On Tue, Jan 09, 2024, Muhammad Usama Anjum wrote:
> Move mmu notification mechanism inside mm lock to prevent race condition
> in other components which depend on it. The notifier will invalidate
> memory range. Depending upon the number of iterations, different memory
> ranges would be invalidated.
>
> The following warning would be removed by this patch:
> WARNING: CPU: 0 PID: 5067 at arch/x86/kvm/../../../virt/kvm/kvm_main.c:734 kvm_mmu_notifier_change_pte+0x860/0x960 arch/x86/kvm/../../../virt/kvm/kvm_main.c:734
>
> There is no behavioural and performance change with this patch when
> there is no component registered with the mmu notifier.
>
> Fixes: 52526ca7fdb9 ("fs/proc/task_mmu: implement IOCTL to get and optionally clear info about PTEs")
> Reported-by: syzbot+81227d2bd69e9dedb802@xxxxxxxxxxxxxxxxxxxxxxxxx
> Link: https://lore.kernel.org/all/000000000000f6d051060c6785bc@xxxxxxxxxx/
> Cc: Sean Christopherson <seanjc@xxxxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Muhammad Usama Anjum <usama.anjum@xxxxxxxxxxxxx>
> ---
> fs/proc/task_mmu.c | 22 ++++++++++++----------
> 1 file changed, 12 insertions(+), 10 deletions(-)
>
> diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
> index 62b16f42d5d2..56c2e7357494 100644
> --- a/fs/proc/task_mmu.c
> +++ b/fs/proc/task_mmu.c
> @@ -2448,13 +2448,6 @@ static long do_pagemap_scan(struct mm_struct *mm, unsigned long uarg)
> if (ret)
> return ret;
>
> - /* Protection change for the range is going to happen. */
> - if (p.arg.flags & PM_SCAN_WP_MATCHING) {
> - mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_VMA, 0,
> - mm, p.arg.start, p.arg.end);
> - mmu_notifier_invalidate_range_start(&range);
> - }
> -
> for (walk_start = p.arg.start; walk_start < p.arg.end;
> walk_start = p.arg.walk_end) {
> long n_out;
Nit, might be worth moving
struct mmu_notifier_range range;
inside the loop to guard against stale usage, but that's definitely optional.
Reviewed-by: Sean Christopherson <seanjc@xxxxxxxxxx>
