On Wed, Feb 14, 2024 at 09:56:23AM +0100, Laurent Vivier wrote:
> if buffer is not aligned use sum_16b() only on the not aligned

Nit: s/if/If/

> part, and then use csum_avx2() on the remaining part
> 
> Remove unneeded now function csum_unaligned().
> 
> Signed-off-by: Laurent Vivier <lvivier@redhat.com>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

> ---
> 
> Notes:
>     v2:
>       - use ROUND_UP() and sizeof(__m256i)
>       - fix function comment
>       - remove csum_unaligned() and use csum() instead
> 
>  checksum.c | 47 ++++++++++++++++++++++++-----------------------
>  1 file changed, 24 insertions(+), 23 deletions(-)
> 
> diff --git a/checksum.c b/checksum.c
> index f21c9b7a14d1..65486b4625ba 100644
> --- a/checksum.c
> +++ b/checksum.c
> @@ -56,6 +56,8 @@
>  #include <linux/udp.h>
>  #include <linux/icmpv6.h>
>  
> +#include "util.h"
> +
>  /* Checksums are optional for UDP over IPv4, so we usually just set
>   * them to 0.  Change this to 1 to calculate real UDP over IPv4
>   * checksums
> @@ -110,20 +112,7 @@ uint16_t csum_fold(uint32_t sum)
>  	return sum;
>  }
>  
> -/**
> - * csum_unaligned() - Compute TCP/IP-style checksum for not 32-byte aligned data
> - * @buf:	Input data
> - * @len:	Input length
> - * @init:	Initial 32-bit checksum, 0 for no pre-computed checksum
> - *
> - * Return: 16-bit IPv4-style checksum
> - */
> -/* NOLINTNEXTLINE(clang-diagnostic-unknown-attributes) */
> -__attribute__((optimize("-fno-strict-aliasing")))	/* See csum_16b() */
> -uint16_t csum_unaligned(const void *buf, size_t len, uint32_t init)
> -{
> -	return (uint16_t)~csum_fold(sum_16b(buf, len) + init);
> -}
> +uint16_t csum(const void *buf, size_t len, uint32_t init);
>  
>  /**
>   * csum_ip4_header() - Calculate and set IPv4 header checksum
> @@ -132,7 +121,7 @@ uint16_t csum_unaligned(const void *buf, size_t len, uint32_t init)
>  void csum_ip4_header(struct iphdr *ip4h)
>  {
>  	ip4h->check = 0;
> -	ip4h->check = csum_unaligned(ip4h, (size_t)ip4h->ihl * 4, 0);
> +	ip4h->check = csum(ip4h, (size_t)ip4h->ihl * 4, 0);
>  }
>  
>  /**
> @@ -159,7 +148,7 @@ void csum_udp4(struct udphdr *udp4hr,
>  			+ htons(IPPROTO_UDP);
>  		/* Add in partial checksum for the UDP header alone */
>  		psum += sum_16b(udp4hr, sizeof(*udp4hr));
> -		udp4hr->check = csum_unaligned(payload, len, psum);
> +		udp4hr->check = csum(payload, len, psum);
>  	}
>  }
>  
> @@ -178,7 +167,7 @@ void csum_icmp4(struct icmphdr *icmp4hr, const void *payload, size_t len)
>  	/* Partial checksum for ICMP header alone */
>  	psum = sum_16b(icmp4hr, sizeof(*icmp4hr));
>  
> -	icmp4hr->checksum = csum_unaligned(payload, len, psum);
> +	icmp4hr->checksum = csum(payload, len, psum);
>  }
>  
>  /**
> @@ -199,7 +188,7 @@ void csum_udp6(struct udphdr *udp6hr,
>  	udp6hr->check = 0;
>  	/* Add in partial checksum for the UDP header alone */
>  	psum += sum_16b(udp6hr, sizeof(*udp6hr));
> -	udp6hr->check = csum_unaligned(payload, len, psum);
> +	udp6hr->check = csum(payload, len, psum);
>  }
>  
>  /**
> @@ -222,7 +211,7 @@ void csum_icmp6(struct icmp6hdr *icmp6hr,
>  	icmp6hr->icmp6_cksum = 0;
>  	/* Add in partial checksum for the ICMPv6 header alone */
>  	psum += sum_16b(icmp6hr, sizeof(*icmp6hr));
> -	icmp6hr->icmp6_cksum = csum_unaligned(payload, len, psum);
> +	icmp6hr->icmp6_cksum = csum(payload, len, psum);
>  }
>  
>  #ifdef __AVX2__
> @@ -397,17 +386,29 @@ less_than_128_bytes:
>  
>  /**
>   * csum() - Compute TCP/IP-style checksum
> - * @buf:	Input buffer, must be aligned to 32-byte boundary
> + * @buf:	Input buffer
>   * @len:	Input length
>   * @init:	Initial 32-bit checksum, 0 for no pre-computed checksum
>   *
> - * Return: 16-bit folded, complemented checksum sum
> + * Return: 16-bit folded, complemented checksum
>   */
>  /* NOLINTNEXTLINE(clang-diagnostic-unknown-attributes) */
>  __attribute__((optimize("-fno-strict-aliasing")))	/* See csum_16b() */
>  uint16_t csum(const void *buf, size_t len, uint32_t init)
>  {
> -	return (uint16_t)~csum_fold(csum_avx2(buf, len, init));
> +	intptr_t align = ROUND_UP((intptr_t)buf, sizeof(__m256i));
> +	unsigned int pad = align - (intptr_t)buf;
> +
> +	if (len < pad)
> +		pad = len;
> +
> +	if (pad)
> +		init += sum_16b(buf, pad);
> +
> +	if (len > pad)
> +		init = csum_avx2((void *)align, len - pad, init);
> +
> +	return (uint16_t)~csum_fold(init);
>  }
>  
>  #else /* __AVX2__ */
> @@ -424,7 +425,7 @@ uint16_t csum(const void *buf, size_t len, uint32_t init)
>  __attribute__((optimize("-fno-strict-aliasing")))	/* See csum_16b() */
>  uint16_t csum(const void *buf, size_t len, uint32_t init)
>  {
> -	return csum_unaligned(buf, len, init);
> +	return (uint16_t)~csum_fold(sum_16b(buf, len) + init);
>  }
>  
>  #endif /* !__AVX2__ */

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson