opcodes: add support for eBPF

This patch adds support for the Linux kernel eBPF architecture to the
opcodes.  The port is based on CGEN.

opcodes/ChangeLog:

2019-05-23  Jose E. Marchesi  <jose.marchesi@oracle.com>

	* configure.ac (SHARED_DEPENDENCIES): Add case for bfd_bpf_arch.
	* configure: Regenerated.
	* Makefile.am: Add rules for the files generated from cpu/bpf.cpu
	and cpu/bpf.opc.
	(HFILES): Add bpf-desc.h and bpf-opc.h.
	(TARGET_LIBOPCODES_CFILES): Add bpf-asm.c, bpf-desc.c, bpf-dis.c,
	bpf-ibld.c and bpf-opc.c.
	(BPF_DEPS): Define.
	* Makefile.in: Regenerated.
	* disassemble.c (ARCH_bpf): Define.
	(disassembler): Add case for bfd_arch_bpf.
	(disassemble_init_for_target): Likewise.
	(enum epbf_isa_attr): Define.
	* disassemble.h: extern print_insn_bpf.
	* bpf-asm.c: Generated.
	* bpf-opc.h: Likewise.
	* bpf-opc.c: Likewise.
	* bpf-ibld.c: Likewise.
	* bpf-dis.c: Likewise.
	* bpf-desc.h: Likewise.
	* bpf-desc.c: Likewise.

1 files changed, 590 insertions, 0 deletions

diff --git a/opcodes/bpf-asm.c b/opcodes/bpf-asm.c
new file mode 100644
index 0000000..4332edc
--- /dev/null
+++ b/opcodes/bpf-asm.c
@@ -0,0 +1,590 @@
+/* DO NOT EDIT!  -*- buffer-read-only: t -*- vi:set ro:  */
+/* Assembler interface for targets using CGEN. -*- C -*-
+   CGEN: Cpu tools GENerator
+
+   THIS FILE IS MACHINE GENERATED WITH CGEN.
+   - the resultant file is machine generated, cgen-asm.in isn't
+
+   Copyright (C) 1996-2019 Free Software Foundation, Inc.
+
+   This file is part of libopcodes.
+
+   This library is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3, or (at your option)
+   any later version.
+
+   It is distributed in the hope that it will be useful, but WITHOUT
+   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+   or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public
+   License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software Foundation, Inc.,
+   51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA.  */
+
+
+/* ??? Eventually more and more of this stuff can go to cpu-independent files.
+   Keep that in mind.  */
+
+#include "sysdep.h"
+#include <stdio.h>
+#include "ansidecl.h"
+#include "bfd.h"
+#include "symcat.h"
+#include "bpf-desc.h"
+#include "bpf-opc.h"
+#include "opintl.h"
+#include "xregex.h"
+#include "libiberty.h"
+#include "safe-ctype.h"
+
+#undef  min
+#define min(a,b) ((a) < (b) ? (a) : (b))
+#undef  max
+#define max(a,b) ((a) > (b) ? (a) : (b))
+
+static const char * parse_insn_normal
+  (CGEN_CPU_DESC, const CGEN_INSN *, const char **, CGEN_FIELDS *);
+
+/* -- assembler routines inserted here.  */
+
+/* -- asm.c */
+
+/* Parse a signed 64-bit immediate.  */
+
+static const char *
+parse_imm64 (CGEN_CPU_DESC cd,
+             const char **strp,
+             int opindex,
+             int64_t *valuep)
+{
+  bfd_vma value;
+  enum cgen_parse_operand_result result;
+  const char *errmsg;
+
+  errmsg = (* cd->parse_operand_fn)
+    (cd, CGEN_PARSE_OPERAND_INTEGER, strp, opindex, BFD_RELOC_NONE,
+     &result, &value);
+  if (!errmsg)
+    *valuep = value;
+
+  return errmsg;
+}
+
+/* Endianness size operands are integer immediates whose values can be
+   16, 32 or 64.  */
+
+static const char *
+parse_endsize (CGEN_CPU_DESC cd,
+               const char **strp,
+               int opindex,
+               unsigned long *valuep)
+{
+  const char *errmsg;
+
+  errmsg = cgen_parse_unsigned_integer (cd, strp, opindex, valuep);
+  if (errmsg)
+    return errmsg;
+
+  switch (*valuep)
+    {
+    case 16:
+    case 32:
+    case 64:
+      break;
+    default:
+      return _("expected 16, 32 or 64 in");
+    }
+
+  return NULL;
+}
+
+/* Special check to ensure that the right instruction variant is used
+   for the given endianness induced by the ISA selected in the CPU.
+   See bpf.cpu for a discussion on how eBPF is really two instruction
+   sets.  */
+
+int
+bpf_cgen_insn_supported (CGEN_CPU_DESC cd, const CGEN_INSN *insn)
+{
+  CGEN_BITSET isas = CGEN_INSN_BITSET_ATTR_VALUE (insn, CGEN_INSN_ISA);
+
+  return cgen_bitset_intersect_p (&isas, cd->isas);
+}
+
+
+/* -- dis.c */
+
+const char * bpf_cgen_parse_operand
+  (CGEN_CPU_DESC, int, const char **, CGEN_FIELDS *);
+
+/* Main entry point for operand parsing.
+
+   This function is basically just a big switch statement.  Earlier versions
+   used tables to look up the function to use, but
+   - if the table contains both assembler and disassembler functions then
+     the disassembler contains much of the assembler and vice-versa,
+   - there's a lot of inlining possibilities as things grow,
+   - using a switch statement avoids the function call overhead.
+
+   This function could be moved into `parse_insn_normal', but keeping it
+   separate makes clear the interface between `parse_insn_normal' and each of
+   the handlers.  */
+
+const char *
+bpf_cgen_parse_operand (CGEN_CPU_DESC cd,
+			   int opindex,
+			   const char ** strp,
+			   CGEN_FIELDS * fields)
+{
+  const char * errmsg = NULL;
+  /* Used by scalar operands that still need to be parsed.  */
+  long junk ATTRIBUTE_UNUSED;
+
+  switch (opindex)
+    {
+    case BPF_OPERAND_DISP16 :
+      errmsg = cgen_parse_signed_integer (cd, strp, BPF_OPERAND_DISP16, (long *) (& fields->f_offset16));
+      break;
+    case BPF_OPERAND_DISP32 :
+      errmsg = cgen_parse_signed_integer (cd, strp, BPF_OPERAND_DISP32, (long *) (& fields->f_imm32));
+      break;
+    case BPF_OPERAND_DSTBE :
+      errmsg = cgen_parse_keyword (cd, strp, & bpf_cgen_opval_h_gpr, & fields->f_dstbe);
+      break;
+    case BPF_OPERAND_DSTLE :
+      errmsg = cgen_parse_keyword (cd, strp, & bpf_cgen_opval_h_gpr, & fields->f_dstle);
+      break;
+    case BPF_OPERAND_ENDSIZE :
+      errmsg = parse_endsize (cd, strp, BPF_OPERAND_ENDSIZE, (unsigned long *) (& fields->f_imm32));
+      break;
+    case BPF_OPERAND_IMM32 :
+      errmsg = cgen_parse_signed_integer (cd, strp, BPF_OPERAND_IMM32, (long *) (& fields->f_imm32));
+      break;
+    case BPF_OPERAND_IMM64 :
+      errmsg = parse_imm64 (cd, strp, BPF_OPERAND_IMM64, (int64_t *) (& fields->f_imm64));
+      break;
+    case BPF_OPERAND_OFFSET16 :
+      errmsg = cgen_parse_signed_integer (cd, strp, BPF_OPERAND_OFFSET16, (long *) (& fields->f_offset16));
+      break;
+    case BPF_OPERAND_SRCBE :
+      errmsg = cgen_parse_keyword (cd, strp, & bpf_cgen_opval_h_gpr, & fields->f_srcbe);
+      break;
+    case BPF_OPERAND_SRCLE :
+      errmsg = cgen_parse_keyword (cd, strp, & bpf_cgen_opval_h_gpr, & fields->f_srcle);
+      break;
+
+    default :
+      /* xgettext:c-format */
+      opcodes_error_handler
+	(_("internal error: unrecognized field %d while parsing"),
+	 opindex);
+      abort ();
+  }
+
+  return errmsg;
+}
+
+cgen_parse_fn * const bpf_cgen_parse_handlers[] =
+{
+  parse_insn_normal,
+};
+
+void
+bpf_cgen_init_asm (CGEN_CPU_DESC cd)
+{
+  bpf_cgen_init_opcode_table (cd);
+  bpf_cgen_init_ibld_table (cd);
+  cd->parse_handlers = & bpf_cgen_parse_handlers[0];
+  cd->parse_operand = bpf_cgen_parse_operand;
+#ifdef CGEN_ASM_INIT_HOOK
+CGEN_ASM_INIT_HOOK
+#endif
+}
+
+
+
+/* Regex construction routine.
+
+   This translates an opcode syntax string into a regex string,
+   by replacing any non-character syntax element (such as an
+   opcode) with the pattern '.*'
+
+   It then compiles the regex and stores it in the opcode, for
+   later use by bpf_cgen_assemble_insn
+
+   Returns NULL for success, an error message for failure.  */
+
+char *
+bpf_cgen_build_insn_regex (CGEN_INSN *insn)
+{
+  CGEN_OPCODE *opc = (CGEN_OPCODE *) CGEN_INSN_OPCODE (insn);
+  const char *mnem = CGEN_INSN_MNEMONIC (insn);
+  char rxbuf[CGEN_MAX_RX_ELEMENTS];
+  char *rx = rxbuf;
+  const CGEN_SYNTAX_CHAR_TYPE *syn;
+  int reg_err;
+
+  syn = CGEN_SYNTAX_STRING (CGEN_OPCODE_SYNTAX (opc));
+
+  /* Mnemonics come first in the syntax string.  */
+  if (! CGEN_SYNTAX_MNEMONIC_P (* syn))
+    return _("missing mnemonic in syntax string");
+  ++syn;
+
+  /* Generate a case sensitive regular expression that emulates case
+     insensitive matching in the "C" locale.  We cannot generate a case
+     insensitive regular expression because in Turkish locales, 'i' and 'I'
+     are not equal modulo case conversion.  */
+
+  /* Copy the literal mnemonic out of the insn.  */
+  for (; *mnem; mnem++)
+    {
+      char c = *mnem;
+
+      if (ISALPHA (c))
+	{
+	  *rx++ = '[';
+	  *rx++ = TOLOWER (c);
+	  *rx++ = TOUPPER (c);
+	  *rx++ = ']';
+	}
+      else
+	*rx++ = c;
+    }
+
+  /* Copy any remaining literals from the syntax string into the rx.  */
+  for(; * syn != 0 && rx <= rxbuf + (CGEN_MAX_RX_ELEMENTS - 7 - 4); ++syn)
+    {
+      if (CGEN_SYNTAX_CHAR_P (* syn))
+	{
+	  char c = CGEN_SYNTAX_CHAR (* syn);
+
+	  switch (c)
+	    {
+	      /* Escape any regex metacharacters in the syntax.  */
+	    case '.': case '[': case '\\':
+	    case '*': case '^': case '$':
+
+#ifdef CGEN_ESCAPE_EXTENDED_REGEX
+	    case '?': case '{': case '}':
+	    case '(': case ')': case '*':
+	    case '|': case '+': case ']':
+#endif
+	      *rx++ = '\\';
+	      *rx++ = c;
+	      break;
+
+	    default:
+	      if (ISALPHA (c))
+		{
+		  *rx++ = '[';
+		  *rx++ = TOLOWER (c);
+		  *rx++ = TOUPPER (c);
+		  *rx++ = ']';
+		}
+	      else
+		*rx++ = c;
+	      break;
+	    }
+	}
+      else
+	{
+	  /* Replace non-syntax fields with globs.  */
+	  *rx++ = '.';
+	  *rx++ = '*';
+	}
+    }
+
+  /* Trailing whitespace ok.  */
+  * rx++ = '[';
+  * rx++ = ' ';
+  * rx++ = '\t';
+  * rx++ = ']';
+  * rx++ = '*';
+
+  /* But anchor it after that.  */
+  * rx++ = '$';
+  * rx = '\0';
+
+  CGEN_INSN_RX (insn) = xmalloc (sizeof (regex_t));
+  reg_err = regcomp ((regex_t *) CGEN_INSN_RX (insn), rxbuf, REG_NOSUB);
+
+  if (reg_err == 0)
+    return NULL;
+  else
+    {
+      static char msg[80];
+
+      regerror (reg_err, (regex_t *) CGEN_INSN_RX (insn), msg, 80);
+      regfree ((regex_t *) CGEN_INSN_RX (insn));
+      free (CGEN_INSN_RX (insn));
+      (CGEN_INSN_RX (insn)) = NULL;
+      return msg;
+    }
+}
+
+
+/* Default insn parser.
+
+   The syntax string is scanned and operands are parsed and stored in FIELDS.
+   Relocs are queued as we go via other callbacks.
+
+   ??? Note that this is currently an all-or-nothing parser.  If we fail to
+   parse the instruction, we return 0 and the caller will start over from
+   the beginning.  Backtracking will be necessary in parsing subexpressions,
+   but that can be handled there.  Not handling backtracking here may get
+   expensive in the case of the m68k.  Deal with later.
+
+   Returns NULL for success, an error message for failure.  */
+
+static const char *
+parse_insn_normal (CGEN_CPU_DESC cd,
+		   const CGEN_INSN *insn,
+		   const char **strp,
+		   CGEN_FIELDS *fields)
+{
+  /* ??? Runtime added insns not handled yet.  */
+  const CGEN_SYNTAX *syntax = CGEN_INSN_SYNTAX (insn);
+  const char *str = *strp;
+  const char *errmsg;
+  const char *p;
+  const CGEN_SYNTAX_CHAR_TYPE * syn;
+#ifdef CGEN_MNEMONIC_OPERANDS
+  /* FIXME: wip */
+  int past_opcode_p;
+#endif
+
+  /* For now we assume the mnemonic is first (there are no leading operands).
+     We can parse it without needing to set up operand parsing.
+     GAS's input scrubber will ensure mnemonics are lowercase, but we may
+     not be called from GAS.  */
+  p = CGEN_INSN_MNEMONIC (insn);
+  while (*p && TOLOWER (*p) == TOLOWER (*str))
+    ++p, ++str;
+
+  if (* p)
+    return _("unrecognized instruction");
+
+#ifndef CGEN_MNEMONIC_OPERANDS
+  if (* str && ! ISSPACE (* str))
+    return _("unrecognized instruction");
+#endif
+
+  CGEN_INIT_PARSE (cd);
+  cgen_init_parse_operand (cd);
+#ifdef CGEN_MNEMONIC_OPERANDS
+  past_opcode_p = 0;
+#endif
+
+  /* We don't check for (*str != '\0') here because we want to parse
+     any trailing fake arguments in the syntax string.  */
+  syn = CGEN_SYNTAX_STRING (syntax);
+
+  /* Mnemonics come first for now, ensure valid string.  */
+  if (! CGEN_SYNTAX_MNEMONIC_P (* syn))
+    abort ();
+
+  ++syn;
+
+  while (* syn != 0)
+    {
+      /* Non operand chars must match exactly.  */
+      if (CGEN_SYNTAX_CHAR_P (* syn))
+	{
+	  /* FIXME: While we allow for non-GAS callers above, we assume the
+	     first char after the mnemonic part is a space.  */
+	  /* FIXME: We also take inappropriate advantage of the fact that
+	     GAS's input scrubber will remove extraneous blanks.  */
+	  if (TOLOWER (*str) == TOLOWER (CGEN_SYNTAX_CHAR (* syn)))
+	    {
+#ifdef CGEN_MNEMONIC_OPERANDS
+	      if (CGEN_SYNTAX_CHAR(* syn) == ' ')
+		past_opcode_p = 1;
+#endif
+	      ++ syn;
+	      ++ str;
+	    }
+	  else if (*str)
+	    {
+	      /* Syntax char didn't match.  Can't be this insn.  */
+	      static char msg [80];
+
+	      /* xgettext:c-format */
+	      sprintf (msg, _("syntax error (expected char `%c', found `%c')"),
+		       CGEN_SYNTAX_CHAR(*syn), *str);
+	      return msg;
+	    }
+	  else
+	    {
+	      /* Ran out of input.  */
+	      static char msg [80];
+
+	      /* xgettext:c-format */
+	      sprintf (msg, _("syntax error (expected char `%c', found end of instruction)"),
+		       CGEN_SYNTAX_CHAR(*syn));
+	      return msg;
+	    }
+	  continue;
+	}
+
+#ifdef CGEN_MNEMONIC_OPERANDS
+      (void) past_opcode_p;
+#endif
+      /* We have an operand of some sort.  */
+      errmsg = cd->parse_operand (cd, CGEN_SYNTAX_FIELD (*syn), &str, fields);
+      if (errmsg)
+	return errmsg;
+
+      /* Done with this operand, continue with next one.  */
+      ++ syn;
+    }
+
+  /* If we're at the end of the syntax string, we're done.  */
+  if (* syn == 0)
+    {
+      /* FIXME: For the moment we assume a valid `str' can only contain
+	 blanks now.  IE: We needn't try again with a longer version of
+	 the insn and it is assumed that longer versions of insns appear
+	 before shorter ones (eg: lsr r2,r3,1 vs lsr r2,r3).  */
+      while (ISSPACE (* str))
+	++ str;
+
+      if (* str != '\0')
+	return _("junk at end of line"); /* FIXME: would like to include `str' */
+
+      return NULL;
+    }
+
+  /* We couldn't parse it.  */
+  return _("unrecognized instruction");
+}
+
+/* Main entry point.
+   This routine is called for each instruction to be assembled.
+   STR points to the insn to be assembled.
+   We assume all necessary tables have been initialized.
+   The assembled instruction, less any fixups, is stored in BUF.
+   Remember that if CGEN_INT_INSN_P then BUF is an int and thus the value
+   still needs to be converted to target byte order, otherwise BUF is an array
+   of bytes in target byte order.
+   The result is a pointer to the insn's entry in the opcode table,
+   or NULL if an error occured (an error message will have already been
+   printed).
+
+   Note that when processing (non-alias) macro-insns,
+   this function recurses.
+
+   ??? It's possible to make this cpu-independent.
+   One would have to deal with a few minor things.
+   At this point in time doing so would be more of a curiosity than useful
+   [for example this file isn't _that_ big], but keeping the possibility in
+   mind helps keep the design clean.  */
+
+const CGEN_INSN *
+bpf_cgen_assemble_insn (CGEN_CPU_DESC cd,
+			   const char *str,
+			   CGEN_FIELDS *fields,
+			   CGEN_INSN_BYTES_PTR buf,
+			   char **errmsg)
+{
+  const char *start;
+  CGEN_INSN_LIST *ilist;
+  const char *parse_errmsg = NULL;
+  const char *insert_errmsg = NULL;
+  int recognized_mnemonic = 0;
+
+  /* Skip leading white space.  */
+  while (ISSPACE (* str))
+    ++ str;
+
+  /* The instructions are stored in hashed lists.
+     Get the first in the list.  */
+  ilist = CGEN_ASM_LOOKUP_INSN (cd, str);
+
+  /* Keep looking until we find a match.  */
+  start = str;
+  for ( ; ilist != NULL ; ilist = CGEN_ASM_NEXT_INSN (ilist))
+    {
+      const CGEN_INSN *insn = ilist->insn;
+      recognized_mnemonic = 1;
+
+#ifdef CGEN_VALIDATE_INSN_SUPPORTED
+      /* Not usually needed as unsupported opcodes
+	 shouldn't be in the hash lists.  */
+      /* Is this insn supported by the selected cpu?  */
+      if (! bpf_cgen_insn_supported (cd, insn))
+	continue;
+#endif
+      /* If the RELAXED attribute is set, this is an insn that shouldn't be
+	 chosen immediately.  Instead, it is used during assembler/linker
+	 relaxation if possible.  */
+      if (CGEN_INSN_ATTR_VALUE (insn, CGEN_INSN_RELAXED) != 0)
+	continue;
+
+      str = start;
+
+      /* Skip this insn if str doesn't look right lexically.  */
+      if (CGEN_INSN_RX (insn) != NULL &&
+	  regexec ((regex_t *) CGEN_INSN_RX (insn), str, 0, NULL, 0) == REG_NOMATCH)
+	continue;
+
+      /* Allow parse/insert handlers to obtain length of insn.  */
+      CGEN_FIELDS_BITSIZE (fields) = CGEN_INSN_BITSIZE (insn);
+
+      parse_errmsg = CGEN_PARSE_FN (cd, insn) (cd, insn, & str, fields);
+      if (parse_errmsg != NULL)
+	continue;
+
+      /* ??? 0 is passed for `pc'.  */
+      insert_errmsg = CGEN_INSERT_FN (cd, insn) (cd, insn, fields, buf,
+						 (bfd_vma) 0);
+      if (insert_errmsg != NULL)
+        continue;
+
+      /* It is up to the caller to actually output the insn and any
+         queued relocs.  */
+      return insn;
+    }
+
+  {
+    static char errbuf[150];
+    const char *tmp_errmsg;
+#ifdef CGEN_VERBOSE_ASSEMBLER_ERRORS
+#define be_verbose 1
+#else
+#define be_verbose 0
+#endif
+
+    if (be_verbose)
+      {
+	/* If requesting verbose error messages, use insert_errmsg.
+	   Failing that, use parse_errmsg.  */
+	tmp_errmsg = (insert_errmsg ? insert_errmsg :
+		      parse_errmsg ? parse_errmsg :
+		      recognized_mnemonic ?
+		      _("unrecognized form of instruction") :
+		      _("unrecognized instruction"));
+
+	if (strlen (start) > 50)
+	  /* xgettext:c-format */
+	  sprintf (errbuf, "%s `%.50s...'", tmp_errmsg, start);
+	else
+	  /* xgettext:c-format */
+	  sprintf (errbuf, "%s `%.50s'", tmp_errmsg, start);
+      }
+    else
+      {
+	if (strlen (start) > 50)
+	  /* xgettext:c-format */
+	  sprintf (errbuf, _("bad instruction `%.50s...'"), start);
+	else
+	  /* xgettext:c-format */
+	  sprintf (errbuf, _("bad instruction `%.50s'"), start);
+      }
+
+    *errmsg = errbuf;
+    return NULL;
+  }
+}