lkml.org 
[lkml]   [2022]   [Jun]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [RFC PATCH v2 1/8] virtio/vsock: rework packet allocation logic
On Fri, Jun 03, 2022 at 05:31:00AM +0000, Arseniy Krasnov wrote:
>To support zerocopy receive, packet's buffer allocation
>is changed: for buffers which could be mapped to user's
>vma we can't use 'kmalloc()'(as kernel restricts to map
>slab pages to user's vma) and raw buddy allocator now
>called. But, for tx packets(such packets won't be mapped
>to user), previous 'kmalloc()' way is used, but with special
>flag in packet's structure which allows to distinguish
>between 'kmalloc()' and raw pages buffers.
>
>Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru>
>---
> include/linux/virtio_vsock.h | 1 +
> net/vmw_vsock/virtio_transport.c | 8 ++++++--
> net/vmw_vsock/virtio_transport_common.c | 9 ++++++++-
> 3 files changed, 15 insertions(+), 3 deletions(-)

Each patch should as much as possible work to not break the
bisectability, and here we are not touching vhost_vsock_alloc_pkt() that
uses kmalloc to allocate the buffer.

I see you updated it in the next patch, that should be fine, but here
you should set slab_buf in vhost_vsock_alloc_pkt(), or you can merge the
two patches.

>
>diff --git a/include/linux/virtio_vsock.h b/include/linux/virtio_vsock.h
>index 35d7eedb5e8e..d02cb7aa922f 100644
>--- a/include/linux/virtio_vsock.h
>+++ b/include/linux/virtio_vsock.h
>@@ -50,6 +50,7 @@ struct virtio_vsock_pkt {
> u32 off;
> bool reply;
> bool tap_delivered;
>+ bool slab_buf;
> };
>
> struct virtio_vsock_pkt_info {
>diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
>index ad64f403536a..19909c1e9ba3 100644
>--- a/net/vmw_vsock/virtio_transport.c
>+++ b/net/vmw_vsock/virtio_transport.c
>@@ -255,16 +255,20 @@ static void virtio_vsock_rx_fill(struct virtio_vsock *vsock)
> vq = vsock->vqs[VSOCK_VQ_RX];
>
> do {
>+ struct page *buf_page;
>+
> pkt = kzalloc(sizeof(*pkt), GFP_KERNEL);
> if (!pkt)
> break;
>
>- pkt->buf = kmalloc(buf_len, GFP_KERNEL);
>- if (!pkt->buf) {
>+ buf_page = alloc_page(GFP_KERNEL);
>+
>+ if (!buf_page) {
> virtio_transport_free_pkt(pkt);
> break;
> }
>
>+ pkt->buf = page_to_virt(buf_page);
> pkt->buf_len = buf_len;
> pkt->len = buf_len;
>
>diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
>index ec2c2afbf0d0..278567f748f2 100644
>--- a/net/vmw_vsock/virtio_transport_common.c
>+++ b/net/vmw_vsock/virtio_transport_common.c
>@@ -69,6 +69,7 @@ virtio_transport_alloc_pkt(struct virtio_vsock_pkt_info *info,
> if (!pkt->buf)
> goto out_pkt;
>
>+ pkt->slab_buf = true;
> pkt->buf_len = len;
>
> err = memcpy_from_msg(pkt->buf, info->msg, len);
>@@ -1342,7 +1343,13 @@ EXPORT_SYMBOL_GPL(virtio_transport_recv_pkt);
>
> void virtio_transport_free_pkt(struct virtio_vsock_pkt *pkt)
> {
>- kfree(pkt->buf);
>+ if (pkt->buf_len) {
>+ if (pkt->slab_buf)
>+ kfree(pkt->buf);
>+ else
>+ free_pages(buf, get_order(pkt->buf_len));
>+ }
>+
> kfree(pkt);
> }
> EXPORT_SYMBOL_GPL(virtio_transport_free_pkt);
>--
>2.25.1

\
 
 \ /
  Last update: 2022-06-09 10:38    [W:0.242 / U:0.640 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site