Date: Sun, 26 Jul 2026 01:01:11 +0000 From: Kevin Bowling <kbowling@FreeBSD.org> To: src-committers@FreeBSD.org, dev-commits-src-all@FreeBSD.org, dev-commits-src-main@FreeBSD.org Subject: git: 072e0983d7bc - main - e1000: count TSO wire segments in the AIM counters Message-ID: <6a655c57.32e3a.4118efce@gitrepo.freebsd.org>
index | next in thread | raw e-mail
The branch main has been updated by kbowling: URL: https://cgit.FreeBSD.org/src/commit/?id=072e0983d7bce80356740324973993393e77023a commit 072e0983d7bce80356740324973993393e77023a Author: Kevin Bowling <kbowling@FreeBSD.org> AuthorDate: 2026-07-25 12:33:38 +0000 Commit: Kevin Bowling <kbowling@FreeBSD.org> CommitDate: 2026-07-26 01:00:02 +0000 e1000: count TSO wire segments in the AIM counters The transmit paths billed one packet of ipi_len bytes per request. For TSO that is the whole unsegmented payload, up to 64KB, so the average size the moderation calculation sees is not a size that appears on the wire. Count the segments the hardware will put on the wire and the header each of them carries. Non-TSO accounting is unchanged. MFC after: 1 week --- sys/dev/e1000/em_txrx.c | 21 +++++++++++++++++++-- sys/dev/e1000/igb_txrx.c | 21 +++++++++++++++++++-- 2 files changed, 38 insertions(+), 4 deletions(-) diff --git a/sys/dev/e1000/em_txrx.c b/sys/dev/e1000/em_txrx.c index 6ac41816b043..b2a515027479 100644 --- a/sys/dev/e1000/em_txrx.c +++ b/sys/dev/e1000/em_txrx.c @@ -461,10 +461,27 @@ em_isc_txd_encap(void *arg, if_pkt_info_t pi) first, pidx_last, i); pi->ipi_new_pidx = i; - /* Sent data accounting for AIM */ + /* + * Sent data accounting for AIM. For TSO, ipi_len is the whole + * unsegmented payload, which is not a size the moderation + * calculation can use. Count the segments the hardware will put on + * the wire and the header each of them carries, so that the average + * it sees is a wire packet. + */ + if (do_tso && pi->ipi_tso_segsz != 0) { + u32 hdrlen, segs; + + hdrlen = pi->ipi_ehdrlen + pi->ipi_ip_hlen + pi->ipi_tcp_hlen; + if (pi->ipi_len > hdrlen) { + segs = howmany(pi->ipi_len - hdrlen, pi->ipi_tso_segsz); + txr->tx_bytes += pi->ipi_len + (segs - 1) * hdrlen; + txr->tx_packets += segs; + return (0); + } + } + txr->tx_bytes += pi->ipi_len; ++txr->tx_packets; - return (0); } diff --git a/sys/dev/e1000/igb_txrx.c b/sys/dev/e1000/igb_txrx.c index f6f462cfd577..57f098d2266d 100644 --- a/sys/dev/e1000/igb_txrx.c +++ b/sys/dev/e1000/igb_txrx.c @@ -289,10 +289,27 @@ igb_isc_txd_encap(void *arg, if_pkt_info_t pi) txd->read.cmd_type_len |= htole32(E1000_TXD_CMD_EOP | txd_flags); pi->ipi_new_pidx = i; - /* Sent data accounting for AIM */ + /* + * Sent data accounting for AIM. For TSO, ipi_len is the whole + * unsegmented payload, which is not a size the moderation calculation + * can use. Count the segments the hardware will put on the wire and + * the header each of them carries, so that the average it sees is a + * wire packet. + */ + if ((pi->ipi_csum_flags & CSUM_TSO) && pi->ipi_tso_segsz != 0) { + u32 hdrlen, segs; + + hdrlen = pi->ipi_ehdrlen + pi->ipi_ip_hlen + pi->ipi_tcp_hlen; + if (pi->ipi_len > hdrlen) { + segs = howmany(pi->ipi_len - hdrlen, pi->ipi_tso_segsz); + txr->tx_bytes += pi->ipi_len + (segs - 1) * hdrlen; + txr->tx_packets += segs; + return (0); + } + } + txr->tx_bytes += pi->ipi_len; ++txr->tx_packets; - return (0); }home | help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?6a655c57.32e3a.4118efce>