Re: [PATCH] mm: Avoid using set_pte_at when updating a present pte

[Linus Torvalds]

Ahh, and I should learn to read all my emails before replying to some of them..

On Thu, Oct 8, 2020 at 2:26 AM Aneesh Kumar K.V
 wrote:
>
> This avoids the below warning
> [..]
 > WARNING: CPU: 0 PID: 30613 at arch/powerpc/mm/pgtable.c:185
set_pte_at+0x2a8/0x3a0 arch/powerpc/mm/pgtable.c:185

.. and I assume this is what triggered the other patch too.

Yes, with the ppc warning, we need to do _something_ about this, and
at that point I think the "something" is to just avoid the pte
wrpritect trick.

   Linus

[PATCH] mm: Avoid using set_pte_at when updating a present pte

[Aneesh Kumar K.V]

This avoids the below warning

WARNING: CPU: 0 PID: 30613 at arch/powerpc/mm/pgtable.c:185 
set_pte_at+0x2a8/0x3a0 arch/powerpc/mm/pgtable.c:185
Kernel panic - not syncing: panic_on_warn set ...
CPU: 0 PID: 30613 Comm: syz-executor.0 Not tainted 
5.9.0-rc8-syzkaller-00156-gc85fb28b6f99 #0
Call Trace:
 [c01cd1f0] panic+0x29c/0x75c kernel/panic.c:231
 [c01cce24] __warn+0x104/0x1b8 kernel/panic.c:600
 [c0d829e4] report_bug+0x1d4/0x380 lib/bug.c:198
 [c0036800] program_check_exception+0x4e0/0x750 
arch/powerpc/kernel/traps.c:1508
 [c00098a8] program_check_common_virt+0x308/0x360
--- interrupt: 700 at set_pte_at+0x2a8/0x3a0 arch/powerpc/mm/pgtable.c:185
LR = set_pte_at+0x2a4/0x3a0 arch/powerpc/mm/pgtable.c:185
 [c05d2a7c] copy_present_page mm/memory.c:857 [inline]
 [c05d2a7c] copy_present_pte mm/memory.c:899 [inline]
 [c05d2a7c] copy_pte_range mm/memory.c:1014 [inline]
 [c05d2a7c] copy_pmd_range mm/memory.c:1092 [inline]
 [c05d2a7c] copy_pud_range mm/memory.c:1127 [inline]
 [c05d2a7c] copy_p4d_range mm/memory.c:1150 [inline]
 [c05d2a7c] copy_page_range+0x1f6c/0x2cc0 mm/memory.c:1212
 [c01c63cc] dup_mmap kernel/fork.c:592 [inline]
 [c01c63cc] dup_mm+0x77c/0xab0 kernel/fork.c:1355
 [c01c8f70] copy_mm kernel/fork.c:1411 [inline]
 [c01c8f70] copy_process+0x1f00/0x2740 kernel/fork.c:2070
 [c01c9b54] _do_fork+0xc4/0x10b0 kernel/fork.c:2429
 [c01caf54] __do_sys_clone3+0x1d4/0x2b0 kernel/fork.c:27

Architecture like ppc64 expects set_pte_at to be not used for updating a
valid pte. This is further explained in
commit 56eecdb912b5 ("mm: Use ptep/pmdp_set_numa() for updating _PAGE_NUMA bit")

Cc: Peter Xu 
Cc: Jason Gunthorpe 
Cc: John Hubbard 
Cc: linux...@kvack.org
Cc: linux-ker...@vger.kernel.org
Cc: Andrew Morton 
Cc: Jan Kara 
Cc: Michal Hocko 
Cc: Kirill Shutemov 
Cc: Hugh Dickins 
Cc: Linus Torvalds 
Signed-off-by: Aneesh Kumar K.V 
---
 mm/memory.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/memory.c b/mm/memory.c
index fcfc4ca36eba..bfe202ef6244 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -854,7 +854,7 @@ copy_present_page(struct mm_struct *dst_mm, struct 
mm_struct *src_mm,
 * source pte back to being writable.
 */
if (pte_write(pte))
-   set_pte_at(src_mm, addr, src_pte, pte);
+   ptep_set_access_flags(vma, addr, src_pte, pte, 1);
 
new_page = *prealloc;
if (!new_page)
-- 
2.26.2