Re: [PATCH v2] mshv: Fix deposit memory in MSHV_ROOT_HVCALL

[Nuno Das Neves <nunodasneves@linux.microsoft.com>]

On 10/17/2025 3:26 PM, Wei Liu wrote:
> On Fri, Oct 17, 2025 at 10:06:55PM +0000, Wei Liu wrote:
>> On Fri, Oct 17, 2025 at 11:58:17AM -0700, Nuno Das Neves wrote:
>>> When the MSHV_ROOT_HVCALL ioctl is executing a hypercall, and gets
>>> HV_STATUS_INSUFFICIENT_MEMORY, it deposits memory and then returns
>>> -EAGAIN to userspace. The expectation is that the VMM will retry.
>>>
>>> However, some VMM code in the wild doesn't do this and simply fails.
>>> Rather than force the VMM to retry, change the ioctl to deposit
>>> memory on demand and immediately retry the hypercall as is done with
>>> all the other hypercall helper functions.
>>>
>>> In addition to making the ioctl easier to use, removing the need for
>>> multiple syscalls improves performance.
>>>
>>> There is a complication: unlike the other hypercall helper functions,
>>> in MSHV_ROOT_HVCALL the input is opaque to the kernel. This is
>>> problematic for rep hypercalls, because the next part of the input
>>> list can't be copied on each loop after depositing pages (this was
>>> the original reason for returning -EAGAIN in this case).
>>>
>>> Introduce hv_do_rep_hypercall_ex(), which adds a 'rep_start'
>>> parameter. This solves the issue, allowing the deposit loop in
>>> MSHV_ROOT_HVCALL to restart a rep hypercall after depositing pages
>>> partway through.
>>>
>>> Signed-off-by: Nuno Das Neves <nunodasneves@linux.microsoft.com>
>>
>> In v1 you said you will add a "Fixes" tag. Where is it?
> 
> I added this:
> 
> Fixes: 621191d709b1 ("Drivers: hv: Introduce mshv_root module to expose /dev/mshv to VMMs")
> 
> Let me know if that's not correct.
> 
Oops! That's correct, thanks.

> Wei