linux/fs/btrfs/tests/extent-io-tests.c

833 lines
21 KiB
C
Raw Permalink Normal View History

// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (C) 2013 Fusion IO. All rights reserved.
*/
#include <linux/pagemap.h>
#include <linux/pagevec.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/sizes.h>
#include "btrfs-tests.h"
#include "../ctree.h"
#include "../extent_io.h"
#include "../disk-io.h"
#include "../btrfs_inode.h"
#define PROCESS_UNLOCK (1 << 0)
#define PROCESS_RELEASE (1 << 1)
#define PROCESS_TEST_LOCKED (1 << 2)
static noinline int process_page_range(struct inode *inode, u64 start, u64 end,
unsigned long flags)
{
int ret;
struct folio_batch fbatch;
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
unsigned long index = start >> PAGE_SHIFT;
unsigned long end_index = end >> PAGE_SHIFT;
int i;
int count = 0;
int loops = 0;
folio_batch_init(&fbatch);
while (index <= end_index) {
ret = filemap_get_folios_contig(inode->i_mapping, &index,
end_index, &fbatch);
for (i = 0; i < ret; i++) {
struct folio *folio = fbatch.folios[i];
if (flags & PROCESS_TEST_LOCKED &&
!folio_test_locked(folio))
count++;
if (flags & PROCESS_UNLOCK && folio_test_locked(folio))
folio_unlock(folio);
if (flags & PROCESS_RELEASE)
folio_put(folio);
}
folio_batch_release(&fbatch);
cond_resched();
loops++;
if (loops > 100000) {
printk(KERN_ERR
"stuck in a loop, start %llu, end %llu, ret %d\n",
start, end, ret);
break;
}
}
return count;
}
#define STATE_FLAG_STR_LEN 256
#define PRINT_ONE_FLAG(state, dest, cur, name) \
({ \
if (state->state & EXTENT_##name) \
cur += scnprintf(dest + cur, STATE_FLAG_STR_LEN - cur, \
"%s" #name, cur == 0 ? "" : "|"); \
})
static void extent_flag_to_str(const struct extent_state *state, char *dest)
{
int cur = 0;
dest[0] = 0;
PRINT_ONE_FLAG(state, dest, cur, DIRTY);
PRINT_ONE_FLAG(state, dest, cur, UPTODATE);
PRINT_ONE_FLAG(state, dest, cur, LOCKED);
PRINT_ONE_FLAG(state, dest, cur, NEW);
PRINT_ONE_FLAG(state, dest, cur, DELALLOC);
PRINT_ONE_FLAG(state, dest, cur, DEFRAG);
PRINT_ONE_FLAG(state, dest, cur, BOUNDARY);
PRINT_ONE_FLAG(state, dest, cur, NODATASUM);
PRINT_ONE_FLAG(state, dest, cur, CLEAR_META_RESV);
PRINT_ONE_FLAG(state, dest, cur, NEED_WAIT);
PRINT_ONE_FLAG(state, dest, cur, NORESERVE);
PRINT_ONE_FLAG(state, dest, cur, QGROUP_RESERVED);
PRINT_ONE_FLAG(state, dest, cur, CLEAR_DATA_RESV);
}
static void dump_extent_io_tree(const struct extent_io_tree *tree)
{
struct rb_node *node;
char flags_str[STATE_FLAG_STR_LEN];
node = rb_first(&tree->state);
test_msg("io tree content:");
while (node) {
struct extent_state *state;
state = rb_entry(node, struct extent_state, rb_node);
extent_flag_to_str(state, flags_str);
test_msg(" start=%llu len=%llu flags=%s", state->start,
state->end + 1 - state->start, flags_str);
node = rb_next(node);
}
}
static int test_find_delalloc(u32 sectorsize, u32 nodesize)
{
struct btrfs_fs_info *fs_info;
struct btrfs_root *root = NULL;
struct inode *inode = NULL;
struct extent_io_tree *tmp;
struct page *page;
struct page *locked_page = NULL;
unsigned long index = 0;
/* In this test we need at least 2 file extents at its maximum size */
u64 max_bytes = BTRFS_MAX_EXTENT_SIZE;
u64 total_dirty = 2 * max_bytes;
u64 start, end, test_start;
bool found;
int ret = -EINVAL;
test_msg("running find delalloc tests");
fs_info = btrfs_alloc_dummy_fs_info(nodesize, sectorsize);
if (!fs_info) {
test_std_err(TEST_ALLOC_FS_INFO);
return -ENOMEM;
}
root = btrfs_alloc_dummy_root(fs_info);
if (IS_ERR(root)) {
test_std_err(TEST_ALLOC_ROOT);
ret = PTR_ERR(root);
goto out;
}
inode = btrfs_new_test_inode();
if (!inode) {
test_std_err(TEST_ALLOC_INODE);
ret = -ENOMEM;
goto out;
}
tmp = &BTRFS_I(inode)->io_tree;
BTRFS_I(inode)->root = root;
/*
* Passing NULL as we don't have fs_info but tracepoints are not used
* at this point
*/
extent_io_tree_init(NULL, tmp, IO_TREE_SELFTEST);
/*
* First go through and create and mark all of our pages dirty, we pin
* everything to make sure our pages don't get evicted and screw up our
* test.
*/
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
for (index = 0; index < (total_dirty >> PAGE_SHIFT); index++) {
page = find_or_create_page(inode->i_mapping, index, GFP_KERNEL);
if (!page) {
test_err("failed to allocate test page");
ret = -ENOMEM;
goto out;
}
SetPageDirty(page);
if (index) {
unlock_page(page);
} else {
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
get_page(page);
locked_page = page;
}
}
/* Test this scenario
* |--- delalloc ---|
* |--- search ---|
*/
set_extent_bit(tmp, 0, sectorsize - 1, EXTENT_DELALLOC, NULL);
start = 0;
btrfs: subpage: avoid potential deadlock with compression and delalloc [BUG] With experimental subpage compression enabled, a simple fsstress can lead to self deadlock on page 720896: mkfs.btrfs -f -s 4k $dev > /dev/null mount $dev -o compress $mnt $fsstress -p 1 -n 100 -w -d $mnt -v -s 1625511156 [CAUSE] If we have a file layout looks like below: 0 32K 64K 96K 128K |//| |///////////////| 4K Then we run delalloc range for the inode, it will: - Call find_lock_delalloc_range() with @delalloc_start = 0 Then we got a delalloc range [0, 4K). This range will be COWed. - Call find_lock_delalloc_range() again with @delalloc_start = 4K Since find_lock_delalloc_range() never cares whether the range is still inside page range [0, 64K), it will return range [64K, 128K). This range meets the condition for subpage compression, will go through async COW path. And async COW path will return @page_started. But that @page_started is now for range [64K, 128K), not for range [0, 64K). - writepage_dellloc() returned 1 for page [0, 64K) Thus page [0, 64K) will not be unlocked, nor its page dirty status will be cleared. Next time when we try to lock page [0, 64K) we will deadlock, as there is no one to release page [0, 64K). This problem will never happen for regular page size as one page only contains one sector. After the first find_lock_delalloc_range() call, the @delalloc_end will go beyond @page_end no matter if we found a delalloc range or not Thus this bug only happens for subpage, as now we need multiple runs to exhaust the delalloc range of a page. [FIX] Fix the problem by ensuring the delalloc range we ran at least started inside @locked_page. So that we will never get incorrect @page_started. And to prevent such problem from happening again: - Make find_lock_delalloc_range() return false if the found range is beyond @end value passed in. Since @end will be utilized now, add an ASSERT() to ensure we pass correct @end into find_lock_delalloc_range(). This also means, for selftests we needs to populate @end before calling find_lock_delalloc_range(). - New ASSERT() in find_lock_delalloc_range() Now we will make sure the @start/@end passed in at least covers part of the page. - New ASSERT() in run_delalloc_range() To make sure the range at least starts inside @locked page. - Use @delalloc_start as proper cursor, while @delalloc_end is always reset to @page_end. Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2021-09-27 07:22:07 +00:00
end = start + PAGE_SIZE - 1;
found = find_lock_delalloc_range(inode, page_folio(locked_page), &start,
&end);
if (!found) {
test_err("should have found at least one delalloc");
goto out_bits;
}
if (start != 0 || end != (sectorsize - 1)) {
test_err("expected start 0 end %u, got start %llu end %llu",
sectorsize - 1, start, end);
goto out_bits;
}
unlock_extent(tmp, start, end, NULL);
unlock_page(locked_page);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
put_page(locked_page);
/*
* Test this scenario
*
* |--- delalloc ---|
* |--- search ---|
*/
test_start = SZ_64M;
locked_page = find_lock_page(inode->i_mapping,
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
test_start >> PAGE_SHIFT);
if (!locked_page) {
test_err("couldn't find the locked page");
goto out_bits;
}
set_extent_bit(tmp, sectorsize, max_bytes - 1, EXTENT_DELALLOC, NULL);
start = test_start;
btrfs: subpage: avoid potential deadlock with compression and delalloc [BUG] With experimental subpage compression enabled, a simple fsstress can lead to self deadlock on page 720896: mkfs.btrfs -f -s 4k $dev > /dev/null mount $dev -o compress $mnt $fsstress -p 1 -n 100 -w -d $mnt -v -s 1625511156 [CAUSE] If we have a file layout looks like below: 0 32K 64K 96K 128K |//| |///////////////| 4K Then we run delalloc range for the inode, it will: - Call find_lock_delalloc_range() with @delalloc_start = 0 Then we got a delalloc range [0, 4K). This range will be COWed. - Call find_lock_delalloc_range() again with @delalloc_start = 4K Since find_lock_delalloc_range() never cares whether the range is still inside page range [0, 64K), it will return range [64K, 128K). This range meets the condition for subpage compression, will go through async COW path. And async COW path will return @page_started. But that @page_started is now for range [64K, 128K), not for range [0, 64K). - writepage_dellloc() returned 1 for page [0, 64K) Thus page [0, 64K) will not be unlocked, nor its page dirty status will be cleared. Next time when we try to lock page [0, 64K) we will deadlock, as there is no one to release page [0, 64K). This problem will never happen for regular page size as one page only contains one sector. After the first find_lock_delalloc_range() call, the @delalloc_end will go beyond @page_end no matter if we found a delalloc range or not Thus this bug only happens for subpage, as now we need multiple runs to exhaust the delalloc range of a page. [FIX] Fix the problem by ensuring the delalloc range we ran at least started inside @locked_page. So that we will never get incorrect @page_started. And to prevent such problem from happening again: - Make find_lock_delalloc_range() return false if the found range is beyond @end value passed in. Since @end will be utilized now, add an ASSERT() to ensure we pass correct @end into find_lock_delalloc_range(). This also means, for selftests we needs to populate @end before calling find_lock_delalloc_range(). - New ASSERT() in find_lock_delalloc_range() Now we will make sure the @start/@end passed in at least covers part of the page. - New ASSERT() in run_delalloc_range() To make sure the range at least starts inside @locked page. - Use @delalloc_start as proper cursor, while @delalloc_end is always reset to @page_end. Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2021-09-27 07:22:07 +00:00
end = start + PAGE_SIZE - 1;
found = find_lock_delalloc_range(inode, page_folio(locked_page), &start,
&end);
if (!found) {
test_err("couldn't find delalloc in our range");
goto out_bits;
}
if (start != test_start || end != max_bytes - 1) {
test_err("expected start %llu end %llu, got start %llu, end %llu",
test_start, max_bytes - 1, start, end);
goto out_bits;
}
if (process_page_range(inode, start, end,
PROCESS_TEST_LOCKED | PROCESS_UNLOCK)) {
test_err("there were unlocked pages in the range");
goto out_bits;
}
unlock_extent(tmp, start, end, NULL);
/* locked_page was unlocked above */
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
put_page(locked_page);
/*
* Test this scenario
* |--- delalloc ---|
* |--- search ---|
*/
test_start = max_bytes + sectorsize;
locked_page = find_lock_page(inode->i_mapping, test_start >>
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
PAGE_SHIFT);
if (!locked_page) {
test_err("couldn't find the locked page");
goto out_bits;
}
start = test_start;
btrfs: subpage: avoid potential deadlock with compression and delalloc [BUG] With experimental subpage compression enabled, a simple fsstress can lead to self deadlock on page 720896: mkfs.btrfs -f -s 4k $dev > /dev/null mount $dev -o compress $mnt $fsstress -p 1 -n 100 -w -d $mnt -v -s 1625511156 [CAUSE] If we have a file layout looks like below: 0 32K 64K 96K 128K |//| |///////////////| 4K Then we run delalloc range for the inode, it will: - Call find_lock_delalloc_range() with @delalloc_start = 0 Then we got a delalloc range [0, 4K). This range will be COWed. - Call find_lock_delalloc_range() again with @delalloc_start = 4K Since find_lock_delalloc_range() never cares whether the range is still inside page range [0, 64K), it will return range [64K, 128K). This range meets the condition for subpage compression, will go through async COW path. And async COW path will return @page_started. But that @page_started is now for range [64K, 128K), not for range [0, 64K). - writepage_dellloc() returned 1 for page [0, 64K) Thus page [0, 64K) will not be unlocked, nor its page dirty status will be cleared. Next time when we try to lock page [0, 64K) we will deadlock, as there is no one to release page [0, 64K). This problem will never happen for regular page size as one page only contains one sector. After the first find_lock_delalloc_range() call, the @delalloc_end will go beyond @page_end no matter if we found a delalloc range or not Thus this bug only happens for subpage, as now we need multiple runs to exhaust the delalloc range of a page. [FIX] Fix the problem by ensuring the delalloc range we ran at least started inside @locked_page. So that we will never get incorrect @page_started. And to prevent such problem from happening again: - Make find_lock_delalloc_range() return false if the found range is beyond @end value passed in. Since @end will be utilized now, add an ASSERT() to ensure we pass correct @end into find_lock_delalloc_range(). This also means, for selftests we needs to populate @end before calling find_lock_delalloc_range(). - New ASSERT() in find_lock_delalloc_range() Now we will make sure the @start/@end passed in at least covers part of the page. - New ASSERT() in run_delalloc_range() To make sure the range at least starts inside @locked page. - Use @delalloc_start as proper cursor, while @delalloc_end is always reset to @page_end. Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2021-09-27 07:22:07 +00:00
end = start + PAGE_SIZE - 1;
found = find_lock_delalloc_range(inode, page_folio(locked_page), &start,
&end);
if (found) {
test_err("found range when we shouldn't have");
goto out_bits;
}
btrfs: subpage: avoid potential deadlock with compression and delalloc [BUG] With experimental subpage compression enabled, a simple fsstress can lead to self deadlock on page 720896: mkfs.btrfs -f -s 4k $dev > /dev/null mount $dev -o compress $mnt $fsstress -p 1 -n 100 -w -d $mnt -v -s 1625511156 [CAUSE] If we have a file layout looks like below: 0 32K 64K 96K 128K |//| |///////////////| 4K Then we run delalloc range for the inode, it will: - Call find_lock_delalloc_range() with @delalloc_start = 0 Then we got a delalloc range [0, 4K). This range will be COWed. - Call find_lock_delalloc_range() again with @delalloc_start = 4K Since find_lock_delalloc_range() never cares whether the range is still inside page range [0, 64K), it will return range [64K, 128K). This range meets the condition for subpage compression, will go through async COW path. And async COW path will return @page_started. But that @page_started is now for range [64K, 128K), not for range [0, 64K). - writepage_dellloc() returned 1 for page [0, 64K) Thus page [0, 64K) will not be unlocked, nor its page dirty status will be cleared. Next time when we try to lock page [0, 64K) we will deadlock, as there is no one to release page [0, 64K). This problem will never happen for regular page size as one page only contains one sector. After the first find_lock_delalloc_range() call, the @delalloc_end will go beyond @page_end no matter if we found a delalloc range or not Thus this bug only happens for subpage, as now we need multiple runs to exhaust the delalloc range of a page. [FIX] Fix the problem by ensuring the delalloc range we ran at least started inside @locked_page. So that we will never get incorrect @page_started. And to prevent such problem from happening again: - Make find_lock_delalloc_range() return false if the found range is beyond @end value passed in. Since @end will be utilized now, add an ASSERT() to ensure we pass correct @end into find_lock_delalloc_range(). This also means, for selftests we needs to populate @end before calling find_lock_delalloc_range(). - New ASSERT() in find_lock_delalloc_range() Now we will make sure the @start/@end passed in at least covers part of the page. - New ASSERT() in run_delalloc_range() To make sure the range at least starts inside @locked page. - Use @delalloc_start as proper cursor, while @delalloc_end is always reset to @page_end. Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2021-09-27 07:22:07 +00:00
if (end != test_start + PAGE_SIZE - 1) {
test_err("did not return the proper end offset");
goto out_bits;
}
/*
* Test this scenario
* [------- delalloc -------|
* [max_bytes]|-- search--|
*
* We are re-using our test_start from above since it works out well.
*/
set_extent_bit(tmp, max_bytes, total_dirty - 1, EXTENT_DELALLOC, NULL);
start = test_start;
btrfs: subpage: avoid potential deadlock with compression and delalloc [BUG] With experimental subpage compression enabled, a simple fsstress can lead to self deadlock on page 720896: mkfs.btrfs -f -s 4k $dev > /dev/null mount $dev -o compress $mnt $fsstress -p 1 -n 100 -w -d $mnt -v -s 1625511156 [CAUSE] If we have a file layout looks like below: 0 32K 64K 96K 128K |//| |///////////////| 4K Then we run delalloc range for the inode, it will: - Call find_lock_delalloc_range() with @delalloc_start = 0 Then we got a delalloc range [0, 4K). This range will be COWed. - Call find_lock_delalloc_range() again with @delalloc_start = 4K Since find_lock_delalloc_range() never cares whether the range is still inside page range [0, 64K), it will return range [64K, 128K). This range meets the condition for subpage compression, will go through async COW path. And async COW path will return @page_started. But that @page_started is now for range [64K, 128K), not for range [0, 64K). - writepage_dellloc() returned 1 for page [0, 64K) Thus page [0, 64K) will not be unlocked, nor its page dirty status will be cleared. Next time when we try to lock page [0, 64K) we will deadlock, as there is no one to release page [0, 64K). This problem will never happen for regular page size as one page only contains one sector. After the first find_lock_delalloc_range() call, the @delalloc_end will go beyond @page_end no matter if we found a delalloc range or not Thus this bug only happens for subpage, as now we need multiple runs to exhaust the delalloc range of a page. [FIX] Fix the problem by ensuring the delalloc range we ran at least started inside @locked_page. So that we will never get incorrect @page_started. And to prevent such problem from happening again: - Make find_lock_delalloc_range() return false if the found range is beyond @end value passed in. Since @end will be utilized now, add an ASSERT() to ensure we pass correct @end into find_lock_delalloc_range(). This also means, for selftests we needs to populate @end before calling find_lock_delalloc_range(). - New ASSERT() in find_lock_delalloc_range() Now we will make sure the @start/@end passed in at least covers part of the page. - New ASSERT() in run_delalloc_range() To make sure the range at least starts inside @locked page. - Use @delalloc_start as proper cursor, while @delalloc_end is always reset to @page_end. Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2021-09-27 07:22:07 +00:00
end = start + PAGE_SIZE - 1;
found = find_lock_delalloc_range(inode, page_folio(locked_page), &start,
&end);
if (!found) {
test_err("didn't find our range");
goto out_bits;
}
if (start != test_start || end != total_dirty - 1) {
test_err("expected start %llu end %llu, got start %llu end %llu",
test_start, total_dirty - 1, start, end);
goto out_bits;
}
if (process_page_range(inode, start, end,
PROCESS_TEST_LOCKED | PROCESS_UNLOCK)) {
test_err("pages in range were not all locked");
goto out_bits;
}
unlock_extent(tmp, start, end, NULL);
/*
* Now to test where we run into a page that is no longer dirty in the
* range we want to find.
*/
page = find_get_page(inode->i_mapping,
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
(max_bytes + SZ_1M) >> PAGE_SHIFT);
if (!page) {
test_err("couldn't find our page");
goto out_bits;
}
ClearPageDirty(page);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
put_page(page);
/* We unlocked it in the previous test */
lock_page(locked_page);
start = test_start;
btrfs: subpage: avoid potential deadlock with compression and delalloc [BUG] With experimental subpage compression enabled, a simple fsstress can lead to self deadlock on page 720896: mkfs.btrfs -f -s 4k $dev > /dev/null mount $dev -o compress $mnt $fsstress -p 1 -n 100 -w -d $mnt -v -s 1625511156 [CAUSE] If we have a file layout looks like below: 0 32K 64K 96K 128K |//| |///////////////| 4K Then we run delalloc range for the inode, it will: - Call find_lock_delalloc_range() with @delalloc_start = 0 Then we got a delalloc range [0, 4K). This range will be COWed. - Call find_lock_delalloc_range() again with @delalloc_start = 4K Since find_lock_delalloc_range() never cares whether the range is still inside page range [0, 64K), it will return range [64K, 128K). This range meets the condition for subpage compression, will go through async COW path. And async COW path will return @page_started. But that @page_started is now for range [64K, 128K), not for range [0, 64K). - writepage_dellloc() returned 1 for page [0, 64K) Thus page [0, 64K) will not be unlocked, nor its page dirty status will be cleared. Next time when we try to lock page [0, 64K) we will deadlock, as there is no one to release page [0, 64K). This problem will never happen for regular page size as one page only contains one sector. After the first find_lock_delalloc_range() call, the @delalloc_end will go beyond @page_end no matter if we found a delalloc range or not Thus this bug only happens for subpage, as now we need multiple runs to exhaust the delalloc range of a page. [FIX] Fix the problem by ensuring the delalloc range we ran at least started inside @locked_page. So that we will never get incorrect @page_started. And to prevent such problem from happening again: - Make find_lock_delalloc_range() return false if the found range is beyond @end value passed in. Since @end will be utilized now, add an ASSERT() to ensure we pass correct @end into find_lock_delalloc_range(). This also means, for selftests we needs to populate @end before calling find_lock_delalloc_range(). - New ASSERT() in find_lock_delalloc_range() Now we will make sure the @start/@end passed in at least covers part of the page. - New ASSERT() in run_delalloc_range() To make sure the range at least starts inside @locked page. - Use @delalloc_start as proper cursor, while @delalloc_end is always reset to @page_end. Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2021-09-27 07:22:07 +00:00
end = start + PAGE_SIZE - 1;
/*
* Currently if we fail to find dirty pages in the delalloc range we
* will adjust max_bytes down to PAGE_SIZE and then re-search. If
* this changes at any point in the future we will need to fix this
* tests expected behavior.
*/
found = find_lock_delalloc_range(inode, page_folio(locked_page), &start,
&end);
if (!found) {
test_err("didn't find our range");
goto out_bits;
}
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
if (start != test_start && end != test_start + PAGE_SIZE - 1) {
test_err("expected start %llu end %llu, got start %llu end %llu",
test_start, test_start + PAGE_SIZE - 1, start, end);
goto out_bits;
}
if (process_page_range(inode, start, end, PROCESS_TEST_LOCKED |
PROCESS_UNLOCK)) {
test_err("pages in range were not all locked");
goto out_bits;
}
ret = 0;
out_bits:
if (ret)
dump_extent_io_tree(tmp);
clear_extent_bits(tmp, 0, total_dirty - 1, (unsigned)-1);
out:
if (locked_page)
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 12:29:47 +00:00
put_page(locked_page);
process_page_range(inode, 0, total_dirty - 1,
PROCESS_UNLOCK | PROCESS_RELEASE);
iput(inode);
btrfs_free_dummy_root(root);
btrfs_free_dummy_fs_info(fs_info);
return ret;
}
static int check_eb_bitmap(unsigned long *bitmap, struct extent_buffer *eb)
{
unsigned long i;
for (i = 0; i < eb->len * BITS_PER_BYTE; i++) {
int bit, bit1;
bit = !!test_bit(i, bitmap);
bit1 = !!extent_buffer_test_bit(eb, 0, i);
if (bit1 != bit) {
u8 has;
u8 expect;
read_extent_buffer(eb, &has, i / BITS_PER_BYTE, 1);
expect = bitmap_get_value8(bitmap, ALIGN(i, BITS_PER_BYTE));
test_err(
"bits do not match, start byte 0 bit %lu, byte %lu has 0x%02x expect 0x%02x",
i, i / BITS_PER_BYTE, has, expect);
return -EINVAL;
}
bit1 = !!extent_buffer_test_bit(eb, i / BITS_PER_BYTE,
i % BITS_PER_BYTE);
if (bit1 != bit) {
u8 has;
u8 expect;
read_extent_buffer(eb, &has, i / BITS_PER_BYTE, 1);
expect = bitmap_get_value8(bitmap, ALIGN(i, BITS_PER_BYTE));
test_err(
"bits do not match, start byte %lu bit %lu, byte %lu has 0x%02x expect 0x%02x",
i / BITS_PER_BYTE, i % BITS_PER_BYTE,
i / BITS_PER_BYTE, has, expect);
return -EINVAL;
}
}
return 0;
}
static int test_bitmap_set(const char *name, unsigned long *bitmap,
struct extent_buffer *eb,
unsigned long byte_start, unsigned long bit_start,
unsigned long bit_len)
{
int ret;
bitmap_set(bitmap, byte_start * BITS_PER_BYTE + bit_start, bit_len);
extent_buffer_bitmap_set(eb, byte_start, bit_start, bit_len);
ret = check_eb_bitmap(bitmap, eb);
if (ret < 0)
test_err("%s test failed", name);
return ret;
}
static int test_bitmap_clear(const char *name, unsigned long *bitmap,
struct extent_buffer *eb,
unsigned long byte_start, unsigned long bit_start,
unsigned long bit_len)
{
int ret;
bitmap_clear(bitmap, byte_start * BITS_PER_BYTE + bit_start, bit_len);
extent_buffer_bitmap_clear(eb, byte_start, bit_start, bit_len);
ret = check_eb_bitmap(bitmap, eb);
if (ret < 0)
test_err("%s test failed", name);
return ret;
}
static int __test_eb_bitmaps(unsigned long *bitmap, struct extent_buffer *eb)
{
unsigned long i, j;
unsigned long byte_len = eb->len;
u32 x;
int ret;
ret = test_bitmap_clear("clear all run 1", bitmap, eb, 0, 0,
byte_len * BITS_PER_BYTE);
if (ret < 0)
return ret;
ret = test_bitmap_set("set all", bitmap, eb, 0, 0, byte_len * BITS_PER_BYTE);
if (ret < 0)
return ret;
ret = test_bitmap_clear("clear all run 2", bitmap, eb, 0, 0,
byte_len * BITS_PER_BYTE);
if (ret < 0)
return ret;
ret = test_bitmap_set("same byte set", bitmap, eb, 0, 2, 4);
if (ret < 0)
return ret;
ret = test_bitmap_clear("same byte partial clear", bitmap, eb, 0, 4, 1);
if (ret < 0)
return ret;
ret = test_bitmap_set("cross byte set", bitmap, eb, 2, 4, 8);
if (ret < 0)
return ret;
ret = test_bitmap_set("cross multi byte set", bitmap, eb, 4, 4, 24);
if (ret < 0)
return ret;
ret = test_bitmap_clear("cross byte clear", bitmap, eb, 2, 6, 4);
if (ret < 0)
return ret;
ret = test_bitmap_clear("cross multi byte clear", bitmap, eb, 4, 6, 20);
if (ret < 0)
return ret;
/* Straddling pages test */
if (byte_len > PAGE_SIZE) {
ret = test_bitmap_set("cross page set", bitmap, eb,
PAGE_SIZE - sizeof(long) / 2, 0,
sizeof(long) * BITS_PER_BYTE);
if (ret < 0)
return ret;
ret = test_bitmap_set("cross page set all", bitmap, eb, 0, 0,
byte_len * BITS_PER_BYTE);
if (ret < 0)
return ret;
ret = test_bitmap_clear("cross page clear", bitmap, eb,
PAGE_SIZE - sizeof(long) / 2, 0,
sizeof(long) * BITS_PER_BYTE);
if (ret < 0)
return ret;
}
/*
* Generate a wonky pseudo-random bit pattern for the sake of not using
* something repetitive that could miss some hypothetical off-by-n bug.
*/
x = 0;
ret = test_bitmap_clear("clear all run 3", bitmap, eb, 0, 0,
byte_len * BITS_PER_BYTE);
if (ret < 0)
return ret;
for (i = 0; i < byte_len * BITS_PER_BYTE / 32; i++) {
x = (0x19660dULL * (u64)x + 0x3c6ef35fULL) & 0xffffffffU;
for (j = 0; j < 32; j++) {
if (x & (1U << j)) {
bitmap_set(bitmap, i * 32 + j, 1);
extent_buffer_bitmap_set(eb, 0, i * 32 + j, 1);
}
}
}
ret = check_eb_bitmap(bitmap, eb);
if (ret) {
test_err("random bit pattern failed");
return ret;
}
return 0;
}
static int test_eb_bitmaps(u32 sectorsize, u32 nodesize)
{
struct btrfs_fs_info *fs_info;
unsigned long *bitmap = NULL;
struct extent_buffer *eb = NULL;
int ret;
test_msg("running extent buffer bitmap tests");
btrfs: tests: remove invalid extent-io test In extent-io-test, there are two invalid tests: - Invalid nodesize for test_eb_bitmaps() Instead of the sectorsize and nodesize combination passed in, we're always using hand-crafted nodesize, e.g: len = (sectorsize < BTRFS_MAX_METADATA_BLOCKSIZE) ? sectorsize * 4 : sectorsize; In above case, if we have 32K page size, then we will get a length of 128K, which is beyond max node size, and obviously invalid. The common page size goes up to 64K so we haven't hit that - Invalid extent buffer bytenr For 64K page size, the only combination we're going to test is sectorsize = nodesize = 64K. However, in that case we will try to test an eb which bytenr is not sectorsize aligned: /* Do it over again with an extent buffer which isn't page-aligned. */ eb = __alloc_dummy_extent_buffer(fs_info, nodesize / 2, len); Sector alignment is a hard requirement for any sector size. The only exception is superblock. But anything else should follow sector size alignment. This is definitely an invalid test case. This patch will fix both problems by: - Honor the sectorsize/nodesize combination Now we won't bother to hand-craft the length and use it as nodesize. - Use sectorsize as the 2nd run extent buffer start This would test the case where extent buffer is aligned to sectorsize but not always aligned to nodesize. Please note that, later subpage related cleanup will reduce extent_buffer::pages[] to exactly what we need, making the sector unaligned extent buffer operations cause problems. Since only extent_io self tests utilize this, this patch is required for all later cleanup/refactoring. Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2020-11-13 12:51:27 +00:00
fs_info = btrfs_alloc_dummy_fs_info(nodesize, sectorsize);
if (!fs_info) {
test_std_err(TEST_ALLOC_FS_INFO);
return -ENOMEM;
}
btrfs: tests: remove invalid extent-io test In extent-io-test, there are two invalid tests: - Invalid nodesize for test_eb_bitmaps() Instead of the sectorsize and nodesize combination passed in, we're always using hand-crafted nodesize, e.g: len = (sectorsize < BTRFS_MAX_METADATA_BLOCKSIZE) ? sectorsize * 4 : sectorsize; In above case, if we have 32K page size, then we will get a length of 128K, which is beyond max node size, and obviously invalid. The common page size goes up to 64K so we haven't hit that - Invalid extent buffer bytenr For 64K page size, the only combination we're going to test is sectorsize = nodesize = 64K. However, in that case we will try to test an eb which bytenr is not sectorsize aligned: /* Do it over again with an extent buffer which isn't page-aligned. */ eb = __alloc_dummy_extent_buffer(fs_info, nodesize / 2, len); Sector alignment is a hard requirement for any sector size. The only exception is superblock. But anything else should follow sector size alignment. This is definitely an invalid test case. This patch will fix both problems by: - Honor the sectorsize/nodesize combination Now we won't bother to hand-craft the length and use it as nodesize. - Use sectorsize as the 2nd run extent buffer start This would test the case where extent buffer is aligned to sectorsize but not always aligned to nodesize. Please note that, later subpage related cleanup will reduce extent_buffer::pages[] to exactly what we need, making the sector unaligned extent buffer operations cause problems. Since only extent_io self tests utilize this, this patch is required for all later cleanup/refactoring. Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2020-11-13 12:51:27 +00:00
bitmap = kmalloc(nodesize, GFP_KERNEL);
if (!bitmap) {
test_err("couldn't allocate test bitmap");
ret = -ENOMEM;
goto out;
}
btrfs: tests: remove invalid extent-io test In extent-io-test, there are two invalid tests: - Invalid nodesize for test_eb_bitmaps() Instead of the sectorsize and nodesize combination passed in, we're always using hand-crafted nodesize, e.g: len = (sectorsize < BTRFS_MAX_METADATA_BLOCKSIZE) ? sectorsize * 4 : sectorsize; In above case, if we have 32K page size, then we will get a length of 128K, which is beyond max node size, and obviously invalid. The common page size goes up to 64K so we haven't hit that - Invalid extent buffer bytenr For 64K page size, the only combination we're going to test is sectorsize = nodesize = 64K. However, in that case we will try to test an eb which bytenr is not sectorsize aligned: /* Do it over again with an extent buffer which isn't page-aligned. */ eb = __alloc_dummy_extent_buffer(fs_info, nodesize / 2, len); Sector alignment is a hard requirement for any sector size. The only exception is superblock. But anything else should follow sector size alignment. This is definitely an invalid test case. This patch will fix both problems by: - Honor the sectorsize/nodesize combination Now we won't bother to hand-craft the length and use it as nodesize. - Use sectorsize as the 2nd run extent buffer start This would test the case where extent buffer is aligned to sectorsize but not always aligned to nodesize. Please note that, later subpage related cleanup will reduce extent_buffer::pages[] to exactly what we need, making the sector unaligned extent buffer operations cause problems. Since only extent_io self tests utilize this, this patch is required for all later cleanup/refactoring. Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2020-11-13 12:51:27 +00:00
eb = __alloc_dummy_extent_buffer(fs_info, 0, nodesize);
if (!eb) {
test_std_err(TEST_ALLOC_ROOT);
ret = -ENOMEM;
goto out;
}
ret = __test_eb_bitmaps(bitmap, eb);
if (ret)
goto out;
free_extent_buffer(eb);
btrfs: tests: remove invalid extent-io test In extent-io-test, there are two invalid tests: - Invalid nodesize for test_eb_bitmaps() Instead of the sectorsize and nodesize combination passed in, we're always using hand-crafted nodesize, e.g: len = (sectorsize < BTRFS_MAX_METADATA_BLOCKSIZE) ? sectorsize * 4 : sectorsize; In above case, if we have 32K page size, then we will get a length of 128K, which is beyond max node size, and obviously invalid. The common page size goes up to 64K so we haven't hit that - Invalid extent buffer bytenr For 64K page size, the only combination we're going to test is sectorsize = nodesize = 64K. However, in that case we will try to test an eb which bytenr is not sectorsize aligned: /* Do it over again with an extent buffer which isn't page-aligned. */ eb = __alloc_dummy_extent_buffer(fs_info, nodesize / 2, len); Sector alignment is a hard requirement for any sector size. The only exception is superblock. But anything else should follow sector size alignment. This is definitely an invalid test case. This patch will fix both problems by: - Honor the sectorsize/nodesize combination Now we won't bother to hand-craft the length and use it as nodesize. - Use sectorsize as the 2nd run extent buffer start This would test the case where extent buffer is aligned to sectorsize but not always aligned to nodesize. Please note that, later subpage related cleanup will reduce extent_buffer::pages[] to exactly what we need, making the sector unaligned extent buffer operations cause problems. Since only extent_io self tests utilize this, this patch is required for all later cleanup/refactoring. Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2020-11-13 12:51:27 +00:00
/*
* Test again for case where the tree block is sectorsize aligned but
* not nodesize aligned.
*/
eb = __alloc_dummy_extent_buffer(fs_info, sectorsize, nodesize);
if (!eb) {
test_std_err(TEST_ALLOC_ROOT);
ret = -ENOMEM;
goto out;
}
ret = __test_eb_bitmaps(bitmap, eb);
out:
free_extent_buffer(eb);
kfree(bitmap);
btrfs_free_dummy_fs_info(fs_info);
return ret;
}
static int test_find_first_clear_extent_bit(void)
{
struct extent_io_tree tree;
u64 start, end;
int ret = -EINVAL;
test_msg("running find_first_clear_extent_bit test");
btrfs: Correctly handle empty trees in find_first_clear_extent_bit Raviu reported that running his regular fs_trim segfaulted with the following backtrace: [ 237.525947] assertion failed: prev, in ../fs/btrfs/extent_io.c:1595 [ 237.525984] ------------[ cut here ]------------ [ 237.525985] kernel BUG at ../fs/btrfs/ctree.h:3117! [ 237.525992] invalid opcode: 0000 [#1] SMP PTI [ 237.525998] CPU: 4 PID: 4423 Comm: fstrim Tainted: G U OE 5.4.14-8-vanilla #1 [ 237.526001] Hardware name: ASUSTeK COMPUTER INC. [ 237.526044] RIP: 0010:assfail.constprop.58+0x18/0x1a [btrfs] [ 237.526079] Call Trace: [ 237.526120] find_first_clear_extent_bit+0x13d/0x150 [btrfs] [ 237.526148] btrfs_trim_fs+0x211/0x3f0 [btrfs] [ 237.526184] btrfs_ioctl_fitrim+0x103/0x170 [btrfs] [ 237.526219] btrfs_ioctl+0x129a/0x2ed0 [btrfs] [ 237.526227] ? filemap_map_pages+0x190/0x3d0 [ 237.526232] ? do_filp_open+0xaf/0x110 [ 237.526238] ? _copy_to_user+0x22/0x30 [ 237.526242] ? cp_new_stat+0x150/0x180 [ 237.526247] ? do_vfs_ioctl+0xa4/0x640 [ 237.526278] ? btrfs_ioctl_get_supported_features+0x30/0x30 [btrfs] [ 237.526283] do_vfs_ioctl+0xa4/0x640 [ 237.526288] ? __do_sys_newfstat+0x3c/0x60 [ 237.526292] ksys_ioctl+0x70/0x80 [ 237.526297] __x64_sys_ioctl+0x16/0x20 [ 237.526303] do_syscall_64+0x5a/0x1c0 [ 237.526310] entry_SYSCALL_64_after_hwframe+0x49/0xbe That was due to btrfs_fs_device::aloc_tree being empty. Initially I thought this wasn't possible and as a percaution have put the assert in find_first_clear_extent_bit. Turns out this is indeed possible and could happen when a file system with SINGLE data/metadata profile has a 2nd device added. Until balance is run or a new chunk is allocated on this device it will be completely empty. In this case find_first_clear_extent_bit should return the full range [0, -1ULL] and let the caller handle this i.e for trim the end will be capped at the size of actual device. Link: https://lore.kernel.org/linux-btrfs/izW2WNyvy1dEDweBICizKnd2KDwDiDyY2EYQr4YCwk7pkuIpthx-JRn65MPBde00ND6V0_Lh8mW0kZwzDiLDv25pUYWxkskWNJnVP0kgdMA=@protonmail.com/ Fixes: 45bfcfc168f8 ("btrfs: Implement find_first_clear_extent_bit") CC: stable@vger.kernel.org # 5.2+ Signed-off-by: Nikolay Borisov <nborisov@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2020-01-27 09:59:26 +00:00
extent_io_tree_init(NULL, &tree, IO_TREE_SELFTEST);
btrfs: Correctly handle empty trees in find_first_clear_extent_bit Raviu reported that running his regular fs_trim segfaulted with the following backtrace: [ 237.525947] assertion failed: prev, in ../fs/btrfs/extent_io.c:1595 [ 237.525984] ------------[ cut here ]------------ [ 237.525985] kernel BUG at ../fs/btrfs/ctree.h:3117! [ 237.525992] invalid opcode: 0000 [#1] SMP PTI [ 237.525998] CPU: 4 PID: 4423 Comm: fstrim Tainted: G U OE 5.4.14-8-vanilla #1 [ 237.526001] Hardware name: ASUSTeK COMPUTER INC. [ 237.526044] RIP: 0010:assfail.constprop.58+0x18/0x1a [btrfs] [ 237.526079] Call Trace: [ 237.526120] find_first_clear_extent_bit+0x13d/0x150 [btrfs] [ 237.526148] btrfs_trim_fs+0x211/0x3f0 [btrfs] [ 237.526184] btrfs_ioctl_fitrim+0x103/0x170 [btrfs] [ 237.526219] btrfs_ioctl+0x129a/0x2ed0 [btrfs] [ 237.526227] ? filemap_map_pages+0x190/0x3d0 [ 237.526232] ? do_filp_open+0xaf/0x110 [ 237.526238] ? _copy_to_user+0x22/0x30 [ 237.526242] ? cp_new_stat+0x150/0x180 [ 237.526247] ? do_vfs_ioctl+0xa4/0x640 [ 237.526278] ? btrfs_ioctl_get_supported_features+0x30/0x30 [btrfs] [ 237.526283] do_vfs_ioctl+0xa4/0x640 [ 237.526288] ? __do_sys_newfstat+0x3c/0x60 [ 237.526292] ksys_ioctl+0x70/0x80 [ 237.526297] __x64_sys_ioctl+0x16/0x20 [ 237.526303] do_syscall_64+0x5a/0x1c0 [ 237.526310] entry_SYSCALL_64_after_hwframe+0x49/0xbe That was due to btrfs_fs_device::aloc_tree being empty. Initially I thought this wasn't possible and as a percaution have put the assert in find_first_clear_extent_bit. Turns out this is indeed possible and could happen when a file system with SINGLE data/metadata profile has a 2nd device added. Until balance is run or a new chunk is allocated on this device it will be completely empty. In this case find_first_clear_extent_bit should return the full range [0, -1ULL] and let the caller handle this i.e for trim the end will be capped at the size of actual device. Link: https://lore.kernel.org/linux-btrfs/izW2WNyvy1dEDweBICizKnd2KDwDiDyY2EYQr4YCwk7pkuIpthx-JRn65MPBde00ND6V0_Lh8mW0kZwzDiLDv25pUYWxkskWNJnVP0kgdMA=@protonmail.com/ Fixes: 45bfcfc168f8 ("btrfs: Implement find_first_clear_extent_bit") CC: stable@vger.kernel.org # 5.2+ Signed-off-by: Nikolay Borisov <nborisov@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2020-01-27 09:59:26 +00:00
/* Test correct handling of empty tree */
find_first_clear_extent_bit(&tree, 0, &start, &end, CHUNK_TRIMMED);
if (start != 0 || end != -1) {
test_err(
"error getting a range from completely empty tree: start %llu end %llu",
start, end);
goto out;
}
/*
* Set 1M-4M alloc/discard and 32M-64M thus leaving a hole between
* 4M-32M
*/
set_extent_bit(&tree, SZ_1M, SZ_4M - 1,
CHUNK_TRIMMED | CHUNK_ALLOCATED, NULL);
find_first_clear_extent_bit(&tree, SZ_512K, &start, &end,
CHUNK_TRIMMED | CHUNK_ALLOCATED);
if (start != 0 || end != SZ_1M - 1) {
test_err("error finding beginning range: start %llu end %llu",
start, end);
goto out;
}
/* Now add 32M-64M so that we have a hole between 4M-32M */
set_extent_bit(&tree, SZ_32M, SZ_64M - 1,
CHUNK_TRIMMED | CHUNK_ALLOCATED, NULL);
/*
* Request first hole starting at 12M, we should get 4M-32M
*/
find_first_clear_extent_bit(&tree, 12 * SZ_1M, &start, &end,
CHUNK_TRIMMED | CHUNK_ALLOCATED);
if (start != SZ_4M || end != SZ_32M - 1) {
test_err("error finding trimmed range: start %llu end %llu",
start, end);
goto out;
}
/*
* Search in the middle of allocated range, should get the next one
* available, which happens to be unallocated -> 4M-32M
*/
find_first_clear_extent_bit(&tree, SZ_2M, &start, &end,
CHUNK_TRIMMED | CHUNK_ALLOCATED);
if (start != SZ_4M || end != SZ_32M - 1) {
test_err("error finding next unalloc range: start %llu end %llu",
start, end);
goto out;
}
/*
* Set 64M-72M with CHUNK_ALLOC flag, then search for CHUNK_TRIMMED flag
* being unset in this range, we should get the entry in range 64M-72M
*/
set_extent_bit(&tree, SZ_64M, SZ_64M + SZ_8M - 1, CHUNK_ALLOCATED, NULL);
find_first_clear_extent_bit(&tree, SZ_64M + SZ_1M, &start, &end,
CHUNK_TRIMMED);
if (start != SZ_64M || end != SZ_64M + SZ_8M - 1) {
test_err("error finding exact range: start %llu end %llu",
start, end);
goto out;
}
find_first_clear_extent_bit(&tree, SZ_64M - SZ_8M, &start, &end,
CHUNK_TRIMMED);
/*
* Search in the middle of set range whose immediate neighbour doesn't
* have the bits set so it must be returned
*/
if (start != SZ_64M || end != SZ_64M + SZ_8M - 1) {
test_err("error finding next alloc range: start %llu end %llu",
start, end);
goto out;
}
/*
* Search beyond any known range, shall return after last known range
* and end should be -1
*/
find_first_clear_extent_bit(&tree, -1, &start, &end, CHUNK_TRIMMED);
if (start != SZ_64M + SZ_8M || end != -1) {
test_err(
"error handling beyond end of range search: start %llu end %llu",
start, end);
goto out;
}
ret = 0;
out:
if (ret)
dump_extent_io_tree(&tree);
Btrfs: fix memory leaks in the test test_find_first_clear_extent_bit The test creates an extent io tree and sets several ranges with the CHUNK_ALLOCATED and CHUNK_TRIMMED bits, resulting in the allocation of several extent state structures. However the test never clears those ranges, resulting in memory leaks of the extent state structures. This is detected when CONFIG_BTRFS_DEBUG is set once we remove the btrfs module (rmmod btrfs): [57399.787918] BTRFS: state leak: start 67108864 end 75497471 state 1 in tree 1 refs 1 [57399.790155] BTRFS: state leak: start 33554432 end 67108863 state 33 in tree 1 refs 1 [57399.791941] BTRFS: state leak: start 1048576 end 4194303 state 33 in tree 1 refs 1 [57399.793753] BTRFS: state leak: start 67108864 end 75497471 state 1 in tree 1 refs 1 [57399.795188] BTRFS: state leak: start 33554432 end 67108863 state 33 in tree 1 refs 1 [57399.796453] BTRFS: state leak: start 1048576 end 4194303 state 33 in tree 1 refs 1 [57399.797765] BTRFS: state leak: start 67108864 end 75497471 state 1 in tree 1 refs 1 [57399.799049] BTRFS: state leak: start 33554432 end 67108863 state 33 in tree 1 refs 1 [57399.800142] BTRFS: state leak: start 1048576 end 4194303 state 33 in tree 1 refs 1 [57399.801126] BTRFS: state leak: start 67108864 end 75497471 state 1 in tree 1 refs 1 [57399.802106] BTRFS: state leak: start 33554432 end 67108863 state 33 in tree 1 refs 1 [57399.803119] BTRFS: state leak: start 1048576 end 4194303 state 33 in tree 1 refs 1 [57399.804153] BTRFS: state leak: start 67108864 end 75497471 state 1 in tree 1 refs 1 [57399.805196] BTRFS: state leak: start 33554432 end 67108863 state 33 in tree 1 refs 1 [57399.806191] BTRFS: state leak: start 1048576 end 4194303 state 33 in tree 1 refs 1 The start and end offsets reported correspond exactly to the ranges used by the test. So fix that by clearing all the ranges when the test finishes. Fixes: 1eaebb341d2b41 ("btrfs: Don't trim returned range based on input value in find_first_clear_extent_bit") Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
2019-08-03 08:53:16 +00:00
clear_extent_bits(&tree, 0, (u64)-1, CHUNK_TRIMMED | CHUNK_ALLOCATED);
return ret;
}
static void dump_eb_and_memory_contents(struct extent_buffer *eb, void *memory,
const char *test_name)
{
for (int i = 0; i < eb->len; i++) {
struct page *page = folio_page(eb->folios[i >> PAGE_SHIFT], 0);
void *addr = page_address(page) + offset_in_page(i);
if (memcmp(addr, memory + i, 1) != 0) {
test_err("%s failed", test_name);
test_err("eb and memory diffs at byte %u, eb has 0x%02x memory has 0x%02x",
i, *(u8 *)addr, *(u8 *)(memory + i));
return;
}
}
}
static int verify_eb_and_memory(struct extent_buffer *eb, void *memory,
const char *test_name)
{
for (int i = 0; i < (eb->len >> PAGE_SHIFT); i++) {
void *eb_addr = folio_address(eb->folios[i]);
if (memcmp(memory + (i << PAGE_SHIFT), eb_addr, PAGE_SIZE) != 0) {
dump_eb_and_memory_contents(eb, memory, test_name);
return -EUCLEAN;
}
}
return 0;
}
/*
* Init both memory and extent buffer contents to the same randomly generated
* contents.
*/
static void init_eb_and_memory(struct extent_buffer *eb, void *memory)
{
get_random_bytes(memory, eb->len);
write_extent_buffer(eb, memory, 0, eb->len);
}
static int test_eb_mem_ops(u32 sectorsize, u32 nodesize)
{
struct btrfs_fs_info *fs_info;
struct extent_buffer *eb = NULL;
void *memory = NULL;
int ret;
test_msg("running extent buffer memory operation tests");
fs_info = btrfs_alloc_dummy_fs_info(nodesize, sectorsize);
if (!fs_info) {
test_std_err(TEST_ALLOC_FS_INFO);
return -ENOMEM;
}
memory = kvzalloc(nodesize, GFP_KERNEL);
if (!memory) {
test_err("failed to allocate memory");
ret = -ENOMEM;
goto out;
}
eb = __alloc_dummy_extent_buffer(fs_info, SZ_1M, nodesize);
if (!eb) {
test_std_err(TEST_ALLOC_EXTENT_BUFFER);
ret = -ENOMEM;
goto out;
}
init_eb_and_memory(eb, memory);
ret = verify_eb_and_memory(eb, memory, "full eb write");
if (ret < 0)
goto out;
memcpy(memory, memory + 16, 16);
memcpy_extent_buffer(eb, 0, 16, 16);
ret = verify_eb_and_memory(eb, memory, "same page non-overlapping memcpy 1");
if (ret < 0)
goto out;
memcpy(memory, memory + 2048, 16);
memcpy_extent_buffer(eb, 0, 2048, 16);
ret = verify_eb_and_memory(eb, memory, "same page non-overlapping memcpy 2");
if (ret < 0)
goto out;
memcpy(memory, memory + 2048, 2048);
memcpy_extent_buffer(eb, 0, 2048, 2048);
ret = verify_eb_and_memory(eb, memory, "same page non-overlapping memcpy 3");
if (ret < 0)
goto out;
memmove(memory + 512, memory + 256, 512);
memmove_extent_buffer(eb, 512, 256, 512);
ret = verify_eb_and_memory(eb, memory, "same page overlapping memcpy 1");
if (ret < 0)
goto out;
memmove(memory + 2048, memory + 512, 2048);
memmove_extent_buffer(eb, 2048, 512, 2048);
ret = verify_eb_and_memory(eb, memory, "same page overlapping memcpy 2");
if (ret < 0)
goto out;
memmove(memory + 512, memory + 2048, 2048);
memmove_extent_buffer(eb, 512, 2048, 2048);
ret = verify_eb_and_memory(eb, memory, "same page overlapping memcpy 3");
if (ret < 0)
goto out;
if (nodesize > PAGE_SIZE) {
memcpy(memory, memory + 4096 - 128, 256);
memcpy_extent_buffer(eb, 0, 4096 - 128, 256);
ret = verify_eb_and_memory(eb, memory, "cross page non-overlapping memcpy 1");
if (ret < 0)
goto out;
memcpy(memory + 4096 - 128, memory + 4096 + 128, 256);
memcpy_extent_buffer(eb, 4096 - 128, 4096 + 128, 256);
ret = verify_eb_and_memory(eb, memory, "cross page non-overlapping memcpy 2");
if (ret < 0)
goto out;
memmove(memory + 4096 - 128, memory + 4096 - 64, 256);
memmove_extent_buffer(eb, 4096 - 128, 4096 - 64, 256);
ret = verify_eb_and_memory(eb, memory, "cross page overlapping memcpy 1");
if (ret < 0)
goto out;
memmove(memory + 4096 - 64, memory + 4096 - 128, 256);
memmove_extent_buffer(eb, 4096 - 64, 4096 - 128, 256);
ret = verify_eb_and_memory(eb, memory, "cross page overlapping memcpy 2");
if (ret < 0)
goto out;
}
out:
free_extent_buffer(eb);
kvfree(memory);
btrfs_free_dummy_fs_info(fs_info);
return ret;
}
int btrfs_test_extent_io(u32 sectorsize, u32 nodesize)
{
int ret;
test_msg("running extent I/O tests");
ret = test_find_delalloc(sectorsize, nodesize);
if (ret)
goto out;
ret = test_find_first_clear_extent_bit();
if (ret)
goto out;
ret = test_eb_bitmaps(sectorsize, nodesize);
if (ret)
goto out;
ret = test_eb_mem_ops(sectorsize, nodesize);
out:
return ret;
}