In order to synchronize changes that can affect the thread callback,
introduce an interface to force a flush of the inject workqueue.  The
irqfd pointer is only valid under spinlock, but the workqueue cannot
be flushed under spinlock.  Therefore the flush work for the irqfd is
queued under spinlock.  The vfio_irqfd_cleanup_wq workqueue is re-used
for queuing this work such that flushing the workqueue is also ordered
relative to shutdown.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
---
 drivers/vfio/virqfd.c | 21 +++++++++++++++++++++
 include/linux/vfio.h  |  2 ++
 2 files changed, 23 insertions(+)

diff --git a/drivers/vfio/virqfd.c b/drivers/vfio/virqfd.c
index 29c564b7a6e1..532269133801 100644
--- a/drivers/vfio/virqfd.c
+++ b/drivers/vfio/virqfd.c
@@ -101,6 +101,13 @@ static void virqfd_inject(struct work_struct *work)
 		virqfd->thread(virqfd->opaque, virqfd->data);
 }
 
+static void virqfd_flush_inject(struct work_struct *work)
+{
+	struct virqfd *virqfd = container_of(work, struct virqfd, flush_inject);
+
+	flush_work(&virqfd->inject);
+}
+
 int vfio_virqfd_enable(void *opaque,
 		       int (*handler)(void *, void *),
 		       void (*thread)(void *, void *),
@@ -124,6 +131,7 @@ int vfio_virqfd_enable(void *opaque,
 
 	INIT_WORK(&virqfd->shutdown, virqfd_shutdown);
 	INIT_WORK(&virqfd->inject, virqfd_inject);
+	INIT_WORK(&virqfd->flush_inject, virqfd_flush_inject);
 
 	irqfd = fdget(fd);
 	if (!irqfd.file) {
@@ -213,3 +221,16 @@ void vfio_virqfd_disable(struct virqfd **pvirqfd)
 	flush_workqueue(vfio_irqfd_cleanup_wq);
 }
 EXPORT_SYMBOL_GPL(vfio_virqfd_disable);
+
+void vfio_virqfd_flush_thread(struct virqfd **pvirqfd)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&virqfd_lock, flags);
+	if (*pvirqfd && (*pvirqfd)->thread)
+		queue_work(vfio_irqfd_cleanup_wq, &(*pvirqfd)->flush_inject);
+	spin_unlock_irqrestore(&virqfd_lock, flags);
+
+	flush_workqueue(vfio_irqfd_cleanup_wq);
+}
+EXPORT_SYMBOL_GPL(vfio_virqfd_flush_thread);
diff --git a/include/linux/vfio.h b/include/linux/vfio.h
index 89b265bc6ec3..8b1a29820409 100644
--- a/include/linux/vfio.h
+++ b/include/linux/vfio.h
@@ -356,6 +356,7 @@ struct virqfd {
 	wait_queue_entry_t		wait;
 	poll_table		pt;
 	struct work_struct	shutdown;
+	struct work_struct	flush_inject;
 	struct virqfd		**pvirqfd;
 };
 
@@ -363,5 +364,6 @@ int vfio_virqfd_enable(void *opaque, int (*handler)(void *, void *),
 		       void (*thread)(void *, void *), void *data,
 		       struct virqfd **pvirqfd, int fd);
 void vfio_virqfd_disable(struct virqfd **pvirqfd);
+void vfio_virqfd_flush_thread(struct virqfd **pvirqfd);
 
 #endif /* VFIO_H */
-- 
2.43.2

Hi Alex,

On 3/6/2024 1:14 PM, Alex Williamson wrote:
> In order to synchronize changes that can affect the thread callback,
> introduce an interface to force a flush of the inject workqueue.  The
> irqfd pointer is only valid under spinlock, but the workqueue cannot
> be flushed under spinlock.  Therefore the flush work for the irqfd is
> queued under spinlock.  The vfio_irqfd_cleanup_wq workqueue is re-used
> for queuing this work such that flushing the workqueue is also ordered
> relative to shutdown.
> 
> Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
> ---

Thank you very much.

Reviewed-by: Reinette Chatre <reinette.chatre@intel.com>

Reinette

> From: Alex Williamson <alex.williamson@redhat.com>
> Sent: Thursday, March 7, 2024 5:15 AM
> 
> In order to synchronize changes that can affect the thread callback,
> introduce an interface to force a flush of the inject workqueue.  The
> irqfd pointer is only valid under spinlock, but the workqueue cannot
> be flushed under spinlock.  Therefore the flush work for the irqfd is
> queued under spinlock.  The vfio_irqfd_cleanup_wq workqueue is re-used
> for queuing this work such that flushing the workqueue is also ordered
> relative to shutdown.
> 
> Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

Reviewed-by: Kevin Tian <kevin.tian@intel.com>