Date: Thu, 23 Jun 2016 23:34:11 +0200 From: Jilles Tjoelker <jilles@stack.nl> To: Brooks Davis <brooks@FreeBSD.org> Cc: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r302092 - in head/lib/libc: aarch64/sys amd64/sys arm/sys i386/sys mips/sys powerpc/sys powerpc64/sys riscv/sys sparc64/sys sys Message-ID: <20160623213411.GB5381@stack.nl> In-Reply-To: <201606222111.u5MLBRHn090827@repo.freebsd.org> References: <201606222111.u5MLBRHn090827@repo.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Jun 22, 2016 at 09:11:27PM +0000, Brooks Davis wrote: > Author: brooks > Date: Wed Jun 22 21:11:27 2016 > New Revision: 302092 > URL: https://svnweb.freebsd.org/changeset/base/302092 > Log: > Replace use of the pipe(2) system call with pipe2(2) with a zero flags > value. > This eliminates the need for machine dependant assembly wrappers for > pipe(2). I like it. > It also make passing an invalid address to pipe(2) return EFAULT rather > than triggering a segfault. Document this behavior (which was already > true for pipe2(2), but undocumented). It is valid to promise [EFAULT] in this specific case since the kernel undoes the call's effects when the copyout fails (by closing the descriptors) and it seems unlikely that we will need a userland wrapper around pipe() or pipe2() that needs to access the created file descriptors. However, POSIX does not require it: passing invalid pointers to functions specified by the standard is undefined behaviour. See r241001 for some man pages where I removed the [EFAULT] condition because it is wrong for both reasons mentioned above (something changes even after copyout failure and userland wrappers dereference pointers). -- Jilles Tjoelker
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20160623213411.GB5381>