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

Groups > linux.kernel > #1587611

[PATCH -stable 4.1 1/2] perf: Restructure perf syscall point of no return

Show all headers | View raw

From: Peter Zijlstra <peterz@infradead.org>

[ Upstream commit f55fc2a57cc9ca3b1bb4fb8eb25b6e1989e5b993 ]

The exclusive_event_installable() stuff only works because its
exclusive with the grouping bits.

Rework the code such that there is a sane place to error out before we
go do things we cannot undo.

CVE 2017-6001

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Wang Nan <wangnan0@huawei.com>
---
 kernel/events/core.c | 49 ++++++++++++++++++++++++++++++++-----------------
 1 file changed, 32 insertions(+), 17 deletions(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index 6da64f0..77be116 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -8124,13 +8124,30 @@ SYSCALL_DEFINE5(perf_event_open,
 
 	if (move_group) {
 		gctx = group_leader->ctx;
+		mutex_lock_double(&gctx->mutex, &ctx->mutex);
+	} else {
+		mutex_lock(&ctx->mutex);
+	}
+
+	/*
+	 * Must be under the same ctx::mutex as perf_install_in_context(),
+	 * because we need to serialize with concurrent event creation.
+	 */
+	if (!exclusive_event_installable(event, ctx)) {
+		/* exclusive and group stuff are assumed mutually exclusive */
+		WARN_ON_ONCE(move_group);
+
+		err = -EBUSY;
+		goto err_locked;
+	}
 
+	WARN_ON_ONCE(ctx->parent_ctx);
+
+	if (move_group) {
 		/*
 		 * See perf_event_ctx_lock() for comments on the details
 		 * of swizzling perf_event::ctx.
 		 */
-		mutex_lock_double(&gctx->mutex, &ctx->mutex);
-
 		perf_remove_from_context(group_leader, false);
 
 		list_for_each_entry(sibling, &group_leader->sibling_list,
@@ -8138,13 +8155,7 @@ SYSCALL_DEFINE5(perf_event_open,
 			perf_remove_from_context(sibling, false);
 			put_ctx(gctx);
 		}
-	} else {
-		mutex_lock(&ctx->mutex);
-	}
-
-	WARN_ON_ONCE(ctx->parent_ctx);
 
-	if (move_group) {
 		/*
 		 * Wait for everybody to stop referencing the events through
 		 * the old lists, before installing it on new lists.
@@ -8176,22 +8187,20 @@ SYSCALL_DEFINE5(perf_event_open,
 		perf_event__state_init(group_leader);
 		perf_install_in_context(ctx, group_leader, group_leader->cpu);
 		get_ctx(ctx);
-	}
 
-	if (!exclusive_event_installable(event, ctx)) {
-		err = -EBUSY;
-		mutex_unlock(&ctx->mutex);
-		fput(event_file);
-		goto err_context;
+		/*
+		 * Now that all events are installed in @ctx, nothing
+		 * references @gctx anymore, so drop the last reference we have
+		 * on it.
+		 */
+		put_ctx(gctx);
 	}
 
 	perf_install_in_context(ctx, event, event->cpu);
 	perf_unpin_context(ctx);
 
-	if (move_group) {
+	if (move_group)
 		mutex_unlock(&gctx->mutex);
-		put_ctx(gctx);
-	}
 	mutex_unlock(&ctx->mutex);
 
 	put_online_cpus();
@@ -8218,6 +8227,12 @@ SYSCALL_DEFINE5(perf_event_open,
 	fd_install(event_fd, event_file);
 	return event_fd;
 
+err_locked:
+	if (move_group)
+		mutex_unlock(&gctx->mutex);
+	mutex_unlock(&ctx->mutex);
+/* err_file: */
+	fput(event_file);
 err_context:
 	perf_unpin_context(ctx);
 	put_ctx(ctx);
-- 
2.10.1

Back to linux.kernel | Previous | Next — Previous in thread | Find similar | Unroll thread

Thread

[PATCH -stable 4.1 0/2] perf/core: Fix CVE-2017-6001 Wang Nan <wangnan0@huawei.com> - 2017-02-24 14:10 +0100
  [PATCH -stable 4.1 2/2] perf/core: Fix concurrent sys_perf_event_open() vs. 'move_group' race Wang Nan <wangnan0@huawei.com> - 2017-02-24 14:10 +0100
  [PATCH -stable 4.1 1/2] perf: Restructure perf syscall point of no return Wang Nan <wangnan0@huawei.com> - 2017-02-24 14:10 +0100

csiph-web