Re: [PATCH net-next v3] skbuff: Introduce slab_build_skb()
From: Feng Tang
Date: Thu Dec 08 2022 - 05:23:22 EST
On Thu, Dec 08, 2022 at 09:13:41AM +0100, Vlastimil Babka wrote:
> On 12/8/22 07:02, Kees Cook wrote:
> > syzkaller reported:
> >
> > BUG: KASAN: slab-out-of-bounds in __build_skb_around+0x235/0x340 net/core/skbuff.c:294
> > Write of size 32 at addr ffff88802aa172c0 by task syz-executor413/5295
> >
> > For bpf_prog_test_run_skb(), which uses a kmalloc()ed buffer passed to
> > build_skb().
> >
> > When build_skb() is passed a frag_size of 0, it means the buffer came
> > from kmalloc. In these cases, ksize() is used to find its actual size,
> > but since the allocation may not have been made to that size, actually
> > perform the krealloc() call so that all the associated buffer size
> > checking will be correctly notified (and use the "new" pointer so that
> > compiler hinting works correctly). Split this logic out into a new
> > interface, slab_build_skb(), but leave the original 0 checking for now
> > to catch any stragglers.
> >
> > Reported-by: syzbot+fda18eaa8c12534ccb3b@xxxxxxxxxxxxxxxxxxxxxxxxx
> > Link: https://groups.google.com/g/syzkaller-bugs/c/UnIKxTtU5-0/m/-wbXinkgAQAJ
> > Fixes: 38931d8989b5 ("mm: Make ksize() a reporting-only function")
> > Cc: Jakub Kicinski <kuba@xxxxxxxxxx>
> > Cc: Eric Dumazet <edumazet@xxxxxxxxxx>
> > Cc: "David S. Miller" <davem@xxxxxxxxxxxxx>
> > Cc: Paolo Abeni <pabeni@xxxxxxxxxx>
> > Cc: Pavel Begunkov <asml.silence@xxxxxxxxx>
> > Cc: pepsipu <soopthegoop@xxxxxxxxx>
> > Cc: syzbot+fda18eaa8c12534ccb3b@xxxxxxxxxxxxxxxxxxxxxxxxx
> > Cc: Vlastimil Babka <vbabka@xxxxxxx>
> > Cc: kasan-dev <kasan-dev@xxxxxxxxxxxxxxxx>
> > Cc: Andrii Nakryiko <andrii@xxxxxxxxxx>
> > Cc: ast@xxxxxxxxxx
> > Cc: bpf <bpf@xxxxxxxxxxxxxxx>
> > Cc: Daniel Borkmann <daniel@xxxxxxxxxxxxx>
> > Cc: Hao Luo <haoluo@xxxxxxxxxx>
> > Cc: Jesper Dangaard Brouer <hawk@xxxxxxxxxx>
> > Cc: John Fastabend <john.fastabend@xxxxxxxxx>
> > Cc: jolsa@xxxxxxxxxx
> > Cc: KP Singh <kpsingh@xxxxxxxxxx>
> > Cc: martin.lau@xxxxxxxxx
> > Cc: Stanislav Fomichev <sdf@xxxxxxxxxx>
> > Cc: song@xxxxxxxxxx
> > Cc: Yonghong Song <yhs@xxxxxx>
> > Cc: netdev@xxxxxxxxxxxxxxx
> > Cc: LKML <linux-kernel@xxxxxxxxxxxxxxx>
> > Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
> > ---
> > v3:
> > - make sure "resized" is passed back so compiler hints survive
> > - update kerndoc (kuba)
> > v2: https://lore.kernel.org/lkml/20221208000209.gonna.368-kees@xxxxxxxxxx
> > v1: https://lore.kernel.org/netdev/20221206231659.never.929-kees@xxxxxxxxxx/
> > ---
> > drivers/net/ethernet/broadcom/bnx2.c | 2 +-
> > drivers/net/ethernet/qlogic/qed/qed_ll2.c | 2 +-
> > include/linux/skbuff.h | 1 +
> > net/bpf/test_run.c | 2 +-
> > net/core/skbuff.c | 70 ++++++++++++++++++++---
> > 5 files changed, 66 insertions(+), 11 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/broadcom/bnx2.c b/drivers/net/ethernet/broadcom/bnx2.c
> > index fec57f1982c8..b2230a4a2086 100644
> > --- a/drivers/net/ethernet/broadcom/bnx2.c
> > +++ b/drivers/net/ethernet/broadcom/bnx2.c
> > @@ -3045,7 +3045,7 @@ bnx2_rx_skb(struct bnx2 *bp, struct bnx2_rx_ring_info *rxr, u8 *data,
> >
> > dma_unmap_single(&bp->pdev->dev, dma_addr, bp->rx_buf_use_size,
> > DMA_FROM_DEVICE);
> > - skb = build_skb(data, 0);
> > + skb = slab_build_skb(data);
> > if (!skb) {
> > kfree(data);
> > goto error;
> > diff --git a/drivers/net/ethernet/qlogic/qed/qed_ll2.c b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
> > index ed274f033626..e5116a86cfbc 100644
> > --- a/drivers/net/ethernet/qlogic/qed/qed_ll2.c
> > +++ b/drivers/net/ethernet/qlogic/qed/qed_ll2.c
> > @@ -200,7 +200,7 @@ static void qed_ll2b_complete_rx_packet(void *cxt,
> > dma_unmap_single(&cdev->pdev->dev, buffer->phys_addr,
> > cdev->ll2->rx_size, DMA_FROM_DEVICE);
> >
> > - skb = build_skb(buffer->data, 0);
> > + skb = slab_build_skb(buffer->data);
> > if (!skb) {
> > DP_INFO(cdev, "Failed to build SKB\n");
> > kfree(buffer->data);
> > diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
> > index 7be5bb4c94b6..0b391b635430 100644
> > --- a/include/linux/skbuff.h
> > +++ b/include/linux/skbuff.h
> > @@ -1253,6 +1253,7 @@ struct sk_buff *build_skb_around(struct sk_buff *skb,
> > void skb_attempt_defer_free(struct sk_buff *skb);
> >
> > struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
> > +struct sk_buff *slab_build_skb(void *data);
> >
> > /**
> > * alloc_skb - allocate a network buffer
> > diff --git a/net/bpf/test_run.c b/net/bpf/test_run.c
> > index 13d578ce2a09..611b1f4082cf 100644
> > --- a/net/bpf/test_run.c
> > +++ b/net/bpf/test_run.c
> > @@ -1130,7 +1130,7 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
> > }
> > sock_init_data(NULL, sk);
> >
> > - skb = build_skb(data, 0);
> > + skb = slab_build_skb(data);
> > if (!skb) {
> > kfree(data);
> > kfree(ctx);
> > diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> > index 1d9719e72f9d..ae5a6f7db37b 100644
> > --- a/net/core/skbuff.c
> > +++ b/net/core/skbuff.c
> > @@ -269,12 +269,10 @@ static struct sk_buff *napi_skb_cache_get(void)
> > return skb;
> > }
> >
> > -/* Caller must provide SKB that is memset cleared */
> > -static void __build_skb_around(struct sk_buff *skb, void *data,
> > - unsigned int frag_size)
> > +static inline void __finalize_skb_around(struct sk_buff *skb, void *data,
> > + unsigned int size)
> > {
> > struct skb_shared_info *shinfo;
> > - unsigned int size = frag_size ? : ksize(data);
> >
> > size -= SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
> >
> > @@ -296,15 +294,71 @@ static void __build_skb_around(struct sk_buff *skb, void *data,
> > skb_set_kcov_handle(skb, kcov_common_handle());
> > }
> >
> > +static inline void *__slab_build_skb(struct sk_buff *skb, void *data,
> > + unsigned int *size)
> > +{
> > + void *resized;
> > +
> > + /* Must find the allocation size (and grow it to match). */
> > + *size = ksize(data);
> > + /* krealloc() will immediately return "data" when
> > + * "ksize(data)" is requested: it is the existing upper
> > + * bounds. As a result, GFP_ATOMIC will be ignored. Note
> > + * that this "new" pointer needs to be passed back to the
> > + * caller for use so the __alloc_size hinting will be
> > + * tracked correctly.
> > + */
> > + resized = krealloc(data, *size, GFP_ATOMIC);
>
> Hmm, I just realized, this trick will probably break the new kmalloc size
> tracking from Feng Tang (CC'd)? We need to make krealloc() update the stored
> size, right? And even worse if slab_debug redzoning is enabled and after
> commit 946fa0dbf2d8 ("mm/slub: extend redzone check to extra allocated
> kmalloc space than requested") where the lack of update will result in
> redzone check failures.
I think it's still safe, as currently we skip the kmalloc redzone check
by calling skip_orig_size_check() inside __ksize(). But as we have plan
to remove this skip_orig_size_check() after all ksize() usage has been
sanitized, we need to cover this krealloc() case.
Thanks,
Feng