Skip to contentSkip to conversation

btrfs: make extent buffer memory continuous

PATCH In review Qu Wenruo 5 participants 9 messages Aug 24, 2023 – Sep 6, 2023 Btrfs file system

lore
Share & tools
lore

Qu Wenruo

author
#0 lore raw

cover btrfs: make extent buffer memory continuous

[CHANGELOG]
RFC->v1:
- Rebased to the latest misc-next branch Just a small conflicts in extent_buffer_memmove().

- Further cleanup the extent buffer bitmap operations

[REPO]
https://github.com/adam900710/linux/tree/eb_page_cleanups

This includes the submitted extent buffer accessors cleanup as the dependency.

[BACKGROUND]
We have a lot of extent buffer code addressing the cross-page accesses, on the other hand, other filesystems like XFS is mapping its xfs_buf into kernel virtual address space, so that they can access the content of xfs_buf without bothering the page boundaries.

[OBJECTIVE]
This patchset is mostly learning from the xfs_buf, to greatly simplify the extent buffer accessors.

Now all the extent buffer accessors are turned into wrappers of memcpy()/memcmp()/memmove().

For now, it can pass test cases from btrfs group without new regressions.

Qu Wenruo (3):
btrfs: warn on tree blocks which are not nodesize aligned
btrfs: map uncontinuous extent buffer pages into virtual address space
btrfs: utilize the physically/virtually continuous extent buffer
memory

 fs/btrfs/disk-io.c   |  18 +--
 fs/btrfs/extent_io.c | 360 +++++++++++++------------------------------
 fs/btrfs/extent_io.h |  17 ++
 fs/btrfs/fs.h        |   7 +
 4 files changed, 139 insertions(+), 263 deletions(-)
Signature · 1 line
2.41.0
author
#1 lore raw

PATCH 1/3 btrfs: warn on tree blocks which are not nodesize aligned

A long time ago, we have some metadata chunks which starts at sector boundary but not aligned at nodesize boundary.

This led to some older fs which can have tree blocks only aligned to sectorsize, but not nodesize.

Later btrfs check gained the ability to detect and warn about such tree blocks, and kernel fixed the chunk allocation behavior, nowadays those tree blocks should be pretty rare.

But in the future, if we want to migrate metadata to folio, we can not have such tree blocks, as filemap_add_folio() requires the page index to be aligned with the folio number of pages.
(AKA, such unaligned tree blocks can lead to VM_BUG_ON().)

So this patch adds extra warning for those unaligned tree blocks, as a preparation for the future folio migration.

---
 fs/btrfs/extent_io.c | 8 ++++++++
 fs/btrfs/fs.h        | 7 +++++++
 2 files changed, 15 insertions(+)
fs/btrfs/extent_io.c +8 -0
index ac3fca5a5e41..f13211975e0b 100644
@@ -3462,6 +3462,14 @@ static int check_eb_alignment(struct btrfs_fs_info *fs_info, u64 start)
 			  start, fs_info->nodesize);
 		return -EINVAL;
 	}
+	if (!IS_ALIGNED(start, fs_info->nodesize) &&
+	    !test_and_set_bit(BTRFS_FS_UNALIGNED_TREE_BLOCK,
+			      &fs_info->flags)) {
+		btrfs_warn(fs_info,
+		"tree block not nodesize aligned, start %llu nodesize %u",
+			      start, fs_info->nodesize);
+		btrfs_warn(fs_info, "this can be solved by a full metadata balance");
+	}
 	return 0;
 }
 
fs/btrfs/fs.h +7 -0
index a523d64d5491..4dc16d74437c 100644
@@ -139,6 +139,13 @@ enum {
 	 */
 	BTRFS_FS_FEATURE_CHANGED,
 
+	/*
+	 * Indicate if we have tree block which is only aligned to sectorsize,
+	 * but not to nodesize.
+	 * This should be rare nowadays.
+	 */
+	BTRFS_FS_UNALIGNED_TREE_BLOCK,
+
 #if BITS_PER_LONG == 32
 	/* Indicate if we have error/warn message printed on 32bit systems */
 	BTRFS_FS_32BIT_ERROR,
Signature · 1 line
2.41.0
Reviewed-by on v1 1/3
#2 lore raw

Re: [PATCH 1/3] btrfs: warn on tree blocks which are not nodesize aligned

fs/btrfs/extent_io.c static int check_eb_alignment(struct btrfs_fs_info *fs_info, u64 start) · line 3473

Quoted 12 lines · Qu Wenruo

A long time ago, we have some metadata chunks which starts at sector boundary but not aligned at nodesize boundary.

fs/btrfs/extent_io.c +8 -0
+	if (!IS_ALIGNED(start, fs_info->nodesize) &&
+	    !test_and_set_bit(BTRFS_FS_UNALIGNED_TREE_BLOCK,
+			      &fs_info->flags)) {
+		btrfs_warn(fs_info,
+		"tree block not nodesize aligned, start %llu nodesize %u",
+			      start, fs_info->nodesize);
+		btrfs_warn(fs_info, "this can be solved by a full metadata balance");
+	}
  	return 0;

I don't know if ratelimited is required here. But, that shouldn't be a no-go for the patch.

Thanks.

#3 lore raw

Re: [PATCH 1/3] btrfs: warn on tree blocks which are not nodesize aligned

Quoted 4 lines · Anand Jain
Quoted 3 lines · Qu Wenruo

A long time ago, we have some metadata chunks which starts at sector boundary but not aligned at nodesize boundary.

Quoted 12 lines

fs/btrfs/extent_io.c static int check_eb_alignment(struct btrfs_fs_info *fs_info, u64 start) · line 3473

Quoted 9 lines
fs/btrfs/extent_io.c +8 -0
+	if (!IS_ALIGNED(start, fs_info->nodesize) &&
+	    !test_and_set_bit(BTRFS_FS_UNALIGNED_TREE_BLOCK,
+			      &fs_info->flags)) {
+		btrfs_warn(fs_info,
+		"tree block not nodesize aligned, start %llu nodesize %u",
+			      start, fs_info->nodesize);
+		btrfs_warn(fs_info, "this can be solved by a full metadata balance");
+	}
  	return 0;

I don't know if ratelimited is required here. But, that shouldn't be a no-go for the patch.

There will be only one such message as it's tracked by the global state bit BTRFS_FS_UNALIGNED_TREE_BLOCK. However the message should be printed in one go so it's not mixed with other system messages.

author
#4 lore raw

PATCH 2/3 btrfs: map uncontinuous extent buffer pages into virtual address space

In reply to Qu Wenruo

Currently btrfs implements its extent buffer read-write using various helpers doing cross-page handling for the pages array.

However other filesystems like XFS is mapping the pages into kernel virtual address space, greatly simplify the access.

This patch would learn from XFS and map the pages into virtual address space, if and only if the pages are not physically continuous. (Note, a single page counts as physically continuous.)

For now we only do the map, but not yet really utilize the mapped address.

---
 fs/btrfs/extent_io.c | 70 ++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/extent_io.h |  7 +++++
 2 files changed, 77 insertions(+)
fs/btrfs/extent_io.c +70 -0
index f13211975e0b..9f9a3ab82f04 100644
@@ -14,6 +14,7 @@
 #include <linux/pagevec.h>
 #include <linux/prefetch.h>
 #include <linux/fsverity.h>
+#include <linux/vmalloc.h>
 #include "misc.h"
 #include "extent_io.h"
 #include "extent-io-tree.h"
@@ -3153,6 +3154,8 @@ static void btrfs_release_extent_buffer_pages(struct extent_buffer *eb)
 	ASSERT(!extent_buffer_under_io(eb));
 
 	num_pages = num_extent_pages(eb);
+	if (eb->vaddr)
+		vm_unmap_ram(eb->vaddr, num_pages);
 	for (i = 0; i < num_pages; i++) {
 		struct page *page = eb->pages[i];
 
@@ -3202,6 +3205,7 @@ struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src)
 {
 	int i;
 	struct extent_buffer *new;
+	bool pages_contig = true;
 	int num_pages = num_extent_pages(src);
 	int ret;
 
@@ -3226,6 +3230,9 @@ struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src)
 		int ret;
 		struct page *p = new->pages[i];
 
+		if (i && p != new->pages[i - 1] + 1)
+			pages_contig = false;
+
 		ret = attach_extent_buffer_page(new, p, NULL);
 		if (ret < 0) {
 			btrfs_release_extent_buffer(new);
@@ -3233,6 +3240,23 @@ struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src)
 		}
 		WARN_ON(PageDirty(p));
 	}
+	if (!pages_contig) {
+		unsigned int nofs_flag;
+		int retried = 0;
+
+		nofs_flag = memalloc_nofs_save();
+		do {
+			new->vaddr = vm_map_ram(new->pages, num_pages, -1);
+			if (new->vaddr)
+				break;
+			vm_unmap_aliases();
+		} while ((retried++) <= 1);
+		memalloc_nofs_restore(nofs_flag);
+		if (!new->vaddr) {
+			btrfs_release_extent_buffer(new);
+			return NULL;
+		}
+	}
 	copy_extent_buffer_full(new, src);
 	set_extent_buffer_uptodate(new);
 
@@ -3243,6 +3267,7 @@ struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
 						  u64 start, unsigned long len)
 {
 	struct extent_buffer *eb;
+	bool pages_contig = true;
 	int num_pages;
 	int i;
 	int ret;
@@ -3259,11 +3284,29 @@ struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
 	for (i = 0; i < num_pages; i++) {
 		struct page *p = eb->pages[i];
 
+		if (i && p != eb->pages[i - 1] + 1)
+			pages_contig = false;
+
 		ret = attach_extent_buffer_page(eb, p, NULL);
 		if (ret < 0)
 			goto err;
 	}
 
+	if (!pages_contig) {
+		unsigned int nofs_flag;
+		int retried = 0;
+
+		nofs_flag = memalloc_nofs_save();
+		do {
+			eb->vaddr = vm_map_ram(eb->pages, num_pages, -1);
+			if (eb->vaddr)
+				break;
+			vm_unmap_aliases();
+		} while ((retried++) <= 1);
+		memalloc_nofs_restore(nofs_flag);
+		if (!eb->vaddr)
+			goto err;
+	}
 	set_extent_buffer_uptodate(eb);
 	btrfs_set_header_nritems(eb, 0);
 	set_bit(EXTENT_BUFFER_UNMAPPED, &eb->bflags);
@@ -3486,6 +3529,7 @@ struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
 	struct address_space *mapping = fs_info->btree_inode->i_mapping;
 	struct btrfs_subpage *prealloc = NULL;
 	u64 lockdep_owner = owner_root;
+	bool pages_contig = true;
 	int uptodate = 1;
 	int ret;
 
@@ -3558,6 +3602,10 @@ struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
 		/* Should not fail, as we have preallocated the memory */
 		ret = attach_extent_buffer_page(eb, p, prealloc);
 		ASSERT(!ret);
+
+		if (i && p != eb->pages[i - 1] + 1)
+			pages_contig = false;
+
 		/*
 		 * To inform we have extra eb under allocation, so that
 		 * detach_extent_buffer_page() won't release the page private
@@ -3583,6 +3631,28 @@ struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
 		 * we could crash.
 		 */
 	}
+
+	/*
+	 * If pages are not continuous, here we map it into a continuous virtual
+	 * range to make later access easier.
+	 */
+	if (!pages_contig) {
+		unsigned int nofs_flag;
+		int retried = 0;
+
+		nofs_flag = memalloc_nofs_save();
+		do {
+			eb->vaddr = vm_map_ram(eb->pages, num_pages, -1);
+			if (eb->vaddr)
+				break;
+			vm_unmap_aliases();
+		} while ((retried++) <= 1);
+		memalloc_nofs_restore(nofs_flag);
+		if (!eb->vaddr) {
+			exists = ERR_PTR(-ENOMEM);
+			goto free_eb;
+		}
+	}
 	if (uptodate)
 		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
 again:
fs/btrfs/extent_io.h +7 -0
index 68368ba99321..930a2dc38157 100644
@@ -87,6 +87,13 @@ struct extent_buffer {
 
 	struct rw_semaphore lock;
 
+	/*
+	 * For virtually mapped address.
+	 *
+	 * NULL if the pages are physically continuous.
+	 */
+	void *vaddr;
+
 	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
 #ifdef CONFIG_BTRFS_DEBUG
 	struct list_head leak_list;
Signature · 1 line
2.41.0
#5 lore raw

Re: [PATCH 2/3] btrfs: map uncontinuous extent buffer pages into virtual address space

fs/btrfs/extent_io.c struct extent_buffer *btrfs_clone_extent_buffer(const struct extent_buffer *src) · line 3249

Quoted 2 lines · Qu Wenruo
fs/btrfs/extent_io.c +2 -0
+		do {
+			new->vaddr = vm_map_ram(new->pages, num_pages, -1);

Please use NUMA_NO_NODE instead of -1. This makes it easier to read.

author
#6 lore raw

PATCH 3/3 btrfs: utilize the physically/virtually continuous extent buffer memory

In reply to Qu Wenruo

Since the extent buffer pages are either physically or virtually continuous, let's benefit from the new feature.

This involves the following changes:

- Extent buffer accessors
Now read/write/memcpy/memmove_extent_buffer() functions are just a wrapper of memcpy()/memmove().

The cross-page handling are handled by hardware MMU.

- Extent buffer bitmap accessors

- csum_tree_block()
We can directly go crypto_shash_digest(), as we don't need to handle page boundaries anymore.

---
 fs/btrfs/disk-io.c   |  18 +--
 fs/btrfs/extent_io.c | 282 +++++--------------------------------------
 fs/btrfs/extent_io.h |  10 ++
 3 files changed, 47 insertions(+), 263 deletions(-)
fs/btrfs/disk-io.c +4 -14
index 0a96ea8c1d3a..03a423f687b8 100644
@@ -75,24 +75,14 @@ static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info)
 static void csum_tree_block(struct extent_buffer *buf, u8 *result)
 {
 	struct btrfs_fs_info *fs_info = buf->fs_info;
-	const int num_pages = num_extent_pages(buf);
-	const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize);
 	SHASH_DESC_ON_STACK(shash, fs_info->csum_shash);
-	char *kaddr;
-	int i;
+	void *eb_addr = btrfs_get_eb_addr(buf);
 
+	memset(result, 0, BTRFS_CSUM_SIZE);
 	shash->tfm = fs_info->csum_shash;
 	crypto_shash_init(shash);
-	kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start);
-	crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE,
-			    first_page_part - BTRFS_CSUM_SIZE);
-
-	for (i = 1; i < num_pages && INLINE_EXTENT_BUFFER_PAGES > 1; i++) {
-		kaddr = page_address(buf->pages[i]);
-		crypto_shash_update(shash, kaddr, PAGE_SIZE);
-	}
-	memset(result, 0, BTRFS_CSUM_SIZE);
-	crypto_shash_final(shash, result);
+	crypto_shash_digest(shash, eb_addr + BTRFS_CSUM_SIZE,
+			    buf->len - BTRFS_CSUM_SIZE, result);
 }
 
 /*
fs/btrfs/extent_io.c +33 -249

382 lines not on this page · raw message

fs/btrfs/extent_io.h +10 -0
index 930a2dc38157..bfa14457f461 100644
@@ -140,6 +140,16 @@ static inline unsigned long get_eb_page_index(unsigned long offset)
 	return offset >> PAGE_SHIFT;
 }
 
+static inline void *btrfs_get_eb_addr(const struct extent_buffer *eb)
+{
+	/* For fallback vmapped extent buffer. */
+	if (eb->vaddr)
+		return eb->vaddr;
+
+	/* For physically continuous pages and subpage cases. */
+	return page_address(eb->pages[0]) + offset_in_page(eb->start);
+}
+
 /*
  * Structure to record how many bytes and which ranges are set/cleared
  */
Signature · 1 line
2.41.0
#7 lore raw

Re: [PATCH 3/3] btrfs: utilize the physically/virtually continuous extent buffer memory

kernel test robot noticed a 12.0% improvement of filebench.sum_operations/s on:

Hello,

kernel test robot noticed a 12.0% improvement of filebench.sum_operations/s on:

commit: 2fa4ac9754a7fa77bad88aae11ac77ba137d3858 ("[PATCH 3/3] btrfs: utilize the physically/virtually continuous extent buffer memory")
url: https://github.com/intel-lab-lkp/linux/commits/Qu-Wenruo/btrfs-warn-on-tree-blocks-which-are-not-nodesize-aligned/20230824-143628
base: https://git.kernel.org/cgit/linux/kernel/git/kdave/linux.git for-next
patch link: https://lore.kernel.org/all/8bc15bfdaa2805d1d1b660b8b2e07a55aa02027d.1692858397.git.wqu@suse.com/
patch subject: [PATCH 3/3] btrfs: utilize the physically/virtually continuous extent buffer memory

testcase: filebench
test machine: 96 threads 2 sockets (Ice Lake) with 128G memory parameters:

disk: 1HDD
fs: btrfs
fs2: cifs
test: webproxy.f
cpufreq_governor: performance

Details are as below:
-------------------------------------------------------------------------------------------------->

The kernel config and materials to reproduce are available at:
https://download.01.org/0day-ci/archive/20230906/[email protected]

=========================================================================================
compiler/cpufreq_governor/disk/fs2/fs/kconfig/rootfs/tbox_group/test/testcase:
gcc-12/performance/1HDD/cifs/btrfs/x86_64-rhel-8.3/debian-11.1-x86_64-20220510.cgz/lkp-icl-2sp1/webproxy.f/filebench

commit:
19e81514b8 ("btrfs: map uncontinuous extent buffer pages into virtual address space") 2fa4ac9754 ("btrfs: utilize the physically/virtually continuous extent buffer memory")

19e81514b8c09202 2fa4ac9754a7fa77bad88aae11a
---------------- ---------------------------

         %stddev     %change         %stddev
             \          |                \  
     30592 ±194%     -92.3%       2343 ± 24%  sched_debug.cpu.avg_idle.min
      1.38            -5.9%       1.30        iostat.cpu.iowait
      4.63            +8.9%       5.04        iostat.cpu.system
      2.56            +0.5        3.09        mpstat.cpu.all.sys%
      0.54            +0.1        0.61        mpstat.cpu.all.usr%
      1996            +3.3%       2062        vmstat.io.bo
     33480           +13.5%      37993        vmstat.system.cs
    152.67           +12.6%     171.83        turbostat.Avg_MHz
      2562            +4.2%       2670        turbostat.Bzy_MHz
      5.34            +0.5        5.83        turbostat.C1E%
      7.12 ± 12%     -21.6%       5.58 ± 12%  turbostat.Pkg%pc2
    209.72            +1.5%     212.81        turbostat.PkgWatt
      4.92 ± 24%      +3.5        8.37 ± 32%  perf-profile.calltrace.cycles-pp.cpuidle_enter.cpuidle_idle_call.do_idle.cpu_startup_entry.start_secondary
      5.13 ± 28%      +3.6        8.68 ± 31%  perf-profile.calltrace.cycles-pp.cpuidle_idle_call.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64_no_verify
      5.13 ± 28%      +3.8        8.90 ± 30%  perf-profile.calltrace.cycles-pp.cpuidle_enter_state.cpuidle_enter.cpuidle_idle_call.do_idle.cpu_startup_entry
      5.13 ± 28%      +3.8        8.90 ± 30%  perf-profile.children.cycles-pp.cpuidle_enter
      5.13 ± 28%      +3.8        8.90 ± 30%  perf-profile.children.cycles-pp.cpuidle_enter_state
      5.34 ± 34%      +3.9        9.21 ± 28%  perf-profile.children.cycles-pp.cpuidle_idle_call
     13.90            +9.6%      15.23        filebench.sum_bytes_mb/s
    238030           +12.0%     266575        filebench.sum_operations
      3966           +12.0%       4442        filebench.sum_operations/s
      1043           +12.0%       1168        filebench.sum_reads/s
     25.14           -10.7%      22.46        filebench.sum_time_ms/op
    208.83           +11.9%     233.67        filebench.sum_writes/s
    506705            +5.8%     536097        filebench.time.file_system_outputs
      1597 ±  5%     -36.1%       1020 ±  3%  filebench.time.involuntary_context_switches
     61810 ±  2%      +6.0%      65519        filebench.time.minor_page_faults
    157.67 ±  2%     +31.5%     207.33        filebench.time.percent_of_cpu_this_job_got
    117.60 ±  2%     +27.1%     149.48        filebench.time.system_time
    375177           +10.3%     413862        filebench.time.voluntary_context_switches
     18717            +6.5%      19942        proc-vmstat.nr_active_anon
     20206            +1.2%      20445        proc-vmstat.nr_active_file
    298911            +2.2%     305406        proc-vmstat.nr_anon_pages
    132893            +5.6%     140397        proc-vmstat.nr_dirtied
    313040            +2.0%     319443        proc-vmstat.nr_inactive_anon
     32910            +3.4%      34035        proc-vmstat.nr_shmem
     62503            +1.4%      63367        proc-vmstat.nr_slab_unreclaimable
     99471            +3.7%     103159        proc-vmstat.nr_written
     18717            +6.5%      19942        proc-vmstat.nr_zone_active_anon
     20206            +1.2%      20445        proc-vmstat.nr_zone_active_file
    313040            +2.0%     319443        proc-vmstat.nr_zone_inactive_anon
    943632            +3.2%     974142        proc-vmstat.numa_hit
    841654            +3.6%     871757        proc-vmstat.numa_local
    453634 ± 17%     +27.0%     576268 ±  5%  proc-vmstat.numa_pte_updates
     87464            +6.1%      92814        proc-vmstat.pgactivate
   1595438            +2.9%    1641074        proc-vmstat.pgalloc_normal
   1453326            +3.0%    1497530        proc-vmstat.pgfree
     17590 ±  5%     +14.0%      20045 ±  7%  proc-vmstat.pgreuse
    732160            -1.8%     719104        proc-vmstat.unevictable_pgs_scanned
     19.10            -8.1%      17.55        perf-stat.i.MPKI
 2.039e+09           +17.3%  2.393e+09        perf-stat.i.branch-instructions
      1.27 ±  2%      -0.1        1.15        perf-stat.i.branch-miss-rate%
  25600761            +5.8%   27075672        perf-stat.i.branch-misses
   5037721 ±  4%     +11.4%    5612619        perf-stat.i.cache-misses
 1.632e+08            +5.9%  1.729e+08        perf-stat.i.cache-references
     34079           +14.1%      38871        perf-stat.i.context-switches
 1.326e+10           +14.7%  1.521e+10        perf-stat.i.cpu-cycles
    551.02 ±  2%     +21.0%     666.59 ±  3%  perf-stat.i.cpu-migrations
   3953434 ±  2%     +10.8%    4381924 ±  3%  perf-stat.i.dTLB-load-misses
 2.343e+09           +15.4%  2.704e+09        perf-stat.i.dTLB-loads
 1.141e+09           +14.3%  1.303e+09        perf-stat.i.dTLB-stores
 9.047e+09           +14.9%  1.039e+10        perf-stat.i.instructions
      0.69            +2.0%       0.71        perf-stat.i.ipc
      0.14           +14.7%       0.16        perf-stat.i.metric.GHz
     34.94 ±  4%     +11.1%      38.80        perf-stat.i.metric.K/sec
     59.21           +15.6%      68.43        perf-stat.i.metric.M/sec
      3999 ±  3%      +6.3%       4250        perf-stat.i.minor-faults
   1116010 ±  4%     +14.8%    1280875 ±  2%  perf-stat.i.node-load-misses
   1168171 ±  3%      +7.9%    1259922 ±  2%  perf-stat.i.node-stores
      3999 ±  3%      +6.3%       4250        perf-stat.i.page-faults
     18.04            -7.8%      16.64        perf-stat.overall.MPKI
      1.26 ±  2%      -0.1        1.13        perf-stat.overall.branch-miss-rate%
 2.012e+09           +17.3%  2.359e+09        perf-stat.ps.branch-instructions
  25253051            +5.7%   26690222        perf-stat.ps.branch-misses
   4970910 ±  4%     +11.3%    5534021        perf-stat.ps.cache-misses
  1.61e+08            +5.9%  1.705e+08        perf-stat.ps.cache-references
     33628           +14.0%      38332        perf-stat.ps.context-switches
 1.308e+10           +14.6%    1.5e+10        perf-stat.ps.cpu-cycles
    543.73 ±  2%     +20.9%     657.37 ±  3%  perf-stat.ps.cpu-migrations
   3900887 ±  2%     +10.8%    4321011 ±  3%  perf-stat.ps.dTLB-load-misses
 2.312e+09           +15.3%  2.666e+09        perf-stat.ps.dTLB-loads
 1.125e+09           +14.2%  1.285e+09        perf-stat.ps.dTLB-stores
 8.925e+09           +14.8%  1.024e+10        perf-stat.ps.instructions
      3943 ±  3%      +6.2%       4187        perf-stat.ps.minor-faults
   1101275 ±  4%     +14.7%    1263151 ±  2%  perf-stat.ps.node-load-misses
   1152648 ±  3%      +7.7%    1241973 ±  2%  perf-stat.ps.node-stores
      3943 ±  3%      +6.2%       4187        perf-stat.ps.page-faults
 6.777e+11           +10.5%   7.49e+11        perf-stat.total.instructions
      0.01 ±  7%     -28.2%       0.00 ± 26%  perf-sched.sch_delay.avg.ms.schedule_preempt_disabled.rwsem_down_read_slowpath.down_read.__btrfs_tree_read_lock
      0.30 ± 35%     -63.0%       0.11 ± 25%  perf-sched.sch_delay.max.ms.__cond_resched.__kmem_cache_alloc_node.__kmalloc.cifs_strndup_to_utf16.cifs_convert_path_to_utf16
     30.21 ±  3%      -6.2%      28.33 ±  3%  perf-sched.total_wait_and_delay.average.ms
     30.15 ±  3%      -6.2%      28.28 ±  3%  perf-sched.total_wait_time.average.ms
      1.08           -20.5%       0.86 ±  2%  perf-sched.wait_and_delay.avg.ms.io_schedule.folio_wait_bit_common.filemap_update_page.filemap_get_pages
     99.86 ± 27%     +71.6%     171.38 ± 32%  perf-sched.wait_and_delay.avg.ms.kthreadd.ret_from_fork.ret_from_fork_asm
      1.10 ±  2%     -16.3%       0.92        perf-sched.wait_and_delay.avg.ms.schedule_hrtimeout_range_clock.ep_poll.do_epoll_wait.__x64_sys_epoll_wait
      1.41 ±  5%     -87.1%       0.18 ±223%  perf-sched.wait_and_delay.avg.ms.schedule_preempt_disabled.__mutex_lock.constprop.0.cifs_call_async
      0.21           -13.4%       0.18        perf-sched.wait_and_delay.avg.ms.schedule_timeout.wait_woken.sk_wait_data.tcp_recvmsg_locked
    195.95 ± 10%     -18.4%     159.83 ± 12%  perf-sched.wait_and_delay.avg.ms.wait_for_response.compound_send_recv.cifs_send_recv.__SMB2_close
      2.60           -23.5%       1.99        perf-sched.wait_and_delay.avg.ms.wait_for_response.compound_send_recv.cifs_send_recv.query_info
     20.46           -13.7%      17.66 ±  4%  perf-sched.wait_and_delay.avg.ms.wait_for_response.compound_send_recv.smb2_compound_op.smb2_query_path_info
      3.35 ± 66%    +342.5%      14.82 ± 20%  perf-sched.wait_and_delay.avg.ms.wait_for_response.compound_send_recv.smb2_compound_op.smb2_unlink
      2103           +10.0%       2312 ±  3%  perf-sched.wait_and_delay.count.__lock_sock.sk_wait_data.tcp_recvmsg_locked.tcp_recvmsg
      1025           +14.8%       1176        perf-sched.wait_and_delay.count.io_schedule.folio_wait_bit_common.folio_wait_writeback.__filemap_fdatawait_range
      9729 ±  2%     +21.1%      11779        perf-sched.wait_and_delay.count.schedule_hrtimeout_range_clock.ep_poll.do_epoll_wait.__x64_sys_epoll_wait
      2349 ±  9%     +29.3%       3038 ± 10%  perf-sched.wait_and_delay.count.schedule_preempt_disabled.__mutex_lock.constprop.0.compound_send_recv
    998.00           +14.3%       1140        perf-sched.wait_and_delay.count.schedule_preempt_disabled.rwsem_down_write_slowpath.down_write.do_unlinkat
      1026           +15.0%       1181        perf-sched.wait_and_delay.count.schedule_preempt_disabled.rwsem_down_write_slowpath.down_write.open_last_lookups
     18409           +12.5%      20714 ±  4%  perf-sched.wait_and_delay.count.schedule_timeout.wait_woken.sk_wait_data.tcp_recvmsg_locked
      1011           +14.8%       1160        perf-sched.wait_and_delay.count.wait_for_response.compound_send_recv.cifs_send_recv.query_info
      1013           +14.5%       1160        perf-sched.wait_and_delay.count.wait_for_response.compound_send_recv.smb2_compound_op.smb2_unlink
      2.68 ±  4%     -19.6%       2.16 ±  7%  perf-sched.wait_and_delay.max.ms.__lock_sock.sk_wait_data.tcp_recvmsg_locked.tcp_recvmsg
    282.00 ±  3%     -11.3%     250.07 ±  4%  perf-sched.wait_and_delay.max.ms.schedule_preempt_disabled.rwsem_down_write_slowpath.down_write.do_unlinkat
    280.97 ±  2%     -12.8%     244.97 ±  2%  perf-sched.wait_and_delay.max.ms.schedule_preempt_disabled.rwsem_down_write_slowpath.down_write.open_last_lookups
      0.49 ±125%     -97.2%       0.01 ±198%  perf-sched.wait_time.avg.ms.exit_to_user_mode_loop.exit_to_user_mode_prepare.irqentry_exit_to_user_mode.asm_sysvec_call_function_single
      1.05           -20.9%       0.83 ±  2%  perf-sched.wait_time.avg.ms.io_schedule.folio_wait_bit_common.filemap_update_page.filemap_get_pages
      2.14 ±  4%     +19.1%       2.55 ±  8%  perf-sched.wait_time.avg.ms.io_schedule.rq_qos_wait.wbt_wait.__rq_qos_throttle
     99.82 ± 27%     +69.8%     169.46 ± 31%  perf-sched.wait_time.avg.ms.kthreadd.ret_from_fork.ret_from_fork_asm
      1.08 ±  2%     -16.6%       0.90        perf-sched.wait_time.avg.ms.schedule_hrtimeout_range_clock.ep_poll.do_epoll_wait.__x64_sys_epoll_wait
      1.37 ±  5%     -24.5%       1.03 ±  5%  perf-sched.wait_time.avg.ms.schedule_preempt_disabled.__mutex_lock.constprop.0.cifs_call_async
      0.20           -14.2%       0.17        perf-sched.wait_time.avg.ms.schedule_timeout.wait_woken.sk_wait_data.tcp_recvmsg_locked
    195.53 ± 10%     -18.4%     159.54 ± 12%  perf-sched.wait_time.avg.ms.wait_for_response.compound_send_recv.cifs_send_recv.__SMB2_close
      2.54           -24.0%       1.93        perf-sched.wait_time.avg.ms.wait_for_response.compound_send_recv.cifs_send_recv.query_info
     20.44           -13.8%      17.63 ±  4%  perf-sched.wait_time.avg.ms.wait_for_response.compound_send_recv.smb2_compound_op.smb2_query_path_info
      3.32 ± 67%    +345.6%      14.78 ± 20%  perf-sched.wait_time.avg.ms.wait_for_response.compound_send_recv.smb2_compound_op.smb2_unlink
    245.89 ±  9%     -11.8%     216.92 ±  6%  perf-sched.wait_time.max.ms.__cond_resched.__kmem_cache_alloc_node.__kmalloc.cifs_strndup_to_utf16.cifs_convert_path_to_utf16
      3.14 ±  9%     -43.6%       1.77 ± 40%  perf-sched.wait_time.max.ms.__cond_resched.dput.terminate_walk.path_openat.do_filp_open
      2.65 ±  3%     -19.9%       2.12 ±  6%  perf-sched.wait_time.max.ms.__lock_sock.sk_wait_data.tcp_recvmsg_locked.tcp_recvmsg
      0.57 ±101%     -91.5%       0.05 ±213%  perf-sched.wait_time.max.ms.exit_to_user_mode_loop.exit_to_user_mode_prepare.irqentry_exit_to_user_mode.asm_sysvec_call_function_single
      1.79 ± 82%     -86.4%       0.24 ± 58%  perf-sched.wait_time.max.ms.exit_to_user_mode_loop.exit_to_user_mode_prepare.irqentry_exit_to_user_mode.asm_sysvec_reschedule_ipi
    281.92 ±  3%     -11.3%     249.99 ±  4%  perf-sched.wait_time.max.ms.schedule_preempt_disabled.rwsem_down_write_slowpath.down_write.do_unlinkat
    280.90 ±  2%     -12.8%     244.88 ±  2%  perf-sched.wait_time.max.ms.schedule_preempt_disabled.rwsem_down_write_slowpath.down_write.open_last_lookups

Disclaimer:
Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance.

Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
#8 lore raw
In reply to Qu Wenruo
Quoted 26 lines · Qu Wenruo

[CHANGELOG]
RFC->v1:
- Rebased to the latest misc-next branch Just a small conflicts in extent_buffer_memmove().

- Further cleanup the extent buffer bitmap operations

[REPO]
https://github.com/adam900710/linux/tree/eb_page_cleanups

This includes the submitted extent buffer accessors cleanup as the dependency.

[BACKGROUND]
We have a lot of extent buffer code addressing the cross-page accesses, on the other hand, other filesystems like XFS is mapping its xfs_buf into kernel virtual address space, so that they can access the content of xfs_buf without bothering the page boundaries.

[OBJECTIVE]
This patchset is mostly learning from the xfs_buf, to greatly simplify the extent buffer accessors.

Now all the extent buffer accessors are turned into wrappers of memcpy()/memcmp()/memmove().

Quoted 8 lines

For now, it can pass test cases from btrfs group without new regressions.

Qu Wenruo (3):
btrfs: warn on tree blocks which are not nodesize aligned
btrfs: map uncontinuous extent buffer pages into virtual address space
btrfs: utilize the physically/virtually continuous extent buffer
memory

My objections stand and we can continue the discussion under the RFC series, but for testing purposes I'll add the series to for-next.