On 9/6/18 10:31 AM, Eric Sandeen wrote: > fscache_set_key() seems to have 2 issues related to the memory > read for hashing in fscache_set_key. > > The first reported was a KASAN error, > > BUG: KASAN: slab-out-of-bounds in fscache_alloc_cookie+0x5b3/0x680 [fscache] > Read of size 4 at addr ffff88084ff056d4 by task mount.nfs/32615 > > also reported by syzbot at https://lkml.org/lkml/2018/7/8/236 > > BUG: KASAN: slab-out-of-bounds in fscache_set_key fs/fscache/cookie.c:120 [inline] > BUG: KASAN: slab-out-of-bounds in fscache_alloc_cookie+0x7a9/0x880 fs/fscache/cookie.c:171 > Read of size 4 at addr ffff8801d3cc8bb4 by task syz-executor907/4466 > > This happens for any index_key_len which is not divisible by 4, and is > larger than the size of the inline key, because the code allocates exactly > index_key_len for the key buffer, but the hashing loop is stepping through > it 4 bytes (u32) at a time in the buf[] array. > > When looking over this, it also appears that the inline key is > insufficiently initialized, zeroing only 3 of the 4 slots. Hence > an index_key_len between 13 and 15 bytes will end up hashing uninitialized > memory because the memcpy only partially fills the last buf[] element. > > Fix this by calculating how many u32 buffers we'll need by using > DIV_ROUND_UP, and memset them all to zero before copying in the index_key, > then using that same count as the hashing index limit. > > Reported-by: syzbot+a95b989b2dde8e806af8@xxxxxxxxxxxxxxxxxxxxxxxxx > Signed-off-by: Eric Sandeen <sandeen@xxxxxxxxxx> > --- Seems like a real bug to me. Any thoughts? -Eric > nb: compile-tested only, sorry. > > diff --git a/fs/fscache/cookie.c b/fs/fscache/cookie.c > index 83bfe04..bc74ae3 100644 > --- a/fs/fscache/cookie.c > +++ b/fs/fscache/cookie.c > @@ -83,7 +83,7 @@ void fscache_cookie_init_once(void *_cookie) > } > > /* > - * Set the index key in a cookie. The cookie struct has space for a 12-byte > + * Set the index key in a cookie. The cookie struct has space for a 16-byte > * key plus length and hash, but if that's not big enough, it's instead a > * pointer to a buffer containing 3 bytes of hash, 1 byte of length and then > * the key data. > @@ -93,22 +93,22 @@ static int fscache_set_key(struct fscache_cookie *cookie, > { > unsigned long long h; > u32 *buf; > + int bufs; > int i; > > cookie->key_len = index_key_len; > + bufs = DIV_ROUND_UP(index_key_len, sizeof(*buf)); > > if (index_key_len > sizeof(cookie->inline_key)) { > - buf = kzalloc(index_key_len, GFP_KERNEL); > + buf = kmalloc_array(bufs, sizeof(*buf), GFP_KERNEL); > if (!buf) > return -ENOMEM; > cookie->key = buf; > } else { > buf = (u32 *)cookie->inline_key; > - buf[0] = 0; > - buf[1] = 0; > - buf[2] = 0; > } > > + memset(buf, 0, bufs * sizeof(*buf)); > memcpy(buf, index_key, index_key_len); > > /* Calculate a hash and combine this with the length in the first word > @@ -116,7 +116,8 @@ static int fscache_set_key(struct fscache_cookie *cookie, > */ > h = (unsigned long)cookie->parent; > h += index_key_len + cookie->type; > - for (i = 0; i < (index_key_len + sizeof(u32) - 1) / sizeof(u32); i++) > + > + for (i = 0; i < bufs; i++) > h += buf[i]; > > cookie->key_hash = h ^ (h >> 32); >
