From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-wm0-f67.google.com (mail-wm0-f67.google.com [74.125.82.67]) by dpdk.org (Postfix) with ESMTP id 8763628EE for ; Thu, 30 Nov 2017 10:05:06 +0100 (CET) Received: by mail-wm0-f67.google.com with SMTP id f9so10993561wmh.0 for ; Thu, 30 Nov 2017 01:05:06 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=from:to:cc:subject:date:message-id; bh=HkyzVHt5djuprpijDB7U9xssjA6gblj6ieJR4XOE6C4=; b=F3WF4HuwLeIDCHWi/+ewYJx23jvtCnatvpt5meydmuk9/O8C76fXFLUqV6R7YyiLQb 5bUzrIbxCR6zGX0dUOsUXhD7LIXvwJFmoycd/0uegyVixUTOxuFzcGuULX6chqjc2gEt kWqJRJ+JplinmXVe2W4KZP0wUWR9HptWUo4bSPEkjc+NkGX7h3UbVZ47r9/C4My+Uurf Q1Fetd7MwMVi7qvLoTrTqw40zl1/cTlpgC/Nash39yUbyjHbe8c4giFre2uhLLPY2twl EkGpbFfqDU0tGCeqVOkd2X/nyEVT3Xts5TfOdv44QZBtIrZQ+Ds4nCR4GkMEYfv/ZKuW yKVQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id; bh=HkyzVHt5djuprpijDB7U9xssjA6gblj6ieJR4XOE6C4=; b=iaurJ/jS3VBmo/t0LhShCn6reTYecn04iGoprN+VcT8JyNgAlv+ZmViIYhVQYaVIMD NOAfPHURDMV38urKqIc3UjXq23Y9TZlcSEzgMnjdG+Fd+H2uN6CKZyskP7JgT1WmBQ5J YXMpmfsdgRmGan+7y0Xoz/IMOFsJU3XEFHBeaCJ0dxOgCHtWvso86j1VNiESsW2ST+O8 Ric8pe3aPDztSaicYycPBMtnKEZkTapNFj3WiG/DByDBlfsnYJq5AOVlylAe/zbD9qzH Wpa1sfozH9zxK6Xz0AeOCP3zm16H4tQ73aZr17N03JGJEtfW4ostUrMt9cyZkPSRwXTG 2Zpw== X-Gm-Message-State: AJaThX7f3XzDEUrl9NxW9K+YQPM19cyDEM4m37ezXFjr4uzsydTLNdIA pzDQanmcF/KAh/zx9AjqMlrs X-Google-Smtp-Source: AGs4zMbJLzWq2LXJB/ZojcLcdHoqRL4L5LKunryJB82UQC91ATlKzxNemMDkSQiFwKHj3RvD2rm4Aw== X-Received: by 10.80.153.210 with SMTP id n18mr12127373edb.281.1512032704944; Thu, 30 Nov 2017 01:05:04 -0800 (PST) Received: from bra-l27t7p12.vyatta.net ([213.251.34.151]) by smtp.gmail.com with ESMTPSA id x5sm2622230eda.8.2017.11.30.01.05.03 (version=TLS1_2 cipher=ECDHE-RSA-AES128-SHA bits=128/128); Thu, 30 Nov 2017 01:05:04 -0800 (PST) From: alangordondewar@gmail.com X-Google-Original-From: alan.dewar@att.com To: cristian.dumitrescu@intel.com Cc: dev@dpdk.org, Alan Dewar Date: Thu, 30 Nov 2017 09:04:56 +0000 Message-Id: <1512032696-30765-1-git-send-email-alan.dewar@att.com> X-Mailer: git-send-email 2.1.4 Subject: [dpdk-dev] [PATCH v2] sched: fix overflow errors in WRR weighting code X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 30 Nov 2017 09:05:06 -0000 From: Alan Dewar Revised patch - this version fixes an issue when a small wrr_cost is shifted so far right that its value becomes zero. The WRR code calculates the lowest common denominator between the four WRR weights as a uint32_t value and divides the LCD by each of the WRR weights and casts the results as a uint8_t. This casting can cause the ratios of the computed wrr costs to be wrong. For example with WRR weights of 3, 5, 7 and 11, the LCD is computed to be 1155. The WRR costs get computed as: 1155/3 = 385, 1155/5 = 231, 1155/7 = 165, 1155/11 = 105. When the value 385 is cast into an uint8_t it ends up as 129. Rather than casting straight into a uint8_t, this patch shifts the computed WRR costs right so that the largest value is only eight bits wide. In grinder_schedule, the packet length is multiplied by the WRR cost and added to the grinder's wrr_tokens value. The grinder's wrr_tokens field is a uint16_t, so combination of a packet length of 1500 bytes and a wrr cost of 44 will overflow this field on the first packet. This patch increases the width of the grinder's wrr_tokens and wrr_mask fields from uint16_t to uint32_t. In grinder_wrr_store, the remaining tokens in the grinder's wrr_tokens array are copied to the appropriate pipe's wrr_tokens array. However the pipe's wrr_tokens array is only a uint8_t array so unused tokens were quite frequently lost which upsets the balance of traffic across the four WRR queues. This patch increases the width of the pipe's wrr_tokens array from a uint8_t to uint32_t. Signed-off-by: Alan Dewar --- v2 - fixed bug in the wrr_cost calculation code that could result in a zero wrr_cost lib/librte_sched/rte_sched.c | 59 +++++++++++++++++++++++++++++-------- lib/librte_sched/rte_sched_common.h | 15 ++++++++++ 2 files changed, 61 insertions(+), 13 deletions(-) diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c index 7252f85..324743d 100644 --- a/lib/librte_sched/rte_sched.c +++ b/lib/librte_sched/rte_sched.c @@ -130,7 +130,7 @@ struct rte_sched_pipe { uint32_t tc_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; /* Weighted Round Robin (WRR) */ - uint8_t wrr_tokens[RTE_SCHED_QUEUES_PER_PIPE]; + uint32_t wrr_tokens[RTE_SCHED_QUEUES_PER_PIPE]; /* TC oversubscription */ uint32_t tc_ov_credits; @@ -205,8 +205,8 @@ struct rte_sched_grinder { struct rte_mbuf *pkt; /* WRR */ - uint16_t wrr_tokens[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS]; - uint16_t wrr_mask[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS]; + uint32_t wrr_tokens[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS]; + uint32_t wrr_mask[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS]; uint8_t wrr_cost[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS]; }; @@ -542,6 +542,17 @@ rte_sched_time_ms_to_bytes(uint32_t time_ms, uint32_t rate) return time; } +static uint32_t rte_sched_reduce_to_byte(uint32_t value) +{ + uint32_t shift = 0; + + while (value & 0xFFFFFF00) { + value >>= 1; + shift++; + } + return shift; +} + static void rte_sched_port_config_pipe_profile_table(struct rte_sched_port *port, struct rte_sched_port_params *params) { @@ -583,6 +594,8 @@ rte_sched_port_config_pipe_profile_table(struct rte_sched_port *port, struct rte uint32_t wrr_cost[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS]; uint32_t lcd, lcd1, lcd2; uint32_t qindex; + uint32_t low_pos; + uint32_t shift; qindex = j * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS; @@ -594,12 +607,28 @@ rte_sched_port_config_pipe_profile_table(struct rte_sched_port *port, struct rte lcd1 = rte_get_lcd(wrr_cost[0], wrr_cost[1]); lcd2 = rte_get_lcd(wrr_cost[2], wrr_cost[3]); lcd = rte_get_lcd(lcd1, lcd2); + low_pos = rte_min_pos_4_u32(wrr_cost); wrr_cost[0] = lcd / wrr_cost[0]; wrr_cost[1] = lcd / wrr_cost[1]; wrr_cost[2] = lcd / wrr_cost[2]; wrr_cost[3] = lcd / wrr_cost[3]; + shift = rte_sched_reduce_to_byte(wrr_cost[low_pos]); + wrr_cost[0] >>= shift; + wrr_cost[1] >>= shift; + wrr_cost[2] >>= shift; + wrr_cost[3] >>= shift; + + if (wrr_cost[0] == 0) + wrr_cost[0]++; + if (wrr_cost[1] == 0) + wrr_cost[1]++; + if (wrr_cost[2] == 0) + wrr_cost[2]++; + if (wrr_cost[3] == 0) + wrr_cost[3]++; + dst->wrr_cost[qindex] = (uint8_t) wrr_cost[0]; dst->wrr_cost[qindex + 1] = (uint8_t) wrr_cost[1]; dst->wrr_cost[qindex + 2] = (uint8_t) wrr_cost[2]; @@ -1941,15 +1970,19 @@ grinder_wrr_load(struct rte_sched_port *port, uint32_t pos) qindex = tc_index * 4; - grinder->wrr_tokens[0] = ((uint16_t) pipe->wrr_tokens[qindex]) << RTE_SCHED_WRR_SHIFT; - grinder->wrr_tokens[1] = ((uint16_t) pipe->wrr_tokens[qindex + 1]) << RTE_SCHED_WRR_SHIFT; - grinder->wrr_tokens[2] = ((uint16_t) pipe->wrr_tokens[qindex + 2]) << RTE_SCHED_WRR_SHIFT; - grinder->wrr_tokens[3] = ((uint16_t) pipe->wrr_tokens[qindex + 3]) << RTE_SCHED_WRR_SHIFT; + grinder->wrr_tokens[0] = pipe->wrr_tokens[qindex] << + RTE_SCHED_WRR_SHIFT; + grinder->wrr_tokens[1] = pipe->wrr_tokens[qindex + 1] << + RTE_SCHED_WRR_SHIFT; + grinder->wrr_tokens[2] = pipe->wrr_tokens[qindex + 2] << + RTE_SCHED_WRR_SHIFT; + grinder->wrr_tokens[3] = pipe->wrr_tokens[qindex + 3] << + RTE_SCHED_WRR_SHIFT; - grinder->wrr_mask[0] = (qmask & 0x1) * 0xFFFF; - grinder->wrr_mask[1] = ((qmask >> 1) & 0x1) * 0xFFFF; - grinder->wrr_mask[2] = ((qmask >> 2) & 0x1) * 0xFFFF; - grinder->wrr_mask[3] = ((qmask >> 3) & 0x1) * 0xFFFF; + grinder->wrr_mask[0] = (qmask & 0x1) * 0xFFFFFFFF; + grinder->wrr_mask[1] = ((qmask >> 1) & 0x1) * 0xFFFFFFFF; + grinder->wrr_mask[2] = ((qmask >> 2) & 0x1) * 0xFFFFFFFF; + grinder->wrr_mask[3] = ((qmask >> 3) & 0x1) * 0xFFFFFFFF; grinder->wrr_cost[0] = pipe_params->wrr_cost[qindex]; grinder->wrr_cost[1] = pipe_params->wrr_cost[qindex + 1]; @@ -1981,14 +2014,14 @@ static inline void grinder_wrr(struct rte_sched_port *port, uint32_t pos) { struct rte_sched_grinder *grinder = port->grinder + pos; - uint16_t wrr_tokens_min; + uint32_t wrr_tokens_min; grinder->wrr_tokens[0] |= ~grinder->wrr_mask[0]; grinder->wrr_tokens[1] |= ~grinder->wrr_mask[1]; grinder->wrr_tokens[2] |= ~grinder->wrr_mask[2]; grinder->wrr_tokens[3] |= ~grinder->wrr_mask[3]; - grinder->qpos = rte_min_pos_4_u16(grinder->wrr_tokens); + grinder->qpos = rte_min_pos_4_u32(grinder->wrr_tokens); wrr_tokens_min = grinder->wrr_tokens[grinder->qpos]; grinder->wrr_tokens[0] -= wrr_tokens_min; diff --git a/lib/librte_sched/rte_sched_common.h b/lib/librte_sched/rte_sched_common.h index aed144b..a3c6bc2 100644 --- a/lib/librte_sched/rte_sched_common.h +++ b/lib/librte_sched/rte_sched_common.h @@ -77,6 +77,21 @@ rte_min_pos_4_u16(uint16_t *x) return pos0; } +static inline uint32_t +rte_min_pos_4_u32(uint32_t *x) +{ + uint32_t pos0 = 0; + uint32_t pos1 = 2; + + if (x[1] <= x[0]) + pos0 = 1; + if (x[3] <= x[2]) + pos1 = 3; + if (x[pos1] <= x[pos0]) + pos0 = pos1; + + return pos0; +} #endif /* -- 2.1.4