Re: [PATCH v2] virtio_ring: use alloc_pages_node for NUMA-aware allocation

From: Shile Zhang
Date: Mon Jul 27 2020 - 09:10:17 EST

On 2020/7/21 19:28, Shile Zhang wrote:

On 2020/7/21 16:18, Michael S. Tsirkin wrote:
On Tue, Jul 21, 2020 at 03:00:13PM +0800, Shile Zhang wrote:
Use alloc_pages_node() allocate memory for vring queue with proper
NUMA affinity.

Reported-by: kernel test robot <lkp@xxxxxxxxx>
Suggested-by: Jiang Liu <liuj97@xxxxxxxxx>
Signed-off-by: Shile Zhang <shile.zhang@xxxxxxxxxxxxxxxxx>

Do you observe any performance gains from this patch?

Thanks for your comments!
Yes, the bandwidth can boost more than doubled (from 30Gbps to 80GBps) with this changes in my test env (8 numa nodes), with netperf test.

I also wonder why isn't the probe code run on the correct numa node?
That would fix a wide class of issues like this without need to tweak

Good point, I'll check this, thanks!

Sorry, I have no idea about how the probe code to grab the appropriate NUMA node.

Bjorn, what do you think? Was this considered?

Hi Bjorn, Could you please give any comments about this issue?

v1 -> v2:
- fixed compile warning reported by LKP.
 drivers/virtio/virtio_ring.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 58b96baa8d48..d38fd6872c8c 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -276,9 +276,11 @@ static void *vring_alloc_queue(struct virtio_device *vdev, size_t size,
ÂÂÂÂÂÂÂÂÂ return dma_alloc_coherent(vdev->dev.parent, size,
ÂÂÂÂÂ } else {
-ÂÂÂÂÂÂÂ void *queue = alloc_pages_exact(PAGE_ALIGN(size), flag);
-ÂÂÂÂÂÂÂ if (queue) {
+ÂÂÂÂÂÂÂ void *queue = NULL;
+ÂÂÂÂÂÂÂ struct page *page = alloc_pages_node(dev_to_node(vdev->dev.parent),
+ÂÂÂÂÂÂÂ if (page) {
+ÂÂÂÂÂÂÂÂÂÂÂ queue = page_address(page);
ÂÂÂÂÂÂÂÂÂÂÂÂÂ phys_addr_t phys_addr = virt_to_phys(queue);
ÂÂÂÂÂÂÂÂÂÂÂÂÂ *dma_handle = (dma_addr_t)phys_addr;
@@ -308,7 +310,7 @@ static void vring_free_queue(struct virtio_device *vdev, size_t size,
ÂÂÂÂÂ if (vring_use_dma_api(vdev))
ÂÂÂÂÂÂÂÂÂ dma_free_coherent(vdev->dev.parent, size, queue, dma_handle);
ÂÂÂÂÂ else
-ÂÂÂÂÂÂÂ free_pages_exact(queue, PAGE_ALIGN(size));
+ÂÂÂÂÂÂÂ free_pages((unsigned long)queue, get_order(size));