net/tap: fix eBPF handling of non-RSS flows

The eBPF classifier (section "cls_q" in tap_bpf_program.c) is tracing
marked packets in which skb->cb[1] contains an RSS queue number, and
redirects those packets to the matched queue.
It is expected that skb->cb[1] has been previously set with a valid RSS
queue number during an eBPF action (section "l3_l4" in tap_bpf_program.c).
However, for non-RSS flows, skb->cb[1] may contain a random unset value,
which could falsely be interpreted as a valid RSS queue.
To avoid this potential error, tap_bpf_program.c has been updated as
follows:
1. After calculating the RSS queue number, it is added a unique offset in
order to uniquely identify it as a valid RSS queue number.
2. After matching an RSS queue to a packet, skb->cb[1] is set to 0.

Fixes: cdc07e83bb24 ("net/tap: add eBPF program file")
Fixes: aabe70df73a3 ("net/tap: add eBPF bytes code")

Signed-off-by: Ophir Munk <ophirmu@mellanox.com>
Acked-by: Pascal Mazon <pascal.mazon@6wind.com>

diff --git a/drivers/net/tap/tap_bpf_insns.h b/drivers/net/tap/tap_bpf_insns.h
index c406f78..89873b6 100644 (file)
--- a/drivers/net/tap/tap_bpf_insns.h
+++ b/drivers/net/tap/tap_bpf_insns.h
@@ -6,17 +6,20 @@
 
 /* bpf_insn array matching cls_q section. See tap_bpf_program.c file */
 struct bpf_insn cls_q_insns[] = {
-       {0x61,    1,    1,       52, 0x00000000},
-       {0x18,    2,    0,        0, 0xdeadbeef},
+       {0x61,    2,    1,       52, 0x00000000},
+       {0x18,    3,    0,        0, 0xdeadbeef},
        {0x00,    0,    0,        0, 0x00000000},
-       {0x63,   10,    2,       -4, 0x00000000},
-       {0x61,    2,   10,       -4, 0x00000000},
-       {0x07,    2,    0,        0, 0x00000001},
-       {0x67,    2,    0,        0, 0x00000020},
-       {0x77,    2,    0,        0, 0x00000020},
-       {0xb7,    0,    0,        0, 0xffffffff},
-       {0x1d,    1,    2,        1, 0x00000000},
+       {0x63,   10,    3,       -4, 0x00000000},
        {0xb7,    0,    0,        0, 0x00000000},
+       {0x61,    3,   10,       -4, 0x00000000},
+       {0x07,    3,    0,        0, 0x7cafe800},
+       {0x67,    3,    0,        0, 0x00000020},
+       {0x77,    3,    0,        0, 0x00000020},
+       {0x5d,    2,    3,        4, 0x00000000},
+       {0xb7,    2,    0,        0, 0x00000000},
+       {0x63,    1,    2,       52, 0x00000000},
+       {0x18,    0,    0,        0, 0xffffffff},
+       {0x00,    0,    0,        0, 0x00000000},
        {0x95,    0,    0,        0, 0x00000000},
 };
 
@@ -1685,7 +1688,7 @@ struct bpf_insn l3_l4_hash_insns[] = {
        {0x4f,    3,    2,        0, 0x00000000},
        {0x67,    3,    0,        0, 0x00000010},
        {0x4f,    3,    1,        0, 0x00000000},
-       {0x07,    3,    0,        0, 0x00000001},
+       {0x07,    3,    0,        0, 0x7cafe800},
        {0x63,    5,    3,       52, 0x00000000},
        {0xb7,    7,    0,        0, 0x00000001},
        {0xbf,    0,    7,        0, 0x00000000},

diff --git a/drivers/net/tap/tap_bpf_program.c b/drivers/net/tap/tap_bpf_program.c
index 848c50b..8abb3b7 100644 (file)
--- a/drivers/net/tap/tap_bpf_program.c
+++ b/drivers/net/tap/tap_bpf_program.c
@@ -28,10 +28,10 @@
                ((b) & 0xff))
 
 /*
- * The queue number is offset by 1, to distinguish packets that have
- * gone through this rule (skb->cb[1] != 0) from others.
+ * The queue number is offset by a unique QUEUE_OFFSET, to distinguish
+ * packets that have gone through this rule (skb->cb[1] != 0) from others.
  */
-#define QUEUE_OFFSET           1
+#define QUEUE_OFFSET           0x7cafe800
 #define PIN_GLOBAL_NS          2
 
 #define KEY_IDX                        0
@@ -63,6 +63,9 @@ match_q(struct __sk_buff *skb)
 
        if (queue != match_queue)
                return TC_ACT_OK;
+
+       /* queue match */
+       skb->cb[1] = 0;
        return TC_ACT_UNSPEC;
 }