From 0342454f8aa6fc55e515bad26425533e10b58085 Mon Sep 17 00:00:00 2001
From: Paolo Bonzini <pbonzini@redhat.com>
Date: Fri, 2 Sep 2016 17:36:23 +0200
Subject: [PATCH] optionrom: do not rely on compiler's bswap optimization

Recent compilers can detect and inline manually-written bswap code,
but GCC 4.2.1 (the last GPLv2 version) cannot and generates really
awful code.  Depending on how the compiler is configured, it might
also not want to generate bswap because it was not in i386.  Using
asm is fine because TCG knows about bswap and all processors with
virtualization extensions also do.

Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 pc-bios/linuxboot_dma.bin         | Bin 1536 -> 1536 bytes
 pc-bios/optionrom/linuxboot_dma.c |  18 ++++--------------
 2 files changed, 4 insertions(+), 14 deletions(-)

diff --git a/pc-bios/linuxboot_dma.bin b/pc-bios/linuxboot_dma.bin
index 238a195d3869995067f158d243d852778d38a736..218d3ab4a29bfb5ab7125ec7a4d29dad1860c673 100644
GIT binary patch
delta 339
zc-m{=Y2cX<?p@B%Tf)%0mw_QIIIZ(?TJsx$G{Z}22OkTh@t;WJKb1Bqt@$)p8vn^O
z{?qAcoh~XIKr%#y2T1y;umDLJm9$+FtU&FZ4<`OnXBS9|PCJ}7VX^_Eq+#*@|Nqm|
zdR<fm(sr^i0F~?#2azwL!Q$O4Ae}$~Akzn^<b})s$yJPwl8i82A~0RtU|laKUtyeH
zANUU{%K?*R2g_Q*#3f+j9AI%pusF~HiL_mU5X(S%UvR?I^T5;#gVlfh3l*1viHm^6
fZ-T|Ug+Lm1iNahhllJ1^=08l#%##b46Bro*mK%S6

delta 431
zc-l3Nze@sP9LC?vi%z@H@}k#pylD1M2pZj@9HOG-acOjEI>qTF4ibl!mZGIOmkt_U
zgO-q@A&DR|LT?D7O^Cd{_v*VmynLP?-?wU4?c<WyWqyVEH;iEuqJt8IM7o15h&ybc
z)?+MUsP#!13a?0C+;fO7WZ!~lLw;~1-yjDWB%e`#NMVIh1BEBlzZ^=pDK>-*Icftn
zHAj9b4$E%~vgcBNScWLN7HP$`Ne^6;)N~=<5%6io8t1EN2iw@i9(Hi+w+KDv3fU-%
z%JA3S46aR@MX(fW_&bXE5e)>K06mKo2|q{8Wn(ROHkz~?-I`S`(X1FJ-p}8}wX{2<
z+!xKc=9c18aXP>e%#V-7-SAT9ZJp1md_(iR=8ocr<DG@L5G?9n=MnB-olYeZhW8IM
C5viU4

diff --git a/pc-bios/optionrom/linuxboot_dma.c b/pc-bios/optionrom/linuxboot_dma.c
index 75497977322..4754282ad73 100644
--- a/pc-bios/optionrom/linuxboot_dma.c
+++ b/pc-bios/optionrom/linuxboot_dma.c
@@ -122,24 +122,14 @@ static inline void writel_es(uint16_t offset, uint32_t val)
 
 static inline uint32_t bswap32(uint32_t x)
 {
-    return
-        ((x & 0x000000ffU) << 24) |
-        ((x & 0x0000ff00U) <<  8) |
-        ((x & 0x00ff0000U) >>  8) |
-        ((x & 0xff000000U) >> 24);
+    asm("bswapl %0" : "=r" (x) : "0" (x));
+    return x;
 }
 
 static inline uint64_t bswap64(uint64_t x)
 {
-    return
-        ((x & 0x00000000000000ffULL) << 56) |
-        ((x & 0x000000000000ff00ULL) << 40) |
-        ((x & 0x0000000000ff0000ULL) << 24) |
-        ((x & 0x00000000ff000000ULL) <<  8) |
-        ((x & 0x000000ff00000000ULL) >>  8) |
-        ((x & 0x0000ff0000000000ULL) >> 24) |
-        ((x & 0x00ff000000000000ULL) >> 40) |
-        ((x & 0xff00000000000000ULL) >> 56);
+    asm("bswapl %%eax; bswapl %%edx; xchg %%eax, %%edx" : "=A" (x) : "0" (x));
+    return x;
 }
 
 static inline uint64_t cpu_to_be64(uint64_t x)
-- 
2.39.5