[PATCH v3 4/5] vhost_user: support function defines for back-end
Bathija, Pravin
Pravin.Bathija at dell.com
Thu Nov 13 03:16:35 CET 2025
Comments inline.
Internal Use - Confidential
> -----Original Message-----
> From: fengchengwen <fengchengwen at huawei.com>
> Sent: Tuesday, November 11, 2025 5:21 PM
> To: Bathija, Pravin <Pravin.Bathija at dell.com>; dev at dpdk.org
> Cc: pravin.m.bathija.dev at gmail.com
> Subject: Re: [PATCH v3 4/5] vhost_user: support function defines for back-end
>
>
> [EXTERNAL EMAIL]
>
> On 11/11/2025 7:31 PM, Bathija, Pravin wrote:
> > Responses inline.
> >
> >
> > Internal Use - Confidential
> >> -----Original Message-----
> >> From: fengchengwen <fengchengwen at huawei.com>
> >> Sent: Tuesday, November 4, 2025 12:06 AM
> >> To: Bathija, Pravin <Pravin.Bathija at dell.com>; dev at dpdk.org
> >> Cc: pravin.m.bathija.dev at gmail.com
> >> Subject: Re: [PATCH v3 4/5] vhost_user: support function defines for
> >> back-end
> >>
> >>
> >> [EXTERNAL EMAIL]
> >>
> >> On 11/4/2025 12:21 PM, Pravin M Bathija wrote:
> >>> Here we define support functions which are called from the various
> >>> vhost-user back-end message functions like set memory table, get
> >>> memory slots, add memory region, remove memory region. These are
> >>> essetially common functions to initialize memory, unmap a set of
> >>> memory regions, perform register copy and align memory addresses.
> >>>
> >>> Signed-off-by: Pravin M Bathija <pravin.bathija at dell.com>
> >>> ---
> >>> lib/vhost/vhost_user.c | 80
> >>> +++++++++++++++++++++++++++++++++++-------
> >>> 1 file changed, 68 insertions(+), 12 deletions(-)
> >>>
> >>> diff --git a/lib/vhost/vhost_user.c b/lib/vhost/vhost_user.c index
> >>> 168432e7d1..9a85f2fc92 100644
> >>> --- a/lib/vhost/vhost_user.c
> >>> +++ b/lib/vhost/vhost_user.c
> >>> @@ -228,7 +228,17 @@ async_dma_map(struct virtio_net *dev, bool
> >>> do_map) }
> >>>
> >>> static void
> >>> -free_mem_region(struct virtio_net *dev)
> >>> +free_mem_region(struct rte_vhost_mem_region *reg) {
> >>> + if (reg != NULL && reg->host_user_addr) {
> >>> + munmap(reg->mmap_addr, reg->mmap_size);
> >>> + close(reg->fd);
> >>> + memset(reg, 0, sizeof(struct rte_vhost_mem_region));
> >>> + }
> >>> +}
> >>> +
> >>> +static void
> >>> +free_all_mem_regions(struct virtio_net *dev)
> >>> {
> >>> uint32_t i;
> >>> struct rte_vhost_mem_region *reg; @@ -239,12 +249,10 @@
> >>> free_mem_region(struct virtio_net *dev)
> >>> if (dev->async_copy && rte_vfio_is_enabled("vfio"))
> >>> async_dma_map(dev, false);
> >>>
> >>> - for (i = 0; i < dev->mem->nregions; i++) {
> >>> + for (i = 0; i < VHOST_MEMORY_MAX_NREGIONS; i++) {
> >>> reg = &dev->mem->regions[i];
> >>> - if (reg->host_user_addr) {
> >>> - munmap(reg->mmap_addr, reg->mmap_size);
> >>> - close(reg->fd);
> >>> - }
> >>> + if (reg->mmap_addr)
> >>> + free_mem_region(reg);
> >>> }
> >>> }
> >>>
> >>> @@ -258,7 +266,7 @@ vhost_backend_cleanup(struct virtio_net *dev)
> >>> vdpa_dev->ops->dev_cleanup(dev->vid);
> >>>
> >>> if (dev->mem) {
> >>> - free_mem_region(dev);
> >>> + free_all_mem_regions(dev);
> >>> rte_free(dev->mem);
> >>> dev->mem = NULL;
> >>> }
> >>> @@ -707,7 +715,7 @@ numa_realloc(struct virtio_net **pdev, struct
> >> vhost_virtqueue **pvq)
> >>> vhost_devices[dev->vid] = dev;
> >>>
> >>> mem_size = sizeof(struct rte_vhost_memory) +
> >>> - sizeof(struct rte_vhost_mem_region) * dev->mem->nregions;
> >>> + sizeof(struct rte_vhost_mem_region) *
> >> VHOST_MEMORY_MAX_NREGIONS;
> >>> mem = rte_realloc_socket(dev->mem, mem_size, 0, node);
> >>> if (!mem) {
> >>> VHOST_CONFIG_LOG(dev->ifname, ERR, @@ -811,8 +819,10 @@
> >>> hua_to_alignment(struct rte_vhost_memory
> >> *mem, void *ptr)
> >>> uint32_t i;
> >>> uintptr_t hua = (uintptr_t)ptr;
> >>>
> >>> - for (i = 0; i < mem->nregions; i++) {
> >>> + for (i = 0; i < VHOST_MEMORY_MAX_NREGIONS; i++) {
> >>> r = &mem->regions[i];
> >>> + if (r->host_user_addr == 0)
> >>> + continue;
> >>> if (hua >= r->host_user_addr &&
> >>> hua < r->host_user_addr + r->size) {
> >>> return get_blk_size(r->fd); @@ -1250,9 +1260,13
> >>> @@ vhost_user_postcopy_register(struct virtio_net
> >> *dev, int main_fd,
> >>> * retrieve the region offset when handling userfaults.
> >>> */
> >>> memory = &ctx->msg.payload.memory;
> >>> - for (i = 0; i < memory->nregions; i++) {
> >>> + for (i = 0; i < VHOST_MEMORY_MAX_NREGIONS; i++) {
> >>
> >> I think the using MAX_NREGIONS are most for convienent, but it will
> >> impact the performance, because the rte_vhost_va_from_guest_pa()
> >> should iter the entire array.
> >>
> >
> > Replaced VHOST_MEMORY_MAX_NREGIONS with memory->nregions.
> Please review v4.
> >
> >> I think we should keep the original impl: make sure the nregions
> >> entry of memory-region is always valid.
> >>
> >> Beside, where is the modification for rte_vhost_va_from_guest_pa()???
> >
> > Could you please provide more detail ? rte_vhost_va_from_guest_pa() was
> never called from here before.
>
> Because rte_vhost_va_from_guest_pa() use mem-nregions as for uplimit:
> for (i = 0; i < mem->nregions; i++) {
>
> this function was never called, but I think we need to make each commit
> complete.
>From what I gather, both of your latest comments, for patches 4 and 5 are related to performance impact because of 128 memory regions. My thoughts on this are, I will work on an algorithm to optimize as you suggested.
Meanwhile we could approve patches 1-4. The rest of the implementation does not need 128 memory regions. The existing 8 regions are fine. I put in this change as I felt it makes the system scale more and was also suggested by Maxime. I think patches 1-4 stand by themselves and have been thoroughly tested. They could be merged in unless you have some other recommendations or changes you would like me to consider. Thoughts ?
>
> >
> >>
> >>> + int reg_msg_index = 0;
> >>> reg = &dev->mem->regions[i];
> >>> - memory->regions[i].userspace_addr = reg->host_user_addr;
> >>> + if (reg->host_user_addr == 0)
> >>> + continue;
> >>> + memory->regions[reg_msg_index].userspace_addr = reg-
> >>> host_user_addr;
> >>> + reg_msg_index++;
> >>> }
> >>>
> >>> /* Send the addresses back to qemu */ @@ -1279,8 +1293,10 @@
> >>> vhost_user_postcopy_register(struct virtio_net *dev, int main_fd,
> >>> }
> >>>
> >>> /* Now userfault register and we can use the memory */
> >>> - for (i = 0; i < memory->nregions; i++) {
> >>> + for (i = 0; i < VHOST_MEMORY_MAX_NREGIONS; i++) {
> >>> reg = &dev->mem->regions[i];
> >>> + if (reg->host_user_addr == 0)
> >>> + continue;
> >>> if (vhost_user_postcopy_region_register(dev, reg) < 0)
> >>> return -1;
> >>> }
> >>> @@ -1385,6 +1401,46 @@ vhost_user_mmap_region(struct virtio_net
> *dev,
> >>> return 0;
> >>> }
> >>>
> >>> +static int
> >>> +vhost_user_initialize_memory(struct virtio_net **pdev)
> >>
> >> This function should be part of 3/5, else the 3/5 will compile fail
> >
> > I have moved support functions to patch-3. Please review v4.
> >
> >>
> >>> +{
> >>> + struct virtio_net *dev = *pdev;
> >>> + int numa_node = SOCKET_ID_ANY;
> >>> +
> >>> + /*
> >>> + * If VQ 0 has already been allocated, try to allocate on the same
> >>> + * NUMA node. It can be reallocated later in numa_realloc().
> >>> + */
> >>> + if (dev->nr_vring > 0)
> >>> + numa_node = dev->virtqueue[0]->numa_node;
> >>> +
> >>> + dev->nr_guest_pages = 0;
> >>> + if (dev->guest_pages == NULL) {
> >>> + dev->max_guest_pages = 8;
> >>
> >> It should be VHOST_MEMORY_MAX_NREGIONS
> >
> > Done. Please review v4.
> >
> >>
> >>> + dev->guest_pages = rte_zmalloc_socket(NULL,
> >>> + dev->max_guest_pages *
> >>> + sizeof(struct guest_page),
> >>> + RTE_CACHE_LINE_SIZE,
> >>> + numa_node);
> >>> + if (dev->guest_pages == NULL) {
> >>> + VHOST_CONFIG_LOG(dev->ifname, ERR,
> >>> + "failed to allocate memory for dev-
> >>> guest_pages");
> >>> + return -1;
> >>> + }
> >>> + }
> >>> +
> >>> + dev->mem = rte_zmalloc_socket("vhost-mem-table", sizeof(struct
> >> rte_vhost_memory) +
> >>> + sizeof(struct rte_vhost_mem_region) *
> >> VHOST_MEMORY_MAX_NREGIONS, 0, numa_node);
> >>> + if (dev->mem == NULL) {
> >>> + VHOST_CONFIG_LOG(dev->ifname, ERR, "failed to allocate
> >> memory for dev->mem");
> >>> + rte_free(dev->guest_pages);
> >>> + dev->guest_pages = NULL;
> >>> + return -1;
> >>> + }
> >>> +
> >>> + return 0;
> >>> +}
> >>> +
> >>> static int
> >>> vhost_user_set_mem_table(struct virtio_net **pdev,
> >>> struct vhu_msg_context *ctx,
> >
> >
More information about the dev
mailing list