Rebase to 1.8.4-2 for performance and compression fixes

Resolves: RHEL-72588

Signed-off-by: Neal Gompa <ngompa@centosproject.org>

.gitignore

@@ -1,2 +1,3 @@
 /erofs-utils-1.7.1.tar.gz
 /erofs-utils-1.8.2.tar.gz
+/erofs-utils-1.8.4.tar.gz

erofs-utils-1.8.2-fix-all-fragments.patch

@@ -1,51 +0,0 @@
-From 882ad1c3157f7544bd4d004e3b6d744f0cbe3ffc Mon Sep 17 00:00:00 2001
-From: Gao Xiang <hsiangkao@linux.alibaba.com>
-Date: Sat, 12 Oct 2024 11:52:13 +0800
-Subject: erofs-utils: mkfs: fix `-Eall-fragments` for multi-threaded
- compression
-
-If `-Eall-fragments` is specified when multi-threaded compression is
-enabled, it should only apply to the packed inode instead of other
-real inodes for now.
-
-Fixes: 10c1590c0920 ("erofs-utils: enable multi-threaded support for `-Eall-fragments`")
-Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
-Link: https://lore.kernel.org/r/20241012035213.3729725-1-hsiangkao@linux.alibaba.com
----
- lib/compress.c | 13 +++++++------
- 1 file changed, 7 insertions(+), 6 deletions(-)
-
-diff --git a/lib/compress.c b/lib/compress.c
-index 5d6fb2a..cbd4620 100644
---- a/lib/compress.c
-+++ b/lib/compress.c
-@@ -1456,12 +1456,8 @@ void *erofs_begin_compressed_file(struct erofs_inode *inode, int fd, u64 fpos)
- 	inode->idata_size = 0;
- 	inode->fragment_size = 0;
- 
--	if (z_erofs_mt_enabled) {
--		ictx = malloc(sizeof(*ictx));
--		if (!ictx)
--			return ERR_PTR(-ENOMEM);
--		ictx->fd = dup(fd);
--	} else {
-+	if (!z_erofs_mt_enabled ||
-+	    (cfg.c_all_fragments && !erofs_is_packed_inode(inode))) {
- #ifdef EROFS_MT_ENABLED
- 		pthread_mutex_lock(&g_ictx.mutex);
- 		if (g_ictx.seg_num)
-@@ -1471,6 +1467,11 @@ void *erofs_begin_compressed_file(struct erofs_inode *inode, int fd, u64 fpos)
- #endif
- 		ictx = &g_ictx;
- 		ictx->fd = fd;
-+	} else {
-+		ictx = malloc(sizeof(*ictx));
-+		if (!ictx)
-+			return ERR_PTR(-ENOMEM);
-+		ictx->fd = dup(fd);
- 	}
- 
- 	ictx->ccfg = &erofs_ccfg[inode->z_algorithmtype[0]];
--- 
-cgit 1.2.3-korg
-

erofs-utils-1.8.4-fragdedupe-inode-support.patch

@@ -0,0 +1,102 @@
+From 06875b3f2182eab24b81083dfde542f778b201cc Mon Sep 17 00:00:00 2001
+From: Gao Xiang <hsiangkao@linux.alibaba.com>
+Date: Fri, 3 Jan 2025 10:40:11 +0800
+Subject: erofs-utils: mkfs: support `-Efragdedupe=inode`
+
+If the entire inode can be deduplicated against an existing fragment,
+simply reuse it.
+
+Multi-threading can still be applied for `-Efragdedupe=inode` with
+the current codebase:
+
+Fedora Linux 39 (Workstation Edition) LiveCD results:
+ -zlzma,level=6,dictsize=131072 -C65536 -Eall-fragments
+
+   `-E^fragdedupe`         2,003,587,072 bytes (1911 MiB)
+   `-Efragdedupe=inode`    1,970,577,408 bytes (1880 MiB)
+
+Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
+Link: https://lore.kernel.org/r/20250103024011.198163-1-hsiangkao@linux.alibaba.com
+---
+ include/erofs/config.h |  8 +++++++-
+ lib/compress.c         |  9 +++++++--
+ mkfs/main.c            | 13 ++++++++++---
+ 3 files changed, 24 insertions(+), 6 deletions(-)
+
+diff --git a/include/erofs/config.h b/include/erofs/config.h
+index 47e4d00..92c1467 100644
+--- a/include/erofs/config.h
++++ b/include/erofs/config.h
+@@ -33,6 +33,12 @@ enum {
+ 	TIMESTAMP_CLAMPING,
+ };
+ 
++enum {
++	FRAGDEDUPE_FULL,
++	FRAGDEDUPE_INODE,
++	FRAGDEDUPE_OFF,
++};
++
+ #define EROFS_MAX_COMPR_CFGS		64
+ 
+ struct erofs_compr_opts {
+@@ -53,7 +59,7 @@ struct erofs_configure {
+ 	bool c_fragments;
+ 	bool c_all_fragments;
+ 	bool c_dedupe;
+-	bool c_nofragdedupe;
++	char c_fragdedupe;
+ 	bool c_ignore_mtime;
+ 	bool c_showprogress;
+ 	bool c_extra_ea_name_prefixes;
+diff --git a/lib/compress.c b/lib/compress.c
+index 0e8faad..20ab208 100644
+--- a/lib/compress.c
++++ b/lib/compress.c
+@@ -1527,12 +1527,17 @@ void *erofs_begin_compressed_file(struct erofs_inode *inode, int fd, u64 fpos)
+ 	 * parts into the packed inode.
+ 	 */
+ 	if (cfg.c_fragments && !erofs_is_packed_inode(inode) &&
+-	    !cfg.c_nofragdedupe) {
++	    cfg.c_fragdedupe != FRAGDEDUPE_OFF) {
+ 		ret = z_erofs_fragments_dedupe(inode, fd, &ictx->tof_chksum);
+ 		if (ret < 0)
+ 			goto err_free_ictx;
+-	}
+ 
++		if (cfg.c_fragdedupe == FRAGDEDUPE_INODE &&
++		    inode->fragment_size < inode->i_size) {
++			erofs_dbg("Discard the sub-inode tail fragment @ nid %llu", inode->nid);
++			inode->fragment_size = 0;
++		}
++	}
+ 	ictx->inode = inode;
+ 	ictx->fpos = fpos;
+ 	init_list_head(&ictx->extents);
+diff --git a/mkfs/main.c b/mkfs/main.c
+index 3f74fa2..0f6a32b 100644
+--- a/mkfs/main.c
++++ b/mkfs/main.c
+@@ -306,9 +306,16 @@ static int erofs_mkfs_feat_set_dedupe(bool en, const char *val,
+ static int erofs_mkfs_feat_set_fragdedupe(bool en, const char *val,
+ 					  unsigned int vallen)
+ {
+-	if (vallen)
+-		return -EINVAL;
+-	cfg.c_nofragdedupe = !en;
++	if (!en) {
++		if (vallen)
++			return -EINVAL;
++		cfg.c_fragdedupe = FRAGDEDUPE_OFF;
++	} else if (vallen == sizeof("inode") - 1 &&
++		   !memcmp(val, "inode", vallen)) {
++		cfg.c_fragdedupe = FRAGDEDUPE_INODE;
++	} else {
++		cfg.c_fragdedupe = FRAGDEDUPE_FULL;
++	}
+ 	return 0;
+ }
+ 
+-- 
+cgit 1.2.3-korg
+

erofs-utils.spec

@@ -5,11 +5,12 @@
 %bcond qpl      %[ 0%{?fedora} >= 41 && "%{_arch}" == "x86_64" ]
 %bcond selinux  1
 %bcond uuid     1
+%bcond xxhash   %[ 0%{?fedora} || (0%{?rhel} >= 9 && 0%{?rhel} < 10) ]
 %bcond zlib     1
 %bcond zstd     1
 
 Name:           erofs-utils
-Version:        1.8.2
+Version:        1.8.4
 Release:        2%{?dist}
 
 Summary:        Utilities for working with EROFS
@@ -17,18 +18,19 @@ License:        GPL-2.0-only AND GPL-2.0-or-later AND (GPL-2.0-only OR Apache-2.
 URL:            https://erofs.docs.kernel.org/
 
 Source:         https://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs-utils.git/snapshot/%{name}-%{version}.tar.gz
-Patch:          https://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs-utils.git/patch/?id=882ad1c3157f7544bd4d004e3b6d744f0cbe3ffc#/%{name}-1.8.2-fix-all-fragments.patch
+Patch:          https://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs-utils.git/patch/?id=06875b3f2182eab24b81083dfde542f778b201cc#/%{name}-1.8.4-fragdedupe-inode-support.patch
 
 BuildRequires:  %[ "%{toolchain}" == "clang" ? "clang compiler-rt" : "gcc" ]
 BuildRequires:  libtool
 BuildRequires:  make
 %{?with_deflate:BuildRequires:  pkgconfig(libdeflate)}
 %{?with_fuse:BuildRequires:  pkgconfig(fuse3) >= 3.2}
-%{?with_lz4:BuildRequires:  lz4-devel >= 1.9.3}
+%{?with_lz4:BuildRequires:  pkgconfig(liblz4) >= 1.9.3}
 %{?with_lzma:BuildRequires:  pkgconfig(liblzma) >= 5.4}
 %{?with_qpl:BuildRequires:  pkgconfig(qpl) >= 1.5.0}
 %{?with_selinux:BuildRequires:  pkgconfig(libselinux)}
 %{?with_uuid:BuildRequires:  pkgconfig(uuid)}
+%{?with_xxhash:BuildRequires:  pkgconfig(libxxhash)}
 %{?with_zlib:BuildRequires:  pkgconfig(zlib)}
 %{?with_zstd:BuildRequires:  pkgconfig(libzstd) >= 1.4.0}
 
@@ -55,9 +57,9 @@ This package includes erofsfuse to mount EROFS images.
 
 %prep
 %autosetup -p1
-autoreconf -fi
 
 %build
+autoreconf -fi
 %configure \
     --enable-multithreading \
     --%{?with_deflate:with}%{!?with_deflate:without}-libdeflate \
@@ -67,6 +69,7 @@ autoreconf -fi
     --%{?with_qpl:with}%{!?with_qpl:without}-qpl \
     --%{?with_selinux:with}%{!?with_selinux:without}-selinux \
     --%{?with_uuid:with}%{!?with_uuid:without}-uuid \
+    --%{?with_xxhash:with}%{!?with_xxhash:without}-xxhash \
     --%{?with_zlib:with}%{!?with_zlib:without}-zlib \
     --%{?with_zstd:with}%{!?with_zstd:without}-libzstd
 %make_build
@@ -95,6 +98,10 @@ autoreconf -fi
 
 
 %changelog
+* Thu Jan 02 2025 Neal Gompa <ngompa@centosproject.org> - 1.8.4-2
+- Rebase to 1.8.4-2 for performance and compression fixes
+  Resolves: RHEL-72588
+
 * Wed Nov 20 2024 Abhi Das <adas@redhat.com> - 1.8.2-2
 - Rebase to 1.8.2-2 for zstd and multithreading support and
   Remove erroneously added gating.yaml

sources

@@ -1 +1 @@
-SHA512 (erofs-utils-1.8.2.tar.gz) = 474ea30662e498e6ece5d4e5171c333ec040eaffa4f5670d1b728b1461a2a18fa90ea6cc2ad7b26e87732d735985e9108a9c49a7072b4dc4fdfee7916df48593
+SHA512 (erofs-utils-1.8.4.tar.gz) = c941b0a2ab6c650a9aa4c9cadeb277ebc87007dc51354ff013c7cb763e6e8c9d44ed9e4791730ed05088faaba8c612198b924e70f5e52019382cfdf6d2e6b677