppc, struct pool_params *params,
int check)
{
LOG(3, NULL);
int is_btt = ppc->args.pool_type == PMEMPOOL_POOL_TYPE_BTT;
params->type = POOL_TYPE_UNKNOWN;
params->is_poolset = util_is_poolset_file(ppc->path) == 1;
int fd = util_file_open(ppc->path, NULL, 0, O_RDONLY);
if (fd < 0)
return -1;
int ret = 0;
util_stat_t stat_buf;
ret = util_fstat(fd, &stat_buf);
if (ret)
goto out_close;
ASSERT(stat_buf.st_size >= 0);
params->mode = stat_buf.st_mode;
struct pool_set *set;
void *addr;
if (params->is_poolset) {
/*
* Need to close the poolset because it will be opened with
* flock in the following instructions.
*/
close(fd);
fd = -1;
if (check) {
if (pool_set_map(ppc->path, &set, 0))
return -1;
} else {
ret = util_poolset_create_set(&set, ppc->path, 0, 0);
if (ret < 0) {
LOG(2, "cannot open pool set -- '%s'",
ppc->path);
return -1;
}
if (set->remote) {
ERR("poolsets with remote replicas are not "
"supported");
return -1;
}
if (util_pool_open_nocheck(set, 0))
return -1;
}
params->size = set->poolsize;
addr = set->replica[0]->part[0].addr;
/*
* XXX mprotect for device dax with length not aligned to its
* page granularity causes SIGBUS on the next page fault.
* The length argument of this call should be changed to
* set->poolsize once the kernel issue is solved.
*/
if (mprotect(addr, set->replica[0]->repsize,
PROT_READ) < 0) {
ERR("!mprotect");
goto out_unmap;
}
params->is_device_dax = set->replica[0]->part[0].is_dax;
} else if (is_btt) {
params->size = (size_t)stat_buf.st_size;
#ifndef _WIN32
if (params->mode & S_IFBLK)
if (ioctl(fd, BLKGETSIZE64, ¶ms->size)) {
ERR("!ioctl");
goto out_close;
}
#endif
addr = NULL;
} else {
ssize_t s = util_file_get_size(ppc->path);
if (s < 0) {
ret = -1;
goto out_close;
}
params->size = (size_t)s;
addr = mmap(NULL, (uint64_t)params->size, PROT_READ,
MAP_PRIVATE, fd, 0);
if (addr == MAP_FAILED) {
ret = -1;
goto out_close;
}
params->is_device_dax = util_file_is_device_dax(ppc->path);
}
/* stop processing for BTT device */
if (is_btt) {
params->type = POOL_TYPE_BTT;
params->is_part = false;
goto out_close;
}
struct pool_hdr hdr;
memcpy(&hdr, addr, sizeof(hdr));
util_convert2h_hdr_nocheck(&hdr);
pool_params_from_header(params, &hdr);
if (ppc->args.pool_type != PMEMPOOL_POOL_TYPE_DETECT) {
enum pool_type declared_type =
pmempool_check_type_to_pool_type(ppc->args.pool_type);
if ((params->type & ~declared_type) != 0) {
ERR("declared pool type does not match");
ret = 1;
goto out_unmap;
}
}
if (params->type == POOL_TYPE_BLK) {
struct pmemblk pbp;
memcpy(&pbp, addr, sizeof(pbp));
params->blk.bsize = le32toh(pbp.bsize);
} else if (params->type == POOL_TYPE_OBJ) {
struct pmemobjpool pop;
memcpy(&pop, addr, sizeof(pop));
memcpy(params->obj.layout, pop.layout,
PMEMOBJ_MAX_LAYOUT);
}
out_unmap:
if (params->is_poolset) {
ASSERTeq(fd, -1);
ASSERTne(addr, NULL);
util_poolset_close(set, 0);
} else if (!is_btt) {
ASSERTne(fd, -1);
ASSERTne(addr, NULL);
munmap(addr, params->size);
}
out_close:
if (fd != -1)
close(fd);
return ret;
}
/*
* backup_poolset_requirements -- (internal) check backup requirements
*/
static int
backup_poolset_requirements(PMEMpoolcheck *ppc, location *loc)
{
LOG(3, "backup_path %s", ppc->backup_path);
if (ppc->pool->set_file->poolset->nreplicas > 1) {
CHECK_INFO(ppc,
"backup of a poolset with multiple replicas is not supported");
goto err;
}
if (pool_set_parse(&loc->set, ppc->backup_path)) {
CHECK_INFO(ppc, "invalid poolset backup file: %s",
ppc->backup_path);
goto err;
}
if (loc->set->nreplicas > 1) {
CHECK_INFO(ppc,
"backup to a poolset with multiple replicas is not supported");
goto err_poolset;
}
ASSERTeq(loc->set->nreplicas, 1);
struct pool_replica *srep = ppc->pool->set_file->poolset->replica[0];
struct pool_replica *drep = loc->set->replica[0];
if (srep->nparts != drep->nparts) {
CHECK_INFO(ppc,
"number of part files in the backup poolset must match number of part files in the source poolset");
goto err_poolset;
}
int overwrite_required = 0;
for (unsigned p = 0; p < srep->nparts; p++) {
if (srep->part[p].filesize != drep->part[p].filesize) {
CHECK_INFO(ppc,
"size of the part %u of the backup poolset does not match source poolset",
p);
goto err_poolset;
}
if (os_access(drep->part[p].path, F_OK)) {
if (errno == ENOENT) {
errno = 0;
continue;
} else {
CHECK_INFO(ppc,
"unable to access the part of the destination poolset: %s",
ppc->backup_path);
goto err_poolset;
}
}
overwrite_required = true;
if ((size_t)util_file_get_size(drep->part[p].path) !=
srep->part[p].filesize) {
CHECK_INFO(ppc,
"destination of the backup part does not match size of the source part file: %s",
drep->part[p].path);
goto err_poolset;
}
}
if (CHECK_WITHOUT_FIXING(ppc)) {
location_release(loc);
loc->step = CHECK_STEP_COMPLETE;
return 0;
}
if (overwrite_required) {
CHECK_ASK(ppc, Q_OVERWRITE_EXISTING_PARTS,
"part files of the destination poolset of the backup already exist.|"
"Do you want to overwrite them?");
}
return check_questions_sequence_validate(ppc);
err_poolset:
location_release(loc);
err:
ppc->result = CHECK_RESULT_ERROR;
return CHECK_ERR(ppc, "unable to backup poolset");
}
static inline
#endif
void *
pmem_map_fileU(const char *path, size_t len, int flags,
mode_t mode, size_t *mapped_lenp, int *is_pmemp)
{
LOG(3, "path \"%s\" size %zu flags %x mode %o mapped_lenp %p "
"is_pmemp %p", path, len, flags, mode, mapped_lenp, is_pmemp);
int oerrno;
int fd;
int open_flags = O_RDWR;
int delete_on_err = 0;
int file_type = util_file_get_type(path);
#ifdef _WIN32
open_flags |= O_BINARY;
#endif
if (file_type == OTHER_ERROR)
return NULL;
if (flags & ~(PMEM_FILE_ALL_FLAGS)) {
ERR("invalid flag specified %x", flags);
errno = EINVAL;
return NULL;
}
if (file_type == TYPE_DEVDAX) {
if (flags & ~(PMEM_DAX_VALID_FLAGS)) {
ERR("flag unsupported for Device DAX %x", flags);
errno = EINVAL;
return NULL;
} else {
/* we are ignoring all of the flags */
flags = 0;
ssize_t actual_len = util_file_get_size(path);
if (actual_len < 0) {
ERR("unable to read Device DAX size");
errno = EINVAL;
return NULL;
}
if (len != 0 && len != (size_t)actual_len) {
ERR("Device DAX length must be either 0 or "
"the exact size of the device: %zu",
actual_len);
errno = EINVAL;
return NULL;
}
len = 0;
}
}
if (flags & PMEM_FILE_CREATE) {
if ((os_off_t)len < 0) {
ERR("invalid file length %zu", len);
errno = EINVAL;
return NULL;
}
open_flags |= O_CREAT;
}
if (flags & PMEM_FILE_EXCL)
open_flags |= O_EXCL;
if ((len != 0) && !(flags & PMEM_FILE_CREATE)) {
ERR("non-zero 'len' not allowed without PMEM_FILE_CREATE");
errno = EINVAL;
return NULL;
}
if ((len == 0) && (flags & PMEM_FILE_CREATE)) {
ERR("zero 'len' not allowed with PMEM_FILE_CREATE");
errno = EINVAL;
return NULL;
}
if ((flags & PMEM_FILE_TMPFILE) && !(flags & PMEM_FILE_CREATE)) {
ERR("PMEM_FILE_TMPFILE not allowed without PMEM_FILE_CREATE");
errno = EINVAL;
return NULL;
}
if (flags & PMEM_FILE_TMPFILE) {
if ((fd = util_tmpfile(path,
OS_DIR_SEP_STR"pmem.XXXXXX",
open_flags & O_EXCL)) < 0) {
LOG(2, "failed to create temporary file at \"%s\"",
path);
return NULL;
}
} else {
if ((fd = os_open(path, open_flags, mode)) < 0) {
ERR("!open %s", path);
return NULL;
}
if ((flags & PMEM_FILE_CREATE) && (flags & PMEM_FILE_EXCL))
delete_on_err = 1;
}
if (flags & PMEM_FILE_CREATE) {
/*
* Always set length of file to 'len'.
* (May either extend or truncate existing file.)
*/
if (os_ftruncate(fd, (os_off_t)len) != 0) {
ERR("!ftruncate");
goto err;
}
if ((flags & PMEM_FILE_SPARSE) == 0) {
if ((errno = os_posix_fallocate(fd, 0,
(os_off_t)len)) != 0) {
ERR("!posix_fallocate");
goto err;
}
}
} else {
ssize_t actual_size = util_file_get_size(path);
if (actual_size < 0) {
ERR("stat %s: negative size", path);
errno = EINVAL;
goto err;
}
len = (size_t)actual_size;
}
void *addr = pmem_map_register(fd, len, path, file_type == TYPE_DEVDAX);
if (addr == NULL)
goto err;
if (mapped_lenp != NULL)
*mapped_lenp = len;
if (is_pmemp != NULL)
*is_pmemp = pmem_is_pmem(addr, len);
LOG(3, "returning %p", addr);
VALGRIND_REGISTER_PMEM_MAPPING(addr, len);
VALGRIND_REGISTER_PMEM_FILE(fd, addr, len, 0);
(void) os_close(fd);
return addr;
err:
oerrno = errno;
(void) os_close(fd);
if (delete_on_err)
(void) os_unlink(path);
errno = oerrno;
return NULL;
}
