linux-kernel - [PATCH v2] powerpc/32: Discard ffs() function and use builtin

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-Id: <20161122114444.D0CE8678CE@localhost.localdomain>
Date:   Tue, 22 Nov 2016 12:44:44 +0100 (CET)
From:   Christophe Leroy <christophe.leroy@....fr>
To:     Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Paul Mackerras <paulus@...ba.org>,
        Michael Ellerman <mpe@...erman.id.au>,
        Scott Wood <oss@...error.net>
Cc:     linux-kernel@...r.kernel.org, linuxppc-dev@...ts.ozlabs.org
Subject: [PATCH v2] powerpc/32: Discard ffs() function and use builtin_ffs
 instead

With the ffs() function as defined in arch/powerpc/include/asm/bitops.h
GCC will not optimise the code in case of constant parameter, as shown
by the small exemple below.

int ffs_test(void)
{
	return 4 << ffs(31);
}

c000b8f4 <ffs_test>:
c000b8f4:	38 00 00 01 	li      r0,1
c000b8f8:	38 60 00 04 	li      r3,4
c000b8fc:	7c 00 00 34 	cntlzw  r0,r0
c000b900:	20 00 00 20 	subfic  r0,r0,32
c000b904:	7c 63 00 30 	slw     r3,r3,r0
c000b908:	4e 80 00 20 	blr

With this patch, the same function will compile as follows:

c000b8f4 <ffs_test>:
c000b8f4:	38 60 00 08 	li      r3,8
c000b8f8:	4e 80 00 20 	blr

With non constant parameters, the generated code is identical on PPC32
without and with the patch:

int ffs_test2(long x)
{
	return ffs(x);
}

Before:
c000b90c <ffs_test2>:
c000b90c:	7c 03 00 d0 	neg     r0,r3
c000b910:	7c 03 18 38 	and     r3,r0,r3
c000b914:	7c 63 00 34 	cntlzw  r3,r3
c000b918:	20 63 00 20 	subfic  r3,r3,32
c000b91c:	4e 80 00 20 	blr

After:
c000b8fc <ffs_test2>:
c000b8fc:	7c 03 00 d0 	neg     r0,r3
c000b900:	7c 03 18 38 	and     r3,r0,r3
c000b904:	7c 63 00 34 	cntlzw  r3,r3
c000b908:	20 63 00 20 	subfic  r3,r3,32
c000b90c:	4e 80 00 20 	blr

__builtin_ffs() is already used in arch/powerpc/include/asm/page_32.h
__builtin_ffs() has been in GCC since at least gcc-2.95

On PPC64, the generated code is different, so we do this change
only on PPC32:

Before the change:
c000000000027c00 <.ffs_test2>:
c000000000027c00:	7c 03 00 d0 	neg     r0,r3
c000000000027c04:	7c 03 18 38 	and     r3,r0,r3
c000000000027c08:	7c 63 00 74 	cntlzd  r3,r3
c000000000027c0c:	20 63 00 40 	subfic  r3,r3,64
c000000000027c10:	7c 63 07 b4 	extsw   r3,r3
c000000000027c14:	4e 80 00 20 	blr

After the change:
c000000000027bf0 <.ffs_test2>:
c000000000027bf0:	7c 03 00 d0 	neg     r0,r3
c000000000027bf4:	7c 03 18 38 	and     r3,r0,r3
c000000000027bf8:	7c 63 00 34 	cntlzw  r3,r3
c000000000027bfc:	20 63 00 20 	subfic  r3,r3,32
c000000000027c00:	7c 63 07 b4 	extsw   r3,r3
c000000000027c04:	4e 80 00 20 	blr

Signed-off-by: Christophe Leroy <christophe.leroy@....fr>
---
v2: limit the change to PPC32

 arch/powerpc/include/asm/bitops.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/arch/powerpc/include/asm/bitops.h b/arch/powerpc/include/asm/bitops.h
index 59abc62..0e84f0c 100644
--- a/arch/powerpc/include/asm/bitops.h
+++ b/arch/powerpc/include/asm/bitops.h
@@ -223,11 +223,15 @@ static __inline__ unsigned long __ffs(unsigned long x)
  * the libc and compiler builtin ffs routines, therefore
  * differs in spirit from the above ffz (man ffs).
  */
+#ifdef CONFIG_PPC64
 static __inline__ int ffs(int x)
 {
 	unsigned long i = (unsigned long)x;
 	return __ilog2(i & -i) + 1;
 }
+#else
+#include <asm-generic/bitops/builtin-ffs.h>
+#endif
 
 /*
  * fls: find last (most-significant) bit set.
-- 
2.10.1