linux.git/drivers/vdpa/mlx5, branch master

vdpa/mlx5: Postpone MR deletion

2024-09-25T11:07:44+00:00

Currently, when a new MR is set up, the old MR is deleted. MR deletion
is about 30-40% the time of MR creation. As deleting the old MR is not
important for the process of setting up the new MR, this operation
can be postponed.

This series adds a workqueue that does MR garbage collection at a later
point. If the MR lock is taken, the handler will back off and
reschedule. The exception during shutdown: then the handler must
not postpone the work.

Note that this is only a speculative optimization: if there is some
mapping operation that is triggered while the garbage collector handler
has the lock taken, this operation it will have to wait for the handler
to finish.

Signed-off-by: Dragos Tatulea 
Reviewed-by: Cosmin Ratiu 
Message-Id: <20240830105838.2666587-9-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin

vdpa/mlx5: Introduce init/destroy for MR resources

2024-09-25T11:07:44+00:00

There's currently not a lot of action happening during
the init/destroy of MR resources. But more will be added
in the upcoming patches.

As the mr mutex lock init/destroy has been moved to these
new functions, the lifetime has now shifted away from
mlx5_vdpa_alloc_resources() / mlx5_vdpa_free_resources()
into these new functions. However, the lifetime at the
outer scope remains the same:
mlx5_vdpa_dev_add() / mlx5_vdpa_dev_free()

Signed-off-by: Dragos Tatulea 
Reviewed-by: Cosmin Ratiu 
Message-Id: <20240830105838.2666587-8-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin

vdpa/mlx5: Rename mr_mtx -> lock

2024-09-25T11:07:44+00:00

Now that the mr resources have their own namespace in the
struct, give the lock a clearer name.

Signed-off-by: Dragos Tatulea 
Reviewed-by: Cosmin Ratiu 
Acked-by: Eugenio Pérez 
Message-Id: <20240830105838.2666587-7-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin

vdpa/mlx5: Extract mr members in own resource struct

2024-09-25T11:07:43+00:00

Group all mapping related resources into their own structure.

Upcoming patches will add more members in this new structure.

Signed-off-by: Dragos Tatulea 
Reviewed-by: Cosmin Ratiu 
Acked-by: Eugenio Pérez 
Message-Id: <20240830105838.2666587-6-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin

vdpa/mlx5: Rename function

2024-09-25T11:07:43+00:00

A followup patch will use this name for something else.

Signed-off-by: Dragos Tatulea 
Reviewed-by: Cosmin Ratiu 
Message-Id: <20240830105838.2666587-5-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin

vdpa/mlx5: Delete direct MKEYs in parallel

2024-09-25T11:07:43+00:00

Use the async interface to issue MTT MKEY deletion.

This makes destroy_user_mr() on average 8x times faster.
This number is also dependent on the size of the MR being
deleted.

Signed-off-by: Dragos Tatulea 
Reviewed-by: Cosmin Ratiu 
Acked-by: Eugenio Pérez 
Message-Id: <20240830105838.2666587-4-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin

vdpa/mlx5: Create direct MKEYs in parallel

2024-09-25T11:07:43+00:00

Use the async interface to issue MTT MKEY creation.
Extra care is taken at the allocation of FW input commands
due to the MTT tables having variable sizes depending on
MR.

The indirect MKEY is still created synchronously at the
end as the direct MKEYs need to be filled in.

This makes create_user_mr() 3-5x faster, depending on
the size of the MR.

Signed-off-by: Dragos Tatulea 
Reviewed-by: Cosmin Ratiu 
Message-Id: <20240830105838.2666587-3-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin

vdpa/mlx5: Parallelize VQ suspend/resume for CVQ MQ command

2024-09-25T11:07:43+00:00

change_num_qps() is still suspending/resuming VQs one by one.
This change switches to parallel suspend/resume.

When increasing the number of queues the flow has changed a bit for
simplicity: the setup_vq() function will always be called before
resume_vqs(). If the VQ is initialized, setup_vq() will exit early. If
the VQ is not initialized, setup_vq() will create it and resume_vqs()
will resume it.

Signed-off-by: Dragos Tatulea 
Reviewed-by: Tariq Toukan 
Message-Id: <20240816090159.1967650-11-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin 
Acked-by: Eugenio Pérez 
Tested-by: Lei Yang

vdpa/mlx5: Small improvement for change_num_qps()

2024-09-25T11:07:43+00:00

change_num_qps() has a lot of multiplications by 2 to convert
the number of VQ pairs to number of VQs. This patch simplifies
the code by doing the VQP -> VQ count conversion at the beginning
in a variable.

Signed-off-by: Dragos Tatulea 
Reviewed-by: Tariq Toukan 
Message-Id: <20240816090159.1967650-10-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin 
Acked-by: Eugenio Pérez 
Tested-by: Lei Yang

vdpa/mlx5: Keep notifiers during suspend but ignore

2024-09-25T11:07:42+00:00

Unregistering notifiers is a costly operation. Instead of removing
the notifiers during device suspend and adding them back at resume,
simply ignore the call when the device is suspended.

At resume time call queue_link_work() to make sure that the device state
is propagated in case there were changes.

For 1 vDPA device x 32 VQs (16 VQPs) attached to a large VM (256 GB RAM,
32 CPUs x 2 threads per core), the device suspend time is reduced from
~13 ms to ~2.5 ms.

Signed-off-by: Dragos Tatulea 
Reviewed-by: Tariq Toukan 
Acked-by: Eugenio Pérez 
Message-Id: <20240816090159.1967650-9-dtatulea@nvidia.com>
Signed-off-by: Michael S. Tsirkin 
Tested-by: Lei Yang