| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20140930172308.GI4241@worktop.programming.kicks-ass.net>
Date: Tue, 30 Sep 2014 19:23:08 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Sasha Levin <sasha.levin@...cle.com>
Cc: Cong Wang <cwang@...pensource.com>,
Vince Weaver <vincent.weaver@...ne.edu>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
Paul Mackerras <paulus@...ba.org>,
Ingo Molnar <mingo@...hat.com>,
Arnaldo Carvalho de Melo <acme@...nel.org>
Subject: Re: perf: perf_fuzzer triggers instant reboot
How about this then?
---
Subject: perf: Fix unclone_ctx() vs locking
The idiot who did 4a1c0f262f88 forgot to pay attention and fix all
similar cases. Do so now.
In particular, unclone_ctx() must be called while holding ctx->lock,
therefore all such sites are broken for the same reason. Pull the
put_ctx() call out from under ctx->lock.
Reported-by: Sasha Levin <sasha.levin@...cle.com>
Fixes: 4a1c0f262f88 ("perf: Fix lockdep warning on process exit")
Signed-off-by: Peter Zijlstra (Intel) <peterz@...radead.org>
---
kernel/events/core.c | 54 ++++++++++++++++++++++++++++++----------------------
1 file changed, 31 insertions(+), 23 deletions(-)
diff --git a/kernel/events/core.c b/kernel/events/core.c
index a232b4006026..769d8587cf31 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -911,13 +911,23 @@ static void put_ctx(struct perf_event_context *ctx)
}
}
-static void unclone_ctx(struct perf_event_context *ctx)
+/*
+ * This must be done under the ctx->lock, such as to serialize against
+ * context_equiv(), therefore we cannot call put_ctx() since that might end up
+ * calling scheduler related locks and ctx->lock nests inside those.
+ */
+static __must_check struct perf_event_context *
+unclone_ctx(struct perf_event_context *ctx)
{
- if (ctx->parent_ctx) {
- put_ctx(ctx->parent_ctx);
+ struct perf_event_context *parent_ctx = ctx->parent_ctx;
+
+ lockdep_assert_held(&ctx->lock);
+
+ if (parent_ctx)
ctx->parent_ctx = NULL;
- }
ctx->generation++;
+
+ return parent_ctx;
}
static u32 perf_event_pid(struct perf_event *event, struct task_struct *p)
@@ -2264,6 +2274,9 @@ static void ctx_sched_out(struct perf_event_context *ctx,
static int context_equiv(struct perf_event_context *ctx1,
struct perf_event_context *ctx2)
{
+ lockdep_assert_held(&ctx1->lock);
+ lockdep_assert_held(&ctx2->lock);
+
/* Pinning disables the swap optimization */
if (ctx1->pin_count || ctx2->pin_count)
return 0;
@@ -2997,6 +3010,7 @@ static int event_enable_on_exec(struct perf_event *event,
*/
static void perf_event_enable_on_exec(struct perf_event_context *ctx)
{
+ struct perf_event_context *clone_ctx = NULL;
struct perf_event *event;
unsigned long flags;
int enabled = 0;
@@ -3028,7 +3042,7 @@ static void perf_event_enable_on_exec(struct perf_event_context *ctx)
* Unclone this context if we enabled any event.
*/
if (enabled)
- unclone_ctx(ctx);
+ clone_ctx = unclone_ctx(ctx);
raw_spin_unlock(&ctx->lock);
@@ -3038,6 +3052,9 @@ static void perf_event_enable_on_exec(struct perf_event_context *ctx)
perf_event_context_sched_in(ctx, ctx->task);
out:
local_irq_restore(flags);
+
+ if (clone_ctx)
+ put_ctx(clone_ctx);
}
void perf_event_exec(void)
@@ -3190,7 +3207,7 @@ find_lively_task_by_vpid(pid_t vpid)
static struct perf_event_context *
find_get_context(struct pmu *pmu, struct task_struct *task, int cpu)
{
- struct perf_event_context *ctx;
+ struct perf_event_context *ctx, *clone_ctx = NULL;
struct perf_cpu_context *cpuctx;
unsigned long flags;
int ctxn, err;
@@ -3224,9 +3241,12 @@ find_get_context(struct pmu *pmu, struct task_struct *task, int cpu)
retry:
ctx = perf_lock_task_context(task, ctxn, &flags);
if (ctx) {
- unclone_ctx(ctx);
+ clone_ctx = unclone_ctx(ctx);
++ctx->pin_count;
raw_spin_unlock_irqrestore(&ctx->lock, flags);
+
+ if (clone_ctx)
+ put_ctx(clone_ctx);
} else {
ctx = alloc_perf_context(pmu, task);
err = -ENOMEM;
@@ -7651,7 +7671,7 @@ __perf_event_exit_task(struct perf_event *child_event,
static void perf_event_exit_task_context(struct task_struct *child, int ctxn)
{
struct perf_event *child_event, *next;
- struct perf_event_context *child_ctx, *parent_ctx;
+ struct perf_event_context *child_ctx, *clone_ctx = NULL;
unsigned long flags;
if (likely(!child->perf_event_ctxp[ctxn])) {
@@ -7678,28 +7698,16 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn)
child->perf_event_ctxp[ctxn] = NULL;
/*
- * In order to avoid freeing: child_ctx->parent_ctx->task
- * under perf_event_context::lock, grab another reference.
- */
- parent_ctx = child_ctx->parent_ctx;
- if (parent_ctx)
- get_ctx(parent_ctx);
-
- /*
* If this context is a clone; unclone it so it can't get
* swapped to another process while we're removing all
* the events from it.
*/
- unclone_ctx(child_ctx);
+ clone_ctx = unclone_ctx(child_ctx);
update_context_time(child_ctx);
raw_spin_unlock_irqrestore(&child_ctx->lock, flags);
- /*
- * Now that we no longer hold perf_event_context::lock, drop
- * our extra child_ctx->parent_ctx reference.
- */
- if (parent_ctx)
- put_ctx(parent_ctx);
+ if (clone_ctx)
+ put_ctx(clone_ctx);
/*
* Report the task dead after unscheduling the events so that we
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists