On 03/05/2023 12:49, Hui Wang wrote:
On 4/29/23 03:53, Michal Hocko wrote:
On Thu 27-04-23 11:47:10, Hui Wang wrote:
[...]
So Michal,
Don't know if you read the "[PATCH 0/1] mm/oom_kill: system enters a
state
something like hang when running stress-ng", do you know why
out_of_memory()
will return immediately if there is no __GFP_FS, could we drop these
lines
directly:
/*
* The OOM killer does not compensate for IO-less reclaim.
* pagefault_out_of_memory lost its gfp context so we have to
* make sure exclude 0 mask - all other users should have at least
* ___GFP_DIRECT_RECLAIM to get here. But mem_cgroup_oom() has to
* invoke the OOM killer even if it is a GFP_NOFS allocation.
*/
if (oc->gfp_mask && !(oc->gfp_mask & __GFP_FS) &&
!is_memcg_oom(oc))
return true;
The comment is rather hard to grasp without an intimate knowledge of the
memory reclaim. The primary reason is that the allocation context
without __GFP_FS (and also __GFP_IO) cannot perform a full memory
reclaim because fs or the storage subsystem might be holding locks
required for the memory reclaim. This means that a large amount of
reclaimable memory is out of sight of the specific direct reclaim
context. If we allowed oom killer to trigger we could invoke the oom
killer while there is a lot of otherwise reclaimable memory. As you can
imagine not something many users would appreciate as the oom kill is a
very disruptive operation. In this case we rely on kswapd or other
GFP_KERNEL like allocation context to make forward instead. If there is
really nothing reclaimable then the oom killer would eventually hit from
elsewhere.
HTH
Hi Michal,
Understand. Thanks for explanation. So we can't remove those 2 lines
of code.
Here in my patch, letting a kthread allocate a page with GFP_KERNEL,
It could possibly trigger the reclaim and if nothing reclaimable,
trigger the oom killer. Do you think it is a safe workaround for the
issue we are facing currently?
And Hi Phillip,
What is your opinion on it, do you have a direction to solve this
issue from filesystem?
The following patch creates the concept of "squashfs contexts", which
moves all memory dynamically allocated (in a readahead/read_page path)
into a single structure which can be allocated and deleted once. It
then creates a pool of these at filesystem mount time. Threads entering
readahead/read_page will take a context from the pool, and will then
perform no dynamic memory allocation.
The final patch-series will make this a non-default build option for
systems that need this.
Phillip
From e4dd08d0850796132f9874cd7e0808373d86cbf2 Mon Sep 17 00:00:00 2001
From: Phillip Lougher <phillip@xxxxxxxxxxxxxxx>
Date: Tue, 2 May 2023 20:16:55 +0100
Subject: [PATCH] read_context
Signed-off-by: Phillip Lougher <phillip@xxxxxxxxxxxxxxx>
---
fs/squashfs/Makefile | 1 +
fs/squashfs/block.c | 67 ++++-----
fs/squashfs/cache.c | 66 ++++++---
fs/squashfs/decompressor.c | 14 +-
fs/squashfs/decompressor_multi.c | 2 +
fs/squashfs/decompressor_multi_percpu.c | 2 +
fs/squashfs/decompressor_single.c | 2 +
fs/squashfs/file.c | 61 ++++----
fs/squashfs/file_direct.c | 59 ++++----
fs/squashfs/page_actor.c | 37 ++---
fs/squashfs/page_actor.h | 8 +-
fs/squashfs/read_context.c | 178 ++++++++++++++++++++++++
fs/squashfs/read_context.h | 31 +++++
fs/squashfs/squashfs.h | 12 +-
fs/squashfs/squashfs_fs_sb.h | 3 +-
fs/squashfs/super.c | 13 +-
fs/squashfs/symlink.c | 3 +-
17 files changed, 387 insertions(+), 172 deletions(-)
create mode 100644 fs/squashfs/read_context.c
create mode 100644 fs/squashfs/read_context.h
diff --git a/fs/squashfs/Makefile b/fs/squashfs/Makefile
index 477c89a519ee..93511720e94f 100644
--- a/fs/squashfs/Makefile
+++ b/fs/squashfs/Makefile
@@ -6,6 +6,7 @@
obj-$(CONFIG_SQUASHFS) += squashfs.o
squashfs-y += block.o cache.o dir.o export.o file.o fragment.o id.o inode.o
squashfs-y += namei.o super.o symlink.o decompressor.o page_actor.o
+squashfs-y += read_context.o
squashfs-$(CONFIG_SQUASHFS_FILE_CACHE) += file_cache.o
squashfs-$(CONFIG_SQUASHFS_FILE_DIRECT) += file_direct.o
squashfs-$(CONFIG_SQUASHFS_DECOMP_SINGLE) += decompressor_single.o
diff --git a/fs/squashfs/block.c b/fs/squashfs/block.c
index bed3bb8b27fa..110fcfd2a0eb 100644
--- a/fs/squashfs/block.c
+++ b/fs/squashfs/block.c
@@ -23,9 +23,10 @@
#include "squashfs_fs.h"
#include "squashfs_fs_sb.h"
+#include "read_context.h"
+#include "page_actor.h"
#include "squashfs.h"
#include "decompressor.h"
-#include "page_actor.h"
/*
* Returns the amount of bytes copied to the page actor.
@@ -77,7 +78,7 @@ static int copy_bio_to_actor(struct bio *bio,
}
static int squashfs_bio_read(struct super_block *sb, u64 index, int length,
- struct bio **biop, int *block_offset)
+ struct squashfs_context *context, int *block_offset)
{
struct squashfs_sb_info *msblk = sb->s_fs_info;
const u64 read_start = round_down(index, msblk->devblksize);
@@ -87,44 +88,33 @@ static int squashfs_bio_read(struct super_block *sb, u64 index, int length,
int offset = read_start - round_down(index, PAGE_SIZE);
int total_len = (block_end - block) << msblk->devblksize_log2;
const int page_count = DIV_ROUND_UP(total_len + offset, PAGE_SIZE);
- int error, i;
- struct bio *bio;
+ struct bio *bio = context->bio;
+ int error = -EIO, i;
- bio = bio_kmalloc(page_count, GFP_NOIO);
- if (!bio)
- return -ENOMEM;
bio_init(bio, sb->s_bdev, bio->bi_inline_vecs, page_count, REQ_OP_READ);
bio->bi_iter.bi_sector = block * (msblk->devblksize >> SECTOR_SHIFT);
for (i = 0; i < page_count; ++i) {
unsigned int len =
min_t(unsigned int, PAGE_SIZE - offset, total_len);
- struct page *page = alloc_page(GFP_NOIO);
+ int res = bio_add_page(bio, context->bio_page[i], len, offset);
+
+ if (!res)
+ goto out_uninit_bio;
- if (!page) {
- error = -ENOMEM;
- goto out_free_bio;
- }
- if (!bio_add_page(bio, page, len, offset)) {
- error = -EIO;
- goto out_free_bio;
- }
offset = 0;
total_len -= len;
}
error = submit_bio_wait(bio);
if (error)
- goto out_free_bio;
+ goto out_uninit_bio;
- *biop = bio;
*block_offset = index & ((1 << msblk->devblksize_log2) - 1);
return 0;
-out_free_bio:
- bio_free_pages(bio);
- bio_uninit(bio);
- kfree(bio);
+out_uninit_bio:
+ bio_uninit(context->bio);
return error;
}
@@ -138,10 +128,10 @@ static int squashfs_bio_read(struct super_block *sb, u64 index, int length,
* algorithms).
*/
int squashfs_read_data(struct super_block *sb, u64 index, int length,
- u64 *next_index, struct squashfs_page_actor *output)
+ u64 *next_index, struct squashfs_context *context)
{
struct squashfs_sb_info *msblk = sb->s_fs_info;
- struct bio *bio = NULL;
+ struct squashfs_page_actor *output = context->actor;
int compressed;
int res;
int offset;
@@ -166,13 +156,13 @@ int squashfs_read_data(struct super_block *sb, u64 index, int length,
res = -EIO;
goto out;
}
- res = squashfs_bio_read(sb, index, 2, &bio, &offset);
+ res = squashfs_bio_read(sb, index, 2, context, &offset);
if (res)
goto out;
- if (WARN_ON_ONCE(!bio_next_segment(bio, &iter_all))) {
+ if (WARN_ON_ONCE(!bio_next_segment(context->bio, &iter_all))) {
res = -EIO;
- goto out_free_bio;
+ goto out_uninit_bio;
}
/* Extract the length of the metadata block */
data = bvec_virt(bvec);
@@ -180,16 +170,14 @@ int squashfs_read_data(struct super_block *sb, u64 index, int length,
if (offset < bvec->bv_len - 1) {
length |= data[offset + 1] << 8;
} else {
- if (WARN_ON_ONCE(!bio_next_segment(bio, &iter_all))) {
+ if (WARN_ON_ONCE(!bio_next_segment(context->bio, &iter_all))) {
res = -EIO;
- goto out_free_bio;
+ goto out_uninit_bio;
}
data = bvec_virt(bvec);
length |= data[0] << 8;
}
- bio_free_pages(bio);
- bio_uninit(bio);
- kfree(bio);
+ bio_uninit(context->bio);
compressed = SQUASHFS_COMPRESSED(length);
length = SQUASHFS_COMPRESSED_SIZE(length);
@@ -207,24 +195,23 @@ int squashfs_read_data(struct super_block *sb, u64 index, int length,
if (next_index)
*next_index = index + length;
- res = squashfs_bio_read(sb, index, length, &bio, &offset);
+ res = squashfs_bio_read(sb, index, length, context, &offset);
if (res)
goto out;
if (compressed) {
if (!msblk->stream) {
res = -EIO;
- goto out_free_bio;
+ goto out_uninit_bio;
}
- res = msblk->thread_ops->decompress(msblk, bio, offset, length, output);
+ res = msblk->thread_ops->decompress(msblk, context->bio,
+ offset, length, output);
} else {
- res = copy_bio_to_actor(bio, output, offset, length);
+ res = copy_bio_to_actor(context->bio, output, offset, length);
}
-out_free_bio:
- bio_free_pages(bio);
- bio_uninit(bio);
- kfree(bio);
+out_uninit_bio:
+ bio_uninit(context->bio);
out:
if (res < 0) {
ERROR("Failed to read block 0x%llx: %d\n", index, res);
diff --git a/fs/squashfs/cache.c b/fs/squashfs/cache.c
index 5062326d0efb..98af4f696edd 100644
--- a/fs/squashfs/cache.c
+++ b/fs/squashfs/cache.c
@@ -42,19 +42,24 @@
#include "squashfs_fs.h"
#include "squashfs_fs_sb.h"
-#include "squashfs.h"
+#include "read_context.h"
#include "page_actor.h"
+#include "squashfs.h"
/*
* Look-up block in cache, and increment usage count. If not in cache, read
* and decompress it from disk.
*/
struct squashfs_cache_entry *squashfs_cache_get(struct super_block *sb,
- struct squashfs_cache *cache, u64 block, int length)
+ struct squashfs_cache *cache, struct squashfs_contexts *contexts,
+ struct squashfs_context *context, u64 block, int length)
{
- int i, n;
+ int i, n, need_context = context == NULL;
struct squashfs_cache_entry *entry;
+ if (cache->contexts)
+ contexts = cache->contexts;
+
spin_lock(&cache->lock);
while (1) {
@@ -107,8 +112,17 @@ struct squashfs_cache_entry *squashfs_cache_get(struct super_block *sb,
entry->error = 0;
spin_unlock(&cache->lock);
+ if (need_context)
+ context = squashfs_get_context(contexts);
+
+ squashfs_page_actor_init(context->actor, entry->data,
+ cache->pages, 0);
+
entry->length = squashfs_read_data(sb, block, length,
- &entry->next_index, entry->actor);
+ &entry->next_index, context);
+
+ if (need_context)
+ squashfs_put_context(contexts, context);
spin_lock(&cache->lock);
@@ -207,9 +221,9 @@ void squashfs_cache_delete(struct squashfs_cache *cache)
kfree(cache->entry[i].data[j]);
kfree(cache->entry[i].data);
}
- kfree(cache->entry[i].actor);
}
+ squashfs_delete_contexts(cache->contexts);
kfree(cache->entry);
kfree(cache);
}
@@ -221,7 +235,7 @@ void squashfs_cache_delete(struct squashfs_cache *cache)
* is allocated as a sequence of kmalloced PAGE_SIZE buffers.
*/
struct squashfs_cache *squashfs_cache_init(char *name, int entries,
- int block_size)
+ int block_size, int alloc_contexts)
{
int i, j;
struct squashfs_cache *cache = kzalloc(sizeof(*cache), GFP_KERNEL);
@@ -237,6 +251,15 @@ struct squashfs_cache *squashfs_cache_init(char *name, int entries,
goto cleanup;
}
+ if (alloc_contexts) {
+ cache->contexts = squashfs_create_contexts(block_size, entries,
+ 0, 0);
+ if (cache->contexts == NULL) {
+ ERROR("Failed to allocate %s cache\n", name);
+ goto cleanup;
+ }
+ }
+
cache->curr_blk = 0;
cache->next_blk = 0;
cache->unused = entries;
@@ -268,13 +291,6 @@ struct squashfs_cache *squashfs_cache_init(char *name, int entries,
goto cleanup;
}
}
-
- entry->actor = squashfs_page_actor_init(entry->data,
- cache->pages, 0);
- if (entry->actor == NULL) {
- ERROR("Failed to allocate %s cache entry\n", name);
- goto cleanup;
- }
}
return cache;
@@ -341,7 +357,8 @@ int squashfs_read_metadata(struct super_block *sb, void *buffer,
return -EIO;
while (length) {
- entry = squashfs_cache_get(sb, msblk->block_cache, *block, 0);
+ entry = squashfs_cache_get(sb, msblk->block_cache, NULL, NULL,
+ *block, 0);
if (entry->error) {
res = entry->error;
goto error;
@@ -377,12 +394,12 @@ int squashfs_read_metadata(struct super_block *sb, void *buffer,
* filesystem. If necessary read and decompress it from disk.
*/
struct squashfs_cache_entry *squashfs_get_fragment(struct super_block *sb,
- u64 start_block, int length)
+ u64 start_block, int length, struct squashfs_context *context)
{
struct squashfs_sb_info *msblk = sb->s_fs_info;
- return squashfs_cache_get(sb, msblk->fragment_cache, start_block,
- length);
+ return squashfs_cache_get(sb, msblk->fragment_cache, msblk->contexts,
+ context, start_block, length);
}
@@ -396,7 +413,8 @@ struct squashfs_cache_entry *squashfs_get_datablock(struct super_block *sb,
{
struct squashfs_sb_info *msblk = sb->s_fs_info;
- return squashfs_cache_get(sb, msblk->read_page, start_block, length);
+ return squashfs_cache_get(sb, msblk->read_page, msblk->contexts, NULL,
+ start_block, length);
}
@@ -408,7 +426,7 @@ void *squashfs_read_table(struct super_block *sb, u64 block, int length)
int pages = (length + PAGE_SIZE - 1) >> PAGE_SHIFT;
int i, res;
void *table, *buffer, **data;
- struct squashfs_page_actor *actor;
+ struct squashfs_context *context;
table = buffer = kmalloc(length, GFP_KERNEL);
if (table == NULL)
@@ -420,20 +438,22 @@ void *squashfs_read_table(struct super_block *sb, u64 block, int length)
goto failed;
}
- actor = squashfs_page_actor_init(data, pages, length);
- if (actor == NULL) {
+ context = squashfs_create_context(length, 0, 0);
+ if (context == NULL) {
res = -ENOMEM;
goto failed2;
}
+ squashfs_page_actor_init(context->actor, data, pages, length);
+
for (i = 0; i < pages; i++, buffer += PAGE_SIZE)
data[i] = buffer;
res = squashfs_read_data(sb, block, length |
- SQUASHFS_COMPRESSED_BIT_BLOCK, NULL, actor);
+ SQUASHFS_COMPRESSED_BIT_BLOCK, NULL, context);
+ squashfs_delete_context(context);
kfree(data);
- kfree(actor);
if (res < 0)
goto failed;
diff --git a/fs/squashfs/decompressor.c b/fs/squashfs/decompressor.c
index 8893cb9b4198..d12255edb129 100644
--- a/fs/squashfs/decompressor.c
+++ b/fs/squashfs/decompressor.c
@@ -15,9 +15,10 @@
#include "squashfs_fs.h"
#include "squashfs_fs_sb.h"
+#include "read_context.h"
+#include "page_actor.h"
#include "decompressor.h"
#include "squashfs.h"
-#include "page_actor.h"
/*
* This file (and decompressor.h) implements a decompressor framework for
@@ -89,7 +90,7 @@ static void *get_comp_opts(struct super_block *sb, unsigned short flags)
{
struct squashfs_sb_info *msblk = sb->s_fs_info;
void *buffer = NULL, *comp_opts;
- struct squashfs_page_actor *actor = NULL;
+ struct squashfs_context *context;
int length = 0;
/*
@@ -102,14 +103,16 @@ static void *get_comp_opts(struct super_block *sb, unsigned short flags)
goto out;
}
- actor = squashfs_page_actor_init(&buffer, 1, 0);
- if (actor == NULL) {
+ context = squashfs_create_context(PAGE_SIZE, 0, 0);
+ if (context == NULL) {
comp_opts = ERR_PTR(-ENOMEM);
goto out;
}
+ squashfs_page_actor_init(context->actor, &buffer, 1, 0);
length = squashfs_read_data(sb,
- sizeof(struct squashfs_super_block), 0, NULL, actor);
+ sizeof(struct squashfs_super_block), 0, NULL, context);
+ squashfs_delete_context(context);
if (length < 0) {
comp_opts = ERR_PTR(length);
@@ -120,7 +123,6 @@ static void *get_comp_opts(struct super_block *sb, unsigned short flags)
comp_opts = squashfs_comp_opts(msblk, buffer, length);
out:
- kfree(actor);
kfree(buffer);
return comp_opts;
}
diff --git a/fs/squashfs/decompressor_multi.c b/fs/squashfs/decompressor_multi.c
index 416c53eedbd1..0a31012abec7 100644
--- a/fs/squashfs/decompressor_multi.c
+++ b/fs/squashfs/decompressor_multi.c
@@ -13,6 +13,8 @@
#include "squashfs_fs.h"
#include "squashfs_fs_sb.h"
+#include "read_context.h"
+#include "page_actor.h"
#include "decompressor.h"
#include "squashfs.h"
diff --git a/fs/squashfs/decompressor_multi_percpu.c b/fs/squashfs/decompressor_multi_percpu.c
index 1dfadf76ed9a..12094aaba19f 100644
--- a/fs/squashfs/decompressor_multi_percpu.c
+++ b/fs/squashfs/decompressor_multi_percpu.c
@@ -12,6 +12,8 @@
#include "squashfs_fs.h"
#include "squashfs_fs_sb.h"
+#include "read_context.h"
+#include "page_actor.h"
#include "decompressor.h"
#include "squashfs.h"
diff --git a/fs/squashfs/decompressor_single.c b/fs/squashfs/decompressor_single.c
index 6f161887710b..8d9b73103165 100644
--- a/fs/squashfs/decompressor_single.c
+++ b/fs/squashfs/decompressor_single.c
@@ -11,6 +11,8 @@
#include "squashfs_fs.h"
#include "squashfs_fs_sb.h"
+#include "read_context.h"
+#include "page_actor.h"
#include "decompressor.h"
#include "squashfs.h"
diff --git a/fs/squashfs/file.c b/fs/squashfs/file.c
index 8ba8c4c50770..d8ca63a3eb0e 100644
--- a/fs/squashfs/file.c
+++ b/fs/squashfs/file.c
@@ -38,8 +38,9 @@
#include "squashfs_fs.h"
#include "squashfs_fs_sb.h"
#include "squashfs_fs_i.h"
-#include "squashfs.h"
+#include "read_context.h"
#include "page_actor.h"
+#include "squashfs.h"
/*
* Locate cache slot in range [offset, index] for specified inode. If
@@ -424,7 +425,7 @@ static int squashfs_readpage_fragment(struct page *page, int expected)
struct inode *inode = page->mapping->host;
struct squashfs_cache_entry *buffer = squashfs_get_fragment(inode->i_sb,
squashfs_i(inode)->fragment_block,
- squashfs_i(inode)->fragment_size);
+ squashfs_i(inode)->fragment_size, NULL);
int res = buffer->error;
if (res)
@@ -497,13 +498,13 @@ static int squashfs_read_folio(struct file *file, struct folio *folio)
return res;
}
-static int squashfs_readahead_fragment(struct page **page,
+static int squashfs_readahead_fragment(struct squashfs_context *context,
unsigned int pages, unsigned int expected)
{
- struct inode *inode = page[0]->mapping->host;
+ struct inode *inode = context->page[0]->mapping->host;
struct squashfs_cache_entry *buffer = squashfs_get_fragment(inode->i_sb,
squashfs_i(inode)->fragment_block,
- squashfs_i(inode)->fragment_size);
+ squashfs_i(inode)->fragment_size, context);
struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
unsigned int n, mask = (1 << (msblk->block_log - PAGE_SHIFT)) - 1;
int error = buffer->error;
@@ -514,18 +515,18 @@ static int squashfs_readahead_fragment(struct page **page,
expected += squashfs_i(inode)->fragment_offset;
for (n = 0; n < pages; n++) {
- unsigned int base = (page[n]->index & mask) << PAGE_SHIFT;
+ unsigned int base = (context->page[n]->index & mask) << PAGE_SHIFT;
unsigned int offset = base + squashfs_i(inode)->fragment_offset;
if (expected > offset) {
unsigned int avail = min_t(unsigned int, expected -
offset, PAGE_SIZE);
- squashfs_fill_page(page[n], buffer, offset, avail);
+ squashfs_fill_page(context->page[n], buffer, offset, avail);
}
- unlock_page(page[n]);
- put_page(page[n]);
+ unlock_page(context->page[n]);
+ put_page(context->page[n]);
}
out:
@@ -541,16 +542,15 @@ static void squashfs_readahead(struct readahead_control *ractl)
unsigned short shift = msblk->block_log - PAGE_SHIFT;
loff_t start = readahead_pos(ractl) & ~mask;
size_t len = readahead_length(ractl) + readahead_pos(ractl) - start;
- struct squashfs_page_actor *actor;
+ struct squashfs_context *context;
unsigned int nr_pages = 0;
- struct page **pages;
int i, file_end = i_size_read(inode) >> msblk->block_log;
unsigned int max_pages = 1UL << shift;
readahead_expand(ractl, start, (len | mask) + 1);
- pages = kmalloc_array(max_pages, sizeof(void *), GFP_KERNEL);
- if (!pages)
+ context = squashfs_get_context(msblk->contexts);
+ if (context == NULL)
return;
for (;;) {
@@ -566,21 +566,21 @@ static void squashfs_readahead(struct readahead_control *ractl)
max_pages = (expected + PAGE_SIZE - 1) >> PAGE_SHIFT;
- nr_pages = __readahead_batch(ractl, pages, max_pages);
+ nr_pages = __readahead_batch(ractl, context->page, max_pages);
if (!nr_pages)
break;
if (readahead_pos(ractl) >= i_size_read(inode))
goto skip_pages;
- index = pages[0]->index >> shift;
+ index = context->page[0]->index >> shift;
- if ((pages[nr_pages - 1]->index >> shift) != index)
+ if ((context->page[nr_pages - 1]->index >> shift) != index)
goto skip_pages;
if (index == file_end && squashfs_i(inode)->fragment_block !=
SQUASHFS_INVALID_BLK) {
- res = squashfs_readahead_fragment(pages, nr_pages,
+ res = squashfs_readahead_fragment(context, nr_pages,
expected);
if (res)
goto skip_pages;
@@ -591,14 +591,13 @@ static void squashfs_readahead(struct readahead_control *ractl)
if (bsize == 0)
goto skip_pages;
- actor = squashfs_page_actor_init_special(msblk, pages, nr_pages,
- expected);
- if (!actor)
- goto skip_pages;
+ squashfs_page_actor_init_special(context, msblk, nr_pages,
+ expected);
- res = squashfs_read_data(inode->i_sb, block, bsize, NULL, actor);
+ res = squashfs_read_data(inode->i_sb, block, bsize, NULL,
+ context);
- last_page = squashfs_page_actor_free(actor);
+ last_page = squashfs_page_actor_free(context->actor);
if (res == expected) {
int bytes;
@@ -610,26 +609,26 @@ static void squashfs_readahead(struct readahead_control *ractl)
PAGE_SIZE - bytes);
for (i = 0; i < nr_pages; i++) {
- flush_dcache_page(pages[i]);
- SetPageUptodate(pages[i]);
+ flush_dcache_page(context->page[i]);
+ SetPageUptodate(context->page[i]);
}
}
for (i = 0; i < nr_pages; i++) {
- unlock_page(pages[i]);
- put_page(pages[i]);
+ unlock_page(context->page[i]);
+ put_page(context->page[i]);
}
}
- kfree(pages);
+ squashfs_put_context(msblk->contexts, context);
return;
skip_pages:
for (i = 0; i < nr_pages; i++) {
- unlock_page(pages[i]);
- put_page(pages[i]);
+ unlock_page(context->page[i]);
+ put_page(context->page[i]);
}
- kfree(pages);
+ squashfs_put_context(msblk->contexts, context);
}
const struct address_space_operations squashfs_aops = {
diff --git a/fs/squashfs/file_direct.c b/fs/squashfs/file_direct.c
index f1ccad519e28..ab7b990cf71d 100644
--- a/fs/squashfs/file_direct.c
+++ b/fs/squashfs/file_direct.c
@@ -15,8 +15,9 @@
#include "squashfs_fs.h"
#include "squashfs_fs_sb.h"
#include "squashfs_fs_i.h"
-#include "squashfs.h"
+#include "read_context.h"
#include "page_actor.h"
+#include "squashfs.h"
/* Read separately compressed datablock directly into page cache */
int squashfs_readpage_block(struct page *target_page, u64 block, int bsize,
@@ -31,8 +32,7 @@ int squashfs_readpage_block(struct page *target_page, u64 block, int bsize,
int start_index = target_page->index & ~mask;
int end_index = start_index | mask;
int i, n, pages, bytes, res = -ENOMEM;
- struct page **page;
- struct squashfs_page_actor *actor;
+ struct squashfs_context *context;
void *pageaddr;
if (end_index > file_end)
@@ -40,21 +40,21 @@ int squashfs_readpage_block(struct page *target_page, u64 block, int bsize,
pages = end_index - start_index + 1;
- page = kmalloc_array(pages, sizeof(void *), GFP_KERNEL);
- if (page == NULL)
+ context = squashfs_get_context(msblk->contexts);
+ if (context == NULL)
return res;
/* Try to grab all the pages covered by the Squashfs block */
for (i = 0, n = start_index; n <= end_index; n++) {
- page[i] = (n == target_page->index) ? target_page :
+ context->page[i] = (n == target_page->index) ? target_page :
grab_cache_page_nowait(target_page->mapping, n);
- if (page[i] == NULL)
+ if (context->page[i] == NULL)
continue;
- if (PageUptodate(page[i])) {
- unlock_page(page[i]);
- put_page(page[i]);
+ if (PageUptodate(context->page[i])) {
+ unlock_page(context->page[i]);
+ put_page(context->page[i]);
continue;
}
@@ -64,17 +64,15 @@ int squashfs_readpage_block(struct page *target_page, u64 block, int bsize,
pages = i;
/*
- * Create a "page actor" which will kmap and kunmap the
+ * Initialise "page actor" which will kmap and kunmap the
* page cache pages appropriately within the decompressor
*/
- actor = squashfs_page_actor_init_special(msblk, page, pages, expected);
- if (actor == NULL)
- goto out;
+ squashfs_page_actor_init_special(context, msblk, pages, expected);
/* Decompress directly into the page cache buffers */
- res = squashfs_read_data(inode->i_sb, block, bsize, NULL, actor);
+ res = squashfs_read_data(inode->i_sb, block, bsize, NULL, context);
- squashfs_page_actor_free(actor);
+ squashfs_page_actor_free(context->actor);
if (res < 0)
goto mark_errored;
@@ -86,22 +84,22 @@ int squashfs_readpage_block(struct page *target_page, u64 block, int bsize,
/* Last page (if present) may have trailing bytes not filled */
bytes = res % PAGE_SIZE;
- if (page[pages - 1]->index == end_index && bytes) {
- pageaddr = kmap_local_page(page[pages - 1]);
+ if (context->page[pages - 1]->index == end_index && bytes) {
+ pageaddr = kmap_local_page(context->page[pages - 1]);
memset(pageaddr + bytes, 0, PAGE_SIZE - bytes);
kunmap_local(pageaddr);
}
/* Mark pages as uptodate, unlock and release */
for (i = 0; i < pages; i++) {
- flush_dcache_page(page[i]);
- SetPageUptodate(page[i]);
- unlock_page(page[i]);
- if (page[i] != target_page)
- put_page(page[i]);
+ flush_dcache_page(context->page[i]);
+ SetPageUptodate(context->page[i]);
+ unlock_page(context->page[i]);
+ if (context->page[i] != target_page)
+ put_page(context->page[i]);
}
- kfree(page);
+ squashfs_put_context(msblk->contexts, context);
return 0;
@@ -110,15 +108,14 @@ int squashfs_readpage_block(struct page *target_page, u64 block, int bsize,
* dealt with by the caller
*/
for (i = 0; i < pages; i++) {
- if (page[i] == NULL || page[i] == target_page)
+ if (context->page[i] == NULL || context->page[i] == target_page)
continue;
- flush_dcache_page(page[i]);
- SetPageError(page[i]);
- unlock_page(page[i]);
- put_page(page[i]);
+ flush_dcache_page(context->page[i]);
+ SetPageError(context->page[i]);
+ unlock_page(context->page[i]);
+ put_page(context->page[i]);
}
-out:
- kfree(page);
+ squashfs_put_context(msblk->contexts, context);
return res;
}
diff --git a/fs/squashfs/page_actor.c b/fs/squashfs/page_actor.c
index 81af6c4ca115..8ff50034afee 100644
--- a/fs/squashfs/page_actor.c
+++ b/fs/squashfs/page_actor.c
@@ -8,8 +8,9 @@
#include <linux/slab.h>
#include <linux/pagemap.h>
#include "squashfs_fs_sb.h"
-#include "decompressor.h"
+#include "read_context.h"
#include "page_actor.h"
+#include "decompressor.h"
/*
* This file contains implementations of page_actor for decompressing into
@@ -40,14 +41,9 @@ static void cache_finish_page(struct squashfs_page_actor *actor)
/* empty */
}
-struct squashfs_page_actor *squashfs_page_actor_init(void **buffer,
+void squashfs_page_actor_init(struct squashfs_page_actor *actor, void **buffer,
int pages, int length)
{
- struct squashfs_page_actor *actor = kmalloc(sizeof(*actor), GFP_KERNEL);
-
- if (actor == NULL)
- return NULL;
-
actor->length = length ? : pages * PAGE_SIZE;
actor->buffer = buffer;
actor->pages = pages;
@@ -56,7 +52,6 @@ struct squashfs_page_actor *squashfs_page_actor_init(void **buffer,
actor->squashfs_first_page = cache_first_page;
actor->squashfs_next_page = cache_next_page;
actor->squashfs_finish_page = cache_finish_page;
- return actor;
}
/* Implementation of page_actor for decompressing directly into page cache. */
@@ -102,35 +97,23 @@ static void direct_finish_page(struct squashfs_page_actor *actor)
kunmap_local(actor->pageaddr);
}
-struct squashfs_page_actor *squashfs_page_actor_init_special(struct squashfs_sb_info *msblk,
- struct page **page, int pages, int length)
+void squashfs_page_actor_init_special(struct squashfs_context *context,
+ struct squashfs_sb_info *msblk, int pages, int length)
{
- struct squashfs_page_actor *actor = kmalloc(sizeof(*actor), GFP_KERNEL);
-
- if (actor == NULL)
- return NULL;
-
- if (msblk->decompressor->alloc_buffer) {
- actor->tmp_buffer = kmalloc(PAGE_SIZE, GFP_KERNEL);
-
- if (actor->tmp_buffer == NULL) {
- kfree(actor);
- return NULL;
- }
- } else
- actor->tmp_buffer = NULL;
+ struct squashfs_page_actor *actor = context->actor;
+ actor->tmp_buffer = context->hole;
actor->length = length ? : pages * PAGE_SIZE;
- actor->page = page;
+ actor->page = context->page;
actor->pages = pages;
actor->next_page = 0;
actor->returned_pages = 0;
- actor->next_index = page[0]->index & ~((1 << (msblk->block_log - PAGE_SHIFT)) - 1);
+ actor->next_index = context->page[0]->index & ~((1 <<
+ (msblk->block_log - PAGE_SHIFT)) - 1);
actor->pageaddr = NULL;
actor->last_page = NULL;
actor->alloc_buffer = msblk->decompressor->alloc_buffer;
actor->squashfs_first_page = direct_first_page;
actor->squashfs_next_page = direct_next_page;
actor->squashfs_finish_page = direct_finish_page;
- return actor;
}
diff --git a/fs/squashfs/page_actor.h b/fs/squashfs/page_actor.h
index 97d4983559b1..36fbc93d8907 100644
--- a/fs/squashfs/page_actor.h
+++ b/fs/squashfs/page_actor.h
@@ -25,17 +25,15 @@ struct squashfs_page_actor {
pgoff_t next_index;
};
-extern struct squashfs_page_actor *squashfs_page_actor_init(void **buffer,
+extern void squashfs_page_actor_init(struct squashfs_page_actor *actor, void **buffer,
int pages, int length);
-extern struct squashfs_page_actor *squashfs_page_actor_init_special(
+extern void squashfs_page_actor_init_special(struct squashfs_context *context,
struct squashfs_sb_info *msblk,
- struct page **page, int pages, int length);
+ int pages, int length);
static inline struct page *squashfs_page_actor_free(struct squashfs_page_actor *actor)
{
struct page *last_page = actor->last_page;
- kfree(actor->tmp_buffer);
- kfree(actor);
return last_page;
}
static inline void *squashfs_first_page(struct squashfs_page_actor *actor)
diff --git a/fs/squashfs/read_context.c b/fs/squashfs/read_context.c
new file mode 100644
index 000000000000..6ad151b034ed
--- /dev/null
+++ b/fs/squashfs/read_context.c
@@ -0,0 +1,178 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Squashfs - a compressed read only filesystem for Linux
+ *
+ * Copyright (c) 2023
+ * Phillip Lougher <phillip@xxxxxxxxxxxxxxx>
+ *
+ * read_context.c
+ */
+
+#include <linux/fs.h>
+#include <linux/vfs.h>
+#include <linux/mutex.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/pagemap.h>
+#include <linux/bio.h>
+
+#include "squashfs_fs.h"
+#include "squashfs_fs_sb.h"
+#include "squashfs.h"
+#include "page_actor.h"
+
+
+void squashfs_delete_context(struct squashfs_context *context)
+{
+ int i;
+
+ for (i = 0; i < context->bio_pages; i++)
+ if (context->bio_page[i])
+ __free_page(context->bio_page[i]);
+
+ kfree(context->bio);
+ kfree(context->page);
+ kfree(context->actor);
+ kfree(context->hole);
+}
+
+
+struct squashfs_context *squashfs_create_context(int length, int alloc_hole,
+ int alloc_array)
+{
+ int pages = (length + PAGE_SIZE - 1) >> PAGE_SHIFT;
+ int bio_pages = pages + 1, i;
+ struct squashfs_context *context;
+
+ context = kzalloc(sizeof(struct squashfs_context), GFP_KERNEL);
+ if (context == NULL)
+ return ERR_PTR(-ENOMEM);
+
+ context->bio_page = kcalloc(bio_pages, sizeof(struct bio_pages *),
+ GFP_KERNEL);
+
+ if (context->bio_page == NULL)
+ goto cleanup;
+
+ context->bio_pages = bio_pages;
+
+ for (i = 0; i < bio_pages; i++) {
+ context->bio_page[i] = alloc_page(GFP_KERNEL);
+
+ if (context->bio_page[i] == NULL)
+ goto cleanup;
+ }
+
+ context->bio = bio_kmalloc(bio_pages, GFP_KERNEL);
+ if (context->bio == NULL)
+ goto cleanup;
+
+ context->actor = kmalloc(sizeof(struct squashfs_page_actor),
+ GFP_KERNEL);
+ if (context->actor == NULL)
+ goto cleanup;
+
+ if (alloc_array) {
+ context->page = kmalloc_array(pages, sizeof(struct page *),
+ GFP_KERNEL);
+ if (context->page == NULL)
+ goto cleanup;
+ }
+
+ if (alloc_hole) {
+ context->hole = kmalloc(PAGE_SIZE, GFP_KERNEL);
+ if (context->hole == NULL)
+ goto cleanup;
+ }
+
+ return context;
+
+cleanup:
+ squashfs_delete_context(context);
+ return ERR_PTR(-ENOMEM);
+}
+
+
+struct squashfs_context *squashfs_get_context(struct squashfs_contexts *contexts)
+{
+ struct squashfs_context *context;
+
+ while (1) {
+ mutex_lock(&contexts->mutex);
+
+ if (!list_empty(&contexts->list)) {
+ context = list_entry(contexts->list.prev,
+ struct squashfs_context, list);
+ list_del(&context->list);
+ mutex_unlock(&contexts->mutex);
+ break;
+ }
+
+ mutex_unlock(&contexts->mutex);
+ wait_event(contexts->wait_queue, !list_empty(&contexts->list));
+ }
+
+ return context;
+}
+
+
+void squashfs_put_context(struct squashfs_contexts *contexts,
+ struct squashfs_context *context)
+{
+
+ mutex_lock(&contexts->mutex);
+ list_add(&context->list, &contexts->list);
+ mutex_unlock(&contexts->mutex);
+ wake_up(&contexts->wait_queue);
+}
+
+
+void squashfs_delete_contexts(struct squashfs_contexts *contexts)
+{
+ struct squashfs_context *context;
+
+ if (contexts == NULL)
+ return;
+
+ while (!list_empty(&contexts->list)) {
+ context = list_entry(contexts->list.prev,
+ struct squashfs_context, list);
+ list_del(&context->list);
+ squashfs_delete_context(context);
+ }
+}
+
+
+struct squashfs_contexts *squashfs_create_contexts(int block_size, int entries,
+ int alloc_hole, int alloc_array)
+{
+ int i;
+ struct squashfs_contexts *contexts;
+
+ contexts = kzalloc(sizeof(struct squashfs_contexts), GFP_KERNEL);
+ if (contexts == NULL)
+ return ERR_PTR(-ENOMEM);
+
+ mutex_init(&contexts->mutex);
+ INIT_LIST_HEAD(&contexts->list);
+ init_waitqueue_head(&contexts->wait_queue);
+
+ for (i = 0; i < entries; i++) {
+ struct squashfs_context *context;
+
+ context = squashfs_create_context(block_size, alloc_hole,
+ alloc_array);
+ if (context == NULL)
+ goto cleanup;
+
+ list_add(&context->list, &contexts->list);
+ }
+
+ return contexts;
+
+cleanup:
+ squashfs_delete_contexts(contexts);
+ return ERR_PTR(-ENOMEM);
+}
diff --git a/fs/squashfs/read_context.h b/fs/squashfs/read_context.h
new file mode 100644
index 000000000000..a9b93a43893b
--- /dev/null
+++ b/fs/squashfs/read_context.h
@@ -0,0 +1,31 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#ifndef READ_CONTEXT_H
+#define READ_CONTEXT_H
+/*
+ * Copyright (c) 2023
+ * Phillip Lougher <phillip@xxxxxxxxxxxxxxx>
+ */
+
+struct squashfs_context {
+ int bio_pages;
+ struct bio *bio;
+ struct page **bio_page;
+ struct squashfs_page_actor *actor;
+ struct page **page;
+ void *hole;
+ struct list_head list;
+};
+
+struct squashfs_contexts {
+ struct mutex mutex;
+ struct list_head list;
+ wait_queue_head_t wait_queue;
+};
+
+extern void squashfs_delete_context(struct squashfs_context *context);
+extern struct squashfs_context *squashfs_create_context(int length, int alloc_hole, int alloc_array);
+extern struct squashfs_context *squashfs_get_context(struct squashfs_contexts *contexts);
+extern void squashfs_put_context(struct squashfs_contexts *contexts, struct squashfs_context *context);
+extern void squashfs_delete_contexts(struct squashfs_contexts *contexts);
+extern struct squashfs_contexts *squashfs_create_contexts(int block_size, int entries, int alloc_hole, int alloc_array);
+#endif
diff --git a/fs/squashfs/squashfs.h b/fs/squashfs/squashfs.h
index a6164fdf9435..c5db13907e9f 100644
--- a/fs/squashfs/squashfs.h
+++ b/fs/squashfs/squashfs.h
@@ -8,6 +8,9 @@
* squashfs.h
*/
+#include "read_context.h"
+#include "page_actor.h"
+
#define TRACE(s, args...) pr_debug("SQUASHFS: "s, ## args)
#define ERROR(s, args...) pr_err("SQUASHFS error: "s, ## args)
@@ -16,19 +19,20 @@
/* block.c */
extern int squashfs_read_data(struct super_block *, u64, int, u64 *,
- struct squashfs_page_actor *);
+ struct squashfs_context *);
/* cache.c */
-extern struct squashfs_cache *squashfs_cache_init(char *, int, int);
+extern struct squashfs_cache *squashfs_cache_init(char *, int, int, int);
extern void squashfs_cache_delete(struct squashfs_cache *);
extern struct squashfs_cache_entry *squashfs_cache_get(struct super_block *,
- struct squashfs_cache *, u64, int);
+ struct squashfs_cache *, struct squashfs_contexts *,
+ struct squashfs_context *, u64, int);
extern void squashfs_cache_put(struct squashfs_cache_entry *);
extern int squashfs_copy_data(void *, struct squashfs_cache_entry *, int, int);
extern int squashfs_read_metadata(struct super_block *, void *, u64 *,
int *, int);
extern struct squashfs_cache_entry *squashfs_get_fragment(struct super_block *,
- u64, int);
+ u64, int, struct squashfs_context *);
extern struct squashfs_cache_entry *squashfs_get_datablock(struct super_block *,
u64, int);
extern void *squashfs_read_table(struct super_block *, u64, int);
diff --git a/fs/squashfs/squashfs_fs_sb.h b/fs/squashfs/squashfs_fs_sb.h
index 72f6f4b37863..6cee243be9bb 100644
--- a/fs/squashfs/squashfs_fs_sb.h
+++ b/fs/squashfs/squashfs_fs_sb.h
@@ -23,6 +23,7 @@ struct squashfs_cache {
int pages;
spinlock_t lock;
wait_queue_head_t wait_queue;
+ struct squashfs_contexts *contexts;
struct squashfs_cache_entry *entry;
};
@@ -37,7 +38,6 @@ struct squashfs_cache_entry {
wait_queue_head_t wait_queue;
struct squashfs_cache *cache;
void **data;
- struct squashfs_page_actor *actor;
};
struct squashfs_sb_info {
@@ -47,6 +47,7 @@ struct squashfs_sb_info {
struct squashfs_cache *block_cache;
struct squashfs_cache *fragment_cache;
struct squashfs_cache *read_page;
+ struct squashfs_contexts *contexts;
int next_meta_index;
__le64 *id_table;
__le64 *fragment_index;
diff --git a/fs/squashfs/super.c b/fs/squashfs/super.c
index e090fae48e68..3ba5c2b5c9e0 100644
--- a/fs/squashfs/super.c
+++ b/fs/squashfs/super.c
@@ -33,6 +33,7 @@
#include "squashfs_fs.h"
#include "squashfs_fs_sb.h"
#include "squashfs_fs_i.h"
+#include "read_context.h"
#include "squashfs.h"
#include "decompressor.h"
#include "xattr.h"
@@ -317,18 +318,23 @@ static int squashfs_fill_super(struct super_block *sb, struct fs_context *fc)
err = -ENOMEM;
msblk->block_cache = squashfs_cache_init("metadata",
- SQUASHFS_CACHED_BLKS, SQUASHFS_METADATA_SIZE);
+ SQUASHFS_CACHED_BLKS, SQUASHFS_METADATA_SIZE, 1);
if (msblk->block_cache == NULL)
goto failed_mount;
/* Allocate read_page block */
msblk->read_page = squashfs_cache_init("data",
- msblk->max_thread_num, msblk->block_size);
+ msblk->max_thread_num, msblk->block_size, 0);
if (msblk->read_page == NULL) {
errorf(fc, "Failed to allocate read_page block");
goto failed_mount;
}
+ msblk->contexts = squashfs_create_contexts(msblk->block_size,
+ msblk->max_thread_num, msblk->decompressor->alloc_buffer, 1);
+ if (msblk->contexts == NULL)
+ goto failed_mount;
+
msblk->stream = squashfs_decompressor_setup(sb, flags);
if (IS_ERR(msblk->stream)) {
err = PTR_ERR(msblk->stream);
@@ -392,7 +398,7 @@ static int squashfs_fill_super(struct super_block *sb, struct fs_context *fc)
goto check_directory_table;
msblk->fragment_cache = squashfs_cache_init("fragment",
- SQUASHFS_CACHED_FRAGMENTS, msblk->block_size);
+ SQUASHFS_CACHED_FRAGMENTS, msblk->block_size, 0);
if (msblk->fragment_cache == NULL) {
err = -ENOMEM;
goto failed_mount;
@@ -454,6 +460,7 @@ static int squashfs_fill_super(struct super_block *sb, struct fs_context *fc)
squashfs_cache_delete(msblk->block_cache);
squashfs_cache_delete(msblk->fragment_cache);
squashfs_cache_delete(msblk->read_page);
+ squashfs_delete_contexts(msblk->contexts);
msblk->thread_ops->destroy(msblk);
kfree(msblk->inode_lookup_table);
kfree(msblk->fragment_index);
diff --git a/fs/squashfs/symlink.c b/fs/squashfs/symlink.c
index 2bf977a52c2c..8471a3a5491b 100644
--- a/fs/squashfs/symlink.c
+++ b/fs/squashfs/symlink.c
@@ -69,7 +69,8 @@ static int squashfs_symlink_read_folio(struct file *file, struct folio *folio)
* blocks, we may need to call squashfs_cache_get multiple times.
*/
for (bytes = 0; bytes < length; offset = 0, bytes += copied) {
- entry = squashfs_cache_get(sb, msblk->block_cache, block, 0);
+ entry = squashfs_cache_get(sb, msblk->block_cache, NULL, NULL,
+ block, 0);
if (entry->error) {
ERROR("Unable to read symlink [%llx:%x]\n",
squashfs_i(inode)->start,
--
2.35.1
Thanks,
Hui.
