[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20231018011443.3bd8b366@yea>
Date: Wed, 18 Oct 2023 01:14:43 +0200
From: Erhard Furtner <erhard_f@...lbox.org>
To: "Aneesh Kumar K.V" <aneesh.kumar@...ux.ibm.com>
Cc: "Matthew Wilcox (Oracle)" <willy@...radead.org>,
Juergen Gross <jgross@...e.com>, linuxppc-dev@...ts.ozlabs.org,
linux-kernel@...r.kernel.org, linux-sparc@...r.kernel.org,
David Woodhouse <dwmw2@...radead.org>
Subject: Re: [PATCH 0/2] Allow nesting of lazy MMU mode
On Tue, 17 Oct 2023 11:34:23 +0530
"Aneesh Kumar K.V" <aneesh.kumar@...ux.ibm.com> wrote:
> ie, we can do something like below. The change also make sure we call
> set_pte_filter on all the ptes we are setting via set_ptes(). I haven't
> sent this as a proper patch because we still are not able to fix the
> issue Erhard reported.
>
> diff --git a/arch/powerpc/mm/pgtable.c b/arch/powerpc/mm/pgtable.c
> index 3ba9fe411604..95ab20cca2da 100644
> --- a/arch/powerpc/mm/pgtable.c
> +++ b/arch/powerpc/mm/pgtable.c
> @@ -191,28 +191,35 @@ void set_ptes(struct mm_struct *mm, unsigned long addr, pte_t *ptep,
> pte_t pte, unsigned int nr)
> {
> /*
> - * Make sure hardware valid bit is not set. We don't do
> - * tlb flush for this update.
> + * We don't need to call arch_enter/leave_lazy_mmu_mode()
> + * because we expect set_ptes to be only be used on not present
> + * and not hw_valid ptes. Hence there is not translation cache flush
> + * involved that need to be batched.
> */
> - VM_WARN_ON(pte_hw_valid(*ptep) && !pte_protnone(*ptep));
> + for (;;) {
>
> - /* Note: mm->context.id might not yet have been assigned as
> - * this context might not have been activated yet when this
> - * is called.
> - */
> - pte = set_pte_filter(pte);
> + /*
> + * Make sure hardware valid bit is not set. We don't do
> + * tlb flush for this update.
> + */
> + VM_WARN_ON(pte_hw_valid(*ptep) && !pte_protnone(*ptep));
>
> - /* Perform the setting of the PTE */
> - arch_enter_lazy_mmu_mode();
> - for (;;) {
> + /* Note: mm->context.id might not yet have been assigned as
> + * this context might not have been activated yet when this
> + * is called.
> + */
> + pte = set_pte_filter(pte);
> +
> + /* Perform the setting of the PTE */
> __set_pte_at(mm, addr, ptep, pte, 0);
> if (--nr == 0)
> break;
> ptep++;
> - pte = __pte(pte_val(pte) + (1UL << PTE_RPN_SHIFT));
> addr += PAGE_SIZE;
> + /* increment the pfn */
> + pte = __pte(pte_val(pte) + PAGE_SIZE);
> +
> }
> - arch_leave_lazy_mmu_mode();
> }
>
> void unmap_kernel_page(unsigned long va)
Was this a new version of the patch for me to test? Sorry for asking but this was a bit unclear to me.
In any case I tried it on top of v6.6-rc6 and it did not help with the issue I reported.
Regards,
Erhard
Powered by blists - more mailing lists