This series tries to access virtqueue metadata through kernel virtual address instead of copy_user() friends since they had too much overheads like checks, spec barriers or even hardware feature toggling. Test shows about 24% improvement on TX PPS. It should benefit other cases as well. Changes from V2: - fix buggy range overlapping check - tear down MMU notifier during vhost ioctl to make sure invalidation request can read metadata userspace address and vq size without holding vq mutex. Changes from V1: - instead of pinning pages, use MMU notifier to invalidate vmaps and remap duing metadata prefetch - fix build warning on MIPS Jason Wang (5): vhost: generalize adding used elem vhost: fine grain userspace memory accessors vhost: rename vq_iotlb_prefetch() to vq_meta_prefetch() vhost: introduce helpers to get the size of metadata area vhost: access vq metadata through kernel virtual address drivers/vhost/net.c | 4 +- drivers/vhost/vhost.c | 416 +++++++++++++++++++++++++++++++++++++----- drivers/vhost/vhost.h | 15 +- 3 files changed, 384 insertions(+), 51 deletions(-) -- 2.17.1
Use one generic vhost_copy_to_user() instead of two dedicated accessor. This will simplify the conversion to fine grain accessors. About 2% improvement of PPS were seen during vitio-user txonly test. Signed-off-by: Jason Wang <jasowang at redhat.com> --- drivers/vhost/vhost.c | 11 +---------- 1 file changed, 1 insertion(+), 10 deletions(-) diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c index 55e5aa662ad5..f179b5ee14c4 100644 --- a/drivers/vhost/vhost.c +++ b/drivers/vhost/vhost.c @@ -2174,16 +2174,7 @@ static int __vhost_add_used_n(struct vhost_virtqueue *vq, start = vq->last_used_idx & (vq->num - 1); used = vq->used->ring + start; - if (count == 1) { - if (vhost_put_user(vq, heads[0].id, &used->id)) { - vq_err(vq, "Failed to write used id"); - return -EFAULT; - } - if (vhost_put_user(vq, heads[0].len, &used->len)) { - vq_err(vq, "Failed to write used len"); - return -EFAULT; - } - } else if (vhost_copy_to_user(vq, used, heads, count * sizeof *used)) { + if (vhost_copy_to_user(vq, used, heads, count * sizeof *used)) { vq_err(vq, "Failed to write used"); return -EFAULT; } -- 2.17.1
Jason Wang
2018-Dec-29 12:46 UTC
[RFC PATCH V3 2/5] vhost: fine grain userspace memory accessors
This is used to hide the metadata address from virtqueue helpers. This will allow to implement a vmap based fast accessing to metadata. Signed-off-by: Jason Wang <jasowang at redhat.com> --- drivers/vhost/vhost.c | 94 +++++++++++++++++++++++++++++++++++-------- 1 file changed, 77 insertions(+), 17 deletions(-) diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c index f179b5ee14c4..337ce6f5a098 100644 --- a/drivers/vhost/vhost.c +++ b/drivers/vhost/vhost.c @@ -868,6 +868,34 @@ static inline void __user *__vhost_get_user(struct vhost_virtqueue *vq, ret; \ }) +static inline int vhost_put_avail_event(struct vhost_virtqueue *vq) +{ + return vhost_put_user(vq, cpu_to_vhost16(vq, vq->avail_idx), + vhost_avail_event(vq)); +} + +static inline int vhost_put_used(struct vhost_virtqueue *vq, + struct vring_used_elem *head, int idx, + int count) +{ + return vhost_copy_to_user(vq, vq->used->ring + idx, head, + count * sizeof(*head)); +} + +static inline int vhost_put_used_flags(struct vhost_virtqueue *vq) + +{ + return vhost_put_user(vq, cpu_to_vhost16(vq, vq->used_flags), + &vq->used->flags); +} + +static inline int vhost_put_used_idx(struct vhost_virtqueue *vq) + +{ + return vhost_put_user(vq, cpu_to_vhost16(vq, vq->last_used_idx), + &vq->used->idx); +} + #define vhost_get_user(vq, x, ptr, type) \ ({ \ int ret; \ @@ -906,6 +934,43 @@ static void vhost_dev_unlock_vqs(struct vhost_dev *d) mutex_unlock(&d->vqs[i]->mutex); } +static inline int vhost_get_avail_idx(struct vhost_virtqueue *vq, + __virtio16 *idx) +{ + return vhost_get_avail(vq, *idx, &vq->avail->idx); +} + +static inline int vhost_get_avail_head(struct vhost_virtqueue *vq, + __virtio16 *head, int idx) +{ + return vhost_get_avail(vq, *head, + &vq->avail->ring[idx & (vq->num - 1)]); +} + +static inline int vhost_get_avail_flags(struct vhost_virtqueue *vq, + __virtio16 *flags) +{ + return vhost_get_avail(vq, *flags, &vq->avail->flags); +} + +static inline int vhost_get_used_event(struct vhost_virtqueue *vq, + __virtio16 *event) +{ + return vhost_get_avail(vq, *event, vhost_used_event(vq)); +} + +static inline int vhost_get_used_idx(struct vhost_virtqueue *vq, + __virtio16 *idx) +{ + return vhost_get_used(vq, *idx, &vq->used->idx); +} + +static inline int vhost_get_desc(struct vhost_virtqueue *vq, + struct vring_desc *desc, int idx) +{ + return vhost_copy_from_user(vq, desc, vq->desc + idx, sizeof(*desc)); +} + static int vhost_new_umem_range(struct vhost_umem *umem, u64 start, u64 size, u64 end, u64 userspace_addr, int perm) @@ -1761,8 +1826,7 @@ EXPORT_SYMBOL_GPL(vhost_log_write); static int vhost_update_used_flags(struct vhost_virtqueue *vq) { void __user *used; - if (vhost_put_user(vq, cpu_to_vhost16(vq, vq->used_flags), - &vq->used->flags) < 0) + if (vhost_put_used_flags(vq)) return -EFAULT; if (unlikely(vq->log_used)) { /* Make sure the flag is seen before log. */ @@ -1780,8 +1844,7 @@ static int vhost_update_used_flags(struct vhost_virtqueue *vq) static int vhost_update_avail_event(struct vhost_virtqueue *vq, u16 avail_event) { - if (vhost_put_user(vq, cpu_to_vhost16(vq, vq->avail_idx), - vhost_avail_event(vq))) + if (vhost_put_avail_event(vq)) return -EFAULT; if (unlikely(vq->log_used)) { void __user *used; @@ -1818,7 +1881,7 @@ int vhost_vq_init_access(struct vhost_virtqueue *vq) r = -EFAULT; goto err; } - r = vhost_get_used(vq, last_used_idx, &vq->used->idx); + r = vhost_get_used_idx(vq, &last_used_idx); if (r) { vq_err(vq, "Can't access used idx at %p\n", &vq->used->idx); @@ -2017,7 +2080,7 @@ int vhost_get_vq_desc(struct vhost_virtqueue *vq, last_avail_idx = vq->last_avail_idx; if (vq->avail_idx == vq->last_avail_idx) { - if (unlikely(vhost_get_avail(vq, avail_idx, &vq->avail->idx))) { + if (unlikely(vhost_get_avail_idx(vq, &avail_idx))) { vq_err(vq, "Failed to access avail idx at %p\n", &vq->avail->idx); return -EFAULT; @@ -2044,8 +2107,7 @@ int vhost_get_vq_desc(struct vhost_virtqueue *vq, /* Grab the next descriptor number they're advertising, and increment * the index we've seen. */ - if (unlikely(vhost_get_avail(vq, ring_head, - &vq->avail->ring[last_avail_idx & (vq->num - 1)]))) { + if (unlikely(vhost_get_avail_head(vq, &ring_head, last_avail_idx))) { vq_err(vq, "Failed to read head: idx %d address %p\n", last_avail_idx, &vq->avail->ring[last_avail_idx % vq->num]); @@ -2080,8 +2142,7 @@ int vhost_get_vq_desc(struct vhost_virtqueue *vq, i, vq->num, head); return -EINVAL; } - ret = vhost_copy_from_user(vq, &desc, vq->desc + i, - sizeof desc); + ret = vhost_get_desc(vq, &desc, i); if (unlikely(ret)) { vq_err(vq, "Failed to get descriptor: idx %d addr %p\n", i, vq->desc + i); @@ -2174,7 +2235,7 @@ static int __vhost_add_used_n(struct vhost_virtqueue *vq, start = vq->last_used_idx & (vq->num - 1); used = vq->used->ring + start; - if (vhost_copy_to_user(vq, used, heads, count * sizeof *used)) { + if (vhost_put_used(vq, heads, start, count)) { vq_err(vq, "Failed to write used"); return -EFAULT; } @@ -2218,8 +2279,7 @@ int vhost_add_used_n(struct vhost_virtqueue *vq, struct vring_used_elem *heads, /* Make sure buffer is written before we update index. */ smp_wmb(); - if (vhost_put_user(vq, cpu_to_vhost16(vq, vq->last_used_idx), - &vq->used->idx)) { + if (vhost_put_used_idx(vq)) { vq_err(vq, "Failed to increment used idx"); return -EFAULT; } @@ -2253,7 +2313,7 @@ static bool vhost_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq) if (!vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX)) { __virtio16 flags; - if (vhost_get_avail(vq, flags, &vq->avail->flags)) { + if (vhost_get_avail_flags(vq, &flags)) { vq_err(vq, "Failed to get flags"); return true; } @@ -2267,7 +2327,7 @@ static bool vhost_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq) if (unlikely(!v)) return true; - if (vhost_get_avail(vq, event, vhost_used_event(vq))) { + if (vhost_get_used_event(vq, &event)) { vq_err(vq, "Failed to get used event idx"); return true; } @@ -2312,7 +2372,7 @@ bool vhost_vq_avail_empty(struct vhost_dev *dev, struct vhost_virtqueue *vq) if (vq->avail_idx != vq->last_avail_idx) return false; - r = vhost_get_avail(vq, avail_idx, &vq->avail->idx); + r = vhost_get_avail_idx(vq, &avail_idx); if (unlikely(r)) return false; vq->avail_idx = vhost16_to_cpu(vq, avail_idx); @@ -2348,7 +2408,7 @@ bool vhost_enable_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq) /* They could have slipped one in as we were doing that: make * sure it's written, then check again. */ smp_mb(); - r = vhost_get_avail(vq, avail_idx, &vq->avail->idx); + r = vhost_get_avail_idx(vq, &avail_idx); if (r) { vq_err(vq, "Failed to check avail idx at %p: %d\n", &vq->avail->idx, r); -- 2.17.1
Jason Wang
2018-Dec-29 12:46 UTC
[RFC PATCH V3 3/5] vhost: rename vq_iotlb_prefetch() to vq_meta_prefetch()
Rename the function to be more accurate since it actually tries to prefetch vq metadata address in IOTLB. And this will be used by following patch to prefetch metadata virtual addresses. Signed-off-by: Jason Wang <jasowang at redhat.com> --- drivers/vhost/net.c | 4 ++-- drivers/vhost/vhost.c | 4 ++-- drivers/vhost/vhost.h | 2 +- 3 files changed, 5 insertions(+), 5 deletions(-) diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c index 36f3d0f49e60..0b4b3deab5aa 100644 --- a/drivers/vhost/net.c +++ b/drivers/vhost/net.c @@ -971,7 +971,7 @@ static void handle_tx(struct vhost_net *net) if (!sock) goto out; - if (!vq_iotlb_prefetch(vq)) + if (!vq_meta_prefetch(vq)) goto out; vhost_disable_notify(&net->dev, vq); @@ -1140,7 +1140,7 @@ static void handle_rx(struct vhost_net *net) if (!sock) goto out; - if (!vq_iotlb_prefetch(vq)) + if (!vq_meta_prefetch(vq)) goto out; vhost_disable_notify(&net->dev, vq); diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c index 337ce6f5a098..27b5c03feaac 100644 --- a/drivers/vhost/vhost.c +++ b/drivers/vhost/vhost.c @@ -1304,7 +1304,7 @@ static bool iotlb_access_ok(struct vhost_virtqueue *vq, return true; } -int vq_iotlb_prefetch(struct vhost_virtqueue *vq) +int vq_meta_prefetch(struct vhost_virtqueue *vq) { size_t s = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0; unsigned int num = vq->num; @@ -1323,7 +1323,7 @@ int vq_iotlb_prefetch(struct vhost_virtqueue *vq) num * sizeof(*vq->used->ring) + s, VHOST_ADDR_USED); } -EXPORT_SYMBOL_GPL(vq_iotlb_prefetch); +EXPORT_SYMBOL_GPL(vq_meta_prefetch); /* Can we log writes? */ /* Caller should have device mutex but not vq mutex */ diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h index 466ef7542291..0d1ff977a43e 100644 --- a/drivers/vhost/vhost.h +++ b/drivers/vhost/vhost.h @@ -206,7 +206,7 @@ bool vhost_enable_notify(struct vhost_dev *, struct vhost_virtqueue *); int vhost_log_write(struct vhost_virtqueue *vq, struct vhost_log *log, unsigned int log_num, u64 len); -int vq_iotlb_prefetch(struct vhost_virtqueue *vq); +int vq_meta_prefetch(struct vhost_virtqueue *vq); struct vhost_msg_node *vhost_new_msg(struct vhost_virtqueue *vq, int type); void vhost_enqueue_msg(struct vhost_dev *dev, -- 2.17.1
Jason Wang
2018-Dec-29 12:46 UTC
[RFC PATCH V3 4/5] vhost: introduce helpers to get the size of metadata area
Signed-off-by: Jason Wang <jasowang at redhat.com> --- drivers/vhost/vhost.c | 46 ++++++++++++++++++++++++++----------------- 1 file changed, 28 insertions(+), 18 deletions(-) diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c index 27b5c03feaac..54b43feef8d9 100644 --- a/drivers/vhost/vhost.c +++ b/drivers/vhost/vhost.c @@ -489,6 +489,27 @@ bool vhost_dev_has_owner(struct vhost_dev *dev) } EXPORT_SYMBOL_GPL(vhost_dev_has_owner); +static size_t vhost_get_avail_size(struct vhost_virtqueue *vq, int num) +{ + size_t event = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0; + + return sizeof(*vq->avail) + + sizeof(*vq->avail->ring) * num + event; +} + +static size_t vhost_get_used_size(struct vhost_virtqueue *vq, int num) +{ + size_t event = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0; + + return sizeof(*vq->used) + + sizeof(*vq->used->ring) * num + event; +} + +static size_t vhost_get_desc_size(struct vhost_virtqueue *vq, int num) +{ + return sizeof(*vq->desc) * num; +} + /* Caller should have device mutex */ long vhost_dev_set_owner(struct vhost_dev *dev) { @@ -1248,13 +1269,9 @@ static bool vq_access_ok(struct vhost_virtqueue *vq, unsigned int num, struct vring_used __user *used) { - size_t s = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0; - - return access_ok(VERIFY_READ, desc, num * sizeof *desc) && - access_ok(VERIFY_READ, avail, - sizeof *avail + num * sizeof *avail->ring + s) && - access_ok(VERIFY_WRITE, used, - sizeof *used + num * sizeof *used->ring + s); + return access_ok(VERIFY_READ, desc, vhost_get_desc_size(vq, num)) && + access_ok(VERIFY_READ, avail, vhost_get_avail_size(vq, num)) && + access_ok(VERIFY_WRITE, used, vhost_get_used_size(vq, num)); } static void vhost_vq_meta_update(struct vhost_virtqueue *vq, @@ -1306,22 +1323,18 @@ static bool iotlb_access_ok(struct vhost_virtqueue *vq, int vq_meta_prefetch(struct vhost_virtqueue *vq) { - size_t s = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0; unsigned int num = vq->num; if (!vq->iotlb) return 1; return iotlb_access_ok(vq, VHOST_ACCESS_RO, (u64)(uintptr_t)vq->desc, - num * sizeof(*vq->desc), VHOST_ADDR_DESC) && + vhost_get_desc_size(vq, num), VHOST_ADDR_DESC) && iotlb_access_ok(vq, VHOST_ACCESS_RO, (u64)(uintptr_t)vq->avail, - sizeof *vq->avail + - num * sizeof(*vq->avail->ring) + s, + vhost_get_avail_size(vq, num), VHOST_ADDR_AVAIL) && iotlb_access_ok(vq, VHOST_ACCESS_WO, (u64)(uintptr_t)vq->used, - sizeof *vq->used + - num * sizeof(*vq->used->ring) + s, - VHOST_ADDR_USED); + vhost_get_used_size(vq, num), VHOST_ADDR_USED); } EXPORT_SYMBOL_GPL(vq_meta_prefetch); @@ -1338,13 +1351,10 @@ EXPORT_SYMBOL_GPL(vhost_log_access_ok); static bool vq_log_access_ok(struct vhost_virtqueue *vq, void __user *log_base) { - size_t s = vhost_has_feature(vq, VIRTIO_RING_F_EVENT_IDX) ? 2 : 0; - return vq_memory_access_ok(log_base, vq->umem, vhost_has_feature(vq, VHOST_F_LOG_ALL)) && (!vq->log_used || log_access_ok(log_base, vq->log_addr, - sizeof *vq->used + - vq->num * sizeof *vq->used->ring + s)); + vhost_get_used_size(vq, vq->num))); } /* Can we start vq? */ -- 2.17.1
Jason Wang
2018-Dec-29 12:46 UTC
[RFC PATCH V3 5/5] vhost: access vq metadata through kernel virtual address
It was noticed that the copy_user() friends that was used to access virtqueue metdata tends to be very expensive for dataplane implementation like vhost since it involves lots of software checks, speculation barrier, hardware feature toggling (e.g SMAP). The extra cost will be more obvious when transferring small packets since the time spent on metadata accessing become significant.. This patch tries to eliminate those overhead by accessing them through kernel virtual address by vmap(). To make the pages can be migrated, instead of pinning them through GUP, we use mmu notifiers to invalidate vmaps and re-establish vmaps during each round of metadata prefetching in necessary. For devices that doesn't use metadata prefetching, the memory acessors fallback to normal copy_user() implementation gracefully. The invalidation was synchronized with datapath through vq mutex, and in order to avoid hold vq mutex during range checking, MMU notifier was teared down when trying to modify vq metadata. Note that this was only done when device IOTLB is not enabled. We could use similar method to optimize it in the future. Tests shows about ~24% improvement on TX PPS when using virtio-user + vhost_net + xdp1 on TAP: Before: ~5.0Mpps After: ~6.1Mpps Signed-off-by: Jason Wang <jasowang at redhat.com> --- drivers/vhost/vhost.c | 263 +++++++++++++++++++++++++++++++++++++++++- drivers/vhost/vhost.h | 13 +++ 2 files changed, 274 insertions(+), 2 deletions(-) diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c index 54b43feef8d9..e1ecb8acf8a3 100644 --- a/drivers/vhost/vhost.c +++ b/drivers/vhost/vhost.c @@ -440,6 +440,9 @@ void vhost_dev_init(struct vhost_dev *dev, vq->indirect = NULL; vq->heads = NULL; vq->dev = dev; + memset(&vq->avail_ring, 0, sizeof(vq->avail_ring)); + memset(&vq->used_ring, 0, sizeof(vq->used_ring)); + memset(&vq->desc_ring, 0, sizeof(vq->desc_ring)); mutex_init(&vq->mutex); vhost_vq_reset(dev, vq); if (vq->handle_kick) @@ -510,6 +513,73 @@ static size_t vhost_get_desc_size(struct vhost_virtqueue *vq, int num) return sizeof(*vq->desc) * num; } +static void vhost_uninit_vmap(struct vhost_vmap *map) +{ + if (map->addr) + vunmap(map->unmap_addr); + + map->addr = NULL; + map->unmap_addr = NULL; +} + +static int vhost_invalidate_vmap(struct vhost_virtqueue *vq, + struct vhost_vmap *map, + unsigned long ustart, + size_t size, + unsigned long start, + unsigned long end, + bool blockable) +{ + if (end < ustart || start > ustart - 1 + size) + return 0; + + if (!blockable) + return -EAGAIN; + + mutex_lock(&vq->mutex); + vhost_uninit_vmap(map); + mutex_unlock(&vq->mutex); + + return 0; +} + +static int vhost_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, + struct mm_struct *mm, + unsigned long start, + unsigned long end, + bool blockable) +{ + struct vhost_dev *dev = container_of(mn, struct vhost_dev, + mmu_notifier); + int i; + + for (i = 0; i < dev->nvqs; i++) { + struct vhost_virtqueue *vq = dev->vqs[i]; + + if (vhost_invalidate_vmap(vq, &vq->avail_ring, + (unsigned long)vq->avail, + vhost_get_avail_size(vq, vq->num), + start, end, blockable)) + return -EAGAIN; + if (vhost_invalidate_vmap(vq, &vq->desc_ring, + (unsigned long)vq->desc, + vhost_get_desc_size(vq, vq->num), + start, end, blockable)) + return -EAGAIN; + if (vhost_invalidate_vmap(vq, &vq->used_ring, + (unsigned long)vq->used, + vhost_get_used_size(vq, vq->num), + start, end, blockable)) + return -EAGAIN; + } + + return 0; +} + +static const struct mmu_notifier_ops vhost_mmu_notifier_ops = { + .invalidate_range_start = vhost_mmu_notifier_invalidate_range_start, +}; + /* Caller should have device mutex */ long vhost_dev_set_owner(struct vhost_dev *dev) { @@ -541,7 +611,14 @@ long vhost_dev_set_owner(struct vhost_dev *dev) if (err) goto err_cgroup; + dev->mmu_notifier.ops = &vhost_mmu_notifier_ops; + err = mmu_notifier_register(&dev->mmu_notifier, dev->mm); + if (err) + goto err_mmu_notifier; + return 0; +err_mmu_notifier: + vhost_dev_free_iovecs(dev); err_cgroup: kthread_stop(worker); dev->worker = NULL; @@ -632,6 +709,72 @@ static void vhost_clear_msg(struct vhost_dev *dev) spin_unlock(&dev->iotlb_lock); } +static int vhost_init_vmap(struct vhost_vmap *map, unsigned long uaddr, + size_t size, int write) +{ + struct page **pages; + int npages = DIV_ROUND_UP(size, PAGE_SIZE); + int npinned; + void *vaddr; + int err = 0; + + pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL); + if (!pages) + return -ENOMEM; + + npinned = get_user_pages_fast(uaddr, npages, write, pages); + if (npinned != npages) { + err = -EFAULT; + goto err; + } + + vaddr = vmap(pages, npages, VM_MAP, PAGE_KERNEL); + if (!vaddr) { + err = EFAULT; + goto err; + } + + map->addr = vaddr + (uaddr & (PAGE_SIZE - 1)); + map->unmap_addr = vaddr; + +err: + /* Don't pin pages, mmu notifier will notify us about page + * migration. + */ + if (npinned > 0) + release_pages(pages, npinned); + kfree(pages); + return err; +} + +static void vhost_clean_vmaps(struct vhost_virtqueue *vq) +{ + vhost_uninit_vmap(&vq->avail_ring); + vhost_uninit_vmap(&vq->desc_ring); + vhost_uninit_vmap(&vq->used_ring); +} + +static int vhost_setup_avail_vmap(struct vhost_virtqueue *vq, + unsigned long avail) +{ + return vhost_init_vmap(&vq->avail_ring, avail, + vhost_get_avail_size(vq, vq->num), false); +} + +static int vhost_setup_desc_vmap(struct vhost_virtqueue *vq, + unsigned long desc) +{ + return vhost_init_vmap(&vq->desc_ring, desc, + vhost_get_desc_size(vq, vq->num), false); +} + +static int vhost_setup_used_vmap(struct vhost_virtqueue *vq, + unsigned long used) +{ + return vhost_init_vmap(&vq->used_ring, used, + vhost_get_used_size(vq, vq->num), true); +} + void vhost_dev_cleanup(struct vhost_dev *dev) { int i; @@ -661,8 +804,12 @@ void vhost_dev_cleanup(struct vhost_dev *dev) kthread_stop(dev->worker); dev->worker = NULL; } - if (dev->mm) + if (dev->mm) { + mmu_notifier_unregister(&dev->mmu_notifier, dev->mm); mmput(dev->mm); + } + for (i = 0; i < dev->nvqs; i++) + vhost_clean_vmaps(dev->vqs[i]); dev->mm = NULL; } EXPORT_SYMBOL_GPL(vhost_dev_cleanup); @@ -891,6 +1038,16 @@ static inline void __user *__vhost_get_user(struct vhost_virtqueue *vq, static inline int vhost_put_avail_event(struct vhost_virtqueue *vq) { + if (!vq->iotlb) { + struct vring_used *used = vq->used_ring.addr; + + if (likely(used)) { + *((__virtio16 *)&used->ring[vq->num]) + cpu_to_vhost16(vq, vq->avail_idx); + return 0; + } + } + return vhost_put_user(vq, cpu_to_vhost16(vq, vq->avail_idx), vhost_avail_event(vq)); } @@ -899,6 +1056,16 @@ static inline int vhost_put_used(struct vhost_virtqueue *vq, struct vring_used_elem *head, int idx, int count) { + if (!vq->iotlb) { + struct vring_used *used = vq->used_ring.addr; + + if (likely(used)) { + memcpy(used->ring + idx, head, + count * sizeof(*head)); + return 0; + } + } + return vhost_copy_to_user(vq, vq->used->ring + idx, head, count * sizeof(*head)); } @@ -906,6 +1073,15 @@ static inline int vhost_put_used(struct vhost_virtqueue *vq, static inline int vhost_put_used_flags(struct vhost_virtqueue *vq) { + if (!vq->iotlb) { + struct vring_used *used = vq->used_ring.addr; + + if (likely(used)) { + used->flags = cpu_to_vhost16(vq, vq->used_flags); + return 0; + } + } + return vhost_put_user(vq, cpu_to_vhost16(vq, vq->used_flags), &vq->used->flags); } @@ -913,6 +1089,15 @@ static inline int vhost_put_used_flags(struct vhost_virtqueue *vq) static inline int vhost_put_used_idx(struct vhost_virtqueue *vq) { + if (!vq->iotlb) { + struct vring_used *used = vq->used_ring.addr; + + if (likely(used)) { + used->idx = cpu_to_vhost16(vq, vq->last_used_idx); + return 0; + } + } + return vhost_put_user(vq, cpu_to_vhost16(vq, vq->last_used_idx), &vq->used->idx); } @@ -958,12 +1143,30 @@ static void vhost_dev_unlock_vqs(struct vhost_dev *d) static inline int vhost_get_avail_idx(struct vhost_virtqueue *vq, __virtio16 *idx) { + if (!vq->iotlb) { + struct vring_avail *avail = vq->avail_ring.addr; + + if (likely(avail)) { + *idx = avail->idx; + return 0; + } + } + return vhost_get_avail(vq, *idx, &vq->avail->idx); } static inline int vhost_get_avail_head(struct vhost_virtqueue *vq, __virtio16 *head, int idx) { + if (!vq->iotlb) { + struct vring_avail *avail = vq->avail_ring.addr; + + if (likely(avail)) { + *head = avail->ring[idx & (vq->num - 1)]; + return 0; + } + } + return vhost_get_avail(vq, *head, &vq->avail->ring[idx & (vq->num - 1)]); } @@ -971,24 +1174,60 @@ static inline int vhost_get_avail_head(struct vhost_virtqueue *vq, static inline int vhost_get_avail_flags(struct vhost_virtqueue *vq, __virtio16 *flags) { + if (!vq->iotlb) { + struct vring_avail *avail = vq->avail_ring.addr; + + if (likely(avail)) { + *flags = avail->flags; + return 0; + } + } + return vhost_get_avail(vq, *flags, &vq->avail->flags); } static inline int vhost_get_used_event(struct vhost_virtqueue *vq, __virtio16 *event) { + if (!vq->iotlb) { + struct vring_avail *avail = vq->avail_ring.addr; + + if (likely(avail)) { + *event = (__virtio16)avail->ring[vq->num]; + return 0; + } + } + return vhost_get_avail(vq, *event, vhost_used_event(vq)); } static inline int vhost_get_used_idx(struct vhost_virtqueue *vq, __virtio16 *idx) { + if (!vq->iotlb) { + struct vring_used *used = vq->used_ring.addr; + + if (likely(used)) { + *idx = used->idx; + return 0; + } + } + return vhost_get_used(vq, *idx, &vq->used->idx); } static inline int vhost_get_desc(struct vhost_virtqueue *vq, struct vring_desc *desc, int idx) { + if (!vq->iotlb) { + struct vring_desc *d = vq->desc_ring.addr; + + if (likely(d)) { + *desc = *(d + idx); + return 0; + } + } + return vhost_copy_from_user(vq, desc, vq->desc + idx, sizeof(*desc)); } @@ -1325,8 +1564,16 @@ int vq_meta_prefetch(struct vhost_virtqueue *vq) { unsigned int num = vq->num; - if (!vq->iotlb) + if (!vq->iotlb) { + if (unlikely(!vq->avail_ring.addr)) + vhost_setup_avail_vmap(vq, (unsigned long)vq->avail); + if (unlikely(!vq->desc_ring.addr)) + vhost_setup_desc_vmap(vq, (unsigned long)vq->desc); + if (unlikely(!vq->used_ring.addr)) + vhost_setup_used_vmap(vq, (unsigned long)vq->used); + return 1; + } return iotlb_access_ok(vq, VHOST_ACCESS_RO, (u64)(uintptr_t)vq->desc, vhost_get_desc_size(vq, num), VHOST_ADDR_DESC) && @@ -1478,6 +1725,13 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg mutex_lock(&vq->mutex); + /* Unregister MMU notifer to allow invalidation callback + * can access vq->avail, vq->desc , vq->used and vq->num + * without holding vq->mutex. + */ + if (d->mm) + mmu_notifier_unregister(&d->mmu_notifier, d->mm); + switch (ioctl) { case VHOST_SET_VRING_NUM: /* Resizing ring with an active backend? @@ -1494,6 +1748,7 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg r = -EINVAL; break; } + vhost_clean_vmaps(vq); vq->num = s.num; break; case VHOST_SET_VRING_BASE: @@ -1571,6 +1826,8 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg } } + vhost_clean_vmaps(vq); + vq->log_used = !!(a.flags & (0x1 << VHOST_VRING_F_LOG)); vq->desc = (void __user *)(unsigned long)a.desc_user_addr; vq->avail = (void __user *)(unsigned long)a.avail_user_addr; @@ -1651,6 +1908,8 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg if (pollstart && vq->handle_kick) r = vhost_poll_start(&vq->poll, vq->kick); + if (d->mm) + mmu_notifier_register(&d->mmu_notifier, d->mm); mutex_unlock(&vq->mutex); if (pollstop && vq->handle_kick) diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h index 0d1ff977a43e..00f016a4f198 100644 --- a/drivers/vhost/vhost.h +++ b/drivers/vhost/vhost.h @@ -12,6 +12,8 @@ #include <linux/virtio_config.h> #include <linux/virtio_ring.h> #include <linux/atomic.h> +#include <linux/pagemap.h> +#include <linux/mmu_notifier.h> struct vhost_work; typedef void (*vhost_work_fn_t)(struct vhost_work *work); @@ -80,6 +82,11 @@ enum vhost_uaddr_type { VHOST_NUM_ADDRS = 3, }; +struct vhost_vmap { + void *addr; + void *unmap_addr; +}; + /* The virtqueue structure describes a queue attached to a device. */ struct vhost_virtqueue { struct vhost_dev *dev; @@ -90,6 +97,11 @@ struct vhost_virtqueue { struct vring_desc __user *desc; struct vring_avail __user *avail; struct vring_used __user *used; + + struct vhost_vmap avail_ring; + struct vhost_vmap desc_ring; + struct vhost_vmap used_ring; + const struct vhost_umem_node *meta_iotlb[VHOST_NUM_ADDRS]; struct file *kick; struct eventfd_ctx *call_ctx; @@ -158,6 +170,7 @@ struct vhost_msg_node { struct vhost_dev { struct mm_struct *mm; + struct mmu_notifier mmu_notifier; struct mutex mutex; struct vhost_virtqueue **vqs; int nvqs; -- 2.17.1
On Sat, Dec 29, 2018 at 08:46:51PM +0800, Jason Wang wrote:> This series tries to access virtqueue metadata through kernel virtual > address instead of copy_user() friends since they had too much > overheads like checks, spec barriers or even hardware feature > toggling.Will review, thanks! One questions that comes to mind is whether it's all about bypassing stac/clac. Could you please include a performance comparison with nosmap?> > Test shows about 24% improvement on TX PPS. It should benefit other > cases as well. > > Changes from V2: > - fix buggy range overlapping check > - tear down MMU notifier during vhost ioctl to make sure invalidation > request can read metadata userspace address and vq size without > holding vq mutex. > Changes from V1: > - instead of pinning pages, use MMU notifier to invalidate vmaps and > remap duing metadata prefetch > - fix build warning on MIPS > > Jason Wang (5): > vhost: generalize adding used elem > vhost: fine grain userspace memory accessors > vhost: rename vq_iotlb_prefetch() to vq_meta_prefetch() > vhost: introduce helpers to get the size of metadata area > vhost: access vq metadata through kernel virtual address > > drivers/vhost/net.c | 4 +- > drivers/vhost/vhost.c | 416 +++++++++++++++++++++++++++++++++++++----- > drivers/vhost/vhost.h | 15 +- > 3 files changed, 384 insertions(+), 51 deletions(-) > > -- > 2.17.1
Michael S. Tsirkin
2019-Jan-04 21:29 UTC
[RFC PATCH V3 1/5] vhost: generalize adding used elem
On Sat, Dec 29, 2018 at 08:46:52PM +0800, Jason Wang wrote:> Use one generic vhost_copy_to_user() instead of two dedicated > accessor. This will simplify the conversion to fine grain > accessors. About 2% improvement of PPS were seen during vitio-user > txonly test. > > Signed-off-by: Jason Wang <jasowang at redhat.com>I don't hve a problem with this patch but do you have any idea how come removing what's supposed to be an optimization speeds things up?> --- > drivers/vhost/vhost.c | 11 +---------- > 1 file changed, 1 insertion(+), 10 deletions(-) > > diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c > index 55e5aa662ad5..f179b5ee14c4 100644 > --- a/drivers/vhost/vhost.c > +++ b/drivers/vhost/vhost.c > @@ -2174,16 +2174,7 @@ static int __vhost_add_used_n(struct vhost_virtqueue *vq, > > start = vq->last_used_idx & (vq->num - 1); > used = vq->used->ring + start; > - if (count == 1) { > - if (vhost_put_user(vq, heads[0].id, &used->id)) { > - vq_err(vq, "Failed to write used id"); > - return -EFAULT; > - } > - if (vhost_put_user(vq, heads[0].len, &used->len)) { > - vq_err(vq, "Failed to write used len"); > - return -EFAULT; > - } > - } else if (vhost_copy_to_user(vq, used, heads, count * sizeof *used)) { > + if (vhost_copy_to_user(vq, used, heads, count * sizeof *used)) { > vq_err(vq, "Failed to write used"); > return -EFAULT; > } > -- > 2.17.1
Michael S. Tsirkin
2019-Jan-04 21:34 UTC
[RFC PATCH V3 5/5] vhost: access vq metadata through kernel virtual address
On Sat, Dec 29, 2018 at 08:46:56PM +0800, Jason Wang wrote:> It was noticed that the copy_user() friends that was used to access > virtqueue metdata tends to be very expensive for dataplane > implementation like vhost since it involves lots of software checks, > speculation barrier, hardware feature toggling (e.g SMAP). The > extra cost will be more obvious when transferring small packets since > the time spent on metadata accessing become significant.. > > This patch tries to eliminate those overhead by accessing them through > kernel virtual address by vmap(). To make the pages can be migrated, > instead of pinning them through GUP, we use mmu notifiers to > invalidate vmaps and re-establish vmaps during each round of metadata > prefetching in necessary. For devices that doesn't use metadata > prefetching, the memory acessors fallback to normal copy_user() > implementation gracefully. The invalidation was synchronized with > datapath through vq mutex, and in order to avoid hold vq mutex during > range checking, MMU notifier was teared down when trying to modify vq > metadata. > > Note that this was only done when device IOTLB is not enabled. We > could use similar method to optimize it in the future. > > Tests shows about ~24% improvement on TX PPS when using virtio-user + > vhost_net + xdp1 on TAP: > > Before: ~5.0Mpps > After: ~6.1Mpps > > Signed-off-by: Jason Wang <jasowang at redhat.com> > --- > drivers/vhost/vhost.c | 263 +++++++++++++++++++++++++++++++++++++++++- > drivers/vhost/vhost.h | 13 +++ > 2 files changed, 274 insertions(+), 2 deletions(-) > > diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c > index 54b43feef8d9..e1ecb8acf8a3 100644 > --- a/drivers/vhost/vhost.c > +++ b/drivers/vhost/vhost.c > @@ -440,6 +440,9 @@ void vhost_dev_init(struct vhost_dev *dev, > vq->indirect = NULL; > vq->heads = NULL; > vq->dev = dev; > + memset(&vq->avail_ring, 0, sizeof(vq->avail_ring)); > + memset(&vq->used_ring, 0, sizeof(vq->used_ring)); > + memset(&vq->desc_ring, 0, sizeof(vq->desc_ring)); > mutex_init(&vq->mutex); > vhost_vq_reset(dev, vq); > if (vq->handle_kick) > @@ -510,6 +513,73 @@ static size_t vhost_get_desc_size(struct vhost_virtqueue *vq, int num) > return sizeof(*vq->desc) * num; > } > > +static void vhost_uninit_vmap(struct vhost_vmap *map) > +{ > + if (map->addr) > + vunmap(map->unmap_addr); > + > + map->addr = NULL; > + map->unmap_addr = NULL; > +} > + > +static int vhost_invalidate_vmap(struct vhost_virtqueue *vq, > + struct vhost_vmap *map, > + unsigned long ustart, > + size_t size, > + unsigned long start, > + unsigned long end, > + bool blockable) > +{ > + if (end < ustart || start > ustart - 1 + size) > + return 0; > + > + if (!blockable) > + return -EAGAIN; > + > + mutex_lock(&vq->mutex); > + vhost_uninit_vmap(map); > + mutex_unlock(&vq->mutex); > + > + return 0; > +} > + > +static int vhost_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, > + struct mm_struct *mm, > + unsigned long start, > + unsigned long end, > + bool blockable) > +{ > + struct vhost_dev *dev = container_of(mn, struct vhost_dev, > + mmu_notifier); > + int i; > + > + for (i = 0; i < dev->nvqs; i++) { > + struct vhost_virtqueue *vq = dev->vqs[i]; > + > + if (vhost_invalidate_vmap(vq, &vq->avail_ring, > + (unsigned long)vq->avail, > + vhost_get_avail_size(vq, vq->num), > + start, end, blockable)) > + return -EAGAIN; > + if (vhost_invalidate_vmap(vq, &vq->desc_ring, > + (unsigned long)vq->desc, > + vhost_get_desc_size(vq, vq->num), > + start, end, blockable)) > + return -EAGAIN; > + if (vhost_invalidate_vmap(vq, &vq->used_ring, > + (unsigned long)vq->used, > + vhost_get_used_size(vq, vq->num), > + start, end, blockable)) > + return -EAGAIN; > + } > + > + return 0; > +} > + > +static const struct mmu_notifier_ops vhost_mmu_notifier_ops = { > + .invalidate_range_start = vhost_mmu_notifier_invalidate_range_start, > +}; > + > /* Caller should have device mutex */ > long vhost_dev_set_owner(struct vhost_dev *dev) > { > @@ -541,7 +611,14 @@ long vhost_dev_set_owner(struct vhost_dev *dev) > if (err) > goto err_cgroup; > > + dev->mmu_notifier.ops = &vhost_mmu_notifier_ops; > + err = mmu_notifier_register(&dev->mmu_notifier, dev->mm); > + if (err) > + goto err_mmu_notifier; > + > return 0; > +err_mmu_notifier: > + vhost_dev_free_iovecs(dev); > err_cgroup: > kthread_stop(worker); > dev->worker = NULL; > @@ -632,6 +709,72 @@ static void vhost_clear_msg(struct vhost_dev *dev) > spin_unlock(&dev->iotlb_lock); > } > > +static int vhost_init_vmap(struct vhost_vmap *map, unsigned long uaddr, > + size_t size, int write) > +{ > + struct page **pages; > + int npages = DIV_ROUND_UP(size, PAGE_SIZE); > + int npinned; > + void *vaddr; > + int err = 0; > + > + pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL); > + if (!pages) > + return -ENOMEM; > + > + npinned = get_user_pages_fast(uaddr, npages, write, pages); > + if (npinned != npages) { > + err = -EFAULT; > + goto err; > + } > + > + vaddr = vmap(pages, npages, VM_MAP, PAGE_KERNEL); > + if (!vaddr) { > + err = EFAULT; > + goto err; > + } > + > + map->addr = vaddr + (uaddr & (PAGE_SIZE - 1)); > + map->unmap_addr = vaddr; > + > +err: > + /* Don't pin pages, mmu notifier will notify us about page > + * migration. > + */ > + if (npinned > 0) > + release_pages(pages, npinned); > + kfree(pages); > + return err; > +} > + > +static void vhost_clean_vmaps(struct vhost_virtqueue *vq) > +{ > + vhost_uninit_vmap(&vq->avail_ring); > + vhost_uninit_vmap(&vq->desc_ring); > + vhost_uninit_vmap(&vq->used_ring); > +} > + > +static int vhost_setup_avail_vmap(struct vhost_virtqueue *vq, > + unsigned long avail) > +{ > + return vhost_init_vmap(&vq->avail_ring, avail, > + vhost_get_avail_size(vq, vq->num), false); > +} > + > +static int vhost_setup_desc_vmap(struct vhost_virtqueue *vq, > + unsigned long desc) > +{ > + return vhost_init_vmap(&vq->desc_ring, desc, > + vhost_get_desc_size(vq, vq->num), false); > +} > + > +static int vhost_setup_used_vmap(struct vhost_virtqueue *vq, > + unsigned long used) > +{ > + return vhost_init_vmap(&vq->used_ring, used, > + vhost_get_used_size(vq, vq->num), true); > +} > + > void vhost_dev_cleanup(struct vhost_dev *dev) > { > int i; > @@ -661,8 +804,12 @@ void vhost_dev_cleanup(struct vhost_dev *dev) > kthread_stop(dev->worker); > dev->worker = NULL; > } > - if (dev->mm) > + if (dev->mm) { > + mmu_notifier_unregister(&dev->mmu_notifier, dev->mm); > mmput(dev->mm); > + } > + for (i = 0; i < dev->nvqs; i++) > + vhost_clean_vmaps(dev->vqs[i]); > dev->mm = NULL; > } > EXPORT_SYMBOL_GPL(vhost_dev_cleanup); > @@ -891,6 +1038,16 @@ static inline void __user *__vhost_get_user(struct vhost_virtqueue *vq, > > static inline int vhost_put_avail_event(struct vhost_virtqueue *vq) > { > + if (!vq->iotlb) {Do we have to limit this to !iotlb?> + struct vring_used *used = vq->used_ring.addr; > + > + if (likely(used)) { > + *((__virtio16 *)&used->ring[vq->num]) > + cpu_to_vhost16(vq, vq->avail_idx);So here we are modifying userspace memory without marking it dirty. Is this OK? And why?> + return 0; > + } > + } > + > return vhost_put_user(vq, cpu_to_vhost16(vq, vq->avail_idx), > vhost_avail_event(vq)); > } > @@ -899,6 +1056,16 @@ static inline int vhost_put_used(struct vhost_virtqueue *vq, > struct vring_used_elem *head, int idx, > int count) > { > + if (!vq->iotlb) { > + struct vring_used *used = vq->used_ring.addr; > + > + if (likely(used)) { > + memcpy(used->ring + idx, head, > + count * sizeof(*head)); > + return 0;Same here.> + } > + } > + > return vhost_copy_to_user(vq, vq->used->ring + idx, head, > count * sizeof(*head)); > } > @@ -906,6 +1073,15 @@ static inline int vhost_put_used(struct vhost_virtqueue *vq, > static inline int vhost_put_used_flags(struct vhost_virtqueue *vq) > > { > + if (!vq->iotlb) { > + struct vring_used *used = vq->used_ring.addr; > + > + if (likely(used)) { > + used->flags = cpu_to_vhost16(vq, vq->used_flags); > + return 0; > + } > + } > + > return vhost_put_user(vq, cpu_to_vhost16(vq, vq->used_flags), > &vq->used->flags); > } > @@ -913,6 +1089,15 @@ static inline int vhost_put_used_flags(struct vhost_virtqueue *vq) > static inline int vhost_put_used_idx(struct vhost_virtqueue *vq) > > { > + if (!vq->iotlb) { > + struct vring_used *used = vq->used_ring.addr; > + > + if (likely(used)) { > + used->idx = cpu_to_vhost16(vq, vq->last_used_idx); > + return 0; > + } > + } > + > return vhost_put_user(vq, cpu_to_vhost16(vq, vq->last_used_idx), > &vq->used->idx); > } > @@ -958,12 +1143,30 @@ static void vhost_dev_unlock_vqs(struct vhost_dev *d) > static inline int vhost_get_avail_idx(struct vhost_virtqueue *vq, > __virtio16 *idx) > { > + if (!vq->iotlb) { > + struct vring_avail *avail = vq->avail_ring.addr; > + > + if (likely(avail)) { > + *idx = avail->idx; > + return 0; > + } > + } > + > return vhost_get_avail(vq, *idx, &vq->avail->idx); > } > > static inline int vhost_get_avail_head(struct vhost_virtqueue *vq, > __virtio16 *head, int idx) > { > + if (!vq->iotlb) { > + struct vring_avail *avail = vq->avail_ring.addr; > + > + if (likely(avail)) { > + *head = avail->ring[idx & (vq->num - 1)]; > + return 0; > + } > + } > + > return vhost_get_avail(vq, *head, > &vq->avail->ring[idx & (vq->num - 1)]); > } > @@ -971,24 +1174,60 @@ static inline int vhost_get_avail_head(struct vhost_virtqueue *vq, > static inline int vhost_get_avail_flags(struct vhost_virtqueue *vq, > __virtio16 *flags) > { > + if (!vq->iotlb) { > + struct vring_avail *avail = vq->avail_ring.addr; > + > + if (likely(avail)) { > + *flags = avail->flags; > + return 0; > + } > + } > + > return vhost_get_avail(vq, *flags, &vq->avail->flags); > } > > static inline int vhost_get_used_event(struct vhost_virtqueue *vq, > __virtio16 *event) > { > + if (!vq->iotlb) { > + struct vring_avail *avail = vq->avail_ring.addr; > + > + if (likely(avail)) { > + *event = (__virtio16)avail->ring[vq->num]; > + return 0; > + } > + } > + > return vhost_get_avail(vq, *event, vhost_used_event(vq)); > } > > static inline int vhost_get_used_idx(struct vhost_virtqueue *vq, > __virtio16 *idx) > { > + if (!vq->iotlb) { > + struct vring_used *used = vq->used_ring.addr; > + > + if (likely(used)) { > + *idx = used->idx; > + return 0; > + } > + } > + > return vhost_get_used(vq, *idx, &vq->used->idx); > } > > static inline int vhost_get_desc(struct vhost_virtqueue *vq, > struct vring_desc *desc, int idx) > { > + if (!vq->iotlb) { > + struct vring_desc *d = vq->desc_ring.addr; > + > + if (likely(d)) { > + *desc = *(d + idx); > + return 0; > + } > + } > + > return vhost_copy_from_user(vq, desc, vq->desc + idx, sizeof(*desc)); > } > > @@ -1325,8 +1564,16 @@ int vq_meta_prefetch(struct vhost_virtqueue *vq) > { > unsigned int num = vq->num; > > - if (!vq->iotlb) > + if (!vq->iotlb) { > + if (unlikely(!vq->avail_ring.addr)) > + vhost_setup_avail_vmap(vq, (unsigned long)vq->avail); > + if (unlikely(!vq->desc_ring.addr)) > + vhost_setup_desc_vmap(vq, (unsigned long)vq->desc); > + if (unlikely(!vq->used_ring.addr)) > + vhost_setup_used_vmap(vq, (unsigned long)vq->used); > + > return 1; > + } > > return iotlb_access_ok(vq, VHOST_ACCESS_RO, (u64)(uintptr_t)vq->desc, > vhost_get_desc_size(vq, num), VHOST_ADDR_DESC) && > @@ -1478,6 +1725,13 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg > > mutex_lock(&vq->mutex); > > + /* Unregister MMU notifer to allow invalidation callback > + * can access vq->avail, vq->desc , vq->used and vq->num > + * without holding vq->mutex. > + */ > + if (d->mm) > + mmu_notifier_unregister(&d->mmu_notifier, d->mm); > + > switch (ioctl) { > case VHOST_SET_VRING_NUM: > /* Resizing ring with an active backend? > @@ -1494,6 +1748,7 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg > r = -EINVAL; > break; > } > + vhost_clean_vmaps(vq); > vq->num = s.num; > break; > case VHOST_SET_VRING_BASE: > @@ -1571,6 +1826,8 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg > } > } > > + vhost_clean_vmaps(vq); > + > vq->log_used = !!(a.flags & (0x1 << VHOST_VRING_F_LOG)); > vq->desc = (void __user *)(unsigned long)a.desc_user_addr; > vq->avail = (void __user *)(unsigned long)a.avail_user_addr; > @@ -1651,6 +1908,8 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg > if (pollstart && vq->handle_kick) > r = vhost_poll_start(&vq->poll, vq->kick); > > + if (d->mm) > + mmu_notifier_register(&d->mmu_notifier, d->mm); > mutex_unlock(&vq->mutex); > > if (pollstop && vq->handle_kick) > diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h > index 0d1ff977a43e..00f016a4f198 100644 > --- a/drivers/vhost/vhost.h > +++ b/drivers/vhost/vhost.h > @@ -12,6 +12,8 @@ > #include <linux/virtio_config.h> > #include <linux/virtio_ring.h> > #include <linux/atomic.h> > +#include <linux/pagemap.h> > +#include <linux/mmu_notifier.h> > > struct vhost_work; > typedef void (*vhost_work_fn_t)(struct vhost_work *work); > @@ -80,6 +82,11 @@ enum vhost_uaddr_type { > VHOST_NUM_ADDRS = 3, > }; > > +struct vhost_vmap { > + void *addr; > + void *unmap_addr; > +}; > +How about using actual types like struct vring_used etc so we get type safety and do not need to cast on access?> /* The virtqueue structure describes a queue attached to a device. */ > struct vhost_virtqueue { > struct vhost_dev *dev; > @@ -90,6 +97,11 @@ struct vhost_virtqueue { > struct vring_desc __user *desc; > struct vring_avail __user *avail; > struct vring_used __user *used; > + > + struct vhost_vmap avail_ring; > + struct vhost_vmap desc_ring; > + struct vhost_vmap used_ring; > + > const struct vhost_umem_node *meta_iotlb[VHOST_NUM_ADDRS]; > struct file *kick; > struct eventfd_ctx *call_ctx; > @@ -158,6 +170,7 @@ struct vhost_msg_node { > > struct vhost_dev { > struct mm_struct *mm; > + struct mmu_notifier mmu_notifier; > struct mutex mutex; > struct vhost_virtqueue **vqs; > int nvqs; > -- > 2.17.1
On Sat, Dec 29, 2018 at 08:46:51PM +0800, Jason Wang wrote:> This series tries to access virtqueue metadata through kernel virtual > address instead of copy_user() friends since they had too much > overheads like checks, spec barriers or even hardware feature > toggling.I think it's a reasonable approach. However I need to look at whether and which mmu notifiers are invoked before writeback. Do you know?> Test shows about 24% improvement on TX PPS. It should benefit other > cases as well. > > Changes from V2: > - fix buggy range overlapping check > - tear down MMU notifier during vhost ioctl to make sure invalidation > request can read metadata userspace address and vq size without > holding vq mutex. > Changes from V1: > - instead of pinning pages, use MMU notifier to invalidate vmaps and > remap duing metadata prefetch > - fix build warning on MIPS > > Jason Wang (5): > vhost: generalize adding used elem > vhost: fine grain userspace memory accessors > vhost: rename vq_iotlb_prefetch() to vq_meta_prefetch() > vhost: introduce helpers to get the size of metadata area > vhost: access vq metadata through kernel virtual address > > drivers/vhost/net.c | 4 +- > drivers/vhost/vhost.c | 416 +++++++++++++++++++++++++++++++++++++----- > drivers/vhost/vhost.h | 15 +- > 3 files changed, 384 insertions(+), 51 deletions(-) > > -- > 2.17.1
On 2019/1/3 ??4:47, Michael S. Tsirkin wrote:> On Sat, Dec 29, 2018 at 08:46:51PM +0800, Jason Wang wrote: >> This series tries to access virtqueue metadata through kernel virtual >> address instead of copy_user() friends since they had too much >> overheads like checks, spec barriers or even hardware feature >> toggling. > Will review, thanks! > One questions that comes to mind is whether it's all about bypassing > stac/clac. Could you please include a performance comparison with > nosmap? >On machine without SMAP (Sandy Bridge): Before: 4.8Mpps After: 5.2Mpps On machine with SMAP (Broadwell): Before: 5.0Mpps After: 6.1Mpps No smap: 7.5Mpps Thanks
On 2019/1/5 ??5:41, Michael S. Tsirkin wrote:> On Sat, Dec 29, 2018 at 08:46:51PM +0800, Jason Wang wrote: >> This series tries to access virtqueue metadata through kernel virtual >> address instead of copy_user() friends since they had too much >> overheads like checks, spec barriers or even hardware feature >> toggling. > > I think it's a reasonable approach. > However I need to look at whether and which mmu notifiers are invoked before > writeback. Do you know?I don't know but just looking at the MMU notifier ops definition, there's no such callback if my understanding is correct. Thanks> >> Test shows about 24% improvement on TX PPS. It should benefit other >> cases as well. >> >> Changes from V2: >> - fix buggy range overlapping check >> - tear down MMU notifier during vhost ioctl to make sure invalidation >> request can read metadata userspace address and vq size without >> holding vq mutex. >> Changes from V1: >> - instead of pinning pages, use MMU notifier to invalidate vmaps and >> remap duing metadata prefetch >> - fix build warning on MIPS >> >> Jason Wang (5): >> vhost: generalize adding used elem >> vhost: fine grain userspace memory accessors >> vhost: rename vq_iotlb_prefetch() to vq_meta_prefetch() >> vhost: introduce helpers to get the size of metadata area >> vhost: access vq metadata through kernel virtual address >> >> drivers/vhost/net.c | 4 +- >> drivers/vhost/vhost.c | 416 +++++++++++++++++++++++++++++++++++++----- >> drivers/vhost/vhost.h | 15 +- >> 3 files changed, 384 insertions(+), 51 deletions(-) >> >> -- >> 2.17.1
On 2019/1/5 ??8:33, Sean Christopherson wrote:> On Fri, Jan 04, 2019 at 04:29:34PM -0500, Michael S. Tsirkin wrote: >> On Sat, Dec 29, 2018 at 08:46:52PM +0800, Jason Wang wrote: >>> Use one generic vhost_copy_to_user() instead of two dedicated >>> accessor. This will simplify the conversion to fine grain >>> accessors. About 2% improvement of PPS were seen during vitio-user >>> txonly test. >>> >>> Signed-off-by: Jason Wang <jasowang at redhat.com> >> I don't hve a problem with this patch but do you have >> any idea how come removing what's supposed to be >> an optimization speeds things up? > With SMAP, the 2x vhost_put_user() will also mean an extra STAC/CLAC pair, > which is probably slower than the overhead of CALL+RET to whatever flavor > of copy_user_generic() gets used. CALL+RET is really the only overhead > since all variants of copy_user_generic() unroll accesses smaller than > 64 bytes, e.g. on a 64-bit system, __copy_to_user() will write all 8 > bytes in a single MOV. > > Removing the special casing also eliminates a few hundred bytes of code > as well as the need for hardware to predict count==1 vs. count>1. >Yes, I don't measure, but STAC/CALC is pretty expensive when we are do very small copies based on the result of nosmap PPS. Thanks
Seemingly Similar Threads
- [PATCH net-next 1/3] vhost: generalize adding used elem
- [RFC PATCH V3 1/5] vhost: generalize adding used elem
- [PATCH net-next 1/3] vhost: generalize adding used elem
- [RFC PATCH V2 2/5] vhost: fine grain userspace memory accessors
- [RFC PATCH V2 2/5] vhost: fine grain userspace memory accessors