Skip to contentSkip to conversation

ice: switch: debloat packet templates code

PATCH In review Alexander Lobakin 3 participants 13 messages Mar 21, 2022 – Mar 30, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v4 cover ice: switch: debloat packet templates code

This hunts down several places around packet templates/dummies for switch rules which are either repetitive, fragile or just not really readable code.
It's a common need to add new packet templates and to review such changes as well, try to simplify both with the help of a pair macros and aliases.
ice_find_dummy_packet() became very complex at this point with tons of nested if-elses. It clearly showed this approach does not scale, so convert its logics to the simple mask-match + static const array.

bloat-o-meter is happy about that (built w/ LLVM 13):

add/remove: 0/1 grow/shrink: 1/1 up/down: 2/-1058 (-1056)

Function                                     old     new   delta
ice_fill_adv_dummy_packet                    289     291      +2
ice_adv_add_update_vsi_list                  201       -    -201

ice_add_adv_rule 2950 2093 -857
Total: Before=414512, After=413456, chg -0.25% add/remove: 53/52 grow/shrink: 0/0 up/down: 4660/-3988 (672)
RO Data old new delta
ice_dummy_pkt_profiles - 672 +672
Total: Before=37895, After=38567, chg +1.77%

Diffstat also looks nice, and adding new packet templates now takes less lines.

We'll probably come out with dynamic template crafting in a while, but for now let's improve what we have currently.

From v3[0]:
- change u64:48 + u64:16 -> u32 + u16 to fix issues on 32-bit platforms (and make object code a bit simpler);
- mention ice_find_dummy_packet() conversion in the cover letter.

From v2[1]:
- rebase on top of the GTP changes;
- new: convert template search code to a rodata array (-1000 bytes from .text, -400 bytes from ice.ko);
- collect Reviewed-by and Tested-by (Marcin, Michal).

From v1[2]:
- rebase on top of the latest next-queue (to fix #3 not applying);
- adjust the kdoc accordingly to the function proto changes in #3;
- no functional changes.

[0] https://lore.kernel.org/netdev/20220318161713.680436-1-alexandr.lobakin at intel.com
[1] https://lore.kernel.org/netdev/20220127154009.623304-1-alexandr.lobakin at intel.com
[2] https://lore.kernel.org/netdev/20220124173116.739083-1-alexandr.lobakin at intel.com

Alexander Lobakin (5):
ice: switch: add and use u16[] aliases to ice_adv_lkup_elem::{h,m}_u
ice: switch: unobscurify bitops loop in ice_fill_adv_dummy_packet()
ice: switch: use a struct to pass packet template params
ice: switch: use convenience macros to declare dummy pkt templates
ice: switch: convert packet template match code to rodata

 drivers/net/ethernet/intel/ice/ice_switch.c | 489 ++++++++------------
 drivers/net/ethernet/intel/ice/ice_switch.h |  12 +-
 2 files changed, 211 insertions(+), 290 deletions(-)
Signature · 1 line
2.35.1
#1 lore raw

PATCH v4 1/5 ice: switch: add and use u16[] aliases to ice_adv_lkup_elem::{h,m}_u

ice_adv_lkup_elem fields h_u and m_u are being accessed as raw u16 arrays in several places.
To reduce cast and braces burden, add permanent array-of-u16 aliases with the same size as the union ice_prot_hdr itself via anonymous unions to the actual struct declaration, and just access them directly.

This:
- removes the need to cast the union to u16[] and then dereference it each time -> reduces the horizon for potential bugs;
- improves -Warray-bounds coverage -- the array size is now known at compilation time;
- addresses cppcheck complaints.

---
 drivers/net/ethernet/intel/ice/ice_switch.c | 15 +++++++--------
 drivers/net/ethernet/intel/ice/ice_switch.h | 12 ++++++++++--
 2 files changed, 17 insertions(+), 10 deletions(-)
drivers/net/ethernet/intel/ice/ice_switch.c +7 -8
index 25b8f6f726eb..075df2474688 100644
@@ -5811,12 +5811,12 @@ ice_fill_adv_dummy_packet(struct ice_adv_lkup_elem *lkups, u16 lkups_cnt,
 		 * over any significant packet data.
 		 */
 		for (j = 0; j < len / sizeof(u16); j++)
-			if (((u16 *)&lkups[i].m_u)[j])
+			if (lkups[i].m_raw[j])
 				((u16 *)(pkt + offset))[j] =
 					(((u16 *)(pkt + offset))[j] &
-					 ~((u16 *)&lkups[i].m_u)[j]) |
-					(((u16 *)&lkups[i].h_u)[j] &
-					 ((u16 *)&lkups[i].m_u)[j]);
+					 ~lkups[i].m_raw[j]) |
+					(lkups[i].h_raw[j] &
+					 lkups[i].m_raw[j]);
 	}
 
 	s_rule->pdata.lkup_tx_rx.hdr_len = cpu_to_le16(pkt_len);
@@ -6065,11 +6065,10 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 	/* get # of words we need to match */
 	word_cnt = 0;
 	for (i = 0; i < lkups_cnt; i++) {
-		u16 j, *ptr;
+		u16 j;
 
-		ptr = (u16 *)&lkups[i].m_u;
-		for (j = 0; j < sizeof(lkups->m_u) / sizeof(u16); j++)
-			if (ptr[j] != 0)
+		for (j = 0; j < ARRAY_SIZE(lkups->m_raw); j++)
+			if (lkups[i].m_raw[j])
 				word_cnt++;
 	}
 
drivers/net/ethernet/intel/ice/ice_switch.h +10 -2
index ed3d1d03befa..ecac75e71395 100644
@@ -138,8 +138,16 @@ struct ice_update_recipe_lkup_idx_params {
 
 struct ice_adv_lkup_elem {
 	enum ice_protocol_type type;
-	union ice_prot_hdr h_u;	/* Header values */
-	union ice_prot_hdr m_u;	/* Mask of header values to match */
+	union {
+		union ice_prot_hdr h_u;	/* Header values */
+		/* Used to iterate over the headers */
+		u16 h_raw[sizeof(union ice_prot_hdr) / sizeof(u16)];
+	};
+	union {
+		union ice_prot_hdr m_u;	/* Mask of header values to match */
+		/* Used to iterate over header mask */
+		u16 m_raw[sizeof(union ice_prot_hdr) / sizeof(u16)];
+	};
 };
 
 struct ice_sw_act_ctrl {
Signature · 1 line
2.35.1
Tested-by on v4 1/5
#2 lore raw

RE: [Intel-wired-lan] [PATCH v4 net-next 1/5] ice: switch: add and use u16[] aliases to ice_adv_lkup_elem::{h, m}_u

Quoted 4 lines

-----Original Message-----
From: Intel-wired-lan <[email protected]> On Behalf Of Alexander Lobakin
Sent: Monday, March 21, 2022 4:30 PM

To: [email protected]

Quoted 20 lines

Cc: Szapar-Mudlaw, Martyna <[email protected]>; netdev at vger.kernel.org; linux-kernel at vger.kernel.org; Jakub Kicinski <[email protected]>; Paolo Abeni <[email protected]>; David S. Miller <[email protected]>
Subject: [Intel-wired-lan] [PATCH v4 net-next 1/5] ice: switch: add and use u16[] aliases to ice_adv_lkup_elem::{h, m}_u

ice_adv_lkup_elem fields h_u and m_u are being accessed as raw u16 arrays in several places.
To reduce cast and braces burden, add permanent array-of-u16 aliases with the same size as the union ice_prot_hdr itself via anonymous unions to the actual struct declaration, and just access them directly.

This:
- removes the need to cast the union to u16[] and then dereference it each time -> reduces the horizon for potential bugs;
- improves -Warray-bounds coverage -- the array size is now known at compilation time;
- addresses cppcheck complaints.

Quoted 8 lines

---
drivers/net/ethernet/intel/ice/ice_switch.c | 15 +++++++-------- drivers/net/ethernet/intel/ice/ice_switch.h | 12 ++++++++++--

2 files changed, 17 insertions(+), 10 deletions(-)
#3 lore raw

PATCH v4 2/5 ice: switch: unobscurify bitops loop in ice_fill_adv_dummy_packet()

In reply to Alexander Lobakin

A loop performing header modification according to the provided mask in ice_fill_adv_dummy_packet() is very cryptic (and error-prone). Replace two identical cast-deferences with a variable. Replace three struct-member-array-accesses with a variable. Invert the condition, reduce the indentation by one -> eliminate line wraps.

---
 drivers/net/ethernet/intel/ice/ice_switch.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)
drivers/net/ethernet/intel/ice/ice_switch.c +9 -7
index 075df2474688..0936d39de70c 100644
@@ -5810,13 +5810,15 @@ ice_fill_adv_dummy_packet(struct ice_adv_lkup_elem *lkups, u16 lkups_cnt,
 		 * indicated by the mask to make sure we don't improperly write
 		 * over any significant packet data.
 		 */
-		for (j = 0; j < len / sizeof(u16); j++)
-			if (lkups[i].m_raw[j])
-				((u16 *)(pkt + offset))[j] =
-					(((u16 *)(pkt + offset))[j] &
-					 ~lkups[i].m_raw[j]) |
-					(lkups[i].h_raw[j] &
-					 lkups[i].m_raw[j]);
+		for (j = 0; j < len / sizeof(u16); j++) {
+			u16 *ptr = (u16 *)(pkt + offset);
+			u16 mask = lkups[i].m_raw[j];
+
+			if (!mask)
+				continue;
+
+			ptr[j] = (ptr[j] & ~mask) | (lkups[i].h_raw[j] & mask);
+		}
 	}
 
 	s_rule->pdata.lkup_tx_rx.hdr_len = cpu_to_le16(pkt_len);
Signature · 1 line
2.35.1
Tested-by on v4 2/5
#4 lore raw

RE: [Intel-wired-lan] [PATCH v4 net-next 2/5] ice: switch: unobscurify bitops loop in ice_fill_adv_dummy_packet()

Quoted 4 lines

-----Original Message-----
From: Intel-wired-lan <[email protected]> On Behalf Of Alexander Lobakin
Sent: Monday, March 21, 2022 4:30 PM

To: [email protected]

Quoted 13 lines

Cc: Szapar-Mudlaw, Martyna <[email protected]>; netdev at vger.kernel.org; linux-kernel at vger.kernel.org; Jakub Kicinski <[email protected]>; Paolo Abeni <[email protected]>; David S. Miller <[email protected]>
Subject: [Intel-wired-lan] [PATCH v4 net-next 2/5] ice: switch: unobscurify bitops loop in ice_fill_adv_dummy_packet()

A loop performing header modification according to the provided mask in ice_fill_adv_dummy_packet() is very cryptic (and error-prone). Replace two identical cast-deferences with a variable. Replace three struct- member-array-accesses with a variable. Invert the condition, reduce the indentation by one -> eliminate line wraps.

Quoted 7 lines

---
drivers/net/ethernet/intel/ice/ice_switch.c | 16 +++++++++-------

1 file changed, 9 insertions(+), 7 deletions(-)
#5 lore raw

PATCH v4 3/5 ice: switch: use a struct to pass packet template params

In reply to Alexander Lobakin

ice_find_dummy_packet() contains a lot of boilerplate code and a nice room for copy-paste mistakes.
Instead of passing 3 separate pointers back and forth to get packet template (dummy) params, directly return a structure containing them. Then, use a macro to compose compound literals and avoid code duplication on return path.
Now, dummy packet type/name is needed only once to return a full correct triple pkt-pkt_len-offsets, and those are all one-liners. dummy_ipv4_gtpu_ipv4_packet_offsets is just moved around and renamed (as well as dummy_ipv6_gtp_packet_offsets) with no function changes.

---
 drivers/net/ethernet/intel/ice/ice_switch.c | 268 +++++++-------------
 1 file changed, 94 insertions(+), 174 deletions(-)
drivers/net/ethernet/intel/ice/ice_switch.c +94 -174
index 0936d39de70c..4697eb8b4c66 100644
@@ -35,6 +35,20 @@ struct ice_dummy_pkt_offsets {
 	u16 offset; /* ICE_PROTOCOL_LAST indicates end of list */
 };
 
+struct ice_dummy_pkt_profile {
+	const struct ice_dummy_pkt_offsets *offsets;
+	const u8 *pkt;
+	u16 pkt_len;
+};
+
+#define ICE_PKT_PROFILE(type) ({					\
+	(struct ice_dummy_pkt_profile){					\
+		.pkt		= dummy_##type##_packet,		\
+		.pkt_len	= sizeof(dummy_##type##_packet),	\
+		.offsets	= dummy_##type##_packet_offsets,	\
+	};								\
+})
+
 static const struct ice_dummy_pkt_offsets dummy_gre_tcp_packet_offsets[] = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
@@ -1141,6 +1155,15 @@ static const u8 dummy_ipv6_gtpu_ipv6_udp_packet[] = {
 	0x00, 0x00, /* 2 bytes for 4 byte alignment */
 };
 
+static const
+struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv4_packet_offsets[] = {
+	{ ICE_MAC_OFOS,		0 },
+	{ ICE_IPV4_OFOS,	14 },
+	{ ICE_UDP_OF,		34 },
+	{ ICE_GTP_NO_PAY,	42 },
+	{ ICE_PROTOCOL_LAST,	0 },
+};
+
 static const u8 dummy_ipv4_gtpu_ipv4_packet[] = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
@@ -1172,16 +1195,7 @@ static const u8 dummy_ipv4_gtpu_ipv4_packet[] = {
 };
 
 static const
-struct ice_dummy_pkt_offsets dummy_ipv4_gtp_no_pay_packet_offsets[] = {
-	{ ICE_MAC_OFOS,		0 },
-	{ ICE_IPV4_OFOS,	14 },
-	{ ICE_UDP_OF,		34 },
-	{ ICE_GTP_NO_PAY,	42 },
-	{ ICE_PROTOCOL_LAST,	0 },
-};
-
-static const
-struct ice_dummy_pkt_offsets dummy_ipv6_gtp_no_pay_packet_offsets[] = {
+struct ice_dummy_pkt_offsets dummy_ipv6_gtp_packet_offsets[] = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV6_OFOS,	14 },
 	{ ICE_UDP_OF,		54 },
@@ -5501,15 +5515,12 @@ ice_add_adv_recipe(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
  *	   structure per protocol header
  * @lkups_cnt: number of protocols
  * @tun_type: tunnel type
- * @pkt: dummy packet to fill according to filter match criteria
- * @pkt_len: packet length of dummy packet
- * @offsets: pointer to receive the pointer to the offsets for the packet
+ *
+ * Returns the &ice_dummy_pkt_profile corresponding to these lookup params.
  */
-static void
+static struct ice_dummy_pkt_profile
 ice_find_dummy_packet(struct ice_adv_lkup_elem *lkups, u16 lkups_cnt,
-		      enum ice_sw_tunnel_type tun_type,
-		      const u8 **pkt, u16 *pkt_len,
-		      const struct ice_dummy_pkt_offsets **offsets)
+		      enum ice_sw_tunnel_type tun_type)
 {
 	bool inner_tcp = false, inner_udp = false, outer_ipv6 = false;
 	bool vlan = false, inner_ipv6 = false, gtp_no_pay = false;
@@ -5545,168 +5556,86 @@ ice_find_dummy_packet(struct ice_adv_lkup_elem *lkups, u16 lkups_cnt,
 	if (tun_type == ICE_SW_TUN_GTPU) {
 		if (outer_ipv6) {
 			if (gtp_no_pay) {
-				*pkt = dummy_ipv6_gtp_packet;
-				*pkt_len = sizeof(dummy_ipv6_gtp_packet);
-				*offsets = dummy_ipv6_gtp_no_pay_packet_offsets;
+				return ICE_PKT_PROFILE(ipv6_gtp);
 			} else if (inner_ipv6) {
-				if (inner_udp) {
-					*pkt = dummy_ipv6_gtpu_ipv6_udp_packet;
-					*pkt_len = sizeof(dummy_ipv6_gtpu_ipv6_udp_packet);
-					*offsets = dummy_ipv6_gtpu_ipv6_udp_packet_offsets;
-				} else {
-					*pkt = dummy_ipv6_gtpu_ipv6_tcp_packet;
-					*pkt_len = sizeof(dummy_ipv6_gtpu_ipv6_tcp_packet);
-					*offsets = dummy_ipv6_gtpu_ipv6_tcp_packet_offsets;
-				}
+				if (inner_udp)
+					return ICE_PKT_PROFILE(ipv6_gtpu_ipv6_udp);
+				else
+					return ICE_PKT_PROFILE(ipv6_gtpu_ipv6_tcp);
 			} else {
-				if (inner_udp) {
-					*pkt = dummy_ipv6_gtpu_ipv4_udp_packet;
-					*pkt_len = sizeof(dummy_ipv6_gtpu_ipv4_udp_packet);
-					*offsets = dummy_ipv6_gtpu_ipv4_udp_packet_offsets;
-				} else {
-					*pkt = dummy_ipv6_gtpu_ipv4_tcp_packet;
-					*pkt_len = sizeof(dummy_ipv6_gtpu_ipv4_tcp_packet);
-					*offsets = dummy_ipv6_gtpu_ipv4_tcp_packet_offsets;
-				}
+				if (inner_udp)
+					return ICE_PKT_PROFILE(ipv6_gtpu_ipv4_udp);
+				else
+					return ICE_PKT_PROFILE(ipv6_gtpu_ipv4_tcp);
 			}
 		} else {
 			if (gtp_no_pay) {
-				*pkt = dummy_ipv4_gtpu_ipv4_packet;
-				*pkt_len = sizeof(dummy_ipv4_gtpu_ipv4_packet);
-				*offsets = dummy_ipv4_gtp_no_pay_packet_offsets;
+				return ICE_PKT_PROFILE(ipv4_gtpu_ipv4);
 			} else if (inner_ipv6) {
-				if (inner_udp) {
-					*pkt = dummy_ipv4_gtpu_ipv6_udp_packet;
-					*pkt_len = sizeof(dummy_ipv4_gtpu_ipv6_udp_packet);
-					*offsets = dummy_ipv4_gtpu_ipv6_udp_packet_offsets;
-				} else {
-					*pkt = dummy_ipv4_gtpu_ipv6_tcp_packet;
-					*pkt_len = sizeof(dummy_ipv4_gtpu_ipv6_tcp_packet);
-					*offsets = dummy_ipv4_gtpu_ipv6_tcp_packet_offsets;
-				}
+				if (inner_udp)
+					return ICE_PKT_PROFILE(ipv4_gtpu_ipv6_udp);
+				else
+					return ICE_PKT_PROFILE(ipv4_gtpu_ipv6_tcp);
 			} else {
-				if (inner_udp) {
-					*pkt = dummy_ipv4_gtpu_ipv4_udp_packet;
-					*pkt_len = sizeof(dummy_ipv4_gtpu_ipv4_udp_packet);
-					*offsets = dummy_ipv4_gtpu_ipv4_udp_packet_offsets;
-				} else {
-					*pkt = dummy_ipv4_gtpu_ipv4_tcp_packet;
-					*pkt_len = sizeof(dummy_ipv4_gtpu_ipv4_tcp_packet);
-					*offsets = dummy_ipv4_gtpu_ipv4_tcp_packet_offsets;
-				}
+				if (inner_udp)
+					return ICE_PKT_PROFILE(ipv4_gtpu_ipv4_udp);
+				else
+					return ICE_PKT_PROFILE(ipv4_gtpu_ipv4_tcp);
 			}
 		}
-		return;
 	}
 
 	if (tun_type == ICE_SW_TUN_GTPC) {
-		if (outer_ipv6) {
-			*pkt = dummy_ipv6_gtp_packet;
-			*pkt_len = sizeof(dummy_ipv6_gtp_packet);
-			*offsets = dummy_ipv6_gtp_no_pay_packet_offsets;
-		} else {
-			*pkt = dummy_ipv4_gtpu_ipv4_packet;
-			*pkt_len = sizeof(dummy_ipv4_gtpu_ipv4_packet);
-			*offsets = dummy_ipv4_gtp_no_pay_packet_offsets;
-		}
-		return;
+		if (outer_ipv6)
+			return ICE_PKT_PROFILE(ipv6_gtp);
+		else
+			return ICE_PKT_PROFILE(ipv4_gtpu_ipv4);
 	}
 
 	if (tun_type == ICE_SW_TUN_NVGRE) {
-		if (inner_tcp && inner_ipv6) {
-			*pkt = dummy_gre_ipv6_tcp_packet;
-			*pkt_len = sizeof(dummy_gre_ipv6_tcp_packet);
-			*offsets = dummy_gre_ipv6_tcp_packet_offsets;
-			return;
-		}
-		if (inner_tcp) {
-			*pkt = dummy_gre_tcp_packet;
-			*pkt_len = sizeof(dummy_gre_tcp_packet);
-			*offsets = dummy_gre_tcp_packet_offsets;
-			return;
-		}
-		if (inner_ipv6) {
-			*pkt = dummy_gre_ipv6_udp_packet;
-			*pkt_len = sizeof(dummy_gre_ipv6_udp_packet);
-			*offsets = dummy_gre_ipv6_udp_packet_offsets;
-			return;
-		}
-		*pkt = dummy_gre_udp_packet;
-		*pkt_len = sizeof(dummy_gre_udp_packet);
-		*offsets = dummy_gre_udp_packet_offsets;
-		return;
+		if (inner_tcp && inner_ipv6)
+			return ICE_PKT_PROFILE(gre_ipv6_tcp);
+		else if (inner_tcp)
+			return ICE_PKT_PROFILE(gre_tcp);
+		else if (inner_ipv6)
+			return ICE_PKT_PROFILE(gre_ipv6_udp);
+		else
+			return ICE_PKT_PROFILE(gre_udp);
 	}
 
 	if (tun_type == ICE_SW_TUN_VXLAN ||
 	    tun_type == ICE_SW_TUN_GENEVE) {
-		if (inner_tcp && inner_ipv6) {
-			*pkt = dummy_udp_tun_ipv6_tcp_packet;
-			*pkt_len = sizeof(dummy_udp_tun_ipv6_tcp_packet);
-			*offsets = dummy_udp_tun_ipv6_tcp_packet_offsets;
-			return;
-		}
-		if (inner_tcp) {
-			*pkt = dummy_udp_tun_tcp_packet;
-			*pkt_len = sizeof(dummy_udp_tun_tcp_packet);
-			*offsets = dummy_udp_tun_tcp_packet_offsets;
-			return;
-		}
-		if (inner_ipv6) {
-			*pkt = dummy_udp_tun_ipv6_udp_packet;
-			*pkt_len = sizeof(dummy_udp_tun_ipv6_udp_packet);
-			*offsets = dummy_udp_tun_ipv6_udp_packet_offsets;
-			return;
-		}
-		*pkt = dummy_udp_tun_udp_packet;
-		*pkt_len = sizeof(dummy_udp_tun_udp_packet);
-		*offsets = dummy_udp_tun_udp_packet_offsets;
-		return;
+		if (inner_tcp && inner_ipv6)
+			return ICE_PKT_PROFILE(udp_tun_ipv6_tcp);
+		else if (inner_tcp)
+			return ICE_PKT_PROFILE(udp_tun_tcp);
+		else if (inner_ipv6)
+			return ICE_PKT_PROFILE(udp_tun_ipv6_udp);
+		else
+			return ICE_PKT_PROFILE(udp_tun_udp);
 	}
 
 	if (inner_udp && !outer_ipv6) {
-		if (vlan) {
-			*pkt = dummy_vlan_udp_packet;
-			*pkt_len = sizeof(dummy_vlan_udp_packet);
-			*offsets = dummy_vlan_udp_packet_offsets;
-			return;
-		}
-		*pkt = dummy_udp_packet;
-		*pkt_len = sizeof(dummy_udp_packet);
-		*offsets = dummy_udp_packet_offsets;
-		return;
+		if (vlan)
+			return ICE_PKT_PROFILE(vlan_udp);
+		else
+			return ICE_PKT_PROFILE(udp);
 	} else if (inner_udp && outer_ipv6) {
-		if (vlan) {
-			*pkt = dummy_vlan_udp_ipv6_packet;
-			*pkt_len = sizeof(dummy_vlan_udp_ipv6_packet);
-			*offsets = dummy_vlan_udp_ipv6_packet_offsets;
-			return;
-		}
-		*pkt = dummy_udp_ipv6_packet;
-		*pkt_len = sizeof(dummy_udp_ipv6_packet);
-		*offsets = dummy_udp_ipv6_packet_offsets;
-		return;
+		if (vlan)
+			return ICE_PKT_PROFILE(vlan_udp_ipv6);
+		else
+			return ICE_PKT_PROFILE(udp_ipv6);
 	} else if ((inner_tcp && outer_ipv6) || outer_ipv6) {
-		if (vlan) {
-			*pkt = dummy_vlan_tcp_ipv6_packet;
-			*pkt_len = sizeof(dummy_vlan_tcp_ipv6_packet);
-			*offsets = dummy_vlan_tcp_ipv6_packet_offsets;
-			return;
-		}
-		*pkt = dummy_tcp_ipv6_packet;
-		*pkt_len = sizeof(dummy_tcp_ipv6_packet);
-		*offsets = dummy_tcp_ipv6_packet_offsets;
-		return;
+		if (vlan)
+			return ICE_PKT_PROFILE(vlan_tcp_ipv6);
+		else
+			return ICE_PKT_PROFILE(tcp_ipv6);
 	}
 
-	if (vlan) {
-		*pkt = dummy_vlan_tcp_packet;
-		*pkt_len = sizeof(dummy_vlan_tcp_packet);
-		*offsets = dummy_vlan_tcp_packet_offsets;
-	} else {
-		*pkt = dummy_tcp_packet;
-		*pkt_len = sizeof(dummy_tcp_packet);
-		*offsets = dummy_tcp_packet_offsets;
-	}
+	if (vlan)
+		return ICE_PKT_PROFILE(vlan_tcp);
+
+	return ICE_PKT_PROFILE(tcp);
 }
 
 /**
@@ -5716,15 +5645,12 @@ ice_find_dummy_packet(struct ice_adv_lkup_elem *lkups, u16 lkups_cnt,
  *	   structure per protocol header
  * @lkups_cnt: number of protocols
  * @s_rule: stores rule information from the match criteria
- * @dummy_pkt: dummy packet to fill according to filter match criteria
- * @pkt_len: packet length of dummy packet
- * @offsets: offset info for the dummy packet
+ * @profile: dummy packet profile (the template, its size and header offsets)
  */
 static int
 ice_fill_adv_dummy_packet(struct ice_adv_lkup_elem *lkups, u16 lkups_cnt,
 			  struct ice_aqc_sw_rules_elem *s_rule,
-			  const u8 *dummy_pkt, u16 pkt_len,
-			  const struct ice_dummy_pkt_offsets *offsets)
+			  const struct ice_dummy_pkt_profile *profile)
 {
 	u8 *pkt;
 	u16 i;
@@ -5734,9 +5660,10 @@ ice_fill_adv_dummy_packet(struct ice_adv_lkup_elem *lkups, u16 lkups_cnt,
 	 */
 	pkt = s_rule->pdata.lkup_tx_rx.hdr;
 
-	memcpy(pkt, dummy_pkt, pkt_len);
+	memcpy(pkt, profile->pkt, profile->pkt_len);
 
 	for (i = 0; i < lkups_cnt; i++) {
+		const struct ice_dummy_pkt_offsets *offsets = profile->offsets;
 		enum ice_protocol_type type;
 		u16 offset = 0, len = 0, j;
 		bool found = false;
@@ -5821,7 +5748,7 @@ ice_fill_adv_dummy_packet(struct ice_adv_lkup_elem *lkups, u16 lkups_cnt,
 		}
 	}
 
-	s_rule->pdata.lkup_tx_rx.hdr_len = cpu_to_le16(pkt_len);
+	s_rule->pdata.lkup_tx_rx.hdr_len = cpu_to_le16(profile->pkt_len);
 
 	return 0;
 }
@@ -6044,12 +5971,11 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 		 struct ice_rule_query_data *added_entry)
 {
 	struct ice_adv_fltr_mgmt_list_entry *m_entry, *adv_fltr = NULL;
-	u16 rid = 0, i, pkt_len, rule_buf_sz, vsi_handle;
-	const struct ice_dummy_pkt_offsets *pkt_offsets;
 	struct ice_aqc_sw_rules_elem *s_rule = NULL;
+	u16 rid = 0, i, rule_buf_sz, vsi_handle;
+	struct ice_dummy_pkt_profile profile;
 	struct list_head *rule_head;
 	struct ice_switch_info *sw;
-	const u8 *pkt = NULL;
 	u16 word_cnt;
 	u32 act = 0;
 	int status;
@@ -6077,13 +6003,8 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 	if (!word_cnt || word_cnt > ICE_MAX_CHAIN_WORDS)
 		return -EINVAL;
 
-	/* make sure that we can locate a dummy packet */
-	ice_find_dummy_packet(lkups, lkups_cnt, rinfo->tun_type, &pkt, &pkt_len,
-			      &pkt_offsets);
-	if (!pkt) {
-		status = -EINVAL;
-		goto err_ice_add_adv_rule;
-	}
+	/* locate a dummy packet */
+	profile = ice_find_dummy_packet(lkups, lkups_cnt, rinfo->tun_type);
 
 	if (!(rinfo->sw_act.fltr_act == ICE_FWD_TO_VSI ||
 	      rinfo->sw_act.fltr_act == ICE_FWD_TO_Q ||
@@ -6124,7 +6045,7 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 		}
 		return status;
 	}
-	rule_buf_sz = ICE_SW_RULE_RX_TX_NO_HDR_SIZE + pkt_len;
+	rule_buf_sz = ICE_SW_RULE_RX_TX_NO_HDR_SIZE + profile.pkt_len;
 	s_rule = kzalloc(rule_buf_sz, GFP_KERNEL);
 	if (!s_rule)
 		return -ENOMEM;
@@ -6184,8 +6105,7 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 	s_rule->pdata.lkup_tx_rx.recipe_id = cpu_to_le16(rid);
 	s_rule->pdata.lkup_tx_rx.act = cpu_to_le32(act);
 
-	status = ice_fill_adv_dummy_packet(lkups, lkups_cnt, s_rule, pkt,
-					   pkt_len, pkt_offsets);
+	status = ice_fill_adv_dummy_packet(lkups, lkups_cnt, s_rule, &profile);
 	if (status)
 		goto err_ice_add_adv_rule;
 
@@ -6193,7 +6113,7 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 	    rinfo->tun_type != ICE_SW_TUN_AND_NON_TUN) {
 		status = ice_fill_adv_packet_tun(hw, rinfo->tun_type,
 						 s_rule->pdata.lkup_tx_rx.hdr,
-						 pkt_offsets);
+						 profile.offsets);
 		if (status)
 			goto err_ice_add_adv_rule;
 	}
Signature · 1 line
2.35.1
Tested-by on v4 3/5
#6 lore raw

RE: [Intel-wired-lan] [PATCH v4 net-next 3/5] ice: switch: use a struct to pass packet template params

Quoted 4 lines

-----Original Message-----
From: Intel-wired-lan <[email protected]> On Behalf Of Alexander Lobakin
Sent: Monday, March 21, 2022 4:30 PM

To: [email protected]

Quoted 18 lines

Cc: Szapar-Mudlaw, Martyna <[email protected]>; netdev at vger.kernel.org; linux-kernel at vger.kernel.org; Jakub Kicinski <[email protected]>; Paolo Abeni <[email protected]>; David S. Miller <[email protected]>
Subject: [Intel-wired-lan] [PATCH v4 net-next 3/5] ice: switch: use a struct to pass packet template params

ice_find_dummy_packet() contains a lot of boilerplate code and a nice room for copy-paste mistakes.
Instead of passing 3 separate pointers back and forth to get packet template (dummy) params, directly return a structure containing them. Then, use a macro to compose compound literals and avoid code duplication on return path.
Now, dummy packet type/name is needed only once to return a full correct triple pkt-pkt_len-offsets, and those are all one-liners. dummy_ipv4_gtpu_ipv4_packet_offsets is just moved around and renamed (as well as dummy_ipv6_gtp_packet_offsets) with no function changes.

Quoted 7 lines

---
drivers/net/ethernet/intel/ice/ice_switch.c | 268 +++++++-------------

1 file changed, 94 insertions(+), 174 deletions(-)
#7 lore raw

PATCH v4 4/5 ice: switch: use convenience macros to declare dummy pkt templates

In reply to Alexander Lobakin

Declarations of dummy/template packet headers and offsets can be minified to improve readability and simplify adding new templates. Move all the repetitive constructions into two macros and let them do the name and type expansions.
Linewrap removal is yet another positive side effect.

---
 drivers/net/ethernet/intel/ice/ice_switch.c | 133 +++++++++-----------
 1 file changed, 62 insertions(+), 71 deletions(-)
drivers/net/ethernet/intel/ice/ice_switch.c +62 -71
index 4697eb8b4c66..cde9e480ea89 100644
@@ -41,15 +41,22 @@ struct ice_dummy_pkt_profile {
 	u16 pkt_len;
 };
 
+#define ICE_DECLARE_PKT_OFFSETS(type)					\
+	static const struct ice_dummy_pkt_offsets			\
+	ice_dummy_##type##_packet_offsets[]
+
+#define ICE_DECLARE_PKT_TEMPLATE(type)					\
+	static const u8 ice_dummy_##type##_packet[]
+
 #define ICE_PKT_PROFILE(type) ({					\
 	(struct ice_dummy_pkt_profile){					\
-		.pkt		= dummy_##type##_packet,		\
-		.pkt_len	= sizeof(dummy_##type##_packet),	\
-		.offsets	= dummy_##type##_packet_offsets,	\
+		.pkt		= ice_dummy_##type##_packet,		\
+		.pkt_len	= sizeof(ice_dummy_##type##_packet),	\
+		.offsets	= ice_dummy_##type##_packet_offsets,	\
 	};								\
 })
 
-static const struct ice_dummy_pkt_offsets dummy_gre_tcp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(gre_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV4_OFOS,	14 },
@@ -61,7 +68,7 @@ static const struct ice_dummy_pkt_offsets dummy_gre_tcp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_gre_tcp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(gre_tcp) = {
 	0x00, 0x00, 0x00, 0x00,	/* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -96,7 +103,7 @@ static const u8 dummy_gre_tcp_packet[] = {
 	0x00, 0x00, 0x00, 0x00
 };
 
-static const struct ice_dummy_pkt_offsets dummy_gre_udp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(gre_udp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV4_OFOS,	14 },
@@ -108,7 +115,7 @@ static const struct ice_dummy_pkt_offsets dummy_gre_udp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_gre_udp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(gre_udp) = {
 	0x00, 0x00, 0x00, 0x00,	/* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -140,7 +147,7 @@ static const u8 dummy_gre_udp_packet[] = {
 	0x00, 0x08, 0x00, 0x00,
 };
 
-static const struct ice_dummy_pkt_offsets dummy_udp_tun_tcp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(udp_tun_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV4_OFOS,	14 },
@@ -155,7 +162,7 @@ static const struct ice_dummy_pkt_offsets dummy_udp_tun_tcp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_udp_tun_tcp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(udp_tun_tcp) = {
 	0x00, 0x00, 0x00, 0x00,  /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -193,7 +200,7 @@ static const u8 dummy_udp_tun_tcp_packet[] = {
 	0x00, 0x00, 0x00, 0x00
 };
 
-static const struct ice_dummy_pkt_offsets dummy_udp_tun_udp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(udp_tun_udp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV4_OFOS,	14 },
@@ -208,7 +215,7 @@ static const struct ice_dummy_pkt_offsets dummy_udp_tun_udp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_udp_tun_udp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(udp_tun_udp) = {
 	0x00, 0x00, 0x00, 0x00,  /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -243,8 +250,7 @@ static const u8 dummy_udp_tun_udp_packet[] = {
 	0x00, 0x08, 0x00, 0x00,
 };
 
-static const struct ice_dummy_pkt_offsets
-dummy_gre_ipv6_tcp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(gre_ipv6_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV4_OFOS,	14 },
@@ -256,7 +262,7 @@ dummy_gre_ipv6_tcp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_gre_ipv6_tcp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(gre_ipv6_tcp) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -296,8 +302,7 @@ static const u8 dummy_gre_ipv6_tcp_packet[] = {
 	0x00, 0x00, 0x00, 0x00
 };
 
-static const struct ice_dummy_pkt_offsets
-dummy_gre_ipv6_udp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(gre_ipv6_udp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV4_OFOS,	14 },
@@ -309,7 +314,7 @@ dummy_gre_ipv6_udp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_gre_ipv6_udp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(gre_ipv6_udp) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -346,8 +351,7 @@ static const u8 dummy_gre_ipv6_udp_packet[] = {
 	0x00, 0x08, 0x00, 0x00,
 };
 
-static const struct ice_dummy_pkt_offsets
-dummy_udp_tun_ipv6_tcp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(udp_tun_ipv6_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV4_OFOS,	14 },
@@ -362,7 +366,7 @@ dummy_udp_tun_ipv6_tcp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_udp_tun_ipv6_tcp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(udp_tun_ipv6_tcp) = {
 	0x00, 0x00, 0x00, 0x00,  /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -405,8 +409,7 @@ static const u8 dummy_udp_tun_ipv6_tcp_packet[] = {
 	0x00, 0x00, 0x00, 0x00
 };
 
-static const struct ice_dummy_pkt_offsets
-dummy_udp_tun_ipv6_udp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(udp_tun_ipv6_udp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV4_OFOS,	14 },
@@ -421,7 +424,7 @@ dummy_udp_tun_ipv6_udp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_udp_tun_ipv6_udp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(udp_tun_ipv6_udp) = {
 	0x00, 0x00, 0x00, 0x00,  /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -462,7 +465,7 @@ static const u8 dummy_udp_tun_ipv6_udp_packet[] = {
 };
 
 /* offset info for MAC + IPv4 + UDP dummy packet */
-static const struct ice_dummy_pkt_offsets dummy_udp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(udp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV4_OFOS,	14 },
@@ -471,7 +474,7 @@ static const struct ice_dummy_pkt_offsets dummy_udp_packet_offsets[] = {
 };
 
 /* Dummy packet for MAC + IPv4 + UDP */
-static const u8 dummy_udp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(udp) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -491,7 +494,7 @@ static const u8 dummy_udp_packet[] = {
 };
 
 /* offset info for MAC + VLAN + IPv4 + UDP dummy packet */
-static const struct ice_dummy_pkt_offsets dummy_vlan_udp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(vlan_udp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_VLAN_OFOS,	12 },
 	{ ICE_ETYPE_OL,		16 },
@@ -501,7 +504,7 @@ static const struct ice_dummy_pkt_offsets dummy_vlan_udp_packet_offsets[] = {
 };
 
 /* C-tag (801.1Q), IPv4:UDP dummy packet */
-static const u8 dummy_vlan_udp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(vlan_udp) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -523,7 +526,7 @@ static const u8 dummy_vlan_udp_packet[] = {
 };
 
 /* offset info for MAC + IPv4 + TCP dummy packet */
-static const struct ice_dummy_pkt_offsets dummy_tcp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(tcp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV4_OFOS,	14 },
@@ -532,7 +535,7 @@ static const struct ice_dummy_pkt_offsets dummy_tcp_packet_offsets[] = {
 };
 
 /* Dummy packet for MAC + IPv4 + TCP */
-static const u8 dummy_tcp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(tcp) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -555,7 +558,7 @@ static const u8 dummy_tcp_packet[] = {
 };
 
 /* offset info for MAC + VLAN (C-tag, 802.1Q) + IPv4 + TCP dummy packet */
-static const struct ice_dummy_pkt_offsets dummy_vlan_tcp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(vlan_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_VLAN_OFOS,	12 },
 	{ ICE_ETYPE_OL,		16 },
@@ -565,7 +568,7 @@ static const struct ice_dummy_pkt_offsets dummy_vlan_tcp_packet_offsets[] = {
 };
 
 /* C-tag (801.1Q), IPv4:TCP dummy packet */
-static const u8 dummy_vlan_tcp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(vlan_tcp) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -589,7 +592,7 @@ static const u8 dummy_vlan_tcp_packet[] = {
 	0x00, 0x00,	/* 2 bytes for 4 byte alignment */
 };
 
-static const struct ice_dummy_pkt_offsets dummy_tcp_ipv6_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(tcp_ipv6) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV6_OFOS,	14 },
@@ -597,7 +600,7 @@ static const struct ice_dummy_pkt_offsets dummy_tcp_ipv6_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_tcp_ipv6_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(tcp_ipv6) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -625,8 +628,7 @@ static const u8 dummy_tcp_ipv6_packet[] = {
 };
 
 /* C-tag (802.1Q): IPv6 + TCP */
-static const struct ice_dummy_pkt_offsets
-dummy_vlan_tcp_ipv6_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(vlan_tcp_ipv6) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_VLAN_OFOS,	12 },
 	{ ICE_ETYPE_OL,		16 },
@@ -636,7 +638,7 @@ dummy_vlan_tcp_ipv6_packet_offsets[] = {
 };
 
 /* C-tag (802.1Q), IPv6 + TCP dummy packet */
-static const u8 dummy_vlan_tcp_ipv6_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(vlan_tcp_ipv6) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -666,7 +668,7 @@ static const u8 dummy_vlan_tcp_ipv6_packet[] = {
 };
 
 /* IPv6 + UDP */
-static const struct ice_dummy_pkt_offsets dummy_udp_ipv6_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(udp_ipv6) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_ETYPE_OL,		12 },
 	{ ICE_IPV6_OFOS,	14 },
@@ -675,7 +677,7 @@ static const struct ice_dummy_pkt_offsets dummy_udp_ipv6_packet_offsets[] = {
 };
 
 /* IPv6 + UDP dummy packet */
-static const u8 dummy_udp_ipv6_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(udp_ipv6) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -703,8 +705,7 @@ static const u8 dummy_udp_ipv6_packet[] = {
 };
 
 /* C-tag (802.1Q): IPv6 + UDP */
-static const struct ice_dummy_pkt_offsets
-dummy_vlan_udp_ipv6_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(vlan_udp_ipv6) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_VLAN_OFOS,	12 },
 	{ ICE_ETYPE_OL,		16 },
@@ -714,7 +715,7 @@ dummy_vlan_udp_ipv6_packet_offsets[] = {
 };
 
 /* C-tag (802.1Q), IPv6 + UDP dummy packet */
-static const u8 dummy_vlan_udp_ipv6_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(vlan_udp_ipv6) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -741,8 +742,7 @@ static const u8 dummy_vlan_udp_ipv6_packet[] = {
 };
 
 /* Outer IPv4 + Outer UDP + GTP + Inner IPv4 + Inner TCP */
-static const
-struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv4_tcp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(ipv4_gtpu_ipv4_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV4_OFOS,	14 },
 	{ ICE_UDP_OF,		34 },
@@ -752,7 +752,7 @@ struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv4_tcp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_ipv4_gtpu_ipv4_tcp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(ipv4_gtpu_ipv4_tcp) = {
 	0x00, 0x00, 0x00, 0x00, /* Ethernet 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -790,8 +790,7 @@ static const u8 dummy_ipv4_gtpu_ipv4_tcp_packet[] = {
 };
 
 /* Outer IPv4 + Outer UDP + GTP + Inner IPv4 + Inner UDP */
-static const
-struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv4_udp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(ipv4_gtpu_ipv4_udp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV4_OFOS,	14 },
 	{ ICE_UDP_OF,		34 },
@@ -801,7 +800,7 @@ struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv4_udp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_ipv4_gtpu_ipv4_udp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(ipv4_gtpu_ipv4_udp) = {
 	0x00, 0x00, 0x00, 0x00, /* Ethernet 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -836,8 +835,7 @@ static const u8 dummy_ipv4_gtpu_ipv4_udp_packet[] = {
 };
 
 /* Outer IPv6 + Outer UDP + GTP + Inner IPv4 + Inner TCP */
-static const
-struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv6_tcp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(ipv4_gtpu_ipv6_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV4_OFOS,	14 },
 	{ ICE_UDP_OF,		34 },
@@ -847,7 +845,7 @@ struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv6_tcp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_ipv4_gtpu_ipv6_tcp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(ipv4_gtpu_ipv6_tcp) = {
 	0x00, 0x00, 0x00, 0x00, /* Ethernet 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -889,8 +887,7 @@ static const u8 dummy_ipv4_gtpu_ipv6_tcp_packet[] = {
 	0x00, 0x00, /* 2 bytes for 4 byte alignment */
 };
 
-static const
-struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv6_udp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(ipv4_gtpu_ipv6_udp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV4_OFOS,	14 },
 	{ ICE_UDP_OF,		34 },
@@ -900,7 +897,7 @@ struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv6_udp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_ipv4_gtpu_ipv6_udp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(ipv4_gtpu_ipv6_udp) = {
 	0x00, 0x00, 0x00, 0x00, /* Ethernet 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -939,8 +936,7 @@ static const u8 dummy_ipv4_gtpu_ipv6_udp_packet[] = {
 	0x00, 0x00, /* 2 bytes for 4 byte alignment */
 };
 
-static const
-struct ice_dummy_pkt_offsets dummy_ipv6_gtpu_ipv4_tcp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(ipv6_gtpu_ipv4_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV6_OFOS,	14 },
 	{ ICE_UDP_OF,		54 },
@@ -950,7 +946,7 @@ struct ice_dummy_pkt_offsets dummy_ipv6_gtpu_ipv4_tcp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_ipv6_gtpu_ipv4_tcp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(ipv6_gtpu_ipv4_tcp) = {
 	0x00, 0x00, 0x00, 0x00, /* Ethernet 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -992,8 +988,7 @@ static const u8 dummy_ipv6_gtpu_ipv4_tcp_packet[] = {
 	0x00, 0x00, /* 2 bytes for 4 byte alignment */
 };
 
-static const
-struct ice_dummy_pkt_offsets dummy_ipv6_gtpu_ipv4_udp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(ipv6_gtpu_ipv4_udp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV6_OFOS,	14 },
 	{ ICE_UDP_OF,		54 },
@@ -1003,7 +998,7 @@ struct ice_dummy_pkt_offsets dummy_ipv6_gtpu_ipv4_udp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_ipv6_gtpu_ipv4_udp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(ipv6_gtpu_ipv4_udp) = {
 	0x00, 0x00, 0x00, 0x00, /* Ethernet 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -1042,8 +1037,7 @@ static const u8 dummy_ipv6_gtpu_ipv4_udp_packet[] = {
 	0x00, 0x00, /* 2 bytes for 4 byte alignment */
 };
 
-static const
-struct ice_dummy_pkt_offsets dummy_ipv6_gtpu_ipv6_tcp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(ipv6_gtpu_ipv6_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV6_OFOS,	14 },
 	{ ICE_UDP_OF,		54 },
@@ -1053,7 +1047,7 @@ struct ice_dummy_pkt_offsets dummy_ipv6_gtpu_ipv6_tcp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_ipv6_gtpu_ipv6_tcp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(ipv6_gtpu_ipv6_tcp) = {
 	0x00, 0x00, 0x00, 0x00, /* Ethernet 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -1100,8 +1094,7 @@ static const u8 dummy_ipv6_gtpu_ipv6_tcp_packet[] = {
 	0x00, 0x00, /* 2 bytes for 4 byte alignment */
 };
 
-static const
-struct ice_dummy_pkt_offsets dummy_ipv6_gtpu_ipv6_udp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(ipv6_gtpu_ipv6_udp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV6_OFOS,	14 },
 	{ ICE_UDP_OF,		54 },
@@ -1111,7 +1104,7 @@ struct ice_dummy_pkt_offsets dummy_ipv6_gtpu_ipv6_udp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_ipv6_gtpu_ipv6_udp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(ipv6_gtpu_ipv6_udp) = {
 	0x00, 0x00, 0x00, 0x00, /* Ethernet 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -1155,8 +1148,7 @@ static const u8 dummy_ipv6_gtpu_ipv6_udp_packet[] = {
 	0x00, 0x00, /* 2 bytes for 4 byte alignment */
 };
 
-static const
-struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv4_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(ipv4_gtpu_ipv4) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV4_OFOS,	14 },
 	{ ICE_UDP_OF,		34 },
@@ -1164,7 +1156,7 @@ struct ice_dummy_pkt_offsets dummy_ipv4_gtpu_ipv4_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_ipv4_gtpu_ipv4_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(ipv4_gtpu_ipv4) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
@@ -1194,8 +1186,7 @@ static const u8 dummy_ipv4_gtpu_ipv4_packet[] = {
 	0x00, 0x00,
 };
 
-static const
-struct ice_dummy_pkt_offsets dummy_ipv6_gtp_packet_offsets[] = {
+ICE_DECLARE_PKT_OFFSETS(ipv6_gtp) = {
 	{ ICE_MAC_OFOS,		0 },
 	{ ICE_IPV6_OFOS,	14 },
 	{ ICE_UDP_OF,		54 },
@@ -1203,7 +1194,7 @@ struct ice_dummy_pkt_offsets dummy_ipv6_gtp_packet_offsets[] = {
 	{ ICE_PROTOCOL_LAST,	0 },
 };
 
-static const u8 dummy_ipv6_gtp_packet[] = {
+ICE_DECLARE_PKT_TEMPLATE(ipv6_gtp) = {
 	0x00, 0x00, 0x00, 0x00, /* ICE_MAC_OFOS 0 */
 	0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x00,
Signature · 1 line
2.35.1
Tested-by on v4 4/5
#8 lore raw

RE: [Intel-wired-lan] [PATCH v4 net-next 4/5] ice: switch: use convenience macros to declare dummy pkt templates

Quoted 4 lines

-----Original Message-----
From: Intel-wired-lan <[email protected]> On Behalf Of Alexander Lobakin
Sent: Monday, March 21, 2022 4:30 PM

To: [email protected]

Quoted 13 lines

Cc: Szapar-Mudlaw, Martyna <[email protected]>; netdev at vger.kernel.org; linux-kernel at vger.kernel.org; Jakub Kicinski <[email protected]>; Paolo Abeni <[email protected]>; David S. Miller <[email protected]>
Subject: [Intel-wired-lan] [PATCH v4 net-next 4/5] ice: switch: use convenience macros to declare dummy pkt templates

Declarations of dummy/template packet headers and offsets can be minified to improve readability and simplify adding new templates. Move all the repetitive constructions into two macros and let them do the name and type expansions.
Linewrap removal is yet another positive side effect.

Quoted 7 lines

---
drivers/net/ethernet/intel/ice/ice_switch.c | 133 +++++++++-----------

1 file changed, 62 insertions(+), 71 deletions(-)
#9 lore raw

PATCH v4 5/5 ice: switch: convert packet template match code to rodata

In reply to Alexander Lobakin

Trade text size for rodata size and replace tons of nested if-elses to the const mask match based structs. The almost entire ice_find_dummy_packet() now becomes just one plain while-increment loop. The order in ice_dummy_pkt_profiles[] should be same with the if-elses order previously, as masks become less and less strict through the array to follow the original code flow. Apart from removing 80 locs of 4-level if-elses, it brings a solid text size optimization:

add/remove: 0/1 grow/shrink: 1/1 up/down: 2/-1058 (-1056)

Function                                     old     new   delta
ice_fill_adv_dummy_packet                    289     291      +2
ice_adv_add_update_vsi_list                  201       -    -201

ice_add_adv_rule 2950 2093 -857
Total: Before=414512, After=413456, chg -0.25% add/remove: 53/52 grow/shrink: 0/0 up/down: 4660/-3988 (672)
RO Data old new delta
ice_dummy_pkt_profiles - 672 +672
Total: Before=37895, After=38567, chg +1.77%

---
 drivers/net/ethernet/intel/ice/ice_switch.c | 215 ++++++++++----------
 1 file changed, 108 insertions(+), 107 deletions(-)
drivers/net/ethernet/intel/ice/ice_switch.c +108 -107
index cde9e480ea89..ed7130b7abfe 100644
@@ -30,6 +30,19 @@ static const u8 dummy_eth_header[DUMMY_ETH_HDR_LEN] = { 0x2, 0, 0, 0, 0, 0,
 							0x2, 0, 0, 0, 0, 0,
 							0x81, 0, 0, 0};
 
+enum {
+	ICE_PKT_VLAN		= BIT(0),
+	ICE_PKT_OUTER_IPV6	= BIT(1),
+	ICE_PKT_TUN_GTPC	= BIT(2),
+	ICE_PKT_TUN_GTPU	= BIT(3),
+	ICE_PKT_TUN_NVGRE	= BIT(4),
+	ICE_PKT_TUN_UDP		= BIT(5),
+	ICE_PKT_INNER_IPV6	= BIT(6),
+	ICE_PKT_INNER_TCP	= BIT(7),
+	ICE_PKT_INNER_UDP	= BIT(8),
+	ICE_PKT_GTP_NOPAY	= BIT(9),
+};
+
 struct ice_dummy_pkt_offsets {
 	enum ice_protocol_type type;
 	u16 offset; /* ICE_PROTOCOL_LAST indicates end of list */
@@ -38,23 +51,23 @@ struct ice_dummy_pkt_offsets {
 struct ice_dummy_pkt_profile {
 	const struct ice_dummy_pkt_offsets *offsets;
 	const u8 *pkt;
+	u32 match;
 	u16 pkt_len;
 };
 
-#define ICE_DECLARE_PKT_OFFSETS(type)					\
-	static const struct ice_dummy_pkt_offsets			\
+#define ICE_DECLARE_PKT_OFFSETS(type)				\
+	static const struct ice_dummy_pkt_offsets		\
 	ice_dummy_##type##_packet_offsets[]
 
-#define ICE_DECLARE_PKT_TEMPLATE(type)					\
+#define ICE_DECLARE_PKT_TEMPLATE(type)				\
 	static const u8 ice_dummy_##type##_packet[]
 
-#define ICE_PKT_PROFILE(type) ({					\
-	(struct ice_dummy_pkt_profile){					\
-		.pkt		= ice_dummy_##type##_packet,		\
-		.pkt_len	= sizeof(ice_dummy_##type##_packet),	\
-		.offsets	= ice_dummy_##type##_packet_offsets,	\
-	};								\
-})
+#define ICE_PKT_PROFILE(type, m) {				\
+	.match		= (m),					\
+	.pkt		= ice_dummy_##type##_packet,		\
+	.pkt_len	= sizeof(ice_dummy_##type##_packet),	\
+	.offsets	= ice_dummy_##type##_packet_offsets,	\
+}
 
 ICE_DECLARE_PKT_OFFSETS(gre_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
@@ -1220,6 +1233,55 @@ ICE_DECLARE_PKT_TEMPLATE(ipv6_gtp) = {
 	0x00, 0x00,
 };
 
+static const struct ice_dummy_pkt_profile ice_dummy_pkt_profiles[] = {
+	ICE_PKT_PROFILE(ipv6_gtp, ICE_PKT_TUN_GTPU | ICE_PKT_OUTER_IPV6 |
+				  ICE_PKT_GTP_NOPAY),
+	ICE_PKT_PROFILE(ipv6_gtpu_ipv6_udp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_OUTER_IPV6 |
+					    ICE_PKT_INNER_IPV6 |
+					    ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(ipv6_gtpu_ipv6_tcp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_OUTER_IPV6 |
+					    ICE_PKT_INNER_IPV6),
+	ICE_PKT_PROFILE(ipv6_gtpu_ipv4_udp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_OUTER_IPV6 |
+					    ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(ipv6_gtpu_ipv4_tcp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_OUTER_IPV6),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv4, ICE_PKT_TUN_GTPU | ICE_PKT_GTP_NOPAY),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv6_udp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_INNER_IPV6 |
+					    ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv6_tcp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_INNER_IPV6),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv4_udp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv4_tcp, ICE_PKT_TUN_GTPU),
+	ICE_PKT_PROFILE(ipv6_gtp, ICE_PKT_TUN_GTPC | ICE_PKT_OUTER_IPV6),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv4, ICE_PKT_TUN_GTPC),
+	ICE_PKT_PROFILE(gre_ipv6_tcp, ICE_PKT_TUN_NVGRE | ICE_PKT_INNER_IPV6 |
+				      ICE_PKT_INNER_TCP),
+	ICE_PKT_PROFILE(gre_tcp, ICE_PKT_TUN_NVGRE | ICE_PKT_INNER_TCP),
+	ICE_PKT_PROFILE(gre_ipv6_udp, ICE_PKT_TUN_NVGRE | ICE_PKT_INNER_IPV6),
+	ICE_PKT_PROFILE(gre_udp, ICE_PKT_TUN_NVGRE),
+	ICE_PKT_PROFILE(udp_tun_ipv6_tcp, ICE_PKT_TUN_UDP |
+					  ICE_PKT_INNER_IPV6 |
+					  ICE_PKT_INNER_TCP),
+	ICE_PKT_PROFILE(udp_tun_tcp, ICE_PKT_TUN_UDP | ICE_PKT_INNER_TCP),
+	ICE_PKT_PROFILE(udp_tun_ipv6_udp, ICE_PKT_TUN_UDP |
+					  ICE_PKT_INNER_IPV6),
+	ICE_PKT_PROFILE(udp_tun_udp, ICE_PKT_TUN_UDP),
+	ICE_PKT_PROFILE(vlan_udp_ipv6, ICE_PKT_OUTER_IPV6 | ICE_PKT_INNER_UDP |
+				       ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(udp_ipv6, ICE_PKT_OUTER_IPV6 | ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(vlan_udp, ICE_PKT_INNER_UDP | ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(udp, ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(vlan_tcp_ipv6, ICE_PKT_INNER_IPV6 | ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(tcp_ipv6, ICE_PKT_INNER_IPV6),
+	ICE_PKT_PROFILE(vlan_tcp, ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(tcp, 0),
+};
+
 #define ICE_SW_RULE_RX_TX_ETH_HDR_SIZE \
 	(offsetof(struct ice_aqc_sw_rules_elem, pdata.lkup_tx_rx.hdr) + \
 	 (DUMMY_ETH_HDR_LEN * \
@@ -5509,124 +5571,63 @@ ice_add_adv_recipe(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
  *
  * Returns the &ice_dummy_pkt_profile corresponding to these lookup params.
  */
-static struct ice_dummy_pkt_profile
+static const struct ice_dummy_pkt_profile *
 ice_find_dummy_packet(struct ice_adv_lkup_elem *lkups, u16 lkups_cnt,
 		      enum ice_sw_tunnel_type tun_type)
 {
-	bool inner_tcp = false, inner_udp = false, outer_ipv6 = false;
-	bool vlan = false, inner_ipv6 = false, gtp_no_pay = false;
+	const struct ice_dummy_pkt_profile *ret = ice_dummy_pkt_profiles;
+	u32 match = 0;
 	u16 i;
 
+	switch (tun_type) {
+	case ICE_SW_TUN_GTPC:
+		match |= ICE_PKT_TUN_GTPC;
+		break;
+	case ICE_SW_TUN_GTPU:
+		match |= ICE_PKT_TUN_GTPU;
+		break;
+	case ICE_SW_TUN_NVGRE:
+		match |= ICE_PKT_TUN_NVGRE;
+		break;
+	case ICE_SW_TUN_GENEVE:
+	case ICE_SW_TUN_VXLAN:
+		match |= ICE_PKT_TUN_UDP;
+		break;
+	default:
+		break;
+	}
+
 	for (i = 0; i < lkups_cnt; i++) {
 		if (lkups[i].type == ICE_UDP_ILOS)
-			inner_udp = true;
+			match |= ICE_PKT_INNER_UDP;
 		else if (lkups[i].type == ICE_TCP_IL)
-			inner_tcp = true;
+			match |= ICE_PKT_INNER_TCP;
 		else if (lkups[i].type == ICE_IPV6_OFOS)
-			outer_ipv6 = true;
+			match |= ICE_PKT_OUTER_IPV6;
 		else if (lkups[i].type == ICE_VLAN_OFOS)
-			vlan = true;
+			match |= ICE_PKT_VLAN;
 		else if (lkups[i].type == ICE_ETYPE_OL &&
 			 lkups[i].h_u.ethertype.ethtype_id ==
 				cpu_to_be16(ICE_IPV6_ETHER_ID) &&
 			 lkups[i].m_u.ethertype.ethtype_id ==
 				cpu_to_be16(0xFFFF))
-			outer_ipv6 = true;
+			match |= ICE_PKT_OUTER_IPV6;
 		else if (lkups[i].type == ICE_ETYPE_IL &&
 			 lkups[i].h_u.ethertype.ethtype_id ==
 				cpu_to_be16(ICE_IPV6_ETHER_ID) &&
 			 lkups[i].m_u.ethertype.ethtype_id ==
 				cpu_to_be16(0xFFFF))
-			inner_ipv6 = true;
+			match |= ICE_PKT_INNER_IPV6;
 		else if (lkups[i].type == ICE_IPV6_IL)
-			inner_ipv6 = true;
+			match |= ICE_PKT_INNER_IPV6;
 		else if (lkups[i].type == ICE_GTP_NO_PAY)
-			gtp_no_pay = true;
-	}
-
-	if (tun_type == ICE_SW_TUN_GTPU) {
-		if (outer_ipv6) {
-			if (gtp_no_pay) {
-				return ICE_PKT_PROFILE(ipv6_gtp);
-			} else if (inner_ipv6) {
-				if (inner_udp)
-					return ICE_PKT_PROFILE(ipv6_gtpu_ipv6_udp);
-				else
-					return ICE_PKT_PROFILE(ipv6_gtpu_ipv6_tcp);
-			} else {
-				if (inner_udp)
-					return ICE_PKT_PROFILE(ipv6_gtpu_ipv4_udp);
-				else
-					return ICE_PKT_PROFILE(ipv6_gtpu_ipv4_tcp);
-			}
-		} else {
-			if (gtp_no_pay) {
-				return ICE_PKT_PROFILE(ipv4_gtpu_ipv4);
-			} else if (inner_ipv6) {
-				if (inner_udp)
-					return ICE_PKT_PROFILE(ipv4_gtpu_ipv6_udp);
-				else
-					return ICE_PKT_PROFILE(ipv4_gtpu_ipv6_tcp);
-			} else {
-				if (inner_udp)
-					return ICE_PKT_PROFILE(ipv4_gtpu_ipv4_udp);
-				else
-					return ICE_PKT_PROFILE(ipv4_gtpu_ipv4_tcp);
-			}
-		}
+			match |= ICE_PKT_GTP_NOPAY;
 	}
 
-	if (tun_type == ICE_SW_TUN_GTPC) {
-		if (outer_ipv6)
-			return ICE_PKT_PROFILE(ipv6_gtp);
-		else
-			return ICE_PKT_PROFILE(ipv4_gtpu_ipv4);
-	}
-
-	if (tun_type == ICE_SW_TUN_NVGRE) {
-		if (inner_tcp && inner_ipv6)
-			return ICE_PKT_PROFILE(gre_ipv6_tcp);
-		else if (inner_tcp)
-			return ICE_PKT_PROFILE(gre_tcp);
-		else if (inner_ipv6)
-			return ICE_PKT_PROFILE(gre_ipv6_udp);
-		else
-			return ICE_PKT_PROFILE(gre_udp);
-	}
+	while (ret->match && (match & ret->match) != ret->match)
+		ret++;
 
-	if (tun_type == ICE_SW_TUN_VXLAN ||
-	    tun_type == ICE_SW_TUN_GENEVE) {
-		if (inner_tcp && inner_ipv6)
-			return ICE_PKT_PROFILE(udp_tun_ipv6_tcp);
-		else if (inner_tcp)
-			return ICE_PKT_PROFILE(udp_tun_tcp);
-		else if (inner_ipv6)
-			return ICE_PKT_PROFILE(udp_tun_ipv6_udp);
-		else
-			return ICE_PKT_PROFILE(udp_tun_udp);
-	}
-
-	if (inner_udp && !outer_ipv6) {
-		if (vlan)
-			return ICE_PKT_PROFILE(vlan_udp);
-		else
-			return ICE_PKT_PROFILE(udp);
-	} else if (inner_udp && outer_ipv6) {
-		if (vlan)
-			return ICE_PKT_PROFILE(vlan_udp_ipv6);
-		else
-			return ICE_PKT_PROFILE(udp_ipv6);
-	} else if ((inner_tcp && outer_ipv6) || outer_ipv6) {
-		if (vlan)
-			return ICE_PKT_PROFILE(vlan_tcp_ipv6);
-		else
-			return ICE_PKT_PROFILE(tcp_ipv6);
-	}
-
-	if (vlan)
-		return ICE_PKT_PROFILE(vlan_tcp);
-
-	return ICE_PKT_PROFILE(tcp);
+	return ret;
 }
 
 /**
@@ -5963,8 +5964,8 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 {
 	struct ice_adv_fltr_mgmt_list_entry *m_entry, *adv_fltr = NULL;
 	struct ice_aqc_sw_rules_elem *s_rule = NULL;
+	const struct ice_dummy_pkt_profile *profile;
 	u16 rid = 0, i, rule_buf_sz, vsi_handle;
-	struct ice_dummy_pkt_profile profile;
 	struct list_head *rule_head;
 	struct ice_switch_info *sw;
 	u16 word_cnt;
@@ -6036,7 +6037,7 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 		}
 		return status;
 	}
-	rule_buf_sz = ICE_SW_RULE_RX_TX_NO_HDR_SIZE + profile.pkt_len;
+	rule_buf_sz = ICE_SW_RULE_RX_TX_NO_HDR_SIZE + profile->pkt_len;
 	s_rule = kzalloc(rule_buf_sz, GFP_KERNEL);
 	if (!s_rule)
 		return -ENOMEM;
@@ -6096,7 +6097,7 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 	s_rule->pdata.lkup_tx_rx.recipe_id = cpu_to_le16(rid);
 	s_rule->pdata.lkup_tx_rx.act = cpu_to_le32(act);
 
-	status = ice_fill_adv_dummy_packet(lkups, lkups_cnt, s_rule, &profile);
+	status = ice_fill_adv_dummy_packet(lkups, lkups_cnt, s_rule, profile);
 	if (status)
 		goto err_ice_add_adv_rule;
 
@@ -6104,7 +6105,7 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 	    rinfo->tun_type != ICE_SW_TUN_AND_NON_TUN) {
 		status = ice_fill_adv_packet_tun(hw, rinfo->tun_type,
 						 s_rule->pdata.lkup_tx_rx.hdr,
-						 profile.offsets);
+						 profile->offsets);
 		if (status)
 			goto err_ice_add_adv_rule;
 	}
Signature · 1 line
2.35.1
Tested-by on v4 5/5
#10 lore raw

RE: [Intel-wired-lan] [PATCH v4 net-next 5/5] ice: switch: convert packet template match code to rodata

Quoted 4 lines

-----Original Message-----
From: Intel-wired-lan <[email protected]> On Behalf Of Alexander Lobakin
Sent: Monday, March 21, 2022 4:30 PM

To: [email protected]

Quoted 27 lines

Cc: Szapar-Mudlaw, Martyna <[email protected]>; netdev at vger.kernel.org; linux-kernel at vger.kernel.org; Jakub Kicinski <[email protected]>; Paolo Abeni <[email protected]>; David S. Miller <[email protected]>
Subject: [Intel-wired-lan] [PATCH v4 net-next 5/5] ice: switch: convert packet template match code to rodata

Trade text size for rodata size and replace tons of nested if-elses to the const mask match based structs. The almost entire
ice_find_dummy_packet() now becomes just one plain while-increment loop. The order in ice_dummy_pkt_profiles[] should be same with the if-elses order previously, as masks become less and less strict through the array to follow the original code flow.
Apart from removing 80 locs of 4-level if-elses, it brings a solid text size optimization:

add/remove: 0/1 grow/shrink: 1/1 up/down: 2/-1058 (-1056)

Function                                     old     new   delta
ice_fill_adv_dummy_packet                    289     291      +2
ice_adv_add_update_vsi_list                  201       -    -201

ice_add_adv_rule 2950 2093 -857
Total: Before=414512, After=413456, chg -0.25% add/remove: 53/52 grow/shrink: 0/0 up/down: 4660/-3988 (672)
RO Data old new delta
ice_dummy_pkt_profiles - 672 +672
Total: Before=37895, After=38567, chg +1.77%

Quoted 7 lines

---
drivers/net/ethernet/intel/ice/ice_switch.c | 215 ++++++++++----------

1 file changed, 108 insertions(+), 107 deletions(-)
#11 lore raw

RE: [PATCH v4 net-next 5/5] ice: switch: convert packet template match code to rodata

In reply to Alexander Lobakin

Hi Alex,

Quoted 135 lines

-----Original Message-----
From: Lobakin, Alexandr <[email protected]>
Sent: poniedzia?ek, 21 marca 2022 12:00
To: intel-wired-lan at lists.osuosl.org
Cc: Lobakin, Alexandr <[email protected]>; Fijalkowski, Maciej <[email protected]>; Michal Swiatkowski <[email protected]>; Drewek, Wojciech <[email protected]>; Marcin Szycik <[email protected]>; Szapar-Mudlaw, Martyna <[email protected]>; David S. Miller <[email protected]>; Jakub Kicinski <[email protected]>; Paolo Abeni <[email protected]>; netdev at vger.kernel.org; linux- kernel at vger.kernel.org
Subject: [PATCH v4 net-next 5/5] ice: switch: convert packet template match code to rodata

Trade text size for rodata size and replace tons of nested if-elses to the const mask match based structs. The almost entire ice_find_dummy_packet() now becomes just one plain while-increment loop. The order in ice_dummy_pkt_profiles[] should be same with the if-elses order previously, as masks become less and less strict through the array to follow the original code flow. Apart from removing 80 locs of 4-level if-elses, it brings a solid text size optimization:

add/remove: 0/1 grow/shrink: 1/1 up/down: 2/-1058 (-1056)

Function                                     old     new   delta
ice_fill_adv_dummy_packet                    289     291      +2
ice_adv_add_update_vsi_list                  201       -    -201

ice_add_adv_rule 2950 2093 -857
Total: Before=414512, After=413456, chg -0.25% add/remove: 53/52 grow/shrink: 0/0 up/down: 4660/-3988 (672)
RO Data old new delta
ice_dummy_pkt_profiles - 672 +672
Total: Before=37895, After=38567, chg +1.77%

---
 drivers/net/ethernet/intel/ice/ice_switch.c | 215 ++++++++++----------
 1 file changed, 108 insertions(+), 107 deletions(-)
drivers/net/ethernet/intel/ice/ice_switch.c +58 -10
index cde9e480ea89..ed7130b7abfe 100644
@@ -30,6 +30,19 @@ static const u8 dummy_eth_header[DUMMY_ETH_HDR_LEN] = { 0x2, 0, 0, 0, 0, 0,
 							0x2, 0, 0, 0, 0, 0,
 							0x81, 0, 0, 0};
 
+enum {
+	ICE_PKT_VLAN		= BIT(0),
+	ICE_PKT_OUTER_IPV6	= BIT(1),
+	ICE_PKT_TUN_GTPC	= BIT(2),
+	ICE_PKT_TUN_GTPU	= BIT(3),
+	ICE_PKT_TUN_NVGRE	= BIT(4),
+	ICE_PKT_TUN_UDP		= BIT(5),
+	ICE_PKT_INNER_IPV6	= BIT(6),
+	ICE_PKT_INNER_TCP	= BIT(7),
+	ICE_PKT_INNER_UDP	= BIT(8),
+	ICE_PKT_GTP_NOPAY	= BIT(9),
+};
+
 struct ice_dummy_pkt_offsets {
 	enum ice_protocol_type type;
 	u16 offset; /* ICE_PROTOCOL_LAST indicates end of list */
@@ -38,23 +51,23 @@ struct ice_dummy_pkt_offsets {
 struct ice_dummy_pkt_profile {
 	const struct ice_dummy_pkt_offsets *offsets;
 	const u8 *pkt;
+	u32 match;
 	u16 pkt_len;
 };
 
-#define ICE_DECLARE_PKT_OFFSETS(type)					\
-	static const struct ice_dummy_pkt_offsets			\
+#define ICE_DECLARE_PKT_OFFSETS(type)				\
+	static const struct ice_dummy_pkt_offsets		\
 	ice_dummy_##type##_packet_offsets[]
 
-#define ICE_DECLARE_PKT_TEMPLATE(type)					\
+#define ICE_DECLARE_PKT_TEMPLATE(type)				\
 	static const u8 ice_dummy_##type##_packet[]
 
-#define ICE_PKT_PROFILE(type) ({					\
-	(struct ice_dummy_pkt_profile){					\
-		.pkt		= ice_dummy_##type##_packet,		\
-		.pkt_len	= sizeof(ice_dummy_##type##_packet),	\
-		.offsets	= ice_dummy_##type##_packet_offsets,	\
-	};								\
-})
+#define ICE_PKT_PROFILE(type, m) {				\
+	.match		= (m),					\
+	.pkt		= ice_dummy_##type##_packet,		\
+	.pkt_len	= sizeof(ice_dummy_##type##_packet),	\
+	.offsets	= ice_dummy_##type##_packet_offsets,	\
+}
 
 ICE_DECLARE_PKT_OFFSETS(gre_tcp) = {
 	{ ICE_MAC_OFOS,		0 },
@@ -1220,6 +1233,55 @@ ICE_DECLARE_PKT_TEMPLATE(ipv6_gtp) = {
 	0x00, 0x00,
 };
 
+static const struct ice_dummy_pkt_profile ice_dummy_pkt_profiles[] = {
+	ICE_PKT_PROFILE(ipv6_gtp, ICE_PKT_TUN_GTPU | ICE_PKT_OUTER_IPV6 |
+				  ICE_PKT_GTP_NOPAY),
+	ICE_PKT_PROFILE(ipv6_gtpu_ipv6_udp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_OUTER_IPV6 |
+					    ICE_PKT_INNER_IPV6 |
+					    ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(ipv6_gtpu_ipv6_tcp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_OUTER_IPV6 |
+					    ICE_PKT_INNER_IPV6),
+	ICE_PKT_PROFILE(ipv6_gtpu_ipv4_udp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_OUTER_IPV6 |
+					    ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(ipv6_gtpu_ipv4_tcp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_OUTER_IPV6),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv4, ICE_PKT_TUN_GTPU | ICE_PKT_GTP_NOPAY),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv6_udp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_INNER_IPV6 |
+					    ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv6_tcp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_INNER_IPV6),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv4_udp, ICE_PKT_TUN_GTPU |
+					    ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv4_tcp, ICE_PKT_TUN_GTPU),
+	ICE_PKT_PROFILE(ipv6_gtp, ICE_PKT_TUN_GTPC | ICE_PKT_OUTER_IPV6),
+	ICE_PKT_PROFILE(ipv4_gtpu_ipv4, ICE_PKT_TUN_GTPC),
+	ICE_PKT_PROFILE(gre_ipv6_tcp, ICE_PKT_TUN_NVGRE | ICE_PKT_INNER_IPV6 |
+				      ICE_PKT_INNER_TCP),
+	ICE_PKT_PROFILE(gre_tcp, ICE_PKT_TUN_NVGRE | ICE_PKT_INNER_TCP),
+	ICE_PKT_PROFILE(gre_ipv6_udp, ICE_PKT_TUN_NVGRE | ICE_PKT_INNER_IPV6),
+	ICE_PKT_PROFILE(gre_udp, ICE_PKT_TUN_NVGRE),
+	ICE_PKT_PROFILE(udp_tun_ipv6_tcp, ICE_PKT_TUN_UDP |
+					  ICE_PKT_INNER_IPV6 |
+					  ICE_PKT_INNER_TCP),
+	ICE_PKT_PROFILE(udp_tun_tcp, ICE_PKT_TUN_UDP | ICE_PKT_INNER_TCP),

drivers/net/ethernet/intel/ice/ice_switch.c ICE_DECLARE_PKT_TEMPLATE(ipv6_gtp) = { · line 1280

Quoted 10 lines
drivers/net/ethernet/intel/ice/ice_switch.c +10 -0
+	ICE_PKT_PROFILE(udp_tun_ipv6_udp, ICE_PKT_TUN_UDP |
+					  ICE_PKT_INNER_IPV6),
+	ICE_PKT_PROFILE(udp_tun_udp, ICE_PKT_TUN_UDP),
+	ICE_PKT_PROFILE(vlan_udp_ipv6, ICE_PKT_OUTER_IPV6 | ICE_PKT_INNER_UDP |
+				       ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(udp_ipv6, ICE_PKT_OUTER_IPV6 | ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(vlan_udp, ICE_PKT_INNER_UDP | ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(udp, ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(vlan_tcp_ipv6, ICE_PKT_INNER_IPV6 | ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(tcp_ipv6, ICE_PKT_INNER_IPV6),

I think that in both "vlan_tcp_ipv6" and "tcp_ipv6" we should use ICE_PKT_OUTER_IPV6 instead of ICE_PKT_INNER_IPV6. We are not dealing with tunnels in those cases so inner addresses are incorrect here.

Thanks,
Wojtek

Quoted 203 lines
drivers/net/ethernet/intel/ice/ice_switch.c +40 -98
+	ICE_PKT_PROFILE(vlan_tcp, ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(tcp, 0),
+};
+
 #define ICE_SW_RULE_RX_TX_ETH_HDR_SIZE \
 	(offsetof(struct ice_aqc_sw_rules_elem, pdata.lkup_tx_rx.hdr) + \
 	 (DUMMY_ETH_HDR_LEN * \
@@ -5509,124 +5571,63 @@ ice_add_adv_recipe(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
  *
  * Returns the &ice_dummy_pkt_profile corresponding to these lookup params.
  */
-static struct ice_dummy_pkt_profile
+static const struct ice_dummy_pkt_profile *
 ice_find_dummy_packet(struct ice_adv_lkup_elem *lkups, u16 lkups_cnt,
 		      enum ice_sw_tunnel_type tun_type)
 {
-	bool inner_tcp = false, inner_udp = false, outer_ipv6 = false;
-	bool vlan = false, inner_ipv6 = false, gtp_no_pay = false;
+	const struct ice_dummy_pkt_profile *ret = ice_dummy_pkt_profiles;
+	u32 match = 0;
 	u16 i;
 
+	switch (tun_type) {
+	case ICE_SW_TUN_GTPC:
+		match |= ICE_PKT_TUN_GTPC;
+		break;
+	case ICE_SW_TUN_GTPU:
+		match |= ICE_PKT_TUN_GTPU;
+		break;
+	case ICE_SW_TUN_NVGRE:
+		match |= ICE_PKT_TUN_NVGRE;
+		break;
+	case ICE_SW_TUN_GENEVE:
+	case ICE_SW_TUN_VXLAN:
+		match |= ICE_PKT_TUN_UDP;
+		break;
+	default:
+		break;
+	}
+
 	for (i = 0; i < lkups_cnt; i++) {
 		if (lkups[i].type == ICE_UDP_ILOS)
-			inner_udp = true;
+			match |= ICE_PKT_INNER_UDP;
 		else if (lkups[i].type == ICE_TCP_IL)
-			inner_tcp = true;
+			match |= ICE_PKT_INNER_TCP;
 		else if (lkups[i].type == ICE_IPV6_OFOS)
-			outer_ipv6 = true;
+			match |= ICE_PKT_OUTER_IPV6;
 		else if (lkups[i].type == ICE_VLAN_OFOS)
-			vlan = true;
+			match |= ICE_PKT_VLAN;
 		else if (lkups[i].type == ICE_ETYPE_OL &&
 			 lkups[i].h_u.ethertype.ethtype_id ==
 				cpu_to_be16(ICE_IPV6_ETHER_ID) &&
 			 lkups[i].m_u.ethertype.ethtype_id ==
 				cpu_to_be16(0xFFFF))
-			outer_ipv6 = true;
+			match |= ICE_PKT_OUTER_IPV6;
 		else if (lkups[i].type == ICE_ETYPE_IL &&
 			 lkups[i].h_u.ethertype.ethtype_id ==
 				cpu_to_be16(ICE_IPV6_ETHER_ID) &&
 			 lkups[i].m_u.ethertype.ethtype_id ==
 				cpu_to_be16(0xFFFF))
-			inner_ipv6 = true;
+			match |= ICE_PKT_INNER_IPV6;
 		else if (lkups[i].type == ICE_IPV6_IL)
-			inner_ipv6 = true;
+			match |= ICE_PKT_INNER_IPV6;
 		else if (lkups[i].type == ICE_GTP_NO_PAY)
-			gtp_no_pay = true;
-	}
-
-	if (tun_type == ICE_SW_TUN_GTPU) {
-		if (outer_ipv6) {
-			if (gtp_no_pay) {
-				return ICE_PKT_PROFILE(ipv6_gtp);
-			} else if (inner_ipv6) {
-				if (inner_udp)
-					return ICE_PKT_PROFILE(ipv6_gtpu_ipv6_udp);
-				else
-					return ICE_PKT_PROFILE(ipv6_gtpu_ipv6_tcp);
-			} else {
-				if (inner_udp)
-					return ICE_PKT_PROFILE(ipv6_gtpu_ipv4_udp);
-				else
-					return ICE_PKT_PROFILE(ipv6_gtpu_ipv4_tcp);
-			}
-		} else {
-			if (gtp_no_pay) {
-				return ICE_PKT_PROFILE(ipv4_gtpu_ipv4);
-			} else if (inner_ipv6) {
-				if (inner_udp)
-					return ICE_PKT_PROFILE(ipv4_gtpu_ipv6_udp);
-				else
-					return ICE_PKT_PROFILE(ipv4_gtpu_ipv6_tcp);
-			} else {
-				if (inner_udp)
-					return ICE_PKT_PROFILE(ipv4_gtpu_ipv4_udp);
-				else
-					return ICE_PKT_PROFILE(ipv4_gtpu_ipv4_tcp);
-			}
-		}
+			match |= ICE_PKT_GTP_NOPAY;
 	}
 
-	if (tun_type == ICE_SW_TUN_GTPC) {
-		if (outer_ipv6)
-			return ICE_PKT_PROFILE(ipv6_gtp);
-		else
-			return ICE_PKT_PROFILE(ipv4_gtpu_ipv4);
-	}
-
-	if (tun_type == ICE_SW_TUN_NVGRE) {
-		if (inner_tcp && inner_ipv6)
-			return ICE_PKT_PROFILE(gre_ipv6_tcp);
-		else if (inner_tcp)
-			return ICE_PKT_PROFILE(gre_tcp);
-		else if (inner_ipv6)
-			return ICE_PKT_PROFILE(gre_ipv6_udp);
-		else
-			return ICE_PKT_PROFILE(gre_udp);
-	}
+	while (ret->match && (match & ret->match) != ret->match)
+		ret++;
 
-	if (tun_type == ICE_SW_TUN_VXLAN ||
-	    tun_type == ICE_SW_TUN_GENEVE) {
-		if (inner_tcp && inner_ipv6)
-			return ICE_PKT_PROFILE(udp_tun_ipv6_tcp);
-		else if (inner_tcp)
-			return ICE_PKT_PROFILE(udp_tun_tcp);
-		else if (inner_ipv6)
-			return ICE_PKT_PROFILE(udp_tun_ipv6_udp);
-		else
-			return ICE_PKT_PROFILE(udp_tun_udp);
-	}
-
-	if (inner_udp && !outer_ipv6) {
-		if (vlan)
-			return ICE_PKT_PROFILE(vlan_udp);
-		else
-			return ICE_PKT_PROFILE(udp);
-	} else if (inner_udp && outer_ipv6) {
-		if (vlan)
-			return ICE_PKT_PROFILE(vlan_udp_ipv6);
-		else
-			return ICE_PKT_PROFILE(udp_ipv6);
-	} else if ((inner_tcp && outer_ipv6) || outer_ipv6) {
-		if (vlan)
-			return ICE_PKT_PROFILE(vlan_tcp_ipv6);
-		else
-			return ICE_PKT_PROFILE(tcp_ipv6);
-	}
-
-	if (vlan)
-		return ICE_PKT_PROFILE(vlan_tcp);
-
-	return ICE_PKT_PROFILE(tcp);
+	return ret;
 }
 
 /**
@@ -5963,8 +5964,8 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 {
 	struct ice_adv_fltr_mgmt_list_entry *m_entry, *adv_fltr = NULL;
 	struct ice_aqc_sw_rules_elem *s_rule = NULL;
+	const struct ice_dummy_pkt_profile *profile;
 	u16 rid = 0, i, rule_buf_sz, vsi_handle;
-	struct ice_dummy_pkt_profile profile;
 	struct list_head *rule_head;
 	struct ice_switch_info *sw;
 	u16 word_cnt;
@@ -6036,7 +6037,7 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 		}
 		return status;
 	}
-	rule_buf_sz = ICE_SW_RULE_RX_TX_NO_HDR_SIZE + profile.pkt_len;
+	rule_buf_sz = ICE_SW_RULE_RX_TX_NO_HDR_SIZE + profile->pkt_len;
 	s_rule = kzalloc(rule_buf_sz, GFP_KERNEL);
 	if (!s_rule)
 		return -ENOMEM;
@@ -6096,7 +6097,7 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 	s_rule->pdata.lkup_tx_rx.recipe_id = cpu_to_le16(rid);
 	s_rule->pdata.lkup_tx_rx.act = cpu_to_le32(act);
 
-	status = ice_fill_adv_dummy_packet(lkups, lkups_cnt, s_rule, &profile);
+	status = ice_fill_adv_dummy_packet(lkups, lkups_cnt, s_rule, profile);
 	if (status)
 		goto err_ice_add_adv_rule;
 
@@ -6104,7 +6105,7 @@ ice_add_adv_rule(struct ice_hw *hw, struct ice_adv_lkup_elem *lkups,
 	    rinfo->tun_type != ICE_SW_TUN_AND_NON_TUN) {
 		status = ice_fill_adv_packet_tun(hw, rinfo->tun_type,
 						 s_rule->pdata.lkup_tx_rx.hdr,
-						 profile.offsets);
+						 profile->offsets);
 		if (status)
 			goto err_ice_add_adv_rule;
 	}
--

2.35.1

#12 lore raw

Re: [PATCH v4 net-next 5/5] ice: switch: convert packet template match code to rodata

From: Wojciech Drewek <[email protected]>
Date: Tue, 29 Mar 2022 15:12:44 +0000

Quoted 39 lines

Hi Alex,

Quoted 37 lines

-----Original Message-----
From: Lobakin, Alexandr <[email protected]>
Sent: poniedzialek, 21 marca 2022 12:00
To: intel-wired-lan at lists.osuosl.org
Cc: Lobakin, Alexandr <[email protected]>; Fijalkowski, Maciej <[email protected]>; Michal Swiatkowski <[email protected]>; Drewek, Wojciech <[email protected]>; Marcin Szycik <[email protected]>; Szapar-Mudlaw, Martyna <[email protected]>; David S. Miller <[email protected]>; Jakub Kicinski <[email protected]>; Paolo Abeni <[email protected]>; netdev at vger.kernel.org; linux- kernel at vger.kernel.org
Subject: [PATCH v4 net-next 5/5] ice: switch: convert packet template match code to rodata

Trade text size for rodata size and replace tons of nested if-elses to the const mask match based structs. The almost entire ice_find_dummy_packet() now becomes just one plain while-increment loop. The order in ice_dummy_pkt_profiles[] should be same with the if-elses order previously, as masks become less and less strict through the array to follow the original code flow. Apart from removing 80 locs of 4-level if-elses, it brings a solid text size optimization:

add/remove: 0/1 grow/shrink: 1/1 up/down: 2/-1058 (-1056)

Function                                     old     new   delta
ice_fill_adv_dummy_packet                    289     291      +2
ice_adv_add_update_vsi_list                  201       -    -201

ice_add_adv_rule 2950 2093 -857
Total: Before=414512, After=413456, chg -0.25% add/remove: 53/52 grow/shrink: 0/0 up/down: 4660/-3988 (672)
RO Data old new delta
ice_dummy_pkt_profiles - 672 +672
Total: Before=37895, After=38567, chg +1.77%

---
 drivers/net/ethernet/intel/ice/ice_switch.c | 215 ++++++++++----------
 1 file changed, 108 insertions(+), 107 deletions(-)

--- 8< ---

Quoted 8 lines

drivers/net/ethernet/intel/ice/ice_switch.c ICE_DECLARE_PKT_TEMPLATE(ipv6_gtp) = { · line 1280

Quoted 4 lines
drivers/net/ethernet/intel/ice/ice_switch.c +4 -0
+	ICE_PKT_PROFILE(vlan_udp, ICE_PKT_INNER_UDP | ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(udp, ICE_PKT_INNER_UDP),
+	ICE_PKT_PROFILE(vlan_tcp_ipv6, ICE_PKT_INNER_IPV6 | ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(tcp_ipv6, ICE_PKT_INNER_IPV6),

I think that in both "vlan_tcp_ipv6" and "tcp_ipv6" we should use ICE_PKT_OUTER_IPV6 instead of ICE_PKT_INNER_IPV6. We are not dealing with tunnels in those cases so inner addresses are incorrect here.

Oh, some copy'n'paste braino indeed.
I'll send a fixup to Tony in a moment.

Quoted 3 lines

Thanks,
Wojtek

Great catch, thanks for noticing!

Quoted 4 lines

drivers/net/ethernet/intel/ice/ice_switch.c ICE_DECLARE_PKT_TEMPLATE(ipv6_gtp) = { · line 1283

Quoted 4 lines
drivers/net/ethernet/intel/ice/ice_switch.c +3 -0
+	ICE_PKT_PROFILE(vlan_tcp, ICE_PKT_VLAN),
+	ICE_PKT_PROFILE(tcp, 0),
+};

--- 8< ---

Quoted 2 lines
Quoted 2 lines

--
2.35.1

Al