From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Qiu, Michael" Subject: Re: [PATCH 0/3] Enable uio_pci_generic support Date: Thu, 29 Jan 2015 09:34:55 +0000 Message-ID: <533710CFB86FA344BFBF2D6802E60286CCFD9C@SHSMSX101.ccr.corp.intel.com> References: <1422523699-17181-1-git-send-email-danny.zhou@intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable To: "Zhou, Danny" , "dev-VfR2kkLFssw@public.gmane.org" Return-path: Content-Language: en-US List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces-VfR2kkLFssw@public.gmane.org Sender: "dev" On 1/29/2015 5:28 PM, Danny Zhou wrote:=0A= > Linux kernel provides UIO as well as VFIO mechanism to support writing us= er=0A= > space device driver. Comparing to UIO which is available since 2.6.32 ker= nel,=0A= > the VFIO is introduced into kernel since version 3.6.0 with better interr= upt=0A= > and memory protection (build on top of Intel VT-d technology) supports.= =0A= =0A= I would like to say that, VFIO is not only for X86.=0A= =0A= Thanks,=0A= Michael=0A= > Basically, UIO and VFIO do two common things below:=0A= > 1) Map PCIe device's I/O memory space to user space driver=0A= > 2) Support device interrupt notification mechanism that notifies user spa= ce=0A= > driver/application when a device interrupt triggers.=0A= >=0A= > To run an DPDK application and make use of VFIO, two in_kernel modules=0A= > vfio and vfio-pci module must be loaded. To use UIO, a DPDK kernel=0A= > module igb_uio, which was there since DPDK is invented, must be loaded to= =0A= > attach to in_kernel uio module. As an solution to deprecate igb_uio, =0A= > this patch serials leverage the uio_pci_generic in_kernel module to suppo= rt=0A= > DPDK user space PMD in a generic fashion (similar to how VFIO works), to= =0A= > remove user space DPDK dependency on GPL code igb_uio in kernel. =0A= >=0A= > Example to bind Network Ports to uio_pci_generic:=0A= > modprobe uio=0A= > modprobe uio_pci_generic=0A= > /* to bind device 08:00.0, to the uio_pci_generic driver */=0A= > ./tools/dpdk_nic_bind.py -b uio_pci_generic 08:00.0=0A= >=0A= > Note: this patch set does not remove igb_uio support due to igb_uio suppo= rts=0A= > creating maximum number of SR-IOV VFs (Virtual Functions) by using max_vf= s =0A= > kernel parameter on older kernels (kernel 3.7.x and below).=0A= > Specifically, igb_uio explicitly calls pci_enable_sriov() to create VFs, = while=0A= > it is not invoked in either uio or uio_pci_generic kernel modules. On ker= nel 3.8.x=0A= > and above, user can use the standard sysfs to enable VFs. For examples:= =0A= >=0A= > #echo $num_vf_enabled > /sys/class/net/$dev/device/sriov_numvfs //enabl= e VFs=0A= > #echo 0 > /sys/class/net/$dev/device/sriov_numvfs //disab= le VFs=0A= >=0A= > Danny Zhou (3):=0A= > eal: add interrupt enable/disable routines for uio_pci_generic=0A= > eal: enable uio_pci_generic support=0A= > tools: enable binding NIC device to uio_pci_generic=0A= >=0A= > lib/librte_eal/common/include/rte_pci.h | 1 +=0A= > lib/librte_eal/linuxapp/eal/eal_interrupts.c | 68 +++++--=0A= > lib/librte_eal/linuxapp/eal/eal_pci.c | 2 +-=0A= > lib/librte_eal/linuxapp/eal/eal_pci_uio.c | 202 +++++++++++----= ------=0A= > .../linuxapp/eal/include/exec-env/rte_interrupts.h | 10 +-=0A= > tools/dpdk_nic_bind.py | 2 +-=0A= > 6 files changed, 172 insertions(+), 113 deletions(-)=0A= >=0A= =0A=