From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrew Rybchenko Subject: [PATCH 20/53] net/sfc/base: clarify meaning of Rx desc lbits in PS mode Date: Thu, 16 Nov 2017 08:04:08 +0000 Message-ID: <1510819481-6809-21-git-send-email-arybchenko@solarflare.com> References: <1510819481-6809-1-git-send-email-arybchenko@solarflare.com> Mime-Version: 1.0 Content-Type: text/plain Cc: Andrew Rybchenko To: Return-path: Received: from dispatch1-us1.ppe-hosted.com (dispatch1-us1.ppe-hosted.com [67.231.154.164]) by dpdk.org (Postfix) with ESMTP id 268941B27E for ; Thu, 16 Nov 2017 09:05:00 +0100 (CET) In-Reply-To: <1510819481-6809-1-git-send-email-arybchenko@solarflare.com> List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" From: Andrew Rybchenko There is no point to add mask plus one before mask applying since it still does not help to avoid overflow on subtract. Modulo mask arithmetic works perfectly for unsigned integers of the same type. Signed-off-by: Andrew Rybchenko --- drivers/net/sfc/base/ef10_ev.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-) diff --git a/drivers/net/sfc/base/ef10_ev.c b/drivers/net/sfc/base/ef10_ev.c index eb9ec75..9ed22f8 100644 --- a/drivers/net/sfc/base/ef10_ev.c +++ b/drivers/net/sfc/base/ef10_ev.c @@ -774,7 +774,7 @@ ef10_ev_rx_packed_stream( __in_opt void *arg) { uint32_t label; - uint32_t next_read_lbits; + uint32_t pkt_count_lbits; uint16_t flags; boolean_t should_abort; efx_evq_rxq_state_t *eersp; @@ -782,15 +782,22 @@ ef10_ev_rx_packed_stream( unsigned int current_id; boolean_t new_buffer; - next_read_lbits = EFX_QWORD_FIELD(*eqp, ESF_DZ_RX_DSC_PTR_LBITS); + pkt_count_lbits = EFX_QWORD_FIELD(*eqp, ESF_DZ_RX_DSC_PTR_LBITS); label = EFX_QWORD_FIELD(*eqp, ESF_DZ_RX_QLABEL); new_buffer = EFX_QWORD_FIELD(*eqp, ESF_DZ_RX_EV_ROTATE); flags = 0; eersp = &eep->ee_rxq_state[label]; - pkt_count = (EFX_MASK32(ESF_DZ_RX_DSC_PTR_LBITS) + 1 + - next_read_lbits - eersp->eers_rx_stream_npackets) & + + /* + * RX_DSC_PTR_LBITS has least significant bits of the global + * (not per-buffer) packet counter. It is guaranteed that + * maximum number of completed packets fits in lbits-mask. + * So, modulo lbits-mask arithmetic should be used to calculate + * packet counter increment. + */ + pkt_count = (pkt_count_lbits - eersp->eers_rx_stream_npackets) & EFX_MASK32(ESF_DZ_RX_DSC_PTR_LBITS); eersp->eers_rx_stream_npackets += pkt_count; -- 2.7.4