Blackfin arch: defines and provides entry points for certain user space functions at fixed addresses

This patch defines (and provides) entry points for certain user space functions
at fixed addresses.  The Blackfin has no usable atomic instructions, but we can
ensure that these code sequences appear atomic from a user space point of view
by detecting when we're in the process of executing them during the interrupt
handler return path.  This allows much more efficient pthread lock
implementations than the bfin_spinlock syscall we're currently using.

Also provided is a small sys_rt_sigreturn stub which can be used by the signal
handler setup code.  The signal.c part will be committed separately.

Signed-off-by: Bernd Schmidt <bernd.schmidt@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>

1 files changed, 132 insertions, 0 deletions

diff --git a/arch/blackfin/kernel/fixed_code.S b/arch/blackfin/kernel/fixed_code.S
new file mode 100644
index 000000000000..99ea296c82c7
--- /dev/null
+++ b/arch/blackfin/kernel/fixed_code.S
@@ -0,0 +1,132 @@
+/*
+ * This file contains sequences of code that will be copied to a
+ * fixed location, defined in <asm/atomic_seq.h>.  The interrupt
+ * handlers ensure that these sequences appear to be atomic when
+ * executed from userspace.
+ * These are aligned to 16 bytes, so that we have some space to replace
+ * these sequences with something else (e.g. kernel traps if we ever do
+ * BF561 SMP).
+ */
+#include <linux/linkage.h>
+#include <asm/entry.h>
+#include <asm/unistd.h>
+
+.text
+ENTRY(_fixed_code_start)
+
+.align 16
+ENTRY(_sigreturn_stub)
+	P0 = __NR_rt_sigreturn;
+	EXCPT 0;
+	/* Speculative execution paranoia.  */
+0:	JUMP.S 0b;
+ENDPROC (_sigreturn_stub)
+
+.align 16
+	/*
+	 * Atomic swap, 8 bit.
+	 * Inputs:	P0: memory address to use
+	 *		R1: value to store
+	 * Output:	R0: old contents of the memory address, zero extended.
+	 */
+ENTRY(_atomic_xchg32)
+	R0 = [P0];
+	[P0] = R1;
+	rts;
+ENDPROC (_atomic_xchg32)
+
+.align 16
+	/*
+	 * Compare and swap, 32 bit.
+	 * Inputs:	P0: memory address to use
+	 *		R1: compare value
+	 *		R2: new value to store
+	 * The new value is stored if the contents of the memory
+	 * address is equal to the compare value.
+	 * Output:	R0: old contents of the memory address.
+	 */
+ENTRY(_atomic_cas32)
+	R0 = [P0];
+	CC = R0 == R1;
+	IF !CC JUMP 1f;
+	[P0] = R2;
+1:
+	rts;
+ENDPROC (_atomic_cas32)
+
+.align 16
+	/*
+	 * Atomic add, 32 bit.
+	 * Inputs:	P0: memory address to use
+	 *		R0: value to add
+	 * Outputs:	R0: new contents of the memory address.
+	 *		R1: previous contents of the memory address.
+	 */
+ENTRY(_atomic_add32)
+	R1 = [P0];
+	R0 = R1 + R0;
+	[P0] = R0;
+	rts;
+ENDPROC (_atomic_add32)
+
+.align 16
+	/*
+	 * Atomic sub, 32 bit.
+	 * Inputs:	P0: memory address to use
+	 *		R0: value to subtract
+	 * Outputs:	R0: new contents of the memory address.
+	 *		R1: previous contents of the memory address.
+	 */
+ENTRY(_atomic_sub32)
+	R1 = [P0];
+	R0 = R1 - R0;
+	[P0] = R0;
+	rts;
+ENDPROC (_atomic_sub32)
+
+.align 16
+	/*
+	 * Atomic ior, 32 bit.
+	 * Inputs:	P0: memory address to use
+	 *		R0: value to ior
+	 * Outputs:	R0: new contents of the memory address.
+	 *		R1: previous contents of the memory address.
+	 */
+ENTRY(_atomic_ior32)
+	R1 = [P0];
+	R0 = R1 | R0;
+	[P0] = R0;
+	rts;
+ENDPROC (_atomic_ior32)
+
+.align 16
+	/*
+	 * Atomic ior, 32 bit.
+	 * Inputs:	P0: memory address to use
+	 *		R0: value to ior
+	 * Outputs:	R0: new contents of the memory address.
+	 *		R1: previous contents of the memory address.
+	 */
+ENTRY(_atomic_and32)
+	R1 = [P0];
+	R0 = R1 & R0;
+	[P0] = R0;
+	rts;
+ENDPROC (_atomic_ior32)
+
+.align 16
+	/*
+	 * Atomic ior, 32 bit.
+	 * Inputs:	P0: memory address to use
+	 *		R0: value to ior
+	 * Outputs:	R0: new contents of the memory address.
+	 *		R1: previous contents of the memory address.
+	 */
+ENTRY(_atomic_xor32)
+	R1 = [P0];
+	R0 = R1 ^ R0;
+	[P0] = R0;
+	rts;
+ENDPROC (_atomic_ior32)
+
+ENTRY(_fixed_code_end)