Received-SPF: pass (google.com: domain of linux-kernel+bounces-94588-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240306190822.390086-1-dtatulea@nvidia.com>
In-Reply-To: <20240306190822.390086-1-dtatulea@nvidia.com>
From: Mina Almasry <almasrymina@google.com>
Date: Wed, 6 Mar 2024 13:15:07 -0800
Message-ID: <CAHS8izPinALpWgq_3m7C_VCTHwOMiAVgUXkAY=WK4ZO2DCtXKg@mail.gmail.com>
Subject: Re: [RFC PATCH v2] net: esp: fix bad handling of pages from page_pool
To: Dragos Tatulea <dtatulea@nvidia.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>, Herbert Xu <herbert@gondor.apana.org.au>, 
	"David S. Miller" <davem@davemloft.net>, David Ahern <dsahern@kernel.org>, 
	Eric Dumazet <edumazet@google.com>, Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>, 
	leonro@nvidia.com, gal@nvidia.com, 
	"Anatoli N . Chechelnickiy" <Anatoli.Chechelnickiy@m.interpipe.biz>, Ian Kumlien <ian.kumlien@gmail.com>, 
	linux-kernel@vger.kernel.org, netdev@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Wed, Mar 6, 2024 at 11:08=E2=80=AFAM Dragos Tatulea <dtatulea@nvidia.com=
> wrote:
>
> When the skb is reorganized during esp_output (!esp->inline), the pages
> coming from the original skb fragments are supposed to be released back
> to the system through put_page. But if the skb fragment pages are
> originating from a page_pool, calling put_page on them will trigger a
> page_pool leak which will eventually result in a crash.
>
> This leak can be easily observed when using CONFIG_DEBUG_VM and doing
> ipsec + gre (non offloaded) forwarding:
>
>   BUG: Bad page state in process ksoftirqd/16  pfn:1451b6
>   page:00000000de2b8d32 refcount:0 mapcount:0 mapping:0000000000000000 in=
dex:0x1451b6000 pfn:0x1451b6
>   flags: 0x200000000000000(node=3D0|zone=3D2)
>   page_type: 0xffffffff()
>   raw: 0200000000000000 dead000000000040 ffff88810d23c000 000000000000000=
0
>   raw: 00000001451b6000 0000000000000001 00000000ffffffff 000000000000000=
0
>   page dumped because: page_pool leak
>   Modules linked in: ip_gre gre mlx5_ib mlx5_core xt_conntrack xt_MASQUER=
ADE nf_conntrack_netlink nfnetlink iptable_nat nf_nat xt_addrtype br_netfil=
ter rpcrdma rdma_ucm ib_iser libiscsi scsi_transport_iscsi ib_umad rdma_cm =
ib_ipoib iw_cm ib_cm ib_uverbs ib_core overlay zram zsmalloc fuse [last unl=
oaded: mlx5_core]
>   CPU: 16 PID: 96 Comm: ksoftirqd/16 Not tainted 6.8.0-rc4+ #22
>   Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-g=
f21b5a4aeb02-prebuilt.qemu.org 04/01/2014
>   Call Trace:
>    <TASK>
>    dump_stack_lvl+0x36/0x50
>    bad_page+0x70/0xf0
>    free_unref_page_prepare+0x27a/0x460
>    free_unref_page+0x38/0x120
>    esp_ssg_unref.isra.0+0x15f/0x200
>    esp_output_tail+0x66d/0x780
>    esp_xmit+0x2c5/0x360
>    validate_xmit_xfrm+0x313/0x370
>    ? validate_xmit_skb+0x1d/0x330
>    validate_xmit_skb_list+0x4c/0x70
>    sch_direct_xmit+0x23e/0x350
>    __dev_queue_xmit+0x337/0xba0
>    ? nf_hook_slow+0x3f/0xd0
>    ip_finish_output2+0x25e/0x580
>    iptunnel_xmit+0x19b/0x240
>    ip_tunnel_xmit+0x5fb/0xb60
>    ipgre_xmit+0x14d/0x280 [ip_gre]
>    dev_hard_start_xmit+0xc3/0x1c0
>    __dev_queue_xmit+0x208/0xba0
>    ? nf_hook_slow+0x3f/0xd0
>    ip_finish_output2+0x1ca/0x580
>    ip_sublist_rcv_finish+0x32/0x40
>    ip_sublist_rcv+0x1b2/0x1f0
>    ? ip_rcv_finish_core.constprop.0+0x460/0x460
>    ip_list_rcv+0x103/0x130
>    __netif_receive_skb_list_core+0x181/0x1e0
>    netif_receive_skb_list_internal+0x1b3/0x2c0
>    napi_gro_receive+0xc8/0x200
>    gro_cell_poll+0x52/0x90
>    __napi_poll+0x25/0x1a0
>    net_rx_action+0x28e/0x300
>    __do_softirq+0xc3/0x276
>    ? sort_range+0x20/0x20
>    run_ksoftirqd+0x1e/0x30
>    smpboot_thread_fn+0xa6/0x130
>    kthread+0xcd/0x100
>    ? kthread_complete_and_exit+0x20/0x20
>    ret_from_fork+0x31/0x50
>    ? kthread_complete_and_exit+0x20/0x20
>    ret_from_fork_asm+0x11/0x20
>    </TASK>
>
> The suggested fix is to use a new wrapper that covers page refcounting
> for page_pool pages as well.
>
> Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
> Reported-by: Anatoli N.Chechelnickiy <Anatoli.Chechelnickiy@m.interpipe.b=
iz>
> Reported-by: Ian Kumlien <ian.kumlien@gmail.com>

Maybe you want to Cc: stable on this & add fixes tag? Or maybe you
skipped that because it's still an RFC patch.

I think it maybe also needs to be  [PATCH net] or [RFC PATCH net]
subject to designate to net and not net-next.

Looks good to me with very minor comment so, FWIW,

Reviewed-by: Mina Almasry <almasrymina@google.com>

> ---
>
> Changes in v2:
> - Added napi_page_unref api based on discussion in v1 [0].
>
> [0]https://lore.kernel.org/netdev/CAHS8izOoO-EovwMwAm9tLYetwikNPxC0FKyVGu=
1TPJWSz4bGoA@mail.gmail.com/T/#t
>
> ---
>  include/linux/skbuff.h | 10 +++++++---
>  net/ipv4/esp4.c        | 16 ++++++++++------
>  net/ipv6/esp6.c        | 16 ++++++++++------
>  3 files changed, 27 insertions(+), 15 deletions(-)
>
> diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
> index 696e7680656f..009603db2a43 100644
> --- a/include/linux/skbuff.h
> +++ b/include/linux/skbuff.h
> @@ -3453,10 +3453,8 @@ int skb_cow_data_for_xdp(struct page_pool *pool, s=
truct sk_buff **pskb,
>  bool napi_pp_put_page(struct page *page, bool napi_safe);
>
>  static inline void
> -napi_frag_unref(skb_frag_t *frag, bool recycle, bool napi_safe)
> +napi_page_unref(struct page *page, bool recycle, bool napi_safe)
>  {
> -       struct page *page =3D skb_frag_page(frag);
> -
>  #ifdef CONFIG_PAGE_POOL
>         if (recycle && napi_pp_put_page(page, napi_safe))

Ah, it looks like we don't need is_pp_page() after all because
napi_pp_put_page() already checks for that internally. Good find, much
better than what I suggested.

>                 return;
> @@ -3464,6 +3462,12 @@ napi_frag_unref(skb_frag_t *frag, bool recycle, bo=
ol napi_safe)
>         put_page(page);
>  }
>
> +static inline void
> +napi_frag_unref(skb_frag_t *frag, bool recycle, bool napi_safe)
> +{
> +       napi_page_unref(skb_frag_page(frag), recycle, napi_safe);
> +}
> +
>  /**
>   * __skb_frag_unref - release a reference on a paged fragment.
>   * @frag: the paged fragment
> diff --git a/net/ipv4/esp4.c b/net/ipv4/esp4.c
> index 4dd9e5040672..3126aeb0588d 100644
> --- a/net/ipv4/esp4.c
> +++ b/net/ipv4/esp4.c
> @@ -95,7 +95,7 @@ static inline struct scatterlist *esp_req_sg(struct cry=
pto_aead *aead,
>                              __alignof__(struct scatterlist));
>  }
>
> -static void esp_ssg_unref(struct xfrm_state *x, void *tmp)
> +static void esp_ssg_unref(struct xfrm_state *x, void *tmp, struct sk_buf=
f *skb)
>  {
>         struct crypto_aead *aead =3D x->data;
>         int extralen =3D 0;
> @@ -112,9 +112,13 @@ static void esp_ssg_unref(struct xfrm_state *x, void=
 *tmp)
>         /* Unref skb_frag_pages in the src scatterlist if necessary.
>          * Skip the first sg which comes from skb->data.
>          */
> -       if (req->src !=3D req->dst)
> -               for (sg =3D sg_next(req->src); sg; sg =3D sg_next(sg))
> -                       put_page(sg_page(sg));
> +       if (req->src !=3D req->dst) {
> +               for (sg =3D sg_next(req->src); sg; sg =3D sg_next(sg)) {
> +                       struct page *page =3D sg_page(sg);
> +
> +                       napi_page_unref(page, skb->pp_recycle, false);

[minor annoying nit, feel free to ignore]: I would use sg_page(sg)
inline here and below, since page is not reused elsewhere. But I'm
fine either way.

> +               }
> +       }
>  }
>
>  #ifdef CONFIG_INET_ESPINTCP
> @@ -260,7 +264,7 @@ static void esp_output_done(void *data, int err)
>         }
>
>         tmp =3D ESP_SKB_CB(skb)->tmp;
> -       esp_ssg_unref(x, tmp);
> +       esp_ssg_unref(x, tmp, skb);
>         kfree(tmp);
>
>         if (xo && (xo->flags & XFRM_DEV_RESUME)) {
> @@ -639,7 +643,7 @@ int esp_output_tail(struct xfrm_state *x, struct sk_b=
uff *skb, struct esp_info *
>         }
>
>         if (sg !=3D dsg)
> -               esp_ssg_unref(x, tmp);
> +               esp_ssg_unref(x, tmp, skb);
>
>         if (!err && x->encap && x->encap->encap_type =3D=3D TCP_ENCAP_ESP=
INTCP)
>                 err =3D esp_output_tail_tcp(x, skb);
> diff --git a/net/ipv6/esp6.c b/net/ipv6/esp6.c
> index 6e6efe026cdc..7bbe19e74cf0 100644
> --- a/net/ipv6/esp6.c
> +++ b/net/ipv6/esp6.c
> @@ -112,7 +112,7 @@ static inline struct scatterlist *esp_req_sg(struct c=
rypto_aead *aead,
>                              __alignof__(struct scatterlist));
>  }
>
> -static void esp_ssg_unref(struct xfrm_state *x, void *tmp)
> +static void esp_ssg_unref(struct xfrm_state *x, void *tmp, struct sk_buf=
f *skb)
>  {
>         struct crypto_aead *aead =3D x->data;
>         int extralen =3D 0;
> @@ -129,9 +129,13 @@ static void esp_ssg_unref(struct xfrm_state *x, void=
 *tmp)
>         /* Unref skb_frag_pages in the src scatterlist if necessary.
>          * Skip the first sg which comes from skb->data.
>          */
> -       if (req->src !=3D req->dst)
> -               for (sg =3D sg_next(req->src); sg; sg =3D sg_next(sg))
> -                       put_page(sg_page(sg));
> +       if (req->src !=3D req->dst) {
> +               for (sg =3D sg_next(req->src); sg; sg =3D sg_next(sg)) {
> +                       struct page *page =3D sg_page(sg);
> +
> +                       napi_page_unref(page, skb->pp_recycle, false);
> +               }
> +       }
>  }
>
>  #ifdef CONFIG_INET6_ESPINTCP
> @@ -294,7 +298,7 @@ static void esp_output_done(void *data, int err)
>         }
>
>         tmp =3D ESP_SKB_CB(skb)->tmp;
> -       esp_ssg_unref(x, tmp);
> +       esp_ssg_unref(x, tmp, skb);
>         kfree(tmp);
>
>         esp_output_encap_csum(skb);
> @@ -677,7 +681,7 @@ int esp6_output_tail(struct xfrm_state *x, struct sk_=
buff *skb, struct esp_info
>         }
>
>         if (sg !=3D dsg)
> -               esp_ssg_unref(x, tmp);
> +               esp_ssg_unref(x, tmp, skb);
>
>         if (!err && x->encap && x->encap->encap_type =3D=3D TCP_ENCAP_ESP=
INTCP)
>                 err =3D esp_output_tail_tcp(x, skb);
> --
> 2.42.0
>


--=20
Thanks,
Mina