/*
* Remaps all hugepages into single file segments
*/
static int
remap_all_hugepages(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi)
{
int fd;
unsigned i = 0, j, num_pages, page_idx = 0;
void *vma_addr = NULL, *old_addr = NULL, *page_addr = NULL;
size_t vma_len = 0;
size_t hugepage_sz = hpi->hugepage_sz;
size_t total_size, offset;
char filepath[MAX_HUGEPAGE_PATH];
phys_addr_t physaddr;
int socket;
while (i < hpi->num_pages[0]) {
#ifndef RTE_ARCH_64
/* for 32-bit systems, don't remap 1G pages and 16G pages,
* just reuse original map address as final map address.
*/
if ((hugepage_sz == RTE_PGSIZE_1G)
|| (hugepage_sz == RTE_PGSIZE_16G)) {
hugepg_tbl[i].final_va = hugepg_tbl[i].orig_va;
hugepg_tbl[i].orig_va = NULL;
i++;
continue;
}
#endif
/* reserve a virtual area for next contiguous
* physical block: count the number of
* contiguous physical pages. */
for (j = i+1; j < hpi->num_pages[0] ; j++) {
#ifdef RTE_ARCH_PPC_64
/* The physical addresses are sorted in descending
* order on PPC64 */
if (hugepg_tbl[j].physaddr !=
hugepg_tbl[j-1].physaddr - hugepage_sz)
break;
#else
if (hugepg_tbl[j].physaddr !=
hugepg_tbl[j-1].physaddr + hugepage_sz)
break;
#endif
}
num_pages = j - i;
vma_len = num_pages * hugepage_sz;
socket = hugepg_tbl[i].socket_id;
/* get the biggest virtual memory area up to
* vma_len. If it fails, vma_addr is NULL, so
* let the kernel provide the address. */
vma_addr = get_virtual_area(&vma_len, hpi->hugepage_sz);
/* If we can't find a big enough virtual area, work out how many pages
* we are going to get */
if (vma_addr == NULL)
j = i + 1;
else if (vma_len != num_pages * hugepage_sz) {
num_pages = vma_len / hugepage_sz;
j = i + num_pages;
}
hugepg_tbl[page_idx].file_id = page_idx;
eal_get_hugefile_path(filepath,
sizeof(filepath),
hpi->hugedir,
hugepg_tbl[page_idx].file_id);
/* try to create hugepage file */
fd = open(filepath, O_CREAT | O_RDWR, 0755);
if (fd < 0) {
RTE_LOG(ERR, EAL, "%s(): open failed: %s\n", __func__, strerror(errno));
return -1;
}
total_size = 0;
for (;i < j; i++) {
/* unmap current segment */
if (total_size > 0)
munmap(vma_addr, total_size);
/* unmap original page */
munmap(hugepg_tbl[i].orig_va, hugepage_sz);
unlink(hugepg_tbl[i].filepath);
total_size += hugepage_sz;
old_addr = vma_addr;
/* map new, bigger segment */
vma_addr = mmap(vma_addr, total_size,
PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
if (vma_addr == MAP_FAILED || vma_addr != old_addr) {
RTE_LOG(ERR, EAL, "%s(): mmap failed: %s\n", __func__, strerror(errno));
close(fd);
return -1;
}
/* touch the page. this is needed because kernel postpones mapping
* creation until the first page fault. with this, we pin down
* the page and it is marked as used and gets into process' pagemap.
*/
for (offset = 0; offset < total_size; offset += hugepage_sz)
*((volatile uint8_t*) RTE_PTR_ADD(vma_addr, offset));
}
/* set shared flock on the file. */
if (flock(fd, LOCK_SH | LOCK_NB) == -1) {
RTE_LOG(ERR, EAL, "%s(): Locking file failed:%s \n",
__func__, strerror(errno));
close(fd);
return -1;
}
snprintf(hugepg_tbl[page_idx].filepath, MAX_HUGEPAGE_PATH, "%s",
filepath);
physaddr = rte_mem_virt2phy(vma_addr);
if (physaddr == RTE_BAD_PHYS_ADDR)
return -1;
hugepg_tbl[page_idx].final_va = vma_addr;
hugepg_tbl[page_idx].physaddr = physaddr;
hugepg_tbl[page_idx].repeated = num_pages;
hugepg_tbl[page_idx].socket_id = socket;
close(fd);
/* verify the memory segment - that is, check that every VA corresponds
* to the physical address we expect to see
*/
for (offset = 0; offset < vma_len; offset += hugepage_sz) {
uint64_t expected_physaddr;
expected_physaddr = hugepg_tbl[page_idx].physaddr + offset;
page_addr = RTE_PTR_ADD(vma_addr, offset);
physaddr = rte_mem_virt2phy(page_addr);
if (physaddr != expected_physaddr) {
RTE_LOG(ERR, EAL, "Segment sanity check failed: wrong physaddr "
"at %p (offset 0x%" PRIx64 ": 0x%" PRIx64
" (expected 0x%" PRIx64 ")\n",
page_addr, offset, physaddr, expected_physaddr);
return -1;
}
}
/* zero out the whole segment */
memset(hugepg_tbl[page_idx].final_va, 0, total_size);
page_idx++;
}
/* zero out the rest */
memset(&hugepg_tbl[page_idx], 0, (hpi->num_pages[0] - page_idx) * sizeof(struct hugepage_file));
return page_idx;
}
/*
* Mmap all hugepages of hugepage table: it first open a file in
* hugetlbfs, then mmap() hugepage_sz data in it. If orig is set, the
* virtual address is stored in hugepg_tbl[i].orig_va, else it is stored
* in hugepg_tbl[i].final_va. The second mapping (when orig is 0) tries to
* map continguous physical blocks in contiguous virtual blocks.
*/
static int
map_all_hugepages(struct hugepage *hugepg_tbl,
struct hugepage_info *hpi, int orig)
{
int fd;
unsigned i;
void *virtaddr;
void *vma_addr = NULL;
uint64_t vma_len = 0;
for (i = 0; i < hpi->num_pages[0]; i++) {
uint64_t hugepage_sz = hpi->hugepage_sz;
if (orig) {
hugepg_tbl[i].file_id = i;
hugepg_tbl[i].size = hugepage_sz;
eal_get_hugefile_path(hugepg_tbl[i].filepath,
sizeof(hugepg_tbl[i].filepath), hpi->hugedir,
hugepg_tbl[i].file_id);
hugepg_tbl[i].filepath[sizeof(hugepg_tbl[i].filepath) - 1] = '\0';
}
#ifndef RTE_ARCH_X86_64
/* for 32-bit systems, don't remap 1G pages, just reuse original
* map address as final map address.
*/
else if (hugepage_sz == RTE_PGSIZE_1G){
hugepg_tbl[i].final_va = hugepg_tbl[i].orig_va;
hugepg_tbl[i].orig_va = NULL;
continue;
}
#endif
else if (vma_len == 0) {
unsigned j, num_pages;
/* reserve a virtual area for next contiguous
* physical block: count the number of
* contiguous physical pages. */
for (j = i+1; j < hpi->num_pages[0] ; j++) {
if (hugepg_tbl[j].physaddr !=
hugepg_tbl[j-1].physaddr + hugepage_sz)
break;
}
num_pages = j - i;
vma_len = num_pages * hugepage_sz;
/* get the biggest virtual memory area up to
* vma_len. If it fails, vma_addr is NULL, so
* let the kernel provide the address. */
vma_addr = get_virtual_area(&vma_len, hpi->hugepage_sz);
if (vma_addr == NULL)
vma_len = hugepage_sz;
}
/* try to create hugepage file */
fd = open(hugepg_tbl[i].filepath, O_CREAT | O_RDWR, 0755);
if (fd < 0) {
RTE_LOG(ERR, EAL, "%s(): open failed: %s\n", __func__,
strerror(errno));
return -1;
}
virtaddr = mmap(vma_addr, hugepage_sz, PROT_READ | PROT_WRITE,
MAP_SHARED, fd, 0);
if (virtaddr == MAP_FAILED) {
RTE_LOG(ERR, EAL, "%s(): mmap failed: %s\n", __func__,
strerror(errno));
close(fd);
return -1;
}
if (orig) {
hugepg_tbl[i].orig_va = virtaddr;
memset(virtaddr, 0, hugepage_sz);
}
else {
hugepg_tbl[i].final_va = virtaddr;
}
/* close the file descriptor, files will be locked later */
close(fd);
vma_addr = (char *)vma_addr + hugepage_sz;
vma_len -= hugepage_sz;
}
return 0;
}
/*
* Mmap all hugepages of hugepage table: it first open a file in
* hugetlbfs, then mmap() hugepage_sz data in it. If orig is set, the
* virtual address is stored in hugepg_tbl[i].orig_va, else it is stored
* in hugepg_tbl[i].final_va. The second mapping (when orig is 0) tries to
* map continguous physical blocks in contiguous virtual blocks.
*/
static int
map_all_hugepages(struct hugepage_file *hugepg_tbl,
struct hugepage_info *hpi, int orig)
{
int fd;
unsigned i;
void *virtaddr;
void *vma_addr = NULL;
size_t vma_len = 0;
#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
RTE_SET_USED(vma_len);
#endif
for (i = 0; i < hpi->num_pages[0]; i++) {
uint64_t hugepage_sz = hpi->hugepage_sz;
if (orig) {
hugepg_tbl[i].file_id = i;
hugepg_tbl[i].size = hugepage_sz;
#ifdef RTE_EAL_SINGLE_FILE_SEGMENTS
eal_get_hugefile_temp_path(hugepg_tbl[i].filepath,
sizeof(hugepg_tbl[i].filepath), hpi->hugedir,
hugepg_tbl[i].file_id);
#else
eal_get_hugefile_path(hugepg_tbl[i].filepath,
sizeof(hugepg_tbl[i].filepath), hpi->hugedir,
hugepg_tbl[i].file_id);
#endif
hugepg_tbl[i].filepath[sizeof(hugepg_tbl[i].filepath) - 1] = '\0';
}
#ifndef RTE_ARCH_64
/* for 32-bit systems, don't remap 1G and 16G pages, just reuse
* original map address as final map address.
*/
else if ((hugepage_sz == RTE_PGSIZE_1G)
|| (hugepage_sz == RTE_PGSIZE_16G)) {
hugepg_tbl[i].final_va = hugepg_tbl[i].orig_va;
hugepg_tbl[i].orig_va = NULL;
continue;
}
#endif
#ifndef RTE_EAL_SINGLE_FILE_SEGMENTS
else if (vma_len == 0) {
unsigned j, num_pages;
/* reserve a virtual area for next contiguous
* physical block: count the number of
* contiguous physical pages. */
for (j = i+1; j < hpi->num_pages[0] ; j++) {
#ifdef RTE_ARCH_PPC_64
/* The physical addresses are sorted in
* descending order on PPC64 */
if (hugepg_tbl[j].physaddr !=
hugepg_tbl[j-1].physaddr - hugepage_sz)
break;
#else
if (hugepg_tbl[j].physaddr !=
hugepg_tbl[j-1].physaddr + hugepage_sz)
break;
#endif
}
num_pages = j - i;
vma_len = num_pages * hugepage_sz;
/* get the biggest virtual memory area up to
* vma_len. If it fails, vma_addr is NULL, so
* let the kernel provide the address. */
vma_addr = get_virtual_area(&vma_len, hpi->hugepage_sz);
if (vma_addr == NULL)
vma_len = hugepage_sz;
}
#endif
/* try to create hugepage file */
fd = open(hugepg_tbl[i].filepath, O_CREAT | O_RDWR, 0755);
if (fd < 0) {
RTE_LOG(ERR, EAL, "%s(): open failed: %s\n", __func__,
strerror(errno));
return -1;
}
virtaddr = mmap(vma_addr, hugepage_sz, PROT_READ | PROT_WRITE,
MAP_SHARED, fd, 0);
if (virtaddr == MAP_FAILED) {
RTE_LOG(ERR, EAL, "%s(): mmap failed: %s\n", __func__,
strerror(errno));
close(fd);
return -1;
}
if (orig) {
hugepg_tbl[i].orig_va = virtaddr;
memset(virtaddr, 0, hugepage_sz);
}
else {
hugepg_tbl[i].final_va = virtaddr;
}
/* set shared flock on the file. */
if (flock(fd, LOCK_SH | LOCK_NB) == -1) {
RTE_LOG(ERR, EAL, "%s(): Locking file failed:%s \n",
__func__, strerror(errno));
close(fd);
return -1;
}
close(fd);
vma_addr = (char *)vma_addr + hugepage_sz;
vma_len -= hugepage_sz;
}
return 0;
}
/*
* Mmap all hugepages of hugepage table: it first open a file in
* hugetlbfs, then mmap() hugepage_sz data in it. If orig is set, the
* virtual address is stored in hugepg_tbl[i].orig_va, else it is stored
* in hugepg_tbl[i].final_va. The second mapping (when orig is 0) tries to
* map continguous physical blocks in contiguous virtual blocks.
*/
static unsigned
map_all_hugepages(struct hugepage_file *hugepg_tbl,
struct hugepage_info *hpi, int orig)
{
int fd;
unsigned i;
void *virtaddr;
void *vma_addr = NULL;
size_t vma_len = 0;
for (i = 0; i < hpi->num_pages[0]; i++) {
uint64_t hugepage_sz = hpi->hugepage_sz;
if (orig) {
hugepg_tbl[i].file_id = i;
hugepg_tbl[i].size = hugepage_sz;
eal_get_hugefile_path(hugepg_tbl[i].filepath,
sizeof(hugepg_tbl[i].filepath), hpi->hugedir,
hugepg_tbl[i].file_id);
hugepg_tbl[i].filepath[sizeof(hugepg_tbl[i].filepath) - 1] = '\0';
}
#ifndef RTE_ARCH_64
/* for 32-bit systems, don't remap 1G and 16G pages, just reuse
* original map address as final map address.
*/
else if ((hugepage_sz == RTE_PGSIZE_1G)
|| (hugepage_sz == RTE_PGSIZE_16G)) {
hugepg_tbl[i].final_va = hugepg_tbl[i].orig_va;
hugepg_tbl[i].orig_va = NULL;
continue;
}
#endif
else if (vma_len == 0) {
unsigned j, num_pages;
/* reserve a virtual area for next contiguous
* physical block: count the number of
* contiguous physical pages. */
for (j = i+1; j < hpi->num_pages[0] ; j++) {
#ifdef RTE_ARCH_PPC_64
/* The physical addresses are sorted in
* descending order on PPC64 */
if (hugepg_tbl[j].physaddr !=
hugepg_tbl[j-1].physaddr - hugepage_sz)
break;
#else
if (hugepg_tbl[j].physaddr !=
hugepg_tbl[j-1].physaddr + hugepage_sz)
break;
#endif
}
num_pages = j - i;
vma_len = num_pages * hugepage_sz;
/* get the biggest virtual memory area up to
* vma_len. If it fails, vma_addr is NULL, so
* let the kernel provide the address. */
vma_addr = get_virtual_area(&vma_len, hpi->hugepage_sz);
if (vma_addr == NULL)
vma_len = hugepage_sz;
}
/* try to create hugepage file */
fd = open(hugepg_tbl[i].filepath, O_CREAT | O_RDWR, 0600);
if (fd < 0) {
RTE_LOG(DEBUG, EAL, "%s(): open failed: %s\n", __func__,
strerror(errno));
return i;
}
/* map the segment, and populate page tables,
* the kernel fills this segment with zeros */
virtaddr = mmap(vma_addr, hugepage_sz, PROT_READ | PROT_WRITE,
MAP_SHARED | MAP_POPULATE, fd, 0);
if (virtaddr == MAP_FAILED) {
RTE_LOG(DEBUG, EAL, "%s(): mmap failed: %s\n", __func__,
strerror(errno));
close(fd);
return i;
}
if (orig) {
hugepg_tbl[i].orig_va = virtaddr;
}
else {
hugepg_tbl[i].final_va = virtaddr;
}
if (orig) {
/* In linux, hugetlb limitations, like cgroup, are
* enforced at fault time instead of mmap(), even
* with the option of MAP_POPULATE. Kernel will send
* a SIGBUS signal. To avoid to be killed, save stack
* environment here, if SIGBUS happens, we can jump
* back here.
*/
if (huge_wrap_sigsetjmp()) {
RTE_LOG(DEBUG, EAL, "SIGBUS: Cannot mmap more "
"hugepages of size %u MB\n",
(unsigned)(hugepage_sz / 0x100000));
munmap(virtaddr, hugepage_sz);
close(fd);
unlink(hugepg_tbl[i].filepath);
return i;
}
*(int *)virtaddr = 0;
}
/* set shared flock on the file. */
if (flock(fd, LOCK_SH | LOCK_NB) == -1) {
RTE_LOG(DEBUG, EAL, "%s(): Locking file failed:%s \n",
__func__, strerror(errno));
close(fd);
return i;
}
close(fd);
vma_addr = (char *)vma_addr + hugepage_sz;
vma_len -= hugepage_sz;
}
return i;
}
