When mb-xdp is set and return is XDP_PASS, packet is converted from
xdp_buff to sk_buff with xdp_update_skb_shared_info() in
bnxt_xdp_build_skb().
bnxt_xdp_build_skb() passes incorrect truesize argument to
xdp_update_skb_shared_info().
The truesize is calculated as BNXT_RX_PAGE_SIZE * sinfo->nr_frags but
the skb_shared_info was wiped by napi_build_skb() before.
So it stores skb_shared_info before bnxt_xdp_build_skb() and use it
instead of getting skb_shared_info from xdp_get_shared_info_from_buff().

Splat looks like:
 ------------[ cut here ]------------
 WARNING: CPU: 2 PID: 0 at net/core/skbuff.c:6072 skb_try_coalesce+0x504/0x590
 Modules linked in: xt_nat xt_tcpudp veth af_packet xt_conntrack nft_chain_nat 
xt_MASQUERADE nf_conntrack_netlink xfrm_user xt_addrtype nft_coms
 CPU: 2 UID: 0 PID: 0 Comm: swapper/2 Not tainted 6.14.0-rc2+ #3
 RIP: 0010:skb_try_coalesce+0x504/0x590
 Code: 4b fd ff ff 49 8b 34 24 40 80 e6 40 0f 84 3d fd ff ff 49 8b 74 24 48 40 
f6 c6 01 0f 84 2e fd ff ff 48 8d 4e ff e9 25 fd ff ff <0f> 0b e99
 RSP: 0018:ffffb62c4120caa8 EFLAGS: 00010287
 RAX: 0000000000000003 RBX: ffffb62c4120cb14 RCX: 0000000000000ec0
 RDX: 0000000000001000 RSI: ffffa06e5d7dc000 RDI: 0000000000000003
 RBP: ffffa06e5d7ddec0 R08: ffffa06e6120a800 R09: ffffa06e7a119900
 R10: 0000000000002310 R11: ffffa06e5d7dcec0 R12: ffffe4360575f740
 R13: ffffe43600000000 R14: 0000000000000002 R15: 0000000000000002
 FS:  0000000000000000(0000) GS:ffffa0755f700000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00007f147b76b0f8 CR3: 00000001615d4000 CR4: 00000000007506f0
 PKRU: 55555554
 Call Trace:
  <IRQ>
  ? __warn+0x84/0x130
  ? skb_try_coalesce+0x504/0x590
  ? report_bug+0x18a/0x1a0
  ? handle_bug+0x53/0x90
  ? exc_invalid_op+0x14/0x70
  ? asm_exc_invalid_op+0x16/0x20
  ? skb_try_coalesce+0x504/0x590
  inet_frag_reasm_finish+0x11f/0x2e0
  ip_defrag+0x37a/0x900
  ip_local_deliver+0x51/0x120
  ip_sublist_rcv_finish+0x64/0x70
  ip_sublist_rcv+0x179/0x210
  ip_list_rcv+0xf9/0x130

How to reproduce:
<Node A>
ip link set $interface1 xdp obj xdp_pass.o
ip link set $interface1 mtu 9000 up
ip a a 10.0.0.1/24 dev $interface1
<Node B>
ip link set $interfac2 mtu 9000 up
ip a a 10.0.0.2/24 dev $interface2
ping 10.0.0.1 -s 65000

Following ping.py patch adds xdp-mb-pass case. so ping.py is going to be
able to reproduce this issue.

Fixes: 1dc4c557bfed ("bnxt: adding bnxt_xdp_build_skb to build skb from 
multibuffer xdp_buff")
Signed-off-by: Taehee Yoo <ap420...@gmail.com>
---

v2:
 - Do not use num_frags in the bnxt_xdp_build_skb().

 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 30 +++++++++----------
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c |  7 ++---
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h |  4 +--
 3 files changed, 20 insertions(+), 21 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c 
b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 7b8b5b39c7bb..13c9be49216a 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -2040,6 +2040,7 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct 
bnxt_cp_ring_info *cpr,
        u16 cons, prod, cp_cons = RING_CMP(tmp_raw_cons);
        struct bnxt_sw_rx_bd *rx_buf;
        unsigned int len;
+       struct skb_shared_info sinfo = {0};
        u8 *data_ptr, agg_bufs, cmp_type;
        bool xdp_active = false;
        dma_addr_t dma_addr;
@@ -2166,13 +2167,12 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct 
bnxt_cp_ring_info *cpr,
                                goto oom_next_rx;
                }
                xdp_active = true;
-       }
-
-       if (xdp_active) {
                if (bnxt_rx_xdp(bp, rxr, cons, &xdp, data, &data_ptr, &len, 
event)) {
                        rc = 1;
                        goto next_rx;
                }
+               memcpy(&sinfo, xdp_get_shared_info_from_buff(&xdp),
+                      sizeof(struct skb_shared_info));
        }
 
        if (len <= bp->rx_copybreak) {
@@ -2204,18 +2204,18 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct 
bnxt_cp_ring_info *cpr,
                        goto oom_next_rx;
        }
 
-       if (agg_bufs) {
-               if (!xdp_active) {
-                       skb = bnxt_rx_agg_pages_skb(bp, cpr, skb, cp_cons, 
agg_bufs, false);
-                       if (!skb)
-                               goto oom_next_rx;
-               } else {
-                       skb = bnxt_xdp_build_skb(bp, skb, agg_bufs, 
rxr->page_pool, &xdp, rxcmp1);
-                       if (!skb) {
-                               /* we should be able to free the old skb here */
-                               bnxt_xdp_buff_frags_free(rxr, &xdp);
-                               goto oom_next_rx;
-                       }
+       if (!xdp_active && agg_bufs) {
+               skb = bnxt_rx_agg_pages_skb(bp, cpr, skb, cp_cons, agg_bufs,
+                                           false);
+               if (!skb)
+                       goto oom_next_rx;
+       } else if (xdp_active && xdp_buff_has_frags(&xdp)) {
+               skb = bnxt_xdp_build_skb(bp, skb, &sinfo, rxr->page_pool, &xdp,
+                                        rxcmp1);
+               if (!skb) {
+                       /* we should be able to free the old skb here */
+                       bnxt_xdp_buff_frags_free(rxr, &xdp);
+                       goto oom_next_rx;
                }
        }
 
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c 
b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
index e6c64e4bd66c..77860848e4f9 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
@@ -459,12 +459,11 @@ int bnxt_xdp(struct net_device *dev, struct netdev_bpf 
*xdp)
 }
 
 struct sk_buff *
-bnxt_xdp_build_skb(struct bnxt *bp, struct sk_buff *skb, u8 num_frags,
+bnxt_xdp_build_skb(struct bnxt *bp, struct sk_buff *skb,
+                  struct skb_shared_info *sinfo,
                   struct page_pool *pool, struct xdp_buff *xdp,
                   struct rx_cmp_ext *rxcmp1)
 {
-       struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp);
-
        if (!skb)
                return NULL;
        skb_checksum_none_assert(skb);
@@ -474,7 +473,7 @@ bnxt_xdp_build_skb(struct bnxt *bp, struct sk_buff *skb, u8 
num_frags,
                        skb->csum_level = RX_CMP_ENCAP(rxcmp1);
                }
        }
-       xdp_update_skb_shared_info(skb, num_frags,
+       xdp_update_skb_shared_info(skb, sinfo->nr_frags,
                                   sinfo->xdp_frags_size,
                                   BNXT_RX_PAGE_SIZE * sinfo->nr_frags,
                                   xdp_buff_is_frag_pfmemalloc(xdp));
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h 
b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h
index 0122782400b8..c1974bffafe5 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h
@@ -32,7 +32,7 @@ void bnxt_xdp_buff_init(struct bnxt *bp, struct 
bnxt_rx_ring_info *rxr,
 void bnxt_xdp_buff_frags_free(struct bnxt_rx_ring_info *rxr,
                              struct xdp_buff *xdp);
 struct sk_buff *bnxt_xdp_build_skb(struct bnxt *bp, struct sk_buff *skb,
-                                  u8 num_frags, struct page_pool *pool,
-                                  struct xdp_buff *xdp,
+                                  struct skb_shared_info *sinfo,
+                                  struct page_pool *pool, struct xdp_buff *xdp,
                                   struct rx_cmp_ext *rxcmp1);
 #endif
-- 
2.34.1


Reply via email to