Re: [PATCH net-next v5 3/3] net: ethernet: rmnet: Add support for MAPv5 egress packets

From: sharathv
Date: Wed May 26 2021 - 10:33:54 EST

Next message: Heikki Krogerus: "Re: [PATCH 24/24] usb: typec: ucsi: Fix copy/paste issue for 'ucsi_set_drvdata()'"
Previous message: Yang Yingliang: "[PATCH -next] cred: add missing return error code when set_cred_ucounts() failed"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On 2021-04-23 18:18, Alex Elder wrote:

On 4/23/21 7:19 AM, Sharath Chandra Vurukala wrote:

Adding Support for MAPv5 egress packets.
Based on the configuration Request HW for csum offload
by setting the csum_valid_required of Mapv5 packet.

Please try to re-word this description. I'm not sure
I understand what it means.

I see what I think is a bug below. Please either
fix or explain.

Acked-by: Subash Abhinov Kasiviswanathan <subashab@xxxxxxxxxxxxxx>
Acked-by: Alex Elder <elder@xxxxxxxxxx>

I did not acknowledge this patch.

I will rewrite the description with more details and clear wordings. Thanks.

Signed-off-by: Sharath Chandra Vurukala <sharathv@xxxxxxxxxxxxxx>
---
drivers/net/ethernet/qualcomm/rmnet/rmnet_config.h | 4 +-
.../net/ethernet/qualcomm/rmnet/rmnet_handlers.c | 14 +++-
drivers/net/ethernet/qualcomm/rmnet/rmnet_map.h | 8 +-
.../net/ethernet/qualcomm/rmnet/rmnet_map_data.c | 93 ++++++++++++++++++++--
drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c | 3 +-
include/uapi/linux/if_link.h | 1 +
6 files changed, 109 insertions(+), 14 deletions(-)

diff --git a/drivers/net/ethernet/qualcomm/rmnet/rmnet_config.h b/drivers/net/ethernet/qualcomm/rmnet/rmnet_config.h
index 8d8d469..8e64ca9 100644
--- a/drivers/net/ethernet/qualcomm/rmnet/rmnet_config.h
+++ b/drivers/net/ethernet/qualcomm/rmnet/rmnet_config.h
@@ -1,5 +1,6 @@
/* SPDX-License-Identifier: GPL-2.0-only */
-/* Copyright (c) 2013-2014, 2016-2018 The Linux Foundation. All rights reserved.
+/* Copyright (c) 2013-2014, 2016-2018, 2021 The Linux Foundation.
+ * All rights reserved.
*
* RMNET Data configuration engine
*/
@@ -56,6 +57,7 @@ struct rmnet_priv_stats {
u64 csum_fragmented_pkt;
u64 csum_skipped;
u64 csum_sw;
+ u64 csum_hw;

Why is this new statistic type added? Would it be
meaningful to use before--with only QMAPv4? Or is
there something different about QMAPv5 (inline) checksum
offload that makes this necessary or desirable?

This is something new that ought to be at least
mentioned in the description at the top. And for
future reference, this could likely have been
defined in a separate patch, before this one.

Will update the description with details about this stat.

};
struct rmnet_priv {
diff --git a/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c b/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c
index 706a225..51a2e94 100644
--- a/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c
+++ b/drivers/net/ethernet/qualcomm/rmnet/rmnet_handlers.c
@@ -133,7 +133,7 @@ static int rmnet_map_egress_handler(struct sk_buff *skb,
struct rmnet_port *port, u8 mux_id,
struct net_device *orig_dev)
{
- int required_headroom, additional_header_len;
+ int required_headroom, additional_header_len, csum_type = 0;
struct rmnet_map_header *map_header;
additional_header_len = 0;
@@ -142,6 +142,10 @@ static int rmnet_map_egress_handler(struct sk_buff *skb,
if (port->data_format & RMNET_FLAGS_EGRESS_MAP_CKSUMV4) {
additional_header_len = sizeof(struct rmnet_map_ul_csum_header);
required_headroom += additional_header_len;
+ csum_type = RMNET_FLAGS_EGRESS_MAP_CKSUMV4;
+ } else if (port->data_format & RMNET_FLAGS_EGRESS_MAP_CKSUMV5) {
+ additional_header_len = sizeof(struct rmnet_map_v5_csum_header);
+ csum_type = RMNET_FLAGS_EGRESS_MAP_CKSUMV5;
}

Does additional_header_len need to be added to required_headroom,
as it is for QMAPv4 above?

If so, this is a bug and must be fixed.

What I tested last week (and verified work for IPA v3.5.1 and
IPA v4.2) looked like this:

if (port->data_format & RMNET_FLAGS_EGRESS_MAP_CKSUMV4) {
additional_header_len = sizeof(struct rmnet_map_ul_csum_header);
csum_type = RMNET_FLAGS_EGRESS_MAP_CKSUMV4;
} else if (port->data_format & RMNET_FLAGS_EGRESS_MAP_CKSUMV5) {
additional_header_len = sizeof(struct rmnet_map_v5_csum_header);
csum_type = RMNET_FLAGS_EGRESS_MAP_CKSUMV5;
}
required_headroom += additional_header_len;

-Alex

Thanks a lot in identifying the bug, I messed up with the patches before raising them for review.
Will share the corrected version in subsequent patch.

if (skb_headroom(skb) < required_headroom) {
@@ -149,10 +153,12 @@ static int rmnet_map_egress_handler(struct sk_buff *skb,
return -ENOMEM;
}
- if (port->data_format & RMNET_FLAGS_EGRESS_MAP_CKSUMV4)
- rmnet_map_checksum_uplink_packet(skb, orig_dev);
+ if (csum_type)
+ rmnet_map_checksum_uplink_packet(skb, port, orig_dev,
+ csum_type);
- map_header = rmnet_map_add_map_header(skb, additional_header_len, 0);
+ map_header = rmnet_map_add_map_header(skb, additional_header_len,
+ port, 0);
if (!map_header)
return -ENOMEM;
diff --git a/drivers/net/ethernet/qualcomm/rmnet/rmnet_map.h b/drivers/net/ethernet/qualcomm/rmnet/rmnet_map.h
index 1a399bf..e5a0b38 100644
--- a/drivers/net/ethernet/qualcomm/rmnet/rmnet_map.h
+++ b/drivers/net/ethernet/qualcomm/rmnet/rmnet_map.h
@@ -43,11 +43,15 @@ enum rmnet_map_commands {
struct sk_buff *rmnet_map_deaggregate(struct sk_buff *skb,
struct rmnet_port *port);
struct rmnet_map_header *rmnet_map_add_map_header(struct sk_buff *skb,
- int hdrlen, int pad);
+ int hdrlen,
+ struct rmnet_port *port,
+ int pad);
void rmnet_map_command(struct sk_buff *skb, struct rmnet_port *port);
int rmnet_map_checksum_downlink_packet(struct sk_buff *skb, u16 len);
void rmnet_map_checksum_uplink_packet(struct sk_buff *skb,
- struct net_device *orig_dev);
+ struct rmnet_port *port,
+ struct net_device *orig_dev,
+ int csum_type);
int rmnet_map_process_next_hdr_packet(struct sk_buff *skb, u16 len);
#endif /* _RMNET_MAP_H_ */
diff --git a/drivers/net/ethernet/qualcomm/rmnet/rmnet_map_data.c b/drivers/net/ethernet/qualcomm/rmnet/rmnet_map_data.c
index 43813cf..339d964 100644
--- a/drivers/net/ethernet/qualcomm/rmnet/rmnet_map_data.c
+++ b/drivers/net/ethernet/qualcomm/rmnet/rmnet_map_data.c
@@ -12,6 +12,7 @@
#include "rmnet_config.h"
#include "rmnet_map.h"
#include "rmnet_private.h"
+#include <linux/bitfield.h>
#define RMNET_MAP_DEAGGR_SPACING 64
#define RMNET_MAP_DEAGGR_HEADROOM (RMNET_MAP_DEAGGR_SPACING / 2)
@@ -251,12 +252,69 @@ rmnet_map_ipv6_ul_csum_header(void *ip6hdr,
}
#endif
+static void rmnet_map_v5_checksum_uplink_packet(struct sk_buff *skb,
+ struct rmnet_port *port,
+ struct net_device *orig_dev)
+{
+ struct rmnet_priv *priv = netdev_priv(orig_dev);
+ struct rmnet_map_v5_csum_header *ul_header;
+
+ if (!(port->data_format & RMNET_FLAGS_EGRESS_MAP_CKSUMV5))
+ return;
+
+ ul_header = skb_push(skb, sizeof(*ul_header));
+ memset(ul_header, 0, sizeof(*ul_header));
+ ul_header->header_info = RMNET_MAP_HEADER_TYPE_CSUM_OFFLOAD <<
+ MAPV5_HDRINFO_HDR_TYPE_SHIFT;
+
+ if (skb->ip_summed == CHECKSUM_PARTIAL) {
+ void *iph = (char *)ul_header + sizeof(*ul_header);
+ __sum16 *check;
+ void *trans;
+ u8 proto;
+
+ if (skb->protocol == htons(ETH_P_IP)) {
+ u16 ip_len = ((struct iphdr *)iph)->ihl * 4;
+
+ proto = ((struct iphdr *)iph)->protocol;
+ trans = iph + ip_len;
+ } else if (skb->protocol == htons(ETH_P_IPV6)) {
+#if IS_ENABLED(CONFIG_IPV6)
+ u16 ip_len = sizeof(struct ipv6hdr);
+
+ proto = ((struct ipv6hdr *)iph)->nexthdr;
+ trans = iph + ip_len;
+#else
+ priv->stats.csum_err_invalid_ip_version++;
+ goto sw_csum;
+#endif /* CONFIG_IPV6 */
+ } else {
+ priv->stats.csum_err_invalid_ip_version++;
+ goto sw_csum;
+ }
+
+ check = rmnet_map_get_csum_field(proto, trans);
+ if (check) {
+ skb->ip_summed = CHECKSUM_NONE;
+ /* Ask for checksum offloading */
+ ul_header->csum_info |= MAPV5_CSUMINFO_VALID_FLAG;
+ priv->stats.csum_hw++;
+ return;
+ }
+ }
+
+sw_csum:
+ priv->stats.csum_sw++;
+}
+
/* Adds MAP header to front of skb->data
* Padding is calculated and set appropriately in MAP header. Mux ID is
* initialized to 0.
*/
struct rmnet_map_header *rmnet_map_add_map_header(struct sk_buff *skb,
- int hdrlen, int pad)
+ int hdrlen,
+ struct rmnet_port *port,
+ int pad)
{
struct rmnet_map_header *map_header;
u32 padding, map_datalen;
@@ -267,6 +325,10 @@ struct rmnet_map_header *rmnet_map_add_map_header(struct sk_buff *skb,
skb_push(skb, sizeof(struct rmnet_map_header));
memset(map_header, 0, sizeof(struct rmnet_map_header));
+ /* Set next_hdr bit for csum offload packets */
+ if (port->data_format & RMNET_FLAGS_EGRESS_MAP_CKSUMV5)
+ map_header->flags |= MAP_NEXT_HEADER_FLAG;
+
if (pad == RMNET_MAP_NO_PAD_BYTES) {
map_header->pkt_len = htons(map_datalen);
return map_header;
@@ -394,11 +456,8 @@ int rmnet_map_checksum_downlink_packet(struct sk_buff *skb, u16 len)
return 0;
}
-/* Generates UL checksum meta info header for IPv4 and IPv6 over TCP and UDP
- * packets that are supported for UL checksum offload.
- */
-void rmnet_map_checksum_uplink_packet(struct sk_buff *skb,
- struct net_device *orig_dev)
+static void rmnet_map_v4_checksum_uplink_packet(struct sk_buff *skb,
+ struct net_device *orig_dev)
{
struct rmnet_priv *priv = netdev_priv(orig_dev);
struct rmnet_map_ul_csum_header *ul_header;
@@ -417,10 +476,12 @@ void rmnet_map_checksum_uplink_packet(struct sk_buff *skb,
if (skb->protocol == htons(ETH_P_IP)) {
rmnet_map_ipv4_ul_csum_header(iphdr, ul_header, skb);
+ priv->stats.csum_hw++;
return;
} else if (skb->protocol == htons(ETH_P_IPV6)) {
#if IS_ENABLED(CONFIG_IPV6)
rmnet_map_ipv6_ul_csum_header(iphdr, ul_header, skb);
+ priv->stats.csum_hw++;
return;
#else
priv->stats.csum_err_invalid_ip_version++;
@@ -437,6 +498,26 @@ void rmnet_map_checksum_uplink_packet(struct sk_buff *skb,
priv->stats.csum_sw++;
}
+/* Generates UL checksum meta info header for IPv4 and IPv6 over TCP and UDP
+ * packets that are supported for UL checksum offload.
+ */
+void rmnet_map_checksum_uplink_packet(struct sk_buff *skb,
+ struct rmnet_port *port,
+ struct net_device *orig_dev,
+ int csum_type)
+{
+ switch (csum_type) {
+ case RMNET_FLAGS_EGRESS_MAP_CKSUMV4:
+ rmnet_map_v4_checksum_uplink_packet(skb, orig_dev);
+ break;
+ case RMNET_FLAGS_EGRESS_MAP_CKSUMV5:
+ rmnet_map_v5_checksum_uplink_packet(skb, port, orig_dev);
+ break;
+ default:
+ break;
+ }
+}
+
/* Process a MAPv5 packet header */
int rmnet_map_process_next_hdr_packet(struct sk_buff *skb,
u16 len)
diff --git a/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c b/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
index 41fbd2c..bc6d6ac 100644
--- a/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
+++ b/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
@@ -174,6 +174,7 @@ static const char rmnet_gstrings_stats[][ETH_GSTRING_LEN] = {
"Checksum skipped on ip fragment",
"Checksum skipped",
"Checksum computed in software",
+ "Checksum computed in hardware",
};
static void rmnet_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
@@ -354,4 +355,4 @@ int rmnet_vnd_update_dev_mtu(struct rmnet_port *port,
}
return 0;
-}
\ No newline at end of file
+}
diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h
index 21529b3..1691f3a 100644
--- a/include/uapi/linux/if_link.h
+++ b/include/uapi/linux/if_link.h
@@ -1236,6 +1236,7 @@ enum {
#define RMNET_FLAGS_INGRESS_MAP_CKSUMV4 (1U << 2)
#define RMNET_FLAGS_EGRESS_MAP_CKSUMV4 (1U << 3)
#define RMNET_FLAGS_INGRESS_MAP_CKSUMV5 (1U << 4)
+#define RMNET_FLAGS_EGRESS_MAP_CKSUMV5 (1U << 5)
enum {
IFLA_RMNET_UNSPEC,

Next message: Heikki Krogerus: "Re: [PATCH 24/24] usb: typec: ucsi: Fix copy/paste issue for 'ucsi_set_drvdata()'"
Previous message: Yang Yingliang: "[PATCH -next] cred: add missing return error code when set_cred_ucounts() failed"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]