[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] x86: moke show_page_walk() more robust



I very much think it is worthwhile: In the case given, a supposed guest crash
resulted in a host crash - the expected behavior would have been to do the
page table walk, then forward the fault to the guest.

Jan

>>> Keir Fraser <keir.fraser@xxxxxxxxxxxxx> 17.07.09 21:48 >>>
Is this worthwhile? It's kind of game over by this point, and we already
dumped out the useful stuff. And the ensuing crash makes it very clear that
the page walk discovered an invalid mfn (rather than this alternative of
silently truncating the walk).

 -- Keir

On 17/07/2009 10:20, "Jan Beulich" <JBeulich@xxxxxxxxxx> wrote:

> This should fix the host crash reported in the thread "d273:v3:
> reserved bit in page table (ec=0019)" (but it specifically doesn't deal
> with the yet unknown original cause of the problem).
> 
> Also add in a missing line in x86-64's do_page_walk().
> 
> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxxxx>
> 
> --- 2009-07-10.orig/xen/arch/x86/x86_32/traps.c 2009-07-10 13:58:56.000000000
> +0200
> +++ 2009-07-10/xen/arch/x86/x86_32/traps.c 2009-07-17 11:12:42.000000000 +0200
> @@ -165,7 +165,8 @@ void show_page_walk(unsigned long addr)
>      printk(" L3[0x%03lx] = %"PRIpte" %08lx\n",
>             l3_table_offset(addr), l3e_get_intpte(l3e), pfn);
>      unmap_domain_page(l3t);
> -    if ( !(l3e_get_flags(l3e) & _PAGE_PRESENT) )
> +    if ( !(l3e_get_flags(l3e) & _PAGE_PRESENT) ||
> +         !mfn_valid(mfn) )
>          return;
>  
>      l2t = map_domain_page(mfn);
> @@ -178,7 +179,8 @@ void show_page_walk(unsigned long addr)
>             (l2e_get_flags(l2e) & _PAGE_PSE) ? "(PSE)" : "");
>      unmap_domain_page(l2t);
>      if ( !(l2e_get_flags(l2e) & _PAGE_PRESENT) ||
> -         (l2e_get_flags(l2e) & _PAGE_PSE) )
> +         (l2e_get_flags(l2e) & _PAGE_PSE) ||
> +         !mfn_valid(mfn) )
>          return;
>  
>      l1t = map_domain_page(mfn);
> --- 2009-07-10.orig/xen/arch/x86/x86_64/mm.c 2009-07-10 13:58:56.000000000
> +0200
> +++ 2009-07-10/xen/arch/x86/x86_64/mm.c 2009-07-17 11:10:29.000000000 +0200
> @@ -124,6 +124,7 @@ void *do_page_walk(struct vcpu *v, unsig
>      l3e = l3t[l3_table_offset(addr)];
>      mfn = l3e_get_pfn(l3e);
>      if ( !(l3e_get_flags(l3e) & _PAGE_PRESENT) || !mfn_valid(mfn) )
> +        return NULL;
>      if ( (l3e_get_flags(l3e) & _PAGE_PSE) )
>          return mfn_to_virt(mfn) + (addr & ((1UL << L3_PAGETABLE_SHIFT) - 1));
>  
> --- 2009-07-10.orig/xen/arch/x86/x86_64/traps.c 2009-07-10 13:58:56.000000000
> +0200
> +++ 2009-07-10/xen/arch/x86/x86_64/traps.c 2009-07-17 11:11:59.000000000 +0200
> @@ -180,7 +180,8 @@ void show_page_walk(unsigned long addr)
>      pfn = mfn_valid(mfn) ? get_gpfn_from_mfn(mfn) : INVALID_M2P_ENTRY;
>      printk(" L4[0x%03lx] = %"PRIpte" %016lx\n",
>             l4_table_offset(addr), l4e_get_intpte(l4e), pfn);
> -    if ( !(l4e_get_flags(l4e) & _PAGE_PRESENT) )
> +    if ( !(l4e_get_flags(l4e) & _PAGE_PRESENT) ||
> +         !mfn_valid(mfn) )
>          return;
>  
>      l3t = mfn_to_virt(mfn);
> @@ -192,7 +193,8 @@ void show_page_walk(unsigned long addr)
>             l3_table_offset(addr), l3e_get_intpte(l3e), pfn,
>             (l3e_get_flags(l3e) & _PAGE_PSE) ? " (PSE)" : "");
>      if ( !(l3e_get_flags(l3e) & _PAGE_PRESENT) ||
> -         (l3e_get_flags(l3e) & _PAGE_PSE) )
> +         (l3e_get_flags(l3e) & _PAGE_PSE) ||
> +         !mfn_valid(mfn) )
>          return;
>  
>      l2t = mfn_to_virt(mfn);
> @@ -204,7 +206,8 @@ void show_page_walk(unsigned long addr)
>             l2_table_offset(addr), l2e_get_intpte(l2e), pfn,
>             (l2e_get_flags(l2e) & _PAGE_PSE) ? "(PSE)" : "");
>      if ( !(l2e_get_flags(l2e) & _PAGE_PRESENT) ||
> -         (l2e_get_flags(l2e) & _PAGE_PSE) )
> +         (l2e_get_flags(l2e) & _PAGE_PSE) ||
> +         !mfn_valid(mfn) )
>          return;
>  
>      l1t = mfn_to_virt(mfn);
> 
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx 
> http://lists.xensource.com/xen-devel 



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.