btrfs: reduce compression workspace buffer space to block size

Currently the compression workspace buffer size is always based on
PAGE_SIZE, but btrfs has support subpage sized block size for some time.

This means for one-shot compression algorithm like lzo, we're wasting
quite some memory if the block size is smaller than page size, as the
LZO only works on one block (thus one-shot).

On 64K page sized systems with 4K block size, it means we only need at
most 8K buffer space for lzo, but in reality we're allocating 64K
buffer.

So to reduce the memory usage, change all workspace buffer to base its
size based on block size other than page size.

Signed-off-by: Qu Wenruo <[email protected]>
Reviewed-by: David Sterba <[email protected]>
Signed-off-by: David Sterba <[email protected]>

Author: adam900710

fs/btrfs/lzo.c

@@ -58,16 +58,22 @@
  * 0x1000   | SegHdr N+1| Data payload N+1 ...                |
  */
 
-#define WORKSPACE_BUF_LENGTH	(lzo1x_worst_compress(PAGE_SIZE))
-#define WORKSPACE_CBUF_LENGTH	(lzo1x_worst_compress(PAGE_SIZE))
-
 struct workspace {
 	void *mem;
 	void *buf;	/* where decompressed data goes */
 	void *cbuf;	/* where compressed data goes */
 	struct list_head list;
 };
 
+static u32 workspace_buf_length(const struct btrfs_fs_info *fs_info)
+{
+	return lzo1x_worst_compress(fs_info->sectorsize);
+}
+static u32 workspace_cbuf_length(const struct btrfs_fs_info *fs_info)
+{
+	return lzo1x_worst_compress(fs_info->sectorsize);
+}
+
 void lzo_free_workspace(struct list_head *ws)
 {
 	struct workspace *workspace = list_entry(ws, struct workspace, list);
@@ -87,8 +93,8 @@ struct list_head *lzo_alloc_workspace(struct btrfs_fs_info *fs_info)
 		return ERR_PTR(-ENOMEM);
 
 	workspace->mem = kvmalloc(LZO1X_MEM_COMPRESS, GFP_KERNEL | __GFP_NOWARN);
-	workspace->buf = kvmalloc(WORKSPACE_BUF_LENGTH, GFP_KERNEL | __GFP_NOWARN);
-	workspace->cbuf = kvmalloc(WORKSPACE_CBUF_LENGTH, GFP_KERNEL | __GFP_NOWARN);
+	workspace->buf = kvmalloc(workspace_buf_length(fs_info), GFP_KERNEL | __GFP_NOWARN);
+	workspace->cbuf = kvmalloc(workspace_cbuf_length(fs_info), GFP_KERNEL | __GFP_NOWARN);
 	if (!workspace->mem || !workspace->buf || !workspace->cbuf)
 		goto fail;
 
@@ -384,7 +390,7 @@ int lzo_decompress_bio(struct list_head *ws, struct compressed_bio *cb)
 		kunmap_local(kaddr);
 		cur_in += LZO_LEN;
 
-		if (unlikely(seg_len > WORKSPACE_CBUF_LENGTH)) {
+		if (unlikely(seg_len > workspace_cbuf_length(fs_info))) {
 			struct btrfs_inode *inode = cb->bbio.inode;
 
 			/*
@@ -444,7 +450,7 @@ int lzo_decompress(struct list_head *ws, const u8 *data_in,
 	const u32 sectorsize = fs_info->sectorsize;
 	size_t in_len;
 	size_t out_len;
-	size_t max_segment_len = WORKSPACE_BUF_LENGTH;
+	size_t max_segment_len = workspace_buf_length(fs_info);
 	int ret = 0;
 
 	if (srclen < LZO_LEN || srclen > max_segment_len + LZO_LEN * 2)

fs/btrfs/zlib.c

@@ -55,6 +55,7 @@ void zlib_free_workspace(struct list_head *ws)
 
 struct list_head *zlib_alloc_workspace(struct btrfs_fs_info *fs_info, unsigned int level)
 {
+	const u32 blocksize = fs_info->sectorsize;
 	struct workspace *workspace;
 	int workspacesize;
 
@@ -78,8 +79,8 @@ struct list_head *zlib_alloc_workspace(struct btrfs_fs_info *fs_info, unsigned i
 		workspace->buf_size = ZLIB_DFLTCC_BUF_SIZE;
 	}
 	if (!workspace->buf) {
-		workspace->buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
-		workspace->buf_size = PAGE_SIZE;
+		workspace->buf = kmalloc(blocksize, GFP_KERNEL);
+		workspace->buf_size = blocksize;
 	}
 	if (!workspace->strm.workspace || !workspace->buf)
 		goto fail;

fs/btrfs/zstd.c

@@ -370,6 +370,7 @@ void zstd_free_workspace(struct list_head *ws)
 
 struct list_head *zstd_alloc_workspace(struct btrfs_fs_info *fs_info, int level)
 {
+	const u32 blocksize = fs_info->sectorsize;
 	struct workspace *workspace;
 
 	workspace = kzalloc(sizeof(*workspace), GFP_KERNEL);
@@ -382,7 +383,7 @@ struct list_head *zstd_alloc_workspace(struct btrfs_fs_info *fs_info, int level)
 	workspace->req_level = level;
 	workspace->last_used = jiffies;
 	workspace->mem = kvmalloc(workspace->size, GFP_KERNEL | __GFP_NOWARN);
-	workspace->buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
+	workspace->buf = kmalloc(blocksize, GFP_KERNEL);
 	if (!workspace->mem || !workspace->buf)
 		goto fail;
 
@@ -590,6 +591,7 @@ int zstd_decompress_bio(struct list_head *ws, struct compressed_bio *cb)
 	size_t srclen = cb->compressed_len;
 	zstd_dstream *stream;
 	int ret = 0;
+	const u32 blocksize = cb_to_fs_info(cb)->sectorsize;
 	unsigned long folio_in_index = 0;
 	unsigned long total_folios_in = DIV_ROUND_UP(srclen, PAGE_SIZE);
 	unsigned long buf_start;
@@ -613,7 +615,7 @@ int zstd_decompress_bio(struct list_head *ws, struct compressed_bio *cb)
 
 	workspace->out_buf.dst = workspace->buf;
 	workspace->out_buf.pos = 0;
-	workspace->out_buf.size = PAGE_SIZE;
+	workspace->out_buf.size = blocksize;
 
 	while (1) {
 		size_t ret2;