[RFC 2/4] squashfs: use percpu for lzo decompression

[Stephen Hemminger <shemminger@vyatta.com>]

[-- Attachment #1: squashfs-lz-percpu.patch --]
[-- Type: text/plain, Size: 3047 bytes --]

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>


--- a/fs/squashfs/lzo_wrapper.c	2011-04-22 13:15:28.349003198 -0700
+++ b/fs/squashfs/lzo_wrapper.c	2011-04-22 13:21:03.320648456 -0700
@@ -22,6 +22,7 @@
  */
 
 #include <linux/mutex.h>
+#include <linux/percpu.h>
 #include <linux/buffer_head.h>
 #include <linux/slab.h>
 #include <linux/vmalloc.h>
@@ -40,51 +41,71 @@ struct squashfs_lzo {
 static void *lzo_init(struct squashfs_sb_info *msblk, void *buff, int len)
 {
 	int block_size = max_t(int, msblk->block_size, SQUASHFS_METADATA_SIZE);
+	struct squashfs_lzo __percpu *percpu;
+	struct squashfs_lzo *stream;
+	int cpu, cpu0;
+
+	percpu = alloc_percpu(struct squashfs_lzo);
+	if (!percpu) {
+		ERROR("Failed to allocate per cpu stream\n");
+		return ERR_PTR(-ENOMEM);
+	}
 
-	struct squashfs_lzo *stream = kzalloc(sizeof(*stream), GFP_KERNEL);
-	if (stream == NULL)
-		goto failed;
-	stream->input = vmalloc(block_size);
-	if (stream->input == NULL)
-		goto failed;
-	stream->output = vmalloc(block_size);
-	if (stream->output == NULL)
-		goto failed2;
+	for_each_possible_cpu(cpu) {
+		stream = per_cpu_ptr(percpu, cpu);
+
+		stream->input = vmalloc(block_size);
+		if (stream->input == NULL)
+			goto failed;
+
+		stream->output = vmalloc(block_size);
+		if (stream->output == NULL)
+			goto failed2;
+	}
 
-	return stream;
+	return (__force void *) percpu;
 
 failed2:
 	vfree(stream->input);
 failed:
+	for_each_possible_cpu(cpu0) {
+		if (cpu0 == cpu)
+			break;
+		stream = per_cpu_ptr(percpu, cpu);
+		vfree(stream->output);
+		vfree(stream->input);
+	}
 	ERROR("Failed to allocate lzo workspace\n");
-	kfree(stream);
+	free_percpu(percpu);
 	return ERR_PTR(-ENOMEM);
 }
 
 
 static void lzo_free(void *strm)
 {
-	struct squashfs_lzo *stream = strm;
+	struct squashfs_lzo __percpu *percpu
+		= (struct squashfs_lzo __percpu *) strm;
+	int cpu;
 
-	if (stream) {
+	for_each_possible_cpu(cpu) {
+		struct squashfs_lzo *stream = per_cpu_ptr(percpu, cpu);
 		vfree(stream->input);
 		vfree(stream->output);
 	}
-	kfree(stream);
+	free_percpu(percpu);
 }
 
-
 static int lzo_uncompress(struct squashfs_sb_info *msblk, void **buffer,
 	struct buffer_head **bh, int b, int offset, int length, int srclength,
 	int pages)
 {
-	struct squashfs_lzo *stream = msblk->stream;
+	struct squashfs_lzo __percpu *percpu
+		= (struct squashfs_lzo __percpu *)msblk->stream;
+	struct squashfs_lzo *stream = get_cpu_ptr(percpu);
 	void *buff = stream->input;
 	int avail, i, bytes = length, res;
 	size_t out_len = srclength;
 
-	mutex_lock(&msblk->read_data_mutex);
-
 	for (i = 0; i < b; i++) {
 		wait_on_buffer(bh[i]);
 		if (!buffer_uptodate(bh[i]))
@@ -111,7 +132,7 @@ static int lzo_uncompress(struct squashf
 		bytes -= avail;
 	}
 
-	mutex_unlock(&msblk->read_data_mutex);
+	put_cpu_ptr(stream);
 	return res;
 
 block_release:
@@ -119,7 +140,7 @@ block_release:
 		put_bh(bh[i]);
 
 failed:
-	mutex_unlock(&msblk->read_data_mutex);
+	put_cpu_ptr(stream);
 
 	ERROR("lzo decompression failed, data probably corrupt\n");
 	return -EIO;