1 | The following changes since commit 23895cbd82be95428e90168b12e925d0d3ca2f06: | 1 | The following changes since commit 7dd9d7e0bd29abf590d1ac235c0a00606ef81153: |
---|---|---|---|
2 | 2 | ||
3 | Merge remote-tracking branch 'remotes/awilliam/tags/vfio-update-20201123.0' into staging (2020-11-23 18:51:13 +0000) | 3 | Merge tag 'pull-ppc-20220831' of https://gitlab.com/danielhb/qemu into staging (2022-09-01 13:53:20 -0400) |
4 | 4 | ||
5 | are available in the git repository at: | 5 | are available in the git repository at: |
6 | 6 | ||
7 | https://github.com/jasowang/qemu.git tags/net-pull-request | 7 | https://github.com/jasowang/qemu.git tags/net-pull-request |
8 | 8 | ||
9 | for you to fetch changes up to 9925990d01a92564af55f6f69d0f5f59b47609b1: | 9 | for you to fetch changes up to 36a894aeb64a2e02871016da1c37d4a4ca109182: |
10 | 10 | ||
11 | net: Use correct default-path macro for downscript (2020-11-24 10:40:17 +0800) | 11 | net: tulip: Restrict DMA engine to memories (2022-09-02 10:22:39 +0800) |
12 | 12 | ||
13 | ---------------------------------------------------------------- | 13 | ---------------------------------------------------------------- |
14 | 14 | ||
15 | ---------------------------------------------------------------- | 15 | ---------------------------------------------------------------- |
16 | Keqian Zhu (1): | 16 | Eugenio Pérez (19): |
17 | net: Use correct default-path macro for downscript | 17 | vdpa: Skip the maps not in the iova tree |
18 | vdpa: do not save failed dma maps in SVQ iova tree | ||
19 | util: accept iova_tree_remove_parameter by value | ||
20 | vdpa: Remove SVQ vring from iova_tree at shutdown | ||
21 | vdpa: Make SVQ vring unmapping return void | ||
22 | vhost: Always store new kick fd on vhost_svq_set_svq_kick_fd | ||
23 | vdpa: Use ring hwaddr at vhost_vdpa_svq_unmap_ring | ||
24 | vhost: stop transfer elem ownership in vhost_handle_guest_kick | ||
25 | vhost: use SVQ element ndescs instead of opaque data for desc validation | ||
26 | vhost: Delete useless read memory barrier | ||
27 | vhost: Do not depend on !NULL VirtQueueElement on vhost_svq_flush | ||
28 | vhost_net: Add NetClientInfo start callback | ||
29 | vhost_net: Add NetClientInfo stop callback | ||
30 | vdpa: add net_vhost_vdpa_cvq_info NetClientInfo | ||
31 | vdpa: Move command buffers map to start of net device | ||
32 | vdpa: extract vhost_vdpa_net_cvq_add from vhost_vdpa_net_handle_ctrl_avail | ||
33 | vhost_net: add NetClientState->load() callback | ||
34 | vdpa: Add virtio-net mac address via CVQ at start | ||
35 | vdpa: Delete CVQ migration blocker | ||
18 | 36 | ||
19 | Paolo Bonzini (1): | 37 | Zhang Chen (1): |
20 | net: do not exit on "netdev_add help" monitor command | 38 | net/colo.c: Fix the pointer issue reported by Coverity. |
21 | 39 | ||
22 | Prasad J Pandit (1): | 40 | Zheyu Ma (1): |
23 | hw/net/e1000e: advance desc_offset in case of null descriptor | 41 | net: tulip: Restrict DMA engine to memories |
24 | 42 | ||
25 | Yuri Benditovich (1): | 43 | hw/i386/intel_iommu.c | 6 +- |
26 | net: purge queued rx packets on queue deletion | 44 | hw/net/tulip.c | 4 +- |
27 | 45 | hw/net/vhost_net.c | 17 +++ | |
28 | yuanjungong (1): | 46 | hw/virtio/vhost-iova-tree.c | 2 +- |
29 | tap: fix a memory leak | 47 | hw/virtio/vhost-iova-tree.h | 2 +- |
30 | 48 | hw/virtio/vhost-shadow-virtqueue.c | 31 +++-- | |
31 | hw/net/e1000e_core.c | 8 +++--- | 49 | hw/virtio/vhost-vdpa.c | 90 +++++++-------- |
32 | include/net/net.h | 1 + | 50 | include/hw/virtio/vhost-vdpa.h | 1 - |
33 | monitor/hmp-cmds.c | 6 ++++ | 51 | include/net/net.h | 6 + |
34 | net/net.c | 80 +++++++++++++++++++++++++++------------------------- | 52 | include/qemu/iova-tree.h | 2 +- |
35 | net/tap.c | 5 +++- | 53 | net/colo.c | 25 ++-- |
36 | 5 files changed, 57 insertions(+), 43 deletions(-) | 54 | net/colo.h | 1 + |
55 | net/trace-events | 2 +- | ||
56 | net/vhost-vdpa.c | 228 +++++++++++++++++++++++-------------- | ||
57 | util/iova-tree.c | 4 +- | ||
58 | 15 files changed, 248 insertions(+), 173 deletions(-) | ||
37 | 59 | ||
38 | 60 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | Next patch will skip the registering of dma maps that the vdpa device | ||
4 | rejects in the iova tree. We need to consider that here or we cause a | ||
5 | SIGSEGV accessing result. | ||
6 | |||
7 | Reported-by: Lei Yang <leiyang@redhat.com> | ||
8 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
9 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
10 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
11 | --- | ||
12 | hw/virtio/vhost-vdpa.c | 4 ++++ | ||
13 | 1 file changed, 4 insertions(+) | ||
14 | |||
15 | diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c | ||
16 | index XXXXXXX..XXXXXXX 100644 | ||
17 | --- a/hw/virtio/vhost-vdpa.c | ||
18 | +++ b/hw/virtio/vhost-vdpa.c | ||
19 | @@ -XXX,XX +XXX,XX @@ static void vhost_vdpa_listener_region_del(MemoryListener *listener, | ||
20 | }; | ||
21 | |||
22 | result = vhost_iova_tree_find_iova(v->iova_tree, &mem_region); | ||
23 | + if (!result) { | ||
24 | + /* The memory listener map wasn't mapped */ | ||
25 | + return; | ||
26 | + } | ||
27 | iova = result->iova; | ||
28 | vhost_iova_tree_remove(v->iova_tree, result); | ||
29 | } | ||
30 | -- | ||
31 | 2.7.4 | ||
32 | |||
33 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | If a map fails for whatever reason, it must not be saved in the tree. | ||
4 | Otherwise, qemu will try to unmap it in cleanup, leaving to more errors. | ||
5 | |||
6 | Fixes: 34e3c94eda ("vdpa: Add custom IOTLB translations to SVQ") | ||
7 | Reported-by: Lei Yang <leiyang@redhat.com> | ||
8 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
9 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
10 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
11 | --- | ||
12 | hw/virtio/vhost-vdpa.c | 20 +++++++++++++------- | ||
13 | 1 file changed, 13 insertions(+), 7 deletions(-) | ||
14 | |||
15 | diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c | ||
16 | index XXXXXXX..XXXXXXX 100644 | ||
17 | --- a/hw/virtio/vhost-vdpa.c | ||
18 | +++ b/hw/virtio/vhost-vdpa.c | ||
19 | @@ -XXX,XX +XXX,XX @@ static void vhost_vdpa_listener_commit(MemoryListener *listener) | ||
20 | static void vhost_vdpa_listener_region_add(MemoryListener *listener, | ||
21 | MemoryRegionSection *section) | ||
22 | { | ||
23 | + DMAMap mem_region = {}; | ||
24 | struct vhost_vdpa *v = container_of(listener, struct vhost_vdpa, listener); | ||
25 | hwaddr iova; | ||
26 | Int128 llend, llsize; | ||
27 | @@ -XXX,XX +XXX,XX @@ static void vhost_vdpa_listener_region_add(MemoryListener *listener, | ||
28 | |||
29 | llsize = int128_sub(llend, int128_make64(iova)); | ||
30 | if (v->shadow_vqs_enabled) { | ||
31 | - DMAMap mem_region = { | ||
32 | - .translated_addr = (hwaddr)(uintptr_t)vaddr, | ||
33 | - .size = int128_get64(llsize) - 1, | ||
34 | - .perm = IOMMU_ACCESS_FLAG(true, section->readonly), | ||
35 | - }; | ||
36 | + int r; | ||
37 | |||
38 | - int r = vhost_iova_tree_map_alloc(v->iova_tree, &mem_region); | ||
39 | + mem_region.translated_addr = (hwaddr)(uintptr_t)vaddr, | ||
40 | + mem_region.size = int128_get64(llsize) - 1, | ||
41 | + mem_region.perm = IOMMU_ACCESS_FLAG(true, section->readonly), | ||
42 | + | ||
43 | + r = vhost_iova_tree_map_alloc(v->iova_tree, &mem_region); | ||
44 | if (unlikely(r != IOVA_OK)) { | ||
45 | error_report("Can't allocate a mapping (%d)", r); | ||
46 | goto fail; | ||
47 | @@ -XXX,XX +XXX,XX @@ static void vhost_vdpa_listener_region_add(MemoryListener *listener, | ||
48 | vaddr, section->readonly); | ||
49 | if (ret) { | ||
50 | error_report("vhost vdpa map fail!"); | ||
51 | - goto fail; | ||
52 | + goto fail_map; | ||
53 | } | ||
54 | |||
55 | return; | ||
56 | |||
57 | +fail_map: | ||
58 | + if (v->shadow_vqs_enabled) { | ||
59 | + vhost_iova_tree_remove(v->iova_tree, &mem_region); | ||
60 | + } | ||
61 | + | ||
62 | fail: | ||
63 | /* | ||
64 | * On the initfn path, store the first error in the container so we | ||
65 | -- | ||
66 | 2.7.4 | ||
67 | |||
68 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | It's convenient to call iova_tree_remove from a map returned from | ||
4 | iova_tree_find or iova_tree_find_iova. With the current code this is not | ||
5 | possible, since we will free it, and then we will try to search for it | ||
6 | again. | ||
7 | |||
8 | Fix it making accepting the map by value, forcing a copy of the | ||
9 | argument. Not applying a fixes tag, since there is no use like that at | ||
10 | the moment. | ||
11 | |||
12 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
13 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
14 | --- | ||
15 | hw/i386/intel_iommu.c | 6 +++--- | ||
16 | hw/virtio/vhost-iova-tree.c | 2 +- | ||
17 | hw/virtio/vhost-iova-tree.h | 2 +- | ||
18 | hw/virtio/vhost-vdpa.c | 6 +++--- | ||
19 | include/qemu/iova-tree.h | 2 +- | ||
20 | net/vhost-vdpa.c | 4 ++-- | ||
21 | util/iova-tree.c | 4 ++-- | ||
22 | 7 files changed, 13 insertions(+), 13 deletions(-) | ||
23 | |||
24 | diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c | ||
25 | index XXXXXXX..XXXXXXX 100644 | ||
26 | --- a/hw/i386/intel_iommu.c | ||
27 | +++ b/hw/i386/intel_iommu.c | ||
28 | @@ -XXX,XX +XXX,XX @@ static int vtd_page_walk_one(IOMMUTLBEvent *event, vtd_page_walk_info *info) | ||
29 | return ret; | ||
30 | } | ||
31 | /* Drop any existing mapping */ | ||
32 | - iova_tree_remove(as->iova_tree, &target); | ||
33 | + iova_tree_remove(as->iova_tree, target); | ||
34 | /* Recover the correct type */ | ||
35 | event->type = IOMMU_NOTIFIER_MAP; | ||
36 | entry->perm = cache_perm; | ||
37 | @@ -XXX,XX +XXX,XX @@ static int vtd_page_walk_one(IOMMUTLBEvent *event, vtd_page_walk_info *info) | ||
38 | trace_vtd_page_walk_one_skip_unmap(entry->iova, entry->addr_mask); | ||
39 | return 0; | ||
40 | } | ||
41 | - iova_tree_remove(as->iova_tree, &target); | ||
42 | + iova_tree_remove(as->iova_tree, target); | ||
43 | } | ||
44 | |||
45 | trace_vtd_page_walk_one(info->domain_id, entry->iova, | ||
46 | @@ -XXX,XX +XXX,XX @@ static void vtd_address_space_unmap(VTDAddressSpace *as, IOMMUNotifier *n) | ||
47 | |||
48 | map.iova = n->start; | ||
49 | map.size = size; | ||
50 | - iova_tree_remove(as->iova_tree, &map); | ||
51 | + iova_tree_remove(as->iova_tree, map); | ||
52 | } | ||
53 | |||
54 | static void vtd_address_space_unmap_all(IntelIOMMUState *s) | ||
55 | diff --git a/hw/virtio/vhost-iova-tree.c b/hw/virtio/vhost-iova-tree.c | ||
56 | index XXXXXXX..XXXXXXX 100644 | ||
57 | --- a/hw/virtio/vhost-iova-tree.c | ||
58 | +++ b/hw/virtio/vhost-iova-tree.c | ||
59 | @@ -XXX,XX +XXX,XX @@ int vhost_iova_tree_map_alloc(VhostIOVATree *tree, DMAMap *map) | ||
60 | * @iova_tree: The vhost iova tree | ||
61 | * @map: The map to remove | ||
62 | */ | ||
63 | -void vhost_iova_tree_remove(VhostIOVATree *iova_tree, const DMAMap *map) | ||
64 | +void vhost_iova_tree_remove(VhostIOVATree *iova_tree, DMAMap map) | ||
65 | { | ||
66 | iova_tree_remove(iova_tree->iova_taddr_map, map); | ||
67 | } | ||
68 | diff --git a/hw/virtio/vhost-iova-tree.h b/hw/virtio/vhost-iova-tree.h | ||
69 | index XXXXXXX..XXXXXXX 100644 | ||
70 | --- a/hw/virtio/vhost-iova-tree.h | ||
71 | +++ b/hw/virtio/vhost-iova-tree.h | ||
72 | @@ -XXX,XX +XXX,XX @@ G_DEFINE_AUTOPTR_CLEANUP_FUNC(VhostIOVATree, vhost_iova_tree_delete); | ||
73 | const DMAMap *vhost_iova_tree_find_iova(const VhostIOVATree *iova_tree, | ||
74 | const DMAMap *map); | ||
75 | int vhost_iova_tree_map_alloc(VhostIOVATree *iova_tree, DMAMap *map); | ||
76 | -void vhost_iova_tree_remove(VhostIOVATree *iova_tree, const DMAMap *map); | ||
77 | +void vhost_iova_tree_remove(VhostIOVATree *iova_tree, DMAMap map); | ||
78 | |||
79 | #endif | ||
80 | diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c | ||
81 | index XXXXXXX..XXXXXXX 100644 | ||
82 | --- a/hw/virtio/vhost-vdpa.c | ||
83 | +++ b/hw/virtio/vhost-vdpa.c | ||
84 | @@ -XXX,XX +XXX,XX @@ static void vhost_vdpa_listener_region_add(MemoryListener *listener, | ||
85 | |||
86 | fail_map: | ||
87 | if (v->shadow_vqs_enabled) { | ||
88 | - vhost_iova_tree_remove(v->iova_tree, &mem_region); | ||
89 | + vhost_iova_tree_remove(v->iova_tree, mem_region); | ||
90 | } | ||
91 | |||
92 | fail: | ||
93 | @@ -XXX,XX +XXX,XX @@ static void vhost_vdpa_listener_region_del(MemoryListener *listener, | ||
94 | return; | ||
95 | } | ||
96 | iova = result->iova; | ||
97 | - vhost_iova_tree_remove(v->iova_tree, result); | ||
98 | + vhost_iova_tree_remove(v->iova_tree, *result); | ||
99 | } | ||
100 | vhost_vdpa_iotlb_batch_begin_once(v); | ||
101 | ret = vhost_vdpa_dma_unmap(v, iova, int128_get64(llsize)); | ||
102 | @@ -XXX,XX +XXX,XX @@ static bool vhost_vdpa_svq_map_ring(struct vhost_vdpa *v, DMAMap *needle, | ||
103 | needle->perm == IOMMU_RO); | ||
104 | if (unlikely(r != 0)) { | ||
105 | error_setg_errno(errp, -r, "Cannot map region to device"); | ||
106 | - vhost_iova_tree_remove(v->iova_tree, needle); | ||
107 | + vhost_iova_tree_remove(v->iova_tree, *needle); | ||
108 | } | ||
109 | |||
110 | return r == 0; | ||
111 | diff --git a/include/qemu/iova-tree.h b/include/qemu/iova-tree.h | ||
112 | index XXXXXXX..XXXXXXX 100644 | ||
113 | --- a/include/qemu/iova-tree.h | ||
114 | +++ b/include/qemu/iova-tree.h | ||
115 | @@ -XXX,XX +XXX,XX @@ int iova_tree_insert(IOVATree *tree, const DMAMap *map); | ||
116 | * all the mappings that are included in the provided range will be | ||
117 | * removed from the tree. Here map->translated_addr is meaningless. | ||
118 | */ | ||
119 | -void iova_tree_remove(IOVATree *tree, const DMAMap *map); | ||
120 | +void iova_tree_remove(IOVATree *tree, DMAMap map); | ||
121 | |||
122 | /** | ||
123 | * iova_tree_find: | ||
124 | diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c | ||
125 | index XXXXXXX..XXXXXXX 100644 | ||
126 | --- a/net/vhost-vdpa.c | ||
127 | +++ b/net/vhost-vdpa.c | ||
128 | @@ -XXX,XX +XXX,XX @@ static void vhost_vdpa_cvq_unmap_buf(struct vhost_vdpa *v, void *addr) | ||
129 | error_report("Device cannot unmap: %s(%d)", g_strerror(r), r); | ||
130 | } | ||
131 | |||
132 | - vhost_iova_tree_remove(tree, map); | ||
133 | + vhost_iova_tree_remove(tree, *map); | ||
134 | } | ||
135 | |||
136 | static size_t vhost_vdpa_net_cvq_cmd_len(void) | ||
137 | @@ -XXX,XX +XXX,XX @@ static bool vhost_vdpa_cvq_map_buf(struct vhost_vdpa *v, | ||
138 | return true; | ||
139 | |||
140 | dma_map_err: | ||
141 | - vhost_iova_tree_remove(v->iova_tree, &map); | ||
142 | + vhost_iova_tree_remove(v->iova_tree, map); | ||
143 | return false; | ||
144 | } | ||
145 | |||
146 | diff --git a/util/iova-tree.c b/util/iova-tree.c | ||
147 | index XXXXXXX..XXXXXXX 100644 | ||
148 | --- a/util/iova-tree.c | ||
149 | +++ b/util/iova-tree.c | ||
150 | @@ -XXX,XX +XXX,XX @@ void iova_tree_foreach(IOVATree *tree, iova_tree_iterator iterator) | ||
151 | g_tree_foreach(tree->tree, iova_tree_traverse, iterator); | ||
152 | } | ||
153 | |||
154 | -void iova_tree_remove(IOVATree *tree, const DMAMap *map) | ||
155 | +void iova_tree_remove(IOVATree *tree, DMAMap map) | ||
156 | { | ||
157 | const DMAMap *overlap; | ||
158 | |||
159 | - while ((overlap = iova_tree_find(tree, map))) { | ||
160 | + while ((overlap = iova_tree_find(tree, &map))) { | ||
161 | g_tree_remove(tree->tree, overlap); | ||
162 | } | ||
163 | } | ||
164 | -- | ||
165 | 2.7.4 | ||
166 | |||
167 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | Although the device will be reset before usage, the right thing to do is | ||
4 | to clean it. | ||
5 | |||
6 | Reported-by: Lei Yang <leiyang@redhat.com> | ||
7 | Fixes: 34e3c94eda ("vdpa: Add custom IOTLB translations to SVQ") | ||
8 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
9 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
10 | --- | ||
11 | hw/virtio/vhost-vdpa.c | 6 ++++++ | ||
12 | 1 file changed, 6 insertions(+) | ||
13 | |||
14 | diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c | ||
15 | index XXXXXXX..XXXXXXX 100644 | ||
16 | --- a/hw/virtio/vhost-vdpa.c | ||
17 | +++ b/hw/virtio/vhost-vdpa.c | ||
18 | @@ -XXX,XX +XXX,XX @@ static bool vhost_vdpa_svq_unmap_ring(struct vhost_vdpa *v, | ||
19 | |||
20 | size = ROUND_UP(result->size, qemu_real_host_page_size()); | ||
21 | r = vhost_vdpa_dma_unmap(v, result->iova, size); | ||
22 | + if (unlikely(r < 0)) { | ||
23 | + error_report("Unable to unmap SVQ vring: %s (%d)", g_strerror(-r), -r); | ||
24 | + return false; | ||
25 | + } | ||
26 | + | ||
27 | + vhost_iova_tree_remove(v->iova_tree, *result); | ||
28 | return r == 0; | ||
29 | } | ||
30 | |||
31 | -- | ||
32 | 2.7.4 | ||
33 | |||
34 | diff view generated by jsdifflib |
1 | From: Yuri Benditovich <yuri.benditovich@daynix.com> | 1 | From: Eugenio Pérez <eperezma@redhat.com> |
---|---|---|---|
2 | 2 | ||
3 | https://bugzilla.redhat.com/show_bug.cgi?id=1829272 | 3 | Nothing actually reads the return value, but an error in cleaning some |
4 | When deleting queue pair, purge pending RX packets if any. | 4 | entries could cause device stop to abort, making a restart impossible. |
5 | Example of problematic flow: | 5 | Better ignore explicitely the return value. |
6 | 1. Bring up q35 VM with tap (vhost off) and virtio-net or e1000e | ||
7 | 2. Run ping flood to the VM NIC ( 1 ms interval) | ||
8 | 3. Hot unplug the NIC device (device_del) | ||
9 | During unplug process one or more packets come, the NIC | ||
10 | can't receive, tap disables read_poll | ||
11 | 4. Hot plug the device (device_add) with the same netdev | ||
12 | The tap stays with read_poll disabled and does not receive | ||
13 | any packets anymore (tap_send never triggered) | ||
14 | 6 | ||
15 | Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com> | 7 | Reported-by: Lei Yang <leiyang@redhat.com> |
8 | Fixes: 34e3c94eda ("vdpa: Add custom IOTLB translations to SVQ") | ||
9 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
10 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
16 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 11 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
17 | --- | 12 | --- |
18 | net/net.c | 12 ++++++++---- | 13 | hw/virtio/vhost-vdpa.c | 32 ++++++++++---------------------- |
19 | 1 file changed, 8 insertions(+), 4 deletions(-) | 14 | 1 file changed, 10 insertions(+), 22 deletions(-) |
20 | 15 | ||
21 | diff --git a/net/net.c b/net/net.c | 16 | diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c |
22 | index XXXXXXX..XXXXXXX 100644 | 17 | index XXXXXXX..XXXXXXX 100644 |
23 | --- a/net/net.c | 18 | --- a/hw/virtio/vhost-vdpa.c |
24 | +++ b/net/net.c | 19 | +++ b/hw/virtio/vhost-vdpa.c |
25 | @@ -XXX,XX +XXX,XX @@ void qemu_del_nic(NICState *nic) | 20 | @@ -XXX,XX +XXX,XX @@ static int vhost_vdpa_svq_set_fds(struct vhost_dev *dev, |
26 | 21 | /** | |
27 | qemu_macaddr_set_free(&nic->conf->macaddr); | 22 | * Unmap a SVQ area in the device |
28 | 23 | */ | |
29 | - /* If this is a peer NIC and peer has already been deleted, free it now. */ | 24 | -static bool vhost_vdpa_svq_unmap_ring(struct vhost_vdpa *v, |
30 | - if (nic->peer_deleted) { | 25 | +static void vhost_vdpa_svq_unmap_ring(struct vhost_vdpa *v, |
31 | - for (i = 0; i < queues; i++) { | 26 | const DMAMap *needle) |
32 | - qemu_free_net_client(qemu_get_subqueue(nic, i)->peer); | 27 | { |
33 | + for (i = 0; i < queues; i++) { | 28 | const DMAMap *result = vhost_iova_tree_find_iova(v->iova_tree, needle); |
34 | + NetClientState *nc = qemu_get_subqueue(nic, i); | 29 | @@ -XXX,XX +XXX,XX @@ static bool vhost_vdpa_svq_unmap_ring(struct vhost_vdpa *v, |
35 | + /* If this is a peer NIC and peer has already been deleted, free it now. */ | 30 | |
36 | + if (nic->peer_deleted) { | 31 | if (unlikely(!result)) { |
37 | + qemu_free_net_client(nc->peer); | 32 | error_report("Unable to find SVQ address to unmap"); |
38 | + } else if (nc->peer) { | 33 | - return false; |
39 | + /* if there are RX packets pending, complete them */ | 34 | + return; |
40 | + qemu_purge_queued_packets(nc->peer); | 35 | } |
36 | |||
37 | size = ROUND_UP(result->size, qemu_real_host_page_size()); | ||
38 | r = vhost_vdpa_dma_unmap(v, result->iova, size); | ||
39 | if (unlikely(r < 0)) { | ||
40 | error_report("Unable to unmap SVQ vring: %s (%d)", g_strerror(-r), -r); | ||
41 | - return false; | ||
42 | + return; | ||
43 | } | ||
44 | |||
45 | vhost_iova_tree_remove(v->iova_tree, *result); | ||
46 | - return r == 0; | ||
47 | } | ||
48 | |||
49 | -static bool vhost_vdpa_svq_unmap_rings(struct vhost_dev *dev, | ||
50 | +static void vhost_vdpa_svq_unmap_rings(struct vhost_dev *dev, | ||
51 | const VhostShadowVirtqueue *svq) | ||
52 | { | ||
53 | DMAMap needle = {}; | ||
54 | struct vhost_vdpa *v = dev->opaque; | ||
55 | struct vhost_vring_addr svq_addr; | ||
56 | - bool ok; | ||
57 | |||
58 | vhost_svq_get_vring_addr(svq, &svq_addr); | ||
59 | |||
60 | needle.translated_addr = svq_addr.desc_user_addr; | ||
61 | - ok = vhost_vdpa_svq_unmap_ring(v, &needle); | ||
62 | - if (unlikely(!ok)) { | ||
63 | - return false; | ||
64 | - } | ||
65 | + vhost_vdpa_svq_unmap_ring(v, &needle); | ||
66 | |||
67 | needle.translated_addr = svq_addr.used_user_addr; | ||
68 | - return vhost_vdpa_svq_unmap_ring(v, &needle); | ||
69 | + vhost_vdpa_svq_unmap_ring(v, &needle); | ||
70 | } | ||
71 | |||
72 | /** | ||
73 | @@ -XXX,XX +XXX,XX @@ err: | ||
74 | return false; | ||
75 | } | ||
76 | |||
77 | -static bool vhost_vdpa_svqs_stop(struct vhost_dev *dev) | ||
78 | +static void vhost_vdpa_svqs_stop(struct vhost_dev *dev) | ||
79 | { | ||
80 | struct vhost_vdpa *v = dev->opaque; | ||
81 | |||
82 | if (!v->shadow_vqs) { | ||
83 | - return true; | ||
84 | + return; | ||
85 | } | ||
86 | |||
87 | for (unsigned i = 0; i < v->shadow_vqs->len; ++i) { | ||
88 | VhostShadowVirtqueue *svq = g_ptr_array_index(v->shadow_vqs, i); | ||
89 | - bool ok = vhost_vdpa_svq_unmap_rings(dev, svq); | ||
90 | - if (unlikely(!ok)) { | ||
91 | - return false; | ||
92 | - } | ||
93 | + vhost_vdpa_svq_unmap_rings(dev, svq); | ||
94 | } | ||
95 | |||
96 | if (v->migration_blocker) { | ||
97 | migrate_del_blocker(v->migration_blocker); | ||
98 | } | ||
99 | - return true; | ||
100 | } | ||
101 | |||
102 | static int vhost_vdpa_dev_start(struct vhost_dev *dev, bool started) | ||
103 | @@ -XXX,XX +XXX,XX @@ static int vhost_vdpa_dev_start(struct vhost_dev *dev, bool started) | ||
41 | } | 104 | } |
105 | vhost_vdpa_set_vring_ready(dev); | ||
106 | } else { | ||
107 | - ok = vhost_vdpa_svqs_stop(dev); | ||
108 | - if (unlikely(!ok)) { | ||
109 | - return -1; | ||
110 | - } | ||
111 | + vhost_vdpa_svqs_stop(dev); | ||
112 | vhost_vdpa_host_notifiers_uninit(dev, dev->nvqs); | ||
42 | } | 113 | } |
43 | 114 | ||
44 | -- | 115 | -- |
45 | 2.7.4 | 116 | 2.7.4 |
46 | 117 | ||
47 | 118 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | We can unbind twice a file descriptor if we call twice | ||
4 | vhost_svq_set_svq_kick_fd because of this. Since it comes from vhost and | ||
5 | not from SVQ, that file descriptor could be a different thing that | ||
6 | guest's vhost notifier. | ||
7 | |||
8 | Likewise, it can happens the same if a guest start and stop the device | ||
9 | multiple times. | ||
10 | |||
11 | Reported-by: Lei Yang <leiyang@redhat.com> | ||
12 | Fixes: dff4426fa6 ("vhost: Add Shadow VirtQueue kick forwarding capabilities") | ||
13 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
14 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
15 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
16 | --- | ||
17 | hw/virtio/vhost-shadow-virtqueue.c | 4 ++-- | ||
18 | 1 file changed, 2 insertions(+), 2 deletions(-) | ||
19 | |||
20 | diff --git a/hw/virtio/vhost-shadow-virtqueue.c b/hw/virtio/vhost-shadow-virtqueue.c | ||
21 | index XXXXXXX..XXXXXXX 100644 | ||
22 | --- a/hw/virtio/vhost-shadow-virtqueue.c | ||
23 | +++ b/hw/virtio/vhost-shadow-virtqueue.c | ||
24 | @@ -XXX,XX +XXX,XX @@ void vhost_svq_set_svq_kick_fd(VhostShadowVirtqueue *svq, int svq_kick_fd) | ||
25 | event_notifier_set_handler(svq_kick, NULL); | ||
26 | } | ||
27 | |||
28 | + event_notifier_init_fd(svq_kick, svq_kick_fd); | ||
29 | /* | ||
30 | * event_notifier_set_handler already checks for guest's notifications if | ||
31 | * they arrive at the new file descriptor in the switch, so there is no | ||
32 | * need to explicitly check for them. | ||
33 | */ | ||
34 | if (poll_start) { | ||
35 | - event_notifier_init_fd(svq_kick, svq_kick_fd); | ||
36 | event_notifier_set(svq_kick); | ||
37 | event_notifier_set_handler(svq_kick, vhost_handle_guest_kick_notifier); | ||
38 | } | ||
39 | @@ -XXX,XX +XXX,XX @@ void vhost_svq_start(VhostShadowVirtqueue *svq, VirtIODevice *vdev, | ||
40 | */ | ||
41 | void vhost_svq_stop(VhostShadowVirtqueue *svq) | ||
42 | { | ||
43 | - event_notifier_set_handler(&svq->svq_kick, NULL); | ||
44 | + vhost_svq_set_svq_kick_fd(svq, VHOST_FILE_UNBIND); | ||
45 | g_autofree VirtQueueElement *next_avail_elem = NULL; | ||
46 | |||
47 | if (!svq->vq) { | ||
48 | -- | ||
49 | 2.7.4 | ||
50 | |||
51 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | Reduce code duplication. | ||
4 | |||
5 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
6 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
7 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
8 | --- | ||
9 | hw/virtio/vhost-vdpa.c | 17 ++++++++--------- | ||
10 | 1 file changed, 8 insertions(+), 9 deletions(-) | ||
11 | |||
12 | diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c | ||
13 | index XXXXXXX..XXXXXXX 100644 | ||
14 | --- a/hw/virtio/vhost-vdpa.c | ||
15 | +++ b/hw/virtio/vhost-vdpa.c | ||
16 | @@ -XXX,XX +XXX,XX @@ static int vhost_vdpa_svq_set_fds(struct vhost_dev *dev, | ||
17 | /** | ||
18 | * Unmap a SVQ area in the device | ||
19 | */ | ||
20 | -static void vhost_vdpa_svq_unmap_ring(struct vhost_vdpa *v, | ||
21 | - const DMAMap *needle) | ||
22 | +static void vhost_vdpa_svq_unmap_ring(struct vhost_vdpa *v, hwaddr addr) | ||
23 | { | ||
24 | - const DMAMap *result = vhost_iova_tree_find_iova(v->iova_tree, needle); | ||
25 | + const DMAMap needle = { | ||
26 | + .translated_addr = addr, | ||
27 | + }; | ||
28 | + const DMAMap *result = vhost_iova_tree_find_iova(v->iova_tree, &needle); | ||
29 | hwaddr size; | ||
30 | int r; | ||
31 | |||
32 | @@ -XXX,XX +XXX,XX @@ static void vhost_vdpa_svq_unmap_ring(struct vhost_vdpa *v, | ||
33 | static void vhost_vdpa_svq_unmap_rings(struct vhost_dev *dev, | ||
34 | const VhostShadowVirtqueue *svq) | ||
35 | { | ||
36 | - DMAMap needle = {}; | ||
37 | struct vhost_vdpa *v = dev->opaque; | ||
38 | struct vhost_vring_addr svq_addr; | ||
39 | |||
40 | vhost_svq_get_vring_addr(svq, &svq_addr); | ||
41 | |||
42 | - needle.translated_addr = svq_addr.desc_user_addr; | ||
43 | - vhost_vdpa_svq_unmap_ring(v, &needle); | ||
44 | + vhost_vdpa_svq_unmap_ring(v, svq_addr.desc_user_addr); | ||
45 | |||
46 | - needle.translated_addr = svq_addr.used_user_addr; | ||
47 | - vhost_vdpa_svq_unmap_ring(v, &needle); | ||
48 | + vhost_vdpa_svq_unmap_ring(v, svq_addr.used_user_addr); | ||
49 | } | ||
50 | |||
51 | /** | ||
52 | @@ -XXX,XX +XXX,XX @@ static bool vhost_vdpa_svq_map_rings(struct vhost_dev *dev, | ||
53 | ok = vhost_vdpa_svq_map_ring(v, &device_region, errp); | ||
54 | if (unlikely(!ok)) { | ||
55 | error_prepend(errp, "Cannot create vq device region: "); | ||
56 | - vhost_vdpa_svq_unmap_ring(v, &driver_region); | ||
57 | + vhost_vdpa_svq_unmap_ring(v, driver_region.translated_addr); | ||
58 | } | ||
59 | addr->used_user_addr = device_region.iova; | ||
60 | |||
61 | -- | ||
62 | 2.7.4 | ||
63 | |||
64 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | It was easier to allow vhost_svq_add to handle the memory. Now that we | ||
4 | will allow qemu to add elements to a SVQ without the guest's knowledge, | ||
5 | it's better to handle it in the caller. | ||
6 | |||
7 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
8 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
9 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
10 | --- | ||
11 | hw/virtio/vhost-shadow-virtqueue.c | 10 ++++------ | ||
12 | 1 file changed, 4 insertions(+), 6 deletions(-) | ||
13 | |||
14 | diff --git a/hw/virtio/vhost-shadow-virtqueue.c b/hw/virtio/vhost-shadow-virtqueue.c | ||
15 | index XXXXXXX..XXXXXXX 100644 | ||
16 | --- a/hw/virtio/vhost-shadow-virtqueue.c | ||
17 | +++ b/hw/virtio/vhost-shadow-virtqueue.c | ||
18 | @@ -XXX,XX +XXX,XX @@ static void vhost_svq_kick(VhostShadowVirtqueue *svq) | ||
19 | /** | ||
20 | * Add an element to a SVQ. | ||
21 | * | ||
22 | - * The caller must check that there is enough slots for the new element. It | ||
23 | - * takes ownership of the element: In case of failure not ENOSPC, it is free. | ||
24 | - * | ||
25 | * Return -EINVAL if element is invalid, -ENOSPC if dev queue is full | ||
26 | */ | ||
27 | int vhost_svq_add(VhostShadowVirtqueue *svq, const struct iovec *out_sg, | ||
28 | @@ -XXX,XX +XXX,XX @@ int vhost_svq_add(VhostShadowVirtqueue *svq, const struct iovec *out_sg, | ||
29 | |||
30 | ok = vhost_svq_add_split(svq, out_sg, out_num, in_sg, in_num, &qemu_head); | ||
31 | if (unlikely(!ok)) { | ||
32 | - g_free(elem); | ||
33 | return -EINVAL; | ||
34 | } | ||
35 | |||
36 | @@ -XXX,XX +XXX,XX @@ static void vhost_handle_guest_kick(VhostShadowVirtqueue *svq) | ||
37 | virtio_queue_set_notification(svq->vq, false); | ||
38 | |||
39 | while (true) { | ||
40 | - VirtQueueElement *elem; | ||
41 | + g_autofree VirtQueueElement *elem; | ||
42 | int r; | ||
43 | |||
44 | if (svq->next_guest_avail_elem) { | ||
45 | @@ -XXX,XX +XXX,XX @@ static void vhost_handle_guest_kick(VhostShadowVirtqueue *svq) | ||
46 | * queue the current guest descriptor and ignore kicks | ||
47 | * until some elements are used. | ||
48 | */ | ||
49 | - svq->next_guest_avail_elem = elem; | ||
50 | + svq->next_guest_avail_elem = g_steal_pointer(&elem); | ||
51 | } | ||
52 | |||
53 | /* VQ is full or broken, just return and ignore kicks */ | ||
54 | return; | ||
55 | } | ||
56 | + /* elem belongs to SVQ or external caller now */ | ||
57 | + elem = NULL; | ||
58 | } | ||
59 | |||
60 | virtio_queue_set_notification(svq->vq, true); | ||
61 | -- | ||
62 | 2.7.4 | ||
63 | |||
64 | diff view generated by jsdifflib |
1 | From: Keqian Zhu <zhukeqian1@huawei.com> | 1 | From: Eugenio Pérez <eperezma@redhat.com> |
---|---|---|---|
2 | 2 | ||
3 | Fixes: 63c4db4c2e6d (net: relocate paths to helpers and scripts) | 3 | Since we're going to allow SVQ to add elements without the guest's |
4 | Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> | 4 | knowledge and without its own VirtQueueElement, it's easier to check if |
5 | an element is a valid head checking a different thing than the | ||
6 | VirtQueueElement. | ||
7 | |||
8 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
9 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
5 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 10 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
6 | --- | 11 | --- |
7 | net/tap.c | 3 ++- | 12 | hw/virtio/vhost-shadow-virtqueue.c | 3 ++- |
8 | 1 file changed, 2 insertions(+), 1 deletion(-) | 13 | 1 file changed, 2 insertions(+), 1 deletion(-) |
9 | 14 | ||
10 | diff --git a/net/tap.c b/net/tap.c | 15 | diff --git a/hw/virtio/vhost-shadow-virtqueue.c b/hw/virtio/vhost-shadow-virtqueue.c |
11 | index XXXXXXX..XXXXXXX 100644 | 16 | index XXXXXXX..XXXXXXX 100644 |
12 | --- a/net/tap.c | 17 | --- a/hw/virtio/vhost-shadow-virtqueue.c |
13 | +++ b/net/tap.c | 18 | +++ b/hw/virtio/vhost-shadow-virtqueue.c |
14 | @@ -XXX,XX +XXX,XX @@ free_fail: | 19 | @@ -XXX,XX +XXX,XX @@ static VirtQueueElement *vhost_svq_get_buf(VhostShadowVirtqueue *svq, |
15 | script = default_script = get_relocated_path(DEFAULT_NETWORK_SCRIPT); | 20 | return NULL; |
16 | } | 21 | } |
17 | if (!downscript) { | 22 | |
18 | - downscript = default_downscript = get_relocated_path(DEFAULT_NETWORK_SCRIPT); | 23 | - if (unlikely(!svq->desc_state[used_elem.id].elem)) { |
19 | + downscript = default_downscript = | 24 | + if (unlikely(!svq->desc_state[used_elem.id].ndescs)) { |
20 | + get_relocated_path(DEFAULT_NETWORK_DOWN_SCRIPT); | 25 | qemu_log_mask(LOG_GUEST_ERROR, |
21 | } | 26 | "Device %s says index %u is used, but it was not available", |
22 | 27 | svq->vdev->name, used_elem.id); | |
23 | if (tap->has_ifname) { | 28 | @@ -XXX,XX +XXX,XX @@ static VirtQueueElement *vhost_svq_get_buf(VhostShadowVirtqueue *svq, |
29 | } | ||
30 | |||
31 | num = svq->desc_state[used_elem.id].ndescs; | ||
32 | + svq->desc_state[used_elem.id].ndescs = 0; | ||
33 | last_used_chain = vhost_svq_last_desc_of_chain(svq, num, used_elem.id); | ||
34 | svq->desc_next[last_used_chain] = svq->free_head; | ||
35 | svq->free_head = used_elem.id; | ||
24 | -- | 36 | -- |
25 | 2.7.4 | 37 | 2.7.4 |
26 | 38 | ||
27 | 39 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | As discussed in previous series [1], this memory barrier is useless with | ||
4 | the atomic read of used idx at vhost_svq_more_used. Deleting it. | ||
5 | |||
6 | [1] https://lists.nongnu.org/archive/html/qemu-devel/2022-07/msg02616.html | ||
7 | |||
8 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
9 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
10 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
11 | --- | ||
12 | hw/virtio/vhost-shadow-virtqueue.c | 3 --- | ||
13 | 1 file changed, 3 deletions(-) | ||
14 | |||
15 | diff --git a/hw/virtio/vhost-shadow-virtqueue.c b/hw/virtio/vhost-shadow-virtqueue.c | ||
16 | index XXXXXXX..XXXXXXX 100644 | ||
17 | --- a/hw/virtio/vhost-shadow-virtqueue.c | ||
18 | +++ b/hw/virtio/vhost-shadow-virtqueue.c | ||
19 | @@ -XXX,XX +XXX,XX @@ size_t vhost_svq_poll(VhostShadowVirtqueue *svq) | ||
20 | if (unlikely(g_get_monotonic_time() - start_us > 10e6)) { | ||
21 | return 0; | ||
22 | } | ||
23 | - | ||
24 | - /* Make sure we read new used_idx */ | ||
25 | - smp_rmb(); | ||
26 | } while (true); | ||
27 | } | ||
28 | |||
29 | -- | ||
30 | 2.7.4 | ||
31 | |||
32 | diff view generated by jsdifflib |
1 | From: yuanjungong <ruc_gongyuanjun@163.com> | 1 | From: Eugenio Pérez <eperezma@redhat.com> |
---|---|---|---|
2 | 2 | ||
3 | Close fd before returning. | 3 | Since QEMU will be able to inject new elements on CVQ to restore the |
4 | state, we need not to depend on a VirtQueueElement to know if a new | ||
5 | element has been used by the device or not. Instead of check that, check | ||
6 | if there are new elements only using used idx on vhost_svq_flush. | ||
4 | 7 | ||
5 | Buglink: https://bugs.launchpad.net/qemu/+bug/1904486 | 8 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> |
6 | 9 | Acked-by: Jason Wang <jasowang@redhat.com> | |
7 | Signed-off-by: yuanjungong <ruc_gongyuanjun@163.com> | ||
8 | Reviewed-by: Peter Maydell <peter.maydell@linaro.org> | ||
9 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 10 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
10 | --- | 11 | --- |
11 | net/tap.c | 2 ++ | 12 | hw/virtio/vhost-shadow-virtqueue.c | 11 +++++++---- |
12 | 1 file changed, 2 insertions(+) | 13 | 1 file changed, 7 insertions(+), 4 deletions(-) |
13 | 14 | ||
14 | diff --git a/net/tap.c b/net/tap.c | 15 | diff --git a/hw/virtio/vhost-shadow-virtqueue.c b/hw/virtio/vhost-shadow-virtqueue.c |
15 | index XXXXXXX..XXXXXXX 100644 | 16 | index XXXXXXX..XXXXXXX 100644 |
16 | --- a/net/tap.c | 17 | --- a/hw/virtio/vhost-shadow-virtqueue.c |
17 | +++ b/net/tap.c | 18 | +++ b/hw/virtio/vhost-shadow-virtqueue.c |
18 | @@ -XXX,XX +XXX,XX @@ int net_init_tap(const Netdev *netdev, const char *name, | 19 | @@ -XXX,XX +XXX,XX @@ static void vhost_svq_flush(VhostShadowVirtqueue *svq, |
19 | if (ret < 0) { | 20 | size_t vhost_svq_poll(VhostShadowVirtqueue *svq) |
20 | error_setg_errno(errp, -ret, "%s: Can't use file descriptor %d", | 21 | { |
21 | name, fd); | 22 | int64_t start_us = g_get_monotonic_time(); |
22 | + close(fd); | 23 | + uint32_t len; |
23 | return -1; | 24 | + |
25 | do { | ||
26 | - uint32_t len; | ||
27 | - VirtQueueElement *elem = vhost_svq_get_buf(svq, &len); | ||
28 | - if (elem) { | ||
29 | - return len; | ||
30 | + if (vhost_svq_more_used(svq)) { | ||
31 | + break; | ||
24 | } | 32 | } |
25 | 33 | ||
26 | @@ -XXX,XX +XXX,XX @@ int net_init_tap(const Netdev *netdev, const char *name, | 34 | if (unlikely(g_get_monotonic_time() - start_us > 10e6)) { |
27 | vhostfdname, vnet_hdr, fd, &err); | 35 | return 0; |
28 | if (err) { | ||
29 | error_propagate(errp, err); | ||
30 | + close(fd); | ||
31 | return -1; | ||
32 | } | 36 | } |
33 | } else if (tap->has_fds) { | 37 | } while (true); |
38 | + | ||
39 | + vhost_svq_get_buf(svq, &len); | ||
40 | + return len; | ||
41 | } | ||
42 | |||
43 | /** | ||
34 | -- | 44 | -- |
35 | 2.7.4 | 45 | 2.7.4 |
36 | 46 | ||
37 | 47 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | This is used by the backend to perform actions before the device is | ||
4 | started. | ||
5 | |||
6 | In particular, vdpa net use it to map CVQ buffers to the device, so it | ||
7 | can send control commands using them. | ||
8 | |||
9 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
10 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
11 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
12 | --- | ||
13 | hw/net/vhost_net.c | 7 +++++++ | ||
14 | include/net/net.h | 2 ++ | ||
15 | 2 files changed, 9 insertions(+) | ||
16 | |||
17 | diff --git a/hw/net/vhost_net.c b/hw/net/vhost_net.c | ||
18 | index XXXXXXX..XXXXXXX 100644 | ||
19 | --- a/hw/net/vhost_net.c | ||
20 | +++ b/hw/net/vhost_net.c | ||
21 | @@ -XXX,XX +XXX,XX @@ static int vhost_net_start_one(struct vhost_net *net, | ||
22 | struct vhost_vring_file file = { }; | ||
23 | int r; | ||
24 | |||
25 | + if (net->nc->info->start) { | ||
26 | + r = net->nc->info->start(net->nc); | ||
27 | + if (r < 0) { | ||
28 | + return r; | ||
29 | + } | ||
30 | + } | ||
31 | + | ||
32 | r = vhost_dev_enable_notifiers(&net->dev, dev); | ||
33 | if (r < 0) { | ||
34 | goto fail_notifiers; | ||
35 | diff --git a/include/net/net.h b/include/net/net.h | ||
36 | index XXXXXXX..XXXXXXX 100644 | ||
37 | --- a/include/net/net.h | ||
38 | +++ b/include/net/net.h | ||
39 | @@ -XXX,XX +XXX,XX @@ typedef struct NICConf { | ||
40 | |||
41 | typedef void (NetPoll)(NetClientState *, bool enable); | ||
42 | typedef bool (NetCanReceive)(NetClientState *); | ||
43 | +typedef int (NetStart)(NetClientState *); | ||
44 | typedef ssize_t (NetReceive)(NetClientState *, const uint8_t *, size_t); | ||
45 | typedef ssize_t (NetReceiveIOV)(NetClientState *, const struct iovec *, int); | ||
46 | typedef void (NetCleanup) (NetClientState *); | ||
47 | @@ -XXX,XX +XXX,XX @@ typedef struct NetClientInfo { | ||
48 | NetReceive *receive_raw; | ||
49 | NetReceiveIOV *receive_iov; | ||
50 | NetCanReceive *can_receive; | ||
51 | + NetStart *start; | ||
52 | NetCleanup *cleanup; | ||
53 | LinkStatusChanged *link_status_changed; | ||
54 | QueryRxFilter *query_rx_filter; | ||
55 | -- | ||
56 | 2.7.4 | ||
57 | |||
58 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | Used by the backend to perform actions after the device is stopped. | ||
4 | |||
5 | In particular, vdpa net use it to unmap CVQ buffers to the device, | ||
6 | cleaning the actions performed in prepare(). | ||
7 | |||
8 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
9 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
10 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
11 | --- | ||
12 | hw/net/vhost_net.c | 3 +++ | ||
13 | include/net/net.h | 2 ++ | ||
14 | 2 files changed, 5 insertions(+) | ||
15 | |||
16 | diff --git a/hw/net/vhost_net.c b/hw/net/vhost_net.c | ||
17 | index XXXXXXX..XXXXXXX 100644 | ||
18 | --- a/hw/net/vhost_net.c | ||
19 | +++ b/hw/net/vhost_net.c | ||
20 | @@ -XXX,XX +XXX,XX @@ static void vhost_net_stop_one(struct vhost_net *net, | ||
21 | net->nc->info->poll(net->nc, true); | ||
22 | } | ||
23 | vhost_dev_stop(&net->dev, dev); | ||
24 | + if (net->nc->info->stop) { | ||
25 | + net->nc->info->stop(net->nc); | ||
26 | + } | ||
27 | vhost_dev_disable_notifiers(&net->dev, dev); | ||
28 | } | ||
29 | |||
30 | diff --git a/include/net/net.h b/include/net/net.h | ||
31 | index XXXXXXX..XXXXXXX 100644 | ||
32 | --- a/include/net/net.h | ||
33 | +++ b/include/net/net.h | ||
34 | @@ -XXX,XX +XXX,XX @@ typedef struct NICConf { | ||
35 | typedef void (NetPoll)(NetClientState *, bool enable); | ||
36 | typedef bool (NetCanReceive)(NetClientState *); | ||
37 | typedef int (NetStart)(NetClientState *); | ||
38 | +typedef void (NetStop)(NetClientState *); | ||
39 | typedef ssize_t (NetReceive)(NetClientState *, const uint8_t *, size_t); | ||
40 | typedef ssize_t (NetReceiveIOV)(NetClientState *, const struct iovec *, int); | ||
41 | typedef void (NetCleanup) (NetClientState *); | ||
42 | @@ -XXX,XX +XXX,XX @@ typedef struct NetClientInfo { | ||
43 | NetReceiveIOV *receive_iov; | ||
44 | NetCanReceive *can_receive; | ||
45 | NetStart *start; | ||
46 | + NetStop *stop; | ||
47 | NetCleanup *cleanup; | ||
48 | LinkStatusChanged *link_status_changed; | ||
49 | QueryRxFilter *query_rx_filter; | ||
50 | -- | ||
51 | 2.7.4 | ||
52 | |||
53 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | Next patches will add a new info callback to restore NIC status through | ||
4 | CVQ. Since only the CVQ vhost device is needed, create it with a new | ||
5 | NetClientInfo. | ||
6 | |||
7 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
8 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
9 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
10 | --- | ||
11 | net/vhost-vdpa.c | 12 +++++++++++- | ||
12 | 1 file changed, 11 insertions(+), 1 deletion(-) | ||
13 | |||
14 | diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c | ||
15 | index XXXXXXX..XXXXXXX 100644 | ||
16 | --- a/net/vhost-vdpa.c | ||
17 | +++ b/net/vhost-vdpa.c | ||
18 | @@ -XXX,XX +XXX,XX @@ static bool vhost_vdpa_net_cvq_map_elem(VhostVDPAState *s, | ||
19 | return true; | ||
20 | } | ||
21 | |||
22 | +static NetClientInfo net_vhost_vdpa_cvq_info = { | ||
23 | + .type = NET_CLIENT_DRIVER_VHOST_VDPA, | ||
24 | + .size = sizeof(VhostVDPAState), | ||
25 | + .receive = vhost_vdpa_receive, | ||
26 | + .cleanup = vhost_vdpa_cleanup, | ||
27 | + .has_vnet_hdr = vhost_vdpa_has_vnet_hdr, | ||
28 | + .has_ufo = vhost_vdpa_has_ufo, | ||
29 | + .check_peer_type = vhost_vdpa_check_peer_type, | ||
30 | +}; | ||
31 | + | ||
32 | /** | ||
33 | * Do not forward commands not supported by SVQ. Otherwise, the device could | ||
34 | * accept it and qemu would not know how to update the device model. | ||
35 | @@ -XXX,XX +XXX,XX @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer, | ||
36 | nc = qemu_new_net_client(&net_vhost_vdpa_info, peer, device, | ||
37 | name); | ||
38 | } else { | ||
39 | - nc = qemu_new_net_control_client(&net_vhost_vdpa_info, peer, | ||
40 | + nc = qemu_new_net_control_client(&net_vhost_vdpa_cvq_info, peer, | ||
41 | device, name); | ||
42 | } | ||
43 | snprintf(nc->info_str, sizeof(nc->info_str), TYPE_VHOST_VDPA); | ||
44 | -- | ||
45 | 2.7.4 | ||
46 | |||
47 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | 1 | From: Eugenio Pérez <eperezma@redhat.com> | |
2 | |||
3 | As this series will reuse them to restore the device state at the end of | ||
4 | a migration (or a device start), let's allocate only once at the device | ||
5 | start so we don't duplicate their map and unmap. | ||
6 | |||
7 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
8 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
9 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
10 | --- | ||
11 | net/vhost-vdpa.c | 123 ++++++++++++++++++++++++++----------------------------- | ||
12 | 1 file changed, 58 insertions(+), 65 deletions(-) | ||
13 | |||
14 | diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c | ||
15 | index XXXXXXX..XXXXXXX 100644 | ||
16 | --- a/net/vhost-vdpa.c | ||
17 | +++ b/net/vhost-vdpa.c | ||
18 | @@ -XXX,XX +XXX,XX @@ static size_t vhost_vdpa_net_cvq_cmd_page_len(void) | ||
19 | return ROUND_UP(vhost_vdpa_net_cvq_cmd_len(), qemu_real_host_page_size()); | ||
20 | } | ||
21 | |||
22 | -/** Copy and map a guest buffer. */ | ||
23 | -static bool vhost_vdpa_cvq_map_buf(struct vhost_vdpa *v, | ||
24 | - const struct iovec *out_data, | ||
25 | - size_t out_num, size_t data_len, void *buf, | ||
26 | - size_t *written, bool write) | ||
27 | +/** Map CVQ buffer. */ | ||
28 | +static int vhost_vdpa_cvq_map_buf(struct vhost_vdpa *v, void *buf, size_t size, | ||
29 | + bool write) | ||
30 | { | ||
31 | DMAMap map = {}; | ||
32 | int r; | ||
33 | |||
34 | - if (unlikely(!data_len)) { | ||
35 | - qemu_log_mask(LOG_GUEST_ERROR, "%s: invalid legnth of %s buffer\n", | ||
36 | - __func__, write ? "in" : "out"); | ||
37 | - return false; | ||
38 | - } | ||
39 | - | ||
40 | - *written = iov_to_buf(out_data, out_num, 0, buf, data_len); | ||
41 | map.translated_addr = (hwaddr)(uintptr_t)buf; | ||
42 | - map.size = vhost_vdpa_net_cvq_cmd_page_len() - 1; | ||
43 | + map.size = size - 1; | ||
44 | map.perm = write ? IOMMU_RW : IOMMU_RO, | ||
45 | r = vhost_iova_tree_map_alloc(v->iova_tree, &map); | ||
46 | if (unlikely(r != IOVA_OK)) { | ||
47 | error_report("Cannot map injected element"); | ||
48 | - return false; | ||
49 | + return r; | ||
50 | } | ||
51 | |||
52 | r = vhost_vdpa_dma_map(v, map.iova, vhost_vdpa_net_cvq_cmd_page_len(), buf, | ||
53 | @@ -XXX,XX +XXX,XX @@ static bool vhost_vdpa_cvq_map_buf(struct vhost_vdpa *v, | ||
54 | goto dma_map_err; | ||
55 | } | ||
56 | |||
57 | - return true; | ||
58 | + return 0; | ||
59 | |||
60 | dma_map_err: | ||
61 | vhost_iova_tree_remove(v->iova_tree, map); | ||
62 | - return false; | ||
63 | + return r; | ||
64 | } | ||
65 | |||
66 | -/** | ||
67 | - * Copy the guest element into a dedicated buffer suitable to be sent to NIC | ||
68 | - * | ||
69 | - * @iov: [0] is the out buffer, [1] is the in one | ||
70 | - */ | ||
71 | -static bool vhost_vdpa_net_cvq_map_elem(VhostVDPAState *s, | ||
72 | - VirtQueueElement *elem, | ||
73 | - struct iovec *iov) | ||
74 | +static int vhost_vdpa_net_cvq_start(NetClientState *nc) | ||
75 | { | ||
76 | - size_t in_copied; | ||
77 | - bool ok; | ||
78 | + VhostVDPAState *s; | ||
79 | + int r; | ||
80 | |||
81 | - iov[0].iov_base = s->cvq_cmd_out_buffer; | ||
82 | - ok = vhost_vdpa_cvq_map_buf(&s->vhost_vdpa, elem->out_sg, elem->out_num, | ||
83 | - vhost_vdpa_net_cvq_cmd_len(), iov[0].iov_base, | ||
84 | - &iov[0].iov_len, false); | ||
85 | - if (unlikely(!ok)) { | ||
86 | - return false; | ||
87 | + assert(nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA); | ||
88 | + | ||
89 | + s = DO_UPCAST(VhostVDPAState, nc, nc); | ||
90 | + if (!s->vhost_vdpa.shadow_vqs_enabled) { | ||
91 | + return 0; | ||
92 | } | ||
93 | |||
94 | - iov[1].iov_base = s->cvq_cmd_in_buffer; | ||
95 | - ok = vhost_vdpa_cvq_map_buf(&s->vhost_vdpa, NULL, 0, | ||
96 | - sizeof(virtio_net_ctrl_ack), iov[1].iov_base, | ||
97 | - &in_copied, true); | ||
98 | - if (unlikely(!ok)) { | ||
99 | + r = vhost_vdpa_cvq_map_buf(&s->vhost_vdpa, s->cvq_cmd_out_buffer, | ||
100 | + vhost_vdpa_net_cvq_cmd_page_len(), false); | ||
101 | + if (unlikely(r < 0)) { | ||
102 | + return r; | ||
103 | + } | ||
104 | + | ||
105 | + r = vhost_vdpa_cvq_map_buf(&s->vhost_vdpa, s->cvq_cmd_in_buffer, | ||
106 | + vhost_vdpa_net_cvq_cmd_page_len(), true); | ||
107 | + if (unlikely(r < 0)) { | ||
108 | vhost_vdpa_cvq_unmap_buf(&s->vhost_vdpa, s->cvq_cmd_out_buffer); | ||
109 | - return false; | ||
110 | } | ||
111 | |||
112 | - iov[1].iov_len = sizeof(virtio_net_ctrl_ack); | ||
113 | - return true; | ||
114 | + return r; | ||
115 | +} | ||
116 | + | ||
117 | +static void vhost_vdpa_net_cvq_stop(NetClientState *nc) | ||
118 | +{ | ||
119 | + VhostVDPAState *s = DO_UPCAST(VhostVDPAState, nc, nc); | ||
120 | + | ||
121 | + assert(nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA); | ||
122 | + | ||
123 | + if (s->vhost_vdpa.shadow_vqs_enabled) { | ||
124 | + vhost_vdpa_cvq_unmap_buf(&s->vhost_vdpa, s->cvq_cmd_out_buffer); | ||
125 | + vhost_vdpa_cvq_unmap_buf(&s->vhost_vdpa, s->cvq_cmd_in_buffer); | ||
126 | + } | ||
127 | } | ||
128 | |||
129 | static NetClientInfo net_vhost_vdpa_cvq_info = { | ||
130 | .type = NET_CLIENT_DRIVER_VHOST_VDPA, | ||
131 | .size = sizeof(VhostVDPAState), | ||
132 | .receive = vhost_vdpa_receive, | ||
133 | + .start = vhost_vdpa_net_cvq_start, | ||
134 | + .stop = vhost_vdpa_net_cvq_stop, | ||
135 | .cleanup = vhost_vdpa_cleanup, | ||
136 | .has_vnet_hdr = vhost_vdpa_has_vnet_hdr, | ||
137 | .has_ufo = vhost_vdpa_has_ufo, | ||
138 | @@ -XXX,XX +XXX,XX @@ static NetClientInfo net_vhost_vdpa_cvq_info = { | ||
139 | * Do not forward commands not supported by SVQ. Otherwise, the device could | ||
140 | * accept it and qemu would not know how to update the device model. | ||
141 | */ | ||
142 | -static bool vhost_vdpa_net_cvq_validate_cmd(const struct iovec *out, | ||
143 | - size_t out_num) | ||
144 | +static bool vhost_vdpa_net_cvq_validate_cmd(const void *out_buf, size_t len) | ||
145 | { | ||
146 | struct virtio_net_ctrl_hdr ctrl; | ||
147 | - size_t n; | ||
148 | |||
149 | - n = iov_to_buf(out, out_num, 0, &ctrl, sizeof(ctrl)); | ||
150 | - if (unlikely(n < sizeof(ctrl))) { | ||
151 | + if (unlikely(len < sizeof(ctrl))) { | ||
152 | qemu_log_mask(LOG_GUEST_ERROR, | ||
153 | - "%s: invalid legnth of out buffer %zu\n", __func__, n); | ||
154 | + "%s: invalid legnth of out buffer %zu\n", __func__, len); | ||
155 | return false; | ||
156 | } | ||
157 | |||
158 | + memcpy(&ctrl, out_buf, sizeof(ctrl)); | ||
159 | switch (ctrl.class) { | ||
160 | case VIRTIO_NET_CTRL_MAC: | ||
161 | switch (ctrl.cmd) { | ||
162 | @@ -XXX,XX +XXX,XX @@ static int vhost_vdpa_net_handle_ctrl_avail(VhostShadowVirtqueue *svq, | ||
163 | VhostVDPAState *s = opaque; | ||
164 | size_t in_len, dev_written; | ||
165 | virtio_net_ctrl_ack status = VIRTIO_NET_ERR; | ||
166 | - /* out and in buffers sent to the device */ | ||
167 | - struct iovec dev_buffers[2] = { | ||
168 | - { .iov_base = s->cvq_cmd_out_buffer }, | ||
169 | - { .iov_base = s->cvq_cmd_in_buffer }, | ||
170 | + /* Out buffer sent to both the vdpa device and the device model */ | ||
171 | + struct iovec out = { | ||
172 | + .iov_base = s->cvq_cmd_out_buffer, | ||
173 | + }; | ||
174 | + /* In buffer sent to the device */ | ||
175 | + const struct iovec dev_in = { | ||
176 | + .iov_base = s->cvq_cmd_in_buffer, | ||
177 | + .iov_len = sizeof(virtio_net_ctrl_ack), | ||
178 | }; | ||
179 | /* in buffer used for device model */ | ||
180 | const struct iovec in = { | ||
181 | @@ -XXX,XX +XXX,XX @@ static int vhost_vdpa_net_handle_ctrl_avail(VhostShadowVirtqueue *svq, | ||
182 | int r = -EINVAL; | ||
183 | bool ok; | ||
184 | |||
185 | - ok = vhost_vdpa_net_cvq_map_elem(s, elem, dev_buffers); | ||
186 | - if (unlikely(!ok)) { | ||
187 | - goto out; | ||
188 | - } | ||
189 | - | ||
190 | - ok = vhost_vdpa_net_cvq_validate_cmd(&dev_buffers[0], 1); | ||
191 | + out.iov_len = iov_to_buf(elem->out_sg, elem->out_num, 0, | ||
192 | + s->cvq_cmd_out_buffer, | ||
193 | + vhost_vdpa_net_cvq_cmd_len()); | ||
194 | + ok = vhost_vdpa_net_cvq_validate_cmd(s->cvq_cmd_out_buffer, out.iov_len); | ||
195 | if (unlikely(!ok)) { | ||
196 | goto out; | ||
197 | } | ||
198 | |||
199 | - r = vhost_svq_add(svq, &dev_buffers[0], 1, &dev_buffers[1], 1, elem); | ||
200 | + r = vhost_svq_add(svq, &out, 1, &dev_in, 1, elem); | ||
201 | if (unlikely(r != 0)) { | ||
202 | if (unlikely(r == -ENOSPC)) { | ||
203 | qemu_log_mask(LOG_GUEST_ERROR, "%s: No space on device queue\n", | ||
204 | @@ -XXX,XX +XXX,XX @@ static int vhost_vdpa_net_handle_ctrl_avail(VhostShadowVirtqueue *svq, | ||
205 | goto out; | ||
206 | } | ||
207 | |||
208 | - memcpy(&status, dev_buffers[1].iov_base, sizeof(status)); | ||
209 | + memcpy(&status, s->cvq_cmd_in_buffer, sizeof(status)); | ||
210 | if (status != VIRTIO_NET_OK) { | ||
211 | goto out; | ||
212 | } | ||
213 | |||
214 | status = VIRTIO_NET_ERR; | ||
215 | - virtio_net_handle_ctrl_iov(svq->vdev, &in, 1, dev_buffers, 1); | ||
216 | + virtio_net_handle_ctrl_iov(svq->vdev, &in, 1, &out, 1); | ||
217 | if (status != VIRTIO_NET_OK) { | ||
218 | error_report("Bad CVQ processing in model"); | ||
219 | } | ||
220 | @@ -XXX,XX +XXX,XX @@ out: | ||
221 | } | ||
222 | vhost_svq_push_elem(svq, elem, MIN(in_len, sizeof(status))); | ||
223 | g_free(elem); | ||
224 | - if (dev_buffers[0].iov_base) { | ||
225 | - vhost_vdpa_cvq_unmap_buf(&s->vhost_vdpa, dev_buffers[0].iov_base); | ||
226 | - } | ||
227 | - if (dev_buffers[1].iov_base) { | ||
228 | - vhost_vdpa_cvq_unmap_buf(&s->vhost_vdpa, dev_buffers[1].iov_base); | ||
229 | - } | ||
230 | return r; | ||
231 | } | ||
232 | |||
233 | -- | ||
234 | 2.7.4 | ||
235 | |||
236 | diff view generated by jsdifflib |
1 | From: Paolo Bonzini <pbonzini@redhat.com> | 1 | From: Eugenio Pérez <eperezma@redhat.com> |
---|---|---|---|
2 | 2 | ||
3 | "netdev_add help" is causing QEMU to exit because the code that | 3 | So we can reuse it to inject state messages. |
4 | invokes show_netdevs is shared between CLI and HMP processing. | ||
5 | Move the check to the callers so that exit(0) remains only | ||
6 | in the CLI flow. | ||
7 | 4 | ||
8 | "netdev_add help" is not fixed by this patch; that is left for | 5 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> |
9 | later work. | 6 | Acked-by: Jason Wang <jasowang@redhat.com> |
7 | -- | ||
8 | v7: | ||
9 | * Remove double free error | ||
10 | 10 | ||
11 | Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> | 11 | v6: |
12 | * Do not assume in buffer sent to the device is sizeof(virtio_net_ctrl_ack) | ||
13 | |||
14 | v5: | ||
15 | * Do not use an artificial !NULL VirtQueueElement | ||
16 | * Use only out size instead of iovec dev_buffers for these functions. | ||
17 | |||
12 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 18 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
13 | --- | 19 | --- |
14 | include/net/net.h | 1 + | 20 | net/vhost-vdpa.c | 59 ++++++++++++++++++++++++++++++++++++-------------------- |
15 | monitor/hmp-cmds.c | 6 +++++ | 21 | 1 file changed, 38 insertions(+), 21 deletions(-) |
16 | net/net.c | 68 +++++++++++++++++++++++++++--------------------------- | ||
17 | 3 files changed, 41 insertions(+), 34 deletions(-) | ||
18 | 22 | ||
19 | diff --git a/include/net/net.h b/include/net/net.h | 23 | diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c |
20 | index XXXXXXX..XXXXXXX 100644 | 24 | index XXXXXXX..XXXXXXX 100644 |
21 | --- a/include/net/net.h | 25 | --- a/net/vhost-vdpa.c |
22 | +++ b/include/net/net.h | 26 | +++ b/net/vhost-vdpa.c |
23 | @@ -XXX,XX +XXX,XX @@ extern const char *host_net_devices[]; | 27 | @@ -XXX,XX +XXX,XX @@ static void vhost_vdpa_net_cvq_stop(NetClientState *nc) |
24 | |||
25 | /* from net.c */ | ||
26 | int net_client_parse(QemuOptsList *opts_list, const char *str); | ||
27 | +void show_netdevs(void); | ||
28 | int net_init_clients(Error **errp); | ||
29 | void net_check_clients(void); | ||
30 | void net_cleanup(void); | ||
31 | diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c | ||
32 | index XXXXXXX..XXXXXXX 100644 | ||
33 | --- a/monitor/hmp-cmds.c | ||
34 | +++ b/monitor/hmp-cmds.c | ||
35 | @@ -XXX,XX +XXX,XX @@ | ||
36 | #include "qemu/option.h" | ||
37 | #include "qemu/timer.h" | ||
38 | #include "qemu/sockets.h" | ||
39 | +#include "qemu/help_option.h" | ||
40 | #include "monitor/monitor-internal.h" | ||
41 | #include "qapi/error.h" | ||
42 | #include "qapi/clone-visitor.h" | ||
43 | @@ -XXX,XX +XXX,XX @@ void hmp_netdev_add(Monitor *mon, const QDict *qdict) | ||
44 | { | ||
45 | Error *err = NULL; | ||
46 | QemuOpts *opts; | ||
47 | + const char *type = qdict_get_try_str(qdict, "type"); | ||
48 | |||
49 | + if (type && is_help_option(type)) { | ||
50 | + show_netdevs(); | ||
51 | + return; | ||
52 | + } | ||
53 | opts = qemu_opts_from_qdict(qemu_find_opts("netdev"), qdict, &err); | ||
54 | if (err) { | ||
55 | goto out; | ||
56 | diff --git a/net/net.c b/net/net.c | ||
57 | index XXXXXXX..XXXXXXX 100644 | ||
58 | --- a/net/net.c | ||
59 | +++ b/net/net.c | ||
60 | @@ -XXX,XX +XXX,XX @@ | ||
61 | #include "qemu/config-file.h" | ||
62 | #include "qemu/ctype.h" | ||
63 | #include "qemu/iov.h" | ||
64 | +#include "qemu/qemu-print.h" | ||
65 | #include "qemu/main-loop.h" | ||
66 | #include "qemu/option.h" | ||
67 | #include "qapi/error.h" | ||
68 | @@ -XXX,XX +XXX,XX @@ static int net_client_init1(const Netdev *netdev, bool is_netdev, Error **errp) | ||
69 | return 0; | ||
70 | } | ||
71 | |||
72 | -static void show_netdevs(void) | ||
73 | +void show_netdevs(void) | ||
74 | { | ||
75 | int idx; | ||
76 | const char *available_netdevs[] = { | ||
77 | @@ -XXX,XX +XXX,XX @@ static void show_netdevs(void) | ||
78 | #endif | ||
79 | }; | ||
80 | |||
81 | - printf("Available netdev backend types:\n"); | ||
82 | + qemu_printf("Available netdev backend types:\n"); | ||
83 | for (idx = 0; idx < ARRAY_SIZE(available_netdevs); idx++) { | ||
84 | - puts(available_netdevs[idx]); | ||
85 | + qemu_printf("%s\n", available_netdevs[idx]); | ||
86 | } | 28 | } |
87 | } | 29 | } |
88 | 30 | ||
89 | @@ -XXX,XX +XXX,XX @@ static int net_client_init(QemuOpts *opts, bool is_netdev, Error **errp) | 31 | +static ssize_t vhost_vdpa_net_cvq_add(VhostVDPAState *s, size_t out_len, |
90 | int ret = -1; | 32 | + size_t in_len) |
91 | Visitor *v = opts_visitor_new(opts); | 33 | +{ |
92 | 34 | + /* Buffers for the device */ | |
93 | - const char *type = qemu_opt_get(opts, "type"); | 35 | + const struct iovec out = { |
94 | - | 36 | + .iov_base = s->cvq_cmd_out_buffer, |
95 | - if (is_netdev && type && is_help_option(type)) { | 37 | + .iov_len = out_len, |
96 | - show_netdevs(); | 38 | + }; |
97 | - exit(0); | 39 | + const struct iovec in = { |
98 | - } else { | 40 | + .iov_base = s->cvq_cmd_in_buffer, |
99 | - /* Parse convenience option format ip6-net=fec0::0[/64] */ | 41 | + .iov_len = sizeof(virtio_net_ctrl_ack), |
100 | - const char *ip6_net = qemu_opt_get(opts, "ipv6-net"); | 42 | + }; |
101 | + /* Parse convenience option format ip6-net=fec0::0[/64] */ | 43 | + VhostShadowVirtqueue *svq = g_ptr_array_index(s->vhost_vdpa.shadow_vqs, 0); |
102 | + const char *ip6_net = qemu_opt_get(opts, "ipv6-net"); | 44 | + int r; |
103 | 45 | + | |
104 | - if (ip6_net) { | 46 | + r = vhost_svq_add(svq, &out, 1, &in, 1, NULL); |
105 | - char *prefix_addr; | 47 | + if (unlikely(r != 0)) { |
106 | - unsigned long prefix_len = 64; /* Default 64bit prefix length. */ | 48 | + if (unlikely(r == -ENOSPC)) { |
107 | + if (ip6_net) { | 49 | + qemu_log_mask(LOG_GUEST_ERROR, "%s: No space on device queue\n", |
108 | + char *prefix_addr; | 50 | + __func__); |
109 | + unsigned long prefix_len = 64; /* Default 64bit prefix length. */ | ||
110 | |||
111 | - substrings = g_strsplit(ip6_net, "/", 2); | ||
112 | - if (!substrings || !substrings[0]) { | ||
113 | - error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net", | ||
114 | - "a valid IPv6 prefix"); | ||
115 | - goto out; | ||
116 | - } | ||
117 | + substrings = g_strsplit(ip6_net, "/", 2); | ||
118 | + if (!substrings || !substrings[0]) { | ||
119 | + error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "ipv6-net", | ||
120 | + "a valid IPv6 prefix"); | ||
121 | + goto out; | ||
122 | + } | 51 | + } |
123 | 52 | + return r; | |
124 | - prefix_addr = substrings[0]; | 53 | + } |
125 | + prefix_addr = substrings[0]; | ||
126 | |||
127 | - /* Handle user-specified prefix length. */ | ||
128 | - if (substrings[1] && | ||
129 | - qemu_strtoul(substrings[1], NULL, 10, &prefix_len)) | ||
130 | - { | ||
131 | - error_setg(errp, QERR_INVALID_PARAMETER_VALUE, | ||
132 | - "ipv6-prefixlen", "a number"); | ||
133 | - goto out; | ||
134 | - } | ||
135 | - | ||
136 | - qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort); | ||
137 | - qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len, | ||
138 | - &error_abort); | ||
139 | - qemu_opt_unset(opts, "ipv6-net"); | ||
140 | + /* Handle user-specified prefix length. */ | ||
141 | + if (substrings[1] && | ||
142 | + qemu_strtoul(substrings[1], NULL, 10, &prefix_len)) | ||
143 | + { | ||
144 | + error_setg(errp, QERR_INVALID_PARAMETER_VALUE, | ||
145 | + "ipv6-prefixlen", "a number"); | ||
146 | + goto out; | ||
147 | } | ||
148 | + | 54 | + |
149 | + qemu_opt_set(opts, "ipv6-prefix", prefix_addr, &error_abort); | 55 | + /* |
150 | + qemu_opt_set_number(opts, "ipv6-prefixlen", prefix_len, | 56 | + * We can poll here since we've had BQL from the time we sent the |
151 | + &error_abort); | 57 | + * descriptor. Also, we need to take the answer before SVQ pulls by itself, |
152 | + qemu_opt_unset(opts, "ipv6-net"); | 58 | + * when BQL is released |
59 | + */ | ||
60 | + return vhost_svq_poll(svq); | ||
61 | +} | ||
62 | + | ||
63 | static NetClientInfo net_vhost_vdpa_cvq_info = { | ||
64 | .type = NET_CLIENT_DRIVER_VHOST_VDPA, | ||
65 | .size = sizeof(VhostVDPAState), | ||
66 | @@ -XXX,XX +XXX,XX @@ static int vhost_vdpa_net_handle_ctrl_avail(VhostShadowVirtqueue *svq, | ||
67 | void *opaque) | ||
68 | { | ||
69 | VhostVDPAState *s = opaque; | ||
70 | - size_t in_len, dev_written; | ||
71 | + size_t in_len; | ||
72 | virtio_net_ctrl_ack status = VIRTIO_NET_ERR; | ||
73 | /* Out buffer sent to both the vdpa device and the device model */ | ||
74 | struct iovec out = { | ||
75 | .iov_base = s->cvq_cmd_out_buffer, | ||
76 | }; | ||
77 | - /* In buffer sent to the device */ | ||
78 | - const struct iovec dev_in = { | ||
79 | - .iov_base = s->cvq_cmd_in_buffer, | ||
80 | - .iov_len = sizeof(virtio_net_ctrl_ack), | ||
81 | - }; | ||
82 | /* in buffer used for device model */ | ||
83 | const struct iovec in = { | ||
84 | .iov_base = &status, | ||
85 | .iov_len = sizeof(status), | ||
86 | }; | ||
87 | - int r = -EINVAL; | ||
88 | + ssize_t dev_written = -EINVAL; | ||
89 | bool ok; | ||
90 | |||
91 | out.iov_len = iov_to_buf(elem->out_sg, elem->out_num, 0, | ||
92 | @@ -XXX,XX +XXX,XX @@ static int vhost_vdpa_net_handle_ctrl_avail(VhostShadowVirtqueue *svq, | ||
93 | goto out; | ||
153 | } | 94 | } |
154 | 95 | ||
155 | /* Create an ID for -net if the user did not specify one */ | 96 | - r = vhost_svq_add(svq, &out, 1, &dev_in, 1, elem); |
156 | @@ -XXX,XX +XXX,XX @@ static int net_init_client(void *dummy, QemuOpts *opts, Error **errp) | 97 | - if (unlikely(r != 0)) { |
157 | 98 | - if (unlikely(r == -ENOSPC)) { | |
158 | static int net_init_netdev(void *dummy, QemuOpts *opts, Error **errp) | 99 | - qemu_log_mask(LOG_GUEST_ERROR, "%s: No space on device queue\n", |
159 | { | 100 | - __func__); |
160 | + const char *type = qemu_opt_get(opts, "type"); | 101 | - } |
161 | + | 102 | + dev_written = vhost_vdpa_net_cvq_add(s, out.iov_len, sizeof(status)); |
162 | + if (type && is_help_option(type)) { | 103 | + if (unlikely(dev_written < 0)) { |
163 | + show_netdevs(); | 104 | goto out; |
164 | + exit(0); | 105 | } |
165 | + } | 106 | |
166 | return net_client_init(opts, true, errp); | 107 | - /* |
108 | - * We can poll here since we've had BQL from the time we sent the | ||
109 | - * descriptor. Also, we need to take the answer before SVQ pulls by itself, | ||
110 | - * when BQL is released | ||
111 | - */ | ||
112 | - dev_written = vhost_svq_poll(svq); | ||
113 | if (unlikely(dev_written < sizeof(status))) { | ||
114 | error_report("Insufficient written data (%zu)", dev_written); | ||
115 | goto out; | ||
116 | @@ -XXX,XX +XXX,XX @@ static int vhost_vdpa_net_handle_ctrl_avail(VhostShadowVirtqueue *svq, | ||
117 | |||
118 | memcpy(&status, s->cvq_cmd_in_buffer, sizeof(status)); | ||
119 | if (status != VIRTIO_NET_OK) { | ||
120 | - goto out; | ||
121 | + return VIRTIO_NET_ERR; | ||
122 | } | ||
123 | |||
124 | status = VIRTIO_NET_ERR; | ||
125 | @@ -XXX,XX +XXX,XX @@ out: | ||
126 | } | ||
127 | vhost_svq_push_elem(svq, elem, MIN(in_len, sizeof(status))); | ||
128 | g_free(elem); | ||
129 | - return r; | ||
130 | + return dev_written < 0 ? dev_written : 0; | ||
167 | } | 131 | } |
168 | 132 | ||
133 | static const VhostShadowVirtqueueOps vhost_vdpa_net_svq_ops = { | ||
169 | -- | 134 | -- |
170 | 2.7.4 | 135 | 2.7.4 |
171 | 136 | ||
172 | 137 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | It allows per-net client operations right after device's successful | ||
4 | start. In particular, to load the device status. | ||
5 | |||
6 | Vhost-vdpa net will use it to add the CVQ buffers to restore the device | ||
7 | status. | ||
8 | |||
9 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
10 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
11 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
12 | --- | ||
13 | hw/net/vhost_net.c | 7 +++++++ | ||
14 | include/net/net.h | 2 ++ | ||
15 | 2 files changed, 9 insertions(+) | ||
16 | |||
17 | diff --git a/hw/net/vhost_net.c b/hw/net/vhost_net.c | ||
18 | index XXXXXXX..XXXXXXX 100644 | ||
19 | --- a/hw/net/vhost_net.c | ||
20 | +++ b/hw/net/vhost_net.c | ||
21 | @@ -XXX,XX +XXX,XX @@ static int vhost_net_start_one(struct vhost_net *net, | ||
22 | } | ||
23 | } | ||
24 | } | ||
25 | + | ||
26 | + if (net->nc->info->load) { | ||
27 | + r = net->nc->info->load(net->nc); | ||
28 | + if (r < 0) { | ||
29 | + goto fail; | ||
30 | + } | ||
31 | + } | ||
32 | return 0; | ||
33 | fail: | ||
34 | file.fd = -1; | ||
35 | diff --git a/include/net/net.h b/include/net/net.h | ||
36 | index XXXXXXX..XXXXXXX 100644 | ||
37 | --- a/include/net/net.h | ||
38 | +++ b/include/net/net.h | ||
39 | @@ -XXX,XX +XXX,XX @@ typedef struct NICConf { | ||
40 | typedef void (NetPoll)(NetClientState *, bool enable); | ||
41 | typedef bool (NetCanReceive)(NetClientState *); | ||
42 | typedef int (NetStart)(NetClientState *); | ||
43 | +typedef int (NetLoad)(NetClientState *); | ||
44 | typedef void (NetStop)(NetClientState *); | ||
45 | typedef ssize_t (NetReceive)(NetClientState *, const uint8_t *, size_t); | ||
46 | typedef ssize_t (NetReceiveIOV)(NetClientState *, const struct iovec *, int); | ||
47 | @@ -XXX,XX +XXX,XX @@ typedef struct NetClientInfo { | ||
48 | NetReceiveIOV *receive_iov; | ||
49 | NetCanReceive *can_receive; | ||
50 | NetStart *start; | ||
51 | + NetLoad *load; | ||
52 | NetStop *stop; | ||
53 | NetCleanup *cleanup; | ||
54 | LinkStatusChanged *link_status_changed; | ||
55 | -- | ||
56 | 2.7.4 | ||
57 | |||
58 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Eugenio Pérez <eperezma@redhat.com> | ||
1 | 2 | ||
3 | This is needed so the destination vdpa device see the same state a the | ||
4 | guest set in the source. | ||
5 | |||
6 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> | ||
7 | Acked-by: Jason Wang <jasowang@redhat.com> | ||
8 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
9 | --- | ||
10 | net/vhost-vdpa.c | 40 ++++++++++++++++++++++++++++++++++++++++ | ||
11 | 1 file changed, 40 insertions(+) | ||
12 | |||
13 | diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c | ||
14 | index XXXXXXX..XXXXXXX 100644 | ||
15 | --- a/net/vhost-vdpa.c | ||
16 | +++ b/net/vhost-vdpa.c | ||
17 | @@ -XXX,XX +XXX,XX @@ static ssize_t vhost_vdpa_net_cvq_add(VhostVDPAState *s, size_t out_len, | ||
18 | return vhost_svq_poll(svq); | ||
19 | } | ||
20 | |||
21 | +static int vhost_vdpa_net_load(NetClientState *nc) | ||
22 | +{ | ||
23 | + VhostVDPAState *s = DO_UPCAST(VhostVDPAState, nc, nc); | ||
24 | + const struct vhost_vdpa *v = &s->vhost_vdpa; | ||
25 | + const VirtIONet *n; | ||
26 | + uint64_t features; | ||
27 | + | ||
28 | + assert(nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA); | ||
29 | + | ||
30 | + if (!v->shadow_vqs_enabled) { | ||
31 | + return 0; | ||
32 | + } | ||
33 | + | ||
34 | + n = VIRTIO_NET(v->dev->vdev); | ||
35 | + features = n->parent_obj.guest_features; | ||
36 | + if (features & BIT_ULL(VIRTIO_NET_F_CTRL_MAC_ADDR)) { | ||
37 | + const struct virtio_net_ctrl_hdr ctrl = { | ||
38 | + .class = VIRTIO_NET_CTRL_MAC, | ||
39 | + .cmd = VIRTIO_NET_CTRL_MAC_ADDR_SET, | ||
40 | + }; | ||
41 | + char *cursor = s->cvq_cmd_out_buffer; | ||
42 | + ssize_t dev_written; | ||
43 | + | ||
44 | + memcpy(cursor, &ctrl, sizeof(ctrl)); | ||
45 | + cursor += sizeof(ctrl); | ||
46 | + memcpy(cursor, n->mac, sizeof(n->mac)); | ||
47 | + | ||
48 | + dev_written = vhost_vdpa_net_cvq_add(s, sizeof(ctrl) + sizeof(n->mac), | ||
49 | + sizeof(virtio_net_ctrl_ack)); | ||
50 | + if (unlikely(dev_written < 0)) { | ||
51 | + return dev_written; | ||
52 | + } | ||
53 | + | ||
54 | + return *((virtio_net_ctrl_ack *)s->cvq_cmd_in_buffer) != VIRTIO_NET_OK; | ||
55 | + } | ||
56 | + | ||
57 | + return 0; | ||
58 | +} | ||
59 | + | ||
60 | static NetClientInfo net_vhost_vdpa_cvq_info = { | ||
61 | .type = NET_CLIENT_DRIVER_VHOST_VDPA, | ||
62 | .size = sizeof(VhostVDPAState), | ||
63 | .receive = vhost_vdpa_receive, | ||
64 | .start = vhost_vdpa_net_cvq_start, | ||
65 | + .load = vhost_vdpa_net_load, | ||
66 | .stop = vhost_vdpa_net_cvq_stop, | ||
67 | .cleanup = vhost_vdpa_cleanup, | ||
68 | .has_vnet_hdr = vhost_vdpa_has_vnet_hdr, | ||
69 | -- | ||
70 | 2.7.4 | ||
71 | |||
72 | diff view generated by jsdifflib |
1 | From: Prasad J Pandit <pjp@fedoraproject.org> | 1 | From: Eugenio Pérez <eperezma@redhat.com> |
---|---|---|---|
2 | 2 | ||
3 | While receiving packets via e1000e_write_packet_to_guest() routine, | 3 | We can restore the device state in the destination via CVQ now. Remove |
4 | 'desc_offset' is advanced only when RX descriptor is processed. And | 4 | the migration blocker. |
5 | RX descriptor is not processed if it has NULL buffer address. | ||
6 | This may lead to an infinite loop condition. Increament 'desc_offset' | ||
7 | to process next descriptor in the ring to avoid infinite loop. | ||
8 | 5 | ||
9 | Reported-by: Cheol-woo Myung <330cjfdn@gmail.com> | 6 | Signed-off-by: Eugenio Pérez <eperezma@redhat.com> |
10 | Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org> | 7 | Acked-by: Jason Wang <jasowang@redhat.com> |
11 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 8 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
12 | --- | 9 | --- |
13 | hw/net/e1000e_core.c | 8 ++++---- | 10 | hw/virtio/vhost-vdpa.c | 15 --------------- |
14 | 1 file changed, 4 insertions(+), 4 deletions(-) | 11 | include/hw/virtio/vhost-vdpa.h | 1 - |
12 | net/vhost-vdpa.c | 2 -- | ||
13 | 3 files changed, 18 deletions(-) | ||
15 | 14 | ||
16 | diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c | 15 | diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c |
17 | index XXXXXXX..XXXXXXX 100644 | 16 | index XXXXXXX..XXXXXXX 100644 |
18 | --- a/hw/net/e1000e_core.c | 17 | --- a/hw/virtio/vhost-vdpa.c |
19 | +++ b/hw/net/e1000e_core.c | 18 | +++ b/hw/virtio/vhost-vdpa.c |
20 | @@ -XXX,XX +XXX,XX @@ e1000e_write_packet_to_guest(E1000ECore *core, struct NetRxPkt *pkt, | 19 | @@ -XXX,XX +XXX,XX @@ static bool vhost_vdpa_svqs_start(struct vhost_dev *dev) |
21 | (const char *) &fcs_pad, e1000x_fcs_len(core->mac)); | 20 | return true; |
22 | } | 21 | } |
23 | } | 22 | |
24 | - desc_offset += desc_size; | 23 | - if (v->migration_blocker) { |
25 | - if (desc_offset >= total_size) { | 24 | - int r = migrate_add_blocker(v->migration_blocker, &err); |
26 | - is_last = true; | 25 | - if (unlikely(r < 0)) { |
27 | - } | 26 | - return false; |
28 | } else { /* as per intel docs; skip descriptors with null buf addr */ | 27 | - } |
29 | trace_e1000e_rx_null_descriptor(); | 28 | - } |
30 | } | 29 | - |
31 | + desc_offset += desc_size; | 30 | for (i = 0; i < v->shadow_vqs->len; ++i) { |
32 | + if (desc_offset >= total_size) { | 31 | VirtQueue *vq = virtio_get_queue(dev->vdev, dev->vq_index + i); |
33 | + is_last = true; | 32 | VhostShadowVirtqueue *svq = g_ptr_array_index(v->shadow_vqs, i); |
34 | + } | 33 | @@ -XXX,XX +XXX,XX @@ err: |
35 | 34 | vhost_svq_stop(svq); | |
36 | e1000e_write_rx_descr(core, desc, is_last ? core->rx_pkt : NULL, | 35 | } |
37 | rss_info, do_ps ? ps_hdr_len : 0, &bastate.written); | 36 | |
37 | - if (v->migration_blocker) { | ||
38 | - migrate_del_blocker(v->migration_blocker); | ||
39 | - } | ||
40 | - | ||
41 | return false; | ||
42 | } | ||
43 | |||
44 | @@ -XXX,XX +XXX,XX @@ static void vhost_vdpa_svqs_stop(struct vhost_dev *dev) | ||
45 | VhostShadowVirtqueue *svq = g_ptr_array_index(v->shadow_vqs, i); | ||
46 | vhost_vdpa_svq_unmap_rings(dev, svq); | ||
47 | } | ||
48 | - | ||
49 | - if (v->migration_blocker) { | ||
50 | - migrate_del_blocker(v->migration_blocker); | ||
51 | - } | ||
52 | } | ||
53 | |||
54 | static int vhost_vdpa_dev_start(struct vhost_dev *dev, bool started) | ||
55 | diff --git a/include/hw/virtio/vhost-vdpa.h b/include/hw/virtio/vhost-vdpa.h | ||
56 | index XXXXXXX..XXXXXXX 100644 | ||
57 | --- a/include/hw/virtio/vhost-vdpa.h | ||
58 | +++ b/include/hw/virtio/vhost-vdpa.h | ||
59 | @@ -XXX,XX +XXX,XX @@ typedef struct vhost_vdpa { | ||
60 | bool shadow_vqs_enabled; | ||
61 | /* IOVA mapping used by the Shadow Virtqueue */ | ||
62 | VhostIOVATree *iova_tree; | ||
63 | - Error *migration_blocker; | ||
64 | GPtrArray *shadow_vqs; | ||
65 | const VhostShadowVirtqueueOps *shadow_vq_ops; | ||
66 | void *shadow_vq_ops_opaque; | ||
67 | diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c | ||
68 | index XXXXXXX..XXXXXXX 100644 | ||
69 | --- a/net/vhost-vdpa.c | ||
70 | +++ b/net/vhost-vdpa.c | ||
71 | @@ -XXX,XX +XXX,XX @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer, | ||
72 | |||
73 | s->vhost_vdpa.shadow_vq_ops = &vhost_vdpa_net_svq_ops; | ||
74 | s->vhost_vdpa.shadow_vq_ops_opaque = s; | ||
75 | - error_setg(&s->vhost_vdpa.migration_blocker, | ||
76 | - "Migration disabled: vhost-vdpa uses CVQ."); | ||
77 | } | ||
78 | ret = vhost_vdpa_add(nc, (void *)&s->vhost_vdpa, queue_pair_index, nvqs); | ||
79 | if (ret) { | ||
38 | -- | 80 | -- |
39 | 2.7.4 | 81 | 2.7.4 |
40 | 82 | ||
41 | 83 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Zhang Chen <chen.zhang@intel.com> | ||
1 | 2 | ||
3 | When enabled the virtio-net-pci, guest network packet will | ||
4 | load the vnet_hdr. In COLO status, the primary VM's network | ||
5 | packet maybe redirect to another VM, it needs filter-redirect | ||
6 | enable the vnet_hdr flag at the same time, COLO-proxy will | ||
7 | correctly parse the original network packet. If have any | ||
8 | misconfiguration here, the vnet_hdr_len is wrong for parse | ||
9 | the packet, the data+offset will point to wrong place. | ||
10 | |||
11 | Signed-off-by: Zhang Chen <chen.zhang@intel.com> | ||
12 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
13 | --- | ||
14 | net/colo.c | 25 ++++++++++++++++--------- | ||
15 | net/colo.h | 1 + | ||
16 | net/trace-events | 2 +- | ||
17 | 3 files changed, 18 insertions(+), 10 deletions(-) | ||
18 | |||
19 | diff --git a/net/colo.c b/net/colo.c | ||
20 | index XXXXXXX..XXXXXXX 100644 | ||
21 | --- a/net/colo.c | ||
22 | +++ b/net/colo.c | ||
23 | @@ -XXX,XX +XXX,XX @@ int parse_packet_early(Packet *pkt) | ||
24 | { | ||
25 | int network_length; | ||
26 | static const uint8_t vlan[] = {0x81, 0x00}; | ||
27 | - uint8_t *data = pkt->data + pkt->vnet_hdr_len; | ||
28 | + uint8_t *data = pkt->data; | ||
29 | uint16_t l3_proto; | ||
30 | ssize_t l2hdr_len; | ||
31 | |||
32 | - if (data == NULL) { | ||
33 | - trace_colo_proxy_main_vnet_info("This packet is not parsed correctly, " | ||
34 | - "pkt->vnet_hdr_len", pkt->vnet_hdr_len); | ||
35 | + assert(data); | ||
36 | + | ||
37 | + /* Check the received vnet_hdr_len then add the offset */ | ||
38 | + if ((pkt->vnet_hdr_len > sizeof(struct virtio_net_hdr_v1_hash)) || | ||
39 | + (pkt->size < sizeof(struct eth_header) + sizeof(struct vlan_header) + | ||
40 | + pkt->vnet_hdr_len)) { | ||
41 | + /* | ||
42 | + * The received remote packet maybe misconfiguration here, | ||
43 | + * Please enable/disable filter module's the vnet_hdr flag at | ||
44 | + * the same time. | ||
45 | + */ | ||
46 | + trace_colo_proxy_main_vnet_info("This received packet load wrong ", | ||
47 | + pkt->vnet_hdr_len, pkt->size); | ||
48 | return 1; | ||
49 | } | ||
50 | - l2hdr_len = eth_get_l2_hdr_length(data); | ||
51 | + data += pkt->vnet_hdr_len; | ||
52 | |||
53 | - if (pkt->size < ETH_HLEN + pkt->vnet_hdr_len) { | ||
54 | - trace_colo_proxy_main("pkt->size < ETH_HLEN"); | ||
55 | - return 1; | ||
56 | - } | ||
57 | + l2hdr_len = eth_get_l2_hdr_length(data); | ||
58 | |||
59 | /* | ||
60 | * TODO: support vlan. | ||
61 | diff --git a/net/colo.h b/net/colo.h | ||
62 | index XXXXXXX..XXXXXXX 100644 | ||
63 | --- a/net/colo.h | ||
64 | +++ b/net/colo.h | ||
65 | @@ -XXX,XX +XXX,XX @@ | ||
66 | #include "qemu/jhash.h" | ||
67 | #include "qemu/timer.h" | ||
68 | #include "net/eth.h" | ||
69 | +#include "standard-headers/linux/virtio_net.h" | ||
70 | |||
71 | #define HASHTABLE_MAX_SIZE 16384 | ||
72 | |||
73 | diff --git a/net/trace-events b/net/trace-events | ||
74 | index XXXXXXX..XXXXXXX 100644 | ||
75 | --- a/net/trace-events | ||
76 | +++ b/net/trace-events | ||
77 | @@ -XXX,XX +XXX,XX @@ vhost_user_event(const char *chr, int event) "chr: %s got event: %d" | ||
78 | |||
79 | # colo.c | ||
80 | colo_proxy_main(const char *chr) ": %s" | ||
81 | -colo_proxy_main_vnet_info(const char *sta, int size) ": %s = %d" | ||
82 | +colo_proxy_main_vnet_info(const char *sta, uint32_t vnet_hdr, int size) ": %s pkt->vnet_hdr_len = %u, pkt->size = %d" | ||
83 | |||
84 | # colo-compare.c | ||
85 | colo_compare_main(const char *chr) ": %s" | ||
86 | -- | ||
87 | 2.7.4 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Zheyu Ma <zheyuma97@gmail.com> | ||
1 | 2 | ||
3 | The DMA engine is started by I/O access and then itself accesses the | ||
4 | I/O registers, triggering a reentrancy bug. | ||
5 | |||
6 | The following log can reveal it: | ||
7 | ==5637==ERROR: AddressSanitizer: stack-overflow | ||
8 | #0 0x5595435f6078 in tulip_xmit_list_update qemu/hw/net/tulip.c:673 | ||
9 | #1 0x5595435f204a in tulip_write qemu/hw/net/tulip.c:805:13 | ||
10 | #2 0x559544637f86 in memory_region_write_accessor qemu/softmmu/memory.c:492:5 | ||
11 | #3 0x5595446379fa in access_with_adjusted_size qemu/softmmu/memory.c:554:18 | ||
12 | #4 0x5595446372fa in memory_region_dispatch_write qemu/softmmu/memory.c | ||
13 | #5 0x55954468b74c in flatview_write_continue qemu/softmmu/physmem.c:2825:23 | ||
14 | #6 0x559544683662 in flatview_write qemu/softmmu/physmem.c:2867:12 | ||
15 | #7 0x5595446833f3 in address_space_write qemu/softmmu/physmem.c:2963:18 | ||
16 | #8 0x5595435fb082 in dma_memory_rw_relaxed qemu/include/sysemu/dma.h:87:12 | ||
17 | #9 0x5595435fb082 in dma_memory_rw qemu/include/sysemu/dma.h:130:12 | ||
18 | #10 0x5595435fb082 in dma_memory_write qemu/include/sysemu/dma.h:171:12 | ||
19 | #11 0x5595435fb082 in stl_le_dma qemu/include/sysemu/dma.h:272:1 | ||
20 | #12 0x5595435fb082 in stl_le_pci_dma qemu/include/hw/pci/pci.h:910:1 | ||
21 | #13 0x5595435fb082 in tulip_desc_write qemu/hw/net/tulip.c:101:9 | ||
22 | #14 0x5595435f7e3d in tulip_xmit_list_update qemu/hw/net/tulip.c:706:9 | ||
23 | #15 0x5595435f204a in tulip_write qemu/hw/net/tulip.c:805:13 | ||
24 | |||
25 | Fix this bug by restricting the DMA engine to memories regions. | ||
26 | |||
27 | Signed-off-by: Zheyu Ma <zheyuma97@gmail.com> | ||
28 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
29 | --- | ||
30 | hw/net/tulip.c | 4 ++-- | ||
31 | 1 file changed, 2 insertions(+), 2 deletions(-) | ||
32 | |||
33 | diff --git a/hw/net/tulip.c b/hw/net/tulip.c | ||
34 | index XXXXXXX..XXXXXXX 100644 | ||
35 | --- a/hw/net/tulip.c | ||
36 | +++ b/hw/net/tulip.c | ||
37 | @@ -XXX,XX +XXX,XX @@ static const VMStateDescription vmstate_pci_tulip = { | ||
38 | static void tulip_desc_read(TULIPState *s, hwaddr p, | ||
39 | struct tulip_descriptor *desc) | ||
40 | { | ||
41 | - const MemTxAttrs attrs = MEMTXATTRS_UNSPECIFIED; | ||
42 | + const MemTxAttrs attrs = { .memory = true }; | ||
43 | |||
44 | if (s->csr[0] & CSR0_DBO) { | ||
45 | ldl_be_pci_dma(&s->dev, p, &desc->status, attrs); | ||
46 | @@ -XXX,XX +XXX,XX @@ static void tulip_desc_read(TULIPState *s, hwaddr p, | ||
47 | static void tulip_desc_write(TULIPState *s, hwaddr p, | ||
48 | struct tulip_descriptor *desc) | ||
49 | { | ||
50 | - const MemTxAttrs attrs = MEMTXATTRS_UNSPECIFIED; | ||
51 | + const MemTxAttrs attrs = { .memory = true }; | ||
52 | |||
53 | if (s->csr[0] & CSR0_DBO) { | ||
54 | stl_be_pci_dma(&s->dev, p, desc->status, attrs); | ||
55 | -- | ||
56 | 2.7.4 | diff view generated by jsdifflib |