Re: [PATCH mm-hotfixes v3 2/4] x86/mm/pat: acquire mmap lock on page table free to avoid ptdump UAF

[Will Deacon]

Hi Lorenzo,

I'm certainly no x86 expert (quite the opposite!), but I was looking at
this with Adrian and got myself confused. See below.

On Tue, Jul 14, 2026 at 06:24:24PM +0100, Lorenzo Stoakes wrote:
> x86 implements page attribute modification using its Change Page
> Attributes (CPA) mechanism.
> 
> This tracks properties of ranges such as cache mode through x86 page
> attributes, and as part of that logic manipulates kernel page tables.
> 
> Since commit 41d88484c71c ("x86/mm/pat: restore large ROX pages after
> fragmentation") ranges of kernel page table entries can be collapsed into
> huge page table entries as part of this logic.
> 
> As part of this collapse, it frees the page tables which the collapsed
> entries previously pointed to, and it does so without any relevant locks
> being held to preclude concurrent kernel page table walkers.
> 
> The only way this code can be reached is if CPA_COLLAPSE is specified, and
> this is only set in set_memory_rox() via:
> 
> set_memory_rox()
> -> change_page_attr_set_clr()
> -> cpa_flush()
> -> cpa_collapse_large_pages()
> 
> Notable users of this are execmem and bpf when manipulating executable
> mappings.
> 
> However, this is problematic for ptdump as it walks ranges it does not own
> and thus runs the risk of a use-after-free on page tables freed underneath
> it.
> 
> Resolve the issue by acquiring the mmap read lock on init_mm which prevents
> a concurrent ptdump as it acquires the write lock.
> 
> It is safe to acquire a sleeping lock as all the callers invoke
> set_memory_rox() from process context and in any case,
> change_page_attr_set_clr() calls vm_unmap_alias() which ultimately takes a
> mutex, disallowing atomic context here.
> 
> Fixes: 41d88484c71c ("x86/mm/pat: restore large ROX pages after fragmentation")
> Cc: stable@xxxxxxxxxxxxxxx
> Reviewed-by: Mike Rapoport (Microsoft) <rppt@xxxxxxxxxx>
> Reviewed-by: Kiryl Shutsemau (Meta) <kas@xxxxxxxxxx>
> Signed-off-by: Lorenzo Stoakes <ljs@xxxxxxxxxx>
> ---
>  arch/x86/mm/pat/set_memory.c | 14 +++++++++++---
>  1 file changed, 11 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/mm/pat/set_memory.c b/arch/x86/mm/pat/set_memory.c
> index d023a40a1e03..4c4b8244502f 100644
> --- a/arch/x86/mm/pat/set_memory.c
> +++ b/arch/x86/mm/pat/set_memory.c
> @@ -22,6 +22,7 @@
>  #include <linux/cc_platform.h>
>  #include <linux/set_memory.h>
>  #include <linux/memregion.h>
> +#include <linux/cleanup.h>
>  
>  #include <asm/e820/api.h>
>  #include <asm/processor.h>
> @@ -436,9 +437,16 @@ static void cpa_collapse_large_pages(struct cpa_data *cpa)
>  
>  	flush_tlb_all();
>  
> -	list_for_each_entry_safe(ptdesc, tmp, &pgtables, pt_list) {
> -		list_del(&ptdesc->pt_list);
> -		pagetable_free(ptdesc);
> +	/*
> +	 * ptdump might read these page tables, so avoid a use-after-free by
> +	 * acquiring the mmap read lock on init_mm (ptdump acquires the mmap
> +	 * write lock).
> +	 */
> +	scoped_guard(mmap_read_lock, &init_mm) {
> +		list_for_each_entry_safe(ptdesc, tmp, &pgtables, pt_list) {
> +			list_del(&ptdesc->pt_list);
> +			pagetable_free(ptdesc);
> +		}

As I understand it, the argument for taking the read lock is that we're
operating on a region that we "wholly own" and therefore we can happily
run concurrently with CPUs walking distinct parts of the page-table.
However, from what I can tell, the CPA collapse logic will operate on
regions outside of the address range being manipulated by its caller
because it rounds up to the PMD size.

As a made-up example, imagine I have a 2MiB aligned region where the
first 1MiB is read-only and the second 1MiB is in the default r/w state.
If one CPU calls set_memory_rw() on the first 1MiB while another CPU is
walking the second 1MiB (via some other API that doesn't take cpa_lock),
it looks to me like the first CPU can collapse the page-table and free
the unused pages under the feet of the other CPU. What prevents that
from happening?

If all concurrent walkers have interrupts disabled, I guess the TLB
invalidation logic would do it, but it would be good to call this out in
the commit message because it's not clear to me why the read_lock is
sufficient for the collapsing case.

Cheers,

Will