nvme: only add a controller to dev_list after it's been fully initialized

Without this we can easily get bad derferences on nvmeq->d_db when the nvme
kthread tries to poll the CQs for controllers that are in half initialized
state.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
This commit is contained in:
Christoph Hellwig 2015-10-22 14:03:33 +02:00 committed by Jens Axboe
parent 749941f236
commit 7385014c07

View File

@ -1994,6 +1994,30 @@ static void nvme_disable_io_queues(struct nvme_dev *dev)
kthread_stop(kworker_task); kthread_stop(kworker_task);
} }
static int nvme_dev_list_add(struct nvme_dev *dev)
{
bool start_thread = false;
spin_lock(&dev_list_lock);
if (list_empty(&dev_list) && IS_ERR_OR_NULL(nvme_thread)) {
start_thread = true;
nvme_thread = NULL;
}
list_add(&dev->node, &dev_list);
spin_unlock(&dev_list_lock);
if (start_thread) {
nvme_thread = kthread_run(nvme_kthread, NULL, "nvme");
wake_up_all(&nvme_kthread_wait);
} else
wait_event_killable(nvme_kthread_wait, nvme_thread);
if (IS_ERR_OR_NULL(nvme_thread))
return nvme_thread ? PTR_ERR(nvme_thread) : -EINTR;
return 0;
}
/* /*
* Remove the node from the device list and check * Remove the node from the device list and check
* for whether or not we need to stop the nvme_thread. * for whether or not we need to stop the nvme_thread.
@ -2109,7 +2133,6 @@ static void nvme_pci_free_ctrl(struct nvme_ctrl *ctrl)
static void nvme_probe_work(struct work_struct *work) static void nvme_probe_work(struct work_struct *work)
{ {
struct nvme_dev *dev = container_of(work, struct nvme_dev, probe_work); struct nvme_dev *dev = container_of(work, struct nvme_dev, probe_work);
bool start_thread = false;
int result; int result;
result = nvme_dev_map(dev); result = nvme_dev_map(dev);
@ -2120,25 +2143,6 @@ static void nvme_probe_work(struct work_struct *work)
if (result) if (result)
goto unmap; goto unmap;
spin_lock(&dev_list_lock);
if (list_empty(&dev_list) && IS_ERR_OR_NULL(nvme_thread)) {
start_thread = true;
nvme_thread = NULL;
}
list_add(&dev->node, &dev_list);
spin_unlock(&dev_list_lock);
if (start_thread) {
nvme_thread = kthread_run(nvme_kthread, NULL, "nvme");
wake_up_all(&nvme_kthread_wait);
} else
wait_event_killable(nvme_kthread_wait, nvme_thread);
if (IS_ERR_OR_NULL(nvme_thread)) {
result = nvme_thread ? PTR_ERR(nvme_thread) : -EINTR;
goto disable;
}
nvme_init_queue(dev->queues[0], 0); nvme_init_queue(dev->queues[0], 0);
result = nvme_alloc_admin_tags(dev); result = nvme_alloc_admin_tags(dev);
if (result) if (result)
@ -2154,6 +2158,10 @@ static void nvme_probe_work(struct work_struct *work)
dev->ctrl.event_limit = 1; dev->ctrl.event_limit = 1;
result = nvme_dev_list_add(dev);
if (result)
goto remove;
/* /*
* Keep the controller around but remove all namespaces if we don't have * Keep the controller around but remove all namespaces if we don't have
* any working I/O queue. * any working I/O queue.
@ -2168,6 +2176,8 @@ static void nvme_probe_work(struct work_struct *work)
return; return;
remove:
nvme_dev_list_remove(dev);
free_tags: free_tags:
nvme_dev_remove_admin(dev); nvme_dev_remove_admin(dev);
blk_put_queue(dev->ctrl.admin_q); blk_put_queue(dev->ctrl.admin_q);
@ -2175,7 +2185,6 @@ static void nvme_probe_work(struct work_struct *work)
dev->queues[0]->tags = NULL; dev->queues[0]->tags = NULL;
disable: disable:
nvme_disable_queue(dev, 0); nvme_disable_queue(dev, 0);
nvme_dev_list_remove(dev);
unmap: unmap:
nvme_dev_unmap(dev); nvme_dev_unmap(dev);
out: out: