Re: [PATCH 1/4] mm: Export flush_vm_area() to sync the PTEs upon construction
Quoting Joerg Roedel (2020-08-21 11:22:04) > On Fri, Aug 21, 2020 at 10:54:22AM +0100, Chris Wilson wrote: > > Ok. I thought it had to be after assigning the *ptep. If we apply the > > sync first, do not have to worry about PGTBL_PTE_MODIFIED from the > > *ptep? > > Hmm, if I understand the code correctly, you are re-implementing some > generic ioremap/vmalloc mapping logic in the i915 driver. I don't know > the reason, but if it is valid you need to manually call > arch_sync_kernel_mappings() from your driver like this to be correct: > > if (ARCH_PAGE_TABLE_SYNC_MASK & PGTBL_PTE_MODIFIED) > arch_sync_kernel_mappings(); > > In practice this is a no-op, because nobody sets PGTBL_PTE_MODIFIED in > ARCH_PAGE_TABLE_SYNC_MASK, so the above code would be optimized away. > > But what you really care about is the tracking in apply_to_page_range(), > as that allocates the !pte levels of your page-table, which needs > synchronization on x86-32. > > Btw, what are the reasons you can't use generic vmalloc/ioremap > interfaces to map the range? ioremap_prot and ioremap_page_range assume a contiguous IO address. So we needed to allocate the vmalloc area [and would then need to iterate over the discontiguous iomem chunks with ioremap_page_range], and since alloc_vm_area returned the ptep, it looked clearer to then assign those according to whether we wanted ioremapping or a plain page. So we ended up with one call to the core to return us a vm_struct and a pte array that worked for either backing store. -Chris
Re: [PATCH 1/4] mm: Export flush_vm_area() to sync the PTEs upon construction
On Fri, Aug 21, 2020 at 10:54:22AM +0100, Chris Wilson wrote: > Ok. I thought it had to be after assigning the *ptep. If we apply the > sync first, do not have to worry about PGTBL_PTE_MODIFIED from the > *ptep? Hmm, if I understand the code correctly, you are re-implementing some generic ioremap/vmalloc mapping logic in the i915 driver. I don't know the reason, but if it is valid you need to manually call arch_sync_kernel_mappings() from your driver like this to be correct: if (ARCH_PAGE_TABLE_SYNC_MASK & PGTBL_PTE_MODIFIED) arch_sync_kernel_mappings(); In practice this is a no-op, because nobody sets PGTBL_PTE_MODIFIED in ARCH_PAGE_TABLE_SYNC_MASK, so the above code would be optimized away. But what you really care about is the tracking in apply_to_page_range(), as that allocates the !pte levels of your page-table, which needs synchronization on x86-32. Btw, what are the reasons you can't use generic vmalloc/ioremap interfaces to map the range? Regards, Joerg
Re: [PATCH 1/4] mm: Export flush_vm_area() to sync the PTEs upon construction
Quoting Joerg Roedel (2020-08-21 10:51:29)
> On Fri, Aug 21, 2020 at 09:50:08AM +0100, Chris Wilson wrote:
> > The alloc_vm_area() is another method for drivers to
> > vmap/map_kernel_range that uses apply_to_page_range() rather than the
> > direct vmalloc walkers. This is missing the page table modification
> > tracking, and the ability to synchronize the PTE updates afterwards.
> > Provide flush_vm_area() for the users of alloc_vm_area() that assumes
> > the worst and ensures that the page directories are correctly flushed
> > upon construction.
> >
> > The impact is most pronounced on x86_32 due to the delayed set_pmd().
> >
> > Reported-by: Pavel Machek
> > References: 2ba3e6947aed ("mm/vmalloc: track which page-table levels were
> > modified")
> > References: 86cf69f1d893 ("x86/mm/32: implement
> > arch_sync_kernel_mappings()")
> > Signed-off-by: Chris Wilson
> > Cc: Andrew Morton
> > Cc: Joerg Roedel
> > Cc: Linus Torvalds
> > Cc: Dave Airlie
> > Cc: Joonas Lahtinen
> > Cc: Rodrigo Vivi
> > Cc: Pavel Machek
> > Cc: David Vrabel
> > Cc: # v5.8+
> > ---
> > include/linux/vmalloc.h | 1 +
> > mm/vmalloc.c| 16
> > 2 files changed, 17 insertions(+)
> >
> > diff --git a/include/linux/vmalloc.h b/include/linux/vmalloc.h
> > index 0221f852a7e1..a253b27df0ac 100644
> > --- a/include/linux/vmalloc.h
> > +++ b/include/linux/vmalloc.h
> > @@ -204,6 +204,7 @@ static inline void set_vm_flush_reset_perms(void *addr)
> >
> > /* Allocate/destroy a 'vmalloc' VM area. */
> > extern struct vm_struct *alloc_vm_area(size_t size, pte_t **ptes);
> > +extern void flush_vm_area(struct vm_struct *area);
> > extern void free_vm_area(struct vm_struct *area);
> >
> > /* for /dev/kmem */
> > diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> > index b482d240f9a2..c41934486031 100644
> > --- a/mm/vmalloc.c
> > +++ b/mm/vmalloc.c
> > @@ -3078,6 +3078,22 @@ struct vm_struct *alloc_vm_area(size_t size, pte_t
> > **ptes)
> > }
> > EXPORT_SYMBOL_GPL(alloc_vm_area);
> >
> > +void flush_vm_area(struct vm_struct *area)
> > +{
> > + unsigned long addr = (unsigned long)area->addr;
> > +
> > + /* apply_to_page_range() doesn't track the damage, assume the worst */
> > + if (ARCH_PAGE_TABLE_SYNC_MASK & (PGTBL_PTE_MODIFIED |
> > + PGTBL_PMD_MODIFIED |
> > + PGTBL_PUD_MODIFIED |
> > + PGTBL_P4D_MODIFIED |
> > + PGTBL_PGD_MODIFIED))
> > + arch_sync_kernel_mappings(addr, addr + area->size);
>
> This should happen in __apply_to_page_range() directly and look like
> this:
Ok. I thought it had to be after assigning the *ptep. If we apply the
sync first, do not have to worry about PGTBL_PTE_MODIFIED from the
*ptep?
-Chris
Re: [PATCH 1/4] mm: Export flush_vm_area() to sync the PTEs upon construction
On Fri, Aug 21, 2020 at 09:50:08AM +0100, Chris Wilson wrote:
> The alloc_vm_area() is another method for drivers to
> vmap/map_kernel_range that uses apply_to_page_range() rather than the
> direct vmalloc walkers. This is missing the page table modification
> tracking, and the ability to synchronize the PTE updates afterwards.
> Provide flush_vm_area() for the users of alloc_vm_area() that assumes
> the worst and ensures that the page directories are correctly flushed
> upon construction.
>
> The impact is most pronounced on x86_32 due to the delayed set_pmd().
>
> Reported-by: Pavel Machek
> References: 2ba3e6947aed ("mm/vmalloc: track which page-table levels were
> modified")
> References: 86cf69f1d893 ("x86/mm/32: implement arch_sync_kernel_mappings()")
> Signed-off-by: Chris Wilson
> Cc: Andrew Morton
> Cc: Joerg Roedel
> Cc: Linus Torvalds
> Cc: Dave Airlie
> Cc: Joonas Lahtinen
> Cc: Rodrigo Vivi
> Cc: Pavel Machek
> Cc: David Vrabel
> Cc: # v5.8+
> ---
> include/linux/vmalloc.h | 1 +
> mm/vmalloc.c| 16
> 2 files changed, 17 insertions(+)
>
> diff --git a/include/linux/vmalloc.h b/include/linux/vmalloc.h
> index 0221f852a7e1..a253b27df0ac 100644
> --- a/include/linux/vmalloc.h
> +++ b/include/linux/vmalloc.h
> @@ -204,6 +204,7 @@ static inline void set_vm_flush_reset_perms(void *addr)
>
> /* Allocate/destroy a 'vmalloc' VM area. */
> extern struct vm_struct *alloc_vm_area(size_t size, pte_t **ptes);
> +extern void flush_vm_area(struct vm_struct *area);
> extern void free_vm_area(struct vm_struct *area);
>
> /* for /dev/kmem */
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index b482d240f9a2..c41934486031 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -3078,6 +3078,22 @@ struct vm_struct *alloc_vm_area(size_t size, pte_t
> **ptes)
> }
> EXPORT_SYMBOL_GPL(alloc_vm_area);
>
> +void flush_vm_area(struct vm_struct *area)
> +{
> + unsigned long addr = (unsigned long)area->addr;
> +
> + /* apply_to_page_range() doesn't track the damage, assume the worst */
> + if (ARCH_PAGE_TABLE_SYNC_MASK & (PGTBL_PTE_MODIFIED |
> + PGTBL_PMD_MODIFIED |
> + PGTBL_PUD_MODIFIED |
> + PGTBL_P4D_MODIFIED |
> + PGTBL_PGD_MODIFIED))
> + arch_sync_kernel_mappings(addr, addr + area->size);
This should happen in __apply_to_page_range() directly and look like
this:
if (ARCH_PAGE_TABLE_SYNC_MASK && create)
arch_sync_kernel_mappings(addr, addr + size);
Or even better, track whether something had to be allocated in the
__apply_to_page_range() path and check for:
if (ARCH_PAGE_TABLE_SYNC_MASK & mask)
