Date: Thu, 16 Mar 2017 01:51:56 +0200 From: Konstantin Belousov <kostikbel@gmail.com> To: John Baldwin <jhb@freebsd.org> Cc: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r315323 - head/sys/kern Message-ID: <20170315235156.GM16105@kib.kiev.ua> In-Reply-To: <3167768.s4jIG6eyGp@ralph.baldwin.cx> References: <201703151823.v2FINWN9083379@repo.freebsd.org> <3167768.s4jIG6eyGp@ralph.baldwin.cx>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Mar 15, 2017 at 12:21:00PM -0700, John Baldwin wrote: > On Wednesday, March 15, 2017 06:23:32 PM John Baldwin wrote: > > Author: jhb > > Date: Wed Mar 15 18:23:32 2017 > > New Revision: 315323 > > URL: https://svnweb.freebsd.org/changeset/base/315323 > > > > Log: > > Use UMA_ALIGN_PTR instead of sizeof(void *) for zone alignment. > > > > uma_zcreate()'s alignment argument is supposed to be sizeof(foo) - 1, > > and uma.h provides a set of helper macros for common types. Passing > > sizeof(void *) results in all of the members being misaligned triggering > > unaligned access faults on certain architectures (notably MIPS). > > > > Reported by: brooks > > Obtained from: CheriBSD > > MFC after: 3 days > > Sponsored by: DARPA / AFRL > > > > Modified: > > head/sys/kern/vfs_lookup.c > > We should perhaps add a KASSERT() to uma_zcreate to catch this sort of > thing. Something like KASSERT(powerof2(align + 1)) (assuming there isn't > a real use case for having non-power-of-2 alignments)? > Will you do it ?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20170315235156.GM16105>