linux-kernel - Re: [PATCH]mm/vmalloc: interchage the implementation of vmalloc_to

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20131201182349.GA1847@hp530>
Date:	Sun, 1 Dec 2013 19:23:55 +0100
From:	Vladimir Murzin <murzin.v@...il.com>
To:	Jianyu Zhan <nasa4836@...il.com>
Cc:	akpm@...ux-foundation.org, iamjoonsoo.kim@....com,
	zhangyanfei@...fujitsu.com, rientjes@...gle.com,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH]mm/vmalloc: interchage the implementation of
 vmalloc_to_{pfn,page}

On Fri, Nov 29, 2013 at 02:11:14AM +0800, Jianyu Zhan wrote:
> 
> Hi, Vladimir,
> 
> On Fri, Nov 29, 2013 at 1:41 AM, Vladimir Murzin <murzin.v@...il.com> wrote:
> >
> > Any numbers for efficiency?
> >
> 
> For the original implementation, vmalloc_to_pfn() wraps the vmalloc_to_page(),
> which means
> 
>      pfn   ------>         struct page      ------>    pfn
>       |                                                          |
>   vmalloc_to_page()                             vmalloc_to_pfn()
> 
> So this patch interchange the implementation, do the dirty page table
> walking code in vmalloc_to_pfn(), and then vmalloc_to_page() uses it, the graph
> now becomes
> 
>      pfn            ------>         struct page        
>        |                                     |
>   vmalloc_to_pfn()         vmalloc_to_page()
> 
> 
> >>  /*
> >> - * Walk a vmap address to the struct page it maps.
> >> + * Walk a vmap address to the physical pfn it maps to.
> >>   */
> >> -struct page *vmalloc_to_page(const void *vmalloc_addr)
> >> +unsigned long vmalloc_to_pfn(const void *vmalloc_addr)
> >>  {
> >>       unsigned long addr = (unsigned long) vmalloc_addr;
> >> -     struct page *page = NULL;
> >> +     unsigned long pfn;
> >
> > uninitialized pfn will lead to a bug.
> >
> 
> Why? The coding pratice has mandates we use it after we initialize it,
> And if we initialize it , to what value will it promise no bug?

Unless you initialize it conditionally. I bet gcc warned you about this ;)

> It is unlikely a rubbish initial value will creep in.
> 
> 
> >>       /*
> >> @@ -244,23 +244,23 @@ struct page *vmalloc_to_page(const void *vmalloc_addr)
> >>                               ptep = pte_offset_map(pmd, addr);
> >>                               pte = *ptep;
> >>                               if (pte_present(pte))
> >> -                                     page = pte_page(pte);
> >> +                                     pfn = pte_page(pte);
> >
> > page_to_pfn is missed here.
> >
> > Have you ever tested there is no functional changes?
> 
> Oh, gods. My fault. It did has no functional changes.
> 
> I just sent the incorrect patch...
> 
> it should be  
>  -   page = pte_page(pte);
>  +  pfn = pte_pfn(pte);;
> 
> Here is the resent patch:
> 

I think it is incorrect too. Originally, vmalloc_to_page might return NULL
under some conditions. With your implementation it will return pfn_to_page(0)
which is not the same as NULL.

Vladimir

> 
> ---
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index 0fdf968..e4f0db2 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -220,12 +220,12 @@ int is_vmalloc_or_module_addr(const void *x)
>  }
>  
>  /*
> - * Walk a vmap address to the struct page it maps.
> + * Walk a vmap address to the physical pfn it maps to.
>   */
> -struct page *vmalloc_to_page(const void *vmalloc_addr)
> +unsigned long vmalloc_to_pfn(const void *vmalloc_addr)
>  {
>  	unsigned long addr = (unsigned long) vmalloc_addr;
> -	struct page *page = NULL;
> +	unsigned long pfn = 0;
>  	pgd_t *pgd = pgd_offset_k(addr);
>  
>  	/*
> @@ -244,23 +244,23 @@ struct page *vmalloc_to_page(const void *vmalloc_addr)
>  				ptep = pte_offset_map(pmd, addr);
>  				pte = *ptep;
>  				if (pte_present(pte))
> -					page = pte_page(pte);
> +					pfn = pte_pfn(pte);
>  				pte_unmap(ptep);
>  			}
>  		}
>  	}
> -	return page;
> +	return pfn;
>  }
> -EXPORT_SYMBOL(vmalloc_to_page);
> +EXPORT_SYMBOL(vmalloc_to_pfn);
>  
>  /*
> - * Map a vmalloc()-space virtual address to the physical page frame number.
> + * Map a vmalloc()-space virtual address to the struct page.
>   */
> -unsigned long vmalloc_to_pfn(const void *vmalloc_addr)
> +struct page *vmalloc_to_page(const void *vmalloc_addr)
>  {
> -	return page_to_pfn(vmalloc_to_page(vmalloc_addr));
> +	return pfn_to_page(vmalloc_to_pfn(vmalloc_addr));
>  }
> -EXPORT_SYMBOL(vmalloc_to_pfn);
> +EXPORT_SYMBOL(vmalloc_to_page);
>  
>  
>  /*** Global kva allocator ***/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/