linux-erofs.lists.ozlabs.org archive mirror
 help / color / mirror / Atom feed
From: gaoxiang25@huawei.com (Gao Xiang)
Subject: [PATCH v6 21/24] erofs: introduce LZ4 decompression inplace
Date: Fri, 2 Aug 2019 20:53:44 +0800	[thread overview]
Message-ID: <20190802125347.166018-22-gaoxiang25@huawei.com> (raw)
In-Reply-To: <20190802125347.166018-1-gaoxiang25@huawei.com>

compressed data will be usually loaded into last pages of
the extent (the last page for 4k) for in-place decompression
(more specifically, in-place IO), as ilustration below,

         start of compressed logical extent
           |                          end of this logical extent
           |                           |
     ______v___________________________v________
... |  page 6  |  page 7  |  page 8  |  page 9  | ...
    |__________|__________|__________|__________|
           .                         ^ .        ^
           .                         |compressed|
           .                         |   data   |
           .                           .        .
           |<          dstsize        >|<margin>|
                                       oend     iend
           op                        ip

Therefore, it's possible to do decompression inplace (thus no
memcpy at all) if the margin is sufficient and safe enough [1],
and it can be implemented only for fixed-size output compression
compared with fixed-size input compression.

No memcpy for most of in-place IO (about 99% of enwik9) after
decompression inplace is implemented and sequential read will
be improved of course (see the following patches for test results).

[1] https://github.com/lz4/lz4/commit/b17f578a919b7e6b078cede2d52be29dd48c8e8c
    https://github.com/lz4/lz4/commit/5997e139f53169fa3a1c1b4418d2452a90b01602

Signed-off-by: Gao Xiang <gaoxiang25 at huawei.com>
---
 fs/erofs/decompressor.c | 36 ++++++++++++++++++++++++++++++++----
 fs/erofs/erofs_fs.h     |  2 +-
 2 files changed, 33 insertions(+), 5 deletions(-)

diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index 244afb29907d..1e9e1359e40c 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -15,6 +15,9 @@
 #endif
 
 #define LZ4_MAX_DISTANCE_PAGES	(DIV_ROUND_UP(LZ4_DISTANCE_MAX, PAGE_SIZE) + 1)
+#ifndef LZ4_DECOMPRESS_INPLACE_MARGIN
+#define LZ4_DECOMPRESS_INPLACE_MARGIN(srcsize)  (((srcsize) >> 8) + 32)
+#endif
 
 struct z_erofs_decompressor {
 	/*
@@ -117,7 +120,7 @@ static int lz4_decompress(struct z_erofs_decompress_req *rq, u8 *out)
 {
 	unsigned int inputmargin, inlen;
 	u8 *src;
-	bool copied;
+	bool copied, support_0padding;
 	int ret;
 
 	if (rq->inputsize > PAGE_SIZE)
@@ -125,13 +128,38 @@ static int lz4_decompress(struct z_erofs_decompress_req *rq, u8 *out)
 
 	src = kmap_atomic(*rq->in);
 	inputmargin = 0;
+	support_0padding = false;
+
+	/* decompression inplace is only safe when 0padding is enabled */
+	if (EROFS_SB(rq->sb)->requirements & EROFS_REQUIREMENT_LZ4_0PADDING) {
+		support_0padding = true;
+
+		while (!src[inputmargin & ~PAGE_MASK])
+			if (!(++inputmargin & ~PAGE_MASK))
+				break;
+
+		if (inputmargin >= rq->inputsize) {
+			kunmap_atomic(src);
+			return -EIO;
+		}
+	}
 
 	copied = false;
 	inlen = rq->inputsize - inputmargin;
 	if (rq->inplace_io) {
-		src = generic_copy_inplace_data(rq, src, inputmargin);
-		inputmargin = 0;
-		copied = true;
+		const uint oend = (rq->pageofs_out +
+				   rq->outputsize) & ~PAGE_MASK;
+		const uint nr = PAGE_ALIGN(rq->pageofs_out +
+					   rq->outputsize) >> PAGE_SHIFT;
+
+		if (rq->partial_decoding || !support_0padding ||
+		    rq->out[nr - 1] != rq->in[0] ||
+		    rq->inputsize - oend <
+		      LZ4_DECOMPRESS_INPLACE_MARGIN(inlen)) {
+			src = generic_copy_inplace_data(rq, src, inputmargin);
+			inputmargin = 0;
+			copied = true;
+		}
 	}
 
 	ret = LZ4_decompress_safe_partial(src + inputmargin, out,
diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index b307060dd220..25bda459f2e6 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -17,7 +17,7 @@
  * incompatible with this kernel version.
  */
 #define EROFS_REQUIREMENT_LZ4_0PADDING	0x00000001
-#define EROFS_ALL_REQUIREMENTS		0
+#define EROFS_ALL_REQUIREMENTS		EROFS_REQUIREMENT_LZ4_0PADDING
 
 struct erofs_super_block {
 /*  0 */__le32 magic;           /* in the little endian */
-- 
2.17.1

  parent reply	other threads:[~2019-08-02 12:53 UTC|newest]

Thread overview: 159+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-08-02 12:53 [PATCH v6 00/24] erofs: promote erofs from staging Gao Xiang
2019-08-02 12:53 ` [PATCH v6 01/24] erofs: add on-disk layout Gao Xiang
2019-08-29  9:59   ` Christoph Hellwig
2019-08-29 10:32     ` Gao Xiang
2019-08-29 10:36       ` Christoph Hellwig
2019-08-29 10:58         ` Gao Xiang
2019-08-29 15:58       ` Joe Perches
2019-08-29 17:26         ` Gao Xiang
2019-08-30 12:07         ` David Sterba
2019-08-30 12:18           ` Gao Xiang via Linux-erofs
2019-09-02  8:43           ` Pavel Machek
2019-09-02 14:07             ` David Sterba
2019-09-03 11:27               ` Pavel Machek
2019-08-29 15:41     ` Gao Xiang
2019-09-01  7:54     ` Gao Xiang via Linux-erofs
2019-09-02 12:45       ` Christoph Hellwig
2019-09-02 13:02         ` Gao Xiang
2019-09-02  8:40     ` Pavel Machek
2019-09-02 10:35       ` Gao Xiang
2019-08-02 12:53 ` [PATCH v6 02/24] erofs: add erofs in-memory stuffs Gao Xiang
2019-08-02 12:53 ` [PATCH v6 03/24] erofs: add super block operations Gao Xiang
2019-08-29 10:15   ` Christoph Hellwig
2019-08-29 10:50     ` Gao Xiang
2019-08-30 16:39       ` Christoph Hellwig
2019-08-30 17:15         ` Gao Xiang
2019-08-31  0:54           ` Gao Xiang
2019-08-31  6:34           ` Amir Goldstein
2019-08-31  6:48             ` Gao Xiang
2019-09-01  8:54     ` Gao Xiang via Linux-erofs
2019-09-02 12:51       ` Christoph Hellwig
2019-09-02 14:43         ` Gao Xiang
2019-09-02 15:19           ` Christoph Hellwig
2019-09-02 15:24             ` Gao Xiang
2019-08-02 12:53 ` [PATCH v6 04/24] erofs: add raw address_space operations Gao Xiang
2019-08-29 10:17   ` Christoph Hellwig
2019-08-29 11:46     ` Gao Xiang
2019-08-30 16:40       ` Christoph Hellwig
2019-08-30 17:23         ` Gao Xiang
2019-08-02 12:53 ` [PATCH v6 05/24] erofs: add inode operations Gao Xiang
2019-08-29 10:24   ` Christoph Hellwig
2019-08-29 11:59     ` Gao Xiang
2019-08-30 16:42       ` Christoph Hellwig
2019-08-30 18:46         ` Gao Xiang
2019-09-01  9:34     ` Gao Xiang via Linux-erofs
2019-09-02 12:53       ` Christoph Hellwig
2019-09-02 13:43       ` David Sterba
2019-09-02 13:55         ` Gao Xiang
2019-08-02 12:53 ` [PATCH v6 06/24] erofs: support special inode Gao Xiang
2019-08-29 10:25   ` Christoph Hellwig
2019-09-01  9:39     ` Gao Xiang via Linux-erofs
2019-08-02 12:53 ` [PATCH v6 07/24] erofs: add directory operations Gao Xiang
2019-08-02 12:53 ` [PATCH v6 08/24] erofs: add namei functions Gao Xiang
2019-08-29 10:28   ` Christoph Hellwig
2019-08-29 11:28     ` Gao Xiang
2019-08-02 12:53 ` [PATCH v6 09/24] erofs: support tracepoint Gao Xiang
2019-08-02 12:53 ` [PATCH v6 10/24] erofs: update Kconfig and Makefile Gao Xiang
2019-08-02 12:53 ` [PATCH v6 11/24] erofs: introduce xattr & posixacl support Gao Xiang
2019-08-02 12:53 ` [PATCH v6 12/24] erofs: introduce tagged pointer Gao Xiang
2019-08-02 12:53 ` [PATCH v6 13/24] erofs: add compression indexes support Gao Xiang
2019-08-02 12:53 ` [PATCH v6 14/24] erofs: introduce superblock registration Gao Xiang
2019-08-02 12:53 ` [PATCH v6 15/24] erofs: introduce erofs shrinker Gao Xiang
2019-08-02 12:53 ` [PATCH v6 16/24] erofs: introduce workstation for decompression Gao Xiang
2019-08-02 12:53 ` [PATCH v6 17/24] erofs: introduce per-CPU buffers implementation Gao Xiang
2019-08-02 12:53 ` [PATCH v6 18/24] erofs: introduce pagevec for decompression subsystem Gao Xiang
2019-08-02 12:53 ` [PATCH v6 19/24] erofs: add erofs_allocpage() Gao Xiang
2019-08-02 12:53 ` [PATCH v6 20/24] erofs: introduce generic decompression backend Gao Xiang
2019-08-02 12:53 ` Gao Xiang [this message]
2019-08-02 12:53 ` [PATCH v6 22/24] erofs: introduce the decompression frontend Gao Xiang
2019-08-02 12:53 ` [PATCH v6 23/24] erofs: introduce cached decompression Gao Xiang
2019-08-02 12:53 ` [PATCH v6 24/24] erofs: add document Gao Xiang
2019-09-01  5:51 ` [PATCH 00/21] erofs: patchset addressing Christoph's comments Gao Xiang via Linux-erofs
2019-09-01  5:51   ` [PATCH 01/21] erofs: remove all the byte offset comments Gao Xiang via Linux-erofs
2019-09-02 12:05     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 02/21] erofs: on-disk format should have explicitly assigned numbers Gao Xiang via Linux-erofs
2019-09-02 12:05     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 03/21] erofs: some macros are much more readable as a function Gao Xiang via Linux-erofs
2019-09-02 12:06     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 04/21] erofs: kill __packed for on-disk structures Gao Xiang via Linux-erofs
2019-09-02 12:06     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 05/21] erofs: update erofs_inode_is_data_compressed helper Gao Xiang via Linux-erofs
2019-09-02 12:07     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 06/21] erofs: kill erofs_{init,exit}_inode_cache Gao Xiang via Linux-erofs
2019-09-02 12:09     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 07/21] erofs: use erofs_inode naming Gao Xiang via Linux-erofs
2019-09-02 12:10     ` Christoph Hellwig
2019-09-02 12:13       ` Gao Xiang
2019-09-02 12:47         ` Christoph Hellwig
2019-09-02 13:33           ` Gao Xiang
2019-09-01  5:51   ` [PATCH 08/21] erofs: update comments in inode.c Gao Xiang via Linux-erofs
2019-09-01  5:51   ` [PATCH 09/21] erofs: update erofs symlink stuffs Gao Xiang via Linux-erofs
2019-09-02 12:11     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 10/21] erofs: kill is_inode_layout_compression() Gao Xiang via Linux-erofs
2019-09-02 12:11     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 11/21] erofs: use dsb instead of layout for ondisk super_block Gao Xiang via Linux-erofs
2019-09-02 12:12     ` Christoph Hellwig
2019-09-02 12:15       ` Gao Xiang
2019-09-01  5:51   ` [PATCH 12/21] erofs: kill verbose debug info in erofs_fill_super Gao Xiang via Linux-erofs
2019-09-02 12:14     ` Christoph Hellwig
2019-09-02 12:18       ` Gao Xiang
2019-09-01  5:51   ` [PATCH 13/21] erofs: simplify erofs_grab_bio() since bio_alloc() never fail Gao Xiang via Linux-erofs
2019-09-02 12:20     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 14/21] erofs: kill prio and nofail of erofs_get_meta_page() Gao Xiang via Linux-erofs
2019-09-02 12:21     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 15/21] erofs: kill __submit_bio() Gao Xiang via Linux-erofs
2019-09-01  5:51   ` [PATCH 16/21] erofs: kill magic underscores Gao Xiang via Linux-erofs
2019-09-02 12:26     ` Christoph Hellwig
2019-09-02 12:39       ` Gao Xiang
2019-09-02 12:54         ` Christoph Hellwig
2019-09-02 13:38           ` Gao Xiang
2019-09-01  5:51   ` [PATCH 17/21] erofs: use a switch statement when dealing with the file modes Gao Xiang via Linux-erofs
2019-09-02 12:27     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 18/21] erofs: add "erofs_" prefix for common and short functions Gao Xiang via Linux-erofs
2019-09-02 12:28     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 19/21] erofs: kill all erofs specific fault injection Gao Xiang via Linux-erofs
2019-09-02 12:28     ` Christoph Hellwig
2019-09-01  5:51   ` [PATCH 20/21] erofs: kill use_vmap module parameter Gao Xiang via Linux-erofs
2019-09-02 12:31     ` Christoph Hellwig
2019-09-02 12:43       ` Gao Xiang
2019-09-01  5:51   ` [PATCH 21/21] erofs: save one level of indentation Gao Xiang via Linux-erofs
2019-09-02 12:31     ` Christoph Hellwig
2019-09-02 12:46   ` [PATCH 00/21] erofs: patchset addressing Christoph's comments Christoph Hellwig
2019-09-02 14:24     ` Gao Xiang
2019-09-02 15:23       ` Christoph Hellwig
2019-09-02 15:50         ` Gao Xiang
2019-09-03  6:58           ` Christoph Hellwig
2019-09-03  8:17             ` Gao Xiang
2019-09-03 15:37               ` Christoph Hellwig
2019-09-03 15:43                 ` Gao Xiang via Linux-erofs
2019-09-04  2:08   ` [PATCH v2 00/25] " Gao Xiang
2019-09-04  2:08     ` [PATCH v2 01/25] erofs: remove all the byte offset comments Gao Xiang
2019-09-04  2:08     ` [PATCH v2 02/25] erofs: on-disk format should have explicitly assigned numbers Gao Xiang
2019-09-04  2:08     ` [PATCH v2 03/25] erofs: some macros are much more readable as a function Gao Xiang
2019-09-04  2:08     ` [PATCH v2 04/25] erofs: kill __packed for on-disk structures Gao Xiang
2019-09-04  2:08     ` [PATCH v2 05/25] erofs: update erofs_inode_is_data_compressed helper Gao Xiang
2019-09-04  2:08     ` [PATCH v2 06/25] erofs: use feature_incompat rather than requirements Gao Xiang
2019-09-04  2:08     ` [PATCH v2 07/25] erofs: better naming for erofs inode related stuffs Gao Xiang
2019-09-04  2:08     ` [PATCH v2 08/25] erofs: kill erofs_{init,exit}_inode_cache Gao Xiang
2019-09-04  2:08     ` [PATCH v2 09/25] erofs: use erofs_inode naming Gao Xiang
2019-09-04  2:08     ` [PATCH v2 10/25] erofs: update erofs_fs.h comments Gao Xiang
2019-09-04  2:08     ` [PATCH v2 11/25] erofs: update comments in inode.c Gao Xiang
2019-09-04  2:08     ` [PATCH v2 12/25] erofs: better erofs symlink stuffs Gao Xiang
2019-09-04  2:09     ` [PATCH v2 13/25] erofs: use dsb instead of layout for ondisk super_block Gao Xiang
2019-09-04  2:09     ` [PATCH v2 14/25] erofs: kill verbose debug info in erofs_fill_super Gao Xiang
2019-09-04  2:09     ` [PATCH v2 15/25] erofs: localize erofs_grab_bio() Gao Xiang
2019-09-04  2:09     ` [PATCH v2 16/25] erofs: kill prio and nofail of erofs_get_meta_page() Gao Xiang
2019-09-04  2:09     ` [PATCH v2 17/25] erofs: kill __submit_bio() Gao Xiang
2019-09-04  2:09     ` [PATCH v2 18/25] erofs: add "erofs_" prefix for common and short functions Gao Xiang
2019-09-04  2:09     ` [PATCH v2 19/25] erofs: kill all erofs specific fault injection Gao Xiang
2019-09-04  2:09     ` [PATCH v2 20/25] erofs: kill use_vmap module parameter Gao Xiang
2019-09-04  2:09     ` [PATCH v2 21/25] erofs: save one level of indentation Gao Xiang
2019-09-04  2:09     ` [PATCH v2 22/25] erofs: rename errln/infoln/debugln to erofs_{err, info, dbg} Gao Xiang
2019-09-04  2:09     ` [PATCH v2 23/25] erofs: use read_mapping_page instead of sb_bread Gao Xiang
2019-09-04  2:09     ` [PATCH v2 24/25] erofs: always use iget5_locked Gao Xiang
2019-09-04  2:09     ` [PATCH v2 25/25] erofs: use read_cache_page_gfp for erofs_get_meta_page Gao Xiang
2019-09-04  3:27     ` [PATCH v2 00/25] erofs: patchset addressing Christoph's comments Chao Yu
2019-09-05  1:03       ` Gao Xiang via Linux-erofs
2019-09-05 11:30         ` Christoph Hellwig
2019-09-04  5:16     ` Christoph Hellwig
2019-09-04  6:08       ` Gao Xiang via Linux-erofs

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=20190802125347.166018-22-gaoxiang25@huawei.com \
    --to=gaoxiang25@huawei.com \
    /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).