gve: add optional metadata descriptor type GVE_TXD_MTD
Allow drivers to pass metadata along with packet data to the device.
Introduce a new metadata descriptor type
* GVE_TXD_MTD
This descriptor is optional. If present it immediate follows the
packet descriptor and precedes the segment descriptor.
This descriptor may be repeated. Multiple metadata descriptors may
follow. There are no immediate uses for this, this is for future
proofing. At present devices allow only 1 MTD descriptor.
The lower four bits of the type_flags field encode GVE_TXD_MTD.
The upper four bits of the type_flags field encodes a *sub*type.
Introduce one such metadata descriptor subtype
* GVE_MTD_SUBTYPE_PATH
This shares path information with the device for network failure
discovery and robust response:
Linux derives ipv6 flowlabel and ECMP multipath from sk->sk_txhash,
and updates this field on error with sk_rethink_txhash. Allow the host
stack to do the same. Pass the tx_hash value if set. Also communicate
whether the path hash is set, or more exactly, what its type is. Define
two common types
GVE_MTD_PATH_HASH_NONE
GVE_MTD_PATH_HASH_L4
Concrete examples of error conditions that are resolved are
mentioned in the commits that add sk_rethink_txhash calls. Such as
commit 7788174e8726 ("tcp: change IPv6 flow-label upon receiving
spurious retransmission").
Experimental results mirror what the theory suggests: where IPv6
FlowLabel is included in path selection (e.g., LAG/ECMP), flowlabel
rotation on TCP timeout avoids the vast majority of TCP disconnects
that would otherwise have occurred during link failures in long-haul
backbones, when an alternative path is available.
Rotation can be applied to various bad connection signals, such as
timeouts and spurious retransmissions. In aggregate, such flow level
signals can help locate network issues. Define initial common states:
GVE_MTD_PATH_STATE_DEFAULT
GVE_MTD_PATH_STATE_TIMEOUT
GVE_MTD_PATH_STATE_CONGESTION
GVE_MTD_PATH_STATE_RETRANSMIT
BUG=b/185177172
TEST=presubmit
RELEASE_NOTE=None
SOURCE=GVNIC
Signed-off-by: Willem de Bruijn <willemb@google.com>
Change-Id: I29bdc2369f9aa0b3d483c2547f145bf0c4b1c384
Reviewed-on: https://cos-review.googlesource.com/c/third_party/kernel/+/16034
Tested-by: Cusky Presubmit Bot <presubmit@cos-infra-prod.iam.gserviceaccount.com>
Reviewed-by: Roy Yang <royyang@google.com>
diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h
index 133c698..7cc649d 100644
--- a/drivers/net/ethernet/google/gve/gve.h
+++ b/drivers/net/ethernet/google/gve/gve.h
@@ -102,6 +102,7 @@
/* A TX desc ring entry */
union gve_tx_desc {
struct gve_tx_pkt_desc pkt; /* first desc for a packet */
+ struct gve_tx_mtd_desc mtd; /* optional metadata descriptor */
struct gve_tx_seg_desc seg; /* subsequent descs for a packet */
};
diff --git a/drivers/net/ethernet/google/gve/gve_desc.h b/drivers/net/ethernet/google/gve/gve_desc.h
index a7da364..61d0c66 100644
--- a/drivers/net/ethernet/google/gve/gve_desc.h
+++ b/drivers/net/ethernet/google/gve/gve_desc.h
@@ -33,6 +33,14 @@
__be64 seg_addr; /* Base address (see note) of this segment */
} __packed;
+struct gve_tx_mtd_desc {
+ u8 type_flags; /* type is lower 4 bits, subtype upper */
+ u8 path_state; /* state is lower 4 bits, hash type upper */
+ __be16 reserved0;
+ __be32 path_hash;
+ __be64 reserved1;
+} __packed;
+
struct gve_tx_seg_desc {
u8 type_flags; /* type is lower 4 bits, flags upper */
u8 l3_offset; /* TSO: 2 byte units to start of IPH */
@@ -46,6 +54,7 @@
#define GVE_TXD_STD (0x0 << 4) /* Std with Host Address */
#define GVE_TXD_TSO (0x1 << 4) /* TSO with Host Address */
#define GVE_TXD_SEG (0x2 << 4) /* Seg with Host Address */
+#define GVE_TXD_MTD (0x3 << 4) /* Metadata */
/* GVE Transmit Descriptor Flags for Std Pkts */
#define GVE_TXF_L4CSUM BIT(0) /* Need csum offload */
@@ -54,6 +63,17 @@
/* GVE Transmit Descriptor Flags for TSO Segs */
#define GVE_TXSF_IPV6 BIT(1) /* IPv6 TSO */
+/* GVE Transmit Descriptor Options for MTD Segs */
+#define GVE_MTD_SUBTYPE_PATH 0
+
+#define GVE_MTD_PATH_STATE_DEFAULT 0
+#define GVE_MTD_PATH_STATE_TIMEOUT 1
+#define GVE_MTD_PATH_STATE_CONGESTION 2
+#define GVE_MTD_PATH_STATE_RETRANSMIT 3
+
+#define GVE_MTD_PATH_HASH_NONE (0x0 << 4)
+#define GVE_MTD_PATH_HASH_L4 (0x1 << 4)
+
/* GVE Receive Packet Descriptor */
/* The start of an ethernet packet comes 2 bytes into the rx buffer.
* gVNIC adds this padding so that both the DMA and the L3/4 protocol header
diff --git a/drivers/net/ethernet/google/gve/gve_tx.c b/drivers/net/ethernet/google/gve/gve_tx.c
index b85b7a1..17fc8b9 100644
--- a/drivers/net/ethernet/google/gve/gve_tx.c
+++ b/drivers/net/ethernet/google/gve/gve_tx.c
@@ -408,6 +408,19 @@
pkt_desc->pkt.seg_addr = cpu_to_be64(addr);
}
+static void gve_tx_fill_mtd_desc(union gve_tx_desc *mtd_desc,
+ struct sk_buff *skb)
+{
+ BUILD_BUG_ON(sizeof(mtd_desc->mtd) != sizeof(mtd_desc->pkt));
+
+ mtd_desc->mtd.type_flags = GVE_TXD_MTD | GVE_MTD_SUBTYPE_PATH;
+ mtd_desc->mtd.path_state = GVE_MTD_PATH_STATE_DEFAULT |
+ GVE_MTD_PATH_HASH_L4;
+ mtd_desc->mtd.path_hash = skb->hash;
+ mtd_desc->mtd.reserved0 = 0;
+ mtd_desc->mtd.reserved1 = 0;
+}
+
static void gve_tx_fill_seg_desc(union gve_tx_desc *seg_desc,
struct sk_buff *skb, bool is_gso,
u16 len, u64 addr)
@@ -444,6 +457,7 @@
int pad_bytes, hlen, hdr_nfrags, payload_nfrags, l4_hdr_offset;
union gve_tx_desc *pkt_desc, *seg_desc;
struct gve_tx_buffer_state *info;
+ int mtd_desc_nr = !!skb->l4_hash;
bool is_gso = skb_is_gso(skb);
u32 idx = tx->req & tx->mask;
int payload_iov = 2;
@@ -475,7 +489,7 @@
&info->iov[payload_iov]);
gve_tx_fill_pkt_desc(pkt_desc, skb, is_gso, l4_hdr_offset,
- 1 + payload_nfrags, hlen,
+ 1 + mtd_desc_nr + payload_nfrags, hlen,
info->iov[hdr_nfrags - 1].iov_offset);
skb_copy_bits(skb, 0,
@@ -486,8 +500,13 @@
info->iov[hdr_nfrags - 1].iov_len);
copy_offset = hlen;
+ if (mtd_desc_nr) {
+ next_idx = (tx->req + 1) & tx->mask;
+ gve_tx_fill_mtd_desc(&tx->desc[next_idx], skb);
+ }
+
for (i = payload_iov; i < payload_nfrags + payload_iov; i++) {
- next_idx = (tx->req + 1 + i - payload_iov) & tx->mask;
+ next_idx = (tx->req + 1 + mtd_desc_nr + i - payload_iov) & tx->mask;
seg_desc = &tx->desc[next_idx];
gve_tx_fill_seg_desc(seg_desc, skb, is_gso,
@@ -503,7 +522,7 @@
copy_offset += info->iov[i].iov_len;
}
- return 1 + payload_nfrags;
+ return 1 + mtd_desc_nr + payload_nfrags;
}
static int gve_tx_add_skb_no_copy(struct gve_priv *priv, struct gve_tx_ring *tx,
@@ -511,8 +530,9 @@
{
const struct skb_shared_info *shinfo = skb_shinfo(skb);
int hlen, payload_nfrags, l4_hdr_offset, seg_idx_bias;
- union gve_tx_desc *pkt_desc, *seg_desc;
+ union gve_tx_desc *pkt_desc, *mtd_desc, *seg_desc;
struct gve_tx_buffer_state *info;
+ int mtd_desc_nr = !!skb->l4_hash;
bool is_gso = skb_is_gso(skb);
u32 idx = tx->req & tx->mask;
struct gve_tx_dma_buf *buf;
@@ -546,23 +566,29 @@
dma_unmap_addr_set(buf, dma, addr);
payload_nfrags = shinfo->nr_frags;
+ if (hlen < len)
+ payload_nfrags++;
+
+ gve_tx_fill_pkt_desc(pkt_desc, skb, is_gso, l4_hdr_offset,
+ 1 + mtd_desc_nr + payload_nfrags, hlen, addr);
+
+ if (mtd_desc_nr) {
+ mtd_desc = &tx->desc[(tx->req + 1) & tx->mask];
+ gve_tx_fill_mtd_desc(mtd_desc, skb);
+ seg_idx_bias = 2;
+ } else {
+ seg_idx_bias = 1;
+ }
+
if (hlen < len) {
/* For gso the rest of the linear portion of the skb needs to
* be in its own descriptor.
*/
- payload_nfrags++;
- gve_tx_fill_pkt_desc(pkt_desc, skb, is_gso, l4_hdr_offset,
- 1 + payload_nfrags, hlen, addr);
-
len -= hlen;
addr += hlen;
- seg_desc = &tx->desc[(tx->req + 1) & tx->mask];
- seg_idx_bias = 2;
+ seg_desc = &tx->desc[(tx->req + seg_idx_bias) & tx->mask];
gve_tx_fill_seg_desc(seg_desc, skb, is_gso, len, addr);
- } else {
- seg_idx_bias = 1;
- gve_tx_fill_pkt_desc(pkt_desc, skb, is_gso, l4_hdr_offset,
- 1 + payload_nfrags, hlen, addr);
+ seg_idx_bias++;
}
for (i = 0; i < payload_nfrags - (seg_idx_bias - 1); i++) {
