On Mon, Sep 18, 2023 at 4:41 PM Jason Wang <jasowang@xxxxxxxxxx> wrote:
>
> On Tue, Sep 12, 2023 at 11:00 AM Cindy Lu <lulu@xxxxxxxxxx> wrote:
> >
> > Add the function vduse_alloc_reconnnect_info_mem
> > and vduse_alloc_reconnnect_info_mem
> > In this 2 function, vduse will get/free (vq_num + 1)*page
> > Page 0 will be used to save the reconnection information, The
> > Userspace App will maintain this. Page 1 ~ vq_num + 1 will save
> > the reconnection information for vqs.
>
> Please explain why this is needed instead of only describing how it is
> implemented. (Code can explain itself).
>
> >
> > Signed-off-by: Cindy Lu <lulu@xxxxxxxxxx>
> > ---
> > drivers/vdpa/vdpa_user/vduse_dev.c | 86 ++++++++++++++++++++++++++++++
> > 1 file changed, 86 insertions(+)
> >
> > diff --git a/drivers/vdpa/vdpa_user/vduse_dev.c b/drivers/vdpa/vdpa_user/vduse_dev.c
> > index 26b7e29cb900..4c256fa31fc4 100644
> > --- a/drivers/vdpa/vdpa_user/vduse_dev.c
> > +++ b/drivers/vdpa/vdpa_user/vduse_dev.c
> > @@ -30,6 +30,10 @@
> > #include <uapi/linux/virtio_blk.h>
> > #include <linux/mod_devicetable.h>
> >
> > +#ifdef CONFIG_X86
> > +#include <asm/set_memory.h>
> > +#endif
> > +
> > #include "iova_domain.h"
> >
> > #define DRV_AUTHOR "Yongji Xie <xieyongji@xxxxxxxxxxxxx>"
> > @@ -41,6 +45,23 @@
> > #define VDUSE_IOVA_SIZE (128 * 1024 * 1024)
> > #define VDUSE_MSG_DEFAULT_TIMEOUT 30
> >
> > +/* struct vdpa_reconnect_info save the page information for reconnection
> > + * kernel will init these information while alloc the pages
> > + * and use these information to free the pages
> > + */
> > +struct vdpa_reconnect_info {
> > + /* Offset (within vm_file) in PAGE_SIZE,
> > + * this just for check, not using
> > + */
> > + u32 index;
> > + /* physical address for this page*/
> > + phys_addr_t addr;
> > + /* virtual address for this page*/
> > + unsigned long vaddr;
>
> If it could be switched by virt_to_phys() why duplicate those fields?
>
yes will remove this part
Thanks
Cindy
> > + /* memory size, here always page_size*/
> > + phys_addr_t size;
>
> If it's always PAGE_SIZE why would we have this?
will remove this
Thanks
Cindy
>
> > +};
> > +
> > struct vduse_virtqueue {
> > u16 index;
> > u16 num_max;
> > @@ -57,6 +78,7 @@ struct vduse_virtqueue {
> > struct vdpa_callback cb;
> > struct work_struct inject;
> > struct work_struct kick;
> > + struct vdpa_reconnect_info reconnect_info;
> > };
> >
> > struct vduse_dev;
> > @@ -106,6 +128,7 @@ struct vduse_dev {
> > u32 vq_align;
> > struct vduse_umem *umem;
> > struct mutex mem_lock;
> > + struct vdpa_reconnect_info reconnect_status;
> > };
> >
> > struct vduse_dev_msg {
> > @@ -1030,6 +1053,65 @@ static int vduse_dev_reg_umem(struct vduse_dev *dev,
> > return ret;
> > }
> >
> > +int vduse_alloc_reconnnect_info_mem(struct vduse_dev *dev)
> > +{
> > + struct vdpa_reconnect_info *info;
> > + struct vduse_virtqueue *vq;
> > + void *addr;
> > +
> > + /*page 0 is use to save status,dpdk will use this to save the information
> > + *needed in reconnection,kernel don't need to maintain this
> > + */
> > + info = &dev->reconnect_status;
> > + addr = (void *)get_zeroed_page(GFP_KERNEL);
> > + if (!addr)
> > + return -1;
>
> -ENOMEM?
>
sure will change this
Thanks
Cidny
> Thanks
>
