linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Stephen Hemminger <shemminger@vyatta.com>
To: Phillip Lougher <phillip@lougher.demon.co.uk>,
	kirk w <kirkpuppy@yahoo.com>
Cc: linux-fsdevel@vger.kernel.org, squashfs-devel@lists.sourceforge.net
Subject: [RFC 1/4] squashfs: use percpu for zlib decompression
Date: Fri, 22 Apr 2011 14:17:25 -0700	[thread overview]
Message-ID: <20110422212037.646381436@vyatta.com> (raw)
In-Reply-To: 20110422211724.400984699@vyatta.com

[-- Attachment #1: squashfs-gzip-percpu.patch --]
[-- Type: text/plain, Size: 3520 bytes --]

Make zlib decompression multi-threaded.

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


--- a/fs/squashfs/zlib_wrapper.c	2011-04-22 10:57:09.450024619 -0700
+++ b/fs/squashfs/zlib_wrapper.c	2011-04-22 13:15:52.509266395 -0700
@@ -26,6 +26,7 @@
 #include <linux/buffer_head.h>
 #include <linux/slab.h>
 #include <linux/zlib.h>
+#include <linux/percpu.h>
 #include <linux/vmalloc.h>
 
 #include "squashfs_fs.h"
@@ -33,31 +34,53 @@
 #include "squashfs.h"
 #include "decompressor.h"
 
+/* don't use per superblock stream anymore. */
 static void *zlib_init(struct squashfs_sb_info *dummy, void *buff, int len)
 {
-	z_stream *stream = kmalloc(sizeof(z_stream), GFP_KERNEL);
-	if (stream == NULL)
-		goto failed;
-	stream->workspace = vmalloc(zlib_inflate_workspacesize());
-	if (stream->workspace == NULL)
-		goto failed;
+	z_stream __percpu *percpu;
+	z_stream *stream;
+	int cpu, cpu0;
+
+	percpu = alloc_percpu(z_stream);
+	if (!percpu) {
+		ERROR("Failed to allocate per cpu stream\n");
+		return ERR_PTR(-ENOMEM);
+	}
+
+	for_each_possible_cpu(cpu) {
+		stream = per_cpu_ptr(percpu, cpu);
+
+		stream->workspace = vmalloc(zlib_inflate_workspacesize());
+		if (stream->workspace == NULL)
+			goto failed;
+	}
 
-	return stream;
+	return (__force void *) percpu;
 
 failed:
-	ERROR("Failed to allocate zlib workspace\n");
-	kfree(stream);
+	for_each_possible_cpu(cpu0) {
+		if (cpu0 == cpu)
+			break;
+		stream = per_cpu_ptr(percpu, cpu);
+		vfree(stream->workspace);
+	}
+	free_percpu(percpu);
+
+	ERROR("Failed to allocate zlib workspaces\n");
 	return ERR_PTR(-ENOMEM);
 }
 
-
 static void zlib_free(void *strm)
 {
-	z_stream *stream = strm;
+	z_stream __percpu *percpu = (z_stream __percpu *) strm;
+	int cpu;
 
-	if (stream)
+	for_each_possible_cpu(cpu) {
+		z_stream *stream = per_cpu_ptr(percpu, cpu);
 		vfree(stream->workspace);
-	kfree(stream);
+
+	}
+	free_percpu(percpu);
 }
 
 
@@ -67,9 +90,8 @@ static int zlib_uncompress(struct squash
 {
 	int zlib_err, zlib_init = 0;
 	int k = 0, page = 0;
-	z_stream *stream = msblk->stream;
-
-	mutex_lock(&msblk->read_data_mutex);
+	z_stream __percpu *percpu = (z_stream __percpu *)msblk->stream;
+	z_stream *stream = get_cpu_ptr(percpu);
 
 	stream->avail_out = 0;
 	stream->avail_in = 0;
@@ -80,7 +102,7 @@ static int zlib_uncompress(struct squash
 			length -= avail;
 			wait_on_buffer(bh[k]);
 			if (!buffer_uptodate(bh[k]))
-				goto release_mutex;
+				goto put_per_cpu;
 
 			stream->next_in = bh[k]->b_data + offset;
 			stream->avail_in = avail;
@@ -98,7 +120,7 @@ static int zlib_uncompress(struct squash
 				ERROR("zlib_inflateInit returned unexpected "
 					"result 0x%x, srclength %d\n",
 					zlib_err, srclength);
-				goto release_mutex;
+				goto put_per_cpu;
 			}
 			zlib_init = 1;
 		}
@@ -111,26 +133,26 @@ static int zlib_uncompress(struct squash
 
 	if (zlib_err != Z_STREAM_END) {
 		ERROR("zlib_inflate error, data probably corrupt\n");
-		goto release_mutex;
+		goto put_per_cpu;
 	}
 
 	zlib_err = zlib_inflateEnd(stream);
 	if (zlib_err != Z_OK) {
 		ERROR("zlib_inflate error, data probably corrupt\n");
-		goto release_mutex;
+		goto put_per_cpu;
 	}
 
 	if (k < b) {
 		ERROR("zlib_uncompress error, data remaining\n");
-		goto release_mutex;
+		goto put_per_cpu;
 	}
 
 	length = stream->total_out;
-	mutex_unlock(&msblk->read_data_mutex);
+	put_cpu_ptr(stream);
 	return length;
 
-release_mutex:
-	mutex_unlock(&msblk->read_data_mutex);
+put_per_cpu:
+	put_cpu_ptr(stream);
 
 	for (; k < b; k++)
 		put_bh(bh[k]);



  reply	other threads:[~2011-04-22 22:06 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-04-22 21:17 [RFC 0/4] Squashfs decompresssion per-cpu Stephen Hemminger
2011-04-22 21:17 ` Stephen Hemminger [this message]
2011-04-22 21:17 ` [RFC 2/4] squashfs: use percpu for lzo decompression Stephen Hemminger
2011-04-22 21:17 ` [RFC 3/4] squashfs: use percpu for xz decompression Stephen Hemminger
2011-04-22 21:17 ` [RFC 4/4] squashfs: eliminate read_data_mutex Stephen Hemminger

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20110422212037.646381436@vyatta.com \
    --to=shemminger@vyatta.com \
    --cc=kirkpuppy@yahoo.com \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=phillip@lougher.demon.co.uk \
    --cc=squashfs-devel@lists.sourceforge.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).