[PATCH v3 0/1] fix race in gmap_make_secure()

[PATCH v3 0/1] fix race in gmap_make_secure()

[Claudio Imbrenda]

This fixes a potential race and, most importantly, removes the last
user of follow_page() without FOLL_GET.

This removes the last user of follow_page without FOLL_GET. Once this
patch is merged I'll send a separate patch to perform FOLL_GET
automatically with follow_page, as suggested by Jason Gunthorpe.

v1->v2:
* fix and improve commit message [thanks Heiko]

Claudio Imbrenda (1):
  KVM: s390: fix race in gmap_make_secure()

 arch/s390/kernel/uv.c | 32 +++++++++++---------------------
 1 file changed, 11 insertions(+), 21 deletions(-)

-- 
2.40.0

[PATCH v3 1/1] KVM: s390: fix race in gmap_make_secure()

[Claudio Imbrenda]

Fix a potential race in gmap_make_secure() and remove the last user of
follow_page() without FOLL_GET.

The old code is locking something it doesn't have a reference to, and
as explained by Jason and David in this discussion:
https://lore.kernel.org/linux-mm/Y9J4P%2FRNvY1Ztn0Q@nvidia.com/
it can lead to all kind of bad things, including the page getting
unmapped (MADV_DONTNEED), freed, reallocated as a larger folio and the
unlock_page() would target the wrong bit.
There is also another race with the FOLL_WRITE, which could race
between the follow_page() and the get_locked_pte().

The main point is to remove the last use of follow_page() without
FOLL_GET or FOLL_PIN, removing the races can be considered a nice
bonus.

Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Link: https://lore.kernel.org/linux-mm/Y9J4P%2FRNvY1Ztn0Q@nvidia.com/
Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Fixes: 214d9bbcd3a6 ("s390/mm: provide memory management functions for protected KVM guests")
---
 arch/s390/kernel/uv.c | 32 +++++++++++---------------------
 1 file changed, 11 insertions(+), 21 deletions(-)

diff --git a/arch/s390/kernel/uv.c b/arch/s390/kernel/uv.c
index 9f18a4af9c13..cb2ee06df286 100644
--- a/arch/s390/kernel/uv.c
+++ b/arch/s390/kernel/uv.c
@@ -192,21 +192,10 @@ static int expected_page_refs(struct page *page)
 	return res;
 }
 
-static int make_secure_pte(pte_t *ptep, unsigned long addr,
-			   struct page *exp_page, struct uv_cb_header *uvcb)
+static int make_page_secure(struct page *page, struct uv_cb_header *uvcb)
 {
-	pte_t entry = READ_ONCE(*ptep);
-	struct page *page;
 	int expected, cc = 0;
 
-	if (!pte_present(entry))
-		return -ENXIO;
-	if (pte_val(entry) & _PAGE_INVALID)
-		return -ENXIO;
-
-	page = pte_page(entry);
-	if (page != exp_page)
-		return -ENXIO;
 	if (PageWriteback(page))
 		return -EAGAIN;
 	expected = expected_page_refs(page);
@@ -304,17 +293,18 @@ int gmap_make_secure(struct gmap *gmap, unsigned long gaddr, void *uvcb)
 		goto out;
 
 	rc = -ENXIO;
-	page = follow_page(vma, uaddr, FOLL_WRITE);
-	if (IS_ERR_OR_NULL(page))
-		goto out;
-
-	lock_page(page);
 	ptep = get_locked_pte(gmap->mm, uaddr, &ptelock);
-	if (should_export_before_import(uvcb, gmap->mm))
-		uv_convert_from_secure(page_to_phys(page));
-	rc = make_secure_pte(ptep, uaddr, page, uvcb);
+	if (pte_present(*ptep) && !(pte_val(*ptep) & _PAGE_INVALID) && pte_write(*ptep)) {
+		page = pte_page(*ptep);
+		rc = -EAGAIN;
+		if (trylock_page(page)) {
+			if (should_export_before_import(uvcb, gmap->mm))
+				uv_convert_from_secure(page_to_phys(page));
+			rc = make_page_secure(page, uvcb);
+			unlock_page(page);
+		}
+	}
 	pte_unmap_unlock(ptep, ptelock);
-	unlock_page(page);
 out:
 	mmap_read_unlock(gmap->mm);
 
-- 
2.40.0

Re: [PATCH v3 1/1] KVM: s390: fix race in gmap_make_secure()

[Jason Gunthorpe]

On Fri, Apr 28, 2023 at 11:27:53AM +0200, Claudio Imbrenda wrote:
> Fix a potential race in gmap_make_secure() and remove the last user of
> follow_page() without FOLL_GET.
> 
> The old code is locking something it doesn't have a reference to, and
> as explained by Jason and David in this discussion:
> https://lore.kernel.org/linux-mm/Y9J4P%2FRNvY1Ztn0Q@nvidia.com/
> it can lead to all kind of bad things, including the page getting
> unmapped (MADV_DONTNEED), freed, reallocated as a larger folio and the
> unlock_page() would target the wrong bit.
> There is also another race with the FOLL_WRITE, which could race
> between the follow_page() and the get_locked_pte().
> 
> The main point is to remove the last use of follow_page() without
> FOLL_GET or FOLL_PIN, removing the races can be considered a nice
> bonus.
> 
> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> Link: https://lore.kernel.org/linux-mm/Y9J4P%2FRNvY1Ztn0Q@nvidia.com/
> Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
> Fixes: 214d9bbcd3a6 ("s390/mm: provide memory management functions for protected KVM guests")
> ---
>  arch/s390/kernel/uv.c | 32 +++++++++++---------------------
>  1 file changed, 11 insertions(+), 21 deletions(-)

Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>

Jason