[tip: perf/core] task_work: Add TWA_NMI_CURRENT as an additional notify mode.

From: tip-bot2 for Sebastian Andrzej Siewior
Date: Tue Jul 09 2024 - 07:43:30 EST


The following commit has been merged into the perf/core branch of tip:

Commit-ID: 466e4d801cd438a1ab2c8a2cce1bef6b65c31bbb
Gitweb: https://git.kernel.org/tip/466e4d801cd438a1ab2c8a2cce1bef6b65c31bbb
Author: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
AuthorDate: Thu, 04 Jul 2024 19:03:36 +02:00
Committer: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
CommitterDate: Tue, 09 Jul 2024 13:26:34 +02:00

task_work: Add TWA_NMI_CURRENT as an additional notify mode.

Adding task_work from NMI context requires the following:
- The kasan_record_aux_stack() is not NMU safe and must be avoided.
- Using TWA_RESUME is NMI safe. If the NMI occurs while the CPU is in
userland then it will continue in userland and not invoke the `work'
callback.

Add TWA_NMI_CURRENT as an additional notify mode. In this mode skip
kasan and use irq_work in hardirq-mode to for needed interrupt. Set
TIF_NOTIFY_RESUME within the irq_work callback due to k[ac]san
instrumentation in test_and_set_bit() which does not look NMI safe in
case of a report.

Suggested-by: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
Link: https://lore.kernel.org/r/20240704170424.1466941-3-bigeasy@xxxxxxxxxxxxx
---
include/linux/task_work.h | 1 +
kernel/task_work.c | 24 +++++++++++++++++++++---
2 files changed, 22 insertions(+), 3 deletions(-)

diff --git a/include/linux/task_work.h b/include/linux/task_work.h
index 26b8a47..cf5e7e8 100644
--- a/include/linux/task_work.h
+++ b/include/linux/task_work.h
@@ -18,6 +18,7 @@ enum task_work_notify_mode {
TWA_RESUME,
TWA_SIGNAL,
TWA_SIGNAL_NO_IPI,
+ TWA_NMI_CURRENT,
};

static inline bool task_work_pending(struct task_struct *task)
diff --git a/kernel/task_work.c b/kernel/task_work.c
index 2134ac8..5c2daa7 100644
--- a/kernel/task_work.c
+++ b/kernel/task_work.c
@@ -1,10 +1,18 @@
// SPDX-License-Identifier: GPL-2.0
+#include <linux/irq_work.h>
#include <linux/spinlock.h>
#include <linux/task_work.h>
#include <linux/resume_user_mode.h>

static struct callback_head work_exited; /* all we need is ->next == NULL */

+static void task_work_set_notify_irq(struct irq_work *entry)
+{
+ test_and_set_tsk_thread_flag(current, TIF_NOTIFY_RESUME);
+}
+static DEFINE_PER_CPU(struct irq_work, irq_work_NMI_resume) =
+ IRQ_WORK_INIT_HARD(task_work_set_notify_irq);
+
/**
* task_work_add - ask the @task to execute @work->func()
* @task: the task which should run the callback
@@ -12,7 +20,7 @@ static struct callback_head work_exited; /* all we need is ->next == NULL */
* @notify: how to notify the targeted task
*
* Queue @work for task_work_run() below and notify the @task if @notify
- * is @TWA_RESUME, @TWA_SIGNAL, or @TWA_SIGNAL_NO_IPI.
+ * is @TWA_RESUME, @TWA_SIGNAL, @TWA_SIGNAL_NO_IPI or @TWA_NMI_CURRENT.
*
* @TWA_SIGNAL works like signals, in that the it will interrupt the targeted
* task and run the task_work, regardless of whether the task is currently
@@ -24,6 +32,8 @@ static struct callback_head work_exited; /* all we need is ->next == NULL */
* kernel anyway.
* @TWA_RESUME work is run only when the task exits the kernel and returns to
* user mode, or before entering guest mode.
+ * @TWA_NMI_CURRENT works like @TWA_RESUME, except it can only be used for the
+ * current @task and if the current context is NMI.
*
* Fails if the @task is exiting/exited and thus it can't process this @work.
* Otherwise @work->func() will be called when the @task goes through one of
@@ -44,8 +54,13 @@ int task_work_add(struct task_struct *task, struct callback_head *work,
{
struct callback_head *head;

- /* record the work call stack in order to print it in KASAN reports */
- kasan_record_aux_stack(work);
+ if (notify == TWA_NMI_CURRENT) {
+ if (WARN_ON_ONCE(task != current))
+ return -EINVAL;
+ } else {
+ /* record the work call stack in order to print it in KASAN reports */
+ kasan_record_aux_stack(work);
+ }

head = READ_ONCE(task->task_works);
do {
@@ -66,6 +81,9 @@ int task_work_add(struct task_struct *task, struct callback_head *work,
case TWA_SIGNAL_NO_IPI:
__set_notify_signal(task);
break;
+ case TWA_NMI_CURRENT:
+ irq_work_queue(this_cpu_ptr(&irq_work_NMI_resume));
+ break;
default:
WARN_ON_ONCE(1);
break;