Date:   Tue, 25 Oct 2022 13:53:54 +0200
From:   Vlastimil Babka <vbabka@...e.cz>
To:     Kees Cook <keescook@...omium.org>,
        Christoph Lameter <cl@...ux.com>,
        Dmitry Vyukov <dvyukov@...gle.com>
Cc:     Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
        Pekka Enberg <penberg@...nel.org>,
        David Rientjes <rientjes@...gle.com>,
        Joonsoo Kim <iamjoonsoo.kim@....com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Roman Gushchin <roman.gushchin@...ux.dev>,
        Hyeonggon Yoo <42.hyeyoo@...il.com>,
        Andrey Ryabinin <ryabinin.a.a@...il.com>,
        Alexander Potapenko <glider@...gle.com>,
        Andrey Konovalov <andreyknvl@...il.com>,
        Vincenzo Frascino <vincenzo.frascino@....com>,
        linux-mm@...ck.org, kasan-dev@...glegroups.com,
        netdev@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-hardening@...r.kernel.org
Subject: Re: [PATCH] mm: Make ksize() a reporting-only function

On 10/22/22 20:08, Kees Cook wrote:
> With all "silently resizing" callers of ksize() refactored, remove the
> logic in ksize() that would allow it to be used to effectively change
> the size of an allocation (bypassing __alloc_size hints, etc). Users
> wanting this feature need to either use kmalloc_size_roundup() before an
> allocation, or use krealloc() directly.
> 
> For kfree_sensitive(), move the unpoisoning logic inline. Replace the
> some of the partially open-coded ksize() in __do_krealloc with ksize()
> now that it doesn't perform unpoisoning.
> 
> Adjust the KUnit tests to match the new ksize() behavior.
> 
> Cc: Dmitry Vyukov <dvyukov@...gle.com>
> Cc: Jakub Kicinski <kuba@...nel.org>
> Cc: Paolo Abeni <pabeni@...hat.com>
> Cc: Christoph Lameter <cl@...ux.com>
> Cc: Pekka Enberg <penberg@...nel.org>
> Cc: David Rientjes <rientjes@...gle.com>
> Cc: Joonsoo Kim <iamjoonsoo.kim@....com>
> Cc: Andrew Morton <akpm@...ux-foundation.org>
> Cc: Vlastimil Babka <vbabka@...e.cz>
> Cc: Roman Gushchin <roman.gushchin@...ux.dev>
> Cc: Hyeonggon Yoo <42.hyeyoo@...il.com>
> Cc: Andrey Ryabinin <ryabinin.a.a@...il.com>
> Cc: Alexander Potapenko <glider@...gle.com>
> Cc: Andrey Konovalov <andreyknvl@...il.com>
> Cc: Vincenzo Frascino <vincenzo.frascino@....com>
> Cc: linux-mm@...ck.org
> Cc: kasan-dev@...glegroups.com
> Cc: netdev@...r.kernel.org
> Signed-off-by: Kees Cook <keescook@...omium.org>

Acked-by: Vlastimil Babka <vbabka@...e.cz>

> ---
> This requires at least this be landed first:
> https://lore.kernel.org/lkml/20221021234713.you.031-kees@kernel.org/

Don't we need all parts to have landed first, even if the skbuff one is the
most prominent?

> I suspect given that is the most central ksize() user, this ksize()
> fix might be best to land through the netdev tree...
> ---
>  mm/kasan/kasan_test.c |  8 +++++---
>  mm/slab_common.c      | 33 ++++++++++++++-------------------
>  2 files changed, 19 insertions(+), 22 deletions(-)
> 
> diff --git a/mm/kasan/kasan_test.c b/mm/kasan/kasan_test.c
> index 0d59098f0876..cb5c54adb503 100644
> --- a/mm/kasan/kasan_test.c
> +++ b/mm/kasan/kasan_test.c
> @@ -783,7 +783,7 @@ static void kasan_global_oob_left(struct kunit *test)
>  	KUNIT_EXPECT_KASAN_FAIL(test, *(volatile char *)p);
>  }
>  
> -/* Check that ksize() makes the whole object accessible. */
> +/* Check that ksize() does NOT unpoison whole object. */
>  static void ksize_unpoisons_memory(struct kunit *test)
>  {
>  	char *ptr;
> @@ -791,15 +791,17 @@ static void ksize_unpoisons_memory(struct kunit *test)
>  
>  	ptr = kmalloc(size, GFP_KERNEL);
>  	KUNIT_ASSERT_NOT_ERR_OR_NULL(test, ptr);
> +
>  	real_size = ksize(ptr);
> +	KUNIT_EXPECT_GT(test, real_size, size);
>  
>  	OPTIMIZER_HIDE_VAR(ptr);
>  
>  	/* This access shouldn't trigger a KASAN report. */
> -	ptr[size] = 'x';
> +	ptr[size - 1] = 'x';
>  
>  	/* This one must. */
> -	KUNIT_EXPECT_KASAN_FAIL(test, ((volatile char *)ptr)[real_size]);
> +	KUNIT_EXPECT_KASAN_FAIL(test, ((volatile char *)ptr)[real_size - 1]);
>  
>  	kfree(ptr);
>  }
> diff --git a/mm/slab_common.c b/mm/slab_common.c
> index 33b1886b06eb..eabd66fcabd0 100644
> --- a/mm/slab_common.c
> +++ b/mm/slab_common.c
> @@ -1333,11 +1333,11 @@ __do_krealloc(const void *p, size_t new_size, gfp_t flags)
>  	void *ret;
>  	size_t ks;
>  
> -	/* Don't use instrumented ksize to allow precise KASAN poisoning. */
> +	/* Check for double-free before calling ksize. */
>  	if (likely(!ZERO_OR_NULL_PTR(p))) {
>  		if (!kasan_check_byte(p))
>  			return NULL;
> -		ks = kfence_ksize(p) ?: __ksize(p);
> +		ks = ksize(p);
>  	} else
>  		ks = 0;
>  
> @@ -1405,8 +1405,10 @@ void kfree_sensitive(const void *p)
>  	void *mem = (void *)p;
>  
>  	ks = ksize(mem);
> -	if (ks)
> +	if (ks) {
> +		kasan_unpoison_range(mem, ks);
>  		memzero_explicit(mem, ks);
> +	}
>  	kfree(mem);
>  }
>  EXPORT_SYMBOL(kfree_sensitive);
> @@ -1415,10 +1417,11 @@ EXPORT_SYMBOL(kfree_sensitive);
>   * ksize - get the actual amount of memory allocated for a given object
>   * @objp: Pointer to the object
>   *
> - * kmalloc may internally round up allocations and return more memory
> + * kmalloc() may internally round up allocations and return more memory
>   * than requested. ksize() can be used to determine the actual amount of
> - * memory allocated. The caller may use this additional memory, even though
> - * a smaller amount of memory was initially specified with the kmalloc call.
> + * allocated memory. The caller may NOT use this additional memory, unless
> + * it calls krealloc(). To avoid an alloc/realloc cycle, callers can use
> + * kmalloc_size_roundup() to find the size of the associated kmalloc bucket.
>   * The caller must guarantee that objp points to a valid object previously
>   * allocated with either kmalloc() or kmem_cache_alloc(). The object
>   * must not be freed during the duration of the call.
> @@ -1427,13 +1430,11 @@ EXPORT_SYMBOL(kfree_sensitive);
>   */
>  size_t ksize(const void *objp)
>  {
> -	size_t size;
> -
>  	/*
> -	 * We need to first check that the pointer to the object is valid, and
> -	 * only then unpoison the memory. The report printed from ksize() is
> -	 * more useful, then when it's printed later when the behaviour could
> -	 * be undefined due to a potential use-after-free or double-free.
> +	 * We need to first check that the pointer to the object is valid.
> +	 * The KASAN report printed from ksize() is more useful, then when
> +	 * it's printed later when the behaviour could be undefined due to
> +	 * a potential use-after-free or double-free.
>  	 *
>  	 * We use kasan_check_byte(), which is supported for the hardware
>  	 * tag-based KASAN mode, unlike kasan_check_read/write().
> @@ -1447,13 +1448,7 @@ size_t ksize(const void *objp)
>  	if (unlikely(ZERO_OR_NULL_PTR(objp)) || !kasan_check_byte(objp))
>  		return 0;
>  
> -	size = kfence_ksize(objp) ?: __ksize(objp);
> -	/*
> -	 * We assume that ksize callers could use whole allocated area,
> -	 * so we need to unpoison this area.
> -	 */
> -	kasan_unpoison_range(objp, size);
> -	return size;
> +	return kfence_ksize(objp) ?: __ksize(objp);
>  }
>  EXPORT_SYMBOL(ksize);
>  

Powered by blists - more mailing lists