From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from dpdk.org (dpdk.org [92.243.14.124]) by inbox.dpdk.org (Postfix) with ESMTP id 853B5A04B5; Mon, 2 Dec 2019 17:09:54 +0100 (CET) Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id 003561BE3D; Mon, 2 Dec 2019 17:09:54 +0100 (CET) Received: from mga18.intel.com (mga18.intel.com [134.134.136.126]) by dpdk.org (Postfix) with ESMTP id DC2CB1BDFD for ; Mon, 2 Dec 2019 17:09:52 +0100 (CET) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga008.jf.intel.com ([10.7.209.65]) by orsmga106.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 02 Dec 2019 08:09:51 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.69,268,1571727600"; d="scan'208";a="204609961" Received: from fyigit-mobl.ger.corp.intel.com (HELO [10.237.221.96]) ([10.237.221.96]) by orsmga008.jf.intel.com with ESMTP; 02 Dec 2019 08:09:50 -0800 To: Alfredo Cardigliano , John McNamara , Marko Kovacevic , Anatoly Burakov Cc: dev@dpdk.org References: <20191015082235.28639-1-cardigliano@ntop.org> <20191015082235.28639-5-cardigliano@ntop.org> From: Ferruh Yigit Autocrypt: addr=ferruh.yigit@intel.com; keydata= mQINBFXZCFABEADCujshBOAaqPZpwShdkzkyGpJ15lmxiSr3jVMqOtQS/sB3FYLT0/d3+bvy qbL9YnlbPyRvZfnP3pXiKwkRoR1RJwEo2BOf6hxdzTmLRtGtwWzI9MwrUPj6n/ldiD58VAGQ +iR1I/z9UBUN/ZMksElA2D7Jgg7vZ78iKwNnd+vLBD6I61kVrZ45Vjo3r+pPOByUBXOUlxp9 GWEKKIrJ4eogqkVNSixN16VYK7xR+5OUkBYUO+sE6etSxCr7BahMPKxH+XPlZZjKrxciaWQb +dElz3Ab4Opl+ZT/bK2huX+W+NJBEBVzjTkhjSTjcyRdxvS1gwWRuXqAml/sh+KQjPV1PPHF YK5LcqLkle+OKTCa82OvUb7cr+ALxATIZXQkgmn+zFT8UzSS3aiBBohg3BtbTIWy51jNlYdy ezUZ4UxKSsFuUTPt+JjHQBvF7WKbmNGS3fCid5Iag4tWOfZoqiCNzxApkVugltxoc6rG2TyX CmI2rP0mQ0GOsGXA3+3c1MCdQFzdIn/5tLBZyKy4F54UFo35eOX8/g7OaE+xrgY/4bZjpxC1 1pd66AAtKb3aNXpHvIfkVV6NYloo52H+FUE5ZDPNCGD0/btFGPWmWRmkPybzColTy7fmPaGz cBcEEqHK4T0aY4UJmE7Ylvg255Kz7s6wGZe6IR3N0cKNv++O7QARAQABtCVGZXJydWggWWln aXQgPGZlcnJ1aC55aWdpdEBpbnRlbC5jb20+iQJUBBMBCgA+AhsDAh4BAheABQsJCAcDBRUK CQgLBRYCAwEAFiEE0jZTh0IuwoTjmYHH+TPrQ98TYR8FAl1meboFCQlupOoACgkQ+TPrQ98T YR9ACBAAv2tomhyxY0Tp9Up7mNGLfEdBu/7joB/vIdqMRv63ojkwr9orQq5V16V/25+JEAD0 60cKodBDM6HdUvqLHatS8fooWRueSXHKYwJ3vxyB2tWDyZrLzLI1jxEvunGodoIzUOtum0Ce gPynnfQCelXBja0BwLXJMplM6TY1wXX22ap0ZViC0m714U5U4LQpzjabtFtjT8qOUR6L7hfy YQ72PBuktGb00UR/N5UrR6GqB0x4W41aZBHXfUQnvWIMmmCrRUJX36hOTYBzh+x86ULgg7H2 1499tA4o6rvE13FiGccplBNWCAIroAe/G11rdoN5NBgYVXu++38gTa/MBmIt6zRi6ch15oLA Ln2vHOdqhrgDuxjhMpG2bpNE36DG/V9WWyWdIRlz3NYPCDM/S3anbHlhjStXHOz1uHOnerXM 1jEjcsvmj1vSyYoQMyRcRJmBZLrekvgZeh7nJzbPHxtth8M7AoqiZ/o/BpYU+0xZ+J5/szWZ aYxxmIRu5ejFf+Wn9s5eXNHmyqxBidpCWvcbKYDBnkw2+Y9E5YTpL0mS0dCCOlrO7gca27ux ybtbj84aaW1g0CfIlUnOtHgMCmz6zPXThb+A8H8j3O6qmPoVqT3qnq3Uhy6GOoH8Fdu2Vchh TWiF5yo+pvUagQP6LpslffufSnu+RKAagkj7/RSuZV25Ag0EV9ZMvgEQAKc0Db17xNqtSwEv mfp4tkddwW9XA0tWWKtY4KUdd/jijYqc3fDD54ESYpV8QWj0xK4YM0dLxnDU2IYxjEshSB1T qAatVWz9WtBYvzalsyTqMKP3w34FciuL7orXP4AibPtrHuIXWQOBECcVZTTOdZYGAzaYzxiA ONzF9eTiwIqe9/oaOjTwTLnOarHt16QApTYQSnxDUQljeNvKYt1lZE/gAUUxNLWsYyTT+22/ vU0GDUahsJxs1+f1yEr+OGrFiEAmqrzpF0lCS3f/3HVTU6rS9cK3glVUeaTF4+1SK5ZNO35p iVQCwphmxa+dwTG/DvvHYCtgOZorTJ+OHfvCnSVjsM4kcXGjJPy3JZmUtyL9UxEbYlrffGPQ I3gLXIGD5AN5XdAXFCjjaID/KR1c9RHd7Oaw0Pdcq9UtMLgM1vdX8RlDuMGPrj5sQrRVbgYH fVU/TQCk1C9KhzOwg4Ap2T3tE1umY/DqrXQgsgH71PXFucVjOyHMYXXugLT8YQ0gcBPHy9mZ qw5mgOI5lCl6d4uCcUT0l/OEtPG/rA1lxz8ctdFBVOQOxCvwRG2QCgcJ/UTn5vlivul+cThi 6ERPvjqjblLncQtRg8izj2qgmwQkvfj+h7Ex88bI8iWtu5+I3K3LmNz/UxHBSWEmUnkg4fJl Rr7oItHsZ0ia6wWQ8lQnABEBAAGJAjwEGAEKACYCGwwWIQTSNlOHQi7ChOOZgcf5M+tD3xNh HwUCXWZ5wAUJB3FgggAKCRD5M+tD3xNhH2O+D/9OEz62YuJQLuIuOfL67eFTIB5/1+0j8Tsu o2psca1PUQ61SZJZOMl6VwNxpdvEaolVdrpnSxUF31kPEvR0Igy8HysQ11pj8AcgH0a9FrvU /8k2Roccd2ZIdpNLkirGFZR7LtRw41Kt1Jg+lafI0efkiHKMT/6D/P1EUp1RxOBNtWGV2hrd 0Yg9ds+VMphHHU69fDH02SwgpvXwG8Qm14Zi5WQ66R4CtTkHuYtA63sS17vMl8fDuTCtvfPF HzvdJLIhDYN3Mm1oMjKLlq4PUdYh68Fiwm+boJoBUFGuregJFlO3hM7uHBDhSEnXQr5mqpPM 6R/7Q5BjAxrwVBisH0yQGjsWlnysRWNfExAE2sRePSl0or9q19ddkRYltl6X4FDUXy2DTXa9 a+Fw4e1EvmcF3PjmTYs9IE3Vc64CRQXkhujcN4ZZh5lvOpU8WgyDxFq7bavFnSS6kx7Tk29/ wNJBp+cf9qsQxLbqhW5kfORuZGecus0TLcmpZEFKKjTJBK9gELRBB/zoN3j41hlEl7uTUXTI JQFLhpsFlEdKLujyvT/aCwP3XWT+B2uZDKrMAElF6ltpTxI53JYi22WO7NH7MR16Fhi4R6vh FHNBOkiAhUpoXRZXaCR6+X4qwA8CwHGqHRBfYFSU/Ulq1ZLR+S3hNj2mbnSx0lBs1eEqe2vh cA== Message-ID: <40e1e2aa-bbbc-9384-a33e-f249b3157104@intel.com> Date: Mon, 2 Dec 2019 16:09:49 +0000 MIME-Version: 1.0 In-Reply-To: <20191015082235.28639-5-cardigliano@ntop.org> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 8bit Subject: Re: [dpdk-dev] [PATCH v2 04/17] net/ionic: register and initialize the adapter 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: , Errors-To: dev-bounces@dpdk.org Sender: "dev" On 10/15/2019 9:22 AM, Alfredo Cardigliano wrote: > Register the Pensando ionic PMD (net_ionic) and define initial probe > and remove callbacks with adapter initialization. > > Signed-off-by: Alfredo Cardigliano > Reviewed-by: Shannon Nelson > --- > doc/guides/nics/features/ionic.ini | 2 + > drivers/net/ionic/Makefile | 3 + > drivers/net/ionic/ionic.h | 63 +++++++++++++ > drivers/net/ionic/ionic_dev.c | 128 ++++++++++++++++++++++++++ > drivers/net/ionic/ionic_dev.h | 138 ++++++++++++++++++++++++++++ > drivers/net/ionic/ionic_ethdev.c | 138 ++++++++++++++++++++++++++++ > drivers/net/ionic/ionic_mac_api.c | 61 +++++++++++++ > drivers/net/ionic/ionic_mac_api.h | 13 +++ > drivers/net/ionic/ionic_main.c | 133 +++++++++++++++++++++++++++ > drivers/net/ionic/ionic_osdep.h | 79 ++++++++++++++++ > drivers/net/ionic/ionic_regs.h | 142 +++++++++++++++++++++++++++++ > drivers/net/ionic/meson.build | 4 + > 12 files changed, 904 insertions(+) > create mode 100644 drivers/net/ionic/ionic.h > create mode 100644 drivers/net/ionic/ionic_dev.c > create mode 100644 drivers/net/ionic/ionic_dev.h > create mode 100644 drivers/net/ionic/ionic_mac_api.c > create mode 100644 drivers/net/ionic/ionic_mac_api.h > create mode 100644 drivers/net/ionic/ionic_main.c > create mode 100644 drivers/net/ionic/ionic_osdep.h > create mode 100644 drivers/net/ionic/ionic_regs.h > <...> > +int > +ionic_dev_setup(struct ionic_adapter *adapter) > +{ > + struct ionic_dev_bar *bar = adapter->bars; > + unsigned int num_bars = adapter->num_bars; > + struct ionic_dev *idev = &adapter->idev; > + uint32_t sig; > + > + /* BAR0: dev_cmd and interrupts */ > + if (num_bars < 1) { > + ionic_init_print(ERR, "No bars found, aborting\n"); > + return -EFAULT; > + } > + > + if (bar->len < IONIC_BAR0_SIZE) { > + ionic_init_print(ERR, > + "Resource bar size %lu too small, aborting\n", > + bar->len); > + return -EFAULT; > + } > + > + idev->dev_info = bar->vaddr + IONIC_BAR0_DEV_INFO_REGS_OFFSET; > + idev->dev_cmd = bar->vaddr + IONIC_BAR0_DEV_CMD_REGS_OFFSET; > + idev->intr_status = bar->vaddr + IONIC_BAR0_INTR_STATUS_OFFSET; > + idev->intr_ctrl = bar->vaddr + IONIC_BAR0_INTR_CTRL_OFFSET; These are causing build error with clang [1], can you please check? [1] .../drivers/net/ionic/ionic_dev.c:33:30: error: arithmetic on a pointer to void is a GNU extension [-Werror,-Wpointer-arith] idev->dev_info = bar->vaddr + IONIC_BAR0_DEV_INFO_REGS_OFFSET; ~~~~~~~~~~ ^ .../drivers/net/ionic/ionic_dev.c:34:29: error: arithmetic on a pointer to void is a GNU extension [-Werror,-Wpointer-arith] idev->dev_cmd = bar->vaddr + IONIC_BAR0_DEV_CMD_REGS_OFFSET; ~~~~~~~~~~ ^ .../drivers/net/ionic/ionic_dev.c:35:33: error: arithmetic on a pointer to void is a GNU extension [-Werror,-Wpointer-arith] idev->intr_status = bar->vaddr + IONIC_BAR0_INTR_STATUS_OFFSET; ~~~~~~~~~~ ^ .../drivers/net/ionic/ionic_dev.c:36:31: error: arithmetic on a pointer to void is a GNU extension [-Werror,-Wpointer-arith] idev->intr_ctrl = bar->vaddr + IONIC_BAR0_INTR_CTRL_OFFSET; ~~~~~~~~~~ ^ <...> > @@ -0,0 +1,138 @@ > +/* SPDX-License-Identifier: GPL-2.0 > + * Copyright(c) 2018-2019 Pensando Systems, Inc. All rights reserved. > + */ > + > +#ifndef _IONIC_DEV_H_ > +#define _IONIC_DEV_H_ > + > +#include "ionic_osdep.h" > +#include "ionic_if.h" > +#include "ionic_regs.h" > + > +/** IONIC_API_VERSION - Version number of this interface. > + * > + * Any interface changes to this interface must also change the version. > + * > + * If drivers are compiled from different sources, > + * they are compatible only if IONIC_API_VERSION is statically the same in both > + * sources. Drivers must have matching values of IONIC_API_VERSION at compile > + * time, to be considered compatible at run time. > + */ > +#define IONIC_API_VERSION "3" Does it make sense to put a table in to the documenation, to document which dpdk version supports which API version? <...> > + * There is no room in struct rte_pci_driver to keep a reference > + * to the adapter, using a static list for the time being. > + */ > +static LIST_HEAD(ionic_pci_adapters_list, ionic_adapter) ionic_pci_adapters = > + LIST_HEAD_INITIALIZER(ionic_pci_adapters); Why this list is used? Will holding the reference in the private data help? > +static rte_spinlock_t ionic_pci_adapters_lock = RTE_SPINLOCK_INITIALIZER; > + > +static int > +eth_ionic_pci_probe(struct rte_pci_driver *pci_drv __rte_unused, > + struct rte_pci_device *pci_dev) > +{ > + struct rte_mem_resource *resource; > + struct ionic_adapter *adapter; > + struct ionic_hw *hw; > + unsigned long i; > + int err; > + > + /* Multi-process not supported */ > + if (rte_eal_process_type() != RTE_PROC_PRIMARY) > + return -EPERM; > + > + ionic_init_print(DEBUG, "Initializing device %s %s", > + pci_dev->device.name, > + rte_eal_process_type() == RTE_PROC_SECONDARY ? > + "[SECONDARY]" : ""); According above check, it can't be secondary. > + > + adapter = rte_zmalloc("ionic", sizeof(*adapter), 0); > + > + if (!adapter) { > + ionic_init_print(ERR, "OOM"); > + return -ENOMEM; > + } > + > + adapter->pci_dev = pci_dev; > + hw = &adapter->hw; > + > + hw->device_id = pci_dev->id.device_id; > + hw->vendor_id = pci_dev->id.vendor_id; > + > + err = ionic_init_mac(hw); > + if (err != 0) { > + ionic_init_print(ERR, "Mac init failed: %d", err); Should some clenaup done here? Same thing for all returns, should they free allocated memory. > + return -EIO; > + } > + > + adapter->is_mgmt_nic = (pci_dev->id.device_id == IONIC_DEV_ID_ETH_MGMT); > + > + adapter->num_bars = 0; > + for (i = 0; i < PCI_MAX_RESOURCE && i < IONIC_BARS_MAX; i++) { > + resource = &pci_dev->mem_resource[i]; > + if (resource->phys_addr == 0 || resource->len == 0) > + continue; > + adapter->bars[adapter->num_bars].vaddr = resource->addr; > + adapter->bars[adapter->num_bars].bus_addr = resource->phys_addr; > + adapter->bars[adapter->num_bars].len = resource->len; > + adapter->num_bars++; > + } > + > + /* Discover ionic dev resources */ > + > + err = ionic_setup(adapter); > + if (err) { > + ionic_init_print(ERR, "Cannot setup device: %d, aborting", err); > + return err; > + } > + > + err = ionic_identify(adapter); > + if (err) { > + ionic_init_print(ERR, "Cannot identify device: %d, aborting", > + err); According dpdk coding convention, new line should have a tap. I can see you are using a mixed format but since this is new code, it is good to have a clean code that applies the coding convention: https://doc.dpdk.org/guides/contributing/coding_style.html For the shared code, shared with other platforms, looks like your 'ionic_if.h' code is like that, we are more flexible to prevent additional overhead of maintaining different versions just for the syntax changes, but for the code that is for the DPDK only please follow the DPDK coding convention. > + return err; > + } > + > + err = ionic_init(adapter); > + if (err) { > + ionic_init_print(ERR, "Cannot init device: %d, aborting", err); > + return err; > + } > + > + rte_spinlock_lock(&ionic_pci_adapters_lock); > + LIST_INSERT_HEAD(&ionic_pci_adapters, adapter, pci_adapters); > + rte_spinlock_unlock(&ionic_pci_adapters_lock); > + > + return 0; > +} > + > +static int > +eth_ionic_pci_remove(struct rte_pci_device *pci_dev) > +{ > + struct ionic_adapter *adapter = NULL; > + > + rte_spinlock_lock(&ionic_pci_adapters_lock); > + LIST_FOREACH(adapter, &ionic_pci_adapters, pci_adapters) { > + if (adapter->pci_dev == pci_dev) > + break; > + > + adapter = NULL; > + } > + if (adapter) > + LIST_REMOVE(adapter, pci_adapters); > + rte_spinlock_unlock(&ionic_pci_adapters_lock); It is possible to get ethdev from pci_dev, and if you strore the adapter in ethdev private data, you can access it without needing list for adapter, please check other pysical device drivers, they have the sample. > + > + if (adapter) > + rte_free(adapter); > + > + return 0; > +} > + > +static struct rte_pci_driver rte_ionic_pmd = { > + .id_table = pci_id_ionic_map, > + .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC, > + .probe = eth_ionic_pci_probe, > + .remove = eth_ionic_pci_remove, > +}; > + > +RTE_PMD_REGISTER_PCI(net_ionic, rte_ionic_pmd); > +RTE_PMD_REGISTER_PCI_TABLE(net_ionic, pci_id_ionic_map); > +RTE_PMD_REGISTER_KMOD_DEP(net_ionic, "* igb_uio | uio_pci_generic | vfio-pci"); > + > RTE_INIT(ionic_init_log) > { > ionic_logtype_init = rte_log_register("pmd.net.ionic.init"); > @@ -14,6 +150,8 @@ RTE_INIT(ionic_init_log) > if (ionic_logtype_init >= 0) > rte_log_set_level(ionic_logtype_init, RTE_LOG_NOTICE); > > + ionic_struct_size_checks(); > + For this check log init function doesn't look like correct place. This fucntion is c constructor, it is called in the beggining of the application, even though there is no ionic device attached at all, why you are making all dpdk users check for it? Also what will happend when check fails? Does it make sense to do this check in probe() and return a fail if checks fail? <...>