Re: hv_hypercall_pg page permissios

From: Peter Zijlstra <peterz@infradead.org>
To: Christoph Hellwig <hch@lst.de>
Cc: Dexuan Cui <decui@microsoft.com>, vkuznets <vkuznets@redhat.com>,
	Stephen Hemminger <stephen@networkplumber.org>,
	Andy Lutomirski <luto@kernel.org>,
	Andy Lutomirski <luto@amacapital.net>,
	Michael Kelley <mikelley@microsoft.com>,
	Ju-Hyoung Lee <juhlee@microsoft.com>,
	"x86@kernel.org" <x86@kernel.org>,
	"linux-hyperv@vger.kernel.org" <linux-hyperv@vger.kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	KY Srinivasan <kys@microsoft.com>,
	Tom Lendacky <thomas.lendacky@amd.com>
Subject: Re: hv_hypercall_pg page permissios
Date: Tue, 16 Jun 2020 13:24:15 +0200	[thread overview]
Message-ID: <20200616112415.GT2531@hirez.programming.kicks-ass.net> (raw)
In-Reply-To: <20200616105200.GA32175@lst.de>

On Tue, Jun 16, 2020 at 12:52:00PM +0200, Christoph Hellwig wrote:

> I think something like this should solve the issue:
> 
> --
> diff --git a/arch/x86/include/asm/module.h b/arch/x86/include/asm/module.h
> index e988bac0a4a1c3..716e4de44a8e78 100644
> --- a/arch/x86/include/asm/module.h
> +++ b/arch/x86/include/asm/module.h
> @@ -13,4 +13,6 @@ struct mod_arch_specific {
>  #endif
>  };
>  
> +void *module_alloc_prot(unsigned long size, pgprot_t prot);
> +
>  #endif /* _ASM_X86_MODULE_H */
> diff --git a/arch/x86/kernel/module.c b/arch/x86/kernel/module.c
> index 34b153cbd4acb4..4db6e655120960 100644
> --- a/arch/x86/kernel/module.c
> +++ b/arch/x86/kernel/module.c
> @@ -65,8 +65,10 @@ static unsigned long int get_module_load_offset(void)
>  }
>  #endif
>  
> -void *module_alloc(unsigned long size)
> +void *module_alloc_prot(unsigned long size, pgprot_t prot)
>  {
> +	unsigned int flags = (pgprot_val(prot) & _PAGE_NX) ?
> +			0 : VM_FLUSH_RESET_PERMS;
>  	void *p;
>  
>  	if (PAGE_ALIGN(size) > MODULES_LEN)
> @@ -75,7 +77,7 @@ void *module_alloc(unsigned long size)
>  	p = __vmalloc_node_range(size, MODULE_ALIGN,
>  				    MODULES_VADDR + get_module_load_offset(),
>  				    MODULES_END, GFP_KERNEL,
> -				    PAGE_KERNEL, 0, NUMA_NO_NODE,
> +				    prot, flags, NUMA_NO_NODE,
>  				    __builtin_return_address(0));
>  	if (p && (kasan_module_alloc(p, size) < 0)) {
>  		vfree(p);

Hurmm.. Yes it would. It just doesn't feel right though. Can't we
unconditionally set the flag? At worst it makes free a little bit more
expensive.

The thing is, I don't think _NX is the only prot that needs restoring.
Any prot other than the default (RW IIRC) needs restoring.