From: "Zhu, Lingshan" Subject: Re: [PATCH V9 9/9] virtio: Intel IFC VF driver for VDPA Date: Thu, 9 Apr 2020 20:49:44 +0800 Message-ID: <4bd16fa9-5436-4cd4-a565-89e11b9f6f92@intel.com> References: <20200326140125.19794-1-jasowang@redhat.com> <20200326140125.19794-10-jasowang@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 8bit Return-path: In-Reply-To: Content-Language: en-US Sender: linux-kernel-owner@vger.kernel.org To: Jason Wang , Arnd Bergmann Cc: "Michael S. Tsirkin" , "linux-kernel@vger.kernel.org" , kvm list , virtualization@lists.linux-foundation.org, Networking , Jason Gunthorpe , maxime.coquelin@redhat.com, cunming.liang@intel.com, zhihong.wang@intel.com, rob.miller@broadcom.com, xiao.w.wang@intel.com, eperezma@redhat.com, lulu@redhat.com, Parav Pandit , kevin.tian@intel.com, Stefan Hajnoczi , Randy Dunlap , Christoph Hellwig , aadam@redhat.com, Jiri Pirko , shahafs@mellanox.com, hanand@xilinx.com, mhabets@solarflare.com, gdawar@xilinx.com, saugatm@xilinx.com, vmireyno@marvell.com, zhangw List-Id: virtualization@lists.linuxfoundation.org On 4/9/2020 8:43 PM, Jason Wang wrote: > > On 2020/4/9 下午6:41, Arnd Bergmann wrote: >> On Thu, Mar 26, 2020 at 3:08 PM Jason Wang wrote: >>> From: Zhu Lingshan >>> >>> This commit introduced two layers to drive IFC VF: >>> >>> (1) ifcvf_base layer, which handles IFC VF NIC hardware operations and >>>      configurations. >>> >>> (2) ifcvf_main layer, which complies to VDPA bus framework, >>>      implemented device operations for VDPA bus, handles device probe, >>>      bus attaching, vring operations, etc. >>> >>> Signed-off-by: Zhu Lingshan >>> Signed-off-by: Bie Tiwei >>> Signed-off-by: Wang Xiao >>> Signed-off-by: Jason Wang >>> + >>> +#define IFCVF_QUEUE_ALIGNMENT  PAGE_SIZE >>> +#define IFCVF_QUEUE_MAX                32768 >>> +static u16 ifcvf_vdpa_get_vq_align(struct vdpa_device *vdpa_dev) >>> +{ >>> +       return IFCVF_QUEUE_ALIGNMENT; >>> +} >> This fails to build on arm64 with 64kb page size (found in linux-next): >> >> /drivers/vdpa/ifcvf/ifcvf_main.c: In function 'ifcvf_vdpa_get_vq_align': >> arch/arm64/include/asm/page-def.h:17:20: error: conversion from 'long >> unsigned int' to 'u16' {aka 'short unsigned int'} changes value from >> '65536' to '0' [-Werror=overflow] >>     17 | #define PAGE_SIZE  (_AC(1, UL) << PAGE_SHIFT) >>        |                    ^ >> drivers/vdpa/ifcvf/ifcvf_base.h:37:31: note: in expansion of macro >> 'PAGE_SIZE' >>     37 | #define IFCVF_QUEUE_ALIGNMENT PAGE_SIZE >>        |                               ^~~~~~~~~ >> drivers/vdpa/ifcvf/ifcvf_main.c:231:9: note: in expansion of macro >> 'IFCVF_QUEUE_ALIGNMENT' >>    231 |  return IFCVF_QUEUE_ALIGNMENT; >>        |         ^~~~~~~~~~~~~~~~~~~~~ >> >> It's probably good enough to just not allow the driver to be built in >> that >> configuration as it's fairly rare but unfortunately there is no >> simple Kconfig >> symbol for it. > > > Or I think the 64KB alignment is probably more than enough. > > Ling Shan, can we use smaller value here? > > Thanks sure, we just need it page aligned, this value is only used in get_vq_align(). I will try to find a arm64 machine and debug on this issue Thanks! > > >> >> In a similar driver, we did >> >> config VMXNET3 >>          tristate "VMware VMXNET3 ethernet driver" >>          depends on PCI && INET >>          depends on !(PAGE_SIZE_64KB || ARM64_64K_PAGES || \ >>                       IA64_PAGE_SIZE_64KB || MICROBLAZE_64K_PAGES || \ >>                       PARISC_PAGE_SIZE_64KB || PPC_64K_PAGES) >> >> I think we should probably make PAGE_SIZE_64KB a global symbol >> in arch/Kconfig and have it selected by the other symbols so drivers >> like yours can add a dependency for it. >> >>           Arnd >> >