vhost-user: add separate memslot counter for vhost-user
PATCH 1/1 vhost-user: add separate memslot counter for vhost-user
Used_memslots is equal to dev->mem->nregions now, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g. used_memslots is 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM because vhost_has_free_slot just returned false, but we can hot plug it safely in fact.
--
ChangeList:
v3:
-make used_memslots a member of struct vhost_dev instead of a global static value
v2:
-eliminating useless used_memslots_exceeded variable and used_memslots_is_exceeded() API
v1:
-vhost-user: add separate memslot counter for vhost-user
--- hw/virtio/vhost-backend.c | 12 ++++++++++ hw/virtio/vhost-user.c | 25 +++++++++++++++++++++ hw/virtio/vhost.c | 37 +++++++++++++++++++++++-------- include/hw/virtio/vhost-backend.h | 5 +++++ include/hw/virtio/vhost.h | 1 + net/vhost-user.c | 7 ++++++ 6 files changed, 78 insertions(+), 9 deletions(-)
hw/virtio/vhost-backend.c +12 -0
@@ -238,6 +238,16 @@ static void vhost_kernel_set_iotlb_callback(struct vhost_dev *dev,qemu_set_fd_handler ((uintptr_t )dev->opaque,NULL ,NULL ,NULL ); } +static void vhost_kernel_set_used_memslots (struct vhost_dev *dev) +{ + dev->used_memslots = dev->mem->nregions; +} + +static unsigned int vhost_kernel_get_used_memslots (struct vhost_dev *dev) +{ +return dev->used_memslots; +} +static const VhostOps kernel_ops = { .backend_type =VHOST_BACKEND_TYPE_KERNEL , .vhost_backend_init = vhost_kernel_init,
@@ -269,6 +279,8 @@ static const VhostOps kernel_ops = {#endif /* CONFIG_VHOST_VSOCK */ .vhost_set_iotlb_callback = vhost_kernel_set_iotlb_callback, .vhost_send_device_iotlb_msg = vhost_kernel_send_device_iotlb_msg, + .vhost_set_used_memslots = vhost_kernel_set_used_memslots, + .vhost_get_used_memslots = vhost_kernel_get_used_memslots, };#endif
hw/virtio/vhost-user.c +25 -0
@@ -2354,6 +2354,29 @@ void vhost_user_cleanup(VhostUserState *user) user->chr =NULL ; } +static void vhost_user_set_used_memslots (struct vhost_dev *dev) +{ +int i; + dev->used_memslots =0 ; + +for (i =0 ; i < dev->mem->nregions; ++i) { +struct vhost_memory_region *reg = dev->mem->regions + i; +ram_addr_t offset; +MemoryRegion *mr; +int fd; + + mr =vhost_user_get_mr_data (reg->userspace_addr, &offset, &fd); +if (mr && fd >0 ) { + dev->used_memslots++; + } + } +} + +static unsigned int vhost_user_get_used_memslots (struct vhost_dev *dev) +{ +return dev->used_memslots; +} +const VhostOps user_ops = { .backend_type =VHOST_BACKEND_TYPE_USER , .vhost_backend_init = vhost_user_backend_init,
@@ -2387,4 +2410,6 @@ const VhostOps user_ops = { .vhost_backend_mem_section_filter = vhost_user_mem_section_filter, .vhost_get_inflight_fd = vhost_user_get_inflight_fd, .vhost_set_inflight_fd = vhost_user_set_inflight_fd, + .vhost_set_used_memslots = vhost_user_set_used_memslots, + .vhost_get_used_memslots = vhost_user_get_used_memslots, };
hw/virtio/vhost.c +28 -9
@@ -45,20 +45,20 @@static struct vhost_log *vhost_log;static struct vhost_log *vhost_log_shm; -static unsigned int used_memslots; static QLIST_HEAD(, vhost_dev) vhost_devices =QLIST_HEAD_INITIALIZER (vhost_devices);bool vhost_has_free_slot (void ) { -unsigned int slots_limit = ~0U ;struct vhost_dev *hdev; QLIST_FOREACH(hdev, &vhost_devices, entry) { -unsigned int r = hdev->vhost_ops->vhost_backend_memslots_limit (hdev); - slots_limit =MIN (slots_limit, r); +if (hdev->vhost_ops->vhost_get_used_memslots (hdev) >= + hdev->vhost_ops->vhost_backend_memslots_limit (hdev)) { +return false; + } } -return slots_limit > used_memslots; +return true; } static void vhost_dev_sync_region(struct vhost_dev *dev,
@@ -502,7 +502,6 @@ static void vhost_commit(MemoryListener *listener) dev->n_mem_sections *sizeof dev->mem->regions[0 ]; dev->mem =g_realloc (dev->mem, regions_size); dev->mem->nregions = dev->n_mem_sections; - used_memslots = dev->mem->nregions;for (i =0 ; i < dev->n_mem_sections; i++) {struct vhost_memory_region *cur_vmr = dev->mem->regions + i;struct MemoryRegionSection *mrs = dev->mem_sections + i;
@@ -678,6 +677,7 @@ static void vhost_region_add_section(struct vhost_dev *dev, dev->tmp_sections[dev->n_tmp_sections -1 ].fv =NULL ;memory_region_ref (section->mr); } + dev->vhost_ops->vhost_set_used_memslots (dev); }/* Used for both add and nop callbacks */
@@ -693,6 +693,17 @@ static void vhost_region_addnop(MemoryListener *listener,vhost_region_add_section (dev, section); } +static void vhost_region_del (MemoryListener *listener, +MemoryRegionSection *section) +{ + struct vhost_dev *dev = container_of(listener, struct vhost_dev, + memory_listener); +if (!vhost_section (dev, section)) { +return ; + } + dev->vhost_ops->vhost_set_used_memslots (dev); +} +static void vhost_iommu_unmap_notify (IOMMUNotifier *n,IOMMUTLBEntry *iotlb) { struct vhost_iommu *iommu = container_of(n, struct vhost_iommu, n);
@@ -1300,6 +1311,7 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque, hdev->memory_listener = (MemoryListener ) { .begin = vhost_begin, .commit = vhost_commit, + .region_del = vhost_region_del, .region_add = vhost_region_addnop, .region_nop = vhost_region_addnop, .log_start = vhost_log_start,
@@ -1346,9 +1358,16 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,memory_listener_register (&hdev->memory_listener, &address_space_memory);QLIST_INSERT_HEAD (&vhost_devices, hdev, entry); -if (used_memslots > hdev->vhost_ops->vhost_backend_memslots_limit (hdev)) { -error_report ("vhost backend memory slots limit is less" -" than current number of present memory slots" ); +/* +* If we started a VM without any vhost device, +* for the first time vhost device hot-plug +* (vhost_get_used_memslots is always 0), +* so it needs to double check here. +*/ +if (hdev->vhost_ops->vhost_get_used_memslots (hdev) > + hdev->vhost_ops->vhost_backend_memslots_limit (hdev)) { +error_report ("vhost backend memory slots limit is less than" +" current number of present memory slots" ); r =-1 ;if (busyloop_timeout) { goto fail_busyloop;
include/hw/virtio/vhost-backend.h +5 -0
@@ -124,6 +124,9 @@ typedef int (*vhost_get_device_id_op)(struct vhost_dev *dev, uint32_t *dev_id);typedef bool (*vhost_force_iommu_op )(struct vhost_dev *dev); +typedef void (*vhost_set_used_memslots_op )(struct vhost_dev *dev); +typedef unsigned int (*vhost_get_used_memslots_op )(struct vhost_dev *dev); + typedef struct VhostOps { VhostBackendType backend_type; vhost_backend_init vhost_backend_init;
@@ -168,6 +171,8 @@ typedef struct VhostOps {vhost_vq_get_addr_op vhost_vq_get_addr;vhost_get_device_id_op vhost_get_device_id;vhost_force_iommu_op vhost_force_iommu; +vhost_set_used_memslots_op vhost_set_used_memslots; +vhost_get_used_memslots_op vhost_get_used_memslots; } VhostOps;extern const VhostOps user_ops;
include/hw/virtio/vhost.h +1 -0
@@ -90,6 +90,7 @@ struct vhost_dev { QLIST_HEAD(, vhost_iommu) iommu_list;IOMMUNotifier n;const VhostDevConfigOps *config_ops; +unsigned int used_memslots; };struct vhost_net {
net/vhost-user.c +7 -0
@@ -20,6 +20,7 @@#include "qemu/error-report.h" #include "qemu/option.h" #include "trace.h" +#include "include/hw/virtio/vhost.h" typedef struct NetVhostUserState { NetClientState nc;
@@ -347,6 +348,12 @@ static int net_vhost_user_init(NetClientState *peer, const char *device,qemu_chr_fe_set_handlers (&s->chr,NULL ,NULL , net_vhost_user_event,NULL , nc0->name,NULL , true); + +if (!vhost_has_free_slot ()) { +error_report ("used memslots exceeded the backend limit, quit " +"loop" ); + goto err; + } }while (!s->started);assert (s->vhost_net);
Signature · 1 line
2.27.0.dirty
Quoted 15 lines · Jiajun Chen
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static value
Quoted 8 lines
v2:
-eliminating uselessused_memslots_exceededvariable andused_memslots_is_exceeded()APIv1:
-vhost-user: add separate memslot counter for vhost-userSigned-off-by Jiajun Chen <[email protected]>Signed-off-by Jianjay Zhou <[email protected]>
I'm happy with this from a vhost/vhost-user perspective. vhost-backend change looks good too. I'm a little confused by what's going on with net/vhost-user.c.
Quoted 23 lines
--- hw/virtio/vhost-backend.c | 12 ++++++++++ hw/virtio/vhost-user.c | 25 +++++++++++++++++++++ hw/virtio/vhost.c | 37 +++++++++++++++++++++++-------- include/hw/virtio/vhost-backend.h | 5 +++++ include/hw/virtio/vhost.h | 1 + net/vhost-user.c | 7 ++++++ 6 files changed, 78 insertions(+), 9 deletions(-)net/vhost-user.c +1
-0@@ -20,6 +20,7 @@#include "qemu/error-report.h" #include "qemu/option.h" #include "trace.h" +#include "include/hw/virtio/vhost.h" typedef struct NetVhostUserState { NetClientState nc;
net/vhost-user.c static int net_vhost_user_init(NetClientState *peer, const char *device, · line 350
Quoted 3 lines
net/vhost-user.c +0
-0@@ -347,6 +348,12 @@ static int net_vhost_user_init(NetClientState *peer, const char *device,qemu_chr_fe_set_handlers (&s->chr,NULL ,NULL , net_vhost_user_event,NULL , nc0->name,NULL , true);
Can you elaborate on this check here? What does it have to do with fixing memslots accounting? Maybe it should be in a separate change?
Quoted 12 lines
net/vhost-user.c +6
-1+ +if (!vhost_has_free_slot ()) { +error_report ("used memslots exceeded the backend limit, quit " +"loop" ); + goto err; + } }while (!s->started);assert (s->vhost_net); --2.27.0.dirty
Quoted 49 lines · Raphael Norwitz
Quoted 15 lines · Jiajun Chen
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static valueQuoted 8 lines
v2:
-eliminating uselessused_memslots_exceededvariable andused_memslots_is_exceeded()APIv1:
-vhost-user: add separate memslot counter for vhost-userSigned-off-by Jiajun Chen <[email protected]>Signed-off-by Jianjay Zhou <[email protected]>I'm happy with this from a vhost/vhost-user perspective. vhost-backend change looks good too. I'm a little confused by what's going on with
net/vhost-user.c.Quoted 20 lines
--- hw/virtio/vhost-backend.c | 12 ++++++++++ hw/virtio/vhost-user.c | 25 +++++++++++++++++++++ hw/virtio/vhost.c | 37 +++++++++++++++++++++++-------- include/hw/virtio/vhost-backend.h | 5 +++++ include/hw/virtio/vhost.h | 1 + net/vhost-user.c | 7 ++++++ 6 files changed, 78 insertions(+), 9 deletions(-)net/vhost-user.c +1
-0@@ -20,6 +20,7 @@#include "qemu/error-report.h" #include "qemu/option.h" #include "trace.h" +#include "include/hw/virtio/vhost.h"
Quoted 10 lines
net/vhost-user.c static int net_vhost_user_init(NetClientState *peer, const char *device, · line 350
Quoted 6 lines
net/vhost-user.c +0
-0typedef struct NetVhostUserState { NetClientState nc;@@ -347,6 +348,12 @@ static int net_vhost_user_init(NetClientState *peer, const char *device,qemu_chr_fe_set_handlers (&s->chr,NULL ,NULL , net_vhost_user_event,NULL , nc0->name,NULL , true);Can you elaborate on this check here? What does it have to do with fixing memslots accounting? Maybe it should be in a separate change?
When the number of virtual machine memslots exceeds the upper limit of the back-end support, QEMU main thread may enters an endless loop and cannot process other requests.
And number of memslots will not automatically decrease, so add a check here to exit from loop in this scenario. For the newly started virtual machine, boot fails; for the hot plug network card, hot plug fails.
Quoted 14 lines
Quoted 10 lines
net/vhost-user.c +6
-1+ +if (!vhost_has_free_slot ()) { +error_report ("used memslots exceeded the backend limit, quit " +"loop" ); + goto err; + } }while (!s->started);assert (s->vhost_net); --Quoted 2 lines
2.27.0.dirty
.
Quoted 63 lines · chenjiajun
Quoted 49 lines · Raphael Norwitz
Quoted 15 lines · Jiajun Chen
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static valueQuoted 8 lines
v2:
-eliminating uselessused_memslots_exceededvariable andused_memslots_is_exceeded()APIv1:
-vhost-user: add separate memslot counter for vhost-userSigned-off-by Jiajun Chen <[email protected]>Signed-off-by Jianjay Zhou <[email protected]>I'm happy with this from a vhost/vhost-user perspective. vhost-backend change looks good too. I'm a little confused by what's going on with
net/vhost-user.c.Quoted 20 lines
--- hw/virtio/vhost-backend.c | 12 ++++++++++ hw/virtio/vhost-user.c | 25 +++++++++++++++++++++ hw/virtio/vhost.c | 37 +++++++++++++++++++++++-------- include/hw/virtio/vhost-backend.h | 5 +++++ include/hw/virtio/vhost.h | 1 + net/vhost-user.c | 7 ++++++ 6 files changed, 78 insertions(+), 9 deletions(-)net/vhost-user.c +1
-0@@ -20,6 +20,7 @@#include "qemu/error-report.h" #include "qemu/option.h" #include "trace.h" +#include "include/hw/virtio/vhost.h" Quoted 10 lines
net/vhost-user.c static int net_vhost_user_init(NetClientState *peer, const char *device, · line 350
Quoted 6 lines
net/vhost-user.c +0
-0typedef struct NetVhostUserState { NetClientState nc;@@ -347,6 +348,12 @@ static int net_vhost_user_init(NetClientState *peer, const char *device,qemu_chr_fe_set_handlers (&s->chr,NULL ,NULL , net_vhost_user_event,NULL , nc0->name,NULL , true);Can you elaborate on this check here? What does it have to do with fixing memslots accounting? Maybe it should be in a separate change?
Quoted 5 lines
When the number of virtual machine memslots exceeds the upper limit of the back-end support, QEMU main thread may enters an endless loop and cannot process other requests.
And number of memslots will not automatically decrease, so add a check here to exit from loop in this scenario. For the newly started virtual machine, boot fails; for the hot plug network card, hot plug fails.
I don't understand what you mean by "number of memslots will not automatically decrease". Where did this happen before and what changes when the new memslots counter is introduced?
Quoted 14 lines
Quoted 14 lines
Quoted 10 lines
net/vhost-user.c +6
-1+ +if (!vhost_has_free_slot ()) { +error_report ("used memslots exceeded the backend limit, quit " +"loop" ); + goto err; + } }while (!s->started);assert (s->vhost_net); --Quoted 2 lines
2.27.0.dirty
.
Quoted 8 lines · On Mon, 28 Sep 2020 21:17:31 +0800 Jiajun Chen <[email protected]> wrote:
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.
I had an impression that all guest RAM has to be shared with vhost, so combination of anon and fd based RAM couldn't work. Am I wrong?
Quoted 5 lines
--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static value
it's global resource, so why?
Quoted 241 lines
v2:
-eliminating uselessused_memslots_exceededvariable andused_memslots_is_exceeded()APIv1:
-vhost-user: add separate memslot counter for vhost-userSigned-off-by Jiajun Chen <[email protected]>Signed-off-by Jianjay Zhou <[email protected]>--- hw/virtio/vhost-backend.c | 12 ++++++++++ hw/virtio/vhost-user.c | 25 +++++++++++++++++++++ hw/virtio/vhost.c | 37 +++++++++++++++++++++++-------- include/hw/virtio/vhost-backend.h | 5 +++++ include/hw/virtio/vhost.h | 1 + net/vhost-user.c | 7 ++++++ 6 files changed, 78 insertions(+), 9 deletions(-)hw/virtio/vhost-backend.c +12
-0@@ -238,6 +238,16 @@ static void vhost_kernel_set_iotlb_callback(struct vhost_dev *dev,qemu_set_fd_handler ((uintptr_t )dev->opaque,NULL ,NULL ,NULL ); } +static void vhost_kernel_set_used_memslots (struct vhost_dev *dev) +{ + dev->used_memslots = dev->mem->nregions; +} + +static unsigned int vhost_kernel_get_used_memslots (struct vhost_dev *dev) +{ +return dev->used_memslots; +} +static const VhostOps kernel_ops = { .backend_type =VHOST_BACKEND_TYPE_KERNEL , .vhost_backend_init = vhost_kernel_init,@@ -269,6 +279,8 @@ static const VhostOps kernel_ops = {#endif /* CONFIG_VHOST_VSOCK */ .vhost_set_iotlb_callback = vhost_kernel_set_iotlb_callback, .vhost_send_device_iotlb_msg = vhost_kernel_send_device_iotlb_msg, + .vhost_set_used_memslots = vhost_kernel_set_used_memslots, + .vhost_get_used_memslots = vhost_kernel_get_used_memslots, };#endif hw/virtio/vhost-user.c +25
-0@@ -2354,6 +2354,29 @@ void vhost_user_cleanup(VhostUserState *user) user->chr =NULL ; } +static void vhost_user_set_used_memslots (struct vhost_dev *dev) +{ +int i; + dev->used_memslots =0 ; + +for (i =0 ; i < dev->mem->nregions; ++i) { +struct vhost_memory_region *reg = dev->mem->regions + i; +ram_addr_t offset; +MemoryRegion *mr; +int fd; + + mr =vhost_user_get_mr_data (reg->userspace_addr, &offset, &fd); +if (mr && fd >0 ) { + dev->used_memslots++; + } + } +} + +static unsigned int vhost_user_get_used_memslots (struct vhost_dev *dev) +{ +return dev->used_memslots; +} +const VhostOps user_ops = { .backend_type =VHOST_BACKEND_TYPE_USER , .vhost_backend_init = vhost_user_backend_init,@@ -2387,4 +2410,6 @@ const VhostOps user_ops = { .vhost_backend_mem_section_filter = vhost_user_mem_section_filter, .vhost_get_inflight_fd = vhost_user_get_inflight_fd, .vhost_set_inflight_fd = vhost_user_set_inflight_fd, + .vhost_set_used_memslots = vhost_user_set_used_memslots, + .vhost_get_used_memslots = vhost_user_get_used_memslots, };hw/virtio/vhost.c +28
-9@@ -45,20 +45,20 @@static struct vhost_log *vhost_log;static struct vhost_log *vhost_log_shm; -static unsigned int used_memslots; static QLIST_HEAD(, vhost_dev) vhost_devices =QLIST_HEAD_INITIALIZER (vhost_devices);bool vhost_has_free_slot (void ) { -unsigned int slots_limit = ~0U ;struct vhost_dev *hdev; QLIST_FOREACH(hdev, &vhost_devices, entry) { -unsigned int r = hdev->vhost_ops->vhost_backend_memslots_limit (hdev); - slots_limit =MIN (slots_limit, r); +if (hdev->vhost_ops->vhost_get_used_memslots (hdev) >= + hdev->vhost_ops->vhost_backend_memslots_limit (hdev)) { +return false; + } } -return slots_limit > used_memslots; +return true; } static void vhost_dev_sync_region(struct vhost_dev *dev,@@ -502,7 +502,6 @@ static void vhost_commit(MemoryListener *listener) dev->n_mem_sections *sizeof dev->mem->regions[0 ]; dev->mem =g_realloc (dev->mem, regions_size); dev->mem->nregions = dev->n_mem_sections; - used_memslots = dev->mem->nregions;for (i =0 ; i < dev->n_mem_sections; i++) {struct vhost_memory_region *cur_vmr = dev->mem->regions + i;struct MemoryRegionSection *mrs = dev->mem_sections + i;@@ -678,6 +677,7 @@ static void vhost_region_add_section(struct vhost_dev *dev, dev->tmp_sections[dev->n_tmp_sections -1 ].fv =NULL ;memory_region_ref (section->mr); } + dev->vhost_ops->vhost_set_used_memslots (dev); }/* Used for both add and nop callbacks */ @@ -693,6 +693,17 @@ static void vhost_region_addnop(MemoryListener *listener,vhost_region_add_section (dev, section); } +static void vhost_region_del (MemoryListener *listener, +MemoryRegionSection *section) +{ + struct vhost_dev *dev = container_of(listener, struct vhost_dev, + memory_listener); +if (!vhost_section (dev, section)) { +return ; + } + dev->vhost_ops->vhost_set_used_memslots (dev); +} +static void vhost_iommu_unmap_notify (IOMMUNotifier *n,IOMMUTLBEntry *iotlb) { struct vhost_iommu *iommu = container_of(n, struct vhost_iommu, n);@@ -1300,6 +1311,7 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque, hdev->memory_listener = (MemoryListener ) { .begin = vhost_begin, .commit = vhost_commit, + .region_del = vhost_region_del, .region_add = vhost_region_addnop, .region_nop = vhost_region_addnop, .log_start = vhost_log_start,@@ -1346,9 +1358,16 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,memory_listener_register (&hdev->memory_listener, &address_space_memory);QLIST_INSERT_HEAD (&vhost_devices, hdev, entry); -if (used_memslots > hdev->vhost_ops->vhost_backend_memslots_limit (hdev)) { -error_report ("vhost backend memory slots limit is less" -" than current number of present memory slots" ); +/* +* If we started a VM without any vhost device, +* for the first time vhost device hot-plug +* (vhost_get_used_memslots is always 0), +* so it needs to double check here. +*/ +if (hdev->vhost_ops->vhost_get_used_memslots (hdev) > + hdev->vhost_ops->vhost_backend_memslots_limit (hdev)) { +error_report ("vhost backend memory slots limit is less than" +" current number of present memory slots" ); r =-1 ;if (busyloop_timeout) { goto fail_busyloop;include/hw/virtio/vhost-backend.h +5
-0@@ -124,6 +124,9 @@ typedef int (*vhost_get_device_id_op)(struct vhost_dev *dev, uint32_t *dev_id);typedef bool (*vhost_force_iommu_op )(struct vhost_dev *dev); +typedef void (*vhost_set_used_memslots_op )(struct vhost_dev *dev); +typedef unsigned int (*vhost_get_used_memslots_op )(struct vhost_dev *dev); + typedef struct VhostOps { VhostBackendType backend_type; vhost_backend_init vhost_backend_init;@@ -168,6 +171,8 @@ typedef struct VhostOps {vhost_vq_get_addr_op vhost_vq_get_addr;vhost_get_device_id_op vhost_get_device_id;vhost_force_iommu_op vhost_force_iommu; +vhost_set_used_memslots_op vhost_set_used_memslots; +vhost_get_used_memslots_op vhost_get_used_memslots; } VhostOps;extern const VhostOps user_ops;include/hw/virtio/vhost.h +1
-0@@ -90,6 +90,7 @@ struct vhost_dev { QLIST_HEAD(, vhost_iommu) iommu_list;IOMMUNotifier n;const VhostDevConfigOps *config_ops; +unsigned int used_memslots; };struct vhost_net {net/vhost-user.c +7
-0@@ -20,6 +20,7 @@#include "qemu/error-report.h" #include "qemu/option.h" #include "trace.h" +#include "include/hw/virtio/vhost.h" typedef struct NetVhostUserState { NetClientState nc;@@ -347,6 +348,12 @@ static int net_vhost_user_init(NetClientState *peer, const char *device,qemu_chr_fe_set_handlers (&s->chr,NULL ,NULL , net_vhost_user_event,NULL , nc0->name,NULL , true); + +if (!vhost_has_free_slot ()) { +error_report ("used memslots exceeded the backend limit, quit " +"loop" ); + goto err; + } }while (!s->started);assert (s->vhost_net);
Quoted 13 lines · Igor Mammedov
Quoted 8 lines · On Mon, 28 Sep 2020 21:17:31 +0800 Jiajun Chen <[email protected]> wrote:
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.
Quoted 3 lines
I had an impression that all guest RAM has to be shared with vhost, so combination of anon and fd based RAM couldn't work. Am I wrong?
I'm not sure about the kernel backend, but I've tested adding anon memory to a VM with a vhost-user-scsi device and it works (eventually the VM crashed, but I could see the guest recognized the anon RAM). The vhost-user code is designed to work with both. I'm not sure I see a use case, but if there is one, this would be a valid issue. Maybe Jiajun or Jianjay can elaborate.
Quoted 7 lines
Quoted 6 lines
--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static valueit's global resource, so why?
I suggested it because I thought it made the code a little cleaner. I'm not opposed to changing it back, or having it stored at the vhost_user level.
Quoted 18 lines · Raphael Norwitz
Quoted 13 lines · Igor Mammedov
Quoted 8 lines · On Mon, 28 Sep 2020 21:17:31 +0800 Jiajun Chen <[email protected]> wrote:
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.Quoted 3 lines
I had an impression that all guest RAM has to be shared with vhost, so combination of anon and fd based RAM couldn't work. Am I wrong?
Quoted 6 lines
I'm not sure about the kernel backend, but I've tested adding anon memory to a VM with a vhost-user-scsi device and it works (eventually the VM crashed, but I could see the guest recognized the anon RAM). The vhost-user code is designed to work with both. I'm not sure I see a use case, but if there is one, this would be a valid issue. Maybe Jiajun or Jianjay can elaborate.
Hmm does not vhost-user skip all regions that do not have an fd?
mr = vhost_user_get_mr_data(reg->userspace_addr, &offset, &fd);
if (fd > 0) {
if (track_ramblocks) {
assert(*fd_num < VHOST_MEMORY_BASELINE_NREGIONS);
trace_vhost_user_set_mem_table_withfd(*fd_num, mr->name,
reg->memory_size,
reg->guest_phys_addr,
reg->userspace_addr,
offset);
u->region_rb_offset[i] = offset;
u->region_rb[i] = mr->ram_block;
} else if (*fd_num == VHOST_MEMORY_BASELINE_NREGIONS) {
error_report("Failed preparing vhost-user memory table msg");
return -1;
}
vhost_user_fill_msg_region(®ion_buffer, reg, offset);
msg->payload.memory.regions[*fd_num] = region_buffer;
fds[(*fd_num)++] = fd;
} else if (track_ramblocks) {
u->region_rb_offset[i] = 0;
u->region_rb[i] = NULL;
}In your test, is it possible that you were lucky and guest did not send any data from anon memory to the device?
Quoted 11 lines
Quoted 7 lines
Quoted 6 lines
--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static valueit's global resource, so why?
I suggested it because I thought it made the code a little cleaner. I'm not opposed to changing it back, or having it stored at the
vhost_userlevel.
Quoted 55 lines · Michael S. Tsirkin
Quoted 18 lines · Raphael Norwitz
Quoted 13 lines · Igor Mammedov
Quoted 8 lines · On Mon, 28 Sep 2020 21:17:31 +0800 Jiajun Chen <[email protected]> wrote:
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.Quoted 3 lines
I had an impression that all guest RAM has to be shared with vhost, so combination of anon and fd based RAM couldn't work. Am I wrong?
Quoted 6 lines
I'm not sure about the kernel backend, but I've tested adding anon memory to a VM with a vhost-user-scsi device and it works (eventually the VM crashed, but I could see the guest recognized the anon RAM). The vhost-user code is designed to work with both. I'm not sure I see a use case, but if there is one, this would be a valid issue. Maybe Jiajun or Jianjay can elaborate.
Hmm does not vhost-user skip all regions that do not have an fd?
mr = vhost_user_get_mr_data(reg->userspace_addr, &offset, &fd); if (fd > 0) { if (track_ramblocks) { assert(*fd_num < VHOST_MEMORY_BASELINE_NREGIONS); trace_vhost_user_set_mem_table_withfd(*fd_num, mr->name, reg->memory_size, reg->guest_phys_addr, reg->userspace_addr, offset); u->region_rb_offset[i] = offset; u->region_rb[i] = mr->ram_block; } else if (*fd_num == VHOST_MEMORY_BASELINE_NREGIONS) { error_report("Failed preparing vhost-user memory table msg"); return -1; } vhost_user_fill_msg_region(®ion_buffer, reg, offset); msg->payload.memory.regions[*fd_num] = region_buffer; fds[(*fd_num)++] = fd; } else if (track_ramblocks) { u->region_rb_offset[i] = 0; u->region_rb[i] = NULL; }
Quoted 2 lines
In your test, is it possible that you were lucky and guest did not send any data from anon memory to the device?
Yes - vhost-user skips the region and does not send anon memory to the device, but it does not fail the hot-add operation.
In my test the fd > 0 check definitely failed and went on to add the memory without sending it to the backend. I understand why this can be problematic (it did eventually crash the VM), but it seems like we allow it as of today. I can't think of a valid reason why you would want anon and FD ram together, but I figured there may be a reason since the vhost-user code allows for it. Should we maybe block that path altogether instead of patching it up?
Quoted 15 lines
Quoted 14 lines
Quoted 10 lines
Quoted 9 lines
--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static valueit's global resource, so why?
I suggested it because I thought it made the code a little cleaner. I'm not opposed to changing it back, or having it stored at the
vhost_userlevel.
Quoted 62 lines · Raphael Norwitz
Quoted 55 lines · Michael S. Tsirkin
Quoted 18 lines · Raphael Norwitz
Quoted 13 lines · Igor Mammedov
Quoted 8 lines · On Mon, 28 Sep 2020 21:17:31 +0800 Jiajun Chen <[email protected]> wrote:
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.Quoted 3 lines
I had an impression that all guest RAM has to be shared with vhost, so combination of anon and fd based RAM couldn't work. Am I wrong?
Quoted 6 lines
I'm not sure about the kernel backend, but I've tested adding anon memory to a VM with a vhost-user-scsi device and it works (eventually the VM crashed, but I could see the guest recognized the anon RAM). The vhost-user code is designed to work with both. I'm not sure I see a use case, but if there is one, this would be a valid issue. Maybe Jiajun or Jianjay can elaborate.
Hmm does not vhost-user skip all regions that do not have an fd?
mr = vhost_user_get_mr_data(reg->userspace_addr, &offset, &fd); if (fd > 0) { if (track_ramblocks) { assert(*fd_num < VHOST_MEMORY_BASELINE_NREGIONS); trace_vhost_user_set_mem_table_withfd(*fd_num, mr->name, reg->memory_size, reg->guest_phys_addr, reg->userspace_addr, offset); u->region_rb_offset[i] = offset; u->region_rb[i] = mr->ram_block; } else if (*fd_num == VHOST_MEMORY_BASELINE_NREGIONS) { error_report("Failed preparing vhost-user memory table msg"); return -1; } vhost_user_fill_msg_region(®ion_buffer, reg, offset); msg->payload.memory.regions[*fd_num] = region_buffer; fds[(*fd_num)++] = fd; } else if (track_ramblocks) { u->region_rb_offset[i] = 0; u->region_rb[i] = NULL; }Quoted 2 lines
In your test, is it possible that you were lucky and guest did not send any data from anon memory to the device?
Yes - vhost-user skips the region and does not send anon memory to the device, but it does not fail the hot-add operation.
Quoted 7 lines
In my test the fd > 0 check definitely failed and went on to add the memory without sending it to the backend. I understand why this can be problematic (it did eventually crash the VM), but it seems like we allow it as of today. I can't think of a valid reason why you would want anon and FD ram together, but I figured there may be a reason since the vhost-user code allows for it. Should we maybe block that path altogether instead of patching it up?
Hmm where do we patch it up? Reason we might have non FD MRs is IIUC due to things like IO regions...
Quoted 15 lines
Quoted 15 lines
Quoted 14 lines
Quoted 10 lines
Quoted 9 lines
--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static valueit's global resource, so why?
I suggested it because I thought it made the code a little cleaner. I'm not opposed to changing it back, or having it stored at the
vhost_userlevel.
Quoted 73 lines · Michael S. Tsirkin
Quoted 62 lines · Raphael Norwitz
Quoted 55 lines · Michael S. Tsirkin
Quoted 18 lines · Raphael Norwitz
Quoted 13 lines · Igor Mammedov
Quoted 8 lines · On Mon, 28 Sep 2020 21:17:31 +0800 Jiajun Chen <[email protected]> wrote:
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.Quoted 3 lines
I had an impression that all guest RAM has to be shared with vhost, so combination of anon and fd based RAM couldn't work. Am I wrong?
Quoted 6 lines
I'm not sure about the kernel backend, but I've tested adding anon memory to a VM with a vhost-user-scsi device and it works (eventually the VM crashed, but I could see the guest recognized the anon RAM). The vhost-user code is designed to work with both. I'm not sure I see a use case, but if there is one, this would be a valid issue. Maybe Jiajun or Jianjay can elaborate.
Hmm does not vhost-user skip all regions that do not have an fd?
mr = vhost_user_get_mr_data(reg->userspace_addr, &offset, &fd); if (fd > 0) { if (track_ramblocks) { assert(*fd_num < VHOST_MEMORY_BASELINE_NREGIONS); trace_vhost_user_set_mem_table_withfd(*fd_num, mr->name, reg->memory_size, reg->guest_phys_addr, reg->userspace_addr, offset); u->region_rb_offset[i] = offset; u->region_rb[i] = mr->ram_block; } else if (*fd_num == VHOST_MEMORY_BASELINE_NREGIONS) { error_report("Failed preparing vhost-user memory table msg"); return -1; } vhost_user_fill_msg_region(®ion_buffer, reg, offset); msg->payload.memory.regions[*fd_num] = region_buffer; fds[(*fd_num)++] = fd; } else if (track_ramblocks) { u->region_rb_offset[i] = 0; u->region_rb[i] = NULL; }Quoted 2 lines
In your test, is it possible that you were lucky and guest did not send any data from anon memory to the device?
Yes - vhost-user skips the region and does not send anon memory to the device, but it does not fail the hot-add operation.
Quoted 7 lines
In my test the fd > 0 check definitely failed and went on to add the memory without sending it to the backend. I understand why this can be problematic (it did eventually crash the VM), but it seems like we allow it as of today. I can't think of a valid reason why you would want anon and FD ram together, but I figured there may be a reason since the vhost-user code allows for it. Should we maybe block that path altogether instead of patching it up?
Quoted 2 lines
Hmm where do we patch it up? Reason we might have non FD MRs is IIUC due to things like IO regions...
The issue is that today such non FD MRs count towards the vhost-user max ramslots limit even though there is no good reason for them to. By "patching it up", I mean accepting this change, which makes it so that the vhost-user max ramslots limit only applies to FD RAM regions.
Quoted 18 lines
Quoted 17 lines
Quoted 17 lines
Quoted 16 lines
Quoted 7 lines
Quoted 7 lines
Quoted 5 lines
Quoted 4 lines
--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static valueit's global resource, so why?
I suggested it because I thought it made the code a little cleaner. I'm not opposed to changing it back, or having it stored at the
vhost_userlevel.
Quoted 74 lines · Raphael Norwitz
Quoted 73 lines · Michael S. Tsirkin
Quoted 69 lines · Raphael Norwitz
Quoted 55 lines · Michael S. Tsirkin
Quoted 18 lines · Raphael Norwitz
Quoted 13 lines · Igor Mammedov
Quoted 8 lines · On Mon, 28 Sep 2020 21:17:31 +0800 Jiajun Chen <[email protected]> wrote:
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.Quoted 3 lines
I had an impression that all guest RAM has to be shared with vhost, so combination of anon and fd based RAM couldn't work. Am I wrong?
Quoted 6 lines
I'm not sure about the kernel backend, but I've tested adding anon memory to a VM with a vhost-user-scsi device and it works (eventually the VM crashed, but I could see the guest recognized the anon RAM). The vhost-user code is designed to work with both. I'm not sure I see a use case, but if there is one, this would be a valid issue. Maybe Jiajun or Jianjay can elaborate.
Hmm does not vhost-user skip all regions that do not have an fd?
mr = vhost_user_get_mr_data(reg->userspace_addr, &offset, &fd); if (fd > 0) { if (track_ramblocks) { assert(*fd_num < VHOST_MEMORY_BASELINE_NREGIONS); trace_vhost_user_set_mem_table_withfd(*fd_num, mr->name, reg->memory_size, reg->guest_phys_addr, reg->userspace_addr, offset); u->region_rb_offset[i] = offset; u->region_rb[i] = mr->ram_block; } else if (*fd_num == VHOST_MEMORY_BASELINE_NREGIONS) { error_report("Failed preparing vhost-user memory table msg"); return -1; } vhost_user_fill_msg_region(®ion_buffer, reg, offset); msg->payload.memory.regions[*fd_num] = region_buffer; fds[(*fd_num)++] = fd; } else if (track_ramblocks) { u->region_rb_offset[i] = 0; u->region_rb[i] = NULL; }Quoted 2 lines
In your test, is it possible that you were lucky and guest did not send any data from anon memory to the device?
Yes - vhost-user skips the region and does not send anon memory to the device, but it does not fail the hot-add operation.
In my test the fd > 0 check definitely failed and went on to add the memory without sending it to the backend. I understand why this can be problematic (it did eventually crash the VM), but it seems like we allow it as of today. I can't think of a valid reason why you would want anon and FD ram together, but I figured there may be a reason since the vhost-user code allows for it. Should we maybe block that path altogether instead of patching it up?
Quoted 7 lines
Quoted 2 lines
Hmm where do we patch it up? Reason we might have non FD MRs is IIUC due to things like IO regions...
The issue is that today such non FD MRs count towards the vhost-user max ramslots limit even though there is no good reason for them to. By "patching it up", I mean accepting this change, which makes it so that the vhost-user max ramslots limit only applies to FD RAM regions.
I don't really remember, maybe one can get these things with things like ROMs ...
Quoted 18 lines
Quoted 18 lines
Quoted 17 lines
Quoted 17 lines
Quoted 16 lines
Quoted 7 lines
Quoted 7 lines
Quoted 5 lines
Quoted 4 lines
--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static valueit's global resource, so why?
I suggested it because I thought it made the code a little cleaner. I'm not opposed to changing it back, or having it stored at the
vhost_userlevel.
Quoted 13 lines · On Wed, 14 Oct 2020 12:11:34 -0400 Raphael Norwitz…
Quoted 11 lines · Michael S. Tsirkin
Quoted 9 lines · Raphael Norwitz
Quoted 8 lines · Igor Mammedov
Quoted 4 lines · On Mon, 28 Sep 2020 21:17:31 +0800 Jiajun Chen <[email protected]> wrote:
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
It is usefully in some scenarios, e.g.
used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.
can you find out what are these extra 3 memory regions are and why they are filtered out from regions that are passed to vhost-user?
Quoted 46 lines
Quoted 40 lines
Quoted 11 lines
Quoted 4 lines
I had an impression that all guest RAM has to be shared with vhost, so combination of anon and fd based RAM couldn't work. Am I wrong?
I'm not sure about the kernel backend, but I've tested adding anon memory to a VM with a vhost-user-scsi device and it works (eventually the VM crashed, but I could see the guest recognized the anon RAM). The vhost-user code is designed to work with both. I'm not sure I see a use case, but if there is one, this would be a valid issue. Maybe Jiajun or Jianjay can elaborate.
Hmm does not vhost-user skip all regions that do not have an fd?
mr = vhost_user_get_mr_data(reg->userspace_addr, &offset, &fd); if (fd > 0) { if (track_ramblocks) { assert(*fd_num < VHOST_MEMORY_BASELINE_NREGIONS); trace_vhost_user_set_mem_table_withfd(*fd_num, mr->name, reg->memory_size, reg->guest_phys_addr, reg->userspace_addr, offset); u->region_rb_offset[i] = offset; u->region_rb[i] = mr->ram_block; } else if (*fd_num == VHOST_MEMORY_BASELINE_NREGIONS) { error_report("Failed preparing vhost-user memory table msg"); return -1; } vhost_user_fill_msg_region(®ion_buffer, reg, offset); msg->payload.memory.regions[*fd_num] = region_buffer; fds[(*fd_num)++] = fd; } else if (track_ramblocks) { u->region_rb_offset[i] = 0; u->region_rb[i] = NULL; }Quoted 2 lines
In your test, is it possible that you were lucky and guest did not send any data from anon memory to the device?
Yes - vhost-user skips the region and does not send anon memory to the device, but it does not fail the hot-add operation.
Quoted 7 lines
In my test the fd > 0 check definitely failed and went on to add the memory without sending it to the backend. I understand why this can be problematic (it did eventually crash the VM), but it seems like we allow it as of today. I can't think of a valid reason why you would want anon and FD ram together, but I figured there may be a reason since the vhost-user code allows for it. Should we maybe block that path altogether instead of patching it up?
I'm more inclined to disabling mixed (provided that's really the case) anon and FD RAM whenever vhost (user) is used or disable hot plugging vhost-user device in case machine has mixed RAM.
Otherwise it's just a time bomb, waiting till guest OS tries to transmit data that it just allocated from anon RAM.
Quoted 13 lines
Quoted 12 lines
Quoted 11 lines
Quoted 7 lines
Quoted 6 lines
--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static valueit's global resource, so why?
I suggested it because I thought it made the code a little cleaner. I'm not opposed to changing it back, or having it stored at the
vhost_userlevel.
Quoted 8 lines · Jiajun Chen
Used_memslotsis equal todev->mem->nregionsnow, it is true for vhost kernel, but not for vhost user, which uses the memory regions that have file descriptor. In fact, not all of the memory regions have file descriptor.
It is usefully in some scenarios, e.g.used_memslotsis 8, and only 5 memory slots can be used by vhost user, it is failed to hot plug a new memory RAM becausevhost_has_free_slotjust returned false, but we can hot plug it safely in fact.
At this point I dropped this, if you are going to resubmit pls include data on qemu invocation that manifests the problem.
Quoted 247 lines
--
ChangeList:
v3:
-makeused_memslotsa member of structvhost_devinstead of a global static valuev2:
-eliminating uselessused_memslots_exceededvariable andused_memslots_is_exceeded()APIv1:
-vhost-user: add separate memslot counter for vhost-userSigned-off-by Jiajun Chen <[email protected]>Signed-off-by Jianjay Zhou <[email protected]>--- hw/virtio/vhost-backend.c | 12 ++++++++++ hw/virtio/vhost-user.c | 25 +++++++++++++++++++++ hw/virtio/vhost.c | 37 +++++++++++++++++++++++-------- include/hw/virtio/vhost-backend.h | 5 +++++ include/hw/virtio/vhost.h | 1 + net/vhost-user.c | 7 ++++++ 6 files changed, 78 insertions(+), 9 deletions(-)hw/virtio/vhost-backend.c +12
-0@@ -238,6 +238,16 @@ static void vhost_kernel_set_iotlb_callback(struct vhost_dev *dev,qemu_set_fd_handler ((uintptr_t )dev->opaque,NULL ,NULL ,NULL ); } +static void vhost_kernel_set_used_memslots (struct vhost_dev *dev) +{ + dev->used_memslots = dev->mem->nregions; +} + +static unsigned int vhost_kernel_get_used_memslots (struct vhost_dev *dev) +{ +return dev->used_memslots; +} +static const VhostOps kernel_ops = { .backend_type =VHOST_BACKEND_TYPE_KERNEL , .vhost_backend_init = vhost_kernel_init,@@ -269,6 +279,8 @@ static const VhostOps kernel_ops = {#endif /* CONFIG_VHOST_VSOCK */ .vhost_set_iotlb_callback = vhost_kernel_set_iotlb_callback, .vhost_send_device_iotlb_msg = vhost_kernel_send_device_iotlb_msg, + .vhost_set_used_memslots = vhost_kernel_set_used_memslots, + .vhost_get_used_memslots = vhost_kernel_get_used_memslots, };#endif hw/virtio/vhost-user.c +25
-0@@ -2354,6 +2354,29 @@ void vhost_user_cleanup(VhostUserState *user) user->chr =NULL ; } +static void vhost_user_set_used_memslots (struct vhost_dev *dev) +{ +int i; + dev->used_memslots =0 ; + +for (i =0 ; i < dev->mem->nregions; ++i) { +struct vhost_memory_region *reg = dev->mem->regions + i; +ram_addr_t offset; +MemoryRegion *mr; +int fd; + + mr =vhost_user_get_mr_data (reg->userspace_addr, &offset, &fd); +if (mr && fd >0 ) { + dev->used_memslots++; + } + } +} + +static unsigned int vhost_user_get_used_memslots (struct vhost_dev *dev) +{ +return dev->used_memslots; +} +const VhostOps user_ops = { .backend_type =VHOST_BACKEND_TYPE_USER , .vhost_backend_init = vhost_user_backend_init,@@ -2387,4 +2410,6 @@ const VhostOps user_ops = { .vhost_backend_mem_section_filter = vhost_user_mem_section_filter, .vhost_get_inflight_fd = vhost_user_get_inflight_fd, .vhost_set_inflight_fd = vhost_user_set_inflight_fd, + .vhost_set_used_memslots = vhost_user_set_used_memslots, + .vhost_get_used_memslots = vhost_user_get_used_memslots, };hw/virtio/vhost.c +28
-9@@ -45,20 +45,20 @@static struct vhost_log *vhost_log;static struct vhost_log *vhost_log_shm; -static unsigned int used_memslots; static QLIST_HEAD(, vhost_dev) vhost_devices =QLIST_HEAD_INITIALIZER (vhost_devices);bool vhost_has_free_slot (void ) { -unsigned int slots_limit = ~0U ;struct vhost_dev *hdev; QLIST_FOREACH(hdev, &vhost_devices, entry) { -unsigned int r = hdev->vhost_ops->vhost_backend_memslots_limit (hdev); - slots_limit =MIN (slots_limit, r); +if (hdev->vhost_ops->vhost_get_used_memslots (hdev) >= + hdev->vhost_ops->vhost_backend_memslots_limit (hdev)) { +return false; + } } -return slots_limit > used_memslots; +return true; } static void vhost_dev_sync_region(struct vhost_dev *dev,@@ -502,7 +502,6 @@ static void vhost_commit(MemoryListener *listener) dev->n_mem_sections *sizeof dev->mem->regions[0 ]; dev->mem =g_realloc (dev->mem, regions_size); dev->mem->nregions = dev->n_mem_sections; - used_memslots = dev->mem->nregions;for (i =0 ; i < dev->n_mem_sections; i++) {struct vhost_memory_region *cur_vmr = dev->mem->regions + i;struct MemoryRegionSection *mrs = dev->mem_sections + i;@@ -678,6 +677,7 @@ static void vhost_region_add_section(struct vhost_dev *dev, dev->tmp_sections[dev->n_tmp_sections -1 ].fv =NULL ;memory_region_ref (section->mr); } + dev->vhost_ops->vhost_set_used_memslots (dev); }/* Used for both add and nop callbacks */ @@ -693,6 +693,17 @@ static void vhost_region_addnop(MemoryListener *listener,vhost_region_add_section (dev, section); } +static void vhost_region_del (MemoryListener *listener, +MemoryRegionSection *section) +{ + struct vhost_dev *dev = container_of(listener, struct vhost_dev, + memory_listener); +if (!vhost_section (dev, section)) { +return ; + } + dev->vhost_ops->vhost_set_used_memslots (dev); +} +static void vhost_iommu_unmap_notify (IOMMUNotifier *n,IOMMUTLBEntry *iotlb) { struct vhost_iommu *iommu = container_of(n, struct vhost_iommu, n);@@ -1300,6 +1311,7 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque, hdev->memory_listener = (MemoryListener ) { .begin = vhost_begin, .commit = vhost_commit, + .region_del = vhost_region_del, .region_add = vhost_region_addnop, .region_nop = vhost_region_addnop, .log_start = vhost_log_start,@@ -1346,9 +1358,16 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,memory_listener_register (&hdev->memory_listener, &address_space_memory);QLIST_INSERT_HEAD (&vhost_devices, hdev, entry); -if (used_memslots > hdev->vhost_ops->vhost_backend_memslots_limit (hdev)) { -error_report ("vhost backend memory slots limit is less" -" than current number of present memory slots" ); +/* +* If we started a VM without any vhost device, +* for the first time vhost device hot-plug +* (vhost_get_used_memslots is always 0), +* so it needs to double check here. +*/ +if (hdev->vhost_ops->vhost_get_used_memslots (hdev) > + hdev->vhost_ops->vhost_backend_memslots_limit (hdev)) { +error_report ("vhost backend memory slots limit is less than" +" current number of present memory slots" ); r =-1 ;if (busyloop_timeout) { goto fail_busyloop;include/hw/virtio/vhost-backend.h +5
-0@@ -124,6 +124,9 @@ typedef int (*vhost_get_device_id_op)(struct vhost_dev *dev, uint32_t *dev_id);typedef bool (*vhost_force_iommu_op )(struct vhost_dev *dev); +typedef void (*vhost_set_used_memslots_op )(struct vhost_dev *dev); +typedef unsigned int (*vhost_get_used_memslots_op )(struct vhost_dev *dev); + typedef struct VhostOps { VhostBackendType backend_type; vhost_backend_init vhost_backend_init;@@ -168,6 +171,8 @@ typedef struct VhostOps {vhost_vq_get_addr_op vhost_vq_get_addr;vhost_get_device_id_op vhost_get_device_id;vhost_force_iommu_op vhost_force_iommu; +vhost_set_used_memslots_op vhost_set_used_memslots; +vhost_get_used_memslots_op vhost_get_used_memslots; } VhostOps;extern const VhostOps user_ops;include/hw/virtio/vhost.h +1
-0@@ -90,6 +90,7 @@ struct vhost_dev { QLIST_HEAD(, vhost_iommu) iommu_list;IOMMUNotifier n;const VhostDevConfigOps *config_ops; +unsigned int used_memslots; };struct vhost_net {net/vhost-user.c +7
-0@@ -20,6 +20,7 @@#include "qemu/error-report.h" #include "qemu/option.h" #include "trace.h" +#include "include/hw/virtio/vhost.h" typedef struct NetVhostUserState { NetClientState nc;@@ -347,6 +348,12 @@ static int net_vhost_user_init(NetClientState *peer, const char *device,qemu_chr_fe_set_handlers (&s->chr,NULL ,NULL , net_vhost_user_event,NULL , nc0->name,NULL , true); + +if (!vhost_has_free_slot ()) { +error_report ("used memslots exceeded the backend limit, quit " +"loop" ); + goto err; + } }while (!s->started);assert (s->vhost_net);Signature · 1 line
2.27.0.dirty