Re: [PATCH] x86: NX protection for kernel data

[Siarhei Liakh <sliakh.lkml@gmail.com>]

On Sun, Jul 19, 2009 at 4:13 PM, Sam Ravnborg<sam@ravnborg.org> wrote:
> On Sun, Jul 19, 2009 at 03:43:06PM -0400, Siarhei Liakh wrote:
>> This patch expands functionality of CONFIG_DEBUG_RODATA to set main
>> (static) kernel data area as NX.
>> The following steps are taken to achieve this:
>> 1. Linker scripts are adjusted so .text always starts and end on a page boundary
>> 2. Linker scripts are adjusted so .rodata and .data always starts and
>> end on a page boundary
>> 3. void mark_nxdata_nx(void) added to arch/x86/mm/init_64.c and
>> arch/x86/mm/init_32.c with actual functionality: NX is set for all
>> pages from _etext through _edata
>> 4. mark_nxdata_nx() called from init_post(void) in init/main.c
>>
>> The patch have been developed for Linux 2.6.30 x86 by Siarhei Liakh
>> <sliakh.lkml@gmail.com> and Xuxian Jiang <jiang@cs.ncsu.edu>.
>
>
> The patch no longer applies.
> The file vmlinux_32.lds.S and vmlinux_64.lds.S has been unified
> into one file.

That is actually a great news. I will get the latest source and
re-write the patch.

>> --- a/arch/x86/kernel/vmlinux_32.lds.S
>> +++ b/arch/x86/kernel/vmlinux_32.lds.S
>> @@ -47,6 +47,7 @@ SECTIONS
>>       IRQENTRY_TEXT
>>       *(.fixup)
>>       *(.gnu.warning)
>> +     . = ALIGN(PAGE_SIZE);   /* .text should occupy whole number of pages */
>>       _etext = .;                     /* End of text section */
>
> So _etext cover until page boundary - makes sense.
>
>>    } :text = 0x9090
>>
>> @@ -93,6 +94,7 @@ SECTIONS
>>       *(.data.read_mostly)
>>       _edata = .;             /* End of data section */
>>    }
>> +  . = ALIGN(PAGE_SIZE);              /* needed so we can set NX for .data */
>
> But here _edata does not cover until page boundary.
> And alignmnet is located outside the output section
> definition.
> It would be better/more consistent to follow the style you use for .text here.

You are correct. _edata should be the last thing in .data, and
alignment should be done before it. However, this brings up a
question: was there any specific reason to leave .data.init_task
beyond the _edata? Should we move _edata into the the last of the
.data.* sections to have poper view of kernel layout?