Skip to contentSkip to conversation

vfs: block_read_full_page allocates array depending on environment

RFC PATCH In review Masayoshi MIZUMA 2 participants 5 messages Jan 29, 2010 – Feb 2, 2010 VFS

lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 vfs: block_read_full_page allocates array depending on environment

Hi,

In block_read_full_page(), buffer_head pointer array is allocated more than the necessity.
For example, the system's page cache size = 16k and filesystem's block size = 4k, the necessary number of pointer array is 4, but it is allocated as 32.

This patch allocates the buffer_head pointer array depending on the system's page cache size and filesystem's block size.

---
 fs/buffer.c |    5 ++---
 1 files changed, 2 insertions(+), 3 deletions(-)
fs/buffer.c +2 -3
index 6fa5302..acdedb8 100644
@@ -2146,13 +2146,12 @@ int block_read_full_page(struct page *page, get_block_t *get_block)
 {
        struct inode *inode = page->mapping->host;
        sector_t iblock, lblock;
-       struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
-       unsigned int blocksize;
+       unsigned int blocksize = 1 << inode->i_blkbits;
+       struct buffer_head *bh, *head, *arr[PAGE_CACHE_SIZE / blocksize];
        int nr, i;
        int fully_mapped = 1;
 
        BUG_ON(!PageLocked(page));
-       blocksize = 1 << inode->i_blkbits;
        if (!page_has_buffers(page))
                create_empty_buffers(page, blocksize, 0);
        head = page_buffers(page);

--
1.6.2.5

Thanks,
Masayoshi

#1 lore raw

fs/buffer.c int block_read_full_page(struct page *page, get_block_t *get_block) · line 2150

Quoted 8 lines · Masayoshi MIZUMA
fs/buffer.c +2 -2
@@ -2146,13 +2146,12 @@ int block_read_full_page(struct page *page, get_block_t *get_block)
 {
        struct inode *inode = page->mapping->host;
        sector_t iblock, lblock;
-       struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
-       unsigned int blocksize;
+       unsigned int blocksize = 1 << inode->i_blkbits;
+       struct buffer_head *bh, *head, *arr[PAGE_CACHE_SIZE / blocksize];

In practice, we avoid to use the variable-length array on the kernel stack. Although I'm not sure how important it is, there are some reasons to avoid. Also sparse will warn it, IIRC.

Thanks.

Quoted 8 lines
fs/buffer.c +0 -1
        int nr, i;
        int fully_mapped = 1;
 
        BUG_ON(!PageLocked(page));
-       blocksize = 1 << inode->i_blkbits;
        if (!page_has_buffers(page))
                create_empty_buffers(page, blocksize, 0);
        head = page_buffers(page);
Signature · 1 line
OGAWA Hirofumi <[email protected]>
#2 lore raw
Quoted 14 lines · OGAWA Hirofumi

fs/buffer.c int block_read_full_page(struct page *page, get_block_t *get_block) · line 2150

Quoted 8 lines · Masayoshi MIZUMA
fs/buffer.c +2 -2
@@ -2146,13 +2146,12 @@ int block_read_full_page(struct page *page, get_block_t *get_block)
 {
        struct inode *inode = page->mapping->host;
        sector_t iblock, lblock;
-       struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
-       unsigned int blocksize;
+       unsigned int blocksize = 1 << inode->i_blkbits;
+       struct buffer_head *bh, *head, *arr[PAGE_CACHE_SIZE / blocksize];

In practice, we avoid to use the variable-length array on the kernel stack. Although I'm not sure how important it is, there are some reasons to avoid. Also sparse will warn it, IIRC.

Umm... I understand the practice itself, but is there specific reason for the practice ? If there isn't, we don't need to avoid use it, I think.

Thanks,
Masayoshi

#3 lore raw
Quoted 17 lines · Masayoshi MIZUMA
Quoted 14 lines · OGAWA Hirofumi

fs/buffer.c int block_read_full_page(struct page *page, get_block_t *get_block) · line 2150

Quoted 8 lines · Masayoshi MIZUMA
fs/buffer.c +2 -2
@@ -2146,13 +2146,12 @@ int block_read_full_page(struct page *page, get_block_t *get_block)
 {
        struct inode *inode = page->mapping->host;
        sector_t iblock, lblock;
-       struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
-       unsigned int blocksize;
+       unsigned int blocksize = 1 << inode->i_blkbits;
+       struct buffer_head *bh, *head, *arr[PAGE_CACHE_SIZE / blocksize];

In practice, we avoid to use the variable-length array on the kernel stack. Although I'm not sure how important it is, there are some reasons to avoid. Also sparse will warn it, IIRC.

Quoted 2 lines

Umm... I understand the practice itself, but is there specific reason for the practice ? If there isn't, we don't need to avoid use it, I think.

For example, from quick google

	http://lkml.org/lkml/2003/9/2/202

I guess you can find more reasons if needed.

Thanks.

Signature · 1 line
OGAWA Hirofumi <[email protected]>
#4 lore raw
Quoted 19 lines · OGAWA Hirofumi
Quoted 17 lines · Masayoshi MIZUMA
Quoted 14 lines · OGAWA Hirofumi

fs/buffer.c int block_read_full_page(struct page *page, get_block_t *get_block) · line 2150

Quoted 8 lines · Masayoshi MIZUMA
fs/buffer.c +2 -2
@@ -2146,13 +2146,12 @@ int block_read_full_page(struct page *page, get_block_t *get_block)
 {
        struct inode *inode = page->mapping->host;
        sector_t iblock, lblock;
-       struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
-       unsigned int blocksize;
+       unsigned int blocksize = 1 << inode->i_blkbits;
+       struct buffer_head *bh, *head, *arr[PAGE_CACHE_SIZE / blocksize];

In practice, we avoid to use the variable-length array on the kernel stack. Although I'm not sure how important it is, there are some reasons to avoid. Also sparse will warn it, IIRC.

Quoted 8 lines
Quoted 2 lines

Umm... I understand the practice itself, but is there specific reason for the practice ? If there isn't, we don't need to avoid use it, I think.

For example, from quick google

	http://lkml.org/lkml/2003/9/2/202

I guess you can find more reasons if needed.

I understand it. Thanks a lot !

Thanks,
Masayoshi