From 3fec86f8aa8c7ae84567a0d1396e84ada96141d8 Mon Sep 17 00:00:00 2001 From: Zi Yan Date: Fri, 7 Mar 2025 12:39:54 -0500 Subject: xarray: add xas_try_split() to split a multi-index entry Patch series "Buddy allocator like (or non-uniform) folio split", v10. This patchset adds a new buddy allocator like (or non-uniform) large folio split from a order-n folio to order-m with m < n. It reduces 1. the total number of after-split folios from 2^(n-m) to n-m+1; 2. the amount of memory needed for multi-index xarray split from 2^(n/6-m/6) to n/6-m/6, assuming XA_CHUNK_SHIFT=6; 3. keep more large folios after a split from all order-m folios to order-(n-1) to order-m folios. For example, to split an order-9 to order-0, folio split generates 10 (or 11 for anonymous memory) folios instead of 512, allocates 1 xa_node instead of 8, and leaves 1 order-8, 1 order-7, ..., 1 order-1 and 2 order-0 folios (or 4 order-0 for anonymous memory) instead of 512 order-0 folios. Instead of duplicating existing split_huge_page*() code, __folio_split() is introduced as the shared backend code for both split_huge_page_to_list_to_order() and folio_split(). __folio_split() can support both uniform split and buddy allocator like (or non-uniform) split. All existing split_huge_page*() users can be gradually converted to use folio_split() if possible. In this patchset, I converted truncate_inode_partial_folio() to use folio_split(). xfstests quick group passed for both tmpfs and xfs. I also semi-replicated Hugh's test[12] and ran it without any issue for almost 24 hours. This patch (of 8): A preparation patch for non-uniform folio split, which always split a folio into half iteratively, and minimal xarray entry split. Currently, xas_split_alloc() and xas_split() always split all slots from a multi-index entry. They cost the same number of xa_node as the to-be-split slots. For example, to split an order-9 entry, which takes 2^(9-6)=8 slots, assuming XA_CHUNK_SHIFT is 6 (!CONFIG_BASE_SMALL), 8 xa_node are needed. Instead xas_try_split() is intended to be used iteratively to split the order-9 entry into 2 order-8 entries, then split one order-8 entry, based on the given index, to 2 order-7 entries, ..., and split one order-1 entry to 2 order-0 entries. When splitting the order-6 entry and a new xa_node is needed, xas_try_split() will try to allocate one if possible. As a result, xas_try_split() would only need 1 xa_node instead of 8. When a new xa_node is needed during the split, xas_try_split() can try to allocate one but no more. -ENOMEM will be return if a node cannot be allocated. -EINVAL will be return if a sibling node is split or cascade split happens, where two or more new nodes are needed, and these are not supported by xas_try_split(). xas_split_alloc() and xas_split() split an order-9 to order-0: --------------------------------- | | | | | | | | | | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | | | | | | | | | | --------------------------------- | | | | ------- --- --- ------- | | ... | | V V V V ----------- ----------- ----------- ----------- | xa_node | | xa_node | ... | xa_node | | xa_node | ----------- ----------- ----------- ----------- xas_try_split() splits an order-9 to order-0: --------------------------------- | | | | | | | | | | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | | | | | | | | | | --------------------------------- | | V ----------- | xa_node | ----------- Link: https://lkml.kernel.org/r/20250307174001.242794-1-ziy@nvidia.com Link: https://lkml.kernel.org/r/20250307174001.242794-2-ziy@nvidia.com Signed-off-by: Zi Yan Cc: Baolin Wang Cc: David Hildenbrand Cc: Hugh Dickins Cc: John Hubbard Cc: Kefeng Wang Cc: Kirill A. Shuemov Cc: Miaohe Lin Cc: Matthew Wilcox Cc: Ryan Roberts Cc: Yang Shi Cc: Yu Zhao Cc: Zi Yan Cc: Kairui Song Signed-off-by: Andrew Morton --- lib/xarray.c | 132 +++++++++++++++++++++++++++++++++++++++++++++++++++++------ 1 file changed, 120 insertions(+), 12 deletions(-) (limited to 'lib/xarray.c') diff --git a/lib/xarray.c b/lib/xarray.c index 116e9286c64e..3bae48558e21 100644 --- a/lib/xarray.c +++ b/lib/xarray.c @@ -278,6 +278,7 @@ void xas_destroy(struct xa_state *xas) xas->xa_alloc = node = next; } } +EXPORT_SYMBOL_GPL(xas_destroy); /** * xas_nomem() - Allocate memory if needed. @@ -1007,6 +1008,26 @@ static void node_set_marks(struct xa_node *node, unsigned int offset, } } +static void __xas_init_node_for_split(struct xa_state *xas, + struct xa_node *node, void *entry) +{ + unsigned int i; + void *sibling = NULL; + unsigned int mask = xas->xa_sibs; + + if (!node) + return; + node->array = xas->xa; + for (i = 0; i < XA_CHUNK_SIZE; i++) { + if ((i & mask) == 0) { + RCU_INIT_POINTER(node->slots[i], entry); + sibling = xa_mk_sibling(i); + } else { + RCU_INIT_POINTER(node->slots[i], sibling); + } + } +} + /** * xas_split_alloc() - Allocate memory for splitting an entry. * @xas: XArray operation state. @@ -1025,7 +1046,6 @@ void xas_split_alloc(struct xa_state *xas, void *entry, unsigned int order, gfp_t gfp) { unsigned int sibs = (1 << (order % XA_CHUNK_SHIFT)) - 1; - unsigned int mask = xas->xa_sibs; /* XXX: no support for splitting really large entries yet */ if (WARN_ON(xas->xa_shift + 2 * XA_CHUNK_SHIFT <= order)) @@ -1034,22 +1054,13 @@ void xas_split_alloc(struct xa_state *xas, void *entry, unsigned int order, return; do { - unsigned int i; - void *sibling = NULL; struct xa_node *node; node = kmem_cache_alloc_lru(radix_tree_node_cachep, xas->xa_lru, gfp); if (!node) goto nomem; - node->array = xas->xa; - for (i = 0; i < XA_CHUNK_SIZE; i++) { - if ((i & mask) == 0) { - RCU_INIT_POINTER(node->slots[i], entry); - sibling = xa_mk_sibling(i); - } else { - RCU_INIT_POINTER(node->slots[i], sibling); - } - } + + __xas_init_node_for_split(xas, node, entry); RCU_INIT_POINTER(node->parent, xas->xa_alloc); xas->xa_alloc = node; } while (sibs-- > 0); @@ -1122,6 +1133,103 @@ void xas_split(struct xa_state *xas, void *entry, unsigned int order) xas_update(xas, node); } EXPORT_SYMBOL_GPL(xas_split); + +/** + * xas_try_split() - Try to split a multi-index entry. + * @xas: XArray operation state. + * @entry: New entry to store in the array. + * @order: Current entry order. + * + * The size of the new entries is set in @xas. The value in @entry is + * copied to all the replacement entries. If and only if one new xa_node is + * needed, the function will use GFP_NOWAIT to get one if xas->xa_alloc is + * NULL. If more new xa_node are needed, the function gives EINVAL error. + * + * Context: Any context. The caller should hold the xa_lock. + */ +void xas_try_split(struct xa_state *xas, void *entry, unsigned int order) +{ + unsigned int sibs = (1 << (order % XA_CHUNK_SHIFT)) - 1; + unsigned int offset, marks; + struct xa_node *node; + void *curr = xas_load(xas); + int values = 0; + gfp_t gfp = GFP_NOWAIT; + + node = xas->xa_node; + if (xas_top(node)) + return; + + if (xas->xa->xa_flags & XA_FLAGS_ACCOUNT) + gfp |= __GFP_ACCOUNT; + + marks = node_get_marks(node, xas->xa_offset); + + offset = xas->xa_offset + sibs; + + if (xas->xa_shift < node->shift) { + struct xa_node *child = xas->xa_alloc; + unsigned int expected_sibs = + (1 << ((order - 1) % XA_CHUNK_SHIFT)) - 1; + + /* + * No support for splitting sibling entries + * (horizontally) or cascade split (vertically), which + * requires two or more new xa_nodes. + * Since if one xa_node allocation fails, + * it is hard to free the prior allocations. + */ + if (sibs || xas->xa_sibs != expected_sibs) { + xas_destroy(xas); + xas_set_err(xas, -EINVAL); + return; + } + + if (!child) { + child = kmem_cache_alloc_lru(radix_tree_node_cachep, + xas->xa_lru, gfp); + if (!child) { + xas_destroy(xas); + xas_set_err(xas, -ENOMEM); + return; + } + RCU_INIT_POINTER(child->parent, xas->xa_alloc); + } + __xas_init_node_for_split(xas, child, entry); + + xas->xa_alloc = rcu_dereference_raw(child->parent); + child->shift = node->shift - XA_CHUNK_SHIFT; + child->offset = offset; + child->count = XA_CHUNK_SIZE; + child->nr_values = xa_is_value(entry) ? + XA_CHUNK_SIZE : 0; + RCU_INIT_POINTER(child->parent, node); + node_set_marks(node, offset, child, xas->xa_sibs, + marks); + rcu_assign_pointer(node->slots[offset], + xa_mk_node(child)); + if (xa_is_value(curr)) + values--; + xas_update(xas, child); + + } else { + do { + unsigned int canon = offset - xas->xa_sibs; + + node_set_marks(node, canon, NULL, 0, marks); + rcu_assign_pointer(node->slots[canon], entry); + while (offset > canon) + rcu_assign_pointer(node->slots[offset--], + xa_mk_sibling(canon)); + values += (xa_is_value(entry) - xa_is_value(curr)) * + (xas->xa_sibs + 1); + } while (offset-- > xas->xa_offset); + } + + node->nr_values += values; + xas_update(xas, node); +} +EXPORT_SYMBOL_GPL(xas_try_split); #endif /** -- cgit v1.2.3-59-g8ed1b From 200a89c159a7a416115e6e309183c82183bf98aa Mon Sep 17 00:00:00 2001 From: Zi Yan Date: Fri, 14 Mar 2025 18:21:12 -0400 Subject: mm/filemap: use xas_try_split() in __filemap_add_folio() Patch series "Minimize xa_node allocation during xarry split", v3. When splitting a multi-index entry in XArray from order-n to order-m, existing xas_split_alloc()+xas_split() approach requires 2^(n % XA_CHUNK_SHIFT) xa_node allocations. But its callers, __filemap_add_folio() and shmem_split_large_entry(), use at most 1 xa_node. To minimize xa_node allocation and remove the limitation of no split from order-12 (or above) to order-0 (or anything between 0 and 5)[1], xas_try_split() was added[2], which allocates (n / XA_CHUNK_SHIFT - m / XA_CHUNK_SHIFT) xa_node. It is used for non-uniform folio split, but can be used by __filemap_add_folio() and shmem_split_large_entry(). xas_split_alloc() and xas_split() split an order-9 to order-0: --------------------------------- | | | | | | | | | | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | | | | | | | | | | --------------------------------- | | | | ------- --- --- ------- | | ... | | V V V V ----------- ----------- ----------- ----------- | xa_node | | xa_node | ... | xa_node | | xa_node | ----------- ----------- ----------- ----------- xas_try_split() splits an order-9 to order-0: --------------------------------- | | | | | | | | | | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | | | | | | | | | | --------------------------------- | | V ----------- | xa_node | ----------- xas_try_split() is designed to be called iteratively with n = m + 1. xas_try_split_mini_order() is added to minmize the number of calls to xas_try_split() by telling the caller the next minimal order to split to instead of n - 1. Splitting order-n to order-m when m= l * XA_CHUNK_SHIFT does not require xa_node allocation and requires 1 xa_node when n=l * XA_CHUNK_SHIFT and m = n - 1, so it is OK to use xas_try_split() with n > m + 1 when no new xa_node is needed. xfstests quick group test passed on xfs and tmpfs. [1] https://lore.kernel.org/linux-mm/Z6YX3RznGLUD07Ao@casper.infradead.org/ [2] https://lore.kernel.org/linux-mm/20250226210032.2044041-1-ziy@nvidia.com/ This patch (of 2): During __filemap_add_folio(), a shadow entry is covering n slots and a folio covers m slots with m < n is to be added. Instead of splitting all n slots, only the m slots covered by the folio need to be split and the remaining n-m shadow entries can be retained with orders ranging from m to n-1. This method only requires (n/XA_CHUNK_SHIFT) - (m/XA_CHUNK_SHIFT) new xa_nodes instead of (n % XA_CHUNK_SHIFT) * ((n/XA_CHUNK_SHIFT) - (m/XA_CHUNK_SHIFT)) new xa_nodes, compared to the original xas_split_alloc() + xas_split() one. For example, to insert an order-0 folio when an order-9 shadow entry is present (assuming XA_CHUNK_SHIFT is 6), 1 xa_node is needed instead of 8. xas_try_split_min_order() is introduced to reduce the number of calls to xas_try_split() during split. Link: https://lkml.kernel.org/r/20250314222113.711703-1-ziy@nvidia.com Link: https://lkml.kernel.org/r/20250314222113.711703-2-ziy@nvidia.com Signed-off-by: Zi Yan Cc: Baolin Wang Cc: Hugh Dickins Cc: Kairui Song Cc: Miaohe Lin Cc: Mattew Wilcox Cc: David Hildenbrand Cc: John Hubbard Cc: Kefeng Wang Cc: Kirill A. Shuemov Cc: Ryan Roberts Cc: Yang Shi Cc: Yu Zhao Signed-off-by: Andrew Morton --- include/linux/xarray.h | 7 +++++++ lib/xarray.c | 25 +++++++++++++++++++++++++ mm/filemap.c | 45 ++++++++++++++++++--------------------------- 3 files changed, 50 insertions(+), 27 deletions(-) (limited to 'lib/xarray.c') diff --git a/include/linux/xarray.h b/include/linux/xarray.h index 4010195201c9..78eede109b1a 100644 --- a/include/linux/xarray.h +++ b/include/linux/xarray.h @@ -1556,6 +1556,7 @@ int xas_get_order(struct xa_state *xas); void xas_split(struct xa_state *, void *entry, unsigned int order); void xas_split_alloc(struct xa_state *, void *entry, unsigned int order, gfp_t); void xas_try_split(struct xa_state *xas, void *entry, unsigned int order); +unsigned int xas_try_split_min_order(unsigned int order); #else static inline int xa_get_order(struct xarray *xa, unsigned long index) { @@ -1582,6 +1583,12 @@ static inline void xas_try_split(struct xa_state *xas, void *entry, unsigned int order) { } + +static inline unsigned int xas_try_split_min_order(unsigned int order) +{ + return 0; +} + #endif /** diff --git a/lib/xarray.c b/lib/xarray.c index 3bae48558e21..9644b18af18d 100644 --- a/lib/xarray.c +++ b/lib/xarray.c @@ -1134,6 +1134,28 @@ void xas_split(struct xa_state *xas, void *entry, unsigned int order) } EXPORT_SYMBOL_GPL(xas_split); +/** + * xas_try_split_min_order() - Minimal split order xas_try_split() can accept + * @order: Current entry order. + * + * xas_try_split() can split a multi-index entry to smaller than @order - 1 if + * no new xa_node is needed. This function provides the minimal order + * xas_try_split() supports. + * + * Return: the minimal order xas_try_split() supports + * + * Context: Any context. + * + */ +unsigned int xas_try_split_min_order(unsigned int order) +{ + if (order % XA_CHUNK_SHIFT == 0) + return order == 0 ? 0 : order - 1; + + return order - (order % XA_CHUNK_SHIFT); +} +EXPORT_SYMBOL_GPL(xas_try_split_min_order); + /** * xas_try_split() - Try to split a multi-index entry. * @xas: XArray operation state. @@ -1145,6 +1167,9 @@ EXPORT_SYMBOL_GPL(xas_split); * needed, the function will use GFP_NOWAIT to get one if xas->xa_alloc is * NULL. If more new xa_node are needed, the function gives EINVAL error. * + * NOTE: use xas_try_split_min_order() to get next split order instead of + * @order - 1 if you want to minmize xas_try_split() calls. + * * Context: Any context. The caller should hold the xa_lock. */ void xas_try_split(struct xa_state *xas, void *entry, unsigned int order) diff --git a/mm/filemap.c b/mm/filemap.c index 152993a86de3..cc69f174f76b 100644 --- a/mm/filemap.c +++ b/mm/filemap.c @@ -857,11 +857,10 @@ EXPORT_SYMBOL_GPL(replace_page_cache_folio); noinline int __filemap_add_folio(struct address_space *mapping, struct folio *folio, pgoff_t index, gfp_t gfp, void **shadowp) { - XA_STATE(xas, &mapping->i_pages, index); - void *alloced_shadow = NULL; - int alloced_order = 0; + XA_STATE_ORDER(xas, &mapping->i_pages, index, folio_order(folio)); bool huge; long nr; + unsigned int forder = folio_order(folio); VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio); VM_BUG_ON_FOLIO(folio_test_swapbacked(folio), folio); @@ -870,7 +869,6 @@ noinline int __filemap_add_folio(struct address_space *mapping, mapping_set_update(&xas, mapping); VM_BUG_ON_FOLIO(index & (folio_nr_pages(folio) - 1), folio); - xas_set_order(&xas, index, folio_order(folio)); huge = folio_test_hugetlb(folio); nr = folio_nr_pages(folio); @@ -880,7 +878,7 @@ noinline int __filemap_add_folio(struct address_space *mapping, folio->index = xas.xa_index; for (;;) { - int order = -1, split_order = 0; + int order = -1; void *entry, *old = NULL; xas_lock_irq(&xas); @@ -898,21 +896,25 @@ noinline int __filemap_add_folio(struct address_space *mapping, order = xas_get_order(&xas); } - /* entry may have changed before we re-acquire the lock */ - if (alloced_order && (old != alloced_shadow || order != alloced_order)) { - xas_destroy(&xas); - alloced_order = 0; - } - if (old) { - if (order > 0 && order > folio_order(folio)) { + if (order > 0 && order > forder) { + unsigned int split_order = max(forder, + xas_try_split_min_order(order)); + /* How to handle large swap entries? */ BUG_ON(shmem_mapping(mapping)); - if (!alloced_order) { - split_order = order; - goto unlock; + + while (order > forder) { + xas_set_order(&xas, index, split_order); + xas_try_split(&xas, old, order); + if (xas_error(&xas)) + goto unlock; + order = split_order; + split_order = + max(xas_try_split_min_order( + split_order), + forder); } - xas_split(&xas, old, order); xas_reset(&xas); } if (shadowp) @@ -936,17 +938,6 @@ noinline int __filemap_add_folio(struct address_space *mapping, unlock: xas_unlock_irq(&xas); - /* split needed, alloc here and retry. */ - if (split_order) { - xas_split_alloc(&xas, old, split_order, gfp); - if (xas_error(&xas)) - goto error; - alloced_shadow = old; - alloced_order = split_order; - xas_reset(&xas); - continue; - } - if (!xas_nomem(&xas, gfp)) break; } -- cgit v1.2.3-59-g8ed1b