From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga06.intel.com (mga06.intel.com [134.134.136.31]) by dpdk.org (Postfix) with ESMTP id 51379568A for ; Wed, 27 Mar 2019 00:46:25 +0100 (CET) X-Amp-Result: UNSCANNABLE X-Amp-File-Uploaded: False Received: from fmsmga004.fm.intel.com ([10.253.24.48]) by orsmga104.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 26 Mar 2019 16:46:24 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.60,274,1549958400"; d="scan'208";a="156078788" Received: from ae13-28.jf.intel.com ([10.166.188.62]) by fmsmga004.fm.intel.com with ESMTP; 26 Mar 2019 16:46:24 -0700 Date: Tue, 26 Mar 2019 16:43:59 -0700 From: Jeff Shaw To: Thomas Monjalon Cc: Jeff Shaw , Stephen Hemminger , Anand Rawat , dev@dpdk.org, pallavi.kadam@intel.com, ranjit.menon@intel.com, bruce.richardson@intel.com Message-ID: <20190326234359.GA86266@ae13-28.jf.intel.com> References: <20190306041634.12976-1-anand.rawat@intel.com> <12533294.MXuU50Ibyh@xps> <20190326223406.GA86204@ae13-28.jf.intel.com> <2223434.7BpSH3hZxr@xps> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <2223434.7BpSH3hZxr@xps> User-Agent: Mutt/1.9.2 (2017-12-15) Subject: Re: [dpdk-dev] [PATCH v5 4/8] eal: sys/queue.h implementation for windows 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: Tue, 26 Mar 2019 23:46:26 -0000 On Wed, Mar 27, 2019 at 12:00:49AM +0100, Thomas Monjalon wrote: > 26/03/2019 23:34, Jeff Shaw: > > On Tue, Mar 26, 2019 at 11:23:50PM +0100, Thomas Monjalon wrote: > > > 26/03/2019 22:54, Jeff Shaw: > > > > On Tue, Mar 26, 2019 at 10:47:54PM +0100, Thomas Monjalon wrote: > > > > > 26/03/2019 22:14, Jeff Shaw: > > > > > > On Tue, Mar 26, 2019 at 09:52:57PM +0100, Thomas Monjalon wrote: > > > > > > > Even better would be to get it as a dependency outside of DPDK. > > > > > > > Where this code come from? > > > > > > > How other projects on Windows get it? > > > > > > > > > > > > It comes from FreeBSD 12.0, specifically > > > > > > https://github.com/freebsd/freebsd/blob/releng/12.0/sys/sys/queue.h > > > > > > > > > > > > It has been modified such that only the parts used by DPDK (i.e. TAILQ) are > > > > > > implemented. The other stuff has been deleted. Windows does not have sys/queue.h, > > > > > > so we reproduce it here. > > > > > > > > > > > > Would it better to have this as a dependency outside of DPDK? I think pulling a file > > > > > > from the internet and applying a patch (where we'd have to maintain a patch file > > > > > > inside of DPDK's repo anyway) would be overkill when we just need a few lines of > > > > > > code that will change very infrequently. > > > > > > > > > > We already try to get the libbsd dependency on Linux. > > > > > Why not mandate libbsd for Windows? > > > > > It has this header file and a lot more: > > > > > https://gitlab.freedesktop.org/libbsd/libbsd/blob/master/include/bsd/sys/queue.h > > > > > > > > > > Relying on libbsd may avoid copying other files for Windows port. > > > > > > > > I like that idea, though it doesn't look like libbsd builds on Windows, do you > > > > know of a Windows version or one that doesn't depend on autotools to build? > > > > > > It seems libbsd is not packaged for Windows. > > > May be worth to ask opinions to libbsd maintainers. > > > > > > Please could you list which other headers are required for the Windows port? > > > > For helloworld the only one is sys/queue.h. > > > > The dpdk-draft-windows repo has at least these (non-empty) ones: > > dirent.h > > getopt.h > > net/ethernet.h > > net/socket.h > > netinet/in.h > > netinet/tcp.h > > pthread.h > > rand48.h > > sched.h > > sys/_iovec.h > > sys/_sockaddr_storage.h > > sys/_termios.h > > sys/_types.h > > sys/cdefs.h > > sys/mman.h > > sys/netbsd/queue.h > > sys/queue.h > > sys/sysctl.h > > syslog.h > > termios.h > > unistd.h > > > > There will likely be more as more libraries are identified with dependencies on UNIX-like > > headers. > > I would like we find a good solution for these headers. I agree. I think the EAL is supposed to do this, however the current implementation generally assums a UNIX OS under the EAL. The libbsd might be a possiblity. > How other cross-platform projects are getting such dependencies? One example is Python. I just briefly reviewed the code and they go through great lengths to abstract the OS and implement custom, OS independent layers wherever required (e.g. sockets, getopt). See Modules/posixmodule.c for a 14K LOC example. Another example is Nginx. The underlying OS is always abstracted with disparate implementations for, e.g. unix & windows. See src/os/unix and src/os/win32. An example is the "socket()" call on unix, nginx "core" would call "ngx_socket()" which is a macro that is defined to use "WSASocketW" on windows, and "socket" on unix. > Is Cygwin a solution? I think the goal is to be a native Windows application. From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from dpdk.org (dpdk.org [92.243.14.124]) by dpdk.space (Postfix) with ESMTP id 56697A05D3 for ; Wed, 27 Mar 2019 00:46:29 +0100 (CET) Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id 08A5C1B1A0; Wed, 27 Mar 2019 00:46:28 +0100 (CET) Received: from mga06.intel.com (mga06.intel.com [134.134.136.31]) by dpdk.org (Postfix) with ESMTP id 51379568A for ; Wed, 27 Mar 2019 00:46:25 +0100 (CET) X-Amp-Result: UNSCANNABLE X-Amp-File-Uploaded: False Received: from fmsmga004.fm.intel.com ([10.253.24.48]) by orsmga104.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 26 Mar 2019 16:46:24 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.60,274,1549958400"; d="scan'208";a="156078788" Received: from ae13-28.jf.intel.com ([10.166.188.62]) by fmsmga004.fm.intel.com with ESMTP; 26 Mar 2019 16:46:24 -0700 Date: Tue, 26 Mar 2019 16:43:59 -0700 From: Jeff Shaw To: Thomas Monjalon Cc: Jeff Shaw , Stephen Hemminger , Anand Rawat , dev@dpdk.org, pallavi.kadam@intel.com, ranjit.menon@intel.com, bruce.richardson@intel.com Message-ID: <20190326234359.GA86266@ae13-28.jf.intel.com> References: <20190306041634.12976-1-anand.rawat@intel.com> <12533294.MXuU50Ibyh@xps> <20190326223406.GA86204@ae13-28.jf.intel.com> <2223434.7BpSH3hZxr@xps> MIME-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Disposition: inline In-Reply-To: <2223434.7BpSH3hZxr@xps> User-Agent: Mutt/1.9.2 (2017-12-15) Subject: Re: [dpdk-dev] [PATCH v5 4/8] eal: sys/queue.h implementation for windows 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" Message-ID: <20190326234359.wniB0bwEv1omW7DZAEA67-b8lxk032kGa9M2lyikr-w@z> On Wed, Mar 27, 2019 at 12:00:49AM +0100, Thomas Monjalon wrote: > 26/03/2019 23:34, Jeff Shaw: > > On Tue, Mar 26, 2019 at 11:23:50PM +0100, Thomas Monjalon wrote: > > > 26/03/2019 22:54, Jeff Shaw: > > > > On Tue, Mar 26, 2019 at 10:47:54PM +0100, Thomas Monjalon wrote: > > > > > 26/03/2019 22:14, Jeff Shaw: > > > > > > On Tue, Mar 26, 2019 at 09:52:57PM +0100, Thomas Monjalon wrote: > > > > > > > Even better would be to get it as a dependency outside of DPDK. > > > > > > > Where this code come from? > > > > > > > How other projects on Windows get it? > > > > > > > > > > > > It comes from FreeBSD 12.0, specifically > > > > > > https://github.com/freebsd/freebsd/blob/releng/12.0/sys/sys/queue.h > > > > > > > > > > > > It has been modified such that only the parts used by DPDK (i.e. TAILQ) are > > > > > > implemented. The other stuff has been deleted. Windows does not have sys/queue.h, > > > > > > so we reproduce it here. > > > > > > > > > > > > Would it better to have this as a dependency outside of DPDK? I think pulling a file > > > > > > from the internet and applying a patch (where we'd have to maintain a patch file > > > > > > inside of DPDK's repo anyway) would be overkill when we just need a few lines of > > > > > > code that will change very infrequently. > > > > > > > > > > We already try to get the libbsd dependency on Linux. > > > > > Why not mandate libbsd for Windows? > > > > > It has this header file and a lot more: > > > > > https://gitlab.freedesktop.org/libbsd/libbsd/blob/master/include/bsd/sys/queue.h > > > > > > > > > > Relying on libbsd may avoid copying other files for Windows port. > > > > > > > > I like that idea, though it doesn't look like libbsd builds on Windows, do you > > > > know of a Windows version or one that doesn't depend on autotools to build? > > > > > > It seems libbsd is not packaged for Windows. > > > May be worth to ask opinions to libbsd maintainers. > > > > > > Please could you list which other headers are required for the Windows port? > > > > For helloworld the only one is sys/queue.h. > > > > The dpdk-draft-windows repo has at least these (non-empty) ones: > > dirent.h > > getopt.h > > net/ethernet.h > > net/socket.h > > netinet/in.h > > netinet/tcp.h > > pthread.h > > rand48.h > > sched.h > > sys/_iovec.h > > sys/_sockaddr_storage.h > > sys/_termios.h > > sys/_types.h > > sys/cdefs.h > > sys/mman.h > > sys/netbsd/queue.h > > sys/queue.h > > sys/sysctl.h > > syslog.h > > termios.h > > unistd.h > > > > There will likely be more as more libraries are identified with dependencies on UNIX-like > > headers. > > I would like we find a good solution for these headers. I agree. I think the EAL is supposed to do this, however the current implementation generally assums a UNIX OS under the EAL. The libbsd might be a possiblity. > How other cross-platform projects are getting such dependencies? One example is Python. I just briefly reviewed the code and they go through great lengths to abstract the OS and implement custom, OS independent layers wherever required (e.g. sockets, getopt). See Modules/posixmodule.c for a 14K LOC example. Another example is Nginx. The underlying OS is always abstracted with disparate implementations for, e.g. unix & windows. See src/os/unix and src/os/win32. An example is the "socket()" call on unix, nginx "core" would call "ngx_socket()" which is a macro that is defined to use "WSASocketW" on windows, and "socket" on unix. > Is Cygwin a solution? I think the goal is to be a native Windows application.