On Thu, Oct 30, 2025 at 04:16:20PM +0000, Pedro Falcato wrote:

On Wed, Oct 29, 2025 at 04:50:31PM +0000, Lorenzo Stoakes wrote:

quoted

Currently, if a user needs to determine if guard regions are present in a
range, they have to scan all VMAs (or have knowledge of which ones might
have guard regions).

Since commit 8e2f2aeb8b48 ("fs/proc/task_mmu: add guard region bit to
pagemap") and the related commit a516403787e0 ("fs/proc: extend the
PAGEMAP_SCAN ioctl to report guard regions"), users can use either
/proc/$pid/pagemap or the PAGEMAP_SCAN functionality to perform this
operation at a virtual address level.

This is not ideal, and it gives no visibility at a /proc/$pid/smaps level
that guard regions exist in ranges.

This patch remedies the situation by establishing a new VMA flag,
VM_MAYBE_GUARD, to indicate that a VMA may contain guard regions (it is
uncertain because we cannot reasonably determine whether a
MADV_GUARD_REMOVE call has removed all of the guard regions in a VMA, and
additionally VMAs may change across merge/split).

We utilise 0x800 for this flag which makes it available to 32-bit
architectures also, a flag that was previously used by VM_DENYWRITE, which
was removed in commit 8d0920bde5eb ("mm: remove VM_DENYWRITE") and hasn't
bee reused yet.

The MADV_GUARD_INSTALL madvise() operation now must take an mmap write
lock (and also VMA write lock) whereas previously it did not, but this
seems a reasonable overhead.

Do you though? Could it be possible to simply atomically set the flag with
the read lock held? This would make it so we can't split the VMA (and tightly

VMA flags are not accessed atomically so no I don't think we can do that in any
workable way.

I also don't think it's at all necessary, see below.

define what "may have a guard page"), but it sounds much better than introducing
lock contention. I don't think it is reasonable to add a write lock to a feature
that may be used by such things as thread stack allocation, malloc, etc.

What lock contention? It's per-VMA so the contention is limited to the VMA in
question, and only over the span of time you are setting the gaurd region.

When allocating thread stacks you'll be mapping things into memory which... take
the write lock. malloc() if it goes to the kernel will also take the write lock.

So I think you're overly worried about an operation that a. isn't going to be
something that happens all that often, b. when it does, it's at a time when
you'd be taking write locks anyway and c. won't contend important stuff like
page faults for any VMA other than the one having the the guard region
installed.

--
Pedro

Thanks, Lorenzo