bpf: Add atomic compare-and-exchange instructions

This patch adds the two remaining BPF v3 atomic instructions:
- BPF_INSN_ACMP{,32}: atomic compare-and-swap
- BPF_INSN_AXCHG{,32}: atomic (non-conditional) exchange

Tests and documentation are also updated.

gas/
	* doc/c-bpf.texi (BPF Instructions): Document atomic exchange and
	atomic compare-and-swap instructions.
	* testsuite/gas/bpf/atomic.s: Test ACMP, ACMP32, AXCHG, AXCGH32
	instructions.
	* testsuite/gas/bpf/atomic.d: Likewise.
	* testsuite/gas/bpf/atomic-be.d: Likewise.
	* testsuite/gas/bpf/atomic-pseudoc.s: Likewise.
	* testsuite/gas/bpf/atomic-pseudoc.d: Likewise.
	* testsuite/gas/bpf/atomic-be-pseudoc.d: Likewise.

include/
	* opcode/bpf.h (BPF_IMM32_ACMP): Fix typo.
	(enum bpf_insn_id): New entries for BPF_INSN_ACMP{,32} and
	BPF_INSN_AXCHG{,32}.

opcodes/
	* bpf-opc.c (bpf_opcodes): Add entries for ACMP{,32} and
	AXCHG{,32} instructions.

gas/doc/c-bpf.texi

@@ -745,8 +745,26 @@ Jump if signed lesser or equal.
 
 @subsection Atomic instructions
 
-Atomic exchange-and-add instructions are provided in two flavors: one
+Atomic exchange instructions are provided in two flavors: one for
-for swapping 64-bit quantities and another for 32-bit quantities.
+compare-and-swap, one for unconditional exchange.
+
+@table @code
+@item acmp [rd + offset16], rs
+@itemx r0 = cmpxchg_64 (rd + offset16, r0, rs)
+Atomic compare-and-swap.  Compares value in @code{r0} to value
+addressed by @code{rd + offset16}.  On match, the value addressed by
+@code{rd + offset16} is replaced with the value in @code{rs}.
+Regardless, the value that was at @code{rd + offset16} is
+zero-extended and loaded into @code{r0}.
+
+@item axchg [rd + offset16], rs
+@itemx rs = xchg_64 (rd + offset16, rs)
+Atomic exchange.  Atomically exchanges the value in @code{rs} with
+the value addressed by @code{rd + offset16}.
+@end table
+
+@noindent
+The following instructions provide atomic arithmetic operations.
 
 @table @code
 @item aadd [rd + offset16], rs
@@ -798,8 +816,26 @@ Alias to @code{aadd}.
 
 @subsection 32-bit atomic instructions
 
-Atomic exchange-and-add instructions are provided in two flavors: one
+32-bit atomic exchange instructions are provided in two flavors: one
-for swapping 32-bit quantities and another for 32-bit quantities.
+for compare-and-swap, one for unconditional exchange.
+
+@table @code
+@item acmp32 [rd + offset16], rs
+@itemx w0 = cmpxchg32_32 (rd + offset16, w0, ws)
+Atomic compare-and-swap.  Compares value in @code{w0} to value
+addressed by @code{rd + offset16}.  On match, the value addressed by
+@code{rd + offset16} is replaced with the value in @code{ws}.
+Regardless, the value that was at @code{rd + offset16} is
+zero-extended and loaded into @code{w0}.
+
+@item axchg [rd + offset16], rs
+@itemx ws = xchg32_32 (rd + offset16, ws)
+Atomic exchange.  Atomically exchanges the value in @code{ws} with
+the value addressed by @code{rd + offset16}.
+@end table
+
+@noindent
+The following instructions provide 32-bit atomic arithmetic operations.
 
 @table @code
 @item aadd32 [rd + offset16], rs

gas/testsuite/gas/bpf/atomic-be-pseudoc.d

@@ -26,3 +26,7 @@ Disassembly of section .text:
   78:	c3 12 1e ef 00 00 00 41 	w2=atomic_fetch_or\(\(u32\*\)\(r1\+0x1eef\),w2\)
   80:	db 12 1e ef 00 00 00 a1 	r2=atomic_fetch_xor\(\(u64\*\)\(r1\+0x1eef\),r2\)
   88:	c3 12 1e ef 00 00 00 a1 	w2=atomic_fetch_xor\(\(u32\*\)\(r1\+0x1eef\),w2\)
+  90:	db 12 00 04 00 00 00 f1 	r0=cmpxchg_64\(r1\+0x4,r0,r2\)
+  98:	c3 23 00 04 00 00 00 f1 	w0=cmpxchg32_32\(r2\+0x4,w0,w3\)
+  a0:	db 12 00 08 00 00 00 e1 	r2=xchg_64\(r1\+0x8,r2\)
+  a8:	c3 13 00 08 00 00 00 e1 	w3=xchg32_32\(r1\+0x8,w3\)

gas/testsuite/gas/bpf/atomic-be.d

@@ -24,3 +24,7 @@ Disassembly of section .text:
   68:	c3 12 1e ef 00 00 00 41 	afor32 \[%r1\+0x1eef\],%r2
   70:	db 12 1e ef 00 00 00 a1 	afxor \[%r1\+0x1eef\],%r2
   78:	c3 12 1e ef 00 00 00 a1 	afxor32 \[%r1\+0x1eef\],%r2
+  80:	db 12 00 04 00 00 00 f1 	acmp \[%r1\+0x4\],%r2
+  88:	c3 23 00 04 00 00 00 f1 	acmp32 \[%r2\+0x4\],%r3
+  90:	db 12 00 08 00 00 00 e1 	axchg \[%r1\+0x8\],%r2
+  98:	c3 13 00 08 00 00 00 e1 	axchg32 \[%r1\+0x8\],%r3

gas/testsuite/gas/bpf/atomic-pseudoc.d

@@ -26,3 +26,7 @@ Disassembly of section .text:
   78:	c3 21 ef 1e 41 00 00 00 	w2=atomic_fetch_or\(\(u32\*\)\(r1\+0x1eef\),w2\)
   80:	db 21 ef 1e a1 00 00 00 	r2=atomic_fetch_xor\(\(u64\*\)\(r1\+0x1eef\),r2\)
   88:	c3 21 ef 1e a1 00 00 00 	w2=atomic_fetch_xor\(\(u32\*\)\(r1\+0x1eef\),w2\)
+  90:	db 21 04 00 f1 00 00 00 	r0=cmpxchg_64\(r1\+0x4,r0,r2\)
+  98:	c3 32 04 00 f1 00 00 00 	w0=cmpxchg32_32\(r2\+0x4,w0,w3\)
+  a0:	db 21 08 00 e1 00 00 00 	r2=xchg_64\(r1\+0x8,r2\)
+  a8:	c3 31 08 00 e1 00 00 00 	w3=xchg32_32\(r1\+0x8,w3\)

gas/testsuite/gas/bpf/atomic-pseudoc.s

@@ -18,3 +18,7 @@
 	w2 = atomic_fetch_or((u32*)(r1+0x1eef),w2)
 	r2 = atomic_fetch_xor((u64*)(r1+0x1eef),r2)
 	w2 = atomic_fetch_xor((u32*)(r1+0x1eef),w2)
+	r0 = cmpxchg_64(r1+0x4,r0,r2)
+	w0 = cmpxchg32_32(r2+0x4,w0,w3)
+	r2 = xchg_64(r1+0x8,r2)
+	w3 = xchg32_32(r1+0x8,w3)

gas/testsuite/gas/bpf/atomic.d

@@ -24,3 +24,7 @@ Disassembly of section .text:
   68:	c3 21 ef 1e 41 00 00 00 	afor32 \[%r1\+0x1eef\],%r2
   70:	db 21 ef 1e a1 00 00 00 	afxor \[%r1\+0x1eef\],%r2
   78:	c3 21 ef 1e a1 00 00 00 	afxor32 \[%r1\+0x1eef\],%r2
+  80:	db 21 04 00 f1 00 00 00 	acmp \[%r1\+0x4\],%r2
+  88:	c3 32 04 00 f1 00 00 00 	acmp32 \[%r2\+0x4\],%r3
+  90:	db 21 08 00 e1 00 00 00 	axchg \[%r1\+0x8\],%r2
+  98:	c3 31 08 00 e1 00 00 00 	axchg32 \[%r1\+0x8\],%r3

gas/testsuite/gas/bpf/atomic.s

@@ -17,3 +17,8 @@
         afor32	[%r1+0x1eef], %r2
         afxor	[%r1+0x1eef], %r2
         afxor32	[%r1+0x1eef], %r2
+
+        acmp	[%r1+4], %r2
+        acmp32	[%r2+4], %r3
+        axchg	[%r1+8], %r2
+        axchg32	[%r1+8], %r3

include/opcode/bpf.h

@@ -153,7 +153,7 @@ typedef uint64_t bpf_insn_word;
 #define BPF_IMM32_AFAND ((uint64_t)0x00000051)
 #define BPF_IMM32_AFXOR ((uint64_t)0x000000a1)
 #define BPF_IMM32_AXCHG ((uint64_t)0x000000e1)
-#define BPF_IMM32_ACMP  ((uint64_t)b0x000000f1)
+#define BPF_IMM32_ACMP  ((uint64_t)0x000000f1)
 
 /* Unique identifiers for BPF instructions.  */
 
@@ -225,6 +225,10 @@ enum bpf_insn_id
   BPF_INSN_AADD32, BPF_INSN_AOR32, BPF_INSN_AAND32, BPF_INSN_AXOR32,
   /* Atomic instructions with fetching (32-bit.)  */
   BPF_INSN_AFADD32, BPF_INSN_AFOR32, BPF_INSN_AFAND32, BPF_INSN_AFXOR32,
+  /* Atomic compare-and-swap, atomic exchange.  */
+  BPF_INSN_ACMP, BPF_INSN_AXCHG,
+  /* Atomic compare-and-swap, atomic exchange (32-bit).  */
+  BPF_INSN_ACMP32, BPF_INSN_AXCHG32,
   /* GNU simulator specific instruction.  */
   BPF_INSN_BRKPT,
 };

opcodes/bpf-opc.c

@@ -403,6 +403,18 @@ const struct bpf_opcode bpf_opcodes[] =
   {BPF_INSN_AFXOR32, "afxor32%W[ %dr %o16 ] , %sr", "%sw = atomic_fetch_xor ( ( u32 * ) ( %dr %o16 ) , %sw )",
    BPF_V3, BPF_CODE|BPF_IMM32, BPF_CLASS_STX|BPF_SIZE_W|BPF_MODE_ATOMIC|BPF_IMM32_AFXOR},
 
+  /* Atomic compare-and-swap, atomic exchange.  */
+  {BPF_INSN_ACMP, "acmp%W[ %dr %o16 ] , %sr", "r0 = cmpxchg_64 ( %dr %o16 , r0 , %sr )",
+   BPF_V3, BPF_CODE|BPF_IMM32, BPF_CLASS_STX|BPF_SIZE_DW|BPF_MODE_ATOMIC|BPF_IMM32_ACMP},
+  {BPF_INSN_AXCHG, "axchg%W[ %dr %o16 ] , %sr", "%sr = xchg_64 ( %dr %o16 , %sr )",
+   BPF_V3, BPF_CODE|BPF_IMM32, BPF_CLASS_STX|BPF_SIZE_DW|BPF_MODE_ATOMIC|BPF_IMM32_AXCHG},
+
+  /* Atomic compare-and-swap, atomic exchange (32-bit).  */
+  {BPF_INSN_ACMP32, "acmp32%W[ %dr %o16 ], %sr", "w0 = cmpxchg32_32 ( %dr %o16 , w0 , %sw )",
+   BPF_V3, BPF_CODE|BPF_IMM32, BPF_CLASS_STX|BPF_SIZE_W|BPF_MODE_ATOMIC|BPF_IMM32_ACMP},
+  {BPF_INSN_AXCHG32, "axchg32%W[ %dr %o16 ], %sr", "%sw = xchg32_32 ( %dr %o16 , %sw )",
+   BPF_V3, BPF_CODE|BPF_IMM32, BPF_CLASS_STX|BPF_SIZE_W|BPF_MODE_ATOMIC|BPF_IMM32_AXCHG},
+
   /* Old versions of aadd and aadd32.  */
   {BPF_INSN_AADD, "xadddw%W[ %dr %o16 ] , %sr", "* ( u64 * ) ( %dr %o16 ) += %sr",
    BPF_V1, BPF_CODE|BPF_IMM32, BPF_CLASS_STX|BPF_SIZE_DW|BPF_MODE_ATOMIC|BPF_IMM32_AADD},