Re: [PATCH] kasan: fix the missing underflow in memmove and memcpy with CONFIG_KASAN_GENERIC=y

From: Walter Wu
Date: Mon Oct 07 2019 - 08:33:05 EST

On Mon, 2019-10-07 at 14:19 +0200, Dmitry Vyukov wrote:
> On Mon, Oct 7, 2019 at 2:03 PM Walter Wu <walter-zh.wu@xxxxxxxxxxxx> wrote:
> > > > > > > > > > On Mon, Oct 7, 2019 at 10:18 AM Walter Wu <walter-zh.wu@xxxxxxxxxxxx> wrote:
> > > > > > > > > > > The patchsets help to produce KASAN report when size is negative numbers
> > > > > > > > > > > in memory operation function. It is helpful for programmer to solve the
> > > > > > > > > > > undefined behavior issue. Patch 1 based on Dmitry's review and
> > > > > > > > > > > suggestion, patch 2 is a test in order to verify the patch 1.
> > > > > > > > > > >
> > > > > > > > > > > [1]https://bugzilla.kernel.org/show_bug.cgi?id=199341
> > > > > > > > > > > [2]https://lore.kernel.org/linux-arm-kernel/20190927034338.15813-1-walter-zh.wu@xxxxxxxxxxxx/
> > > > > > > > > > >
> > > > > > > > > > > Walter Wu (2):
> > > > > > > > > > > kasan: detect invalid size in memory operation function
> > > > > > > > > > > kasan: add test for invalid size in memmove
> > > > > > > > > > >
> > > > > > > > > > > lib/test_kasan.c | 18 ++++++++++++++++++
> > > > > > > > > > > mm/kasan/common.c | 13 ++++++++-----
> > > > > > > > > > > mm/kasan/generic.c | 5 +++++
> > > > > > > > > > > mm/kasan/generic_report.c | 12 ++++++++++++
> > > > > > > > > > > mm/kasan/tags.c | 5 +++++
> > > > > > > > > > > mm/kasan/tags_report.c | 12 ++++++++++++
> > > > > > > > > > > 6 files changed, 60 insertions(+), 5 deletions(-)
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > commit 5b3b68660b3d420fd2bd792f2d9fd3ccb8877ef7
> > > > > > > > > > > Author: Walter-zh Wu <walter-zh.wu@xxxxxxxxxxxx>
> > > > > > > > > > > Date: Fri Oct 4 18:38:31 2019 +0800
> > > > > > > > > > >
> > > > > > > > > > > kasan: detect invalid size in memory operation function
> > > > > > > > > > >
> > > > > > > > > > > It is an undefined behavior to pass a negative numbers to
> > > > > > > > > > > memset()/memcpy()/memmove()
> > > > > > > > > > > , so need to be detected by KASAN.
> > > > > > > > > > >
> > > > > > > > > > > If size is negative numbers, then it has two reasons to be defined
> > > > > > > > > > > as out-of-bounds bug type.
> > > > > > > > > > > 1) Casting negative numbers to size_t would indeed turn up as a
> > > > > > > > > > > large
> > > > > > > > > > > size_t and its value will be larger than ULONG_MAX/2, so that this
> > > > > > > > > > > can
> > > > > > > > > > > qualify as out-of-bounds.
> > > > > > > > > > > 2) Don't generate new bug type in order to prevent duplicate reports
> > > > > > > > > > > by
> > > > > > > > > > > some systems, e.g. syzbot.
> > > > > > > > > > >
> > > > > > > > > > > KASAN report:
> > > > > > > > > > >
> > > > > > > > > > > BUG: KASAN: out-of-bounds in kmalloc_memmove_invalid_size+0x70/0xa0
> > > > > > > > > > > Read of size 18446744073709551608 at addr ffffff8069660904 by task
> > > > > > > > > > > cat/72
> > > > > > > > > > >
> > > > > > > > > > > CPU: 2 PID: 72 Comm: cat Not tainted
> > > > > > > > > > > 5.4.0-rc1-next-20191004ajb-00001-gdb8af2f372b2-dirty #1
> > > > > > > > > > > Hardware name: linux,dummy-virt (DT)
> > > > > > > > > > > Call trace:
> > > > > > > > > > > dump_backtrace+0x0/0x288
> > > > > > > > > > > show_stack+0x14/0x20
> > > > > > > > > > > dump_stack+0x10c/0x164
> > > > > > > > > > > print_address_description.isra.9+0x68/0x378
> > > > > > > > > > > __kasan_report+0x164/0x1a0
> > > > > > > > > > > kasan_report+0xc/0x18
> > > > > > > > > > > check_memory_region+0x174/0x1d0
> > > > > > > > > > > memmove+0x34/0x88
> > > > > > > > > > > kmalloc_memmove_invalid_size+0x70/0xa0
> > > > > > > > > > >
> > > > > > > > > > > [1] https://bugzilla.kernel.org/show_bug.cgi?id=199341
> > > > > > > > > > >
> > > > > > > > > > > Signed-off-by: Walter Wu <walter-zh.wu@xxxxxxxxxxxx>
> > > > > > > > > > > Reported -by: Dmitry Vyukov <dvyukov@xxxxxxxxxx>
> > > > > > > > > > > Suggested-by: Dmitry Vyukov <dvyukov@xxxxxxxxxx>
> > > > > > > > > > >
> > > > > > > > > > > diff --git a/mm/kasan/common.c b/mm/kasan/common.c
> > > > > > > > > > > index 6814d6d6a023..6ef0abd27f06 100644
> > > > > > > > > > > --- a/mm/kasan/common.c
> > > > > > > > > > > +++ b/mm/kasan/common.c
> > > > > > > > > > > @@ -102,7 +102,8 @@ EXPORT_SYMBOL(__kasan_check_write);
> > > > > > > > > > > #undef memset
> > > > > > > > > > > void *memset(void *addr, int c, size_t len)
> > > > > > > > > > > {
> > > > > > > > > > > - check_memory_region((unsigned long)addr, len, true, _RET_IP_);
> > > > > > > > > > > + if (!check_memory_region((unsigned long)addr, len, true, _RET_IP_))
> > > > > > > > > > > + return NULL;
> > > > > > > > > > >
> > > > > > > > > > > return __memset(addr, c, len);
> > > > > > > > > > > }
> > > > > > > > > > > @@ -110,8 +111,9 @@ void *memset(void *addr, int c, size_t len)
> > > > > > > > > > > #undef memmove
> > > > > > > > > > > void *memmove(void *dest, const void *src, size_t len)
> > > > > > > > > > > {
> > > > > > > > > > > - check_memory_region((unsigned long)src, len, false, _RET_IP_);
> > > > > > > > > > > - check_memory_region((unsigned long)dest, len, true, _RET_IP_);
> > > > > > > > > > > + if (!check_memory_region((unsigned long)src, len, false, _RET_IP_) ||
> > > > > > > > > > > + !check_memory_region((unsigned long)dest, len, true, _RET_IP_))
> > > > > > > > > > > + return NULL;
> > > > > > > > > > >
> > > > > > > > > > > return __memmove(dest, src, len);
> > > > > > > > > > > }
> > > > > > > > > > > @@ -119,8 +121,9 @@ void *memmove(void *dest, const void *src, size_t
> > > > > > > > > > > len)
> > > > > > > > > > > #undef memcpy
> > > > > > > > > > > void *memcpy(void *dest, const void *src, size_t len)
> > > > > > > > > > > {
> > > > > > > > > > > - check_memory_region((unsigned long)src, len, false, _RET_IP_);
> > > > > > > > > > > - check_memory_region((unsigned long)dest, len, true, _RET_IP_);
> > > > > > > > > > > + if (!check_memory_region((unsigned long)src, len, false, _RET_IP_) ||
> > > > > > > > > > > + !check_memory_region((unsigned long)dest, len, true, _RET_IP_))
> > > > > > > > > > > + return NULL;
> > > > > > > > > > >
> > > > > > > > > > > return __memcpy(dest, src, len);
> > > > > > > > > > > }
> > > > > > > > > > > diff --git a/mm/kasan/generic.c b/mm/kasan/generic.c
> > > > > > > > > > > index 616f9dd82d12..02148a317d27 100644
> > > > > > > > > > > --- a/mm/kasan/generic.c
> > > > > > > > > > > +++ b/mm/kasan/generic.c
> > > > > > > > > > > @@ -173,6 +173,11 @@ static __always_inline bool
> > > > > > > > > > > check_memory_region_inline(unsigned long addr,
> > > > > > > > > > > if (unlikely(size == 0))
> > > > > > > > > > > return true;
> > > > > > > > > > >
> > > > > > > > > > > + if (unlikely((long)size < 0)) {
> > > > > > > > > > > + kasan_report(addr, size, write, ret_ip);
> > > > > > > > > > > + return false;
> > > > > > > > > > > + }
> > > > > > > > > > > +
> > > > > > > > > > > if (unlikely((void *)addr <
> > > > > > > > > > > kasan_shadow_to_mem((void *)KASAN_SHADOW_START))) {
> > > > > > > > > > > kasan_report(addr, size, write, ret_ip);
> > > > > > > > > > > diff --git a/mm/kasan/generic_report.c b/mm/kasan/generic_report.c
> > > > > > > > > > > index 36c645939bc9..ed0eb94cb811 100644
> > > > > > > > > > > --- a/mm/kasan/generic_report.c
> > > > > > > > > > > +++ b/mm/kasan/generic_report.c
> > > > > > > > > > > @@ -107,6 +107,18 @@ static const char *get_wild_bug_type(struct
> > > > > > > > > > > kasan_access_info *info)
> > > > > > > > > > >
> > > > > > > > > > > const char *get_bug_type(struct kasan_access_info *info)
> > > > > > > > > > > {
> > > > > > > > > > > + /*
> > > > > > > > > > > + * If access_size is negative numbers, then it has two reasons
> > > > > > > > > > > + * to be defined as out-of-bounds bug type.
> > > > > > > > > > > + * 1) Casting negative numbers to size_t would indeed turn up as
> > > > > > > > > > > + * a 'large' size_t and its value will be larger than ULONG_MAX/2,
> > > > > > > > > > > + * so that this can qualify as out-of-bounds.
> > > > > > > > > > > + * 2) Don't generate new bug type in order to prevent duplicate
> > > > > > > > > > > reports
> > > > > > > > > > > + * by some systems, e.g. syzbot.
> > > > > > > > > > > + */
> > > > > > > > > > > + if ((long)info->access_size < 0)
> > > > > > > > > > > + return "out-of-bounds";
> > > > > > > > > >
> > > > > > > > > > "out-of-bounds" is the _least_ frequent KASAN bug type. It won't
> > > > > > > > > > prevent duplicates. "heap-out-of-bounds" is the frequent one.
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > /*
> > > > > > > > > * If access_size is negative numbers, then it has two reasons
> > > > > > > > > * to be defined as out-of-bounds bug type.
> > > > > > > > > * 1) Casting negative numbers to size_t would indeed turn up as
> > > > > > > > > * a "large" size_t and its value will be larger than ULONG_MAX/2,
> > > > > > > > > * so that this can qualify as out-of-bounds.
> > > > > > > > > * 2) Don't generate new bug type in order to prevent duplicate
> > > > > > > > > reports
> > > > > > > > > * by some systems, e.g. syzbot. "out-of-bounds" is the _least_
> > > > > > > > > frequent KASAN bug type.
> > > > > > > > > * It won't prevent duplicates. "heap-out-of-bounds" is the
> > > > > > > > > frequent one.
> > > > > > > > > */
> > > > > > > > >
> > > > > > > > > We directly add it into the comment.
> > > > > > > >
> > > > > > > >
> > > > > > > > OK, let's start from the beginning: why do you return "out-of-bounds" here?
> > > > > > > >
> > > > > > > Uh, comment 1 and 2 should explain it. :)
> > > > > >
> > > > > > The comment says it will cause duplicate reports. It does not explain
> > > > > > why you want syzbot to produce duplicate reports and spam kernel
> > > > > > developers... So why do you want that?
> > > > > >
> > > > > We don't generate new bug type in order to prevent duplicate by some
> > > > > systems, e.g. syzbot. Is it right? If yes, then it should not have
> > > > > duplicate report.
> > > > >
> > > > Sorry, because we don't generate new bug type. it should be duplicate
> > > > report(only one report which may be oob or size invlid),
> > > > the duplicate report goal is that invalid size is oob issue, too.
> > > >
> > > > I would not introduce a new bug type.
> > > > These are parsed and used by some systems, e.g. syzbot. If size is
> > > > user-controllable, then a new bug type for this will mean 2 bug
> > > > reports.
> > >
> > > To prevent duplicates, the new crash title must not just match _any_
> > > crash title that kernel can potentially produce. It must match exactly
> > > the crash that kernel produces for this bug on other input data.
> > >
> > > Consider, userspace passes size=123, KASAN produces "heap-out-of-bounds in foo".
> > > Now userspace passes size=-1 and KASAN produces "invalid-size in foo".
> > > This will be a duplicate bug report.
> > > Now if KASAN will produce "out-of-bounds in foo", it will also lead to
> > > a duplicate report.
> > > Only iff KASAN will produce "heap-out-of-bounds in foo" for size=-1,
> > > it will not lead to a duplicate report.
> >
> > I think it is not easy to avoid the duplicate report(mentioned above).
> > As far as my knowledge is concerned, KASAN is memory corruption detector
> > in kernel space, it should only detect memory corruption and don't
> > distinguish whether it is passed by userspace. if we want to do, then we
> > may need to parse backtrace to check if it has copy_form_user() or other
> > function?
>
> My idea was just to always print "heap-out-of-bounds" and don't
> differentiate if the size come from userspace or not.

Got it.
Would you have any other concern about this patch?