static unsigned int surplus_huge_pages_node[MAX_NUMNODES];
static gfp_t htlb_alloc_mask = GFP_HIGHUSER;
unsigned long hugepages_treat_as_movable;
+unsigned long nr_overcommit_huge_pages;
+static int hugetlb_next_nid;
/*
* Protects updates to hugepage_freelists, nr_huge_pages, and free_huge_pages
static void free_huge_page(struct page *page)
{
int nid = page_to_nid(page);
+ struct address_space *mapping;
+ mapping = (struct address_space *) page_private(page);
BUG_ON(page_count(page));
INIT_LIST_HEAD(&page->lru);
enqueue_huge_page(page);
}
spin_unlock(&hugetlb_lock);
+ if (mapping)
+ hugetlb_put_quota(mapping, 1);
+ set_page_private(page, 0);
}
/*
return ret;
}
-static int alloc_fresh_huge_page(void)
+static struct page *alloc_fresh_huge_page_node(int nid)
{
- static int prev_nid;
struct page *page;
- int nid;
- /*
- * Copy static prev_nid to local nid, work on that, then copy it
- * back to prev_nid afterwards: otherwise there's a window in which
- * a racer might pass invalid nid MAX_NUMNODES to alloc_pages_node.
- * But we don't need to use a spin_lock here: it really doesn't
- * matter if occasionally a racer chooses the same nid as we do.
- */
- nid = next_node(prev_nid, node_online_map);
- if (nid == MAX_NUMNODES)
- nid = first_node(node_online_map);
- prev_nid = nid;
-
- page = alloc_pages_node(nid, htlb_alloc_mask|__GFP_COMP|__GFP_NOWARN,
- HUGETLB_PAGE_ORDER);
+ page = alloc_pages_node(nid,
+ htlb_alloc_mask|__GFP_COMP|__GFP_THISNODE|__GFP_NOWARN,
+ HUGETLB_PAGE_ORDER);
if (page) {
set_compound_page_dtor(page, free_huge_page);
spin_lock(&hugetlb_lock);
nr_huge_pages++;
- nr_huge_pages_node[page_to_nid(page)]++;
+ nr_huge_pages_node[nid]++;
spin_unlock(&hugetlb_lock);
put_page(page); /* free it into the hugepage allocator */
- return 1;
}
- return 0;
+
+ return page;
+}
+
+static int alloc_fresh_huge_page(void)
+{
+ struct page *page;
+ int start_nid;
+ int next_nid;
+ int ret = 0;
+
+ start_nid = hugetlb_next_nid;
+
+ do {
+ page = alloc_fresh_huge_page_node(hugetlb_next_nid);
+ if (page)
+ ret = 1;
+ /*
+ * Use a helper variable to find the next node and then
+ * copy it back to hugetlb_next_nid afterwards:
+ * otherwise there's a window in which a racer might
+ * pass invalid nid MAX_NUMNODES to alloc_pages_node.
+ * But we don't need to use a spin_lock here: it really
+ * doesn't matter if occasionally a racer chooses the
+ * same nid as we do. Move nid forward in the mask even
+ * if we just successfully allocated a hugepage so that
+ * the next caller gets hugepages on the next node.
+ */
+ next_nid = next_node(hugetlb_next_nid, node_online_map);
+ if (next_nid == MAX_NUMNODES)
+ next_nid = first_node(node_online_map);
+ hugetlb_next_nid = next_nid;
+ } while (!page && hugetlb_next_nid != start_nid);
+
+ return ret;
}
static struct page *alloc_buddy_huge_page(struct vm_area_struct *vma,
unsigned long address)
{
struct page *page;
+ unsigned int nid;
+
+ /*
+ * Assume we will successfully allocate the surplus page to
+ * prevent racing processes from causing the surplus to exceed
+ * overcommit
+ *
+ * This however introduces a different race, where a process B
+ * tries to grow the static hugepage pool while alloc_pages() is
+ * called by process A. B will only examine the per-node
+ * counters in determining if surplus huge pages can be
+ * converted to normal huge pages in adjust_pool_surplus(). A
+ * won't be able to increment the per-node counter, until the
+ * lock is dropped by B, but B doesn't drop hugetlb_lock until
+ * no more huge pages can be converted from surplus to normal
+ * state (and doesn't try to convert again). Thus, we have a
+ * case where a surplus huge page exists, the pool is grown, and
+ * the surplus huge page still exists after, even though it
+ * should just have been converted to a normal huge page. This
+ * does not leak memory, though, as the hugepage will be freed
+ * once it is out of use. It also does not allow the counters to
+ * go out of whack in adjust_pool_surplus() as we don't modify
+ * the node values until we've gotten the hugepage and only the
+ * per-node value is checked there.
+ */
+ spin_lock(&hugetlb_lock);
+ if (surplus_huge_pages >= nr_overcommit_huge_pages) {
+ spin_unlock(&hugetlb_lock);
+ return NULL;
+ } else {
+ nr_huge_pages++;
+ surplus_huge_pages++;
+ }
+ spin_unlock(&hugetlb_lock);
page = alloc_pages(htlb_alloc_mask|__GFP_COMP|__GFP_NOWARN,
HUGETLB_PAGE_ORDER);
+
+ spin_lock(&hugetlb_lock);
if (page) {
+ nid = page_to_nid(page);
set_compound_page_dtor(page, free_huge_page);
- spin_lock(&hugetlb_lock);
- nr_huge_pages++;
- nr_huge_pages_node[page_to_nid(page)]++;
- surplus_huge_pages++;
- surplus_huge_pages_node[page_to_nid(page)]++;
- spin_unlock(&hugetlb_lock);
+ /*
+ * We incremented the global counters already
+ */
+ nr_huge_pages_node[nid]++;
+ surplus_huge_pages_node[nid]++;
+ } else {
+ nr_huge_pages--;
+ surplus_huge_pages--;
}
+ spin_unlock(&hugetlb_lock);
return page;
}
list_del(&page->lru);
if ((--needed) >= 0)
enqueue_huge_page(page);
- else
- update_and_free_page(page);
+ else {
+ /*
+ * Decrement the refcount and free the page using its
+ * destructor. This must be done with hugetlb_lock
+ * unlocked which is safe because free_huge_page takes
+ * hugetlb_lock before deciding how to free the page.
+ */
+ spin_unlock(&hugetlb_lock);
+ put_page(page);
+ spin_lock(&hugetlb_lock);
+ }
}
return ret;
* allocated to satisfy the reservation must be explicitly freed if they were
* never used.
*/
-void return_unused_surplus_pages(unsigned long unused_resv_pages)
+static void return_unused_surplus_pages(unsigned long unused_resv_pages)
{
static int nid = -1;
struct page *page;
}
}
-static struct page *alloc_huge_page(struct vm_area_struct *vma,
- unsigned long addr)
+
+static struct page *alloc_huge_page_shared(struct vm_area_struct *vma,
+ unsigned long addr)
{
- struct page *page = NULL;
- int use_reserved_page = vma->vm_flags & VM_MAYSHARE;
+ struct page *page;
spin_lock(&hugetlb_lock);
- if (!use_reserved_page && (free_huge_pages <= resv_huge_pages))
- goto fail;
-
page = dequeue_huge_page(vma, addr);
- if (!page)
- goto fail;
+ spin_unlock(&hugetlb_lock);
+ return page ? page : ERR_PTR(-VM_FAULT_OOM);
+}
+
+static struct page *alloc_huge_page_private(struct vm_area_struct *vma,
+ unsigned long addr)
+{
+ struct page *page = NULL;
+ if (hugetlb_get_quota(vma->vm_file->f_mapping, 1))
+ return ERR_PTR(-VM_FAULT_SIGBUS);
+
+ spin_lock(&hugetlb_lock);
+ if (free_huge_pages > resv_huge_pages)
+ page = dequeue_huge_page(vma, addr);
spin_unlock(&hugetlb_lock);
- set_page_refcounted(page);
+ if (!page) {
+ page = alloc_buddy_huge_page(vma, addr);
+ if (!page) {
+ hugetlb_put_quota(vma->vm_file->f_mapping, 1);
+ return ERR_PTR(-VM_FAULT_OOM);
+ }
+ }
return page;
+}
-fail:
- spin_unlock(&hugetlb_lock);
+static struct page *alloc_huge_page(struct vm_area_struct *vma,
+ unsigned long addr)
+{
+ struct page *page;
+ struct address_space *mapping = vma->vm_file->f_mapping;
- /*
- * Private mappings do not use reserved huge pages so the allocation
- * may have failed due to an undersized hugetlb pool. Try to grab a
- * surplus huge page from the buddy allocator.
- */
- if (!use_reserved_page)
- page = alloc_buddy_huge_page(vma, addr);
+ if (vma->vm_flags & VM_MAYSHARE)
+ page = alloc_huge_page_shared(vma, addr);
+ else
+ page = alloc_huge_page_private(vma, addr);
+ if (!IS_ERR(page)) {
+ set_page_refcounted(page);
+ set_page_private(page, (unsigned long) mapping);
+ }
return page;
}
for (i = 0; i < MAX_NUMNODES; ++i)
INIT_LIST_HEAD(&hugepage_freelists[i]);
+ hugetlb_next_nid = first_node(node_online_map);
+
for (i = 0; i < max_huge_pages; ++i) {
if (!alloc_fresh_huge_page())
break;
for (i = 0; i < MAX_NUMNODES; ++i) {
struct page *page, *next;
list_for_each_entry_safe(page, next, &hugepage_freelists[i], lru) {
+ if (count >= nr_huge_pages)
+ return;
if (PageHighMem(page))
continue;
list_del(&page->lru);
update_and_free_page(page);
free_huge_pages--;
free_huge_pages_node[page_to_nid(page)]--;
- if (count >= nr_huge_pages)
- return;
}
}
}
* Increase the pool size
* First take pages out of surplus state. Then make up the
* remaining difference by allocating fresh huge pages.
+ *
+ * We might race with alloc_buddy_huge_page() here and be unable
+ * to convert a surplus huge page to a normal huge page. That is
+ * not critical, though, it just means the overall size of the
+ * pool might be one hugepage larger than it needs to be, but
+ * within all the constraints specified by the sysctls.
*/
spin_lock(&hugetlb_lock);
while (surplus_huge_pages && count > persistent_huge_pages) {
goto out;
}
- if (count >= persistent_huge_pages)
- goto out;
/*
* Decrease the pool size
* to keep enough around to satisfy reservations). Then place
* pages into surplus state as needed so the pool will shrink
* to the desired size as pages become free.
+ *
+ * By placing pages into the surplus state independent of the
+ * overcommit value, we are allowing the surplus pool size to
+ * exceed overcommit. There are few sane options here. Since
+ * alloc_buddy_huge_page() is checking the global counter,
+ * though, we'll note that we're not allowed to exceed surplus
+ * and won't grow the pool anywhere else. Not until one of the
+ * sysctls are changed, or the surplus pages go out of use.
*/
- min_count = max(count, resv_huge_pages);
+ min_count = resv_huge_pages + nr_huge_pages - free_huge_pages;
+ min_count = max(count, min_count);
try_to_free_low(min_count);
while (min_count < persistent_huge_pages) {
struct page *page = dequeue_huge_page(NULL, 0);
dst_pte = huge_pte_alloc(dst, addr);
if (!dst_pte)
goto nomem;
+
+ /* If the pagetables are shared don't copy or take references */
+ if (dst_pte == src_pte)
+ continue;
+
spin_lock(&dst->page_table_lock);
spin_lock(&src->page_table_lock);
if (!pte_none(*src_pte)) {
page_cache_get(old_page);
new_page = alloc_huge_page(vma, address);
- if (!new_page) {
+ if (IS_ERR(new_page)) {
page_cache_release(old_page);
- return VM_FAULT_OOM;
+ return -PTR_ERR(new_page);
}
spin_unlock(&mm->page_table_lock);
size = i_size_read(mapping->host) >> HPAGE_SHIFT;
if (idx >= size)
goto out;
- if (hugetlb_get_quota(mapping))
- goto out;
page = alloc_huge_page(vma, address);
- if (!page) {
- hugetlb_put_quota(mapping);
- ret = VM_FAULT_OOM;
+ if (IS_ERR(page)) {
+ ret = -PTR_ERR(page);
goto out;
}
clear_huge_page(page, address);
if (vma->vm_flags & VM_SHARED) {
int err;
+ struct inode *inode = mapping->host;
err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
if (err) {
put_page(page);
- hugetlb_put_quota(mapping);
if (err == -EEXIST)
goto retry;
goto out;
}
+
+ spin_lock(&inode->i_lock);
+ inode->i_blocks += BLOCKS_PER_HUGEPAGE;
+ spin_unlock(&inode->i_lock);
} else
lock_page(page);
}
backout:
spin_unlock(&mm->page_table_lock);
- hugetlb_put_quota(mapping);
unlock_page(page);
put_page(page);
goto out;
int follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
struct page **pages, struct vm_area_struct **vmas,
- unsigned long *position, int *length, int i)
+ unsigned long *position, int *length, int i,
+ int write)
{
unsigned long pfn_offset;
unsigned long vaddr = *position;
*/
pte = huge_pte_offset(mm, vaddr & HPAGE_MASK);
- if (!pte || pte_none(*pte)) {
+ if (!pte || pte_none(*pte) || (write && !pte_write(*pte))) {
int ret;
spin_unlock(&mm->page_table_lock);
- ret = hugetlb_fault(mm, vma, vaddr, 0);
+ ret = hugetlb_fault(mm, vma, vaddr, write);
spin_lock(&mm->page_table_lock);
if (!(ret & VM_FAULT_ERROR))
continue;
/* If we are below the current region then a new region is required.
* Subtle, allocate a new region at the position but make it zero
- * size such that we can guarentee to record the reservation. */
+ * size such that we can guarantee to record the reservation. */
if (&rg->link == head || t < rg->from) {
nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
- if (nrg == 0)
+ if (!nrg)
return -ENOMEM;
nrg->from = f;
nrg->to = f;
if (chg < 0)
return chg;
+ if (hugetlb_get_quota(inode->i_mapping, chg))
+ return -ENOSPC;
ret = hugetlb_acct_memory(chg);
- if (ret < 0)
+ if (ret < 0) {
+ hugetlb_put_quota(inode->i_mapping, chg);
return ret;
+ }
region_add(&inode->i_mapping->private_list, from, to);
return 0;
}
void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed)
{
long chg = region_truncate(&inode->i_mapping->private_list, offset);
- hugetlb_acct_memory(freed - chg);
+
+ spin_lock(&inode->i_lock);
+ inode->i_blocks -= BLOCKS_PER_HUGEPAGE * freed;
+ spin_unlock(&inode->i_lock);
+
+ hugetlb_put_quota(inode->i_mapping, (chg - freed));
+ hugetlb_acct_memory(-(chg - freed));
}