Messages in this thread | | | Date | Thu, 03 Feb 2011 14:11:01 -0800 | From | "H. Peter Anvin" <> | Subject | Re: [PATCH] NX protection for kernel data : fix 32 bits S3 suspend |
| |
On 02/01/2011 10:26 PM, Ingo Molnar wrote: > > * castet.matthieu@free.fr <castet.matthieu@free.fr> wrote: > >> Quoting Ingo Molnar <mingo@elte.hu>: >> >>> >>> * matthieu castet <castet.matthieu@free.fr> wrote: >>> >>>> static inline int is_kernel_text(unsigned long addr) >>>> { >>>> +#if defined(CONFIG_X86_32) && defined(CONFIG_ACPI_SLEEP) >>>> + /* >>>> + * We need to make the wakeup trampoline in first 1MB !NX >>>> + */ >>>> + if (addr >= PAGE_OFFSET && addr <= (PAGE_OFFSET + (1<<20))) >>>> + return 1; >>>> +#endif >>> >>> That's pretty ugly. Why not use set_memory_x()/set_memory_nx(), and only for >>> the >>> trampoline itself? Does the whole 1MB need to be marked X? >> >> The previous code was doing that. > > So why not call set_memory_x() in your patch? Mind trying that? >
OK, there seems to be considerable duplication between static_protections() and local invocation. Consider PCI BIOS, which is another X-needed region.
In static_protections() we have:
> /* > * The BIOS area between 640k and 1Mb needs to be executable for > * PCI BIOS based config access (CONFIG_PCI_GOBIOS) support. > */ > #ifdef CONFIG_PCI_BIOS > if (pcibios_enabled && within(pfn, BIOS_BEGIN >> PAGE_SHIFT, BIOS_END >> PAGE_SHIFT)) > pgprot_val(forbidden) |= _PAGE_NX; > #endif
... however, in arch/x86/pci/pcbios.c:
> static inline void set_bios_x(void) > { > pcibios_enabled = 1; > set_memory_x(PAGE_OFFSET + BIOS_BEGIN, (BIOS_END - BIOS_BEGIN) >> PAGE_SHIFT); > if (__supported_pte_mask & _PAGE_NX) > printk(KERN_INFO "PCI : PCI BIOS aera is rw and x. Use pci=nobios if you want it NX.\n"); > }
This is blatant and insanely ugly code duplication! In particular, static_protections() is "action at a distance" which has no business existing at all.
What I want to know is if static_protections() can somehow override set_bios_x() in this context (in which case it's a serious design error), or if it is plain redundant -- in the latter case we should simply use the same technique elsewhere.
-hpa
| |