Swap-out small-sized THP without splitting
Subsystems (5)
Ryan Roberts
reviewercover Swap-out small-sized THP without splitting
Hi All,
This is an RFC for a small series to add support for swapping out small-sized THP without needing to first split the large folio via __split_huge_page(). It closely follows the approach already used by PMD-sized THP.
"Small-sized THP" is an upcoming feature that enables performance improvements by allocating large folios for anonymous memory, where the large folio size is smaller than the traditional PMD-size. See [1].
In some circumstances I've observed a performance regression (see patch 2 for details), and this series is an attempt to fix the regression in advance of merging small-sized THP support.
I've done what I thought was the smallest change possible, and as a result, this approach is only employed when the swap is backed by a non-rotating block device (just as PMD-sized THP is supported today). However, I have a few questions on whether we should consider relaxing those requirements in certain circumstances:
1) block-backed vs file-backed
==============================
The code only attempts to allocate a contiguous set of entries if swap is backed by a block device (i.e. not file-backed). The original commit, f0eea189e8e9 ("mm, THP, swap: don't allocate huge cluster for file backed swap device"), stated "It's hard to write a whole transparent huge page (THP) to a file backed swap device". But didn't state why. Does this imply there is a size limit at which it becomes hard? And does that therefore imply that for "small enough" sizes we should now allow use with file-back swap?
This original commit was subsequently fixed with commit 41663430588c ("mm, THP,
swap: fix allocating cluster for swapfile by mistake"), which said the original commit was using the wrong flag to determine if it was a block device and therefore in some cases was actually doing large allocations for a file-backed swap device, and this was causing file-system corruption. But that implies some sort of correctness issue to me, rather than the performance issue I inferred from the original commit.
If anyone can offer an explanation, that would be helpful in determining if we should allow some large sizes for file-backed swap.
2) rotating vs non-rotating
===========================
I notice that the clustered approach is only used for non-rotating swap. That implies that for rotating media, we will always fail a large allocation, and fall back to splitting THPs to single pages. Which implies that the regression I'm fixing here may still be present on rotating media? Or perhaps rotating disk is so slow that the cost of writing the data out dominates the cost of splitting?
I considered that potentially the free swap entry search algorithm that is used in this case could be modified to look for (small) contiguous runs of entries; Up to ~16 pages (order-4) could be done by doing 2x 64bit reads from map instead of single byte.
I haven't looked into this idea in detail, but wonder if anybody thinks it is worth the effort? Or perhaps it would end up causing bad fragmentation.
Finally on testing, I've run the mm selftests and see no regressions, but I don't think there is anything in there specifically aimed towards swap? Are there any functional or performance tests that I should run? It would certainly be good to confirm I haven't regressed PMD-size THP swap performance.
Thanks,
Ryan
[1] https://lore.kernel.org/linux-mm/[email protected]/
Ryan Roberts (2):
mm: swap: Remove CLUSTER_FLAG_HUGE from swap_cluster_info:flags
mm: swap: Swap-out small-sized THP without splitting
include/linux/swap.h | 17 +++---- mm/huge_memory.c | 3 -- mm/swapfile.c | 105 ++++++++++++++++++++++--------------------- mm/vmscan.c | 10 +++-- 4 files changed, 66 insertions(+), 69 deletions(-)
--
2.25.1
Ryan Roberts
reviewerRFC PATCH 1/2 mm: swap: Remove CLUSTER_FLAG_HUGE from swap_cluster_info:flags
As preparation for supporting small-sized THP in the swap-out path, without first needing to split to order-0, Remove the CLUSTER_FLAG_HUGE, which, when present, always implies PMD-sized THP, which is the same as the cluster size.
The only use of the flag was to determine whether a swap entry refers to a single page or a PMD-sized THP in swap_page_trans_huge_swapped(). Instead of relying on the flag, we now pass in nr_pages, which originates from the folio's number of pages. This allows the logic to work for folios of any order.
The one snag is that one of the swap_page_trans_huge_swapped() call sites does not have the folio. But it was only being called there to avoid bothering to call __try_to_reclaim_swap() in some cases. __try_to_reclaim_swap() gets the folio and (via some other functions) calls swap_page_trans_huge_swapped(). So I've removed the problematic call site and believe the new logic should be equivalent.
Removing CLUSTER_FLAG_HUGE also means we can remove split_swap_cluster() which used to be called during folio splitting, since split_swap_cluster()'s only job was to remove the flag.
--- include/linux/swap.h | 10 ---------- mm/huge_memory.c | 3 --- mm/swapfile.c | 47 ++++++++------------------------------------ 3 files changed, 8 insertions(+), 52 deletions(-)
include/linux/swap.h +0 -10
@@ -259,7 +259,6 @@ struct swap_cluster_info { };#define CLUSTER_FLAG_FREE 1/* This cluster is free */ #define CLUSTER_FLAG_NEXT_NULL 2/* This cluster has no next cluster */ -#define CLUSTER_FLAG_HUGE 4/* This cluster is backing a transparent huge page */ /* * We assign a cluster to each CPU, so each CPU can allocate swap entry from
@@ -595,15 +594,6 @@ static inline int add_swap_extent(struct swap_info_struct *sis, }#endif /* CONFIG_SWAP */ -#ifdef CONFIG_THP_SWAP -extern int split_swap_cluster (swp_entry_t entry); -#else -static inline int split_swap_cluster (swp_entry_t entry) -{ -return 0 ; -} -#endif -#ifdef CONFIG_MEMCG static inline int mem_cgroup_swappiness (struct mem_cgroup *memcg) {
mm/huge_memory.c +0 -3
@@ -2597,9 +2597,6 @@ static void __split_huge_page(struct page *page, struct list_head *list,shmem_uncharge (head->mapping->host, nr_dropped);remap_page (folio, nr); -if (folio_test_swapcache (folio)) -split_swap_cluster (folio->swap); -for (i =0 ; i < nr; i++) {struct page *subpage = head + i;if (subpage == page)
mm/swapfile.c +8 -39
107 lines not on this page · raw message
--
2.25.1
Re: [RFC PATCH v1 1/2] mm: swap: Remove CLUSTER_FLAG_HUGE from swap_cluster_info:flags
Quoted 11 lines · Ryan Roberts
As preparation for supporting small-sized THP in the swap-out path, without first needing to split to order-0, Remove the
CLUSTER_FLAG_HUGE, which, when present, always implies PMD-sized THP, which is the same as the cluster size.The only use of the flag was to determine whether a swap entry refers to a single page or a PMD-sized THP in
swap_page_trans_huge_swapped(). Instead of relying on the flag, we now pass innr_pages, which originates from the folio's number of pages. This allows the logic to work for folios of any order.
Quoted 6 lines
The one snag is that one of the
swap_page_trans_huge_swapped()call sites does not have the folio. But it was only being called there to avoid bothering to call__try_to_reclaim_swap()in some cases.__try_to_reclaim_swap()gets the folio and (via some other functions) callsswap_page_trans_huge_swapped(). So I've removed the problematic call site and believe the new logic should be equivalent.
I believe this should be OK. Better to compare the performance too.
Quoted 4 lines
Removing
CLUSTER_FLAG_HUGEalso means we can removesplit_swap_cluster()which used to be called during folio splitting, since split_swap_cluster()'s only job was to remove the flag.
--
Best Regards,
Huang, Ying
Re: [RFC PATCH v1 1/2] mm: swap: Remove CLUSTER_FLAG_HUGE from swap_cluster_info:flags
In reply to Ryan RobertsQuoted 95 lines · Ryan Roberts
As preparation for supporting small-sized THP in the swap-out path, without first needing to split to order-0, Remove the
CLUSTER_FLAG_HUGE, which, when present, always implies PMD-sized THP, which is the same as the cluster size.The only use of the flag was to determine whether a swap entry refers to a single page or a PMD-sized THP in
swap_page_trans_huge_swapped(). Instead of relying on the flag, we now pass innr_pages, which originates from the folio's number of pages. This allows the logic to work for folios of any order.The one snag is that one of the
swap_page_trans_huge_swapped()call sites does not have the folio. But it was only being called there to avoid bothering to call__try_to_reclaim_swap()in some cases.__try_to_reclaim_swap()gets the folio and (via some other functions) callsswap_page_trans_huge_swapped(). So I've removed the problematic call site and believe the new logic should be equivalent.Removing
CLUSTER_FLAG_HUGEalso means we can removesplit_swap_cluster()which used to be called during folio splitting, since split_swap_cluster()'s only job was to remove the flag.Signed-off-by Ryan Roberts <[email protected]>--- include/linux/swap.h | 10 ---------- mm/huge_memory.c | 3 --- mm/swapfile.c | 47 ++++++++------------------------------------ 3 files changed, 8 insertions(+), 52 deletions(-)include/linux/swap.h +0
-10@@ -259,7 +259,6 @@ struct swap_cluster_info { };#define CLUSTER_FLAG_FREE 1/* This cluster is free */ #define CLUSTER_FLAG_NEXT_NULL 2/* This cluster has no next cluster */ -#define CLUSTER_FLAG_HUGE 4/* This cluster is backing a transparent huge page */ /* * We assign a cluster to each CPU, so each CPU can allocate swap entry from @@ -595,15 +594,6 @@ static inline int add_swap_extent(struct swap_info_struct *sis, }#endif /* CONFIG_SWAP */ -#ifdef CONFIG_THP_SWAP -extern int split_swap_cluster (swp_entry_t entry); -#else -static inline int split_swap_cluster (swp_entry_t entry) -{ -return 0 ; -} -#endif -#ifdef CONFIG_MEMCG static inline int mem_cgroup_swappiness (struct mem_cgroup *memcg) {mm/huge_memory.c +0
-3@@ -2597,9 +2597,6 @@ static void __split_huge_page(struct page *page, struct list_head *list,shmem_uncharge (head->mapping->host, nr_dropped);remap_page (folio, nr); -if (folio_test_swapcache (folio)) -split_swap_cluster (folio->swap); -for (i =0 ; i < nr; i++) {struct page *subpage = head + i;if (subpage == page)mm/swapfile.c +0
-12@@ -342,18 +342,6 @@ static inline void cluster_set_null(struct swap_cluster_info *info) info->data =0 ; } -static inline bool cluster_is_huge (struct swap_cluster_info *info) -{ -if (IS_ENABLED (CONFIG_THP_SWAP )) -return info->flags &CLUSTER_FLAG_HUGE ; -return false; -} - -static inline void cluster_clear_huge (struct swap_cluster_info *info) -{ - info->flags &= ~CLUSTER_FLAG_HUGE ; -} -static inline struct swap_cluster_info *lock_cluster (struct swap_info_struct *si,unsigned long offset) {
mm/swapfile.c static int swap_alloc_cluster(struct swap_info_struct *si, swp_entry_t *slot) · line 1012
Quoted 5 lines
mm/swapfile.c +1
-1@@ -1021,7 +1009,7 @@ static int swap_alloc_cluster(struct swap_info_struct *si, swp_entry_t *slot) offset = idx *SWAPFILE_CLUSTER ; ci =lock_cluster (si, offset);alloc_cluster (si, idx); -cluster_set_count_flag (ci,SWAPFILE_CLUSTER ,CLUSTER_FLAG_HUGE ); +cluster_set_count_flag (ci,SWAPFILE_CLUSTER ,0 );
Maybe just use cluster_set_count() and kill cluster_set_count_flag().
Ryan Roberts
reviewerRe: [RFC PATCH v1 1/2] mm: swap: Remove CLUSTER_FLAG_HUGE from swap_cluster_info:flags
Quoted 103 lines · Kefeng Wang
Quoted 100 lines · Ryan Roberts
As preparation for supporting small-sized THP in the swap-out path, without first needing to split to order-0, Remove the
CLUSTER_FLAG_HUGE, which, when present, always implies PMD-sized THP, which is the same as the cluster size.The only use of the flag was to determine whether a swap entry refers to a single page or a PMD-sized THP in
swap_page_trans_huge_swapped(). Instead of relying on the flag, we now pass innr_pages, which originates from the folio's number of pages. This allows the logic to work for folios of any order.The one snag is that one of the
swap_page_trans_huge_swapped()call sites does not have the folio. But it was only being called there to avoid bothering to call__try_to_reclaim_swap()in some cases.__try_to_reclaim_swap()gets the folio and (via some other functions) callsswap_page_trans_huge_swapped(). So I've removed the problematic call site and believe the new logic should be equivalent.Removing
CLUSTER_FLAG_HUGEalso means we can removesplit_swap_cluster()which used to be called during folio splitting, since split_swap_cluster()'s only job was to remove the flag.Signed-off-by Ryan Roberts <[email protected]>---
include/linux/swap.h| 10 ----------
mm/huge_memory.c | 3 ---
mm/swapfile.c | 47 ++++++++------------------------------------3 files changed, 8 insertions(+), 52 deletions(-)include/linux/swap.h +0
-1@@ -259,7 +259,6 @@ struct swap_cluster_info { }; #define CLUSTER_FLAG_FREE 1 /* This cluster is free */ #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */ -#define CLUSTER_FLAG_HUGE 4/* This cluster is backing a transparent huge page */
/*
* We assign a cluster to each CPU, so each CPU can allocate swap entry frominclude/linux/swap.h +0
-9@@ -595,15 +594,6 @@ static inline int add_swap_extent(struct swap_info_struct *sis, } #endif /* CONFIG_SWAP */ -#ifdef CONFIG_THP_SWAP -extern int split_swap_cluster (swp_entry_t entry); -#else -static inline int split_swap_cluster (swp_entry_t entry) -{ - return 0; -} -#endif - #ifdef CONFIG_MEMCG static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg){
mm/huge_memory.c +0
-3@@ -2597,9 +2597,6 @@ static void __split_huge_page(struct page *page, structlist_head *list, shmem_uncharge(head->mapping->host, nr_dropped); remap_page(folio, nr); - if (folio_test_swapcache(folio)) - split_swap_cluster(folio->swap); - for (i = 0; i < nr; i++) { struct page *subpage = head + i;if (subpage == page)
mm/swapfile.c +0
-12@@ -342,18 +342,6 @@ static inline void cluster_set_null(struct swap_cluster_info *info) info->data = 0; } -static inline bool cluster_is_huge (struct swap_cluster_info *info) -{ - if (IS_ENABLED(CONFIG_THP_SWAP)) - return info->flags & CLUSTER_FLAG_HUGE; - return false; -} - -static inline void cluster_clear_huge (struct swap_cluster_info *info) -{ - info->flags &= ~CLUSTER_FLAG_HUGE; -} -static inline struct
swap_cluster_info*lock_cluster(structswap_info_struct*si,
unsigned long offset)
{mm/swapfile.c +0
-0@@ -1021,7 +1009,7 @@ static int swap_alloc_cluster(struct swap_info_struct
Quoted 8 lines
mm/swapfile.c static int swap_alloc_cluster(struct swap_info_struct *si, swp_entry_t *slot) · line 1012
Quoted 6 lines
mm/swapfile.c +1
-1*si, swp_entry_t *slot) offset = idx * SWAPFILE_CLUSTER; ci = lock_cluster(si, offset); alloc_cluster(si, idx); - cluster_set_count_flag(ci, SWAPFILE_CLUSTER, CLUSTER_FLAG_HUGE); + cluster_set_count_flag(ci, SWAPFILE_CLUSTER, 0);Maybe just use
cluster_set_count()and killcluster_set_count_flag().
Yep good point. I'll do this in the next version - thanks!
Ryan Roberts
reviewerRe: [RFC PATCH v1 1/2] mm: swap: Remove CLUSTER_FLAG_HUGE from swap_cluster_info:flags
In reply to Kefeng WangQuoted 103 lines · Kefeng Wang
Quoted 100 lines · Ryan Roberts
As preparation for supporting small-sized THP in the swap-out path, without first needing to split to order-0, Remove the
CLUSTER_FLAG_HUGE, which, when present, always implies PMD-sized THP, which is the same as the cluster size.The only use of the flag was to determine whether a swap entry refers to a single page or a PMD-sized THP in
swap_page_trans_huge_swapped(). Instead of relying on the flag, we now pass innr_pages, which originates from the folio's number of pages. This allows the logic to work for folios of any order.The one snag is that one of the
swap_page_trans_huge_swapped()call sites does not have the folio. But it was only being called there to avoid bothering to call__try_to_reclaim_swap()in some cases.__try_to_reclaim_swap()gets the folio and (via some other functions) callsswap_page_trans_huge_swapped(). So I've removed the problematic call site and believe the new logic should be equivalent.Removing
CLUSTER_FLAG_HUGEalso means we can removesplit_swap_cluster()which used to be called during folio splitting, since split_swap_cluster()'s only job was to remove the flag.Signed-off-by Ryan Roberts <[email protected]>---
include/linux/swap.h| 10 ----------
mm/huge_memory.c | 3 ---
mm/swapfile.c | 47 ++++++++------------------------------------3 files changed, 8 insertions(+), 52 deletions(-)include/linux/swap.h +0
-1@@ -259,7 +259,6 @@ struct swap_cluster_info { }; #define CLUSTER_FLAG_FREE 1 /* This cluster is free */ #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */ -#define CLUSTER_FLAG_HUGE 4/* This cluster is backing a transparent huge page */
/*
* We assign a cluster to each CPU, so each CPU can allocate swap entry frominclude/linux/swap.h +0
-9@@ -595,15 +594,6 @@ static inline int add_swap_extent(struct swap_info_struct *sis, } #endif /* CONFIG_SWAP */ -#ifdef CONFIG_THP_SWAP -extern int split_swap_cluster (swp_entry_t entry); -#else -static inline int split_swap_cluster (swp_entry_t entry) -{ - return 0; -} -#endif - #ifdef CONFIG_MEMCG static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg){
mm/huge_memory.c +0
-3@@ -2597,9 +2597,6 @@ static void __split_huge_page(struct page *page, structlist_head *list, shmem_uncharge(head->mapping->host, nr_dropped); remap_page(folio, nr); - if (folio_test_swapcache(folio)) - split_swap_cluster(folio->swap); - for (i = 0; i < nr; i++) { struct page *subpage = head + i;if (subpage == page)
mm/swapfile.c +0
-12@@ -342,18 +342,6 @@ static inline void cluster_set_null(struct swap_cluster_info *info) info->data = 0; } -static inline bool cluster_is_huge (struct swap_cluster_info *info) -{ - if (IS_ENABLED(CONFIG_THP_SWAP)) - return info->flags & CLUSTER_FLAG_HUGE; - return false; -} - -static inline void cluster_clear_huge (struct swap_cluster_info *info) -{ - info->flags &= ~CLUSTER_FLAG_HUGE; -} -static inline struct
swap_cluster_info*lock_cluster(structswap_info_struct*si,
unsigned long offset)
{mm/swapfile.c +0
-0@@ -1021,7 +1009,7 @@ static int swap_alloc_cluster(struct swap_info_struct
Quoted 8 lines
mm/swapfile.c static int swap_alloc_cluster(struct swap_info_struct *si, swp_entry_t *slot) · line 1012
Quoted 6 lines
mm/swapfile.c +1
-1*si, swp_entry_t *slot) offset = idx * SWAPFILE_CLUSTER; ci = lock_cluster(si, offset); alloc_cluster(si, idx); - cluster_set_count_flag(ci, SWAPFILE_CLUSTER, CLUSTER_FLAG_HUGE); + cluster_set_count_flag(ci, SWAPFILE_CLUSTER, 0);Maybe just use
cluster_set_count()and killcluster_set_count_flag().
Yes, good point - I'll do this in the next version. Thanks!
Ryan Roberts
reviewerRFC PATCH 2/2 mm: swap: Swap-out small-sized THP without splitting
In reply to Ryan RobertsThe upcoming anonymous small-sized THP feature enables performance improvements by allocating large folios for anonymous memory. However I've observed that on an arm64 system running a parallel workload (e.g. kernel compilation) across many cores, under high memory pressure, the speed regresses. This is due to bottlenecking on the increased number of TLBIs added due to all the extra folio splitting.
Therefore, solve this regression by adding support for swapping out small-sized THP without needing to split the folio, just like is already done for PMD-sized THP. This change only applies when CONFIG_THP_SWAP is enabled, and when the swap backing store is a non-rotating block device
- these are the same constraints as for the existing PMD-sized THP swap-out support.
Note that no attempt is made to swap-in THP here - this is still done page-by-page, like for PMD-sized THP.
The main change here is to improve the swap entry allocator so that it can allocate any power-of-2 number of contiguous entries between [4, (1 << PMD_ORDER)]. This is done by allocating a cluster for each distinct order and allocating sequentially from it until the cluster is full. This ensures that we don't need to search the map and we get no fragmentation due to alignment padding for different orders in the cluster. If there is no current cluster for a given order, we attempt to allocate a free cluster from the list. If there are no free clusters, we fail the allocation and the caller falls back to splitting the folio and allocates individual entries (as per existing PMD-sized THP fallback).
As far as I can tell, this should not cause any extra fragmentation concerns, given how similar it is to the existing PMD-sized THP allocation mechanism. There will be up to (PMD_ORDER-1) clusters in concurrent use though. In practice, the number of orders in use will be small though.
--- include/linux/swap.h | 7 ++++++ mm/swapfile.c | 60 +++++++++++++++++++++++++++++++++----------- mm/vmscan.c | 10 +++++--- 3 files changed, 59 insertions(+), 18 deletions(-)
include/linux/swap.h +7 -0
@@ -320,6 +320,13 @@ struct swap_info_struct {*/ struct work_struct discard_work;/* discard worker */ struct swap_cluster_list discard_clusters;/* discard clusters list */ +unsigned int large_next[PMD_ORDER ];/* +* next free offset within current +* allocation cluster for large +* folios, or UINT_MAX if no current +* cluster. Index is (order - 1). +* Only when cluster_info is used. +*/ struct plist_node avail_lists[];/* * entries in swap_avail_heads, one * entry per node.
mm/swapfile.c +46 -14
101 lines not on this page · raw message
mm/vmscan.c +6 -4
@@ -1212,11 +1212,13 @@ static unsigned int shrink_folio_list(struct list_head *folio_list,if (!can_split_folio (folio,NULL )) goto activate_locked;/* -* Split folios without a PMD map right -* away. Chances are some or all of the -* tail pages can be freed without IO. +* Split PMD-mappable folios without a +* PMD map right away. Chances are some +* or all of the tail pages can be freed +* without IO. */ -if (!folio_entire_mapcount (folio) && +if (folio_test_pmd_mappable (folio) && + !folio_entire_mapcount (folio) &&split_folio_to_list (folio, folio_list)) goto activate_locked;
--
2.25.1
kernel test robot
build botRe: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting
kernel test robot noticed the following build errors:
Hi Ryan,
[This is a private test report for your RFC patch.] kernel test robot noticed the following build errors:
[auto build test ERROR on akpm-mm/mm-everything]
[also build test ERROR on linus/master v6.6-rc5 next-20231010] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Ryan-Roberts/mm-swap-Remove-CLUSTER_FLAG_HUGE-from-swap_cluster_info-flags/20231010-222442 base: https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git mm-everything patch link: https://lore.kernel.org/r/20231010142111.3997780-3-ryan.roberts%40arm.com patch subject: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting config: powerpc-allyesconfig (https://download.01.org/0day-ci/archive/20231011/[email protected]/config) compiler: powerpc64-linux-gcc (GCC) 13.2.0 reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20231011/[email protected]/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
All errors (new ones prefixed by >>):
scripts/genksyms/parse.y: warning: 9 shift/reduce conflicts [-Wconflicts-sr] scripts/genksyms/parse.y: warning: 5 reduce/reduce conflicts [-Wconflicts-rr] scripts/genksyms/parse.y: note: rerun with option '-Wcounterexamples' to generate conflict counterexamples In file included from include/linux/suspend.h:5,
from arch/powerpc/kernel/asm-offsets.c:21:
>> include/linux/swap.h:323:22: error: variably modified 'large_next' at file scope
323 | unsigned int large_next[PMD_ORDER]; /*
| ^~~~~~~~~~
make[3]: *** [scripts/Makefile.build:116: arch/powerpc/kernel/asm-offsets.s] Error 1 make[3]: Target 'prepare' not remade because of errors.
make[2]: *** [Makefile:1202: prepare0] Error 2
make[2]: Target 'prepare' not remade because of errors.
make[1]: *** [Makefile:234: __sub-make] Error 2
make[1]: Target 'prepare' not remade because of errors.
make: *** [Makefile:234: __sub-make] Error 2
make: Target 'prepare' not remade because of errors.
vim +/large_next +323 include/linux/swap.h
277
278 /*
279 * The in-memory structure used to track swap areas.
280 */
281 struct swap_info_struct {
282 struct percpu_ref users; /* indicate and keep swap device valid. */
283 unsigned long flags; /* SWP_USED etc: see above */
284 signed short prio; /* swap priority of this type */
285 struct plist_node list; /* entry in swap_active_head */
286 signed char type; /* strange name for an index */
287 unsigned int max; /* extent of the swap_map */
288 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
289 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
290 struct swap_cluster_list free_clusters; /* free clusters list */
291 unsigned int lowest_bit; /* index of first free in swap_map */
292 unsigned int highest_bit; /* index of last free in swap_map */
293 unsigned int pages; /* total of usable pages of swap */
294 unsigned int inuse_pages; /* number of those currently in use */
295 unsigned int cluster_next; /* likely index for next allocation */
296 unsigned int cluster_nr; /* countdown to next cluster search */
297 unsigned int __percpu *cluster_next_cpu; /*percpu index for next allocation */
298 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
299 struct rb_root swap_extent_root;/* root of the swap extent rbtree */
300 struct block_device *bdev; /* swap device or bdev of swap file */
301 struct file *swap_file; /* seldom referenced */
302 unsigned int old_block_size; /* seldom referenced */
303 struct completion comp; /* seldom referenced */
304 spinlock_t lock; /*
305 * protect map scan related fields like
306 * swap_map, lowest_bit, highest_bit,
307 * inuse_pages, cluster_next,
308 * cluster_nr, lowest_alloc,
309 * highest_alloc, free/discard cluster
310 * list. other fields are only changed
311 * at swapon/swapoff, so are protected
312 * by swap_lock. changing flags need
313 * hold this lock and swap_lock. If
314 * both locks need hold, hold swap_lock
315 * first.
316 */
317 spinlock_t cont_lock; /*
318 * protect swap count continuation page
319 * list.
320 */
321 struct work_struct discard_work; /* discard worker */
322 struct swap_cluster_list discard_clusters; /* discard clusters list */
> 323 unsigned int large_next[PMD_ORDER]; /* 324 * next free offset within current 325 * allocation cluster for large 326 * folios, or UINT_MAX if no current 327 * cluster. Index is (order - 1). 328 * Only when cluster_info is used. 329 */ 330 struct plist_node avail_lists[]; /* 331 * entries in swap_avail_heads, one 332 * entry per node. 333 * Must be last as the number of the 334 * array is nr_node_ids, which is not 335 * a fixed value so have to allocate 336 * dynamically. 337 * And it has to be an array so that 338 * plist_for_each_* can work. 339 */
340 };
341
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki
Ryan Roberts
reviewerRe: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting
In reply to Ryan RobertsQuoted 46 lines · Ryan Roberts
The upcoming anonymous small-sized THP feature enables performance improvements by allocating large folios for anonymous memory. However I've observed that on an arm64 system running a parallel workload (e.g. kernel compilation) across many cores, under high memory pressure, the speed regresses. This is due to bottlenecking on the increased number of TLBIs added due to all the extra folio splitting.
Therefore, solve this regression by adding support for swapping out small-sized THP without needing to split the folio, just like is already done for PMD-sized THP. This change only applies when
CONFIG_THP_SWAPis enabled, and when the swap backing store is a non-rotating block device
- these are the same constraints as for the existing PMD-sized THP swap-out support.Note that no attempt is made to swap-in THP here - this is still done page-by-page, like for PMD-sized THP.
The main change here is to improve the swap entry allocator so that it can allocate any power-of-2 number of contiguous entries between [4, (1 <<
PMD_ORDER)]. This is done by allocating a cluster for each distinct order and allocating sequentially from it until the cluster is full. This ensures that we don't need to search the map and we get no fragmentation due to alignment padding for different orders in the cluster. If there is no current cluster for a given order, we attempt to allocate a free cluster from the list. If there are no free clusters, we fail the allocation and the caller falls back to splitting the folio and allocates individual entries (as per existing PMD-sized THP fallback).As far as I can tell, this should not cause any extra fragmentation concerns, given how similar it is to the existing PMD-sized THP allocation mechanism. There will be up to (PMD_ORDER-1) clusters in concurrent use though. In practice, the number of orders in use will be small though.
Signed-off-by Ryan Roberts <[email protected]>--- include/linux/swap.h | 7 ++++++ mm/swapfile.c | 60 +++++++++++++++++++++++++++++++++----------- mm/vmscan.c | 10 +++++--- 3 files changed, 59 insertions(+), 18 deletions(-)include/linux/swap.h +0
-0
include/linux/swap.h struct swap_info_struct { · line 323
Quoted 4 lines
include/linux/swap.h +1
-0@@ -320,6 +320,13 @@ struct swap_info_struct {*/ struct work_struct discard_work;/* discard worker */ struct swap_cluster_list discard_clusters;/* discard clusters list */ +unsigned int large_next[PMD_ORDER ];/*
Oh dear, I've tripped over this twice in a week now... PMD_ORDER is not a compile-time const on powerpc, so results in build fail. This would have to be allocated at init time, I guess.
Quoted 144 lines
include/linux/swap.h +6
-0+* next free offset within current +* allocation cluster for large +* folios, or UINT_MAX if no current +* cluster. Index is (order - 1). +* Only when cluster_info is used. +*/ struct plist_node avail_lists[];/* * entries in swap_avail_heads, one * entry per node. mm/swapfile.c +46
-14@@ -987,8 +987,10 @@ static int scan_swap_map_slots(struct swap_info_struct *si,return n_ret; } -static int swap_alloc_cluster (struct swap_info_struct *si,swp_entry_t *slot) +static int swap_alloc_large (struct swap_info_struct *si,swp_entry_t *slot, +unsigned int nr_pages) { +int order;unsigned long idx;struct swap_cluster_info *ci;unsigned long offset;@@ -1002,20 +1004,47 @@ static int swap_alloc_cluster(struct swap_info_struct *si, swp_entry_t *slot)return 0 ; } -if (cluster_list_empty (&si->free_clusters)) -return 0 ; +VM_WARN_ON (nr_pages <2 ); +VM_WARN_ON (nr_pages >SWAPFILE_CLUSTER ); +VM_WARN_ON (!is_power_of_2 (nr_pages)); - idx =cluster_list_first (&si->free_clusters); - offset = idx *SWAPFILE_CLUSTER ; - ci =lock_cluster (si, offset); -alloc_cluster (si, idx); -cluster_set_count_flag (ci,SWAPFILE_CLUSTER ,0 ); + order =ilog2 (nr_pages); + offset = si->large_next[order -1 ]; + +if (offset ==UINT_MAX ) { +if (cluster_list_empty (&si->free_clusters)) +return 0 ; -memset (si->swap_map + offset,SWAP_HAS_CACHE ,SWAPFILE_CLUSTER ); + idx =cluster_list_first (&si->free_clusters); + offset = idx *SWAPFILE_CLUSTER ; + + ci =lock_cluster (si, offset); +alloc_cluster (si, idx); +cluster_set_count_flag (ci,SWAPFILE_CLUSTER ,0 ); + +/* +* If scan_swap_map_slots() can't find a free cluster, it will +* check si->swap_map directly. To make sure this standby +* cluster isn't taken by scan_swap_map_slots(), mark the swap +* entries bad (occupied). (same approach as discard). +*/ +memset (si->swap_map + offset + nr_pages,SWAP_MAP_BAD , +SWAPFILE_CLUSTER - nr_pages); + }else { + idx = offset /SWAPFILE_CLUSTER ; + ci =lock_cluster (si, offset); + } + +memset (si->swap_map + offset,SWAP_HAS_CACHE , nr_pages);unlock_cluster (ci); -swap_range_alloc (si, offset,SWAPFILE_CLUSTER ); +swap_range_alloc (si, offset, nr_pages); *slot =swp_entry (si->type, offset); + offset += nr_pages; +if (idx != offset /SWAPFILE_CLUSTER ) + offset =UINT_MAX ; + si->large_next[order -1 ] = offset; +return 1 ; }@@ -1041,7 +1070,7 @@ int get_swap_pages(int n_goal, swp_entry_t swp_entries[], int entry_size)int node;/* Only single cluster request supported */ -WARN_ON_ONCE (n_goal >1 && size ==SWAPFILE_CLUSTER ); +WARN_ON_ONCE (n_goal >1 && size >1 );spin_lock (&swap_avail_lock);@@ -1078,14 +1107,14 @@ int get_swap_pages(int n_goal, swp_entry_t swp_entries[], int entry_size)spin_unlock (&si->lock); goto nextsi; } -if (size ==SWAPFILE_CLUSTER ) { +if (size >1 ) {if (si->flags &SWP_BLKDEV ) - n_ret =swap_alloc_cluster (si, swp_entries); + n_ret =swap_alloc_large (si, swp_entries, size); }else n_ret =scan_swap_map_slots (si,SWAP_HAS_CACHE , n_goal, swp_entries);spin_unlock (&si->lock); -if (n_ret || size ==SWAPFILE_CLUSTER ) +if (n_ret || size >1 ) goto check_out;cond_resched ();@@ -2725,6 +2754,9 @@ static struct swap_info_struct *alloc_swap_info(void)spin_lock_init (&p->cont_lock);init_completion (&p->comp); +for (i =0 ; i <ARRAY_SIZE (p->large_next); i++) + p->large_next[i] =UINT_MAX ; +return p; }mm/vmscan.c +6
-5@@ -1212,11 +1212,13 @@ static unsigned int shrink_folio_list(struct list_head *folio_list,if (!can_split_folio (folio,NULL )) goto activate_locked;/* -* Split folios without a PMD map right -* away. Chances are some or all of the -* tail pages can be freed without IO. +* Split PMD-mappable folios without a +* PMD map right away. Chances are some +* or all of the tail pages can be freed +* without IO. */ -if (!folio_entire_mapcount (folio) && +if (folio_test_pmd_mappable (folio) && + !folio_entire_mapcount (folio) &&split_folio_to_list (folio, folio_list)) goto activate_locked; --2.25.1
Re: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting
In reply to Ryan RobertsQuoted 46 lines · Ryan Roberts
The upcoming anonymous small-sized THP feature enables performance improvements by allocating large folios for anonymous memory. However I've observed that on an arm64 system running a parallel workload (e.g. kernel compilation) across many cores, under high memory pressure, the speed regresses. This is due to bottlenecking on the increased number of TLBIs added due to all the extra folio splitting.
Therefore, solve this regression by adding support for swapping out small-sized THP without needing to split the folio, just like is already done for PMD-sized THP. This change only applies when
CONFIG_THP_SWAPis enabled, and when the swap backing store is a non-rotating block device
- these are the same constraints as for the existing PMD-sized THP swap-out support.Note that no attempt is made to swap-in THP here - this is still done page-by-page, like for PMD-sized THP.
The main change here is to improve the swap entry allocator so that it can allocate any power-of-2 number of contiguous entries between [4, (1 <<
PMD_ORDER)]. This is done by allocating a cluster for each distinct order and allocating sequentially from it until the cluster is full. This ensures that we don't need to search the map and we get no fragmentation due to alignment padding for different orders in the cluster. If there is no current cluster for a given order, we attempt to allocate a free cluster from the list. If there are no free clusters, we fail the allocation and the caller falls back to splitting the folio and allocates individual entries (as per existing PMD-sized THP fallback).As far as I can tell, this should not cause any extra fragmentation concerns, given how similar it is to the existing PMD-sized THP allocation mechanism. There will be up to (PMD_ORDER-1) clusters in concurrent use though. In practice, the number of orders in use will be small though.
Signed-off-by Ryan Roberts <[email protected]>--- include/linux/swap.h | 7 ++++++ mm/swapfile.c | 60 +++++++++++++++++++++++++++++++++----------- mm/vmscan.c | 10 +++++--- 3 files changed, 59 insertions(+), 18 deletions(-)include/linux/swap.h +0
-0
include/linux/swap.h struct swap_info_struct { · line 329
Quoted 10 lines
include/linux/swap.h +7
-0@@ -320,6 +320,13 @@ struct swap_info_struct {*/ struct work_struct discard_work;/* discard worker */ struct swap_cluster_list discard_clusters;/* discard clusters list */ +unsigned int large_next[PMD_ORDER ];/* +* next free offset within current +* allocation cluster for large +* folios, or UINT_MAX if no current +* cluster. Index is (order - 1). +* Only when cluster_info is used. +*/
I think that it is better to make this per-CPU. That is, extend the percpu_cluster mechanism. Otherwise, we may have scalability issue.
And this should be enclosed in CONFIG_THP_SWAP.
include/linux/swap.h struct swap_info_struct { · line 332
Quoted 3 lines
include/linux/swap.h +0
-0struct plist_node avail_lists[];/* * entries in swap_avail_heads, one * entry per node.
--
Best Regards,
Huang, Ying
Ryan Roberts
reviewerRe: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting
Quoted 53 lines · Ying
Quoted 51 lines · Ryan Roberts
The upcoming anonymous small-sized THP feature enables performance improvements by allocating large folios for anonymous memory. However I've observed that on an arm64 system running a parallel workload (e.g. kernel compilation) across many cores, under high memory pressure, the speed regresses. This is due to bottlenecking on the increased number of TLBIs added due to all the extra folio splitting.
Therefore, solve this regression by adding support for swapping out small-sized THP without needing to split the folio, just like is already done for PMD-sized THP. This change only applies when
CONFIG_THP_SWAPis enabled, and when the swap backing store is a non-rotating block device
- these are the same constraints as for the existing PMD-sized THP swap-out support.Note that no attempt is made to swap-in THP here - this is still done page-by-page, like for PMD-sized THP.
The main change here is to improve the swap entry allocator so that it can allocate any power-of-2 number of contiguous entries between [4, (1 <<
PMD_ORDER)]. This is done by allocating a cluster for each distinct order and allocating sequentially from it until the cluster is full. This ensures that we don't need to search the map and we get no fragmentation due to alignment padding for different orders in the cluster. If there is no current cluster for a given order, we attempt to allocate a free cluster from the list. If there are no free clusters, we fail the allocation and the caller falls back to splitting the folio and allocates individual entries (as per existing PMD-sized THP fallback).As far as I can tell, this should not cause any extra fragmentation concerns, given how similar it is to the existing PMD-sized THP allocation mechanism. There will be up to (PMD_ORDER-1) clusters in concurrent use though. In practice, the number of orders in use will be small though.
Signed-off-by Ryan Roberts <[email protected]>--- include/linux/swap.h | 7 ++++++ mm/swapfile.c | 60 +++++++++++++++++++++++++++++++++----------- mm/vmscan.c | 10 +++++--- 3 files changed, 59 insertions(+), 18 deletions(-)include/linux/swap.h +2
-0@@ -320,6 +320,13 @@ struct swap_info_struct {*/ struct work_struct discard_work;/* discard worker */ struct swap_cluster_list discard_clusters;/* discard clusters list */ +unsigned int large_next[PMD_ORDER ];/* +* next free offset within current
Quoted 8 lines
include/linux/swap.h struct swap_info_struct { · line 329
Quoted 5 lines
include/linux/swap.h +5
-0+* allocation cluster for large +* folios, or UINT_MAX if no current +* cluster. Index is (order - 1). +* Only when cluster_info is used. +*/ I think that it is better to make this per-CPU. That is, extend the
percpu_clustermechanism. Otherwise, we may have scalability issue.
Is your concern that the swap_info spinlock will get too contended as its currently written? From briefly looking at percpu_cluster, it looks like that spinlock is always held when accessing the per-cpu structures - presumably that's what's disabling preemption and making sure the thread is not migrated? So I'm not sure what the benefit is currently? Surely you want to just disable preemption but not hold the lock? I'm sure I've missed something crucial...
Quoted 2 lines
And this should be enclosed in
CONFIG_THP_SWAP.
Yes, I'll fix this in the next version.
Thanks for the review!
Quoted 8 lines
include/linux/swap.h struct swap_info_struct { · line 332
Quoted 4 lines
include/linux/swap.h +0
-0struct plist_node avail_lists[];/* * entries in swap_avail_heads, one * entry per node. --
Best Regards,
Huang, Ying
Ryan Roberts
reviewerRe: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting
Quoted 63 lines · Ryan Roberts
Quoted 53 lines · Ying
Quoted 51 lines · Ryan Roberts
The upcoming anonymous small-sized THP feature enables performance improvements by allocating large folios for anonymous memory. However I've observed that on an arm64 system running a parallel workload (e.g. kernel compilation) across many cores, under high memory pressure, the speed regresses. This is due to bottlenecking on the increased number of TLBIs added due to all the extra folio splitting.
Therefore, solve this regression by adding support for swapping out small-sized THP without needing to split the folio, just like is already done for PMD-sized THP. This change only applies when
CONFIG_THP_SWAPis enabled, and when the swap backing store is a non-rotating block device
- these are the same constraints as for the existing PMD-sized THP swap-out support.Note that no attempt is made to swap-in THP here - this is still done page-by-page, like for PMD-sized THP.
The main change here is to improve the swap entry allocator so that it can allocate any power-of-2 number of contiguous entries between [4, (1 <<
PMD_ORDER)]. This is done by allocating a cluster for each distinct order and allocating sequentially from it until the cluster is full. This ensures that we don't need to search the map and we get no fragmentation due to alignment padding for different orders in the cluster. If there is no current cluster for a given order, we attempt to allocate a free cluster from the list. If there are no free clusters, we fail the allocation and the caller falls back to splitting the folio and allocates individual entries (as per existing PMD-sized THP fallback).As far as I can tell, this should not cause any extra fragmentation concerns, given how similar it is to the existing PMD-sized THP allocation mechanism. There will be up to (PMD_ORDER-1) clusters in concurrent use though. In practice, the number of orders in use will be small though.
Signed-off-by Ryan Roberts <[email protected]>--- include/linux/swap.h | 7 ++++++ mm/swapfile.c | 60 +++++++++++++++++++++++++++++++++----------- mm/vmscan.c | 10 +++++--- 3 files changed, 59 insertions(+), 18 deletions(-)include/linux/swap.h +2
-0@@ -320,6 +320,13 @@ struct swap_info_struct {*/ struct work_struct discard_work;/* discard worker */ struct swap_cluster_list discard_clusters;/* discard clusters list */ +unsigned int large_next[PMD_ORDER ];/* +* next free offset within current Quoted 8 lines
include/linux/swap.h struct swap_info_struct { · line 329
Quoted 5 lines
include/linux/swap.h +5
-0+* allocation cluster for large +* folios, or UINT_MAX if no current +* cluster. Index is (order - 1). +* Only when cluster_info is used. +*/ I think that it is better to make this per-CPU. That is, extend the
percpu_clustermechanism. Otherwise, we may have scalability issue.
Quoted 6 lines
Is your concern that the
swap_infospinlock will get too contended as its currently written? From briefly looking atpercpu_cluster, it looks like that spinlock is always held when accessing the per-cpu structures - presumably that's what's disabling preemption and making sure the thread is not migrated? So I'm not sure what the benefit is currently? Surely you want to just disable preemption but not hold the lock? I'm sure I've missed something crucial...
I looked a bit further at how to implement what you are suggesting. get_swap_pages() is currently taking the swap_info lock which it needs to check and update some other parts of the swap_info - I'm not sure that part can be removed. swap_alloc_large() (my new function) is not doing an awful lot of work, so I'm not convinced that you would save too much by releasing the lock for that part. In contrast there is a lot more going on in scan_swap_map_slots() so there is more benefit to releasing the lock and using the percpu stuff - correct me if I've missunderstood.
As an alternative approach, perhaps it makes more sense to beef up the caching layer in swap_slots.c to handle large folios too? Then you avoid taking the swap_info lock at all most of the time, like you currently do for single entry allocations.
What do you think?
Quoted 17 lines
Quoted 3 lines
And this should be enclosed in
CONFIG_THP_SWAP.Yes, I'll fix this in the next version.
Thanks for the review!
Quoted 8 lines
include/linux/swap.h struct swap_info_struct { · line 332
Quoted 4 lines
include/linux/swap.h +0
-0struct plist_node avail_lists[];/* * entries in swap_avail_heads, one * entry per node. --
Best Regards,
Huang, Ying
Re: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting
Quoted 80 lines · Ryan Roberts
Quoted 63 lines · Ryan Roberts
Quoted 53 lines · Ying
Quoted 51 lines · Ryan Roberts
The upcoming anonymous small-sized THP feature enables performance improvements by allocating large folios for anonymous memory. However I've observed that on an arm64 system running a parallel workload (e.g. kernel compilation) across many cores, under high memory pressure, the speed regresses. This is due to bottlenecking on the increased number of TLBIs added due to all the extra folio splitting.
Therefore, solve this regression by adding support for swapping out small-sized THP without needing to split the folio, just like is already done for PMD-sized THP. This change only applies when
CONFIG_THP_SWAPis enabled, and when the swap backing store is a non-rotating block device
- these are the same constraints as for the existing PMD-sized THP swap-out support.Note that no attempt is made to swap-in THP here - this is still done page-by-page, like for PMD-sized THP.
The main change here is to improve the swap entry allocator so that it can allocate any power-of-2 number of contiguous entries between [4, (1 <<
PMD_ORDER)]. This is done by allocating a cluster for each distinct order and allocating sequentially from it until the cluster is full. This ensures that we don't need to search the map and we get no fragmentation due to alignment padding for different orders in the cluster. If there is no current cluster for a given order, we attempt to allocate a free cluster from the list. If there are no free clusters, we fail the allocation and the caller falls back to splitting the folio and allocates individual entries (as per existing PMD-sized THP fallback).As far as I can tell, this should not cause any extra fragmentation concerns, given how similar it is to the existing PMD-sized THP allocation mechanism. There will be up to (PMD_ORDER-1) clusters in concurrent use though. In practice, the number of orders in use will be small though.
Signed-off-by Ryan Roberts <[email protected]>--- include/linux/swap.h | 7 ++++++ mm/swapfile.c | 60 +++++++++++++++++++++++++++++++++----------- mm/vmscan.c | 10 +++++--- 3 files changed, 59 insertions(+), 18 deletions(-)include/linux/swap.h +2
-0@@ -320,6 +320,13 @@ struct swap_info_struct {*/ struct work_struct discard_work;/* discard worker */ struct swap_cluster_list discard_clusters;/* discard clusters list */ +unsigned int large_next[PMD_ORDER ];/* +* next free offset within current Quoted 8 lines
include/linux/swap.h struct swap_info_struct { · line 329
Quoted 5 lines
include/linux/swap.h +5
-0+* allocation cluster for large +* folios, or UINT_MAX if no current +* cluster. Index is (order - 1). +* Only when cluster_info is used. +*/ I think that it is better to make this per-CPU. That is, extend the
percpu_clustermechanism. Otherwise, we may have scalability issue.Quoted 6 lines
Is your concern that the
swap_infospinlock will get too contended as its currently written? From briefly looking atpercpu_cluster, it looks like that spinlock is always held when accessing the per-cpu structures - presumably that's what's disabling preemption and making sure the thread is not migrated? So I'm not sure what the benefit is currently? Surely you want to just disable preemption but not hold the lock? I'm sure I've missed something crucial...I looked a bit further at how to implement what you are suggesting.
get_swap_pages()is currently taking theswap_infolock which it needs to check and update some other parts of theswap_info- I'm not sure that part can be removed.swap_alloc_large()(my new function) is not doing an awful lot of work, so I'm not convinced that you would save too much by releasing the lock for that part. In contrast there is a lot more going on inscan_swap_map_slots()so there is more benefit to releasing the lock and using the percpu stuff - correct me if I've missunderstood.
Quoted 6 lines
As an alternative approach, perhaps it makes more sense to beef up the caching layer in
swap_slots.cto handle large folios too? Then you avoid taking theswap_infolock at all most of the time, like you currently do for single entry allocations.What do you think?
Sorry for late reply.
percpu_cluster is introduced in commit ebc2a1a69111 ("swap: make cluster allocation per-cpu"). Please check the changelog for why it's introduced. Sorry about my incorrect memory about scalability. percpu_cluster is introduced for disk performance mainly instead of scalability.
--
Best Regards,
Huang, Ying
Quoted 12 lines
Quoted 12 lines
Quoted 3 lines
And this should be enclosed in
CONFIG_THP_SWAP.Yes, I'll fix this in the next version.
Thanks for the review!
Quoted 4 lines
Quoted 4 lines
include/linux/swap.h +0
-0struct plist_node avail_lists[];/* * entries in swap_avail_heads, one * entry per node.
Ryan Roberts
reviewerRe: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting
Quoted 89 lines · Ying
Quoted 80 lines · Ryan Roberts
Quoted 63 lines · Ryan Roberts
Quoted 53 lines · Ying
Quoted 51 lines · Ryan Roberts
The upcoming anonymous small-sized THP feature enables performance improvements by allocating large folios for anonymous memory. However I've observed that on an arm64 system running a parallel workload (e.g. kernel compilation) across many cores, under high memory pressure, the speed regresses. This is due to bottlenecking on the increased number of TLBIs added due to all the extra folio splitting.
Therefore, solve this regression by adding support for swapping out small-sized THP without needing to split the folio, just like is already done for PMD-sized THP. This change only applies when
CONFIG_THP_SWAPis enabled, and when the swap backing store is a non-rotating block device
- these are the same constraints as for the existing PMD-sized THP swap-out support.Note that no attempt is made to swap-in THP here - this is still done page-by-page, like for PMD-sized THP.
The main change here is to improve the swap entry allocator so that it can allocate any power-of-2 number of contiguous entries between [4, (1 <<
PMD_ORDER)]. This is done by allocating a cluster for each distinct order and allocating sequentially from it until the cluster is full. This ensures that we don't need to search the map and we get no fragmentation due to alignment padding for different orders in the cluster. If there is no current cluster for a given order, we attempt to allocate a free cluster from the list. If there are no free clusters, we fail the allocation and the caller falls back to splitting the folio and allocates individual entries (as per existing PMD-sized THP fallback).As far as I can tell, this should not cause any extra fragmentation concerns, given how similar it is to the existing PMD-sized THP allocation mechanism. There will be up to (PMD_ORDER-1) clusters in concurrent use though. In practice, the number of orders in use will be small though.
Signed-off-by Ryan Roberts <[email protected]>--- include/linux/swap.h | 7 ++++++ mm/swapfile.c | 60 +++++++++++++++++++++++++++++++++----------- mm/vmscan.c | 10 +++++--- 3 files changed, 59 insertions(+), 18 deletions(-)include/linux/swap.h +2
-0@@ -320,6 +320,13 @@ struct swap_info_struct {*/ struct work_struct discard_work;/* discard worker */ struct swap_cluster_list discard_clusters;/* discard clusters list */ +unsigned int large_next[PMD_ORDER ];/* +* next free offset within current Quoted 8 lines
include/linux/swap.h struct swap_info_struct { · line 329
Quoted 5 lines
include/linux/swap.h +5
-0+* allocation cluster for large +* folios, or UINT_MAX if no current +* cluster. Index is (order - 1). +* Only when cluster_info is used. +*/ I think that it is better to make this per-CPU. That is, extend the
percpu_clustermechanism. Otherwise, we may have scalability issue.Quoted 6 lines
Is your concern that the
swap_infospinlock will get too contended as its currently written? From briefly looking atpercpu_cluster, it looks like that spinlock is always held when accessing the per-cpu structures - presumably that's what's disabling preemption and making sure the thread is not migrated? So I'm not sure what the benefit is currently? Surely you want to just disable preemption but not hold the lock? I'm sure I've missed something crucial...I looked a bit further at how to implement what you are suggesting.
get_swap_pages()is currently taking theswap_infolock which it needs to check and update some other parts of theswap_info- I'm not sure that part can be removed.swap_alloc_large()(my new function) is not doing an awful lot of work, so I'm not convinced that you would save too much by releasing the lock for that part. In contrast there is a lot more going on inscan_swap_map_slots()so there is more benefit to releasing the lock and using the percpu stuff - correct me if I've missunderstood.Quoted 6 lines
As an alternative approach, perhaps it makes more sense to beef up the caching layer in
swap_slots.cto handle large folios too? Then you avoid taking theswap_infolock at all most of the time, like you currently do for single entry allocations.What do you think?
Quoted 7 lines
Sorry for late reply.
percpu_clusteris introduced in commitebc2a1a69111("swap: make cluster allocation per-cpu"). Please check the changelog for why it's introduced. Sorry about my incorrect memory about scalability.percpu_clusteris introduced for disk performance mainly instead of scalability.
Thanks for the pointer. I'm not sure if you are still suggesting that I make my small-sized THP allocation mechanism per-cpu though?
I anticipate that by virtue of allocating multiple contiguous swap entries for a small-sized THP we already get a lot of the benefits that percpu_cluster gives order-0 allocations. (Although obviously it will only give contiguity matching the size of the THP rather than a full cluster). The downside of making this percpu would be keeping more free clusters tied up in the percpu caches, potentially causing a need to scan for free entries more often.
What's your view?
Thanks,
Ryan
Quoted 17 lines
--
Best Regards,
Huang, YingQuoted 12 lines
Quoted 12 lines
Quoted 3 lines
And this should be enclosed in
CONFIG_THP_SWAP.Yes, I'll fix this in the next version.
Thanks for the review!
Quoted 4 lines
Quoted 4 lines
include/linux/swap.h +0
-0struct plist_node avail_lists[];/* * entries in swap_avail_heads, one * entry per node.
Re: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting
Quoted 92 lines · Ryan Roberts
Quoted 91 lines · Ying
Quoted 80 lines · Ryan Roberts
Quoted 63 lines · Ryan Roberts
Quoted 53 lines · Ying
Quoted 51 lines · Ryan Roberts
The upcoming anonymous small-sized THP feature enables performance improvements by allocating large folios for anonymous memory. However I've observed that on an arm64 system running a parallel workload (e.g. kernel compilation) across many cores, under high memory pressure, the speed regresses. This is due to bottlenecking on the increased number of TLBIs added due to all the extra folio splitting.
Therefore, solve this regression by adding support for swapping out small-sized THP without needing to split the folio, just like is already done for PMD-sized THP. This change only applies when
CONFIG_THP_SWAPis enabled, and when the swap backing store is a non-rotating block device
- these are the same constraints as for the existing PMD-sized THP swap-out support.Note that no attempt is made to swap-in THP here - this is still done page-by-page, like for PMD-sized THP.
The main change here is to improve the swap entry allocator so that it can allocate any power-of-2 number of contiguous entries between [4, (1 <<
PMD_ORDER)]. This is done by allocating a cluster for each distinct order and allocating sequentially from it until the cluster is full. This ensures that we don't need to search the map and we get no fragmentation due to alignment padding for different orders in the cluster. If there is no current cluster for a given order, we attempt to allocate a free cluster from the list. If there are no free clusters, we fail the allocation and the caller falls back to splitting the folio and allocates individual entries (as per existing PMD-sized THP fallback).As far as I can tell, this should not cause any extra fragmentation concerns, given how similar it is to the existing PMD-sized THP allocation mechanism. There will be up to (PMD_ORDER-1) clusters in concurrent use though. In practice, the number of orders in use will be small though.
Signed-off-by Ryan Roberts <[email protected]>--- include/linux/swap.h | 7 ++++++ mm/swapfile.c | 60 +++++++++++++++++++++++++++++++++----------- mm/vmscan.c | 10 +++++--- 3 files changed, 59 insertions(+), 18 deletions(-)include/linux/swap.h +2
-0@@ -320,6 +320,13 @@ struct swap_info_struct {*/ struct work_struct discard_work;/* discard worker */ struct swap_cluster_list discard_clusters;/* discard clusters list */ +unsigned int large_next[PMD_ORDER ];/* +* next free offset within current Quoted 8 lines
include/linux/swap.h struct swap_info_struct { · line 329
Quoted 5 lines
include/linux/swap.h +5
-0+* allocation cluster for large +* folios, or UINT_MAX if no current +* cluster. Index is (order - 1). +* Only when cluster_info is used. +*/ I think that it is better to make this per-CPU. That is, extend the
percpu_clustermechanism. Otherwise, we may have scalability issue.Quoted 6 lines
Is your concern that the
swap_infospinlock will get too contended as its currently written? From briefly looking atpercpu_cluster, it looks like that spinlock is always held when accessing the per-cpu structures - presumably that's what's disabling preemption and making sure the thread is not migrated? So I'm not sure what the benefit is currently? Surely you want to just disable preemption but not hold the lock? I'm sure I've missed something crucial...I looked a bit further at how to implement what you are suggesting.
get_swap_pages()is currently taking theswap_infolock which it needs to check and update some other parts of theswap_info- I'm not sure that part can be removed.swap_alloc_large()(my new function) is not doing an awful lot of work, so I'm not convinced that you would save too much by releasing the lock for that part. In contrast there is a lot more going on inscan_swap_map_slots()so there is more benefit to releasing the lock and using the percpu stuff - correct me if I've missunderstood.Quoted 6 lines
As an alternative approach, perhaps it makes more sense to beef up the caching layer in
swap_slots.cto handle large folios too? Then you avoid taking theswap_infolock at all most of the time, like you currently do for single entry allocations.What do you think?
Sorry for late reply.
Quoted 8 lines
Quoted 5 lines
percpu_clusteris introduced in commitebc2a1a69111("swap: make cluster allocation per-cpu"). Please check the changelog for why it's introduced. Sorry about my incorrect memory about scalability.percpu_clusteris introduced for disk performance mainly instead of scalability.Thanks for the pointer. I'm not sure if you are still suggesting that I make my small-sized THP allocation mechanism per-cpu though?
Yes. I think that the reason for that we introduced percpu_cluster still applies now.
Quoted 4 lines
I anticipate that by virtue of allocating multiple contiguous swap entries for a small-sized THP we already get a lot of the benefits that
percpu_clustergives order-0 allocations. (Although obviously it will only give contiguity matching the size of the THP rather than a full cluster).
I think that you will still introduce "interleave disk access" when multiple CPU allocate from and write to swap device simultaneously. Right? Yes, 16KB block is better than 4KB block, but I don't think it solves the problem.
Quoted 3 lines
The downside of making this percpu would be keeping more free clusters tied up in the percpu caches, potentially causing a need to scan for free entries more often.
Yes. We may waste several MB swap space per-CPU. Is this a practical issue given the swap device capacity becomes larger and larger?
--
Best Regards,
Huang, Ying
kernel test robot
build botRe: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting
In reply to Ryan Robertskernel test robot noticed the following build errors:
Hi Ryan,
[This is a private test report for your RFC patch.] kernel test robot noticed the following build errors:
[auto build test ERROR on akpm-mm/mm-everything]
[also build test ERROR on linus/master v6.6-rc5 next-20231013] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Ryan-Roberts/mm-swap-Remove-CLUSTER_FLAG_HUGE-from-swap_cluster_info-flags/20231010-222442 base: https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git mm-everything patch link: https://lore.kernel.org/r/20231010142111.3997780-3-ryan.roberts%40arm.com patch subject: [RFC PATCH v1 2/2] mm: swap: Swap-out small-sized THP without splitting config: arm-randconfig-004-20231013 (https://download.01.org/0day-ci/archive/20231013/[email protected]/config) compiler: arm-linux-gnueabi-gcc (GCC) 13.2.0 reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20231013/[email protected]/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
All errors (new ones prefixed by >>):
In file included from include/linux/mm.h:29,
from include/linux/memcontrol.h:20,
from include/linux/swap.h:9,
from include/linux/suspend.h:5,
from include/linux/regulator/consumer.h:35,
from include/linux/i2c.h:19,
from drivers/iio/accel/mxc4005.c:9:
>> include/linux/pgtable.h:8:26: error: 'PMD_SHIFT' undeclared here (not in a function); did you mean 'P4D_SHIFT'?
8 | #define PMD_ORDER (PMD_SHIFT - PAGE_SHIFT)
| ^~~~~~~~~
include/linux/swap.h:323:33: note: in expansion of macro 'PMD_ORDER'
323 | unsigned int large_next[PMD_ORDER]; /*
| ^~~~~~~~~
drivers/iio/accel/mxc4005.c:472:36: warning: 'mxc4005_acpi_match' defined but not used [-Wunused-const-variable=]
472 | static const struct acpi_device_id mxc4005_acpi_match[] = {
| ^~~~~~~~~~~~~~~~~~
--
In file included from include/linux/mm.h:29,
from include/linux/memcontrol.h:20,
from include/linux/swap.h:9,
from include/linux/suspend.h:5,
from include/linux/regulator/consumer.h:35,
from include/linux/i2c.h:19,
from include/linux/iio/common/st_sensors.h:13,
from drivers/iio/accel/st_accel_core.c:20:
>> include/linux/pgtable.h:8:26: error: 'PMD_SHIFT' undeclared here (not in a function); did you mean 'P4D_SHIFT'? 8 | #define PMD_ORDER (PMD_SHIFT - PAGE_SHIFT)
| ^~~~~~~~~
include/linux/swap.h:323:33: note: in expansion of macro 'PMD_ORDER'
323 | unsigned int large_next[PMD_ORDER]; /*
| ^~~~~~~~~
--
In file included from include/linux/mm.h:29,
from include/linux/memcontrol.h:20,
from include/linux/swap.h:9,
from include/linux/suspend.h:5,
from include/linux/regulator/consumer.h:35,
from include/linux/i2c.h:19,
from drivers/iio/accel/mma9553.c:8:
>> include/linux/pgtable.h:8:26: error: 'PMD_SHIFT' undeclared here (not in a function); did you mean 'P4D_SHIFT'?
8 | #define PMD_ORDER (PMD_SHIFT - PAGE_SHIFT)
| ^~~~~~~~~
include/linux/swap.h:323:33: note: in expansion of macro 'PMD_ORDER'
323 | unsigned int large_next[PMD_ORDER]; /*
| ^~~~~~~~~
drivers/iio/accel/mma9553.c:1229:36: warning: 'mma9553_acpi_match' defined but not used [-Wunused-const-variable=]
1229 | static const struct acpi_device_id mma9553_acpi_match[] = {
| ^~~~~~~~~~~~~~~~~~
--
In file included from include/linux/mm.h:29,
from include/linux/memcontrol.h:20,
from include/linux/swap.h:9,
from include/linux/suspend.h:5,
from include/linux/regulator/consumer.h:35,
from include/linux/i2c.h:19,
from drivers/iio/accel/mxc6255.c:11:
>> include/linux/pgtable.h:8:26: error: 'PMD_SHIFT' undeclared here (not in a function); did you mean 'P4D_SHIFT'?
8 | #define PMD_ORDER (PMD_SHIFT - PAGE_SHIFT)
| ^~~~~~~~~
include/linux/swap.h:323:33: note: in expansion of macro 'PMD_ORDER'
323 | unsigned int large_next[PMD_ORDER]; /*
| ^~~~~~~~~
drivers/iio/accel/mxc6255.c:167:36: warning: 'mxc6255_acpi_match' defined but not used [-Wunused-const-variable=]
167 | static const struct acpi_device_id mxc6255_acpi_match[] = {
| ^~~~~~~~~~~~~~~~~~
--
In file included from include/linux/mm.h:29,
from include/linux/memcontrol.h:20,
from include/linux/swap.h:9,
from include/linux/suspend.h:5,
from include/linux/regulator/consumer.h:35,
from include/linux/i2c.h:19,
from drivers/iio/accel/da280.c:10:
>> include/linux/pgtable.h:8:26: error: 'PMD_SHIFT' undeclared here (not in a function); did you mean 'P4D_SHIFT'?
8 | #define PMD_ORDER (PMD_SHIFT - PAGE_SHIFT)
| ^~~~~~~~~
include/linux/swap.h:323:33: note: in expansion of macro 'PMD_ORDER'
323 | unsigned int large_next[PMD_ORDER]; /*
| ^~~~~~~~~
drivers/iio/accel/da280.c:171:36: warning: 'da280_acpi_match' defined but not used [-Wunused-const-variable=]
171 | static const struct acpi_device_id da280_acpi_match[] = {
| ^~~~~~~~~~~~~~~~
--
In file included from include/linux/mm.h:29,
from include/linux/memcontrol.h:20,
from include/linux/swap.h:9,
from include/linux/suspend.h:5,
from include/linux/regulator/consumer.h:35,
from include/linux/i2c.h:19,
from drivers/iio/accel/stk8ba50.c:11:
>> include/linux/pgtable.h:8:26: error: 'PMD_SHIFT' undeclared here (not in a function); did you mean 'P4D_SHIFT'?
8 | #define PMD_ORDER (PMD_SHIFT - PAGE_SHIFT)
| ^~~~~~~~~
include/linux/swap.h:323:33: note: in expansion of macro 'PMD_ORDER'
323 | unsigned int large_next[PMD_ORDER]; /*
| ^~~~~~~~~
drivers/iio/accel/stk8ba50.c:533:36: warning: 'stk8ba50_acpi_id' defined but not used [-Wunused-const-variable=]
533 | static const struct acpi_device_id stk8ba50_acpi_id[] = {
| ^~~~~~~~~~~~~~~~
--
In file included from include/linux/mm.h:29,
from include/linux/memcontrol.h:20,
from include/linux/swap.h:9,
from include/linux/suspend.h:5,
from include/linux/regulator/consumer.h:35,
from include/linux/i2c.h:19,
from drivers/iio/accel/bmc150-accel-i2c.c:9:
>> include/linux/pgtable.h:8:26: error: 'PMD_SHIFT' undeclared here (not in a function); did you mean 'P4D_SHIFT'?
8 | #define PMD_ORDER (PMD_SHIFT - PAGE_SHIFT)
| ^~~~~~~~~
include/linux/swap.h:323:33: note: in expansion of macro 'PMD_ORDER'
323 | unsigned int large_next[PMD_ORDER]; /*
| ^~~~~~~~~
drivers/iio/accel/bmc150-accel-i2c.c:219:36: warning: 'bmc150_accel_acpi_match' defined but not used [-Wunused-const-variable=]
219 | static const struct acpi_device_id bmc150_accel_acpi_match[] = {
| ^~~~~~~~~~~~~~~~~~~~~~~vim +8 include/linux/pgtable.h
f25748e3c34eb8 include/asm-generic/pgtable.h Dan Williams 2016-01-15 7 051ddcfeb1bdba include/linux/pgtable.h Matthew Wilcox (Oracle 2023-08-18 @8) #define PMD_ORDER (PMD_SHIFT - PAGE_SHIFT) 051ddcfeb1bdba include/linux/pgtable.h Matthew Wilcox (Oracle 2023-08-18 9) #define PUD_ORDER (PUD_SHIFT - PAGE_SHIFT) 051ddcfeb1bdba include/linux/pgtable.h Matthew Wilcox (Oracle 2023-08-18 10)
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki
Quoted 39 lines · Ryan Roberts
Hi All,
This is an RFC for a small series to add support for swapping out small-sized THP without needing to first split the large folio via
__split_huge_page(). It closely follows the approach already used by PMD-sized THP."Small-sized THP" is an upcoming feature that enables performance improvements by allocating large folios for anonymous memory, where the large folio size is smaller than the traditional PMD-size. See [1].
In some circumstances I've observed a performance regression (see patch 2 for details), and this series is an attempt to fix the regression in advance of merging small-sized THP support.
I've done what I thought was the smallest change possible, and as a result, this approach is only employed when the swap is backed by a non-rotating block device (just as PMD-sized THP is supported today). However, I have a few questions on whether we should consider relaxing those requirements in certain circumstances:
1) block-backed vs file-backed
==============================The code only attempts to allocate a contiguous set of entries if swap is backed by a block device (i.e. not file-backed). The original commit,
f0eea189e8e9("mm, THP, swap: don't allocate huge cluster for file backed swap device"), stated "It's hard to write a whole transparent huge page (THP) to a file backed swap device". But didn't state why. Does this imply there is a size limit at which it becomes hard? And does that therefore imply that for "small enough" sizes we should now allow use with file-back swap?This original commit was subsequently fixed with commit
41663430588c("mm, THP,
swap: fix allocating cluster for swapfile by mistake"), which said the original commit was using the wrong flag to determine if it was a block device and therefore in some cases was actually doing large allocations for a file-backed swap device, and this was causing file-system corruption. But that implies some sort of correctness issue to me, rather than the performance issue I inferred from the original commit.
Quoted 2 lines
If anyone can offer an explanation, that would be helpful in determining if we should allow some large sizes for file-backed swap.
swap use 'swap extent' (swap_info_struct.swap_extent_root) to map from swap offset to storage block number. For block-backed swap, the mapping is pure linear. So, you can use arbitrary large page size. But for file-backed swap, only PAGE_SIZE alignment is guaranteed.
Quoted 10 lines
2) rotating vs non-rotating
===========================I notice that the clustered approach is only used for non-rotating swap. That implies that for rotating media, we will always fail a large allocation, and fall back to splitting THPs to single pages. Which implies that the regression I'm fixing here may still be present on rotating media? Or perhaps rotating disk is so slow that the cost of writing the data out dominates the cost of splitting?
Quoted 7 lines
I considered that potentially the free swap entry search algorithm that is used in this case could be modified to look for (small) contiguous runs of entries; Up to ~16 pages (order-4) could be done by doing 2x 64bit reads from map instead of single byte.
I haven't looked into this idea in detail, but wonder if anybody thinks it is worth the effort? Or perhaps it would end up causing bad fragmentation.
I doubt anybody will use rotating storage to back swap now.
Quoted 4 lines
Finally on testing, I've run the mm selftests and see no regressions, but I don't think there is anything in there specifically aimed towards swap? Are there any functional or performance tests that I should run? It would certainly be good to confirm I haven't regressed PMD-size THP swap performance.
I have used swap sub test case of vm-scalbility to test.
https://git.kernel.org/pub/scm/linux/kernel/git/wfg/vm-scalability.git/
--
Best Regards,
Huang, Ying
Ryan Roberts
reviewerQuoted 41 lines · Ying
Quoted 39 lines · Ryan Roberts
Hi All,
This is an RFC for a small series to add support for swapping out small-sized THP without needing to first split the large folio via
__split_huge_page(). It closely follows the approach already used by PMD-sized THP."Small-sized THP" is an upcoming feature that enables performance improvements by allocating large folios for anonymous memory, where the large folio size is smaller than the traditional PMD-size. See [1].
In some circumstances I've observed a performance regression (see patch 2 for details), and this series is an attempt to fix the regression in advance of merging small-sized THP support.
I've done what I thought was the smallest change possible, and as a result, this approach is only employed when the swap is backed by a non-rotating block device (just as PMD-sized THP is supported today). However, I have a few questions on whether we should consider relaxing those requirements in certain circumstances:
1) block-backed vs file-backed
==============================The code only attempts to allocate a contiguous set of entries if swap is backed by a block device (i.e. not file-backed). The original commit,
f0eea189e8e9("mm, THP, swap: don't allocate huge cluster for file backed swap device"), stated "It's hard to write a whole transparent huge page (THP) to a file backed swap device". But didn't state why. Does this imply there is a size limit at which it becomes hard? And does that therefore imply that for "small enough" sizes we should now allow use with file-back swap?This original commit was subsequently fixed with commit
41663430588c("mm, THP,
swap: fix allocating cluster for swapfile by mistake"), which said the original commit was using the wrong flag to determine if it was a block device and therefore in some cases was actually doing large allocations for a file-backed swap device, and this was causing file-system corruption. But that implies some sort of correctness issue to me, rather than the performance issue I inferred from the original commit.
Quoted 7 lines
Quoted 2 lines
If anyone can offer an explanation, that would be helpful in determining if we should allow some large sizes for file-backed swap.
swap use 'swap extent' (swap_info_struct.swap_extent_root) to map from swap offset to storage block number. For block-backed swap, the mapping is pure linear. So, you can use arbitrary large page size. But for file-backed swap, only
PAGE_SIZEalignment is guaranteed.
Ahh, I see, so its a correctness issue then. Thanks!
Quoted 16 lines
Quoted 16 lines
2) rotating vs non-rotating
===========================I notice that the clustered approach is only used for non-rotating swap. That implies that for rotating media, we will always fail a large allocation, and fall back to splitting THPs to single pages. Which implies that the regression I'm fixing here may still be present on rotating media? Or perhaps rotating disk is so slow that the cost of writing the data out dominates the cost of splitting?
I considered that potentially the free swap entry search algorithm that is used in this case could be modified to look for (small) contiguous runs of entries; Up to ~16 pages (order-4) could be done by doing 2x 64bit reads from map instead of single byte.
Quoted 4 lines
Quoted 2 lines
I haven't looked into this idea in detail, but wonder if anybody thinks it is worth the effort? Or perhaps it would end up causing bad fragmentation.
I doubt anybody will use rotating storage to back swap now.
I'm often using a QEMU VM for testing with an Ubuntu install. The disk is enumerating as rotating storage and the swap device is file-backed. But I guess the former issue at least, is me setting up QEMU with the incorrect options.
Quoted 9 lines
Quoted 5 lines
Finally on testing, I've run the mm selftests and see no regressions, but I don't think there is anything in there specifically aimed towards swap? Are there any functional or performance tests that I should run? It would certainly be good to confirm I haven't regressed PMD-size THP swap performance.
I have used swap sub test case of vm-scalbility to test.
https://git.kernel.org/pub/scm/linux/kernel/git/wfg/vm-scalability.git/
Great - I shall take a look!
Quoted 4 lines
--
Best Regards,
Huang, Ying
Ryan Roberts
reviewerQuoted 4 lines · Ying
Ryan Roberts <[email protected]> writes:
[...]
Quoted 8 lines
Quoted 4 lines
Finally on testing, I've run the mm selftests and see no regressions, but I don't think there is anything in there specifically aimed towards swap? Are there any functional or performance tests that I should run? It would certainly be good to confirm I haven't regressed PMD-size THP swap performance.
I have used swap sub test case of vm-scalbility to test.
https://git.kernel.org/pub/scm/linux/kernel/git/wfg/vm-scalability.git/
I ended up using usemem, which is the core of this test suite, but deviated from the pre-canned test case to allow me to use anonymous memory and get numbers for small-sized THP (this is a very useful tool - thanks for pointing it out!)
I've run the tests on Ampere Altra, set up with a 35G block ram device as the swap device and from inside a memcg limited to 40G memory. I've then run usemem with 70 processes (each has its own core), each allocating and writing 1G of memory. I've repeated everything 5 times and taken the mean and stdev:
Mean Performance Improvement vs 4K/baseline
| alloc size | baseline | remove-huge-flag | swap-file-small-thp | | | v6.6-rc4+anonfolio | + patch 1 | + patch 2 | |:-----------|--------------------:|--------------------:|--------------------:| | 4K Page | 0.0% | 2.3% | 9.1% | | 64K THP | -44.1% | -46.3% | 30.6% | | 2M THP | 56.0% | 54.2% | 60.1% |
Standard Deviation as Percentage of Mean
| alloc size | baseline | remove-huge-flag | swap-file-small-thp | | | v6.6-rc4+anonfolio | + patch 1 | + patch 2 | |:-----------|--------------------:|--------------------:|--------------------:| | 4K Page | 3.4% | 7.1% | 1.7% | | 64K THP | 1.9% | 5.6% | 7.7% | | 2M THP | 1.9% | 2.1% | 3.2% |
I don't see any meaningful performance cost to removing the HUGE flag, so hopefully this gives us confidence to move forward with patch 1.
You can indeed see the performance regression in the baseline when THP is configured to allocate small-sized THP only (in this case 64K). And you can see the regression is fixed by patch 2, which avoids splitting the THP and thus avoids the extra TLBIs. This correlates with what I saw in kernel compilation workload.
Huang Ying, based on these results, do you still want me to persue a per-cpu solution to avoid potential contention on the swap info lock? - I proposed in the thread against patch 2 to do this in the swap_slots layer if so, rather than in swapfile.c directly (I'm not sure how your original proposal would actually work?). But based on these results, its not obvious to me that there is a definite problem here, and it might be simpler to avoid the complexity?
Thanks,
Ryan
Quoted 4 lines
--
Best Regards,
Huang, Ying