On 18/10/2023 9:02 am, Tamas K Lengyel wrote:
> When mapping in the shared_info page to a fork the m2p entry wasn't set
> resulting in the shared_info being reset even when the fork reset was called
> with only reset_state and not reset_memory. This results in an extra
> unnecessary TLB flush.
>
> Fixes: 1a0000ac775 ("mem_sharing: map shared_info page to same gfn during
> fork")
> Signed-off-by: Tamas K Lengyel <ta...@tklengyel.com>
> ---
> xen/arch/x86/mm/mem_sharing.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/xen/arch/x86/mm/mem_sharing.c b/xen/arch/x86/mm/mem_sharing.c
> index 94b6b782ef..142258f16a 100644
> --- a/xen/arch/x86/mm/mem_sharing.c
> +++ b/xen/arch/x86/mm/mem_sharing.c
> @@ -1847,6 +1847,8 @@ static int copy_special_pages(struct domain *cd, struct
> domain *d)
> p2m_ram_rw, p2m->default_access, -1);
> if ( rc )
> return rc;
> +
> + set_gpfn_from_mfn(mfn_x(new_mfn), gfn_x(old_gfn));
> }
> }
>
Acked-by: Andrew Cooper <andrew.coop...@citrix.com>
CC Henry. This needs a view about a release ack.
Cons: it's been broken since Xen 4.14 and we're very deep into the 4.18
code freeze.
Pros: it's a bug and would clearly qualify for backport, and is in a
niche feature so isn't plausibly going to adversely affect other users.
~Andrew
