Skip to contentSkip to conversation

Sparse based checker and rule proposal

PATCH Superseded Richard Palethorpe 3 participants 16 messages Jun 29, 2021 – Jul 9, 2021

Subsystems (2)

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

cover Sparse based checker and rule proposal

Hello,

So it turns out that it is quite easy to implement the TST_RET/ERR check in Sparse. It compiles the code into an IR which is just the right abstraction for the test. It is also possible to inspect the AST with Sparse. It appears more difficult to inspect the AST before macros are expanded.

Sparse is not packaged as a shared library, but has few/no dependencies and is easy to compile. It seems like it was meant to be vendored. Including it as a git module seems reasonable to me. Please see tools/sparse/README.md.

There are still a lot of errors/noise when running 'make check' on the entire tree. These are mainly caused by old function definitions and such. They need to be fixed before the tool can be used properly.

Also I have tried to document the rule and created a list of rules. So this can also be taken as a formal proposal for the rule itself.

Thanks,
Richard Palethorpe (7):
Add Sparse based checker and TST_RET/ERR check Add 'make check' to the build system
meltdown: Set CFLAGS for check targets
doc: Add rules and recommendations list
doc: Remind authors and maintainers to run make check
doc: Document TEST macro and state TST_RET/ERR rule LTP-002 Reference LTP-002 rule in Cocci scripts

 .gitmodules                                   |   3 +
 Makefile                                      |   8 +
 doc/c-test-api.txt                            |  47 ++++++
 doc/library-api-writing-guidelines.txt        |  14 ++
 doc/maintainer-patch-review-checklist.txt     |   2 +-
 doc/rules.tsv                                 |   3 +
 doc/test-writing-guidelines.txt               |   6 +
 include/mk/env_post.mk                        |   8 +
 include/mk/generic_leaf_target.inc            |   5 +-
 include/mk/generic_trunk_target.inc           |   7 +-
 include/mk/lib.mk                             |   3 +
 include/mk/module.mk                          |   2 +
 include/mk/rules.mk                           |   9 ++
 include/mk/sparse.mk                          |   9 ++
 include/mk/testcases.mk                       |   1 +
 .../coccinelle/libltp-test-macro-vars.cocci   |   6 +-
 scripts/coccinelle/libltp-test-macro.cocci    |   4 +-
 testcases/cve/Makefile                        |   1 +
 testcases/open_posix_testsuite/Makefile       |   4 +
 tools/sparse/.gitignore                       |   1 +
 tools/sparse/Makefile                         |  20 +++
 tools/sparse/README.md                        |  34 ++++
 tools/sparse/main.c                           | 148 ++++++++++++++++++
 tools/sparse/sparse-src                       |   1 +
 24 files changed, 335 insertions(+), 11 deletions(-)
 create mode 100644 doc/rules.tsv
 create mode 100644 include/mk/sparse.mk
 create mode 100644 tools/sparse/.gitignore
 create mode 100644 tools/sparse/Makefile
 create mode 100644 tools/sparse/README.md
 create mode 100644 tools/sparse/main.c
 create mode 160000 tools/sparse/sparse-src
Signature · 1 line
2.31.1
#1 lore raw

PATCH 1/7 Add Sparse based checker and TST_RET/ERR check

Vendors in Sparse as a git module. Then uses it to check for stores to TST_RET/ERR within the library.

---
 .gitmodules             |   3 +
 tools/sparse/.gitignore |   1 +
 tools/sparse/Makefile   |  20 ++++++
 tools/sparse/README.md  |  34 +++++++++
 tools/sparse/main.c     | 148 ++++++++++++++++++++++++++++++++++++++++
 tools/sparse/sparse-src |   1 +
 6 files changed, 207 insertions(+)
 create mode 100644 tools/sparse/.gitignore
 create mode 100644 tools/sparse/Makefile
 create mode 100644 tools/sparse/README.md
 create mode 100644 tools/sparse/main.c
 create mode 160000 tools/sparse/sparse-src
.gitmodules +3 -0
index 1c9e9c38a..a3c34af4b 100644
@@ -1,3 +1,6 @@
 [submodule "testcases/kernel/mce-test"]
 	path = testcases/kernel/mce-test
 	url = git://git.kernel.org/pub/scm/linux/kernel/git/gong.chen/mce-test.git
+[submodule "tools/sparse/sparse-src"]
+	path = tools/sparse/sparse-src
+	url = git://git.kernel.org/pub/scm/devel/sparse/sparse.git
tools/sparse/.gitignore new file +1 -0
new file mode 100644
index 000000000..ba2906d06
@@ -0,0 +1 @@
+main
tools/sparse/Makefile new file +20 -0
new file mode 100644
index 000000000..cf4ccdb60
@@ -0,0 +1,20 @@
+# SPDX-License-Identifier: GPL-2.0-or-later
+# Copyright (c) 2021 SUSE LLC <[email protected]>
+
+top_srcdir		?= ../..
+
+include $(top_srcdir)/include/mk/env_pre.mk
+include $(top_srcdir)/include/mk/functions.mk
+
+SPARSE_SRC	?= sparse-src
+
+$(SPARSE_SRC)/libsparse.a:
+	$(error "You need to build Sparse. See tools/sparse/README.md")
+
+HOST_MAKE_TARGETS	:= main
+MAKE_DEPS		+= $(SPARSE_SRC)/libsparse.a
+HOST_CFLAGS		+= -I$(SPARSE_SRC)
+HOST_LDLIBS		+= $(SPARSE_SRC)/libsparse.a
+
+
+include $(top_srcdir)/include/mk/generic_leaf_target.mk
tools/sparse/README.md new file +34 -0
new file mode 100644
index 000000000..368cd7d99
@@ -0,0 +1,34 @@
+# Sparse based linting
+
+This tool checks LTP test and library code for common problems.
+
+## Building
+
+The bad news is you must get and build Sparse[^1]. The good news this only
+takes a minute.
+
+If you already have the Sparse source code, then do the
+following. Where `$SRC_PATH` is the path to the Sparse directory.
+
+```sh
+$ cd tools/sparse
+$ make -C $SRC_PATH -j$(nproc) # Optional
+$ make SPARSE_SRC=$SRC_PATH
+```
+If not then you can fetch it via the git submodule
+
+```sh
+$ cd tools/sparse
+$ git submodule update --init
+$ cd sparse-src
+$ make -C sparse-src -j$(nproc)
+$ make
+```
+
+## Usage
+
+It is integrated with the LTP build system. Just run `make check` or
+`make check-a_test01`, where `a_test01` is an arbitrary test
+executable or object file.
+
+[1]: Many distributions have a Sparse package. This only contains some executables. There is no shared library
tools/sparse/main.c new file +148 -0

148 lines not on this page · raw message

tools/sparse/sparse-src new file +1 -0
new file mode 160000
index 000000000..8af243292
@@ -0,0 +1 @@
+Subproject commit 8af2432923486c753ab52cae70b94ee684121080
Signature · 1 line
2.31.1
#2 lore raw

PATCH 1/7 Add Sparse based checker and TST_RET/ERR check

Hi!

Quoted 2 lines

Vendors in Sparse as a git module. Then uses it to check for stores to TST_RET/ERR within the library.

This sounds reasonable enough.

It would probably be even easier to copy the sources to the LTP tree and synchronize them manually, depending on how often is sparse updated it may end up better than a submodule.

Also isn't the submodule frozen on a given commit anyways (unless we add git submodule --remote --merge into the build steps)?

Quoted 81 lines
---
 .gitmodules             |   3 +
 tools/sparse/.gitignore |   1 +
 tools/sparse/Makefile   |  20 ++++++
 tools/sparse/README.md  |  34 +++++++++
 tools/sparse/main.c     | 148 ++++++++++++++++++++++++++++++++++++++++
 tools/sparse/sparse-src |   1 +
 6 files changed, 207 insertions(+)
 create mode 100644 tools/sparse/.gitignore
 create mode 100644 tools/sparse/Makefile
 create mode 100644 tools/sparse/README.md
 create mode 100644 tools/sparse/main.c
 create mode 160000 tools/sparse/sparse-src
.gitmodules +3 -0
index 1c9e9c38a..a3c34af4b 100644
@@ -1,3 +1,6 @@
 [submodule "testcases/kernel/mce-test"]
 	path = testcases/kernel/mce-test
 	url = git://git.kernel.org/pub/scm/linux/kernel/git/gong.chen/mce-test.git
+[submodule "tools/sparse/sparse-src"]
+	path = tools/sparse/sparse-src
+	url = git://git.kernel.org/pub/scm/devel/sparse/sparse.git
tools/sparse/.gitignore new file +1 -0
new file mode 100644
index 000000000..ba2906d06
@@ -0,0 +1 @@
+main
tools/sparse/Makefile new file +20 -0
new file mode 100644
index 000000000..cf4ccdb60
@@ -0,0 +1,20 @@
+# SPDX-License-Identifier: GPL-2.0-or-later
+# Copyright (c) 2021 SUSE LLC <[email protected]>
+
+top_srcdir		?= ../..
+
+include $(top_srcdir)/include/mk/env_pre.mk
+include $(top_srcdir)/include/mk/functions.mk
+
+SPARSE_SRC	?= sparse-src
+
+$(SPARSE_SRC)/libsparse.a:
+	$(error "You need to build Sparse. See tools/sparse/README.md")
+
+HOST_MAKE_TARGETS	:= main
+MAKE_DEPS		+= $(SPARSE_SRC)/libsparse.a
+HOST_CFLAGS		+= -I$(SPARSE_SRC)
+HOST_LDLIBS		+= $(SPARSE_SRC)/libsparse.a
+
+
+include $(top_srcdir)/include/mk/generic_leaf_target.mk
tools/sparse/README.md new file +16 -0
new file mode 100644
index 000000000..368cd7d99
@@ -0,0 +1,34 @@
+# Sparse based linting
+
+This tool checks LTP test and library code for common problems.
+
+## Building
+
+The bad news is you must get and build Sparse[^1]. The good news this only
+takes a minute.
+
+If you already have the Sparse source code, then do the
+following. Where `$SRC_PATH` is the path to the Sparse directory.
+
+```sh
+$ cd tools/sparse
+$ make -C $SRC_PATH -j$(nproc) # Optional
+$ make SPARSE_SRC=$SRC_PATH

tools/sparse/README.md line 26

Quoted 10 lines
tools/sparse/README.md new file +10 -0
+```
+If not then you can fetch it via the git submodule
+
+```sh
+$ cd tools/sparse
+$ git submodule update --init
+$ cd sparse-src
+$ make -C sparse-src -j$(nproc)
+$ make
+```

Can we add a build.sh script into the tools/sparse/ directory that will do exactly these steps?

Quoted 174 lines
tools/sparse/README.md new file +7 -0
+## Usage
+
+It is integrated with the LTP build system. Just run `make check` or
+`make check-a_test01`, where `a_test01` is an arbitrary test
+executable or object file.
+
+[1]: Many distributions have a Sparse package. This only contains some executables. There is no shared library
tools/sparse/main.c new file +148 -0
new file mode 100644
index 000000000..58f9a549c
@@ -0,0 +1,148 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2021 SUSE LLC <[email protected]>
+ *
+ * Sparse allows us to perform checks on the AST (struct symbol) or on
+ * a linearized representation. In the latter case we are given a set
+ * of entry points (functions) containing basic blocks of
+ * instructions.
+ *
+ * The basic blocks contain byte code in SSA form. This is similar to
+ * the the intermediate representation most compilers use during
+ * optimisation.
+ */
+#include <stdarg.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "lib.h"
+#include "allocate.h"
+#include "opcode.h"
+#include "token.h"
+#include "parse.h"
+#include "symbol.h"
+#include "expression.h"
+#include "linearize.h"
+
+/* The rules for test, library and tool code are different */
+enum ltp_tu_kind {
+	LTP_LIB,
+	LTP_OTHER,
+};
+
+static enum ltp_tu_kind tu_kind = LTP_OTHER;
+
+/* Check for LTP-002
+ *
+ * Inspects the destination symbol of each store instruction. If it is
+ * TST_RET or TST_ERR then emit a warning.
+ */
+static void check_lib_sets_TEST_vars(const struct instruction *insn)
+{
+	static struct ident *TST_RES_id, *TST_ERR_id;
+
+	if (!TST_RES_id) {
+		TST_RES_id = built_in_ident("TST_RET");
+		TST_ERR_id = built_in_ident("TST_ERR");
+	}
+
+	if (insn->opcode != OP_STORE)
+		return;
+	if (insn->src->ident != TST_RES_id &&
+	    insn->src->ident != TST_ERR_id)
+		return;
+
+	warning(insn->pos,
+		"LTP-002: Library should not write to TST_RET or TST_ERR");
+}
+
+static void do_basicblock_checks(struct basic_block *bb)
+{
+	struct instruction *insn;
+
+	FOR_EACH_PTR(bb->insns, insn) {
+		if (!bb_reachable(insn->bb))
+			continue;
+
+		if (tu_kind == LTP_LIB)
+			check_lib_sets_TEST_vars(insn);
+	} END_FOR_EACH_PTR(insn);
+}
+
+static void do_entrypoint_checks(struct entrypoint *ep)
+{
+	struct basic_block *bb;
+
+	FOR_EACH_PTR(ep->bbs, bb) {
+		do_basicblock_checks(bb);
+	} END_FOR_EACH_PTR(bb);
+}
+
+/* Compile the AST into a graph of basicblocks */
+static void process_symbols(struct symbol_list *list)
+{
+	struct symbol *sym;
+
+	FOR_EACH_PTR(list, sym) {
+		struct entrypoint *ep;
+
+		expand_symbol(sym);
+		ep = linearize_symbol(sym);
+		if (!ep || !ep->entry)
+			continue;
+
+		do_entrypoint_checks(ep);
+
+		if (dbg_entry)
+			show_entry(ep);
+	} END_FOR_EACH_PTR(sym);
+}
+
+static void collect_info_from_args(const int argc, char *const *const argv)
+{
+	int i;
+
+	for (i = 0; i < argc; i++) {
+		if (!strcmp("-DLTPLIB", argv[i])) {
+			tu_kind = LTP_LIB;
+		}
+	}
+}
+
+int main(int argc, char **argv)
+{
+	struct string_list *filelist = NULL;
+	char *file;
+
+	Waddress_space = 0;
+	Wbitwise = 0;
+	Wcast_truncate = 0;
+	Wcontext = 0;
+	Wdecl = 0;
+	Wexternal_function_has_definition = 0;
+	Wflexible_array_array = 0;
+	Wimplicit_int = 0;
+	Wint_to_pointer_cast = 0;
+	Wmemcpy_max_count = 0;
+	Wnon_pointer_null = 0;
+	Wone_bit_signed_bitfield = 0;
+	Woverride_init = 0;
+	Wpointer_to_int_cast = 0;
+	Wvla = 0;
+
+	do_output = 0;
+
+	collect_info_from_args(argc, argv);
+
+	process_symbols(sparse_initialize(argc, argv, &filelist));
+	FOR_EACH_PTR(filelist, file) {
+		process_symbols(sparse(file));
+	} END_FOR_EACH_PTR(file);
+
+	report_stats();
+	return 0;
+}
tools/sparse/sparse-src new file +1 -0
new file mode 160000
index 000000000..8af243292
@@ -0,0 +1 @@
+Subproject commit 8af2432923486c753ab52cae70b94ee684121080
Signature · 5 lines
2.31.1


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Signature · 2 lines
Cyril Hrubis
[email protected]
#3 lore raw

PATCH 1/7 Add Sparse based checker and TST_RET/ERR check

Hello Cyril,

Quoted 9 lines · Cyril Hrubis

Hi!

Quoted 2 lines

Vendors in Sparse as a git module. Then uses it to check for stores to TST_RET/ERR within the library.

This sounds reasonable enough.

It would probably be even easier to copy the sources to the LTP tree and synchronize them manually, depending on how often is sparse updated it may end up better than a submodule.

It is not updated often. Indeed Smatch did this, but I don't think that has helped adoption of either Smatch or Sparse.

Quoted 4 lines

Also isn't the submodule frozen on a given commit anyways (unless we add git submodule --remote --merge into the build steps)?

Yes. Submodules can be a pain, however I see a few advantages:

* You can easily upstream any changes made to the sub project. (no copy and paste)
* You get the full Git history without a seperate clone of upstream
* If someone already has upstream cloned they can just point the submodule towards it (not that it really matters with Sparse)

It is easy to get in a mess with submodules, but that is mainly just a concern for whoever is updating Sparse.

Quoted 86 lines
Quoted 86 lines
---
 .gitmodules             |   3 +
 tools/sparse/.gitignore |   1 +
 tools/sparse/Makefile   |  20 ++++++
 tools/sparse/README.md  |  34 +++++++++
 tools/sparse/main.c     | 148 ++++++++++++++++++++++++++++++++++++++++
 tools/sparse/sparse-src |   1 +
 6 files changed, 207 insertions(+)
 create mode 100644 tools/sparse/.gitignore
 create mode 100644 tools/sparse/Makefile
 create mode 100644 tools/sparse/README.md
 create mode 100644 tools/sparse/main.c
 create mode 160000 tools/sparse/sparse-src
.gitmodules +3 -0
index 1c9e9c38a..a3c34af4b 100644
@@ -1,3 +1,6 @@
 [submodule "testcases/kernel/mce-test"]
 	path = testcases/kernel/mce-test
 	url = git://git.kernel.org/pub/scm/linux/kernel/git/gong.chen/mce-test.git
+[submodule "tools/sparse/sparse-src"]
+	path = tools/sparse/sparse-src
+	url = git://git.kernel.org/pub/scm/devel/sparse/sparse.git
tools/sparse/.gitignore new file +1 -0
new file mode 100644
index 000000000..ba2906d06
@@ -0,0 +1 @@
+main
tools/sparse/Makefile new file +20 -0
new file mode 100644
index 000000000..cf4ccdb60
@@ -0,0 +1,20 @@
+# SPDX-License-Identifier: GPL-2.0-or-later
+# Copyright (c) 2021 SUSE LLC <[email protected]>
+
+top_srcdir		?= ../..
+
+include $(top_srcdir)/include/mk/env_pre.mk
+include $(top_srcdir)/include/mk/functions.mk
+
+SPARSE_SRC	?= sparse-src
+
+$(SPARSE_SRC)/libsparse.a:
+	$(error "You need to build Sparse. See tools/sparse/README.md")
+
+HOST_MAKE_TARGETS	:= main
+MAKE_DEPS		+= $(SPARSE_SRC)/libsparse.a
+HOST_CFLAGS		+= -I$(SPARSE_SRC)
+HOST_LDLIBS		+= $(SPARSE_SRC)/libsparse.a
+
+
+include $(top_srcdir)/include/mk/generic_leaf_target.mk
tools/sparse/README.md new file +21 -0
new file mode 100644
index 000000000..368cd7d99
@@ -0,0 +1,34 @@
+# Sparse based linting
+
+This tool checks LTP test and library code for common problems.
+
+## Building
+
+The bad news is you must get and build Sparse[^1]. The good news this only
+takes a minute.
+
+If you already have the Sparse source code, then do the
+following. Where `$SRC_PATH` is the path to the Sparse directory.
+
+```sh
+$ cd tools/sparse
+$ make -C $SRC_PATH -j$(nproc) # Optional
+$ make SPARSE_SRC=$SRC_PATH
+```
+If not then you can fetch it via the git submodule
+
+```sh
+$ cd tools/sparse
Quoted 8 lines

tools/sparse/README.md line 26

Quoted 5 lines
tools/sparse/README.md new file +5 -0
+$ git submodule update --init
+$ cd sparse-src
+$ make -C sparse-src -j$(nproc)
+$ make
+```

Can we add a build.sh script into the tools/sparse/ directory that will do exactly these steps?

+1

Quoted 175 lines
Quoted 175 lines
tools/sparse/README.md new file +7 -0
+## Usage
+
+It is integrated with the LTP build system. Just run `make check` or
+`make check-a_test01`, where `a_test01` is an arbitrary test
+executable or object file.
+
+[1]: Many distributions have a Sparse package. This only contains some executables. There is no shared library
tools/sparse/main.c new file +148 -0
new file mode 100644
index 000000000..58f9a549c
@@ -0,0 +1,148 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2021 SUSE LLC <[email protected]>
+ *
+ * Sparse allows us to perform checks on the AST (struct symbol) or on
+ * a linearized representation. In the latter case we are given a set
+ * of entry points (functions) containing basic blocks of
+ * instructions.
+ *
+ * The basic blocks contain byte code in SSA form. This is similar to
+ * the the intermediate representation most compilers use during
+ * optimisation.
+ */
+#include <stdarg.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+#include "lib.h"
+#include "allocate.h"
+#include "opcode.h"
+#include "token.h"
+#include "parse.h"
+#include "symbol.h"
+#include "expression.h"
+#include "linearize.h"
+
+/* The rules for test, library and tool code are different */
+enum ltp_tu_kind {
+	LTP_LIB,
+	LTP_OTHER,
+};
+
+static enum ltp_tu_kind tu_kind = LTP_OTHER;
+
+/* Check for LTP-002
+ *
+ * Inspects the destination symbol of each store instruction. If it is
+ * TST_RET or TST_ERR then emit a warning.
+ */
+static void check_lib_sets_TEST_vars(const struct instruction *insn)
+{
+	static struct ident *TST_RES_id, *TST_ERR_id;
+
+	if (!TST_RES_id) {
+		TST_RES_id = built_in_ident("TST_RET");
+		TST_ERR_id = built_in_ident("TST_ERR");
+	}
+
+	if (insn->opcode != OP_STORE)
+		return;
+	if (insn->src->ident != TST_RES_id &&
+	    insn->src->ident != TST_ERR_id)
+		return;
+
+	warning(insn->pos,
+		"LTP-002: Library should not write to TST_RET or TST_ERR");
+}
+
+static void do_basicblock_checks(struct basic_block *bb)
+{
+	struct instruction *insn;
+
+	FOR_EACH_PTR(bb->insns, insn) {
+		if (!bb_reachable(insn->bb))
+			continue;
+
+		if (tu_kind == LTP_LIB)
+			check_lib_sets_TEST_vars(insn);
+	} END_FOR_EACH_PTR(insn);
+}
+
+static void do_entrypoint_checks(struct entrypoint *ep)
+{
+	struct basic_block *bb;
+
+	FOR_EACH_PTR(ep->bbs, bb) {
+		do_basicblock_checks(bb);
+	} END_FOR_EACH_PTR(bb);
+}
+
+/* Compile the AST into a graph of basicblocks */
+static void process_symbols(struct symbol_list *list)
+{
+	struct symbol *sym;
+
+	FOR_EACH_PTR(list, sym) {
+		struct entrypoint *ep;
+
+		expand_symbol(sym);
+		ep = linearize_symbol(sym);
+		if (!ep || !ep->entry)
+			continue;
+
+		do_entrypoint_checks(ep);
+
+		if (dbg_entry)
+			show_entry(ep);
+	} END_FOR_EACH_PTR(sym);
+}
+
+static void collect_info_from_args(const int argc, char *const *const argv)
+{
+	int i;
+
+	for (i = 0; i < argc; i++) {
+		if (!strcmp("-DLTPLIB", argv[i])) {
+			tu_kind = LTP_LIB;
+		}
+	}
+}
+
+int main(int argc, char **argv)
+{
+	struct string_list *filelist = NULL;
+	char *file;
+
+	Waddress_space = 0;
+	Wbitwise = 0;
+	Wcast_truncate = 0;
+	Wcontext = 0;
+	Wdecl = 0;
+	Wexternal_function_has_definition = 0;
+	Wflexible_array_array = 0;
+	Wimplicit_int = 0;
+	Wint_to_pointer_cast = 0;
+	Wmemcpy_max_count = 0;
+	Wnon_pointer_null = 0;
+	Wone_bit_signed_bitfield = 0;
+	Woverride_init = 0;
+	Wpointer_to_int_cast = 0;
+	Wvla = 0;
+
+	do_output = 0;
+
+	collect_info_from_args(argc, argv);
+
+	process_symbols(sparse_initialize(argc, argv, &filelist));
+	FOR_EACH_PTR(filelist, file) {
+		process_symbols(sparse(file));
+	} END_FOR_EACH_PTR(file);
+
+	report_stats();
+	return 0;
+}
tools/sparse/sparse-src new file +1 -0
new file mode 160000
index 000000000..8af243292
@@ -0,0 +1 @@
+Subproject commit 8af2432923486c753ab52cae70b94ee684121080
Signature · 5 lines
2.31.1


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Signature · 2 lines
Thank you,
Richard.
#4 lore raw

PATCH 2/7 Add 'make check' to the build system

In reply to Richard Palethorpe

Allows one to run 'make check' or 'make check-$TARGET'. Which will execute the CHECK tool with that target's CC arguments and CHECK_FLAGS. By default the check tool is tools/sparse/main.

---
 Makefile                                | 8 ++++++++
 include/mk/env_post.mk                  | 8 ++++++++
 include/mk/generic_leaf_target.inc      | 5 ++++-
 include/mk/generic_trunk_target.inc     | 7 ++++++-
 include/mk/lib.mk                       | 3 +++
 include/mk/module.mk                    | 2 ++
 include/mk/rules.mk                     | 9 +++++++++
 include/mk/sparse.mk                    | 9 +++++++++
 include/mk/testcases.mk                 | 1 +
 testcases/open_posix_testsuite/Makefile | 4 ++++
 10 files changed, 54 insertions(+), 2 deletions(-)
 create mode 100644 include/mk/sparse.mk
Makefile +8 -0
index 56812d77b..3b0ba330d 100644
@@ -79,6 +79,7 @@ BOOTSTRAP_TARGETS	:= $(sort $(COMMON_TARGETS) $(CLEAN_TARGETS) $(INSTALL_TARGETS
 CLEAN_TARGETS		:= $(addsuffix -clean,$(CLEAN_TARGETS))
 INSTALL_TARGETS		:= $(addsuffix -install,$(INSTALL_TARGETS))
 MAKE_TARGETS		:= $(addsuffix -all,$(filter-out lib,$(COMMON_TARGETS)))
+CHECK_TARGETS		:= $(addsuffix -check,testcases lib)
 
 # There's no reason why we should run `all' twice. Otherwise we're just wasting
 # 3+ mins of useful CPU cycles on a modern machine, and even more time on an
@@ -108,6 +109,10 @@ $(MAKE_TARGETS) include-all lib-all libs-all:
 	$(MAKE) -C "$(subst -all,,$@)" \
 		-f "$(abs_top_srcdir)/$(subst -all,,$@)/Makefile" all
 
+$(CHECK_TARGETS): tools-all
+	$(MAKE) -C "$(subst -check,,$@)" \
+		-f "$(abs_top_srcdir)/$(subst -check,,$@)/Makefile" check
+
 # Let's not conflict with ac-clean, maintainer-clean, etc, so.
 $(filter-out include-clean,$(CLEAN_TARGETS))::
 	-$(MAKE) -C "$(subst -clean,,$@)" \
@@ -189,6 +194,9 @@ INSTALL_TARGETS		+= $(addprefix $(DESTDIR)/$(bindir)/,$(BINDIR_INSTALL_SCRIPTS))
 
 $(INSTALL_TARGETS): $(INSTALL_DIR) $(DESTDIR)/$(bindir)
 
+.PHONY: check
+check: $(CHECK_TARGETS)
+
 ## Install
 install: $(INSTALL_TARGETS)
 
include/mk/env_post.mk +8 -0
index 1d22f9c53..0288c30b6 100644
@@ -89,6 +89,14 @@ $(error You must define $$(prefix) before executing install)
 endif # END $(filter-out install,$(MAKECMDGOALS)),$(MAKECMDGOALS)
 endif
 
+CHECK_TARGETS			?= $(addprefix check-,$(notdir $(patsubst %.c,%,$(sort $(wildcard $(abs_srcdir)/*.c)))))
+CHECK_TARGETS			:= $(filter-out $(addprefix check-, $(FILTER_OUT_MAKE_TARGETS)), $(CHECK_TARGETS))
+CHECK				?= $(abs_top_srcdir)/tools/sparse/main
+
+ifeq ($(dir $(CHECK)),$(abs_top_srcdir)/tools/sparse/main)
+CHECK_DEPS			+= $(CHECK)
+endif
+
 include $(top_srcdir)/include/mk/rules.mk
 
 endif
include/mk/generic_leaf_target.inc +4 -1
index 64953f89a..aa092a5a3 100644
@@ -92,7 +92,7 @@
 # INSTALL_DIR			:= $(libdir)
 #
 
-.PHONY: all clean install
+.PHONY: all clean install check
 
 ifneq ($(strip $(MAKE_TARGETS)),)
 $(MAKE_TARGETS) += $(HOST_MAKE_TARGETS)
@@ -109,4 +109,7 @@ $(INSTALL_FILES): | $(INSTALL_DEPS)
 
 install: $(INSTALL_FILES)
 
+$(CHECK_TARGETS): | $(CHECK_DEPS)
+check: $(CHECK_TARGETS)
+
 # vim: syntax=make
include/mk/generic_trunk_target.inc +6 -1
index fc59f944f..32a108fbf 100644
@@ -48,7 +48,7 @@
 
 include $(top_srcdir)/include/mk/functions.mk
 
-RECURSIVE_TARGETS		?= all install
+RECURSIVE_TARGETS		?= all install check
 
 $(eval $(get_make_dirs))
 
@@ -68,6 +68,9 @@ $(INSTALL_FILES): | $(INSTALL_DEPS)
 
 trunk-install: $(INSTALL_FILES)
 
+$(CHECK_TARGETS): | $(CHECK_DEPS)
+trunk-check: $(CHECK_TARGETS)
+
 # Avoid creating duplicate .PHONY references to all, clean, and install. IIRC,
 # I've seen some indeterministic behavior when one does this in the past with
 # GNU Make...
@@ -108,4 +111,6 @@ else
 endif
 endif
 
+check: trunk-check
+
 # vim: syntax=make
include/mk/lib.mk +3 -0
index f9b6c0aff..3bf63bf9e 100644
@@ -26,6 +26,7 @@
 # Makefile to include for libraries.
 
 include $(top_srcdir)/include/mk/env_pre.mk
+include $(top_srcdir)/include/mk/sparse.mk
 
 INSTALL_DIR	:= $(libdir)
 
@@ -57,6 +58,8 @@ LIBSRCS		:= $(filter-out $(FILTER_OUT_LIBSRCS),$(LIBSRCS))
 
 LIBOBJS		:= $(LIBSRCS:.c=.o)
 
+CHECK_TARGETS	:= $(addprefix check-,$(notdir $(LIBSRCS:.c=)))
+
 $(LIB): $(notdir $(LIBOBJS))
 	@if [ -z "$(strip $^)" ] ; then \
 		echo "Cowardly refusing to create empty archive"; \
include/mk/module.mk +2 -0
index 6c8814b96..3bb7350f1 100644
@@ -47,6 +47,8 @@ endif
 
 CLEAN_TARGETS += .dep_modules *.mod built-in.a
 
+CHECK_TARGETS := $(filter-out %.ko, $(CHECK_TARGETS))
+
 MODULE_SOURCES := $(patsubst %.ko,%.c,$(filter %.ko, $(MAKE_TARGETS)))
 
 # Ignoring the exit status of commands is done to be forward compatible with
include/mk/rules.mk +9 -0
index c8f4bbbbe..2a04b2b67 100644
@@ -37,3 +37,12 @@ else
 	@$(CC) $(CPPFLAGS) $(CFLAGS) $(LDFLAGS) $^ $(LTPLDLIBS) $(LDLIBS) -o $@
 	@echo CC $(target_rel_dir)$@
 endif
+
+.PHONY: $(CHECK_TARGETS)
+$(CHECK_TARGETS): check-%: %.c
+ifdef VERBOSE
+	$(CHECK) $(CHECK_FLAGS) $(CPPFLAGS) $(CFLAGS) $<
+else
+	@$(CHECK) $(CHECK_FLAGS) $(CPPFLAGS) $(CFLAGS) $<
+	@echo CHECK $(target_rel_dir)$<
+endif
include/mk/sparse.mk new file +9 -0
new file mode 100644
index 000000000..66711d107
@@ -0,0 +1,9 @@
+# Rules to make sparse tool(s) for inclusion in lib and testcases Makefiles
+
+SPARSE_DIR:= $(abs_top_builddir)/tools/sparse
+
+$(SPARSE_DIR)/main: $(SPARSE_DIR)
+	$(MAKE) -C "$^" all
+
+$(SPARSE_DIR): %:
+	mkdir -p "$@"
include/mk/testcases.mk +1 -0
index 1c81773d0..444020f16 100644
@@ -22,6 +22,7 @@
 
 include $(top_srcdir)/include/mk/env_pre.mk
 include $(top_srcdir)/include/mk/functions.mk
+include $(top_srcdir)/include/mk/sparse.mk
 
 APICMDS_DIR	:= $(abs_top_builddir)/tools/apicmds
 
testcases/open_posix_testsuite/Makefile +4 -0
index 205ecdc00..fea6db14b 100644
@@ -104,3 +104,7 @@ $(CRITICAL_MAKEFILE): \
 	$(top_srcdir)/LDFLAGS			\
 	$(top_srcdir)/LDLIBS
 	@$(MAKE) generate-makefiles
+
+.PHONY: check
+check:
+	@echo "Checker not yet supported by Open POSIX testsuite"
Signature · 1 line
2.31.1
#5 lore raw

PATCH 3/7 meltdown: Set CFLAGS for check targets

In reply to Richard Palethorpe

The include set in the CFLAGS is necessary for Sparse to parse the test source. If a specific target has variables set on it. Then these won't be available to the corresponding check-target. Usually this is not a problem because most variables set on a target are just for linking and the checker doesn't need them.

---
 testcases/cve/Makefile | 1 +
 1 file changed, 1 insertion(+)
testcases/cve/Makefile +1 -0
index ddf8b6fe1..7b011b6c9 100644
@@ -19,6 +19,7 @@ cve-2014-0196:  LDLIBS += -lutil
 endif
 
 meltdown: CFLAGS += -I$(abs_srcdir)/../realtime/include
+check-meltdown: CFLAGS += -I$(abs_srcdir)/../realtime/include
 
 ifneq (,$(filter $(HOST_CPU),x86 x86_64))
 meltdown: CFLAGS += -msse2
Signature · 1 line
2.31.1
#6 lore raw

PATCH 3/7 meltdown: Set CFLAGS for check targets

Hi!

Quoted 16 lines

The include set in the CFLAGS is necessary for Sparse to parse the test source. If a specific target has variables set on it. Then these won't be available to the corresponding check-target. Usually this is not a problem because most variables set on a target are just for linking and the checker doesn't need them.

---
 testcases/cve/Makefile | 1 +
 1 file changed, 1 insertion(+)
testcases/cve/Makefile +0 -0
index ddf8b6fe1..7b011b6c9 100644

testcases/cve/Makefile cve-2014-0196: LDLIBS += -lutil · line 22

Quoted 4 lines
testcases/cve/Makefile +1 -0
@@ -19,6 +19,7 @@ cve-2014-0196:  LDLIBS += -lutil
 endif
 
 meltdown: CFLAGS += -I$(abs_srcdir)/../realtime/include
+check-meltdown: CFLAGS += -I$(abs_srcdir)/../realtime/include

Uff, this will likely bite us in a different ways later on.

I guess that it would be cleaner to move the libtsc.h into top level include so that it's clear that the header is used in several tests.

Signature · 2 lines
Cyril Hrubis
[email protected]
#7 lore raw

PATCH 3/7 meltdown: Set CFLAGS for check targets

Hello,

Quoted 19 lines · Cyril Hrubis

Hi!

Quoted 18 lines

The include set in the CFLAGS is necessary for Sparse to parse the test source. If a specific target has variables set on it. Then these won't be available to the corresponding check-target. Usually this is not a problem because most variables set on a target are just for linking and the checker doesn't need them.

---
 testcases/cve/Makefile | 1 +
 1 file changed, 1 insertion(+)
testcases/cve/Makefile +0 -0
index ddf8b6fe1..7b011b6c9 100644
@@ -19,6 +19,7 @@ cve-2014-0196:  LDLIBS += -lutil
 endif
 
Quoted 7 lines

testcases/cve/Makefile cve-2014-0196: LDLIBS += -lutil · line 22

Quoted 2 lines
testcases/cve/Makefile +1 -0
 meltdown: CFLAGS += -I$(abs_srcdir)/../realtime/include
+check-meltdown: CFLAGS += -I$(abs_srcdir)/../realtime/include

Uff, this will likely bite us in a different ways later on.

I guess that it would be cleaner to move the libtsc.h into top level include so that it's clear that the header is used in several tests.

+1

Signature · 2 lines
Thank you,
Richard.
#8 lore raw

PATCH 4/7 doc: Add rules and recommendations list

In reply to Richard Palethorpe

Create a list of rules which are feasible to check automatically. The file is a table of tab separated values. It's intended to be machine readable.

For now there is just an ID column, with IDs that look similar to CWE IDs or CVEs. For now these can just be used with 'git grep'. Also there is a description column.

Clearly this is not an exhaustive list. It just contains a library rule already stated in the guide and the issue Sparse checks for.

---
 doc/rules.tsv                   | 3 +++
 doc/test-writing-guidelines.txt | 5 +++++
 2 files changed, 8 insertions(+)
 create mode 100644 doc/rules.tsv
doc/rules.tsv new file +3 -0
new file mode 100644
index 000000000..d4081ce0f
@@ -0,0 +1,3 @@
+ID	DESCRIPTION
+LTP-001	Library source files have tst_ prefix
+LTP-002	TST_RET and TST_ERR are never modified by test library functions
doc/test-writing-guidelines.txt +5 -0
index 37624eb7f..1bf00fdf7 100644
@@ -10,6 +10,11 @@ NOTE: See also
       https://github.com/linux-test-project/ltp/wiki/Shell-Test-API[Shell Test API],
       https://github.com/linux-test-project/ltp/wiki/LTP-Library-API-Writing-Guidelines[LTP Library API Writing Guidelines].
 
+Rules and recommendations which are "machine checkable" should be
+tagged with an ID like +LTP-XXX+. There will be a corresponding entry
+in 'doc/rules.tsv'. When you run 'make check' or 'make check-test' it
+will display these IDs as a reference.
+
 1. Guide to clean and understandable code
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
Signature · 1 line
2.31.1
#10 lore raw

PATCH 5/7 doc: Remind authors and maintainers to run make check

In reply to Richard Palethorpe
---
 doc/maintainer-patch-review-checklist.txt | 2 +-
 doc/test-writing-guidelines.txt           | 1 +
 2 files changed, 2 insertions(+), 1 deletion(-)
doc/maintainer-patch-review-checklist.txt +1 -1
index f6682b574..970779acc 100644
@@ -39,7 +39,7 @@ New test should
 * Use new https://github.com/linux-test-project/ltp/wiki/Test-Writing-Guidelines#22-writing-a-test-in-c[C API]
 * Test binaries are added into corresponding '.gitignore' files
 * Check coding style with
-  https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/plain/scripts/checkpatch.pl[checkpatch.pl]
+  https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/plain/scripts/checkpatch.pl[checkpatch.pl] and `make check`
   (more in https://github.com/linux-test-project/ltp/wiki/Test-Writing-Guidelines#131-c-coding-style[C coding style])
 * Docparse documentation
 * If a test is a regression test it should include tags
doc/test-writing-guidelines.txt +1 -0
index 1bf00fdf7..b6ccfdbbd 100644
@@ -308,6 +308,7 @@ NOTE: See also
 3. The runtest entires are in place
 4. Test binaries are added into corresponding '.gitignore' files
 5. Patches apply over the latest git
+6. 'make check' does not emit any warnings
 
 6.1 About .gitignore files
 ~~~~~~~~~~~~~~~~~~~~~~~~~~
Signature · 1 line
2.31.1
#11 lore raw

PATCH 5/7 doc: Remind authors and maintainers to run make check

Hi!
Looks obviously OK if we merge the rest.

Signature · 2 lines
Cyril Hrubis
[email protected]
#12 lore raw

PATCH 6/7 doc: Document TEST macro and state TST_RET/ERR rule LTP-002

In reply to Richard Palethorpe

There are cases where these variables can be safely used by the library. However it is a difficult problem to identify these cases automatically.

Identifying any library code which uses them is relatively easy. In fact it is easier to automate it than by code review. Because it is such a boring thing to repeatedly look for and comment on.

If a test library function needs these variables it can recreate its own private copies.

---
 doc/c-test-api.txt                     | 47 ++++++++++++++++++++++++++
 doc/library-api-writing-guidelines.txt | 14 ++++++++
 2 files changed, 61 insertions(+)
doc/c-test-api.txt +47 -0
index 4cccbdc10..bf41c8a68 100644
@@ -754,6 +754,53 @@ LTP_ALIGN(x, a)
 
 Aligns the x to be next multiple of a. The a must be power of 2.
 
+[source,c]
+-------------------------------------------------------------------------------
+TEST(socket(AF_INET, SOCK_RAW, 1));
+if (TST_RET > -1) {
+	tst_res(TFAIL, "Created raw socket");
+	SAFE_CLOSE(TST_RET);
+} else if (TST_ERR != EPERM) {
+	tst_res(TFAIL | TTERRNO,
+		"Failed to create socket for wrong reason");
+} else {
+	tst_res(TPASS | TTERRNO, "Didn't create raw socket");
+}
+-------------------------------------------------------------------------------
+
+The +TEST+ macro sets +TST_RET+ to its argument's return value and
++TST_ERR+ to +errno+. The +TTERNO+ flag can be used to print the error
+number's symbolic value.
+
+No LTP library function or macro, except those in 'tst_test_macros.h',
+will write to these variables (rule 'LTP-002'). So their values will
+not be changed unexpectedly.
+
+[source,c]
+-------------------------------------------------------------------------------
+TST_EXP_POSITIVE(wait(&status));
+
+if (!TST_PASS)
+	return;
+-------------------------------------------------------------------------------
+
+If the return value of 'wait' is positive. This macro will print a
+pass result and set +TST_PASS+ appropriately. If the return value is
+zero or negative, then it will print fail.
+
+This and similar macros take optional variadic arguments. These begin
+with a format string and then appropriate values to be formatted.
+
+[source,c]
+-------------------------------------------------------------------------------
+TST_EXP_PASS(chown("a/file", uid, gid), "chown(%s,%d,%d)",
+	     "a/file", uid, gid);
+-------------------------------------------------------------------------------
+
+Expects +chown+ to return 0 and emits a pass or a fail. The arguments
+to +chown+ will be printed in either case. There are many similar
+macros, please see 'tst_test_macros.h'.
+
 1.13 Filesystem type detection and skiplist
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
doc/library-api-writing-guidelines.txt +14 -0
index a4393fc54..2819d4177 100644
@@ -21,10 +21,24 @@ Don't forget to update docs when you change the API.
 2. C API
 --------
 
+2.1 LTP-001: Sources have tst_ prefix
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
 API source code is in headers `include/*.h`, `include/lapi/*.h` (backward
 compatibility for old kernel and libc) and C sources in `lib/*.c`. Files have
 'tst_' prefix.
 
+2.2 LTP-002: TST_RET and TST_ERR are not modified
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The test author is guaranteed that the test API will not modify these
+variables. This prevents silent errors where the return value and
+errno are overwritten before the test has chance to check them.
+
+The macros which are clearly intended to update these variables. That
+is +TEST+ and those in 'tst_test_macros.h'. Are of course allowed to
+update these variables.
+
 3. Shell API
 ------------
 
Signature · 1 line
2.31.1
#13 lore raw

PATCH 6/7 doc: Document TEST macro and state TST_RET/ERR rule LTP-002

Quoted 37 lines · Richard Palethorpe via ltp

+[source,c]
+-------------------------------------------------------------------------------

doc/c-test-api.txt +34 -0
+TEST(socket(AF_INET, SOCK_RAW, 1));
+if (TST_RET > -1) {
+	tst_res(TFAIL, "Created raw socket");
+	SAFE_CLOSE(TST_RET);
+} else if (TST_ERR != EPERM) {
+	tst_res(TFAIL | TTERRNO,
+		"Failed to create socket for wrong reason");
+} else {
+	tst_res(TPASS | TTERRNO, "Didn't create raw socket");
+}
+-------------------------------------------------------------------------------
+
+The +TEST+ macro sets +TST_RET+ to its argument's return value and
++TST_ERR+ to +errno+. The +TTERNO+ flag can be used to print the error
+number's symbolic value.
+
+No LTP library function or macro, except those in 'tst_test_macros.h',
+will write to these variables (rule 'LTP-002'). So their values will
+not be changed unexpectedly.
+
+[source,c]
+-------------------------------------------------------------------------------
+TST_EXP_POSITIVE(wait(&status));
+
+if (!TST_PASS)
+	return;
+-------------------------------------------------------------------------------
+
+If the return value of 'wait' is positive. This macro will print a
+pass result and set +TST_PASS+ appropriately. If the return value is
+zero or negative, then it will print fail.
+
+This and similar macros take optional variadic arguments. These begin
+with a format string and then appropriate values to be formatted.

doc/c-test-api.txt LTP_ALIGN(x, a) · line 802

Quoted 10 lines
doc/c-test-api.txt +10 -0
+
+[source,c]
+-------------------------------------------------------------------------------
+TST_EXP_PASS(chown("a/file", uid, gid), "chown(%s,%d,%d)",
+	     "a/file", uid, gid);
+-------------------------------------------------------------------------------
+
+Expects +chown+ to return 0 and emits a pass or a fail. The arguments
+to +chown+ will be printed in either case. There are many similar
+macros, please see 'tst_test_macros.h'.

It appears that TST_EXP* macros (but not TEST) are documented elsewhere. So I should probably combine these comments with the existing docs.

Signature · 2 lines
Thank you,
Richard.
#14 lore raw

PATCH 7/7 Reference LTP-002 rule in Cocci scripts

In reply to Richard Palethorpe

Avoid duplication of rule description

---
 scripts/coccinelle/libltp-test-macro-vars.cocci | 6 +-----
 scripts/coccinelle/libltp-test-macro.cocci      | 4 +---
 2 files changed, 2 insertions(+), 8 deletions(-)
scripts/coccinelle/libltp-test-macro-vars.cocci +1 -5
index ed5459a48..e0fe4e2a7 100644
@@ -1,11 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 // Copyright (c) 2021 SUSE LLC  <[email protected]>
 
-// The TEST macro should not be used in the library because it sets
-// TST_RET and TST_ERR which are global variables. The test author
-// only expects these to be changed if *they* call TEST directly.
-
-// Find all positions where TEST's variables are used
+// Find violations of LTP-002
 @ find_use exists @
 expression E;
 @@
scripts/coccinelle/libltp-test-macro.cocci +1 -3
index 7563d23aa..86c55c6c4 100644
@@ -1,9 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 // Copyright (c) 2021 SUSE LLC  <[email protected]>
 
-// The TEST macro should not be used in the library because it sets
-// TST_RET and TST_ERR which are global variables. The test author
-// only expects these to be changed if *they* call TEST directly.
+// Find and fix violations of rule LTP-002
 
 // Set with -D fix
 virtual fix
Signature · 1 line
2.31.1