Skip to contentSkip to conversation

some fixups for PTRACE_SECCOMP_GET_METADATA

PATCH In review Tycho Andersen 3 participants 6 messages Feb 21, 2018 – Feb 22, 2018

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover some fixups for PTRACE_SECCOMP_GET_METADATA

Hi Kees,

Here are a couple of tweaks/fixes people suggested to the get_metadata functionality, plus a test to ensure that things work the way they're supposed to and stay that way.

Cheers,

Tycho

Tycho Andersen (3):
seccomp, ptrace: switch get_metadata types to arch independent ptrace, seccomp: tweak get_metadata behavior slightly
seccomp: add a selftest for get_metadata

 include/uapi/linux/ptrace.h                   |  4 +-
 kernel/seccomp.c                              |  6 ++-
 tools/testing/selftests/seccomp/seccomp_bpf.c | 61 +++++++++++++++++++++++++++
 3 files changed, 67 insertions(+), 4 deletions(-)
Signature · 1 line
2.14.1
#1 lore raw

PATCH 1/3 seccomp, ptrace: switch get_metadata types to arch independent

Commit 26500475ac1b ("ptrace, seccomp: add support for retrieving seccomp metadata") introduced struct seccomp_metadata, which contained unsigned longs that should be arch independent. The type of the flags member was chosen to match the corresponding argument to seccomp(), and so we need something at least as big as unsigned long. My understanding is that __u64 should fit the bill, so let's switch both types to that.

While this is userspace facing, it was only introduced in 4.16-rc2, and so should be safe assuming it goes in before then.

---
 include/uapi/linux/ptrace.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
include/uapi/linux/ptrace.h +2 -2
index e46d82b91166..d5a1b8a492b9 100644
@@ -69,8 +69,8 @@ struct ptrace_peeksiginfo_args {
 #define PTRACE_SECCOMP_GET_METADATA	0x420d
 
 struct seccomp_metadata {
-	unsigned long filter_off;	/* Input: which filter */
-	unsigned int flags;		/* Output: filter's flags */
+	__u64 filter_off;	/* Input: which filter */
+	__u64 flags;		/* Output: filter's flags */
 };
 
 /* Read signals from a shared (process wide) queue */
Signature · 1 line
2.14.1
Reviewed-by on v1 1/3
#2 lore raw

Re: [PATCH 1/3] seccomp, ptrace: switch get_metadata types to arch independent

Quoted 23 lines · Tycho Andersen

Commit 26500475ac1b ("ptrace, seccomp: add support for retrieving seccomp metadata") introduced struct seccomp_metadata, which contained unsigned longs that should be arch independent. The type of the flags member was chosen to match the corresponding argument to seccomp(), and so we need something at least as big as unsigned long. My understanding is that __u64 should fit the bill, so let's switch both types to that.

While this is userspace facing, it was only introduced in 4.16-rc2, and so should be safe assuming it goes in before then.

---
 include/uapi/linux/ptrace.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
include/uapi/linux/ptrace.h +0 -0
index e46d82b91166..d5a1b8a492b9 100644

include/uapi/linux/ptrace.h struct ptrace_peeksiginfo_args { · line 76

Quoted 10 lines
include/uapi/linux/ptrace.h +2 -2
@@ -69,8 +69,8 @@ struct ptrace_peeksiginfo_args {
 #define PTRACE_SECCOMP_GET_METADATA	0x420d
 
 struct seccomp_metadata {
-	unsigned long filter_off;	/* Input: which filter */
-	unsigned int flags;		/* Output: filter's flags */
+	__u64 filter_off;	/* Input: which filter */
+	__u64 flags;		/* Output: filter's flags */
 };
 
 /* Read signals from a shared (process wide) queue */

That's much better, thanks.

Signature · 1 line
ldv
#3 lore raw

PATCH 2/3 ptrace, seccomp: tweak get_metadata behavior slightly

In reply to Tycho Andersen

Previously if users passed a small size for the input structure size, they would get get odd behavior. It doesn't make sense to pass a structure smaller than at least filter_off size, so let's just give -EINVAL in this case.

This changes userspace visible behavior, but was only introduced in commit 26500475ac1b ("ptrace, seccomp: add support for retrieving seccomp metadata") in 4.16-rc2, so should be safe to change if merged before then.

---
 kernel/seccomp.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
kernel/seccomp.c +4 -2
index 940fa408a288..dc77548167ef 100644
@@ -1076,14 +1076,16 @@ long seccomp_get_metadata(struct task_struct *task,
 
 	size = min_t(unsigned long, size, sizeof(kmd));
 
-	if (copy_from_user(&kmd, data, size))
+	if (size < sizeof(kmd.filter_off))
+		return -EINVAL;
+
+	if (copy_from_user(&kmd.filter_off, data, sizeof(kmd.filter_off)))
 		return -EFAULT;
 
 	filter = get_nth_filter(task, kmd.filter_off);
 	if (IS_ERR(filter))
 		return PTR_ERR(filter);
 
-	memset(&kmd, 0, sizeof(kmd));
 	if (filter->log)
 		kmd.flags |= SECCOMP_FILTER_FLAG_LOG;
 
Signature · 1 line
2.14.1
#4 lore raw

PATCH 3/3 seccomp: add a selftest for get_metadata

In reply to Tycho Andersen

Let's test that we get the flags correctly, and that we preserve the filter index across the ptrace(PTRACE_SECCOMP_GET_METADATA) correctly.

---
 tools/testing/selftests/seccomp/seccomp_bpf.c | 61 +++++++++++++++++++++++++++
 1 file changed, 61 insertions(+)
tools/testing/selftests/seccomp/seccomp_bpf.c +61 -0
index 0b457e8e0f0c..5df609950a66 100644
@@ -141,6 +141,15 @@ struct seccomp_data {
 #define SECCOMP_FILTER_FLAG_LOG 2
 #endif
 
+#ifndef PTRACE_SECCOMP_GET_METADATA
+#define PTRACE_SECCOMP_GET_METADATA	0x420d
+
+struct seccomp_metadata {
+	__u64 filter_off;       /* Input: which filter */
+	__u64 flags;             /* Output: filter's flags */
+};
+#endif
+
 #ifndef seccomp
 int seccomp(unsigned int op, unsigned int flags, void *args)
 {
@@ -2845,6 +2854,58 @@ TEST(get_action_avail)
 	EXPECT_EQ(errno, EOPNOTSUPP);
 }
 
+TEST(get_metadata)
+{
+	pid_t pid;
+	int pipefd[2];
+	char buf;
+	struct seccomp_metadata md;
+
+	ASSERT_EQ(0, pipe(pipefd));
+
+	pid = fork();
+	ASSERT_GE(pid, 0);
+	if (pid == 0) {
+		struct sock_filter filter[] = {
+			BPF_STMT(BPF_RET|BPF_K, SECCOMP_RET_ALLOW),
+		};
+		struct sock_fprog prog = {
+			.len = (unsigned short)ARRAY_SIZE(filter),
+			.filter = filter,
+		};
+
+		/* one with log, one without */
+		ASSERT_EQ(0, seccomp(SECCOMP_SET_MODE_FILTER,
+				     SECCOMP_FILTER_FLAG_LOG, &prog));
+		ASSERT_EQ(0, seccomp(SECCOMP_SET_MODE_FILTER, 0, &prog));
+
+		ASSERT_EQ(0, close(pipefd[0]));
+		ASSERT_EQ(1, write(pipefd[1], "1", 1));
+		ASSERT_EQ(0, close(pipefd[1]));
+
+		while (1)
+			sleep(100);
+	}
+
+	ASSERT_EQ(0, close(pipefd[1]));
+	ASSERT_EQ(1, read(pipefd[0], &buf, 1));
+
+	ASSERT_EQ(0, ptrace(PTRACE_ATTACH, pid));
+	ASSERT_EQ(pid, waitpid(pid, NULL, 0));
+
+	md.filter_off = 0;
+	ASSERT_EQ(sizeof(md), ptrace(PTRACE_SECCOMP_GET_METADATA, pid, sizeof(md), &md));
+	EXPECT_EQ(md.flags, SECCOMP_FILTER_FLAG_LOG);
+	EXPECT_EQ(md.filter_off, 0);
+
+	md.filter_off = 1;
+	ASSERT_EQ(sizeof(md), ptrace(PTRACE_SECCOMP_GET_METADATA, pid, sizeof(md), &md));
+	EXPECT_EQ(md.flags, 0);
+	EXPECT_EQ(md.filter_off, 1);
+
+	ASSERT_EQ(0, kill(pid, SIGKILL));
+}
+
 /*
  * TODO:
  * - add microbenchmarks
Signature · 1 line
2.14.1
#5 lore raw
In reply to Tycho Andersen
Quoted 6 lines · Tycho Andersen

Hi Kees,

Here are a couple of tweaks/fixes people suggested to the get_metadata functionality, plus a test to ensure that things work the way they're supposed to and stay that way.

Quoted 8 lines

Cheers,

Tycho

Tycho Andersen (3):
seccomp, ptrace: switch get_metadata types to arch independent ptrace, seccomp: tweak get_metadata behavior slightly
seccomp: add a selftest for get_metadata

Thanks! I've applied these and asked James to get them pull for 4.16.

-Kees

Signature · 2 lines
Kees Cook
Pixel Security