Re: [PATCH] riscv/mm: Add soft-dirty page tracking support

From: Chunyan Zhang
Date: Tue Jul 16 2024 - 21:16:02 EST

On Tue, 16 Jul 2024 at 21:00, Alexandre Ghiti <alex@xxxxxxxx> wrote:
>
> Hi Chunyan,
>
> On 16/07/2024 04:16, Chunyan Zhang wrote:
> > Hi Alex,
> >
> > On Mon, 15 Jul 2024 at 19:21, Alexandre Ghiti <alex@xxxxxxxx> wrote:
> >> Hi Chunyan,
> >>
> >> On 10/07/2024 05:30, Chunyan Zhang wrote:
> >>> The PTE bit (9) is reserved for software, so we can use it for
> >>> soft-dirty tracking. This patch adds its standard handlers for
> >>> PTE, PMD, and swap entry.
> >>
> >> Unfortunately, ZONE_DEVICE has just used this last bit and should be
> >> merged in 6.11.
> > Yes, I read the patch just now.
> >
> >> I'm currently discussing internally how we can get 2 other PTE bits from
> >> RVI in order to have the same number of available bits as x86 and arm64.
> > Yes I noticed that PTE bits reserved for software are too limited on RISC-V.
> >
> > Besides softdirty, we probably can support uffd write-protect on
> > RISC-V if we will have two PTE bits for use.
>
>
> Indeed, softdirty and uffd-wp will use two PTE bits.
>
>
> >
> >> I guess that for now, if we really have a usecase for softdirty (and I
> >> think we do with CRIU), we'll have to make ZONE_DEVICE and soft-dirty
> >> mutually exclusive.
> > Yes, I also learned that CRIU uses soft-dirty.
> >
> >>> To add swap PTE soft-dirty tracking, we borrow bit (4) which is
> >>> available for swap PTEs on RISC-V systems.
> >>>
> >>> This patch has been tested with the kselftest mm suite in which
> >>> soft-dirty and madv_populate run and pass, and no regressions
> >>> are observed in any of the other tests.
> >>
> >> Did you give CRIU a try?
> > I haven't tried CRIU, actually I found soft-dirty was missing on
> > RISC-V by the way of running mm selftest cases.
>
>
> Since CRIU is the main user (?) of softdirty, it would be really nice if
> you can test it :)

Sure, and will keep you updated with the progress.

> >
> > I can cook a new patch to implement soft-dirty and ZONE_DEVICE share
> > the PTE bit(9), and make both features mutually exclusive if this
> > solution is accepted.
>
>
> I agree with this solution, let's implement both softdirty and uffd-wp
> by sharing the last PTE bit that ZONE_DEVICE stole. At least it will
> allow people to play with them.

Ok, then I will do next.

> Do you intend to work on uffd-wp? This is on my todo list, so up to you.

Yes, this is not hard for me, let me take it.

Thanks for your review,
Chunyan

> > Or not to add soft-dirty until we have more other PTE bits that can be
> > used for software.
> >
> > I'm open to listen to suggestions.
> >
> > Thanks,
> > Chunyan
>
>
> Thanks,
>
> Alex
>
>
> >
> >> Thanks,
> >>
> >> Alex
> >>
> >>
> >>> Signed-off-by: Chunyan Zhang <zhangchunyan@xxxxxxxxxxx>
> >>> ---
> >>> arch/riscv/Kconfig | 1 +
> >>> arch/riscv/include/asm/pgtable-bits.h | 13 ++++++
> >>> arch/riscv/include/asm/pgtable.h | 65 ++++++++++++++++++++++++++-
> >>> 3 files changed, 78 insertions(+), 1 deletion(-)
> >>>
> >>> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> >>> index b94176e25be1..2e3ad2925a6b 100644
> >>> --- a/arch/riscv/Kconfig
> >>> +++ b/arch/riscv/Kconfig
> >>> @@ -118,6 +118,7 @@ config RISCV
> >>> select HAVE_ARCH_MMAP_RND_COMPAT_BITS if COMPAT
> >>> select HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
> >>> select HAVE_ARCH_SECCOMP_FILTER
> >>> + select HAVE_ARCH_SOFT_DIRTY
> >>> select HAVE_ARCH_THREAD_STRUCT_WHITELIST
> >>> select HAVE_ARCH_TRACEHOOK
> >>> select HAVE_ARCH_TRANSPARENT_HUGEPAGE if 64BIT && MMU
> >>> diff --git a/arch/riscv/include/asm/pgtable-bits.h b/arch/riscv/include/asm/pgtable-bits.h
> >>> index 179bd4afece4..bab48f5fd1e2 100644
> >>> --- a/arch/riscv/include/asm/pgtable-bits.h
> >>> +++ b/arch/riscv/include/asm/pgtable-bits.h
> >>> @@ -19,6 +19,19 @@
> >>> #define _PAGE_SOFT (3 << 8) /* Reserved for software */
> >>>
> >>> #define _PAGE_SPECIAL (1 << 8) /* RSW: 0x1 */
> >>> +
> >>> +#ifdef CONFIG_MEM_SOFT_DIRTY
> >>> +#define _PAGE_SOFT_DIRTY (1 << 9) /* RSW: 0x2 for software dirty tracking */
> >>> +/*
> >>> + * BIT 4 is not involved into swap entry computation, so we
> >>> + * can borrow it for swap page soft-dirty tracking.
> >>> + */
> >>> +#define _PAGE_SWP_SOFT_DIRTY _PAGE_USER
> >>> +#else
> >>> +#define _PAGE_SOFT_DIRTY 0
> >>> +#define _PAGE_SWP_SOFT_DIRTY 0
> >>> +#endif /* CONFIG_MEM_SOFT_DIRTY */
> >>> +
> >>> #define _PAGE_TABLE _PAGE_PRESENT
> >>>
> >>> /*
> >>> diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
> >>> index aad8b8ca51f1..46f512f52580 100644
> >>> --- a/arch/riscv/include/asm/pgtable.h
> >>> +++ b/arch/riscv/include/asm/pgtable.h
> >>> @@ -408,7 +408,7 @@ static inline pte_t pte_mkwrite_novma(pte_t pte)
> >>>
> >>> static inline pte_t pte_mkdirty(pte_t pte)
> >>> {
> >>> - return __pte(pte_val(pte) | _PAGE_DIRTY);
> >>> + return __pte(pte_val(pte) | _PAGE_DIRTY | _PAGE_SOFT_DIRTY);
> >>> }
> >>>
> >>> static inline pte_t pte_mkclean(pte_t pte)
> >>> @@ -436,6 +436,36 @@ static inline pte_t pte_mkhuge(pte_t pte)
> >>> return pte;
> >>> }
> >>>
> >>> +static inline int pte_soft_dirty(pte_t pte)
> >>> +{
> >>> + return pte_val(pte) & _PAGE_SOFT_DIRTY;
> >>> +}
> >>> +
> >>> +static inline pte_t pte_mksoft_dirty(pte_t pte)
> >>> +{
> >>> + return __pte(pte_val(pte) | _PAGE_SOFT_DIRTY);
> >>> +}
> >>> +
> >>> +static inline pte_t pte_clear_soft_dirty(pte_t pte)
> >>> +{
> >>> + return __pte(pte_val(pte) & ~(_PAGE_SOFT_DIRTY));
> >>> +}
> >>> +
> >>> +static inline int pte_swp_soft_dirty(pte_t pte)
> >>> +{
> >>> + return pte_val(pte) & _PAGE_SWP_SOFT_DIRTY;
> >>> +}
> >>> +
> >>> +static inline pte_t pte_swp_mksoft_dirty(pte_t pte)
> >>> +{
> >>> + return __pte(pte_val(pte) | _PAGE_SWP_SOFT_DIRTY);
> >>> +}
> >>> +
> >>> +static inline pte_t pte_swp_clear_soft_dirty(pte_t pte)
> >>> +{
> >>> + return __pte(pte_val(pte) & ~(_PAGE_SWP_SOFT_DIRTY));
> >>> +}
> >>> +
> >>> #ifdef CONFIG_RISCV_ISA_SVNAPOT
> >>> #define pte_leaf_size(pte) (pte_napot(pte) ? \
> >>> napot_cont_size(napot_cont_order(pte)) :\
> >>> @@ -721,6 +751,38 @@ static inline pmd_t pmd_mkdirty(pmd_t pmd)
> >>> return pte_pmd(pte_mkdirty(pmd_pte(pmd)));
> >>> }
> >>>
> >>> +static inline int pmd_soft_dirty(pmd_t pmd)
> >>> +{
> >>> + return pte_soft_dirty(pmd_pte(pmd));
> >>> +}
> >>> +
> >>> +static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
> >>> +{
> >>> + return pte_pmd(pte_mksoft_dirty(pmd_pte(pmd)));
> >>> +}
> >>> +
> >>> +static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
> >>> +{
> >>> + return pte_pmd(pte_clear_soft_dirty(pmd_pte(pmd)));
> >>> +}
> >>> +
> >>> +#ifdef CONFIG_ARCH_ENABLE_THP_MIGRATION
> >>> +static inline int pmd_swp_soft_dirty(pmd_t pmd)
> >>> +{
> >>> + return pte_swp_soft_dirty(pmd_pte(pmd));
> >>> +}
> >>> +
> >>> +static inline pmd_t pmd_swp_mksoft_dirty(pmd_t pmd)
> >>> +{
> >>> + return pte_pmd(pte_swp_mksoft_dirty(pmd_pte(pmd)));
> >>> +}
> >>> +
> >>> +static inline pmd_t pmd_swp_clear_soft_dirty(pmd_t pmd)
> >>> +{
> >>> + return pte_pmd(pte_swp_clear_soft_dirty(pmd_pte(pmd)));
> >>> +}
> >>> +#endif /* CONFIG_ARCH_ENABLE_THP_MIGRATION */
> >>> +
> >>> static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
> >>> pmd_t *pmdp, pmd_t pmd)
> >>> {
> >>> @@ -811,6 +873,7 @@ extern pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
> >>> * Format of swap PTE:
> >>> * bit 0: _PAGE_PRESENT (zero)
> >>> * bit 1 to 3: _PAGE_LEAF (zero)
> >>> + * bit 4: _PAGE_SWP_SOFT_DIRTY
> >>> * bit 5: _PAGE_PROT_NONE (zero)
> >>> * bit 6: exclusive marker
> >>> * bits 7 to 11: swap type
> > _______________________________________________
> > linux-riscv mailing list
> > linux-riscv@xxxxxxxxxxxxxxxxxxx
> > http://lists.infradead.org/mailman/listinfo/linux-riscv