From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga07.intel.com (mga07.intel.com [134.134.136.100]) by dpdk.org (Postfix) with ESMTP id 5A80E3B5 for ; Mon, 4 Sep 2017 23:38:36 +0200 (CEST) Received: from orsmga004.jf.intel.com ([10.7.209.38]) by orsmga105.jf.intel.com with ESMTP; 04 Sep 2017 14:38:34 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.41,476,1498546800"; d="scan'208";a="125511447" Received: from ewright1-mobl3.ger.corp.intel.com (HELO [10.252.24.105]) ([10.252.24.105]) by orsmga004.jf.intel.com with ESMTP; 04 Sep 2017 14:38:33 -0700 To: Markus Theil , dev@dpdk.org Cc: stephen@networkplumber.org References: <1503336825-7700-1-git-send-email-markus.theil@tu-ilmenau.de> <1504549066-11067-1-git-send-email-markus.theil@tu-ilmenau.de> <1504549066-11067-5-git-send-email-markus.theil@tu-ilmenau.de> From: Ferruh Yigit Message-ID: Date: Mon, 4 Sep 2017 22:38:33 +0100 User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:52.0) Gecko/20100101 Thunderbird/52.3.0 MIME-Version: 1.0 In-Reply-To: <1504549066-11067-5-git-send-email-markus.theil@tu-ilmenau.de> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 7bit Subject: Re: [dpdk-dev] [PATCH v6 5/5] igb_uio: MSI IRQ mode, use kernel functions for masking MSI and MSI-X 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: Mon, 04 Sep 2017 21:38:36 -0000 On 9/4/2017 7:17 PM, Markus Theil wrote: > This patch adds MSI IRQ mode and in a way, that should > also work on older kernel versions. The base for my patch > was an attempt to do this in cf705bc36c which was later reverted in > d8ee82745a. Compilation was tested on Linux 3.2, 4.10 and 4.12. > > Furthermore the custom MSI-X mask code is ommited and already existing > kernel APIs are used for MSI and MSI-X. Feedback/small corrections > to the previous patch of this series are also incorporated. > > Signed-off-by: Markus Theil > --- > lib/librte_eal/linuxapp/igb_uio/compat.h | 26 +++------- > lib/librte_eal/linuxapp/igb_uio/igb_uio.c | 81 ++++++++++++++++++------------- > 2 files changed, 56 insertions(+), 51 deletions(-) > > diff --git a/lib/librte_eal/linuxapp/igb_uio/compat.h b/lib/librte_eal/linuxapp/igb_uio/compat.h > index 3825933..67a7ab3 100644 > --- a/lib/librte_eal/linuxapp/igb_uio/compat.h > +++ b/lib/librte_eal/linuxapp/igb_uio/compat.h > @@ -15,24 +15,6 @@ > #define HAVE_PTE_MASK_PAGE_IOMAP > #endif > > -#ifndef PCI_MSIX_ENTRY_SIZE > -#define PCI_MSIX_ENTRY_SIZE 16 > -#define PCI_MSIX_ENTRY_LOWER_ADDR 0 > -#define PCI_MSIX_ENTRY_UPPER_ADDR 4 > -#define PCI_MSIX_ENTRY_DATA 8 > -#define PCI_MSIX_ENTRY_VECTOR_CTRL 12 > -#define PCI_MSIX_ENTRY_CTRL_MASKBIT 1 > -#endif > - > -/* > - * for kernels < 2.6.38 and backported patch that moves MSI-X entry definition > - * to pci_regs.h Those kernels has PCI_MSIX_ENTRY_SIZE defined but not > - * PCI_MSIX_ENTRY_CTRL_MASKBIT > - */ > -#ifndef PCI_MSIX_ENTRY_CTRL_MASKBIT > -#define PCI_MSIX_ENTRY_CTRL_MASKBIT 1 > -#endif > - > #if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 34) && \ > (!(defined(RHEL_RELEASE_CODE) && \ > RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(5, 9))) > @@ -127,3 +109,11 @@ static bool pci_check_and_mask_intx(struct pci_dev *pdev) > #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 8, 0) > #define HAVE_ALLOC_IRQ_VECTORS 1 > #endif > + > +#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 19, 0) > +#define HAVE_PCI_MSI_MASK_IRQ 1 > +#endif > + > +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 37) > +#define HAVE_IRQ_DATA 1 > +#endif > diff --git a/lib/librte_eal/linuxapp/igb_uio/igb_uio.c b/lib/librte_eal/linuxapp/igb_uio/igb_uio.c > index c570eed..b578c4a 100644 > --- a/lib/librte_eal/linuxapp/igb_uio/igb_uio.c > +++ b/lib/librte_eal/linuxapp/igb_uio/igb_uio.c > @@ -91,27 +91,6 @@ static struct attribute *dev_attrs[] = { > static const struct attribute_group dev_attr_grp = { > .attrs = dev_attrs, > }; > -/* > - * It masks the msix on/off of generating MSI-X messages. > - */ > -static void > -igbuio_msix_mask_irq(struct msi_desc *desc, int32_t state) > -{ > - u32 mask_bits = desc->masked; > - unsigned offset = desc->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE + > - PCI_MSIX_ENTRY_VECTOR_CTRL; > - > - if (state != 0) > - mask_bits &= ~PCI_MSIX_ENTRY_CTRL_MASKBIT; > - else > - mask_bits |= PCI_MSIX_ENTRY_CTRL_MASKBIT; > - > - if (mask_bits != desc->masked) { > - writel(mask_bits, desc->mask_base + offset); > - readl(desc->mask_base); > - desc->masked = mask_bits; > - } > -} > > /** > * This is the irqcontrol callback to be registered to uio_info. > @@ -132,21 +111,31 @@ igbuio_pci_irqcontrol(struct uio_info *info, s32 irq_state) > struct rte_uio_pci_dev *udev = info->priv; > struct pci_dev *pdev = udev->pdev; > > - pci_cfg_access_lock(pdev); > - if (udev->mode == RTE_INTR_MODE_LEGACY) > - pci_intx(pdev, !!irq_state); > +#ifdef HAVE_IRQ_DATA > + struct irq_data *irq = irq_get_irq_data(udev->info.irq); > +#else > + unsigned int irq = udev->info.irq; > +#endif > > - else if (udev->mode == RTE_INTR_MODE_MSIX) { > - struct msi_desc *desc; > + pci_cfg_access_lock(pdev); > > -#if (LINUX_VERSION_CODE < KERNEL_VERSION(4, 3, 0)) > - list_for_each_entry(desc, &pdev->msi_list, list) > - igbuio_msix_mask_irq(desc, irq_state); > + if (udev->mode == RTE_INTR_MODE_MSIX || udev->mode == RTE_INTR_MODE_MSI) { > +#ifdef HAVE_PCI_MSI_MASK_IRQ > + if (irq_state == 1) > + pci_msi_unmask_irq(irq); > + else > + pci_msi_mask_irq(irq); > #else > - list_for_each_entry(desc, &pdev->dev.msi_list, list) > - igbuio_msix_mask_irq(desc, irq_state); > + if (irq_state == 1) > + unmask_msi_irq(irq); > + else > + mask_msi_irq(irq); Can you please split this patch into two: - first converts igbuio_msix_mask_irq() into Linux functions, with HAVE_IRQ_DATA and HAVE_PCI_MSI_MASK_IRQ checks - second add MSI support, as it is in this patch Thanks, ferruh