Re: [PATCH v1 07/13] mshv: default to safe partition CPU features

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

New issues:
- [High] Functional regression and uninitialized stack memory use on ARM64 for legacy partition creation.

Pre-existing issues:
- [Medium] TOCTOU double-fetch race allows validation and use of flag-gated parameters outside their flag gate.
--

--- Patch [7]: [PATCH v1 07/13] mshv: default to safe partition CPU features ---
commit 85299e9eb98a0db0daae1adea69f8e739d17d8db
Author: Wei Liu <[email protected]>

mshv: default to safe partition CPU features

For legacy create-partition callers, disable all processor and xsave
features first, then enable only the default feature set known to be
supported. Passing an all-zero disabled feature mask can make SEV-SNP
partition creation fail on Hyper-V.

> diff --git a/drivers/hv/mshv_root_main.c b/drivers/hv/mshv_root_main.c
> index cc38507570ac4..11d3c86050b45 100644
> --- a/drivers/hv/mshv_root_main.c
> +++ b/drivers/hv/mshv_root_main.c
> @@ -2333,6 +2333,86 @@ static long mshv_ioctl_process_pt_flags(void __user *user_arg, u64 *pt_flags,
>  	disabled_procs = &cr_props->disabled_processor_features;
>  	disabled_xsave = &cr_props->disabled_processor_xsave_features;
>  
> +	/* Disable all processor features first. */
> +	for (i = 0; i < HV_PARTITION_PROCESSOR_FEATURES_BANKS; i++)
> +		disabled_procs->as_uint64[i] = U64_MAX;
> +
> +#if IS_ENABLED(CONFIG_X86_64)
> +	/* Enable default features that are known to be supported. */

[ ... ]

> +	disabled_xsave->as_uint64 = U64_MAX;
> +	disabled_xsave->xsave_support = 0;
> +	disabled_xsave->xsaveopt_support = 0;
> +	disabled_xsave->avx_support = 0;
> +#endif
> +
>  	/* Check if user provided newer struct with feature fields */
>  	if (args.pt_flags & BIT_ULL(MSHV_PT_BIT_CPU_AND_XSAVE_FEATURES)) {
>  		if (copy_from_user(&args, user_arg, sizeof(args)))
>  			return -EFAULT;

[Severity: Medium]
This is a pre-existing issue, but does this second fetch from userspace
introduce a race condition?

If a concurrent userspace thread clears the MSHV_PT_BIT_CPU_AND_XSAVE_FEATURES
flag between the first and second fetch, args.pt_flags will be overwritten
with the new value. The code does not re-verify the flag before proceeding to
validate and use extended v2 fields like args.pt_num_cpu_fbanks.

This could allow the kernel to operate on uninitialized userspace garbage.

[ ... ]

> @@ -2371,16 +2451,6 @@ static long mshv_ioctl_process_pt_flags(void __user *user_arg, u64 *pt_flags,
>  		if (mshv_field_nonzero(args, pt_rsvd2))
>  			return -EINVAL;
>  #endif
> -	} else {
> -		/*
> -		 * v1 behavior: try to enable everything. The hypervisor will
> -		 * disable features that are not supported. The banks can be
> -		 * queried via the get partition property hypercall.
> -		 */
> -		for (i = 0; i < HV_PARTITION_PROCESSOR_FEATURES_BANKS; i++)
> -			disabled_procs->as_uint64[i] = 0;
> -
> -		disabled_xsave->as_uint64 = 0;
>  	}
>  
>  	/* Only support EXO partitions */

[Severity: High]
Does removing this fallback initialization break legacy v1 callers on ARM64?

With this code removed, disabled_procs->as_uint64[i] is unconditionally set
to U64_MAX earlier in mshv_ioctl_process_pt_flags(), which disables all
processor features. Because the new default initializations are inside an
#if IS_ENABLED(CONFIG_X86_64) block, ARM64 callers never get those defaults
nor the old 0 initialization.

Additionally, disabled_xsave points to the disabled_processor_xsave_features
field of cr_props (which is on the stack in mshv_ioctl_create_partition).
Since it is no longer initialized to 0 here for non-x86 v1 callers, will it
send uninitialized stack memory to the hypervisor?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=7