2013-03-04 10:50:31

by Andrei Vagin

[permalink] [raw]
Subject: [PATCH 1/2] ptrace: add ability to retrieve signals without removing from a queue (v4)

This patch adds a new ptrace request PTRACE_PEEKSIGINFO.

This request is used to retrieve information about pending signals
starting with the specified sequence number. Siginfo_t structures are
copied from the child into the buffer starting at "data".

The argument "addr" is a pointer to struct ptrace_peeksiginfo_args.
struct ptrace_peeksiginfo_args {
u64 off; /* from which siginfo to start */
u32 flags;
s32 nr; /* how may siginfos to take */
};

"nr" has type "s32", because ptrace() returns "long", which has 32 bits
on i386 and a negative values is used for errors.

Currently here is only one flag PTRACE_PEEKSIGINFO_SHARED for dumping
signals from process-wide queue. If this flag is not set, signals are
read from a per-thread queue.

The request PTRACE_PEEKSIGINFO returns a number of dumped signals. If a
signal with the specified sequence number doesn't exist, ptrace returns
zero. The request returns an error, if no signal has been dumped.

Errors:
EINVAL - one or more specified flags are not supported or nr is negative
EFAULT - buf or addr is outside your accessible address space.

A result siginfo contains a kernel part of si_code which usually
striped, but it's required for queuing the same siginfo back during
restore of pending signals.

This functionality is required for checkpointing pending signals.
Pedro Alves suggested using it in "gdb" to peek at pending signals.
gdb already uses PTRACE_GETSIGINFO to get the siginfo for the signal
which was already dequeued. This functionality allows gdb to look at
the pending signals which were not reported yet.

The prototype of this code was developed by Oleg Nesterov.

v2: * don't wrapped on CONFIG_CHECKPOINT_RESTORE. This functionality is
going to be used in gdb.
* use ptrace_peeksiginfo_args, because addr is too small for
encoding a signal number and flags.
v3: incorporate comments received from Andrew Morton
v4: make code a bit more obvious

Cc: Roland McGrath <[email protected]>
Cc: Oleg Nesterov <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: "Paul E. McKenney" <[email protected]>
Cc: David Howells <[email protected]>
Cc: Dave Jones <[email protected]>
Cc: "Michael Kerrisk (man-pages)" <[email protected]>
Cc: Pavel Emelyanov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Pedro Alves <[email protected]>
Signed-off-by: Andrey Vagin <[email protected]>
Signed-off-by: Andrew Vagin <[email protected]>
---
include/uapi/linux/ptrace.h | 12 +++++++
kernel/ptrace.c | 80 +++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 92 insertions(+)

diff --git a/include/uapi/linux/ptrace.h b/include/uapi/linux/ptrace.h
index 022ab18..52ebcc8 100644
--- a/include/uapi/linux/ptrace.h
+++ b/include/uapi/linux/ptrace.h
@@ -5,6 +5,7 @@

/* has the defines to get at the registers. */

+#include <linux/types.h>

#define PTRACE_TRACEME 0
#define PTRACE_PEEKTEXT 1
@@ -52,6 +53,17 @@
#define PTRACE_INTERRUPT 0x4207
#define PTRACE_LISTEN 0x4208

+#define PTRACE_PEEKSIGINFO 0x4209
+
+struct ptrace_peeksiginfo_args {
+ __u64 off; /* from which siginfo to start */
+ __u32 flags;
+ __s32 nr; /* how may siginfos to take */
+};
+
+/* Read signals from a shared (process wide) queue */
+#define PTRACE_PEEKSIGINFO_SHARED (1 << 0)
+
/* Wait extended result codes for the above trace options. */
#define PTRACE_EVENT_FORK 1
#define PTRACE_EVENT_VFORK 2
diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index acbd284..17ae54d 100644
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -24,6 +24,7 @@
#include <linux/regset.h>
#include <linux/hw_breakpoint.h>
#include <linux/cn_proc.h>
+#include <linux/compat.h>


static int ptrace_trapping_sleep_fn(void *flags)
@@ -618,6 +619,81 @@ static int ptrace_setsiginfo(struct task_struct *child, const siginfo_t *info)
return error;
}

+static int ptrace_peek_siginfo(struct task_struct *child,
+ unsigned long addr,
+ unsigned long data)
+{
+ struct ptrace_peeksiginfo_args arg;
+ struct sigpending *pending;
+ struct sigqueue *q;
+ int ret, i;
+
+ ret = copy_from_user(&arg, (void __user *) addr,
+ sizeof(struct ptrace_peeksiginfo_args));
+ if (ret)
+ return -EFAULT;
+
+ if (arg.flags & ~PTRACE_PEEKSIGINFO_SHARED)
+ return -EINVAL; /* unknown flags */
+
+ if (arg.nr < 0)
+ return -EINVAL;
+
+ if (arg.flags & PTRACE_PEEKSIGINFO_SHARED)
+ pending = &child->signal->shared_pending;
+ else
+ pending = &child->pending;
+
+ for (i = 0; i < arg.nr; ) {
+ siginfo_t info;
+ s32 off = arg.off + i;
+
+ spin_lock_irq(&child->sighand->siglock);
+ list_for_each_entry(q, &pending->list, list) {
+ if (!off--) {
+ copy_siginfo(&info, &q->info);
+ break;
+ }
+ }
+ spin_unlock_irq(&child->sighand->siglock);
+
+ if (off >= 0) /* beyond the end of the list */
+ break;
+
+#ifdef CONFIG_COMPAT
+ if (unlikely(is_compat_task())) {
+ compat_siginfo_t __user *uinfo = compat_ptr(data);
+
+ ret = copy_siginfo_to_user32(uinfo, &info);
+ ret |= __put_user(info.si_code, &uinfo->si_code);
+ } else
+#endif
+ {
+ siginfo_t __user *uinfo = (siginfo_t __user *) data;
+
+ ret = copy_siginfo_to_user(uinfo, &info);
+ ret |= __put_user(info.si_code, &uinfo->si_code);
+ }
+
+ if (ret) {
+ ret = -EFAULT;
+ break;
+ }
+
+ data += sizeof(siginfo_t);
+ i++;
+
+ if (signal_pending(current))
+ break;
+
+ cond_resched();
+ }
+
+ if (i > 0)
+ return i;
+
+ return ret;
+}

#ifdef PTRACE_SINGLESTEP
#define is_singlestep(request) ((request) == PTRACE_SINGLESTEP)
@@ -748,6 +824,10 @@ int ptrace_request(struct task_struct *child, long request,
ret = put_user(child->ptrace_message, datalp);
break;

+ case PTRACE_PEEKSIGINFO:
+ ret = ptrace_peek_siginfo(child, addr, data);
+ break;
+
case PTRACE_GETSIGINFO:
ret = ptrace_getsiginfo(child, &siginfo);
if (!ret)
--
1.7.11.7


2013-03-04 10:48:29

by Andrei Vagin

[permalink] [raw]
Subject: [PATCH 2/2] selftest: add a test case for PTRACE_PEEKSIGINFO

* Dump signals from process-wide and per-thread queues with
different sizes of buffers.
* Check error paths for buffers with restricted permissions. A part of
buffer or a whole buffer is for read-only.
* Try to get nonexistent signal.

Cc: Roland McGrath <[email protected]>
Cc: Oleg Nesterov <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: "Paul E. McKenney" <[email protected]>
Cc: David Howells <[email protected]>
Cc: Dave Jones <[email protected]>
Cc: "Michael Kerrisk (man-pages)" <[email protected]>
Cc: Pavel Emelyanov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Pedro Alves <[email protected]>
Signed-off-by: Andrey Vagin <[email protected]>
Signed-off-by: Andrew Vagin <[email protected]>
---
tools/testing/selftests/Makefile | 1 +
tools/testing/selftests/ptrace/Makefile | 10 ++
tools/testing/selftests/ptrace/peeksiginfo.c | 214 +++++++++++++++++++++++++++
3 files changed, 225 insertions(+)
create mode 100644 tools/testing/selftests/ptrace/Makefile
create mode 100644 tools/testing/selftests/ptrace/peeksiginfo.c

diff --git a/tools/testing/selftests/Makefile b/tools/testing/selftests/Makefile
index 3cc0ad7..575ef80 100644
--- a/tools/testing/selftests/Makefile
+++ b/tools/testing/selftests/Makefile
@@ -5,6 +5,7 @@ TARGETS += vm
TARGETS += cpu-hotplug
TARGETS += memory-hotplug
TARGETS += efivarfs
+TARGETS += ptrace

all:
for TARGET in $(TARGETS); do \
diff --git a/tools/testing/selftests/ptrace/Makefile b/tools/testing/selftests/ptrace/Makefile
new file mode 100644
index 0000000..47ae2d3
--- /dev/null
+++ b/tools/testing/selftests/ptrace/Makefile
@@ -0,0 +1,10 @@
+CFLAGS += -iquote../../../../include/uapi -Wall
+peeksiginfo: peeksiginfo.c
+
+all: peeksiginfo
+
+clean:
+ rm -f peeksiginfo
+
+run_tests: all
+ @./peeksiginfo || echo "peeksiginfo selftests: [FAIL]"
diff --git a/tools/testing/selftests/ptrace/peeksiginfo.c b/tools/testing/selftests/ptrace/peeksiginfo.c
new file mode 100644
index 0000000..d46558b
--- /dev/null
+++ b/tools/testing/selftests/ptrace/peeksiginfo.c
@@ -0,0 +1,214 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <signal.h>
+#include <unistd.h>
+#include <errno.h>
+#include <linux/types.h>
+#include <sys/wait.h>
+#include <sys/syscall.h>
+#include <sys/user.h>
+#include <sys/mman.h>
+
+#include "linux/ptrace.h"
+
+static int sys_rt_sigqueueinfo(pid_t tgid, int sig, siginfo_t *uinfo)
+{
+ return syscall(SYS_rt_sigqueueinfo, tgid, sig, uinfo);
+}
+
+static int sys_rt_tgsigqueueinfo(pid_t tgid, pid_t tid,
+ int sig, siginfo_t *uinfo)
+{
+ return syscall(SYS_rt_tgsigqueueinfo, tgid, tid, sig, uinfo);
+}
+
+static int sys_ptrace(int request, pid_t pid, void *addr, void *data)
+{
+ return syscall(SYS_ptrace, request, pid, addr, data);
+}
+
+#define SIGNR 10
+#define TEST_SICODE_PRIV -1
+#define TEST_SICODE_SHARE -2
+
+#define err(fmt, ...) \
+ fprintf(stderr, \
+ "Error (%s:%d): " fmt, \
+ __FILE__, __LINE__, ##__VA_ARGS__)
+
+static int check_error_paths(pid_t child)
+{
+ struct ptrace_peeksiginfo_args arg;
+ int ret, exit_code = -1;
+ void *addr_rw, *addr_ro;
+
+ /*
+ * Allocate two contiguous pages. The first one is for read-write,
+ * another is for read-only.
+ */
+ addr_rw = mmap(NULL, 2 * PAGE_SIZE, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+ if (addr_rw == MAP_FAILED) {
+ err("mmap() failed: %m\n");
+ return 1;
+ }
+
+ addr_ro = mmap(addr_rw + PAGE_SIZE, PAGE_SIZE, PROT_READ,
+ MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
+ if (addr_ro == MAP_FAILED) {
+ err("mmap() failed: %m\n");
+ goto out;
+ }
+
+ arg.nr = SIGNR;
+ arg.off = 0;
+
+ /* Unsupported flags */
+ arg.flags = ~0;
+ ret = sys_ptrace(PTRACE_PEEKSIGINFO, child, &arg, addr_rw);
+ if (ret != -1 || errno != EINVAL) {
+ err("sys_ptrace() returns %d (expected -1),"
+ " errno %d (expected %d): %m\n",
+ ret, errno, EINVAL);
+ goto out;
+ }
+ arg.flags = 0;
+
+ /* A part of the buffer is read-only */
+ ret = sys_ptrace(PTRACE_PEEKSIGINFO, child, &arg,
+ addr_ro - sizeof(siginfo_t) * 2);
+ if (ret != 2) {
+ err("sys_ptrace() returns %d (expected 2): %m\n", ret);
+ goto out;
+ }
+
+ /* Read-only buffer */
+ ret = sys_ptrace(PTRACE_PEEKSIGINFO, child, &arg, addr_ro);
+ if (ret != -1 && errno != EFAULT) {
+ err("sys_ptrace() returns %d (expected -1),"
+ " errno %d (expected %d): %m\n",
+ ret, errno, EFAULT);
+ goto out;
+ }
+
+ exit_code = 0;
+out:
+ munmap(addr_rw, 2 * PAGE_SIZE);
+ return exit_code;
+}
+
+int check_direct_path(pid_t child, int shared, int nr)
+{
+ struct ptrace_peeksiginfo_args arg = {.flags = 0, .nr = nr, .off = 0};
+ int i, j, ret, exit_code = -1;
+ siginfo_t siginfo[SIGNR];
+ int si_code;
+
+ if (shared == 1) {
+ arg.flags = PTRACE_PEEKSIGINFO_SHARED;
+ si_code = TEST_SICODE_SHARE;
+ } else {
+ arg.flags = 0;
+ si_code = TEST_SICODE_PRIV;
+ }
+
+ for (i = 0; i < SIGNR; ) {
+ arg.off = i;
+ ret = sys_ptrace(PTRACE_PEEKSIGINFO, child, &arg, siginfo);
+ if (ret == -1) {
+ err("ptrace() failed: %m\n");
+ goto out;
+ }
+
+ if (ret == 0)
+ break;
+
+ for (j = 0; j < ret; j++, i++) {
+ if (siginfo[j].si_code == si_code &&
+ siginfo[j].si_int == i)
+ continue;
+
+ err("%d: Wrong siginfo i=%d si_code=%d si_int=%d\n",
+ shared, i, siginfo[j].si_code, siginfo[j].si_int);
+ goto out;
+ }
+ }
+
+ if (i != SIGNR) {
+ err("Only %d signals were read\n", i);
+ goto out;
+ }
+
+ exit_code = 0;
+out:
+ return exit_code;
+}
+
+int main(int argc, char *argv[])
+{
+ siginfo_t siginfo[SIGNR];
+ int i, exit_code = 1;
+ sigset_t blockmask;
+ pid_t child;
+
+ sigemptyset(&blockmask);
+ sigaddset(&blockmask, SIGRTMIN);
+ sigprocmask(SIG_BLOCK, &blockmask, NULL);
+
+ child = fork();
+ if (child == -1) {
+ err("fork() failed: %m");
+ return 1;
+ } else if (child == 0) {
+ pid_t ppid = getppid();
+ while (1) {
+ if (ppid != getppid())
+ break;
+ sleep(1);
+ }
+ return 1;
+ }
+
+ /* Send signals in process-wide and per-thread queues */
+ for (i = 0; i < SIGNR; i++) {
+ siginfo->si_code = TEST_SICODE_SHARE;
+ siginfo->si_int = i;
+ sys_rt_sigqueueinfo(child, SIGRTMIN, siginfo);
+
+ siginfo->si_code = TEST_SICODE_PRIV;
+ siginfo->si_int = i;
+ sys_rt_tgsigqueueinfo(child, child, SIGRTMIN, siginfo);
+ }
+
+ if (sys_ptrace(PTRACE_ATTACH, child, NULL, NULL) == -1)
+ return 1;
+
+ waitpid(child, NULL, 0);
+
+ /* Dump signals one by one*/
+ if (check_direct_path(child, 0, 1))
+ goto out;
+ /* Dump all signals for one call */
+ if (check_direct_path(child, 0, SIGNR))
+ goto out;
+
+ /*
+ * Dump signal from the process-wide queue.
+ * The number of signals is not multible to the buffer size
+ */
+ if (check_direct_path(child, 1, 3))
+ goto out;
+
+ if (check_error_paths(child))
+ goto out;
+
+ printf("PASS\n");
+ exit_code = 0;
+out:
+ if (sys_ptrace(PTRACE_KILL, child, NULL, NULL) == -1)
+ return 1;
+
+ waitpid(child, NULL, 0);
+
+ return exit_code;
+}
--
1.7.11.7