Fixed bug 4683 - SDL_atomic infinite recursion on armv6/armv5 w/ thumb

The real problem is that SDL_atomic.c was built in thumb mode instead of ARM mode, which is required to use the mcr instruction on ARM platforms. Added a compiler error to catch this case so we don't generate code that does infinite recursion.

I also added a potentially better way to handle things on Linux ARM platforms, based on comments in the Chromium headers, which we can try out after 2.0.10 ships.
diff --git a/include/SDL_atomic.h b/include/SDL_atomic.h
index 687455f..bea071c 100644
--- a/include/SDL_atomic.h
+++ b/include/SDL_atomic.h
@@ -162,6 +162,22 @@
 #define SDL_MemoryBarrierRelease()   __asm__ __volatile__ ("dmb ish" : : : "memory")
 #define SDL_MemoryBarrierAcquire()   __asm__ __volatile__ ("dmb ish" : : : "memory")
 #elif defined(__GNUC__) && defined(__arm__)
+#if 0 /* defined(__LINUX__) || defined(__ANDROID__) */
+/* Information from:
+   https://chromium.googlesource.com/chromium/chromium/+/trunk/base/atomicops_internals_arm_gcc.h#19
+
+   The Linux kernel provides a helper function which provides the right code for a memory barrier,
+   hard-coded at address 0xffff0fa0
+*/
+typedef void (*SDL_KernelMemoryBarrierFunc)();
+#define SDL_MemoryBarrierRelease()	((SDL_KernelMemoryBarrierFunc)0xffff0fa0)()
+#define SDL_MemoryBarrierAcquire()	((SDL_KernelMemoryBarrierFunc)0xffff0fa0)()
+#elif 0 /* defined(__QNXNTO__) */
+#include <sys/cpuinline.h>
+
+#define SDL_MemoryBarrierRelease()   __cpu_membarrier()
+#define SDL_MemoryBarrierAcquire()   __cpu_membarrier()
+#else
 #if defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7EM__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) || defined(__ARM_ARCH_8A__)
 #define SDL_MemoryBarrierRelease()   __asm__ __volatile__ ("dmb ish" : : : "memory")
 #define SDL_MemoryBarrierAcquire()   __asm__ __volatile__ ("dmb ish" : : : "memory")
@@ -177,6 +193,7 @@
 #else
 #define SDL_MemoryBarrierRelease()   __asm__ __volatile__ ("" : : : "memory")
 #define SDL_MemoryBarrierAcquire()   __asm__ __volatile__ ("" : : : "memory")
+#endif /* __LINUX__ || __ANDROID__ */
 #endif /* __GNUC__ && __arm__ */
 #else
 #if (defined(__SUNPRO_C) && (__SUNPRO_C >= 0x5120))
diff --git a/src/atomic/SDL_atomic.c b/src/atomic/SDL_atomic.c
index d0c915d..a368e30 100644
--- a/src/atomic/SDL_atomic.c
+++ b/src/atomic/SDL_atomic.c
@@ -289,6 +289,10 @@
 #endif
 }
 
+#ifdef __thumb__
+#error This file should be built in arm mode so the mcr instruction is available for memory barriers
+#endif
+
 void
 SDL_MemoryBarrierReleaseFunction(void)
 {