Skip to contentSkip to conversation

cifs: use add_to_page_cache_lru

PATCH In review Nick Piggin 2 participants 4 messages Mar 17, 2010 – Apr 21, 2010 VFS

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 cifs: use add_to_page_cache_lru

cifs: use add_to_page_cache_lru

add_to_page_cache_lru is exported, so it should be used. Benefits over using a private pagevec: neater code, 128 bytes fewer stack used, percpu lru ordering is preserved, and finally don't need to flush pagevec before returning so batching may be shared with other LRU insertions.

---
 fs/cifs/file.c |   13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
fs/cifs/file.c +3 -10
Index: linux-2.6/fs/cifs/file.c
===================================================================
@@ -1907,8 +1907,7 @@ int cifs_file_mmap(struct file *file, st
 
 
 static void cifs_copy_cache_pages(struct address_space *mapping,
-	struct list_head *pages, int bytes_read, char *data,
-	struct pagevec *plru_pvec)
+	struct list_head *pages, int bytes_read, char *data)
 {
 	struct page *page;
 	char *target;
@@ -1920,7 +1919,7 @@ static void cifs_copy_cache_pages(struct
 		page = list_entry(pages->prev, struct page, lru);
 		list_del(&page->lru);
 
-		if (add_to_page_cache(page, mapping, page->index,
+		if (add_to_page_cache_lru(page, mapping, page->index,
 				      GFP_KERNEL)) {
 			page_cache_release(page);
 			cFYI(1, ("Add page cache failed"));
@@ -1946,8 +1945,6 @@ static void cifs_copy_cache_pages(struct
 		flush_dcache_page(page);
 		SetPageUptodate(page);
 		unlock_page(page);
-		if (!pagevec_add(plru_pvec, page))
-			__pagevec_lru_add_file(plru_pvec);
 		data += PAGE_CACHE_SIZE;
 	}
 	return;
@@ -1966,7 +1963,6 @@ static int cifs_readpages(struct file *f
 	unsigned int read_size, i;
 	char *smb_read_data = NULL;
 	struct smb_com_read_rsp *pSMBr;
-	struct pagevec lru_pvec;
 	struct cifsFileInfo *open_file;
 	int buf_type = CIFS_NO_BUFFER;
 
@@ -1980,7 +1976,6 @@ static int cifs_readpages(struct file *f
 	cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
 	pTcon = cifs_sb->tcon;
 
-	pagevec_init(&lru_pvec, 0);
 	cFYI(DBG2, ("rpages: num pages %d", num_pages));
 	for (i = 0; i < num_pages; ) {
 		unsigned contig_pages;
@@ -2049,7 +2044,7 @@ static int cifs_readpages(struct file *f
 			pSMBr = (struct smb_com_read_rsp *)smb_read_data;
 			cifs_copy_cache_pages(mapping, page_list, bytes_read,
 				smb_read_data + 4 /* RFC1001 hdr */ +
-				le16_to_cpu(pSMBr->DataOffset), &lru_pvec);
+				le16_to_cpu(pSMBr->DataOffset));
 
 			i +=  bytes_read >> PAGE_CACHE_SHIFT;
 			cifs_stats_bytes_read(pTcon, bytes_read);
@@ -2082,8 +2077,6 @@ static int cifs_readpages(struct file *f
 		bytes_read = 0;
 	}
 
-	pagevec_lru_add_file(&lru_pvec);
-
 /* need to free smb_read_data buf before exit */
 	if (smb_read_data) {
 		if (buf_type == CIFS_SMALL_BUFFER)
#1 lore raw

Seems reasonable to me (also Shaggy acked this). Any idea how likely it would affect performance (and would thus need to run through some simple iozone and cthon (timing mode) or similar perf testing)

Quoted 79 lines · Nick Piggin

cifs: use add_to_page_cache_lru

add_to_page_cache_lru is exported, so it should be used. Benefits over using a private pagevec: neater code, 128 bytes fewer stack used, percpu lru ordering is preserved, and finally don't need to flush pagevec before returning so batching may be shared with other LRU insertions.

---
 fs/cifs/file.c |   13 +++----------

 1 file changed, 3 insertions(+), 10 deletions(-)

Index: linux-2.6/fs/cifs/file.c
===================================================================

fs/cifs/file.c +2 -7
@@ -1907,8 +1907,7 @@ int cifs_file_mmap(struct file *file, st
 
 
  static void cifs_copy_cache_pages(struct address_space *mapping,
-       struct list_head *pages, int bytes_read, char *data,
-       struct pagevec *plru_pvec)
+       struct list_head *pages, int bytes_read, char *data)
  {
        struct page *page;
        char *target;
@@ -1920,7 +1919,7 @@ static void cifs_copy_cache_pages(struct
                page = list_entry(pages->prev, struct page, lru);
                list_del(&page->lru);
 
-               if (add_to_page_cache(page, mapping, page->index,
+               if (add_to_page_cache_lru(page, mapping, page->index,
                                      GFP_KERNEL)) {
                        page_cache_release(page);
                        cFYI(1, ("Add page cache failed"));
@@ -1946,8 +1945,6 @@ static void cifs_copy_cache_pages(struct
                flush_dcache_page(page);
                SetPageUptodate(page);
                unlock_page(page);
-               if (!pagevec_add(plru_pvec, page))
-                       __pagevec_lru_add_file(plru_pvec);
                data += PAGE_CACHE_SIZE;
        }
        return;
@@ -1966,7 +1963,6 @@ static int cifs_readpages(struct file *f
        unsigned int read_size, i;
        char *smb_read_data = NULL;
        struct smb_com_read_rsp *pSMBr;
-       struct pagevec lru_pvec;
        struct cifsFileInfo *open_file;
        int buf_type = CIFS_NO_BUFFER;
 
@@ -1980,7 +1976,6 @@ static int cifs_readpages(struct file *f
        cifs_sb = CIFS_SB(file->f_path.dentry->d_sb);
        pTcon = cifs_sb->tcon;
 
-       pagevec_init(&lru_pvec, 0);
        cFYI(DBG2, ("rpages: num pages %d", num_pages));
        for (i = 0; i < num_pages; ) {
                unsigned contig_pages;
@@ -2049,7 +2044,7 @@ static int cifs_readpages(struct file *f
                        pSMBr = (struct smb_com_read_rsp *)smb_read_data;
                        cifs_copy_cache_pages(mapping, page_list, bytes_read,

                               smb_read_data + 4 /* RFC1001 hdr */ +
-                               le16_to_cpu(pSMBr->DataOffset), &lru_pvec);
+                               le16_to_cpu(pSMBr->DataOffset));

                       i +=  bytes_read >> PAGE_CACHE_SHIFT; cifs_stats_bytes_read(pTcon, bytes_read);

fs/cifs/file.c +0 -2
@@ -2082,8 +2077,6 @@ static int cifs_readpages(struct file *f
                bytes_read = 0;
        }
 
-       pagevec_lru_add_file(&lru_pvec);
-
  /* need to free smb_read_data buf before exit */
        if (smb_read_data) {
                if (buf_type == CIFS_SMALL_BUFFER)
Signature · 7 lines
Thanks,

Steve
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#2 lore raw
Quoted 3 lines · Steve French

Seems reasonable to me (also Shaggy acked this). Any idea how likely it would affect performance (and would thus need to run through some simple iozone and cthon (timing mode) or similar perf testing)

It should positively affect LRU ordering, and CPU efficiency, but both such a tiny amount it should be in the noise. It wouldn't affect IO patterns (except as a side effect of slightly different pagecache reclaim, but again in the noise).

I wouldn't worry about benchmarking it.

#3 lore raw

queued up for when the merge window opens ...

thx

Quoted 13 lines · Nick Piggin
Quoted 3 lines · Steve French

Seems reasonable to me (also Shaggy acked this).  Any idea how likely it would affect performance (and would thus need to run through some simple iozone and cthon (timing mode) or similar perf testing)

It should positively affect LRU ordering, and CPU efficiency, but both such a tiny amount it should be in the noise. It wouldn't affect IO patterns (except as a side effect of slightly different pagecache reclaim, but again in the noise).

I wouldn't worry about benchmarking it.

Signature · 3 lines
Thanks,

Steve