a question about protection_map[]

a question about protection_map[]

[Xishi Qiu]

Hi,

We can use mprotect to set read only or read/write.

mprotect_fixup()
	vma_set_page_prot()
		vm_pgprot_modify()
			vm_get_page_prot()
				protection_map[vm_flags & (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]

The following code shows that prots from __P001(PROT_READ) and __P010(PROT_WRITE)
are the same, so how does it distinguish read only or read/write from mprotect?

pgprot_t protection_map[16] = {
	__P000, __P001, __P010, __P011, __P100, __P101, __P110, __P111,
	__S000, __S001, __S010, __S011, __S100, __S101, __S110, __S111
};

#define __P001	PAGE_READONLY
#define __P010	PAGE_COPY

#define PAGE_READONLY		__pgprot(_PAGE_PRESENT | _PAGE_USER |	\
					 _PAGE_ACCESSED | _PAGE_NX)

#define PAGE_COPY_NOEXEC	__pgprot(_PAGE_PRESENT | _PAGE_USER |	\
					 _PAGE_ACCESSED | _PAGE_NX)
#define PAGE_COPY		PAGE_COPY_NOEXEC


Thanks,
Xishi Qiu

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

a question about protection_map[]

[Xishi Qiu]

Hi,

We can use mprotect to set read only or read/write.

mprotect_fixup()
	vma_set_page_prot()
		vm_pgprot_modify()
			vm_get_page_prot()
				protection_map[vm_flags & (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]

The following code shows that prots from __P001(PROT_READ) and __P010(PROT_WRITE)
are the same, so how does it distinguish read only or read/write from mprotect?

pgprot_t protection_map[16] = {
	__P000, __P001, __P010, __P011, __P100, __P101, __P110, __P111,
	__S000, __S001, __S010, __S011, __S100, __S101, __S110, __S111
};

#define __P001	PAGE_READONLY
#define __P010	PAGE_COPY

#define PAGE_READONLY		__pgprot(_PAGE_PRESENT | _PAGE_USER |	\
					 _PAGE_ACCESSED | _PAGE_NX)

#define PAGE_COPY_NOEXEC	__pgprot(_PAGE_PRESENT | _PAGE_USER |	\
					 _PAGE_ACCESSED | _PAGE_NX)
#define PAGE_COPY		PAGE_COPY_NOEXEC


Thanks,
Xishi Qiu

Re: a question about protection_map[]

[Kirill A. Shutemov]

On Mon, Jul 11, 2016 at 06:12:30PM +0800, Xishi Qiu wrote:
> Hi,
> 
> We can use mprotect to set read only or read/write.
> 
> mprotect_fixup()
> 	vma_set_page_prot()
> 		vm_pgprot_modify()
> 			vm_get_page_prot()
> 				protection_map[vm_flags & (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]
> 
> The following code shows that prots from __P001(PROT_READ) and __P010(PROT_WRITE)
> are the same, so how does it distinguish read only or read/write from mprotect?

It doesn't.

Write protection will be removed by fault handler on next write access to
the page. Somewhat suboptiomal, but zero page implemenation relies on this
to work properly.

-- 
 Kirill A. Shutemov

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: a question about protection_map[]

[Kirill A. Shutemov]

On Mon, Jul 11, 2016 at 06:12:30PM +0800, Xishi Qiu wrote:
> Hi,
> 
> We can use mprotect to set read only or read/write.
> 
> mprotect_fixup()
> 	vma_set_page_prot()
> 		vm_pgprot_modify()
> 			vm_get_page_prot()
> 				protection_map[vm_flags & (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]
> 
> The following code shows that prots from __P001(PROT_READ) and __P010(PROT_WRITE)
> are the same, so how does it distinguish read only or read/write from mprotect?

It doesn't.

Write protection will be removed by fault handler on next write access to
the page. Somewhat suboptiomal, but zero page implemenation relies on this
to work properly.

-- 
 Kirill A. Shutemov

Re: a question about protection_map[]

[Xishi Qiu]

On 2016/7/11 21:30, Kirill A. Shutemov wrote:

> On Mon, Jul 11, 2016 at 06:12:30PM +0800, Xishi Qiu wrote:
>> Hi,
>>
>> We can use mprotect to set read only or read/write.
>>
>> mprotect_fixup()
>> 	vma_set_page_prot()
>> 		vm_pgprot_modify()
>> 			vm_get_page_prot()
>> 				protection_map[vm_flags & (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]
>>
>> The following code shows that prots from __P001(PROT_READ) and __P010(PROT_WRITE)
>> are the same, so how does it distinguish read only or read/write from mprotect?
> 
> It doesn't.
> 
> Write protection will be removed by fault handler on next write access to
> the page. Somewhat suboptiomal, but zero page implemenation relies on this
> to work properly.
> 

Hi Kirill,

I know, PAGE_READONLY and PAGE_COPY are both missed _PAGE_RW,
so it will cause page fault, then we will set new prot flag from
vma, right?

Thanks,
Xishi Qiu

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: a question about protection_map[]

[Xishi Qiu]

On 2016/7/11 21:30, Kirill A. Shutemov wrote:

> On Mon, Jul 11, 2016 at 06:12:30PM +0800, Xishi Qiu wrote:
>> Hi,
>>
>> We can use mprotect to set read only or read/write.
>>
>> mprotect_fixup()
>> 	vma_set_page_prot()
>> 		vm_pgprot_modify()
>> 			vm_get_page_prot()
>> 				protection_map[vm_flags & (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]
>>
>> The following code shows that prots from __P001(PROT_READ) and __P010(PROT_WRITE)
>> are the same, so how does it distinguish read only or read/write from mprotect?
> 
> It doesn't.
> 
> Write protection will be removed by fault handler on next write access to
> the page. Somewhat suboptiomal, but zero page implemenation relies on this
> to work properly.
> 

Hi Kirill,

I know, PAGE_READONLY and PAGE_COPY are both missed _PAGE_RW,
so it will cause page fault, then we will set new prot flag from
vma, right?

Thanks,
Xishi Qiu

Re: a question about protection_map[]

[Kirill A. Shutemov]

On Tue, Jul 12, 2016 at 09:31:30AM +0800, Xishi Qiu wrote:
> On 2016/7/11 21:30, Kirill A. Shutemov wrote:
> 
> > On Mon, Jul 11, 2016 at 06:12:30PM +0800, Xishi Qiu wrote:
> >> Hi,
> >>
> >> We can use mprotect to set read only or read/write.
> >>
> >> mprotect_fixup()
> >> 	vma_set_page_prot()
> >> 		vm_pgprot_modify()
> >> 			vm_get_page_prot()
> >> 				protection_map[vm_flags & (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]
> >>
> >> The following code shows that prots from __P001(PROT_READ) and __P010(PROT_WRITE)
> >> are the same, so how does it distinguish read only or read/write from mprotect?
> > 
> > It doesn't.
> > 
> > Write protection will be removed by fault handler on next write access to
> > the page. Somewhat suboptiomal, but zero page implemenation relies on this
> > to work properly.
> > 
> 
> Hi Kirill,
> 
> I know, PAGE_READONLY and PAGE_COPY are both missed _PAGE_RW,
> so it will cause page fault, then we will set new prot flag from
> vma, right?

Yes. See wp_page_reuse().

-- 
 Kirill A. Shutemov

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: a question about protection_map[]

[Kirill A. Shutemov]

On Tue, Jul 12, 2016 at 09:31:30AM +0800, Xishi Qiu wrote:
> On 2016/7/11 21:30, Kirill A. Shutemov wrote:
> 
> > On Mon, Jul 11, 2016 at 06:12:30PM +0800, Xishi Qiu wrote:
> >> Hi,
> >>
> >> We can use mprotect to set read only or read/write.
> >>
> >> mprotect_fixup()
> >> 	vma_set_page_prot()
> >> 		vm_pgprot_modify()
> >> 			vm_get_page_prot()
> >> 				protection_map[vm_flags & (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)]
> >>
> >> The following code shows that prots from __P001(PROT_READ) and __P010(PROT_WRITE)
> >> are the same, so how does it distinguish read only or read/write from mprotect?
> > 
> > It doesn't.
> > 
> > Write protection will be removed by fault handler on next write access to
> > the page. Somewhat suboptiomal, but zero page implemenation relies on this
> > to work properly.
> > 
> 
> Hi Kirill,
> 
> I know, PAGE_READONLY and PAGE_COPY are both missed _PAGE_RW,
> so it will cause page fault, then we will set new prot flag from
> vma, right?

Yes. See wp_page_reuse().

-- 
 Kirill A. Shutemov