linux-kernel - Re: [PATCH] virtio_ring: remove unnecessary to

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <7d0d6b08-321e-2e09-3ced-add916e6f07d@linux.alibaba.com>
Date:   Mon, 28 Mar 2022 17:13:16 +0800
From:   Xianting Tian <xianting.tian@...ux.alibaba.com>
To:     Stefano Garzarella <sgarzare@...hat.com>
Cc:     mst@...hat.com, jasowang@...hat.com,
        virtualization@...ts.linux-foundation.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] virtio_ring: remove unnecessary to_vvq call in vring hot
 path

在 2022/3/28 下午5:05, Stefano Garzarella 写道:

> On Thu, Mar 24, 2022 at 03:33:40PM +0800, Xianting Tian wrote:
>> It passes '_vq' to virtqueue_use_indirect(), which still calls
>> to_vvq to get 'vq', let's directly pass 'vq'. It can avoid
>> unnecessary call of to_vvq in hot path.
>
> It seems reasonable to me.
>
>>
>> Other tiny optimization:
>> Add unlikely to "if (vq->vq.num_free < descs_used).
>
> Better to do this change in another patch.
Thanks for the reply, I will send the patchset, which contains the two 
patches, soon.
>
> Thanks,
> Stefano
>
>>
>> Signed-off-by: Xianting Tian <xianting.tian@...ux.alibaba.com>
>> ---
>> drivers/virtio/virtio_ring.c | 10 ++++------
>> 1 file changed, 4 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
>> index 962f1477b1fa..ab6d5f0cb579 100644
>> --- a/drivers/virtio/virtio_ring.c
>> +++ b/drivers/virtio/virtio_ring.c
>> @@ -205,11 +205,9 @@ struct vring_virtqueue {
>>
>> #define to_vvq(_vq) container_of(_vq, struct vring_virtqueue, vq)
>>
>> -static inline bool virtqueue_use_indirect(struct virtqueue *_vq,
>> +static inline bool virtqueue_use_indirect(struct vring_virtqueue *vq,
>>                       unsigned int total_sg)
>> {
>> -    struct vring_virtqueue *vq = to_vvq(_vq);
>> -
>>     /*
>>      * If the host supports indirect descriptor tables, and we have 
>> multiple
>>      * buffers, then go indirect. FIXME: tune this threshold
>> @@ -507,7 +505,7 @@ static inline int virtqueue_add_split(struct 
>> virtqueue *_vq,
>>
>>     head = vq->free_head;
>>
>> -    if (virtqueue_use_indirect(_vq, total_sg))
>> +    if (virtqueue_use_indirect(vq, total_sg))
>>         desc = alloc_indirect_split(_vq, total_sg, gfp);
>>     else {
>>         desc = NULL;
>> @@ -527,7 +525,7 @@ static inline int virtqueue_add_split(struct 
>> virtqueue *_vq,
>>         descs_used = total_sg;
>>     }
>>
>> -    if (vq->vq.num_free < descs_used) {
>> +    if (unlikely(vq->vq.num_free < descs_used)) {
>>         pr_debug("Can't add buf len %i - avail = %i\n",
>>              descs_used, vq->vq.num_free);
>>         /* FIXME: for historical reasons, we force a notify here if
>> @@ -1194,7 +1192,7 @@ static inline int virtqueue_add_packed(struct 
>> virtqueue *_vq,
>>
>>     BUG_ON(total_sg == 0);
>>
>> -    if (virtqueue_use_indirect(_vq, total_sg)) {
>> +    if (virtqueue_use_indirect(vq, total_sg)) {
>>         err = virtqueue_add_indirect_packed(vq, sgs, total_sg, out_sgs,
>>                             in_sgs, data, gfp);
>>         if (err != -ENOMEM) {
>> -- 
>> 2.17.1
>>