Re: [PATCH v2 28/34] kasan, vmalloc: add vmalloc support to HW_TAGS

[Andrey Konovalov <andreyknvl@gmail.com>]

On Mon, Dec 6, 2021 at 10:46 PM <andrey.konovalov@linux.dev> wrote:
>
> From: Andrey Konovalov <andreyknvl@google.com>
>
> This patch adds vmalloc tagging support to HW_TAGS KASAN.
>
> The key difference between HW_TAGS and the other two KASAN modes
> when it comes to vmalloc: HW_TAGS KASAN can only assign tags to
> physical memory. The other two modes have shadow memory covering
> every mapped virtual memory region.
>
> This patch makes __kasan_unpoison_vmalloc() for HW_TAGS KASAN:
>
> - Skip non-VM_ALLOC mappings as HW_TAGS KASAN can only tag a single
>   mapping of normal physical memory; see the comment in the function.
> - Generate a random tag, tag the returned pointer and the allocation,
>   and initialize the allocation at the same time.
> - Propagate the tag into the page stucts to allow accesses through
>   page_address(vmalloc_to_page()).
>
> The rest of vmalloc-related KASAN hooks are not needed:
>
> - The shadow-related ones are fully skipped.
> - __kasan_poison_vmalloc() is kept as a no-op with a comment.
>
> Poisoning and zeroing of physical pages that are backing vmalloc()
> allocations are skipped via __GFP_SKIP_KASAN_UNPOISON and
> __GFP_SKIP_ZERO: __kasan_unpoison_vmalloc() does that instead.
>
> This patch allows enabling CONFIG_KASAN_VMALLOC with HW_TAGS
> and adjusts CONFIG_KASAN_VMALLOC description:
>
> - Mention HW_TAGS support.
> - Remove unneeded internal details: they have no place in Kconfig
>   description and are already explained in the documentation.
>
> Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
> Co-developed-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
>
> ---
>
> Changes v1->v2:
> - Allow enabling CONFIG_KASAN_VMALLOC with HW_TAGS in this patch.
> - Move memory init for page_alloc pages backing vmalloc() into
>   kasan_unpoison_vmalloc().
> ---
>  include/linux/kasan.h | 30 +++++++++++++--
>  lib/Kconfig.kasan     | 20 +++++-----
>  mm/kasan/hw_tags.c    | 89 +++++++++++++++++++++++++++++++++++++++++++
>  mm/kasan/shadow.c     | 11 +++++-
>  mm/vmalloc.c          | 32 +++++++++++++---
>  5 files changed, 162 insertions(+), 20 deletions(-)
>
> diff --git a/include/linux/kasan.h b/include/linux/kasan.h
> index 6a2619759e93..0bdc2b824b9c 100644
> --- a/include/linux/kasan.h
> +++ b/include/linux/kasan.h
> @@ -417,19 +417,40 @@ static inline void kasan_init_hw_tags(void) { }
>
>  #ifdef CONFIG_KASAN_VMALLOC
>
> +#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
> +
>  void kasan_populate_early_vm_area_shadow(void *start, unsigned long size);
>  int kasan_populate_vmalloc(unsigned long addr, unsigned long size);
>  void kasan_release_vmalloc(unsigned long start, unsigned long end,
>                            unsigned long free_region_start,
>                            unsigned long free_region_end);
>
> +#else /* CONFIG_KASAN_GENERIC || CONFIG_KASAN_SW_TAGS */
> +
> +static inline void kasan_populate_early_vm_area_shadow(void *start,
> +                                                      unsigned long size)
> +{ }
> +static inline int kasan_populate_vmalloc(unsigned long start,
> +                                       unsigned long size)
> +{
> +       return 0;
> +}
> +static inline void kasan_release_vmalloc(unsigned long start,
> +                                        unsigned long end,
> +                                        unsigned long free_region_start,
> +                                        unsigned long free_region_end) { }
> +
> +#endif /* CONFIG_KASAN_GENERIC || CONFIG_KASAN_SW_TAGS */
> +
>  void * __must_check __kasan_unpoison_vmalloc(const void *start,
> -                                            unsigned long size);
> +                                            unsigned long size,
> +                                            bool vm_alloc, bool init);
>  static __always_inline void * __must_check kasan_unpoison_vmalloc(
> -                                       const void *start, unsigned long size)
> +                                       const void *start, unsigned long size,
> +                                       bool vm_alloc, bool init)
>  {
>         if (kasan_enabled())
> -               return __kasan_unpoison_vmalloc(start, size);
> +               return __kasan_unpoison_vmalloc(start, size, vm_alloc, init);
>         return (void *)start;
>  }
>
> @@ -456,7 +477,8 @@ static inline void kasan_release_vmalloc(unsigned long start,
>                                          unsigned long free_region_end) { }
>
>  static inline void *kasan_unpoison_vmalloc(const void *start,
> -                                          unsigned long size, bool unique)
> +                                          unsigned long size,
> +                                          bool vm_alloc, bool init)
>  {
>         return (void *)start;
>  }
> diff --git a/lib/Kconfig.kasan b/lib/Kconfig.kasan
> index 3f144a87f8a3..7834c35a7964 100644
> --- a/lib/Kconfig.kasan
> +++ b/lib/Kconfig.kasan
> @@ -178,17 +178,17 @@ config KASAN_TAGS_IDENTIFY
>           memory consumption.
>
>  config KASAN_VMALLOC
> -       bool "Back mappings in vmalloc space with real shadow memory"
> -       depends on (KASAN_GENERIC || KASAN_SW_TAGS) && HAVE_ARCH_KASAN_VMALLOC
> +       bool "Check accesses to vmalloc allocations"
> +       depends on HAVE_ARCH_KASAN_VMALLOC
>         help
> -         By default, the shadow region for vmalloc space is the read-only
> -         zero page. This means that KASAN cannot detect errors involving
> -         vmalloc space.
> -
> -         Enabling this option will hook in to vmap/vmalloc and back those
> -         mappings with real shadow memory allocated on demand. This allows
> -         for KASAN to detect more sorts of errors (and to support vmapped
> -         stacks), but at the cost of higher memory usage.
> +         This mode makes KASAN check accesses to vmalloc allocations for
> +         validity.
> +
> +         With software KASAN modes, checking is done for all types of vmalloc
> +         allocations. Enabling this option leads to higher memory usage.
> +
> +         With hardware tag-based KASAN, only VM_ALLOC mappings are checked.
> +         There is no additional memory usage.
>
>  config KASAN_KUNIT_TEST
>         tristate "KUnit-compatible tests of KASAN bug detection capabilities" if !KUNIT_ALL_TESTS
> diff --git a/mm/kasan/hw_tags.c b/mm/kasan/hw_tags.c
> index 76cf2b6229c7..837c260beec6 100644
> --- a/mm/kasan/hw_tags.c
> +++ b/mm/kasan/hw_tags.c
> @@ -192,6 +192,95 @@ void __init kasan_init_hw_tags(void)
>                 kasan_stack_collection_enabled() ? "on" : "off");
>  }
>
> +#ifdef CONFIG_KASAN_VMALLOC
> +
> +static void unpoison_vmalloc_pages(const void *addr, u8 tag)
> +{
> +       struct vm_struct *area;
> +       int i;
> +
> +       /*
> +        * As hardware tag-based KASAN only tags VM_ALLOC vmalloc allocations
> +        * (see the comment in __kasan_unpoison_vmalloc), all of the pages
> +        * should belong to a single area.
> +        */
> +       area = find_vm_area((void *)addr);
> +       if (WARN_ON(!area))
> +               return;
> +
> +       for (i = 0; i < area->nr_pages; i++) {
> +               struct page *page = area->pages[i];
> +
> +               page_kasan_tag_set(page, tag);
> +       }
> +}
> +
> +void *__kasan_unpoison_vmalloc(const void *start, unsigned long size,
> +                               bool vm_alloc, bool init)
> +{
> +       u8 tag;
> +       unsigned long redzone_start, redzone_size;
> +
> +       if (!is_vmalloc_or_module_addr(start))
> +               return (void *)start;
> +
> +       /* Unpoisoning and pointer tag assignment is skipped for non-VM_ALLOC
> +        * mappings as:
> +        *
> +        * 1. Unlike the software KASAN modes, hardware tag-based KASAN only
> +        *    supports tagging physical memory. Therefore, it can only tag a
> +        *    single mapping of normal physical pages.
> +        * 2. Hardware tag-based KASAN can only tag memory mapped with special
> +        *    mapping protection bits, see arch_vmalloc_pgprot_modify().
> +        *    As non-VM_ALLOC mappings can be mapped outside of vmalloc code,
> +        *    providing these bits would require tracking all non-VM_ALLOC
> +        *    mappers.
> +        *
> +        * Thus, for VM_ALLOC mappings, hardware tag-based KASAN only tags
> +        * the first virtual mapping, which is created by vmalloc().
> +        * Tagging the page_alloc memory backing that vmalloc() allocation is
> +        * skipped, see ___GFP_SKIP_KASAN_UNPOISON.
> +        *
> +        * For non-VM_ALLOC allocations, page_alloc memory is tagged as usual.
> +        */
> +       if (!vm_alloc)
> +               return (void *)start;
> +
> +       tag = kasan_random_tag();
> +       start = set_tag(start, tag);
> +
> +       /* Unpoison and initialize memory up to size. */
> +       kasan_unpoison(start, size, init);
> +
> +       /*
> +        * Explicitly poison and initialize the in-page vmalloc() redzone.
> +        * Unlike software KASAN modes, hardware tag-based KASAN doesn't
> +        * unpoison memory when populating shadow for vmalloc() space.
> +        */
> +       redzone_start = round_up((unsigned long)start + size, KASAN_GRANULE_SIZE);
> +       redzone_size = round_up(redzone_start, PAGE_SIZE) - redzone_start;
> +       kasan_poison((void *)redzone_start, redzone_size, KASAN_TAG_INVALID, init);
> +
> +       /*
> +        * Set per-page tag flags to allow accessing physical memory for the
> +        * vmalloc() mapping through page_address(vmalloc_to_page()).
> +        */
> +       unpoison_vmalloc_pages(start, tag);
> +
> +       return (void *)start;
> +}
> +
> +void __kasan_poison_vmalloc(const void *start, unsigned long size)
> +{
> +       /*
> +        * No tagging here.
> +        * The physical pages backing the vmalloc() allocation are poisoned
> +        * through the usual page_alloc paths.
> +        */
> +}
> +
> +#endif
> +
>  #if IS_ENABLED(CONFIG_KASAN_KUNIT_TEST)
>
>  void kasan_enable_tagging_sync(void)
> diff --git a/mm/kasan/shadow.c b/mm/kasan/shadow.c
> index 4ca280a96fbc..8600dd925f35 100644
> --- a/mm/kasan/shadow.c
> +++ b/mm/kasan/shadow.c
> @@ -475,8 +475,17 @@ void kasan_release_vmalloc(unsigned long start, unsigned long end,
>         }
>  }
>
> -void *__kasan_unpoison_vmalloc(const void *start, unsigned long size)
> +void *__kasan_unpoison_vmalloc(const void *start, unsigned long size,
> +                               bool vm_alloc, bool init)
>  {
> +       /*
> +        * Software tag-based KASAN tags both VM_ALLOC and non-VM_ALLOC
> +        * mappings, so the vm_alloc argument is ignored.
> +        * Software tag-based KASAN can't optimize zeroing memory by combining
> +        * it with setting memory tags, so the init argument is ignored;
> +        * vmalloc() memory is poisoned via page_alloc.
> +        */
> +
>         if (!is_vmalloc_or_module_addr(start))
>                 return (void *)start;
>
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index 82ef1e27e2e4..d48db7cc3358 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2214,8 +2214,12 @@ void *vm_map_ram(struct page **pages, unsigned int count, int node)
>                 return NULL;
>         }
>
> -       /* Mark the pages as accessible after they were mapped in. */
> -       mem = kasan_unpoison_vmalloc(mem, size);
> +       /*
> +        * Mark the pages as accessible after they were mapped in.
> +        * With hardware tag-based KASAN, marking is skipped for
> +        * non-VM_ALLOC mappings, see __kasan_unpoison_vmalloc().
> +        */
> +       mem = kasan_unpoison_vmalloc(mem, size, false, false);
>
>         return mem;
>  }
> @@ -2449,9 +2453,12 @@ static struct vm_struct *__get_vm_area_node(unsigned long size,
>          * accessible after they are mapped in.
>          * Otherwise, as the pages can be mapped outside of vmalloc code,
>          * mark them now as a best-effort approach.
> +        * With hardware tag-based KASAN, marking is skipped for
> +        * non-VM_ALLOC mappings, see __kasan_unpoison_vmalloc().
>          */
>         if (!(flags & VM_ALLOC))
> -               area->addr = kasan_unpoison_vmalloc(area->addr, requested_size);
> +               area->addr = kasan_unpoison_vmalloc(area->addr, requested_size,
> +                                                       false, false);
>
>         return area;
>  }
> @@ -2849,6 +2856,12 @@ vm_area_alloc_pages(gfp_t gfp, int nid,
>         struct page *page;
>         int i;
>
> +       /*
> +        * Skip page_alloc poisoning and zeroing for pages backing VM_ALLOC
> +        * mappings. Only effective in HW_TAGS mode.
> +        */
> +       gfp &= __GFP_SKIP_KASAN_UNPOISON & __GFP_SKIP_ZERO;
> +
>         /*
>          * For order-0 pages we make use of bulk allocator, if
>          * the page array is partly or not at all populated due
> @@ -3027,6 +3040,7 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
>  {
>         struct vm_struct *area;
>         void *addr;
> +       bool init;
>         unsigned long real_size = size;
>         unsigned long real_align = align;
>         unsigned int shift = PAGE_SHIFT;
> @@ -3083,8 +3097,13 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
>         /*
>          * Mark the pages for VM_ALLOC mappings as accessible after they were
>          * mapped in.
> +        * The init condition should match the one in post_alloc_hook()
> +        * (except for the should_skip_init() check) to make sure that memory
> +        * is initialized under the same conditions regardless of the enabled
> +        * KASAN mode.
>          */
> -       addr = kasan_unpoison_vmalloc(addr, real_size);
> +       init = !want_init_on_free() && want_init_on_alloc(gfp_mask);
> +       addr = kasan_unpoison_vmalloc(addr, real_size, true, init);
>
>         /*
>          * In this function, newly allocated vm_struct has VM_UNINITIALIZED
> @@ -3784,10 +3803,13 @@ struct vm_struct **pcpu_get_vm_areas(const unsigned long *offsets,
>          * Mark allocated areas as accessible.
>          * As the pages are mapped outside of vmalloc code,
>          * mark them now as a best-effort approach.
> +        * With hardware tag-based KASAN, marking is skipped for
> +        * non-VM_ALLOC mappings, see __kasan_unpoison_vmalloc().
>          */
>         for (area = 0; area < nr_vms; area++)
>                 vms[area]->addr = kasan_unpoison_vmalloc(vms[area]->addr,
> -                                                        vms[area]->size);
> +                                                        vms[area]->size,
> +                                                        false, false);
>
>         kfree(vas);
>         return vms;
> --
> 2.25.1
>

Hi Vincenzo,

This patch is partially based on an early version of the HW_TAGS
series you had. Could you PTAL and give your sign-off?

Thanks!