Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1484730 > unrolled thread

[PATCH v2 00/11] syscall/tracing: compat syscall support

Back to article view | Back to linux.kernel

Contents

  [PATCH v2 00/11] syscall/tracing: compat syscall support Marcin Nowakowski <marcin.nowakowski@imgtec.com> - 2016-09-16 09:20 +0200
    [PATCH v2 01/11] tracing/syscalls: remove syscall_nr from syscall metadata Marcin Nowakowski <marcin.nowakowski@imgtec.com> - 2016-09-16 09:20 +0200
    [PATCH v2 07/11] arm64/tracing: fix compat syscall handling Marcin Nowakowski <marcin.nowakowski@imgtec.com> - 2016-09-16 09:20 +0200

#1484730 — [PATCH v2 00/11] syscall/tracing: compat syscall support

This patchset adds support syscall event tracing for compat syscalls.

Patch 1 removes the unnecessary syscall_nr field from syscall metadata,
which was one of the obstacles for adding proper support for compat syscalls.

Patch 2 adds a method to distinguish handling of syscalls for compat tasks
if an arch requires that. In disussions about an earlier version of these
patches (http://marc.info/?l=linux-mips&m=147259973128606&w=2) it was suggested
to use audit arch for detecting syscall type. After analysing the code
for various arches it seemed to me that this would add an unnecessary
complexity (as would require extra APIs to enumerate and map all audit
arch types) and I've just simply used compat task status to determine call
type. I cannot see any added value from using the audit arch type in this
context.

Patch 3 add compat syscall metadata - this is mostly a copy of a set of macros
for generating metadata for standard syscalls.

Further patches add arch-specific methods required for differentiating between
standard and compat syscalls as well as for finding syscall addresses from
inside syscall tables.

I have tried to keep the tracing system working for bisections - and most
things work as previously until arch-specific patch is applied. The only
exception here is x86 which had extra methods to prevent incorrect syscall
reporting for compat tasks - this may happen after patch 2 is applied and 
without x86/tracing patch.

Marcin Nowakowski (11):
  tracing/syscalls: remove syscall_nr from syscall metadata
  tracing/syscalls: add handling for compat tasks
  tracing/syscalls: add compat syscall metadata
  syscall/tracing: allow arch to override syscall_get_nr for ftrace
  x86/tracing: fix compat syscall handling
  s390/tracing: fix compat syscall handling
  arm64/tracing: fix compat syscall handling
  powerpc/tracing: fix compat syscall handling
  tile/tracing: fix compat syscall handling
  sparc/tracing: fix compat syscall handling
  parisc/tracing: fix compat syscall handling

 arch/arm64/include/asm/ftrace.h   |  12 +-
 arch/arm64/include/asm/unistd.h   |   1 +
 arch/arm64/kernel/Makefile        |   1 +
 arch/arm64/kernel/ftrace.c        |  16 +++
 arch/mips/kernel/ftrace.c         |   4 +-
 arch/parisc/include/asm/ftrace.h  |  10 ++
 arch/parisc/kernel/Makefile       |   1 +
 arch/parisc/kernel/ftrace.c       |  15 +++
 arch/powerpc/include/asm/ftrace.h |  26 +++-
 arch/powerpc/kernel/ftrace.c      |   6 +-
 arch/s390/include/asm/ftrace.h    |  11 ++
 arch/s390/include/asm/syscall.h   |   1 +
 arch/s390/kernel/Makefile         |   1 +
 arch/s390/kernel/ftrace.c         |  12 ++
 arch/sparc/include/asm/ftrace.h   |  10 ++
 arch/sparc/kernel/Makefile        |   1 +
 arch/sparc/kernel/ftrace.c        |  14 ++
 arch/tile/include/asm/ftrace.h    |  10 ++
 arch/tile/kernel/Makefile         |   1 +
 arch/tile/kernel/ftrace.c         |  13 ++
 arch/x86/include/asm/ftrace.h     |  14 +-
 arch/x86/include/asm/syscall.h    |   9 ++
 arch/x86/kernel/ftrace.c          |  15 +++
 include/linux/compat.h            |  74 +++++++++++
 include/linux/ftrace.h            |   2 +-
 include/linux/syscalls.h          |   1 -
 include/trace/syscall.h           |   2 -
 kernel/trace/trace.h              |  17 ++-
 kernel/trace/trace_syscalls.c     | 260 +++++++++++++++++++++++---------------
 29 files changed, 421 insertions(+), 139 deletions(-)

-- 
2.7.4

[toc] | [next] | [standalone]

#1484731 — [PATCH v2 01/11] tracing/syscalls: remove syscall_nr from syscall metadata

Some architectures map multiple syscall numbers to a single syscall.
This meant that on those platforms, some system calls could not be
properly traced using syscall event tracing mechanism, as a different
number of a syscall was used for registration to the one used by
applications.
We can use syscall lookup together with the syscall metadata table
traversal to register for appropriate events instead. This slightly
increases the overhead during event (un)registration, but does not
impact the trace events themselves, which still use syscall numbers
directly.

Signed-off-by: Marcin Nowakowski <marcin.nowakowski@imgtec.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@redhat.com>
---
 include/linux/syscalls.h      |   1 -
 include/trace/syscall.h       |   2 -
 kernel/trace/trace_syscalls.c | 127 ++++++++++++++++++++++++------------------
 3 files changed, 73 insertions(+), 57 deletions(-)

diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index d022390..c13aadd 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -160,7 +160,6 @@ extern struct trace_event_functions exit_syscall_print_funcs;
 	static struct syscall_metadata __used			\
 	  __syscall_meta_##sname = {				\
 		.name 		= "sys"#sname,			\
-		.syscall_nr	= -1,	/* Filled in at boot */	\
 		.nb_args 	= nb,				\
 		.types		= nb ? types_##sname : NULL,	\
 		.args		= nb ? args_##sname : NULL,	\
diff --git a/include/trace/syscall.h b/include/trace/syscall.h
index 7434f0f..b5fbebe 100644
--- a/include/trace/syscall.h
+++ b/include/trace/syscall.h
@@ -13,7 +13,6 @@
  * A syscall entry in the ftrace syscalls array.
  *
  * @name: name of the syscall
- * @syscall_nr: number of the syscall
  * @nb_args: number of parameters it takes
  * @types: list of types as strings
  * @args: list of args as strings (args[i] matches types[i])
@@ -23,7 +22,6 @@
  */
 struct syscall_metadata {
 	const char	*name;
-	int		syscall_nr;
 	int		nb_args;
 	const char	**types;
 	const char	**args;
diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c
index b2b6efc..86c3288 100644
--- a/kernel/trace/trace_syscalls.c
+++ b/kernel/trace/trace_syscalls.c
@@ -405,16 +405,21 @@ static int reg_event_syscall_enter(struct trace_event_file *file,
 	int ret = 0;
 	int num;
 
-	num = ((struct syscall_metadata *)call->data)->syscall_nr;
-	if (WARN_ON_ONCE(num < 0 || num >= NR_syscalls))
-		return -ENOSYS;
 	mutex_lock(&syscall_trace_lock);
-	if (!tr->sys_refcount_enter)
+	if (!tr->sys_refcount_enter) {
 		ret = register_trace_sys_enter(ftrace_syscall_enter, tr);
-	if (!ret) {
-		rcu_assign_pointer(tr->enter_syscall_files[num], file);
-		tr->sys_refcount_enter++;
+		if (ret)
+			goto out_unlock;
+	}
+
+	for (num = 0; num < NR_syscalls; num++) {
+		if (syscalls_metadata[num] &&
+		   (syscalls_metadata[num] == call->data))
+			rcu_assign_pointer(tr->enter_syscall_files[num], file);
 	}
+	tr->sys_refcount_enter++;
+
+out_unlock:
 	mutex_unlock(&syscall_trace_lock);
 	return ret;
 }
@@ -425,12 +430,13 @@ static void unreg_event_syscall_enter(struct trace_event_file *file,
 	struct trace_array *tr = file->tr;
 	int num;
 
-	num = ((struct syscall_metadata *)call->data)->syscall_nr;
-	if (WARN_ON_ONCE(num < 0 || num >= NR_syscalls))
-		return;
 	mutex_lock(&syscall_trace_lock);
 	tr->sys_refcount_enter--;
-	RCU_INIT_POINTER(tr->enter_syscall_files[num], NULL);
+	for (num = 0; num < NR_syscalls; num++) {
+		if (syscalls_metadata[num] &&
+		   (syscalls_metadata[num] == call->data))
+			RCU_INIT_POINTER(tr->enter_syscall_files[num], NULL);
+	}
 	if (!tr->sys_refcount_enter)
 		unregister_trace_sys_enter(ftrace_syscall_enter, tr);
 	mutex_unlock(&syscall_trace_lock);
@@ -443,16 +449,21 @@ static int reg_event_syscall_exit(struct trace_event_file *file,
 	int ret = 0;
 	int num;
 
-	num = ((struct syscall_metadata *)call->data)->syscall_nr;
-	if (WARN_ON_ONCE(num < 0 || num >= NR_syscalls))
-		return -ENOSYS;
 	mutex_lock(&syscall_trace_lock);
-	if (!tr->sys_refcount_exit)
-		ret = register_trace_sys_exit(ftrace_syscall_exit, tr);
-	if (!ret) {
-		rcu_assign_pointer(tr->exit_syscall_files[num], file);
-		tr->sys_refcount_exit++;
+	if (!tr->sys_refcount_exit) {
+		ret = register_trace_sys_enter(ftrace_syscall_exit, tr);
+		if (ret)
+			goto out_unlock;
 	}
+
+	for (num = 0; num < NR_syscalls; num++) {
+		if (syscalls_metadata[num] &&
+		   (syscalls_metadata[num] == call->data))
+			rcu_assign_pointer(tr->exit_syscall_files[num], file);
+	}
+	tr->sys_refcount_exit++;
+
+out_unlock:
 	mutex_unlock(&syscall_trace_lock);
 	return ret;
 }
@@ -463,12 +474,13 @@ static void unreg_event_syscall_exit(struct trace_event_file *file,
 	struct trace_array *tr = file->tr;
 	int num;
 
-	num = ((struct syscall_metadata *)call->data)->syscall_nr;
-	if (WARN_ON_ONCE(num < 0 || num >= NR_syscalls))
-		return;
 	mutex_lock(&syscall_trace_lock);
 	tr->sys_refcount_exit--;
-	RCU_INIT_POINTER(tr->exit_syscall_files[num], NULL);
+	for (num = 0; num < NR_syscalls; num++) {
+		if (syscalls_metadata[num] &&
+		   (syscalls_metadata[num] == call->data))
+			RCU_INIT_POINTER(tr->exit_syscall_files[num], NULL);
+	}
 	if (!tr->sys_refcount_exit)
 		unregister_trace_sys_exit(ftrace_syscall_exit, tr);
 	mutex_unlock(&syscall_trace_lock);
@@ -477,14 +489,6 @@ static void unreg_event_syscall_exit(struct trace_event_file *file,
 static int __init init_syscall_trace(struct trace_event_call *call)
 {
 	int id;
-	int num;
-
-	num = ((struct syscall_metadata *)call->data)->syscall_nr;
-	if (num < 0 || num >= NR_syscalls) {
-		pr_debug("syscall %s metadata not mapped, disabling ftrace event\n",
-				((struct syscall_metadata *)call->data)->name);
-		return -ENOSYS;
-	}
 
 	if (set_syscall_print_fmt(call) < 0)
 		return -ENOMEM;
@@ -547,7 +551,6 @@ void __init init_ftrace_syscalls(void)
 		if (!meta)
 			continue;
 
-		meta->syscall_nr = i;
 		syscalls_metadata[i] = meta;
 	}
 }
@@ -604,18 +607,24 @@ static int perf_sysenter_enable(struct trace_event_call *call)
 	int ret = 0;
 	int num;
 
-	num = ((struct syscall_metadata *)call->data)->syscall_nr;
-
 	mutex_lock(&syscall_trace_lock);
-	if (!sys_perf_refcount_enter)
+	if (!sys_perf_refcount_enter) {
 		ret = register_trace_sys_enter(perf_syscall_enter, NULL);
-	if (ret) {
-		pr_info("event trace: Could not activate"
+		if (ret) {
+			pr_info("event trace: Could not activate"
 				"syscall entry trace point");
-	} else {
-		set_bit(num, enabled_perf_enter_syscalls);
-		sys_perf_refcount_enter++;
+			goto out_unlock;
+		}
+	}
+
+	for (num = 0; num < NR_syscalls; num++) {
+		if (syscalls_metadata[num] &&
+		   (syscalls_metadata[num] == call->data))
+			set_bit(num, enabled_perf_enter_syscalls);
 	}
+	sys_perf_refcount_enter++;
+
+out_unlock:
 	mutex_unlock(&syscall_trace_lock);
 	return ret;
 }
@@ -624,11 +633,13 @@ static void perf_sysenter_disable(struct trace_event_call *call)
 {
 	int num;
 
-	num = ((struct syscall_metadata *)call->data)->syscall_nr;
-
 	mutex_lock(&syscall_trace_lock);
 	sys_perf_refcount_enter--;
-	clear_bit(num, enabled_perf_enter_syscalls);
+	for (num = 0; num < NR_syscalls; num++) {
+		if (syscalls_metadata[num] &&
+		   (syscalls_metadata[num] == call->data))
+			clear_bit(num, enabled_perf_enter_syscalls);
+	}
 	if (!sys_perf_refcount_enter)
 		unregister_trace_sys_enter(perf_syscall_enter, NULL);
 	mutex_unlock(&syscall_trace_lock);
@@ -676,18 +687,24 @@ static int perf_sysexit_enable(struct trace_event_call *call)
 	int ret = 0;
 	int num;
 
-	num = ((struct syscall_metadata *)call->data)->syscall_nr;
-
 	mutex_lock(&syscall_trace_lock);
-	if (!sys_perf_refcount_exit)
+	if (!sys_perf_refcount_exit) {
 		ret = register_trace_sys_exit(perf_syscall_exit, NULL);
-	if (ret) {
-		pr_info("event trace: Could not activate"
+		if (ret) {
+			pr_info("event trace: Could not activate"
 				"syscall exit trace point");
-	} else {
-		set_bit(num, enabled_perf_exit_syscalls);
-		sys_perf_refcount_exit++;
+			goto out_unlock;
+		}
+	}
+
+	for (num = 0; num < NR_syscalls; num++) {
+		if (syscalls_metadata[num] &&
+		   (syscalls_metadata[num] == call->data))
+			set_bit(num, enabled_perf_exit_syscalls);
 	}
+	sys_perf_refcount_exit++;
+
+out_unlock:
 	mutex_unlock(&syscall_trace_lock);
 	return ret;
 }
@@ -696,11 +713,13 @@ static void perf_sysexit_disable(struct trace_event_call *call)
 {
 	int num;
 
-	num = ((struct syscall_metadata *)call->data)->syscall_nr;
-
 	mutex_lock(&syscall_trace_lock);
 	sys_perf_refcount_exit--;
-	clear_bit(num, enabled_perf_exit_syscalls);
+	for (num = 0; num < NR_syscalls; num++) {
+		if (syscalls_metadata[num] &&
+		   (syscalls_metadata[num] == call->data))
+			clear_bit(num, enabled_perf_exit_syscalls);
+	}
 	if (!sys_perf_refcount_exit)
 		unregister_trace_sys_exit(perf_syscall_exit, NULL);
 	mutex_unlock(&syscall_trace_lock);
-- 
2.7.4

[toc] | [prev] | [next] | [standalone]

#1484732 — [PATCH v2 07/11] arm64/tracing: fix compat syscall handling

Add arch_syscall_addr for arm64 and define NR_compat_syscalls, as the
number of compat syscalls for arm64 exceeds the number defined by
NR_syscalls.

Signed-off-by: Marcin Nowakowski <marcin.nowakowski@imgtec.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
---
 arch/arm64/include/asm/ftrace.h | 12 +-----------
 arch/arm64/include/asm/unistd.h |  1 +
 arch/arm64/kernel/Makefile      |  1 +
 arch/arm64/kernel/ftrace.c      | 16 ++++++++++++++++
 4 files changed, 19 insertions(+), 11 deletions(-)

diff --git a/arch/arm64/include/asm/ftrace.h b/arch/arm64/include/asm/ftrace.h
index caa955f..b57ff7c 100644
--- a/arch/arm64/include/asm/ftrace.h
+++ b/arch/arm64/include/asm/ftrace.h
@@ -41,17 +41,7 @@ static inline unsigned long ftrace_call_adjust(unsigned long addr)
 
 #define ftrace_return_address(n) return_address(n)
 
-/*
- * Because AArch32 mode does not share the same syscall table with AArch64,
- * tracing compat syscalls may result in reporting bogus syscalls or even
- * hang-up, so just do not trace them.
- * See kernel/trace/trace_syscalls.c
- *
- * x86 code says:
- * If the user really wants these, then they should use the
- * raw syscall tracepoints with filtering.
- */
-#define ARCH_TRACE_IGNORE_COMPAT_SYSCALLS
+#define ARCH_COMPAT_SYSCALL_NUMBERS_OVERLAP 1
 static inline bool arch_trace_is_compat_syscall(struct pt_regs *regs)
 {
 	return is_compat_task();
diff --git a/arch/arm64/include/asm/unistd.h b/arch/arm64/include/asm/unistd.h
index e78ac26..276d049 100644
--- a/arch/arm64/include/asm/unistd.h
+++ b/arch/arm64/include/asm/unistd.h
@@ -45,6 +45,7 @@
 #define __ARM_NR_compat_set_tls		(__ARM_NR_COMPAT_BASE+5)
 
 #define __NR_compat_syscalls		394
+#define NR_compat_syscalls (__NR_compat_syscalls)
 #endif
 
 #define __ARCH_WANT_SYS_CLONE
diff --git a/arch/arm64/kernel/Makefile b/arch/arm64/kernel/Makefile
index 14f7b65..64a8c53 100644
--- a/arch/arm64/kernel/Makefile
+++ b/arch/arm64/kernel/Makefile
@@ -28,6 +28,7 @@ $(obj)/%.stub.o: $(obj)/%.o FORCE
 arm64-obj-$(CONFIG_COMPAT)		+= sys32.o kuser32.o signal32.o 	\
 					   sys_compat.o entry32.o
 arm64-obj-$(CONFIG_FUNCTION_TRACER)	+= ftrace.o entry-ftrace.o
+arm64-obj-$(CONFIG_FTRACE_SYSCALLS)	+= ftrace.o
 arm64-obj-$(CONFIG_MODULES)		+= arm64ksyms.o module.o
 arm64-obj-$(CONFIG_ARM64_MODULE_PLTS)	+= module-plts.o
 arm64-obj-$(CONFIG_PERF_EVENTS)		+= perf_regs.o perf_callchain.o
diff --git a/arch/arm64/kernel/ftrace.c b/arch/arm64/kernel/ftrace.c
index ebecf9a..53348d1 100644
--- a/arch/arm64/kernel/ftrace.c
+++ b/arch/arm64/kernel/ftrace.c
@@ -176,4 +176,20 @@ int ftrace_disable_ftrace_graph_caller(void)
 	return ftrace_modify_graph_caller(false);
 }
 #endif /* CONFIG_DYNAMIC_FTRACE */
+
 #endif /* CONFIG_FUNCTION_GRAPH_TRACER */
+
+#if (defined CONFIG_FTRACE_SYSCALLS) && (defined CONFIG_COMPAT)
+
+extern const void *sys_call_table[];
+extern const void *compat_sys_call_table[];
+
+unsigned long __init arch_syscall_addr(int nr, bool compat)
+{
+	if (compat)
+		return (unsigned long)compat_sys_call_table[nr];
+
+	return (unsigned long)sys_call_table[nr];
+}
+
+#endif /* CONFIG_FTRACE_SYSCALLS && CONFIG_COMPAT */
-- 
2.7.4

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web