ARC: unaligned: relax the check for gcc supporting -mno-unaligned-access
authorVineet Gupta <vgupta@synopsys.com>
Tue, 5 Mar 2019 17:16:29 +0000 (09:16 -0800)
committerVineet Gupta <vgupta@synopsys.com>
Tue, 5 Mar 2019 17:16:33 +0000 (09:16 -0800)
Without bleeding edge gcc, kernel builds were tripping everywhere.

So current gcc will generate unaligned code despite
!CONFIG_ARC_USE_UNALIGNED_MEM_ACCESS but that is something we have to
live with.

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
arch/arc/kernel/setup.c

index 15da43d2e416525168386cdc8d0ed0d646d62d1f..a9c88b7e9182f6232b3f319a00865c8c9946b38f 100644 (file)
@@ -314,18 +314,6 @@ static char *arc_cpu_mumbojumbo(int cpu_id, char *buf, int len)
                       IS_AVAIL1(cpu->extn_mpy.ver, mpy_opt),
                       IS_AVAIL1(cpu->isa.div_rem, "div_rem "));
 
-
-#if defined(__ARC_UNALIGNED__) && !defined(CONFIG_ARC_USE_UNALIGNED_MEM_ACCESS)
-       /*
-        * gcc 7.3.1 (GNU 2018.03) onwards generate unaligned access by default
-        * but -mno-unaligned-access to disable that didn't work until gcc 8.2.1
-        * (GNU 2019.03). So landing here implies the interim period, when
-        * despite Kconfig being off, gcc is generating unaligned accesses which
-        * could bomb later on. So better to disallow such broken builds
-        */
-       BUILD_BUG_ON_MSG(1, "gcc doesn't support -mno-unaligned-access");
-#endif
-
        if (cpu->bpu.ver) {
                n += scnprintf(buf + n, len - n,
                              "BPU\t\t: %s%s match, cache:%d, Predict Table:%d Return stk: %d",