Date: Fri, 15 Apr 2016 18:14:33 +1000 (EST) From: Bruce Evans <brde@optusnet.com.au> To: Peter Wemm <peter@wemm.org> Cc: svn-src-head@freebsd.org, Pedro Giffuni <pfg@freebsd.org>, src-committers@freebsd.org, svn-src-all@freebsd.org Subject: Re: svn commit: r297974 - in head/sys: i386/i386 i386/ibcs2 x86/x86 Message-ID: <20160415171314.F6391@besplex.bde.org> In-Reply-To: <208673032.ZrzIEQ39cc@overcee.wemm.org> References: <201604141704.u3EH47Vv063976@repo.freebsd.org> <570FCFE5.3010304@FreeBSD.org> <208673032.ZrzIEQ39cc@overcee.wemm.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 14 Apr 2016, Peter Wemm wrote: > On Thursday, April 14, 2016 12:14:13 PM Pedro Giffuni wrote: >> On 04/14/16 12:04, Pedro F. Giffuni wrote: >>> Author: pfg >>> Date: Thu Apr 14 17:04:06 2016 >>> New Revision: 297974 >>> URL: https://svnweb.freebsd.org/changeset/base/297974 >>> >>> Log: >>> x86: for pointers replace 0 with NULL. >>> >>> These are mostly cosmetical, no functional change. >>> >>> Found with devel/coccinelle. >>> >>> Modified: >>> head/sys/i386/i386/db_disasm.c >>> head/sys/i386/i386/pmap.c >>> head/sys/i386/ibcs2/imgact_coff.c >>> head/sys/x86/x86/nexus.c >> >> ... >> >>> Modified: head/sys/i386/i386/pmap.c >>> ========================================================================== >>> ==== --- head/sys/i386/i386/pmap.c Thu Apr 14 16:32:27 2016 > (r297973) >>> +++ head/sys/i386/i386/pmap.c Thu Apr 14 17:04:06 2016 (r297974) >>> @@ -269,15 +269,15 @@ pt_entry_t *CMAP3; >>> >>> static pd_entry_t *KPTD; >>> caddr_t ptvmmap = 0; >>> caddr_t CADDR3; >>> >>> -struct msgbuf *msgbufp = 0; >>> +struct msgbuf *msgbufp = NULL; >>> >>> /* >>> >>> * Crashdump maps. >>> */ >>> >>> static caddr_t crashdumpmap; >>> >>> -static pt_entry_t *PMAP1 = 0, *PMAP2; >>> -static pt_entry_t *PADDR1 = 0, *PADDR2; >>> +static pt_entry_t *PMAP1 = NULL, *PMAP2; >>> +static pt_entry_t *PADDR1 = NULL, *PADDR2; >>> >>> #ifdef SMP >>> static int PMAP1cpu; >>> static int PMAP1changedcpu; >> >> Hmm .. being static, there is no need to initialize these. > > Several eons ago, at least some of these were initialized to force them into > the data section so that they had known or safe values before the bss zero > pass. I don't know if that was ever an issue on freebsd, or just the upstream > code. You'd have to look well back into ancient 2.0 or earlier vintage code. > > I have a vague memory that our early a.out kernel had to zero its own bss > because the early a.out boot blocks didn't, and these variables would have > been caught in the crossfire. Or something.. I fixed this in 2003. Much later than I thought or could remember without the commit logs. Someone named peter removed clearing of the bss in locore to work around bugs exposed by gcc changes. This broke booting without loader, so someone named bde put back the clearing in the right place in locore so that it didn't clobber variables like cpu_id. It was only a few variables like cpu_id which were affected, and the problem was changing them from nonzero to 0, not setting them to 0. > In any case, I'd be surprised if the compiler didn't put them in the bss > section these days anyway. At least without cc -ffreestanding, anyway. gcc apparently started doing this as late as 2003. -fno-initialized-in-bss can be used to restore the old behaviour, but we never used it. Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20160415171314.F6391>