DPDK patches and discussions
 help / color / mirror / Atom feed
* [PATCH] net/tap: do not force inline or unroll
@ 2025-04-24 16:07 Stephen Hemminger
  2025-06-04  9:27 ` Thomas Monjalon
  0 siblings, 1 reply; 3+ messages in thread
From: Stephen Hemminger @ 2025-04-24 16:07 UTC (permalink / raw)
  To: dev; +Cc: Stephen Hemminger

The BPF code in TAP device was using outdated advice on how
to write BPF programs. Modern BPF verifier has to do less work
if force inlining and unrolling is not done.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/tap/bpf/tap_rss.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/drivers/net/tap/bpf/tap_rss.c b/drivers/net/tap/bpf/tap_rss.c
index 025b831b5c..a76f4bfcb3 100644
--- a/drivers/net/tap/bpf/tap_rss.c
+++ b/drivers/net/tap/bpf/tap_rss.c
@@ -34,14 +34,12 @@ struct {
  * This is same as rte_softrss_be in lib/hash
  * but loop needs to be setup to match BPF restrictions.
  */
-static __always_inline __u32
+static __u32
 softrss_be(const __u32 *input_tuple, __u32 input_len, const __u32 *key)
 {
 	__u32 i, j, hash = 0;
 
-#pragma unroll
 	for (j = 0; j < input_len; j++) {
-#pragma unroll
 		for (i = 0; i < 32; i++) {
 			if (input_tuple[j] & (1U << (31 - i)))
 				hash ^= key[j] << i | key[j + 1] >> (32 - i);
@@ -54,7 +52,7 @@ softrss_be(const __u32 *input_tuple, __u32 input_len, const __u32 *key)
  * Compute RSS hash for IPv4 packet.
  * return in 0 if RSS not specified
  */
-static __always_inline __u32
+static __u32
 parse_ipv4(const struct __sk_buff *skb, __u32 hash_type, const __u32 *key)
 {
 	struct iphdr iph;
@@ -103,7 +101,7 @@ parse_ipv4(const struct __sk_buff *skb, __u32 hash_type, const __u32 *key)
  * Parse Ipv6 extended headers, update offset and return next proto.
  * returns next proto on success, -1 on malformed header
  */
-static __always_inline int
+static int
 skip_ip6_ext(__u16 proto, const struct __sk_buff *skb, __u32 *off, int *frag)
 {
 	struct ext_hdr {
@@ -150,7 +148,7 @@ skip_ip6_ext(__u16 proto, const struct __sk_buff *skb, __u32 *off, int *frag)
  * Compute RSS hash for IPv6 packet.
  * return in 0 if RSS not specified
  */
-static __always_inline __u32
+static __u32
 parse_ipv6(const struct __sk_buff *skb, __u32 hash_type, const __u32 *key)
 {
 	struct {
@@ -207,7 +205,7 @@ parse_ipv6(const struct __sk_buff *skb, __u32 hash_type, const __u32 *key)
  * Scale value to be into range [0, n)
  * Assumes val is large (ie hash covers whole u32 range)
  */
-static __always_inline __u32
+static __u32
 reciprocal_scale(__u32 val, __u32 n)
 {
 	return (__u32)(((__u64)val * n) >> 32);
-- 
2.47.2


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH] net/tap: do not force inline or unroll
  2025-04-24 16:07 [PATCH] net/tap: do not force inline or unroll Stephen Hemminger
@ 2025-06-04  9:27 ` Thomas Monjalon
  2025-06-04 22:05   ` Stephen Hemminger
  0 siblings, 1 reply; 3+ messages in thread
From: Thomas Monjalon @ 2025-06-04  9:27 UTC (permalink / raw)
  To: Stephen Hemminger; +Cc: dev

24/04/2025 18:07, Stephen Hemminger:
> The BPF code in TAP device was using outdated advice on how
> to write BPF programs. Modern BPF verifier has to do less work
> if force inlining and unrolling is not done.

Why do you keep unroll in IPv6 parsing?



^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH] net/tap: do not force inline or unroll
  2025-06-04  9:27 ` Thomas Monjalon
@ 2025-06-04 22:05   ` Stephen Hemminger
  0 siblings, 0 replies; 3+ messages in thread
From: Stephen Hemminger @ 2025-06-04 22:05 UTC (permalink / raw)
  To: Thomas Monjalon; +Cc: dev

On Wed, 04 Jun 2025 11:27:18 +0200
Thomas Monjalon <thomas@monjalon.net> wrote:

> 24/04/2025 18:07, Stephen Hemminger:
> > The BPF code in TAP device was using outdated advice on how
> > to write BPF programs. Modern BPF verifier has to do less work
> > if force inlining and unrolling is not done.  
> 
> Why do you keep unroll in IPv6 parsing?
> 
> 

Good catch.

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2025-06-04 22:05 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-04-24 16:07 [PATCH] net/tap: do not force inline or unroll Stephen Hemminger
2025-06-04  9:27 ` Thomas Monjalon
2025-06-04 22:05   ` Stephen Hemminger

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).