On 2023-03-12 a.m.1:47, Michael S. Tsirkin wrote:
External email: Use caution opening links or attachments
On Sat, Mar 11, 2023 at 05:19:43PM -0500, Feng Liu wrote:
On 2023-03-11 p.m.2:06, Michael S. Tsirkin wrote:
External email: Use caution opening links or attachments
On Fri, Mar 10, 2023 at 08:21:31AM -0500, Feng Liu wrote:
On 2023-03-10 a.m.3:06, Michael S. Tsirkin wrote:
External email: Use caution opening links or attachments
On Fri, Mar 10, 2023 at 07:34:25AM +0200, Feng Liu wrote:
This patch series performs a clean up of the code in virtio_ring and
virtio_pci, modifying it to conform with the Linux kernel coding style
guidance [1]. The modifications ensure the code easy to read and
understand. This small series does few short cleanups in the code.
Patch-1 Allow non power of 2 sizes for virtqueues
Patch-2 Avoid using inline for small functions.
Patch-3 Use const to annotate read-only pointer params.
[1]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.kernel.org%2Fdoc%2Fhtml%2Fv6.2-rc3%2Fprocess%2Fcoding-style.html%23the-inline-disease&data=05%7C01%7Cfeliu%40nvidia.com%7C329325e3773c46e5cadc08db22c5bc87%7C43083d15727340c1b7db39efd9ccc17a%7C0%7C0%7C638142004918816377%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=57roDDMfOEOIuEo%2FqgBgUklMBGHfdSWRI0KTRXQCNgM%3D&reserved=0
All of the patches have been verified based on the kernel code
commit 44889ba56cbb ("Merge tag 'net-6.3-rc2' of
git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net")
verified how?
Hi Michael
1. Applied the patches on lastest kernel source(44889ba56cbb), compile and
install the kernel, and use iperf to test traffic
2. To validate this change, we tested various virtqueue sizes for packed
rings, including 128, 256, 512, 100, 200, 500, and 1000, with
CONFIG_PAGE_POISONING enabled, and test by iperf& ping -f and all tests
passed successfully.
Given split ring does not support non power of 2 how exactly
did you configure non power of 2?
Hi, Michael
We can implement the test by modifying qemu; 1. force the
VIRTIO_F_RING_PACKED feature bit to be set, 2. set
VIRTIO_NET_RX_QUEUE_DEFAULT_SIZE and VIRTIO_NET_TX_QUEUE_DEFAULT_SIZE to the
value of non power_of_2, 3. remove the check of is_power_of virtqueue, then
qemu can create the required virtual device (non power_2 size , packed
virtqueue device) ;In this way, any length and packed ring test can be
performed;
remove the modified code, I can test split vq, and can see that the size of
power_of_2 can load the driver normally, and the size of non power_of_2 will
give an warning and fail to load the driver
Sounds like a plan but what exactly did you do previously?
You indicated you tested non powers of 2.
Hi, Michael
It is not just a plan. Previously, I have already created the
non_power_of_2 packed virtqueue device and the standard split vq device
for test, and after startup, I performed related tests using iperf and
ping -f. This time I will construct a split vq device of non_power_of_2
virtqueue for more abnormal tests
Feng Liu (3):
virtio_pci_modern: Allow non power of 2 sizes for virtqueues
virtio_ring: Avoid using inline for small functions
virtio_ring: Use const to annotate read-only pointer params
drivers/virtio/virtio_pci_modern.c | 5 ----
drivers/virtio/virtio_ring.c | 48 +++++++++++++++---------------
include/linux/virtio.h | 14 ++++-----
3 files changed, 31 insertions(+), 36 deletions(-)
--
2.34.1
_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization
