rpms/qatzip
6
0
Fork 0
Code Issues Pull Requests Releases Activity

Fix large files decompression bug

Browse Source 
This is the upstream Intel bug QATAPP-32528.

Resolves: RHEL-35325

Signed-off-by: Vladis Dronov <vdronov@redhat.com>
This commit is contained in:
Vladis Dronov 2025-04-07 18:55:31 +02:00
parent a78aeed70b
commit 73d1a76fd5
2 changed files with 70 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

63
qatapp-32528-fix-large-files.patch Normal file
View File

@ -0,0 +1,63 @@
From 906ff19a0bc4d770d0d8cd484bdce8d5fbb1843f Mon Sep 17 00:00:00 2001
From: Xinghong Chen <xinghong.chen@intel.com>
Date: Wed, 31 Jul 2024 22:57:09 -0400
Subject: [PATCH] QATAPP-32528: Fixed the qzip decompress large file issue
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Content-type: text/plain
* The “bytes_processed” only updated when “DATA_ERROR/BUF_ERROR”
happened. It’s OK if during each “src buffer”, it would run into
“DATA_ERROR/BUF_ERROR” (This is the majority of cases, for 512M
src buffer, it almost always have “incomplete source buffer”
return). It’s also ok, if for each “src buffer”, it never run into
“DATA_ERROR/BUF_ERROR”, so it wouldn’t call “fseek”,
then bytes_processed is useless. But, if we mix those tow scenarios,
For some “src buffer” run into “DATA_ERROR/BUF_ERROR” and
others don’t. Then “bytes_processed” will miss some process
source for which complete decompression without error issue.
It cause the “fseek” set to incorrect file stream position.
Signed-off-by: Xinghong Chen <xinghong.chen@intel.com>
Signed-off-by: David Qian <david.qian@intel.com>
Signed-off-by: Vladis Dronov <vdronov@redhat.com>
---
utils/qzip.c | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/utils/qzip.c b/utils/qzip.c
index bd69858..63d3702 100644
--- a/utils/qzip.c
+++ b/utils/qzip.c
@@ -255,7 +255,8 @@ void doProcessFile(QzSession_T *sess, const char *src_file_name,
FILE *src_file = NULL;
FILE *dst_file = NULL;
unsigned int bytes_read = 0;
- unsigned long bytes_processed = 0;
+ unsigned int bytes_lookahead = 0;
+ long offset_revert = 0;
unsigned int ratio_idx = 0;
const unsigned int ratio_limit =
sizeof(g_bufsz_expansion_ratio) / sizeof(unsigned int);
@@ -326,14 +327,15 @@ void doProcessFile(QzSession_T *sess, const char *src_file_name,
puts((is_compress) ? "Compressing..." : "Decompressing...");
+ bytes_lookahead = bytes_read;
ret = doProcessBuffer(sess, src_buffer, &bytes_read, dst_buffer,
dst_buffer_size, time_list_head, dst_file,
&dst_file_size, is_compress);
if (QZ_DATA_ERROR == ret || QZ_BUF_ERROR == ret) {
- bytes_processed += bytes_read;
if (0 != bytes_read) {
- if (-1 == fseek(src_file, bytes_processed, SEEK_SET)) {
+ offset_revert = (long)bytes_read - (long)bytes_lookahead;
+ if (-1 == fseek(src_file, offset_revert, SEEK_CUR)) {
ret = ERROR;
goto exit;
}
--
2.49.0

9
qatzip.spec
View File

@ -5,7 +5,7 @@
Name: qatzip
Version: 1.1.2
Release: 2%{?dist}
Release: 3%{?dist}
Summary: Intel QuickAssist Technology (QAT) QATzip Library
License: BSD-3-Clause
URL: https://github.com/intel/%{githubname}
@ -19,6 +19,8 @@ BuildRequires: autoconf automake libtool make lz4-devel
# https://bugzilla.redhat.com/show_bug.cgi?id=1987280
ExclusiveArch: x86_64
Patch1: qatapp-32528-fix-large-files.patch
%description
QATzip is a user space library which builds on top of the Intel
QuickAssist Technology user space library, to provide extended
@ -46,7 +48,7 @@ This package contains headers and libraries required to build
applications that use the QATzip APIs.
%prep
%autosetup -n %{githubname}-%{version}
%autosetup -p1 -n %{githubname}-%{version}
%build
%set_build_flags
@ -86,6 +88,9 @@ rm -vf %{buildroot}%{_mandir}/*.pdf
%{_libdir}/pkgconfig/*.pc
%changelog
* Mon Apr 07 2025 Vladis Dronov <vdronov@redhat.com> - 1.1.2-3
- Fix large files decompression bug QATAPP-32528 (RHEL-35325)
* Mon Nov 20 2023 Vladis Dronov <vdronov@redhat.com> - 1.1.2-2
- Rebuild qatzip 1.1.2 for qatlib 23.11 (RHEL-15637)