On Wed, Nov 4, 2020 at 10:41 AM Jan Kiszka <jan.kis...@siemens.com> wrote:
> From: Jan Kiszka <jan.kis...@siemens.com>
>
> This is not needed, in fact, and it is in conflict with upcoming stable
> changes. Philippe writes:
>
> "IIRC, the reason was to ensure that any mapping duplicated from a
> parent process which has pinned its memory via an earlier call to
> __ipipe_disable_ondemand_mappings() would be immediately un-COWed upon
> fork() in the child process. On second thoughts, this is useless since
> the child process either:
>
> - if it has to, explicitly attaches to the real-time core, which ends up
> calling mlock(), therefore causing these mappings to be un-COWed if
> need be. IOW, there is no inheritance of the Cobalt context at
> fork. Maybe at some point in time there was the assumption that such
> context might be shared between both processes somehow, which has been
> ruled out later on.
>
> - the child process does not attach to the real-time core, in which case
> COW is a non-issue."
>
> Signed-off-by: Jan Kiszka <jan.kis...@siemens.com>
> ---
>
> Greg, needed in case you update to 4.19.149 or later. Not needed for 5.4
> (I already left out that hunks).
>
> mm/memory.c | 61 +++++------------------------------------------------
> 1 file changed, 5 insertions(+), 56 deletions(-)
>
> diff --git a/mm/memory.c b/mm/memory.c
> index f36626dc30d0..797f7fe29454 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -129,11 +129,6 @@ EXPORT_SYMBOL(zero_pfn);
> unsigned long highest_memmap_pfn __read_mostly;
> -static inline void cow_user_page(struct page *dst,
> - struct page *src,
> - unsigned long va,
> - struct vm_area_struct *vma);
> -
> /*
> * CONFIG_MMU architectures set up ZERO_PAGE in their paging_init()
> */
> @@ -950,8 +945,8 @@ struct page *vm_normal_page_pmd(struct vm_area_struct
> *vma, unsigned long addr,
> static inline unsigned long
> copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
> - pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
> - unsigned long addr, int *rss, struct page *uncow_page)
> + pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *vma,
> + unsigned long addr, int *rss)
> {
> unsigned long vm_flags = vma->vm_flags;
> pte_t pte = *src_pte;
> @@ -1029,21 +1024,6 @@ copy_one_pte(struct mm_struct *dst_mm, struct
> mm_struct *src_mm,
> * in the parent and the child
> */
> if (is_cow_mapping(vm_flags) && pte_write(pte)) {
> -#ifdef CONFIG_IPIPE
> - if (uncow_page) {
> - struct page *old_page = vm_normal_page(vma, addr,
> pte);
> - cow_user_page(uncow_page, old_page, addr, vma);
> - pte = mk_pte(uncow_page, vma->vm_page_prot);
> -
> - if (vm_flags & VM_SHARED)
> - pte = pte_mkclean(pte);
> - pte = pte_mkold(pte);
> -
> - page_add_new_anon_rmap(uncow_page, vma, addr,
> false);
> - rss[!!PageAnon(uncow_page)]++;
> - goto out_set_pte;
> - }
> -#endif /* CONFIG_IPIPE */
> ptep_set_wrprotect(src_mm, addr, src_pte);
> pte = pte_wrprotect(pte);
> }
> @@ -1091,27 +1071,13 @@ static int copy_pte_range(struct mm_struct
> *dst_mm, struct mm_struct *src_mm,
> int progress = 0;
> int rss[NR_MM_COUNTERS];
> swp_entry_t entry = (swp_entry_t){0};
> - struct page *uncow_page = NULL;
> -#ifdef CONFIG_IPIPE
> - int do_cow_break = 0;
> -again:
> - if (do_cow_break) {
> - uncow_page = alloc_page_vma(GFP_HIGHUSER, vma, addr);
> - if (uncow_page == NULL)
> - return -ENOMEM;
> - do_cow_break = 0;
> - }
> -#else
> +
> again:
> -#endif
> init_rss_vec(rss);
> dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
> - if (!dst_pte) {
> - if (uncow_page)
> - put_page(uncow_page);
> + if (!dst_pte)
> return -ENOMEM;
> - }
> src_pte = pte_offset_map(src_pmd, addr);
> src_ptl = pte_lockptr(src_mm, src_pmd);
> spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
> @@ -1134,25 +1100,8 @@ static int copy_pte_range(struct mm_struct *dst_mm,
> struct mm_struct *src_mm,
> progress++;
> continue;
> }
> -#ifdef CONFIG_IPIPE
> - if (likely(uncow_page == NULL) &&
> likely(pte_present(*src_pte))) {
> - if (is_cow_mapping(vma->vm_flags) &&
> - test_bit(MMF_VM_PINNED, &src_mm->flags) &&
> - ((vma->vm_flags|src_mm->def_flags) &
> VM_LOCKED)) {
> - arch_leave_lazy_mmu_mode();
> - spin_unlock(src_ptl);
> - pte_unmap(src_pte);
> - add_mm_rss_vec(dst_mm, rss);
> - pte_unmap_unlock(dst_pte, dst_ptl);
> - cond_resched();
> - do_cow_break = 1;
> - goto again;
> - }
> - }
> -#endif
> entry.val = copy_one_pte(dst_mm, src_mm, dst_pte, src_pte,
> - vma, addr, rss, uncow_page);
> - uncow_page = NULL;
> + vma, addr, rss);
> if (entry.val)
> break;
> progress += 8;
> --
> 2.26.2
Ok, still working on 5.4 but I plan to do a final 4.19 release once 5.4 is
stable.
Greg
>
>
