Virtualization - Dec 2022 - [PATCH v2] virtio: vdpa: new SolidNET DPU driver.

Hi Jason,
Thanks for your comment.
> I wonder if vq_num and max_size_host_cfg is not about to be changed by
> the upper layer, any reason not fail the psnet_read_cfg()?
You're right, I can catch this error earlier in psnet_read_cfg.
> So during reset, the irqs were freed, but I don't see how they are
> allocated here.
The IRQs are allocated in probe and are freed in reset.
This assumes that the device won't be revived after reset.
I  will change it and request the IRQs in snet_set_status.
> Any reason for this reset in the remove()?
We must send a  SNET_MSG_DESTROY message to the DPU before removing the device.
This remove() makes sure that if somehow remove was called
before/without reset, the DPU will receive this message.


Alvaro

> We must send a  SNET_MSG_DESTROY message to the DPU before removing the
device.
> This remove() makes sure that if somehow remove was called
> before/without reset, the DPU will receive this message.
I meant that the message must be sent before the PCI device is removed.
The message is sent from the vDPA reset cb, and the reset in the PCI
remove cb makes sure that PCI remove wasn't called without calling
vDPA reset.
I can remove it if you think that this is not required.


Alvaro