diff mbox

[v2,03/13] x86/cpa: Warn if kernel_unmap_pages_in_pgd is used inappropriately

Message ID 3336bb02792d56c49d18e2f7a435fd194a4a22e4.1466192946.git.luto@kernel.org (mailing list archive)
State New, archived
Headers show

Commit Message

Andy Lutomirski June 17, 2016, 8 p.m. UTC
It's currently only used in the EFI code, which is safe AFAICT.
Warn if anyone tries to use it on the normal kernel pgd.

Signed-off-by: Andy Lutomirski <luto@kernel.org>
---
 arch/x86/mm/pageattr.c | 3 +++
 1 file changed, 3 insertions(+)

Comments

Borislav Petkov June 17, 2016, 8:30 p.m. UTC | #1
On Fri, Jun 17, 2016 at 01:00:39PM -0700, Andy Lutomirski wrote:
> It's currently only used in the EFI code, which is safe AFAICT.

"It is basically useful for a pagetable hierarchy which is not init_mm."

> Warn if anyone tries to use it on the normal kernel pgd.
> 
> Signed-off-by: Andy Lutomirski <luto@kernel.org>
> ---
>  arch/x86/mm/pageattr.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
> index 6a8026918bf6..e9b9c5cedbb8 100644
> --- a/arch/x86/mm/pageattr.c
> +++ b/arch/x86/mm/pageattr.c
> @@ -1996,6 +1996,9 @@ out:
>  void kernel_unmap_pages_in_pgd(pgd_t *root, unsigned long address,
>  			       unsigned numpages)
>  {
> +	/* Unmapping kernel entries from init_mm's pgd is not allowed. */
> +	WARN_ON(root == init_mm.pgd);

We can also return and not do the unmapping:

	if (WARN_ON(root == init_mm.pgd))
		return;
Andy Lutomirski June 18, 2016, 10:29 a.m. UTC | #2
On Fri, Jun 17, 2016 at 1:30 PM, Borislav Petkov <bp@alien8.de> wrote:
> On Fri, Jun 17, 2016 at 01:00:39PM -0700, Andy Lutomirski wrote:
>> It's currently only used in the EFI code, which is safe AFAICT.
>
> "It is basically useful for a pagetable hierarchy which is not init_mm."
>
>> Warn if anyone tries to use it on the normal kernel pgd.
>>
>> Signed-off-by: Andy Lutomirski <luto@kernel.org>
>> ---
>>  arch/x86/mm/pageattr.c | 3 +++
>>  1 file changed, 3 insertions(+)
>>
>> diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
>> index 6a8026918bf6..e9b9c5cedbb8 100644
>> --- a/arch/x86/mm/pageattr.c
>> +++ b/arch/x86/mm/pageattr.c
>> @@ -1996,6 +1996,9 @@ out:
>>  void kernel_unmap_pages_in_pgd(pgd_t *root, unsigned long address,
>>                              unsigned numpages)
>>  {
>> +     /* Unmapping kernel entries from init_mm's pgd is not allowed. */
>> +     WARN_ON(root == init_mm.pgd);
>
> We can also return and not do the unmapping:
>
>         if (WARN_ON(root == init_mm.pgd))
>                 return;

I'll do one better: the only function that calls this function is
unused.  I'll just delete it.

--Andy
Borislav Petkov June 18, 2016, 10:37 a.m. UTC | #3
On Sat, Jun 18, 2016 at 03:29:01AM -0700, Andy Lutomirski wrote:
> I'll do one better: the only function that calls this function is
> unused.  I'll just delete it.

Fair enough - Matt is on CC.

Btw, normally I'm almost never talking to you at that time of the day,
what's up? Can't sleep?

:-))
diff mbox

Patch

diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index 6a8026918bf6..e9b9c5cedbb8 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -1996,6 +1996,9 @@  out:
 void kernel_unmap_pages_in_pgd(pgd_t *root, unsigned long address,
 			       unsigned numpages)
 {
+	/* Unmapping kernel entries from init_mm's pgd is not allowed. */
+	WARN_ON(root == init_mm.pgd);
+
 	unmap_pgd_range(root, address, address + (numpages << PAGE_SHIFT));
 }