Add support for generation of interrupts from the device directly to the
VM to the VCPU thus avoiding the overhead on the host CPU.
When supported, the driver will attempt to allocate vectors for each
data virtqueue. If a vector for a virtqueue cannot be provided it will
use the QP mode where notifications go through the driver.
In addition, we add a shutdown callback to make sure allocated
interrupts are released in case of shutdown to allow clean shutdown.
Signed-off-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Message-Id: <20230607190007.290505-1-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Extend the possible list for features that can be supported by firmware.
Note that different versions of firmware may or may not support these
features. The driver is made aware of them by querying the firmware.
While doing this, improve the code so we use enum names instead of hard
coded numerical values.
The new features supported by the driver are the following:
VIRTIO_NET_F_MRG_RXBUF
VIRTIO_NET_F_HOST_ECN
VIRTIO_NET_F_GUEST_ECN
VIRTIO_NET_F_GUEST_TSO6
VIRTIO_NET_F_GUEST_TSO4
Reviewed-by: Si-Wei Liu <si-wei.liu@oracle.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20230321112809.221432-3-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Eugenio Pérez Martin <eperezma@redhat.com>
Following patch adds driver support for VIRTIO_NET_F_MRG_RXBUF.
Current firmware versions show degradation in packet rate when using
MRG_RXBUF. Users who favor memory saving over packet rate could enable
this feature but we want to keep it off by default.
One can still enable it when creating the vdpa device using vdpa tool by
providing features that include it.
For example:
$ vdpa dev add name vdpa0 mgmtdev pci/0000:86:00.2 device_features 0x300cb982b
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20230321112809.221432-2-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Si-Wei Liu <si-wei.liu@oracle.com>
Current code ignores link state updates if VIRTIO_NET_F_STATUS was not
negotiated. However, link state updates could be received before feature
negotiation was completed , therefore causing link state events to be
lost, possibly leaving the link state down.
Modify the code so link state notifier is registered after DRIVER_OK was
negotiated and carry the registration only if
VIRTIO_NET_F_STATUS was negotiated. Unregister the notifier when the
device is reset.
Fixes: 033779a708 ("vdpa/mlx5: make MTU/STATUS presence conditional on feature bits")
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20230417110343.138319-1-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The right place to add the debugfs create is in
setup_driver() and remove it in teardown_driver().
Current code adds the debugfs when creating the device but resetting a
device will remove the debugfs subtree and subsequent set_driver will
not be able to create the files since the debugfs pointer is NULL.
Fixes: 2942210043 ("vdpa/mlx5: Add debugfs subtree")
Signed-off-by: Eli Cohen <elic@nvidia.com>
v3 -> v4:
Fix error flow in setup_driver()
Message-Id: <20230403114039.11102-1-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Otherwise the virtqueue object to instate could point to invalid address
that was unmapped from the MTT:
mlx5_core 0000:41:04.2: mlx5_cmd_out_err:782:(pid 8321):
CREATE_GENERAL_OBJECT(0xa00) op_mod(0xd) failed, status
bad parameter(0x3), syndrome (0x5fa1c), err(-22)
Fixes: cae15c2ed8 ("vdpa/mlx5: Implement susupend virtqueue callback")
Cc: Eli Cohen <elic@nvidia.com>
Signed-off-by: Si-Wei Liu <si-wei.liu@oracle.com>
Reviewed-by: Eli Cohen <elic@nvidia.com>
Message-Id: <1676424640-11673-1-git-send-email-si-wei.liu@oracle.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
This patch implements features provisioning for mlx5_vdpa.
1) Validate the provisioned features are a subset of the parent
features.
2) Clearing features that are not wanted by userspace.
For example:
# vdpa mgmtdev show
pci/0000:41:04.2:
supported_classes net
max_supported_vqs 65
dev_features CSUM GUEST_CSUM MTU MAC HOST_TSO4 HOST_TSO6 STATUS CTRL_VQ CTRL_VLAN MQ CTRL_MAC_ADDR VERSION_1 ACCESS_PLATFORM
1) Provision vDPA device with all features derived from the parent
# vdpa dev add name vdpa1 mgmtdev pci/0000:41:04.2
# vdpa dev config show
vdpa1: mac e4:11:c6:d3:45:f0 link up link_announce false max_vq_pairs 1 mtu 1500
negotiated_features CSUM GUEST_CSUM MTU HOST_TSO4 HOST_TSO6 STATUS CTRL_VQ CTRL_VLAN MQ CTRL_MAC_ADDR VERSION_1 ACCESS_PLATFORM
2) Provision vDPA device with a subset of parent features
# vdpa dev add name vdpa1 mgmtdev pci/0000:41:04.2 device_features 0x300020000
# vdpa dev config show
vdpa1:
negotiated_features CTRL_VQ VERSION_1 ACCESS_PLATFORM
Signed-off-by: Si-Wei Liu <si-wei.liu@oracle.com>
Reviewed-by: Eli Cohen <elic@nvidia.com>
Message-Id: <1675725124-7375-7-git-send-email-si-wei.liu@oracle.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The spec says:
mtu only exists if VIRTIO_NET_F_MTU is set
status only exists if VIRTIO_NET_F_STATUS is set
We should only present MTU and STATUS conditionally depending on
the feature bits.
Signed-off-by: Si-Wei Liu <si-wei.liu@oracle.com>
Reviewed-by: Eli Cohen <elic@nvidia.com>
Message-Id: <1675725124-7375-6-git-send-email-si-wei.liu@oracle.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Initialize itolb spinlock.
Fixes: 5262912ef3 ("vdpa/mlx5: Add support for control VQ and MAC setting")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20230206122016.1149373-1-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Clearing the mr struct erases the lock owner and causes warnings to be
emitted. It is not required to clear the mr so remove the memset call.
Fixes: 94abbccdf2 ("vdpa/mlx5: Add shared memory registration code")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20230206121956.1149356-1-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
When creating a memory key, the key value should be assigned to the
passed pointer and not or'ed to.
No functional issue was observed due to this bug.
Fixes: 29064bfdab ("vdpa/mlx5: Add support library for mlx5 VDPA implementation")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20230205072906.1108194-1-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patch implements per virtqueue dma device for mlx5_vdpa. This is
needed for virtio_vdpa to work for CVQ which is backed by vringh but
not DMA. We simply advertise the vDPA device itself as the DMA device
for CVQ then DMA API can simply use PA so the identical mapping for
CVQ can still be used. Otherwise the identical (1:1) mapping won't
work when platform IOMMU is enabled since the IOVA is allocated on
demand which is not necessarily the PA.
This fixes the following crash when mlx5 vDPA device is bound to
virtio-vdpa with platform IOMMU enabled but not in passthrough mode:
BUG: unable to handle page fault for address: ff2fb3063deb1002
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 1393001067 P4D 1393002067 PUD 0
Oops: 0000 [#1] PREEMPT SMP NOPTI
CPU: 55 PID: 8923 Comm: kworker/u112:3 Kdump: loaded Not tainted 6.1.0+ #7
Hardware name: Dell Inc. PowerEdge R750/0PJ80M, BIOS 1.5.4 12/17/2021
Workqueue: mlx5_vdpa_wq mlx5_cvq_kick_handler [mlx5_vdpa]
RIP: 0010:vringh_getdesc_iotlb+0x93/0x1d0 [vringh]
Code: 14 25 40 ef 01 00 83 82 c0 0a 00 00 01 48 2b 05 93 5a 1b ea 8b 4c 24 14 48 c1 f8 06 48 c1 e0 0c 48 03 05 90 5a 1b ea 48 01 c8 <0f> b7 00 83 aa c0 0a 00 00 01 65 ff 0d bc e4 41 3f 0f 84 05 01 00
RSP: 0018:ff46821ba664fdf8 EFLAGS: 00010282
RAX: ff2fb3063deb1002 RBX: 0000000000000a20 RCX: 0000000000000002
RDX: ff2fb318d2f94380 RSI: 0000000000000002 RDI: 0000000000000001
RBP: ff2fb3065e832410 R08: ff46821ba664fe00 R09: 0000000000000001
R10: 0000000000000000 R11: 000000000000000d R12: ff2fb3065e832488
R13: ff2fb3065e8324a8 R14: ff2fb3065e8324c8 R15: ff2fb3065e8324a8
FS: 0000000000000000(0000) GS:ff2fb3257fac0000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ff2fb3063deb1002 CR3: 0000001392010006 CR4: 0000000000771ee0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
PKRU: 55555554
Call Trace:
<TASK>
mlx5_cvq_kick_handler+0x89/0x2b0 [mlx5_vdpa]
process_one_work+0x1e2/0x3b0
? rescuer_thread+0x390/0x390
worker_thread+0x50/0x3a0
? rescuer_thread+0x390/0x390
kthread+0xd6/0x100
? kthread_complete_and_exit+0x20/0x20
ret_from_fork+0x1f/0x30
</TASK>
Reviewed-by: Eli Cohen <elic@nvidia.com>
Tested-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230119061525.75068-6-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
For each interface, either VLAN tagged or untagged, add two hardware
counters: one for unicast and another for multicast. The counters count
RX packets and bytes and can be read through debugfs:
$ cat /sys/kernel/debug/mlx5/mlx5_core.sf.1/vdpa-0/rx/untagged/mcast/packets
$ cat /sys/kernel/debug/mlx5/mlx5_core.sf.1/vdpa-0/rx/untagged/ucast/bytes
This feature is controlled via the config option
MLX5_VDPA_STEERING_DEBUG. It is off by default as it may have some
impact on performance.
includes a fixup By Yang Yingliang <yangyingliang@huawei.com>:
vdpa/mlx5: fix check wrong pointer in mlx5_vdpa_add_mac_vlan_rules()
The local variable 'rule' is not used anymore, fix return value
check after calling mlx5_add_flow_rules().
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-9-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Message-Id: <20230104074418.1737510-1-yangyingliang@huawei.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Eli Cohen <elic@nvidia.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Add debugfs subtree and expose flow table ID and TIR number. This
information can be used by external tools to do extended
troubleshooting.
The information can be retrieved like so:
$ cat /sys/kernel/debug/mlx5/mlx5_core.sf.1/vdpa-0/rx/table_id
$ cat /sys/kernel/debug/mlx5/mlx5_core.sf.1/vdpa-0/rx/tirn
Reviewed-by: Si-Wei Liu <si-wei.liu@oracle.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-8-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Move some definitions from mlx5_vnet.c to newly added header file
mlx5_vnet.h. We need these definitions for the following patches that
add debugfs tree to expose information vital for debug.
Reviewed-by: Si-Wei Liu <si-wei.liu@oracle.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-7-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Variable i is just being incremented and it's never used
anywhere else. The variable and the increment are redundant so
remove it.
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Message-Id: <20221024133756.2158497-1-colin.i.king@gmail.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
When qemu uses different address spaces for data and control virtqueues,
the current code would overwrite the control virtqueue iotlb through the
dup_iotlb call. Fix this by referring to the address space identifier
and the group to asid mapping to determine which mapping needs to be
updated. We also move the address space logic from mlx5 net to core
directory.
Reported-by: Eugenio Pérez <eperezma@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-6-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Eugenio Pérez <eperezma@redhat.com>
event_handler runs under atomic context and may not acquire reslock. We
can still guarantee that the handler won't be called after suspend by
clearing nb_registered, unregistering the handler and flushing the
workqueue.
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-5-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Delete the old MAC from the table and not the new one which is not there
yet.
Fixes: baf2ad3f6a ("vdpa/mlx5: Add RX MAC VLAN filter support")
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-4-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Check if VIRTIO_NET_F_CTRL_VLAN is negotiated and return error if
control VQ command is received.
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-3-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Eugenio Pérez <eperezma@redhat.com>
Set the VLAN id to the header values field instead of overwriting the
headers criteria field.
Before this fix, VLAN filtering would not really work and tagged packets
would be forwarded unfiltered to the TIR.
Fixes: baf2ad3f6a ("vdpa/mlx5: Add RX MAC VLAN filter support")
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-2-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
RQT objects require that a power of two value be configured for both
rqt_max_size and rqt_actual size.
For create_rqt, make sure to round up to the power of two the value of
given by the user who created the vdpa device and given by
ndev->rqt_size. The actual size is also rounded up to the power of two
using the current number of VQs given by ndev->cur_num_vqs.
Same goes with modify_rqt where we need to make sure act size is power
of two based on the new number of QPs.
Without this patch, attempt to create a device with non power of two QPs
would result in error from firmware.
Fixes: 52893733f2 ("vdpa/mlx5: Add multiqueue support")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20220912125019.833708-1-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Initialize err local variable to return -EAGAIN if the asid cannot be
found thus avoiding returning uninitialized value.
Fixes: 8fcd20c307 ("vdpa/mlx5: Support different address spaces for control and data")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20220811134010.952291-1-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Partition virtqueues to two different address spaces: one for control
virtqueue which is implemented in software, and one for data virtqueues.
Based-on: <20220526124338.36247-1-eperezma@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20220714113927.85729-3-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Implement the suspend callback allowing to suspend the virtqueues so
they stop processing descriptors. This is required to allow to query a
consistent state of the virtqueue while live migration is taking place.
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20220714113927.85729-2-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Using eth_broadcast_addr() to assign broadcast address instead
of memset().
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Xu Qiang <xuqiang36@huawei.com>
Message-Id: <20220704021405.64545-1-xuqiang36@huawei.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Currently, CVQ vringh is initialized inside setup_virtqueues() which is
called every time a memory update is done. This is undesirable since it
resets all the context of the vring, including the available and used
indices.
Move the initialization to mlx5_vdpa_set_status() when
VIRTIO_CONFIG_S_DRIVER_OK is set.
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20220613075958.511064-2-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Eugenio Pérez <eperezma@redhat.com>
The control VQ specific information is stored in the dedicated struct
mlx5_control_vq. When the callback is updated through
mlx5_vdpa_set_vq_cb(), make sure to update the control VQ struct.
Fixes: 5262912ef3 ("vdpa/mlx5: Add support for control VQ and MAC setting")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20220613075958.511064-1-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com)
These lines were supposed to be indented.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Message-Id: <Yp71IYMP+QfuCJ8t@kili>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Eli Cohen <elic@nvidia.com>
Acked-by: Si-Wei Liu <si-wei.liu@oracle.com>
Return success if we were able to delete a vlan. The current code
always returns failure.
Fixes: baf2ad3f6a ("vdpa/mlx5: Add RX MAC VLAN filter support")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Message-Id: <Yp709f1g9NcMBCHg@kili>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Eli Cohen <elic@nvidia.com>
Acked-by: Si-Wei Liu <si-wei.liu@oracle.com>
Delete the redundant word 'is'.
Signed-off-by: Xiang wangx <wangxiang@cdjrlc.com>
Message-Id: <20220604143858.16073-1-wangxiang@cdjrlc.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Support HW offloaded filtering of MAC/VLAN packets.
To allow that, we add a handler to handle VLAN configurations coming
through the control VQ. Two operations are supported.
1. Adding VLAN - in this case, an entry will be added to the RX flow
table that will allow the combination of the MAC/VLAN to be
forwarded to the TIR.
2. Removing VLAN - will remove the entry from the flow table,
effectively blocking such packets from going through.
Currently the control VQ does not propagate changes to the MAC of the
VLAN device so we always use the MAC of the parent device.
Examples:
1. Create vlan device:
$ ip link add link ens1 name ens1.8 type vlan id 8
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20220411122942.225717-4-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
The flow counter has been introduced in early versions of the driver to
aid in debugging. It is no longer needed and can harm performance.
Remove it.
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20220411122942.225717-2-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patches introduces the multiple address spaces support for vDPA
device. This idea is to identify a specific address space via an
dedicated identifier - ASID.
During vDPA device allocation, vDPA device driver needs to report the
number of address spaces supported by the device then the DMA mapping
ops of the vDPA device needs to be extended to support ASID.
This helps to isolate the environments for the virtqueue that will not
be assigned directly. E.g in the case of virtio-net, the control
virtqueue will not be assigned directly to guest.
As a start, simply claim 1 virtqueue groups and 1 address spaces for
all vDPA devices. And vhost-vDPA will simply reject the device with
more than 1 virtqueue groups or address spaces.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Gautam Dawar <gdawar@xilinx.com>
Message-Id: <20220330180436.24644-7-gdawar@xilinx.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patch introduces virtqueue groups to vDPA device. The virtqueue
group is the minimal set of virtqueues that must share an address
space. And the address space identifier could only be attached to
a specific virtqueue group.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Gautam Dawar <gdawar@xilinx.com>
Message-Id: <20220330180436.24644-6-gdawar@xilinx.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reading statistics could be done intensively and by several processes
concurrently. Reader's lock is sufficient in this case.
Change reslock from mutex to a rwsem.
Suggested-by: Si-Wei Liu <si-wei.liu@oracle.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20220518133804.1075129-7-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Implement the get_vq_stats calback of vdpa_config_ops to return the
statistics for a virtqueue.
The statistics are provided as vendor specific statistics where the
driver provides a pair of attribute name and attribute value.
Currently supported are received descriptors and completed descriptors.
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20220518133804.1075129-6-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The current code evaluates RQT size based on the configured number of
virtqueues. This can raise an issue in the following scenario:
Assume MQ was negotiated.
1. mlx5_vdpa_set_map() gets called.
2. handle_ctrl_mq() is called setting cur_num_vqs to some value, lower
than the configured max VQs.
3. A second set_map gets called, but now a smaller number of VQs is used
to evaluate the size of the RQT.
4. handle_ctrl_mq() is called with a value larger than what the RQT can
hold. This will emit errors and the driver state is compromised.
To fix this, we use a new field in struct mlx5_vdpa_net to hold the
required number of entries in the RQT. This value is evaluated in
mlx5_vdpa_set_driver_features() where we have the negotiated features
all set up.
In addition to that, we take into consideration the max capability of RQT
entries early when the device is added so we don't need to take consider
it when creating the RQT.
Last, we remove the use of mlx5_vdpa_max_qps() which just returns the
max_vas / 2 and make the code clearer.
Fixes: 52893733f2 ("vdpa/mlx5: Add multiqueue support")
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Currently, CVQ doesn't have any synchronization with the driver
status. Then CVQ emulation code run in the middle of:
1) device reset
2) device status changed
3) map updating
The will lead several unexpected issue like trying to execute CVQ
command after the driver has been teared down.
Fixing this by using reslock to synchronize CVQ emulation code with
the driver status changing:
- protect the whole device reset, status changing and set_map()
updating with reslock
- protect the CVQ handler with the reslock and check
VIRTIO_CONFIG_S_DRIVER_OK in the CVQ handler
This will guarantee that:
1) CVQ handler won't work if VIRTIO_CONFIG_S_DRIVER_OK is not set
2) CVQ handler will see a consistent state of the driver instead of
the partial one when it is running in the middle of the
teardown_driver() or setup_driver().
Cc: 5262912ef3 ("vdpa/mlx5: Add support for control VQ and MAC setting")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20220329042109.4029-2-jasowang@redhat.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Eli Cohen <elic@nvidia.com>
A userspace triggerable infinite loop could happen in
mlx5_cvq_kick_handler() if userspace keeps sending a huge amount of
cvq requests.
Fixing this by introducing a quota and re-queue the work if we're out
of the budget (currently the implicit budget is one) . While at it,
using a per device work struct to avoid on demand memory allocation
for cvq.
Fixes: 5262912ef3 ("vdpa/mlx5: Add support for control VQ and MAC setting")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20220329042109.4029-1-jasowang@redhat.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Eli Cohen <elic@nvidia.com>
If mlx5_vdpa gets unloaded while a VM is running, the workqueue will be
destroyed. However, vhost might still have reference to the kick
function and might attempt to push new works. This could lead to null
pointer dereference.
To fix this, set mvdev->wq to NULL just before destroying and verify
that the workqueue is not NULL in mlx5_vdpa_kick_vq before attempting to
push a new work.
Fixes: 5262912ef3 ("vdpa/mlx5: Add support for control VQ and MAC setting")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Link: https://lore.kernel.org/r/20220321141303.9586-1-elic@nvidia.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
When MAC Address has been modified in guest, we only re-add the
Mac to mpfs, it is not enough, because the guest network will
not work correctly: the reply package from outside will go
straight away to the host VF net interface.
This patch recreate the flow rules, and make it work correctly.
Signed-off-by: Michael Qiu <qiudayu@archeros.com>
Link: https://lore.kernel.org/r/1648446492-17614-1-git-send-email-08005325@163.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Eli Cohen <elic@nvidia.com>
Allow an admin creating a vdpa device to specify the max MTU for the
net device.
For example, to create a device with max MTU of 1000, the following
command can be used:
$ vdpa dev add name vdpa-a mgmtdev auxiliary/mlx5_core.sf.1 mtu 1000
This configuration mechanism assumes that vdpa is the sole real user of
the function. mlx5_core could theoretically change the mtu of the
function using the ip command on the mlx5_core net device but this
should not be done.
Reviewed-by: Si-Wei Liu<si-wei.liu@oracle.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Link: https://lore.kernel.org/r/20220221121927.194728-1-elic@nvidia.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
When control vq receives a VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET command
request from the driver, presently there is no validation against the
number of queue pairs to configure, or even if multiqueue had been
negotiated or not is unverified. This may lead to kernel panic due to
uninitialized resource for the queues were there any bogus request
sent down by untrusted driver. Tie up the loose ends there.
Fixes: 52893733f2 ("vdpa/mlx5: Add multiqueue support")
Signed-off-by: Si-Wei Liu <si-wei.liu@oracle.com>
Link: https://lore.kernel.org/r/1642206481-30721-4-git-send-email-si-wei.liu@oracle.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Eli Cohen <elic@nvidia.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Per VIRTIO v1.1 specification, section 5.1.3.1 Feature bit requirements:
"VIRTIO_NET_F_MQ Requires VIRTIO_NET_F_CTRL_VQ".
There's assumption in the mlx5_vdpa multiqueue code that MQ must come
together with CTRL_VQ. However, there's nowhere in the upper layer to
guarantee this assumption would hold. Were there an untrusted driver
sending down MQ without CTRL_VQ, it would compromise various spots for
e.g. is_index_valid() and is_ctrl_vq_idx(). Although this doesn't end
up with immediate panic or security loophole as of today's code, the
chance for this to be taken advantage of due to future code change is
not zero.
Harden the crispy assumption by failing the set_driver_features() call
when seeing (MQ && !CTRL_VQ). For that end, verify_min_features() is
renamed to verify_driver_features() to reflect the fact that it now does
more than just validate the minimum features. verify_driver_features()
is now used to accommodate various checks against the driver features
for set_driver_features().
Signed-off-by: Si-Wei Liu <si-wei.liu@oracle.com>
Link: https://lore.kernel.org/r/1642206481-30721-3-git-send-email-si-wei.liu@oracle.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Eli Cohen <elic@nvidia.com>
Acked-by: Jason Wang <jasowang@redhat.com>
partial support for < MAX_ORDER - 1 granularity for virtio-mem
driver_override for vdpa
sysfs ABI documentation for vdpa
multiqueue config support for mlx5 vdpa
Misc fixes, cleanups.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
-----BEGIN PGP SIGNATURE-----
iQFDBAABCAAtFiEEXQn9CHHI+FuUyooNKB8NuNKNVGkFAmHiDHkPHG1zdEByZWRo
YXQuY29tAAoJECgfDbjSjVRpVT4H/3Veixt3uYPOmuLU2tSx+8X+sFTtik81hyiE
okz5fRJrxxA8SqS76FnmO10FS4hlPOGNk0Z5WVhr0yihwFvPLvpCM/xi2Lmrz9I7
pB0sXOIocEL1xApsxukR9K1Twpb2hfYsflbJYUVlRfhS5G0izKJNZp5I7OPrzd80
vVNNDWKW2iLDlfqsavumI4Kvm4nsFuCHG03jzMtcIa7YTXYV3DORD4ZGFFVUOIQN
t5F74TznwHOeYgJeg7TzjFjfPWmXjLetvx10QX1A1uOvwppWW/QY6My0UafTXNXj
VB3gOwJPf+gxXAXl/4bafq4NzM0xys6cpcPpjvhmU+erY4UuyAU=
=Y1eO
-----END PGP SIGNATURE-----
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Pull virtio updates from Michael Tsirkin:
"virtio,vdpa,qemu_fw_cfg: features, cleanups, and fixes.
- partial support for < MAX_ORDER - 1 granularity for virtio-mem
- driver_override for vdpa
- sysfs ABI documentation for vdpa
- multiqueue config support for mlx5 vdpa
- and misc fixes, cleanups"
* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost: (42 commits)
vdpa/mlx5: Fix tracking of current number of VQs
vdpa/mlx5: Fix is_index_valid() to refer to features
vdpa: Protect vdpa reset with cf_mutex
vdpa: Avoid taking cf_mutex lock on get status
vdpa/vdpa_sim_net: Report max device capabilities
vdpa: Use BIT_ULL for bit operations
vdpa/vdpa_sim: Configure max supported virtqueues
vdpa/mlx5: Report max device capabilities
vdpa: Support reporting max device capabilities
vdpa/mlx5: Restore cur_num_vqs in case of failure in change_num_qps()
vdpa: Add support for returning device configuration information
vdpa/mlx5: Support configuring max data virtqueue
vdpa/mlx5: Fix config_attr_mask assignment
vdpa: Allow to configure max data virtqueues
vdpa: Read device configuration only if FEATURES_OK
vdpa: Sync calls set/get config/status with cf_mutex
vdpa/mlx5: Distribute RX virtqueues in RQT object
vdpa: Provide interface to read driver features
vdpa: clean up get_config_size ret value handling
virtio_ring: mark ring unused on error
...
Modify the code such that ndev->cur_num_vqs better reflects the actual
number of data virtqueues. The value can be accurately realized after
features have been negotiated.
This is to prevent possible failures when modifying the RQT object if
the cur_num_vqs bears invalid value.
No issue was actually encountered but this also makes the code more
readable.
Fixes: c5a5cd3d3217 ("vdpa/mlx5: Support configuring max data virtqueue")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Link: https://lore.kernel.org/r/20220111183400.38418-5-elic@nvidia.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Si-Wei Liu<si-wei.liu@oracle.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Make sure the decision whether an index received through a callback is
valid or not consults the negotiated features.
The motivation for this was due to a case encountered where I shut down
the VM. After the reset operation was called features were already
clear, I got get_vq_state() call which caused out array bounds
access since is_index_valid() reported the index value.
So this is more of not hit a bug since the call shouldn't have been made
first place.
Signed-off-by: Eli Cohen <elic@nvidia.com>
Link: https://lore.kernel.org/r/20220111183400.38418-4-elic@nvidia.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Si-Wei Liu<si-wei.liu@oracle.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Configure max supported virtqueues and features on the management
device.
This info can be retrieved using:
$ vdpa mgmtdev show
auxiliary/mlx5_core.sf.1:
supported_classes net
max_supported_vqs 257
dev_features CSUM GUEST_CSUM MTU HOST_TSO4 HOST_TSO6 STATUS CTRL_VQ MQ \
CTRL_MAC_ADDR VERSION_1 ACCESS_PLATFORM
Signed-off-by: Eli Cohen <elic@nvidia.com>
Link: https://lore.kernel.org/r/20220105114646.577224-12-elic@nvidia.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Si-Wei Liu<si-wei.liu@oracle.com>
