Re: [PATCH v2] xen/pv: Call get_cpu_address_sizes to set x86_virt/phys_bits

[Boris Ostrovsky]

x86 maintainers, this needs your ack please.

-boris


On 07/24/2018 08:45 AM, M. Vefa Bicakci wrote:
> Commit d94a155c59c9 ("x86/cpu: Prevent cpuinfo_x86::x86_phys_bits
> adjustment corruption") has moved the query and calculation of the
> x86_virt_bits and x86_phys_bits fields of the cpuinfo_x86 struct
> from the get_cpu_cap function to a new function named
> get_cpu_address_sizes.
>
> One of the call sites related to Xen PV VMs was unfortunately missed
> in the aforementioned commit. This prevents successful boot-up of
> kernel versions 4.17 and up in Xen PV VMs if CONFIG_DEBUG_VIRTUAL
> is enabled, due to the following code path:
>
>   enlighten_pv.c::xen_start_kernel
>     mmu_pv.c::xen_reserve_special_pages
>       page.h::__pa
>         physaddr.c::__phys_addr
>           physaddr.h::phys_addr_valid
>
> phys_addr_valid uses boot_cpu_data.x86_phys_bits to validate physical
> addresses. boot_cpu_data.x86_phys_bits is no longer populated before
> the call to xen_reserve_special_pages due to the aforementioned commit
> though, so the validation performed by phys_addr_valid fails, which
> causes __phys_addr to trigger a BUG, preventing boot-up.
>
> Signed-off-by: M. Vefa Bicakci <m.v.b@xxxxxxxxxx>
> Cc: "Kirill A. Shutemov" <kirill.shutemov@xxxxxxxxxxxxxxx>
> Cc: Andy Lutomirski <luto@xxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: "H. Peter Anvin" <hpa@xxxxxxxxx>
> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Cc: Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>
> Cc: Juergen Gross <jgross@xxxxxxxx>
> Cc: xen-devel@xxxxxxxxxxxxxxxxxxxx
> Cc: x86@xxxxxxxxxx
> Cc: stable@xxxxxxxxxxxxxxx # for v4.17 and up
> Fixes: d94a155c59c9 ("x86/cpu: Prevent cpuinfo_x86::x86_phys_bits adjustment corruption")
>
> ---
>
> Changes since v1:
> - Move the call to get_cpu_address_sizes below the call to
>   x86_configure_nx.
> - Amend the commit message to describe why PV VMs do not boot up
>   successfully when CONFIG_DEBUG_VIRTUAL is enabled.
> ---
>  arch/x86/kernel/cpu/common.c | 2 +-
>  arch/x86/kernel/cpu/cpu.h    | 1 +
>  arch/x86/xen/enlighten_pv.c  | 3 +++
>  3 files changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
> index f73fa6f6d85e..dd282482de09 100644
> --- a/arch/x86/kernel/cpu/common.c
> +++ b/arch/x86/kernel/cpu/common.c
> @@ -911,7 +911,7 @@ void get_cpu_cap(struct cpuinfo_x86 *c)
>  	apply_forced_caps(c);
>  }
>  
> -static void get_cpu_address_sizes(struct cpuinfo_x86 *c)
> +void get_cpu_address_sizes(struct cpuinfo_x86 *c)
>  {
>  	u32 eax, ebx, ecx, edx;
>  
> diff --git a/arch/x86/kernel/cpu/cpu.h b/arch/x86/kernel/cpu/cpu.h
> index 38216f678fc3..12a5f0cec0b2 100644
> --- a/arch/x86/kernel/cpu/cpu.h
> +++ b/arch/x86/kernel/cpu/cpu.h
> @@ -46,6 +46,7 @@ extern const struct cpu_dev *const __x86_cpu_dev_start[],
>  			    *const __x86_cpu_dev_end[];
>  
>  extern void get_cpu_cap(struct cpuinfo_x86 *c);
> +extern void get_cpu_address_sizes(struct cpuinfo_x86 *c);
>  extern void cpu_detect_cache_sizes(struct cpuinfo_x86 *c);
>  extern void init_scattered_cpuid_features(struct cpuinfo_x86 *c);
>  extern u32 get_scattered_cpuid_leaf(unsigned int level,
> diff --git a/arch/x86/xen/enlighten_pv.c b/arch/x86/xen/enlighten_pv.c
> index 105a57d73701..ee3b00c7acda 100644
> --- a/arch/x86/xen/enlighten_pv.c
> +++ b/arch/x86/xen/enlighten_pv.c
> @@ -1256,6 +1256,9 @@ asmlinkage __visible void __init xen_start_kernel(void)
>  	get_cpu_cap(&boot_cpu_data);
>  	x86_configure_nx();
>  
> +	/* Determine virtual and physical address sizes */
> +	get_cpu_address_sizes(&boot_cpu_data);
> +
>  	/* Let's presume PV guests always boot on vCPU with id 0. */
>  	per_cpu(xen_vcpu_id, 0) = 0;
>