add generic/795
Subsystems (2)
ChenXiaoSong
reviewerv2 cover add generic/795
From: ChenXiaoSong <[email protected]>
v1->v2:
- Add patch#01 patch#02.
- Patch#03: some cleanups.
v1: https://lore.kernel.org/linux-cifs/[email protected]/
ChenXiaoSong (3):
src: factor out common stat code
generic/002: clean up the test
generic/795: check nlink returned by fstat()
src/Makefile | 13 ++- src/fstat.c | 42 ++++++++++ src/lstat64.c | 166 +------------------------------------ src/stat_common.c | 185 ++++++++++++++++++++++++++++++++++++++++++ src/stat_common.h | 16 ++++ tests/generic/002 | 16 ++-- tests/generic/002.out | 2 +- tests/generic/795 | 60 ++++++++++++++ tests/generic/795.out | 2 + 9 files changed, 324 insertions(+), 178 deletions(-) create mode 100644 src/fstat.c create mode 100644 src/stat_common.c create mode 100644 src/stat_common.h create mode 100755 tests/generic/795 create mode 100644 tests/generic/795.out
Signature · 1 line
2.43.0
ChenXiaoSong
reviewerPATCH v2 1/3 src: factor out common stat code
From: ChenXiaoSong <[email protected]>
To see only the differences in the code copied from lstat64.c to stat_common.c, run:
git show --find-copies --diff-filter=C <commit-id>
Move command-line parsing and stat output formatting from lstat64 into stat_common so the implementation can be shared by other stat helpers.
No functional change.
--- src/Makefile | 11 ++- src/lstat64.c | 166 +---------------------------------------- src/stat_common.c | 185 ++++++++++++++++++++++++++++++++++++++++++++++ src/stat_common.h | 16 ++++ 4 files changed, 212 insertions(+), 166 deletions(-) create mode 100644 src/stat_common.c create mode 100644 src/stat_common.h
src/Makefile +9 -2
@@ -115,7 +115,8 @@ ifeq ($(NEED_INTERNAL_XFS_IOC_EXCHANGE_RANGE),yes) LCFLAGS += -DNEED_INTERNAL_XFS_IOC_EXCHANGE_RANGE endif -CFILES = $(TARGETS:=.c) +CFILES = $(TARGETS:=.c) stat_common.c +HFILES = stat_common.h LDIRT = $(TARGETS) fssum
@@ -129,7 +130,13 @@ fssum: fssum.c md5.c @echo " [CC] $@" $(Q)$(LTLINK) fssum.c md5.c -o $@ $(CFLAGS) $(LDFLAGS) $(LDLIBS) -$(TARGETS): $(LIBTEST) +STAT_TARGETS = lstat64 + +$(STAT_TARGETS): %: %.c stat_common.c stat_common.h $(LIBTEST) + @echo " [CC] $@" + $(Q)$(LTLINK) [email protected] stat_common.c -o $@ $(CFLAGS) $(LDFLAGS) $(LDLIBS) $(LIBTEST) + +$(filter-out $(STAT_TARGETS),$(TARGETS)): $(LIBTEST) @echo " [CC] $@" $(Q)$(LTLINK) [email protected] -o $@ $(CFLAGS) $(LDFLAGS) $(LDLIBS) $(LIBTEST)
src/lstat64.c +2 -164
174 lines not on this page · raw message
src/stat_common.c new file +185 -0
185 lines not on this page · raw message
src/stat_common.h new file +16 -0
@@ -0,0 +1,16 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Copyright (c) 2000-2002 Silicon Graphics, Inc. +* All Rights Reserved. +*/ + +#ifndef STAT_COMMON_H +#define STAT_COMMON_H + +#include <sys/stat.h> + +typedef int (*get_stat_fn )(const char *path,struct stat64 *sbuf); + +int handle_stat (int argc,char **argv,const char *program,get_stat_fn get_stat); + +#endif /* STAT_COMMON_H */
Signature · 1 line
2.43.0
ChenXiaoSong
reviewerPATCH v2 2/3 generic/002: clean up the test
In reply to ChenXiaoSongFrom: ChenXiaoSong <[email protected]>
Add the hardlink group and use shorter loops. Use _exit() and change the success text to "Silence is golden".
--- tests/generic/002 | 16 ++++++---------- tests/generic/002.out | 2 +- 2 files changed, 7 insertions(+), 11 deletions(-)
tests/generic/002 +6 -10
@@ -2,12 +2,12 @@ # SPDX-License-Identifier: GPL-2.0 # Copyright (c) 2000-2001 Silicon Graphics, Inc. All Rights Reserved. # -# FS QA Test No. 002 +# FS QA Test 002 # # simple inode link count test for a regular file # . ./common/preamble -_begin_fstest metadata udf auto quick +_begin_fstest metadata udf auto quick hardlink # Import common functions. . ./common/filter
@@ -17,14 +17,11 @@ status=0 # success is the default! _require_test _require_hardlinks -echo "Silence is goodness ..." - # ensure target directory exists mkdir `dirname $TEST_DIR/tmp` 2>/dev/null touch $TEST_DIR/tmp.1 -for l in 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 -do +for ((l = 2; l <= 20; l++)); do ln $TEST_DIR/tmp.1 $TEST_DIR/tmp.$l x=`$here/src/lstat64 $TEST_DIR/tmp.1 | sed -n -e '/ Links: /s/.*Links: *//p'` if [ "$l" -ne $x ]
@@ -35,8 +32,7 @@ do fi done -for l in 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 -do +for ((l = 20; l >= 1; l--)); do x=`$here/src/lstat64 $TEST_DIR/tmp.1 | sed -n -e '/ Links: /s/.*Links: *//p'` if [ "$l" -ne $x ] then
@@ -47,5 +43,5 @@ do rm -f $TEST_DIR/tmp.$l done -# success, all done -exit +echo "Silence is golden" +exit $status
tests/generic/002.out +1 -1
@@ -1,2 +1,2 @@ QA output created by 002 -Silence is goodness ... +Silence is golden
Signature · 1 line
2.43.0
ChenXiaoSong
reviewerPATCH v2 3/3 generic/795: check nlink returned by fstat()
In reply to ChenXiaoSongFrom: ChenXiaoSong <[email protected]>
Add a test to verify that fstat(2) returns the expected st_nlink value as hardlinks are created and removed.
--- src/Makefile | 4 +-- src/fstat.c | 42 ++++++++++++++++++++++++++++++ tests/generic/795 | 60 +++++++++++++++++++++++++++++++++++++++++++ tests/generic/795.out | 2 ++ 4 files changed, 106 insertions(+), 2 deletions(-) create mode 100644 src/fstat.c create mode 100755 tests/generic/795 create mode 100644 tests/generic/795.out
src/Makefile +2 -2
@@ -6,7 +6,7 @@ TOPDIR = .. include $(TOPDIR)/include/builddefs -TARGETS = dirstress fill fill2 getpagesize holes lstat64 \ +TARGETS = dirstress fill fill2 getpagesize holes lstat64 fstat \ nametest permname randholes runas truncfile usemem \ mmapcat append_reader append_writer dirperf metaperf \ devzero feature alloc fault fstest t_access_root \
@@ -130,7 +130,7 @@ fssum: fssum.c md5.c @echo " [CC] $@" $(Q)$(LTLINK) fssum.c md5.c -o $@ $(CFLAGS) $(LDFLAGS) $(LDLIBS) -STAT_TARGETS = lstat64 +STAT_TARGETS = lstat64 fstat $(STAT_TARGETS): %: %.c stat_common.c stat_common.h $(LIBTEST) @echo " [CC] $@"
src/fstat.c new file +42 -0
42 lines not on this page · raw message
tests/generic/795 new file +60 -0
60 lines not on this page · raw message
tests/generic/795.out new file +2 -0
@@ -0,0 +1,2 @@ +QA output created by 795 +Silence is golden
Signature · 1 line
2.43.0
Quoted 13 lines · ChenXiaoSong
From: ChenXiaoSong <[email protected]>
v1->v2:
- Add patch#01 patch#02.
- Patch#03: some cleanups.v1: https://lore.kernel.org/linux-cifs/[email protected]/
ChenXiaoSong (3):
src: factor out common stat code
generic/002: clean up the test
generic/795: check nlink returned byfstat()
Quoted 7 lines
src/Makefile | 13 ++- src/fstat.c | 42 ++++++++++ src/lstat64.c | 166 +------------------------------------ src/stat_common.c | 185 ++++++++++++++++++++++++++++++++++++++++++ src/stat_common.h | 16 ++++ tests/generic/002 | 16 ++-- tests/generic/002.out | 2 +-
Hi,
I might not have explained clearly during the review. I wasn't asking you to rewrite g/002 and lstat64.c. I was suggesting checking whether src/fstat.c in your previous patch could be replaced by the stat command in xfs_io [1]. Check man xfs_io for the "stat" command to see if it can fulfill your test requirement.
Due to I noticed you need information like st_nlink from struct stat, so xfs_io -c "stat -r" ... should meet your requirements. Please give it a try, if it works, it will make the whole patch much cleaner.
P.S. For a new test case, please don't use a specific case number (like generic/795) in the subject or commit log, as it might get assigned a different number when merged. You can just use something like:
"[PATCH] generic: new test to ..."
Thanks,
Zorro
[1]
https://lore.kernel.org/fstests/[email protected]/T/#mff79f3e74e8cefbdda71e6a22f6a20f2dd985661
"I'm wondering if a specific src/fstat.c is needed. Can we replace the whole src/fstat.c with xfs_io -c stat ? Can you give it a try?"
Quoted 12 lines
tests/generic/795 | 60 ++++++++++++++ tests/generic/795.out | 2 + 9 files changed, 324 insertions(+), 178 deletions(-) create mode 100644 src/fstat.c create mode 100644 src/stat_common.c create mode 100644 src/stat_common.h create mode 100755 tests/generic/795 create mode 100644 tests/generic/795.outSignature · 1 line
2.43.0
ChenXiaoSong
reviewerHi Zorro,
The stat command uses the statx() syscall rather than fstat(), so we have to use a C program to call fstat().
Quoted 14 lines · Zorro Lang
I might not have explained clearly during the review. I wasn't asking you to rewrite g/002 and
lstat64.c. I was suggesting checking whethersrc/fstat.cin your previous patch could be replaced by the stat command inxfs_io[1]. Checkman xfs_iofor the "stat" command to see if it can fulfill your test requirement.Due to I noticed you need information like
st_nlinkfrom struct stat, soxfs_io -c "stat -r" ...should meet your requirements. Please give it a try, if it works, it will make the whole patch much cleaner.
P.S. For a new test case, please don't use a specific case number (like generic/795) in the subject or commit log, as it might get assigned a different number when merged. You can just use something like:
"[PATCH] generic: new test to ..."
Signature · 3 lines
ChenXiaoSong <[email protected]> Chinese Homepage: https://chenxiaosong.com English Homepage: https://chenxiaosong.com/en
Quoted 4 lines · ChenXiaoSong
Hi Zorro,
The stat command uses the
statx()syscall rather thanfstat(), so we have to use a C program to callfstat().
I'm not sure how you got at that conclusion.
Quoted 10 lines · Zorro Lang
Quoted 8 lines
I might not have explained clearly during the review. I wasn't asking you to rewrite g/002 and
lstat64.c. I was suggesting checking whethersrc/fstat.cin your previous patch could be replaced by the stat command inxfs_io[1]. Checkman xfs_iofor the "stat" command to see if it can fulfill your test requirement.Due to I noticed you need information like
st_nlinkfrom struct stat, soxfs_io -c "stat -r" ...should meet your requirements. Please
^^^^^^^^^^^^^^^^^^
Please note that I have been talking about xfs_io stat subcommand all along, rather than the coreutils' stat command. I suspect there might be a misunderstanding here.
xfstests inherently depends on xfsprogs, particularly the xfs_io command. I just checked this for you, the stat subcommand in xfs_io does indeed use the fstat system call [1][2], not statx, and its -r option will print out fields like st_nlink.
Please double-check this on your end, make sure it works for you.
Thanks,
Zorro
[1]
xfsprogs/io/stat.c:stat_f() :
int
stat_f(
int argc,
char **argv)
{
struct stat st;
int c, verbose = 0, raw = 0;
while ((c = getopt(argc, argv, "rv")) != EOF) {
switch (c) {
case 'r':
raw = 1;
break;
case 'v':
verbose = 1;
break;
default:
exitcode = 1;
return command_usage(&stat_cmd);
}
}
if (fstat(file->fd, &st) < 0) {
perror("fstat");
exitcode = 1;
return 0;
}
if (raw)
return dump_raw_stat(&st);...
[2]
$ strace xfs_io -c "stat -r" README 2>&1|grep -E "statx|fstat"
fstat(3, {st_mode=S_IFREG|0644, st_size=106979, ...}) = 0
fstat(3, {st_mode=S_IFREG|0755, st_size=36680, ...}) = 0
fstat(3, {st_mode=S_IFREG|0755, st_size=236720, ...}) = 0
fstat(3, {st_mode=S_IFREG|0755, st_size=2482096, ...}) = 0
fstat(3, {st_mode=S_IFREG|0755, st_size=187232, ...}) = 0
fstat(3, {st_mode=S_IFREG|0644, st_size=233380912, ...}) = 0
fstat(3, {st_mode=S_IFREG|0444, st_size=0, ...}) = 0
newfstatat(AT_FDCWD, "/", {st_mode=S_IFDIR|0555, st_size=154, ...}, 0) = 0
newfstatat(AT_FDCWD, "/", {st_mode=S_IFDIR|0555, st_size=154, ...}, 0) = 0
newfstatat(AT_FDCWD, "/home", {st_mode=S_IFDIR|0755, st_size=10, ...}, 0) = 0
newfstatat(AT_FDCWD, "/home", {st_mode=S_IFDIR|0755, st_size=10, ...}, 0) = 0
newfstatat(AT_FDCWD, "/boot", {st_mode=S_IFDIR|0555, st_size=4096, ...}, 0) = 0
newfstatat(AT_FDCWD, "/boot", {st_mode=S_IFDIR|0555, st_size=4096, ...}, 0) = 0
newfstatat(AT_FDCWD, "/boot/efi", {st_mode=S_IFDIR|0700, st_size=4096, ...}, 0) = 0
newfstatat(AT_FDCWD, "/boot/efi", {st_mode=S_IFDIR|0700, st_size=4096, ...}, 0) = 0
newfstatat(AT_FDCWD, "README", {st_mode=S_IFREG|0644, st_size=27145, ...}, 0) = 0
fstatfs(3, {f_type=BTRFS_SUPER_MAGIC, f_bsize=4096, f_blocks=249372928, f_bfree=196292374, f_bavail=195356006, f_files=0, f_ffree=0, f_fsid={val=[0xfb10ffbe, 0x278a3f4f]}, f_namelen=255, f_frsize=4096, f_flags=ST_VALID|ST_RELATIME}) = 0
fstat(3, {st_mode=S_IFREG|0644, st_size=27145, ...}) = 0
newfstatat(AT_FDCWD, "README", {st_mode=S_IFREG|0644, st_size=27145, ...}, 0) = 0
fstatfs(3, {f_type=BTRFS_SUPER_MAGIC, f_bsize=4096, f_blocks=249372928, f_bfree=196292374, f_bavail=195356006, f_files=0, f_ffree=0, f_fsid={val=[0xfb10ffbe, 0x278a3f4f]}, f_namelen=255, f_frsize=4096, f_flags=ST_VALID|ST_RELATIME}) = 0
fstat(3, {st_mode=S_IFREG|0644, st_size=27145, ...}) = 0
fstat(4, {st_mode=S_IFREG|0644, st_size=2998, ...}) = 0
fstat(3, {st_mode=S_IFREG|0644, st_size=27145, ...}) = 0
fstat(1, {st_mode=S_IFIFO|0600, st_size=0, ...}) = 0Quoted 12 lines
Quoted 6 lines
give it a try, if it works, it will make the whole patch much cleaner.
P.S. For a new test case, please don't use a specific case number (like generic/795) in the subject or commit log, as it might get assigned a different number when merged. You can just use something like:
"[PATCH] generic: new test to ..."Signature · 3 lines
ChenXiaoSong <[email protected]> Chinese Homepage: https://chenxiaosong.com English Homepage: https://chenxiaosong.com/en
ChenXiaoSong
reviewerYes, I checked the xfs_io source code, and it does indeed call the fstat() system call. Thanks for your suggestion! I will send the next version soon.
https://git.kernel.org/pub/scm/fs/xfs/xfsprogs-dev.git/tree/io/stat.c#n552
https://git.kernel.org/pub/scm/fs/xfs/xfsprogs-dev.git/tree/io/stat.c#n166
Quoted 9 lines · Zorro Lang
Please note that I have been talking about
xfs_iostat subcommand all along, rather than the coreutils' stat command. I suspect there might be a misunderstanding here.xfstests inherently depends on xfsprogs, particularly the
xfs_iocommand. I just checked this for you, the stat subcommand inxfs_iodoes indeed use the fstat system call [1][2], not statx, and its -r option will print out fields likest_nlink.Please double-check this on your end, make sure it works for you.
Signature · 3 lines
ChenXiaoSong <[email protected]> Chinese Homepage: https://chenxiaosong.com English Homepage: https://chenxiaosong.com/en