openhcl_boot: Remove few kernel parameters for uio_hv_generic

[namancse]

Few of the module parameters that were earlier supported in uio_hv_generic driver no longer exist in kernel. Remove the logic which adds these parameters in kernel cmdline from openvmm.

[jstarks]

How does uio_hv_generic know the ring buffer size to use now?

[namancse]

Hi John, sorry for replying late, I missed the email due to some email filters.

This is the patch that adds the support to get ring size from a pre defined table.

https://lore.kernel.org/all/1711788723-8593-3-git-send-email-ssengar@linux.microsoft.com/

[mattkur]

Hi John, sorry for replying late, I missed the email due to some email filters.

This is the patch that adds the support to get ring size from a pre defined table.

https://lore.kernel.org/all/1711788723-8593-3-git-send-email-ssengar@linux.microsoft.com/

Thanks for making this cleanup in openhcl.

Few questions:

1. Do I understand properly that the command line parameters no longer exist, and therefore openhcl's setting them is redundant anyways?
2. Can you confirm that those changes are in the kernel used by openhcl?
3. Does the kernel now choose the same ring buffer size as what openhcl had chosen?

@jstarks : since you already commented ... any concerns?

[jstarks]

I am unhappy that the kernel hard codes ring buffer sizes--we have asked for but did not receive an interface to make this configurable from user mode. The kernel has no way of knowing what an appropriate ring buffer size is.

Having said that, I have no objections to this specific change, since the kernel behavior is what it is, and the old command line args were inflexible (since they applied globally to all channels).

[mattkur]

I am unhappy that the kernel hard codes ring buffer sizes--we have asked for but did not receive an interface to make this configurable from user mode. The kernel has no way of knowing what an appropriate ring buffer size is.

Having said that, I have no objections to this specific change, since the kernel behavior is what it is, and the old command line args were inflexible (since they applied globally to all channels).

Agreed.

Hi John, sorry for replying late, I missed the email due to some email filters.
This is the patch that adds the support to get ring size from a pre defined table.
https://lore.kernel.org/all/1711788723-8593-3-git-send-email-ssengar@linux.microsoft.com/

Thanks for making this cleanup in openhcl.

Few questions:

1. Do I understand properly that the command line parameters no longer exist, and therefore openhcl's setting them is redundant anyways?
2. Can you confirm that those changes are in the kernel used by openhcl?
3. Does the kernel now choose the same ring buffer size as what openhcl had chosen?

@jstarks : since you already commented ... any concerns?

@namancse I'm happy to sign off if you can confirm the questions from above. Thanks again for the feedback!

[namancse]

Hi Matt,
PFB my responses:

Do I understand properly that the command line parameters no longer exist, and therefore openhcl's setting them is redundant anyways?
Naman: Yes, this is true, the settings are redundant now.
Can you confirm that those changes are in the kernel used by openhcl?
Yes, the changes are there in OpenHCL kernel for quite some time now.
https://github.com/microsoft/OHCL-Linux-Kernel/blob/product/hcl-main/6.6/drivers/uio/uio_hv_generic.c#L42

Does the kernel now choose the same ring buffer size as what openhcl had chosen?
Naman: Yes, its same as earlier: 0x11000.
Earlier, we used to have settings which apply to all the devices, but now with the config table, individual devices can choose to have a preferred ring buffer sizes. There is a fallback mechanism to have default size if preferred size is not provided for that device. Below is the table:
https://github.com/microsoft/OHCL-Linux-Kernel/blob/product/hcl-main/6.6/drivers/hv/channel_mgmt.c#L38