+2006-03-28 Jim Blandy <jimb@codesourcery.com>
+
+ * prologue-value.c, prologue-value.h: New files.
+ * Makefile.in (prologue_value_h): New variable.
+ (HFILES_NO_SRCDIR): List prologue-value.h.
+ (SFILES): List prologue-value.c.
+ (COMMON_OBS): List prologue-value.o.
+ (prologue-value.o): New rule.
+
2006-03-27 Michael Snyder <msnyder@redhat.com>
* xstormy16-tdep.c (xstormy16_return_value, xstormy16_push_dummy_call,
objc-exp.y objc-lang.c \
objfiles.c osabi.c observer.c \
p-exp.y p-lang.c p-typeprint.c p-valprint.c parse.c printcmd.c \
+ prologue-value.c \
regcache.c reggroups.c remote.c remote-fileio.c \
scm-exp.c scm-lang.c scm-valprint.c \
sentinel-frame.c \
ppcobsd_tdep_h = ppcobsd-tdep.h
ppc_tdep_h = ppc-tdep.h
proc_utils_h = proc-utils.h
+prologue_value_h = prologue-value.h
regcache_h = regcache.h
reggroups_h = reggroups.h
regset_h = regset.h
symfile.h stabsread.h target.h terminal.h typeprint.h \
xcoffsolib.h \
macrotab.h macroexp.h macroscope.h \
+ prologue-value.h \
ada-lang.h c-lang.h f-lang.h \
jv-lang.h \
m2-lang.h p-lang.h \
proc-service.o: proc-service.c $(defs_h) $(gdb_proc_service_h) $(inferior_h) \
$(symtab_h) $(target_h) $(gregset_h)
proc-why.o: proc-why.c $(defs_h) $(proc_utils_h)
+prologue-value.o: prologue-value.c $(defs_h) $(gdb_string_h) $(gdb_assert_h) \
+ $(prologue_value_h) $(regcache_h)
p-typeprint.o: p-typeprint.c $(defs_h) $(gdb_obstack_h) $(bfd_h) $(symtab_h) \
$(gdbtypes_h) $(expression_h) $(value_h) $(gdbcore_h) $(target_h) \
$(language_h) $(p_lang_h) $(typeprint_h) $(gdb_string_h)
+2006-03-28 Jim Blandy <jimb@codesourcery.com>
+
+ * gdbint.texinfo (Prologue Analysis): New section.
+
2006-03-07 Jim Blandy <jimb@red-bean.com>
* gdb.texinfo (Connecting): Document 'target remote pipe'.
@code{DEPRECATED_INIT_EXTRA_FRAME_INFO} and
@code{DEPRECATED_INIT_FRAME_PC} will be called for the new frame.
+@section Prologue Analysis
+
+@cindex prologue analysis
+@cindex call frame information
+@cindex CFI (call frame information)
+To produce a backtrace and allow the user to manipulate older frames'
+variables and arguments, @value{GDBN} needs to find the base addresses
+of older frames, and discover where those frames' registers have been
+saved. Since a frame's ``callee-saves'' registers get saved by
+younger frames if and when they're reused, a frame's registers may be
+scattered unpredictably across younger frames. This means that
+changing the value of a register-allocated variable in an older frame
+may actually entail writing to a save slot in some younger frame.
+
+Modern versions of GCC emit Dwarf call frame information (``CFI''),
+which describes how to find frame base addresses and saved registers.
+But CFI is not always available, so as a fallback @value{GDBN} uses a
+technique called @dfn{prologue analysis} to find frame sizes and saved
+registers. A prologue analyzer disassembles the function's machine
+code starting from its entry point, and looks for instructions that
+allocate frame space, save the stack pointer in a frame pointer
+register, save registers, and so on. Obviously, this can't be done
+accurately in general, but it's tractible to do well enough to be very
+helpful. Prologue analysis predates the GNU toolchain's support for
+CFI; at one time, prologue analysis was the only mechanism
+@value{GDBN} used for stack unwinding at all, when the function
+calling conventions didn't specify a fixed frame layout.
+
+In the olden days, function prologues were generated by hand-written,
+target-specific code in GCC, and treated as opaque and untouchable by
+optimizers. Looking at this code, it was usually straightforward to
+write a prologue analyzer for @value{GDBN} that would accurately
+understand all the prologues GCC would generate. However, over time
+GCC became more aggressive about instruction scheduling, and began to
+understand more about the semantics of the prologue instructions
+themselves; in response, @value{GDBN}'s analyzers became more complex
+and fragile. Keeping the prologue analyzers working as GCC (and the
+instruction sets themselves) evolved became a substantial task.
+
+@cindex @file{prologue-value.c}
+@cindex abstract interpretation of function prologues
+@cindex pseudo-evaluation of function prologues
+To try to address this problem, the code in @file{prologue-value.h}
+and @file{prologue-value.c} provides a general framework for writing
+prologue analyzers that are simpler and more robust than ad-hoc
+analyzers. When we analyze a prologue using the prologue-value
+framework, we're really doing ``abstract interpretation'' or
+``pseudo-evaluation'': running the function's code in simulation, but
+using conservative approximations of the values registers and memory
+would hold when the code actually runs. For example, if our function
+starts with the instruction:
+
+@example
+addi r1, 42 # add 42 to r1
+@end example
+@noindent
+we don't know exactly what value will be in @code{r1} after executing
+this instruction, but we do know it'll be 42 greater than its original
+value.
+
+If we then see an instruction like:
+
+@example
+addi r1, 22 # add 22 to r1
+@end example
+@noindent
+we still don't know what @code{r1's} value is, but again, we can say
+it is now 64 greater than its original value.
+
+If the next instruction were:
+
+@example
+mov r2, r1 # set r2 to r1's value
+@end example
+@noindent
+then we can say that @code{r2's} value is now the original value of
+@code{r1} plus 64.
+
+It's common for prologues to save registers on the stack, so we'll
+need to track the values of stack frame slots, as well as the
+registers. So after an instruction like this:
+
+@example
+mov (fp+4), r2
+@end example
+@noindent
+then we'd know that the stack slot four bytes above the frame pointer
+holds the original value of @code{r1} plus 64.
+
+And so on.
+
+Of course, this can only go so far before it gets unreasonable. If we
+wanted to be able to say anything about the value of @code{r1} after
+the instruction:
+
+@example
+xor r1, r3 # exclusive-or r1 and r3, place result in r1
+@end example
+@noindent
+then things would get pretty complex. But remember, we're just doing
+a conservative approximation; if exclusive-or instructions aren't
+relevant to prologues, we can just say @code{r1}'s value is now
+``unknown''. We can ignore things that are too complex, if that loss of
+information is acceptable for our application.
+
+So when we say ``conservative approximation'' here, what we mean is an
+approximation that is either accurate, or marked ``unknown'', but
+never inaccurate.
+
+Using this framework, a prologue analyzer is simply an interpreter for
+machine code, but one that uses conservative approximations for the
+contents of registers and memory instead of actual values. Starting
+from the function's entry point, you simulate instructions up to the
+current PC, or an instruction that you don't know how to simulate.
+Now you can examine the state of the registers and stack slots you've
+kept track of.
+
+@itemize @bullet
+
+@item
+To see how large your stack frame is, just check the value of the
+stack pointer register; if it's the original value of the SP
+minus a constant, then that constant is the stack frame's size.
+If the SP's value has been marked as ``unknown'', then that means
+the prologue has done something too complex for us to track, and
+we don't know the frame size.
+
+@item
+To see where we've saved the previous frame's registers, we just
+search the values we've tracked --- stack slots, usually, but
+registers, too, if you want --- for something equal to the register's
+original value. If the calling conventions suggest a standard place
+to save a given register, then we can check there first, but really,
+anything that will get us back the original value will probably work.
+@end itemize
+
+This does take some work. But prologue analyzers aren't
+quick-and-simple pattern patching to recognize a few fixed prologue
+forms any more; they're big, hairy functions. Along with inferior
+function calls, prologue analysis accounts for a substantial portion
+of the time needed to stabilize a @value{GDBN} port. So it's
+worthwhile to look for an approach that will be easier to understand
+and maintain. In the approach described above:
+
+@itemize @bullet
+
+@item
+It's easier to see that the analyzer is correct: you just see
+whether the analyzer properly (albiet conservatively) simulates
+the effect of each instruction.
+
+@item
+It's easier to extend the analyzer: you can add support for new
+instructions, and know that you haven't broken anything that
+wasn't already broken before.
+
+@item
+It's orthogonal: to gather new information, you don't need to
+complicate the code for each instruction. As long as your domain
+of conservative values is already detailed enough to tell you
+what you need, then all the existing instruction simulations are
+already gathering the right data for you.
+
+@end itemize
+
+The file @file{prologue-value.h} contains detailed comments explaining
+the framework and how to use it.
+
+
@section Breakpoint Handling
@cindex breakpoints
--- /dev/null
+/* Prologue value handling for GDB.
+ Copyright 2003, 2004, 2005 Free Software Foundation, Inc.
+
+ This file is part of GDB.
+
+ This program 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 2 of the License, or
+ (at your option) any later version.
+
+ This program 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:
+
+ Free Software Foundation, Inc.
+ 51 Franklin St - Fifth Floor
+ Boston, MA 02110-1301
+ USA */
+
+#include "defs.h"
+#include "gdb_string.h"
+#include "gdb_assert.h"
+#include "prologue-value.h"
+#include "regcache.h"
+
+\f
+/* Constructors. */
+
+pv_t
+pv_unknown (void)
+{
+ pv_t v = { pvk_unknown, 0, 0 };
+
+ return v;
+}
+
+
+pv_t
+pv_constant (CORE_ADDR k)
+{
+ pv_t v;
+
+ v.kind = pvk_constant;
+ v.reg = -1; /* for debugging */
+ v.k = k;
+
+ return v;
+}
+
+
+pv_t
+pv_register (int reg, CORE_ADDR k)
+{
+ pv_t v;
+
+ v.kind = pvk_register;
+ v.reg = reg;
+ v.k = k;
+
+ return v;
+}
+
+
+\f
+/* Arithmetic operations. */
+
+/* If one of *A and *B is a constant, and the other isn't, swap the
+ values as necessary to ensure that *B is the constant. This can
+ reduce the number of cases we need to analyze in the functions
+ below. */
+static void
+constant_last (pv_t *a, pv_t *b)
+{
+ if (a->kind == pvk_constant
+ && b->kind != pvk_constant)
+ {
+ pv_t temp = *a;
+ *a = *b;
+ *b = temp;
+ }
+}
+
+
+pv_t
+pv_add (pv_t a, pv_t b)
+{
+ constant_last (&a, &b);
+
+ /* We can add a constant to a register. */
+ if (a.kind == pvk_register
+ && b.kind == pvk_constant)
+ return pv_register (a.reg, a.k + b.k);
+
+ /* We can add a constant to another constant. */
+ else if (a.kind == pvk_constant
+ && b.kind == pvk_constant)
+ return pv_constant (a.k + b.k);
+
+ /* Anything else we don't know how to add. We don't have a
+ representation for, say, the sum of two registers, or a multiple
+ of a register's value (adding a register to itself). */
+ else
+ return pv_unknown ();
+}
+
+
+pv_t
+pv_add_constant (pv_t v, CORE_ADDR k)
+{
+ /* Rather than thinking of all the cases we can and can't handle,
+ we'll just let pv_add take care of that for us. */
+ return pv_add (v, pv_constant (k));
+}
+
+
+pv_t
+pv_subtract (pv_t a, pv_t b)
+{
+ /* This isn't quite the same as negating B and adding it to A, since
+ we don't have a representation for the negation of anything but a
+ constant. For example, we can't negate { pvk_register, R1, 10 },
+ but we do know that { pvk_register, R1, 10 } minus { pvk_register,
+ R1, 5 } is { pvk_constant, <ignored>, 5 }.
+
+ This means, for example, that we could subtract two stack
+ addresses; they're both relative to the original SP. Since the
+ frame pointer is set based on the SP, its value will be the
+ original SP plus some constant (probably zero), so we can use its
+ value just fine, too. */
+
+ constant_last (&a, &b);
+
+ /* We can subtract two constants. */
+ if (a.kind == pvk_constant
+ && b.kind == pvk_constant)
+ return pv_constant (a.k - b.k);
+
+ /* We can subtract a constant from a register. */
+ else if (a.kind == pvk_register
+ && b.kind == pvk_constant)
+ return pv_register (a.reg, a.k - b.k);
+
+ /* We can subtract a register from itself, yielding a constant. */
+ else if (a.kind == pvk_register
+ && b.kind == pvk_register
+ && a.reg == b.reg)
+ return pv_constant (a.k - b.k);
+
+ /* We don't know how to subtract anything else. */
+ else
+ return pv_unknown ();
+}
+
+
+pv_t
+pv_logical_and (pv_t a, pv_t b)
+{
+ constant_last (&a, &b);
+
+ /* We can 'and' two constants. */
+ if (a.kind == pvk_constant
+ && b.kind == pvk_constant)
+ return pv_constant (a.k & b.k);
+
+ /* We can 'and' anything with the constant zero. */
+ else if (b.kind == pvk_constant
+ && b.k == 0)
+ return pv_constant (0);
+
+ /* We can 'and' anything with ~0. */
+ else if (b.kind == pvk_constant
+ && b.k == ~ (CORE_ADDR) 0)
+ return a;
+
+ /* We can 'and' a register with itself. */
+ else if (a.kind == pvk_register
+ && b.kind == pvk_register
+ && a.reg == b.reg
+ && a.k == b.k)
+ return a;
+
+ /* Otherwise, we don't know. */
+ else
+ return pv_unknown ();
+}
+
+
+\f
+/* Examining prologue values. */
+
+int
+pv_is_identical (pv_t a, pv_t b)
+{
+ if (a.kind != b.kind)
+ return 0;
+
+ switch (a.kind)
+ {
+ case pvk_unknown:
+ return 1;
+ case pvk_constant:
+ return (a.k == b.k);
+ case pvk_register:
+ return (a.reg == b.reg && a.k == b.k);
+ default:
+ gdb_assert (0);
+ }
+}
+
+
+int
+pv_is_constant (pv_t a)
+{
+ return (a.kind == pvk_constant);
+}
+
+
+int
+pv_is_register (pv_t a, int r)
+{
+ return (a.kind == pvk_register
+ && a.reg == r);
+}
+
+
+int
+pv_is_register_k (pv_t a, int r, CORE_ADDR k)
+{
+ return (a.kind == pvk_register
+ && a.reg == r
+ && a.k == k);
+}
+
+
+enum pv_boolean
+pv_is_array_ref (pv_t addr, CORE_ADDR size,
+ pv_t array_addr, CORE_ADDR array_len,
+ CORE_ADDR elt_size,
+ int *i)
+{
+ /* Note that, since .k is a CORE_ADDR, and CORE_ADDR is unsigned, if
+ addr is *before* the start of the array, then this isn't going to
+ be negative... */
+ pv_t offset = pv_subtract (addr, array_addr);
+
+ if (offset.kind == pvk_constant)
+ {
+ /* This is a rather odd test. We want to know if the SIZE bytes
+ at ADDR don't overlap the array at all, so you'd expect it to
+ be an || expression: "if we're completely before || we're
+ completely after". But with unsigned arithmetic, things are
+ different: since it's a number circle, not a number line, the
+ right values for offset.k are actually one contiguous range. */
+ if (offset.k <= -size
+ && offset.k >= array_len * elt_size)
+ return pv_definite_no;
+ else if (offset.k % elt_size != 0
+ || size != elt_size)
+ return pv_maybe;
+ else
+ {
+ *i = offset.k / elt_size;
+ return pv_definite_yes;
+ }
+ }
+ else
+ return pv_maybe;
+}
+
+
+\f
+/* Areas. */
+
+
+/* A particular value known to be stored in an area.
+
+ Entries form a ring, sorted by unsigned offset from the area's base
+ register's value. Since entries can straddle the wrap-around point,
+ unsigned offsets form a circle, not a number line, so the list
+ itself is structured the same way --- there is no inherent head.
+ The entry with the lowest offset simply follows the entry with the
+ highest offset. Entries may abut, but never overlap. The area's
+ 'entry' pointer points to an arbitrary node in the ring. */
+struct area_entry
+{
+ /* Links in the doubly-linked ring. */
+ struct area_entry *prev, *next;
+
+ /* Offset of this entry's address from the value of the base
+ register. */
+ CORE_ADDR offset;
+
+ /* The size of this entry. Note that an entry may wrap around from
+ the end of the address space to the beginning. */
+ CORE_ADDR size;
+
+ /* The value stored here. */
+ pv_t value;
+};
+
+
+struct pv_area
+{
+ /* This area's base register. */
+ int base_reg;
+
+ /* The mask to apply to addresses, to make the wrap-around happen at
+ the right place. */
+ CORE_ADDR addr_mask;
+
+ /* An element of the doubly-linked ring of entries, or zero if we
+ have none. */
+ struct area_entry *entry;
+};
+
+
+struct pv_area *
+make_pv_area (int base_reg)
+{
+ struct pv_area *a = (struct pv_area *) xmalloc (sizeof (*a));
+
+ memset (a, 0, sizeof (*a));
+
+ a->base_reg = base_reg;
+ a->entry = 0;
+
+ /* Remember that shift amounts equal to the type's width are
+ undefined. */
+ a->addr_mask = ((((CORE_ADDR) 1 << (TARGET_ADDR_BIT - 1)) - 1) << 1) | 1;
+
+ return a;
+}
+
+
+/* Delete all entries from AREA. */
+static void
+clear_entries (struct pv_area *area)
+{
+ struct area_entry *e = area->entry;
+
+ if (e)
+ {
+ /* This needs to be a do-while loop, in order to actually
+ process the node being checked for in the terminating
+ condition. */
+ do
+ {
+ struct area_entry *next = e->next;
+ xfree (e);
+ }
+ while (e != area->entry);
+
+ area->entry = 0;
+ }
+}
+
+
+void
+free_pv_area (struct pv_area *area)
+{
+ clear_entries (area);
+ xfree (area);
+}
+
+
+static void
+do_free_pv_area_cleanup (void *arg)
+{
+ free_pv_area ((struct pv_area *) arg);
+}
+
+
+struct cleanup *
+make_cleanup_free_pv_area (struct pv_area *area)
+{
+ return make_cleanup (do_free_pv_area_cleanup, (void *) area);
+}
+
+
+int
+pv_area_store_would_trash (struct pv_area *area, pv_t addr)
+{
+ /* It may seem odd that pvk_constant appears here --- after all,
+ that's the case where we know the most about the address! But
+ pv_areas are always relative to a register, and we don't know the
+ value of the register, so we can't compare entry addresses to
+ constants. */
+ return (addr.kind == pvk_unknown
+ || addr.kind == pvk_constant
+ || (addr.kind == pvk_register && addr.reg != area->base_reg));
+}
+
+
+/* Return a pointer to the first entry we hit in AREA starting at
+ OFFSET and going forward.
+
+ This may return zero, if AREA has no entries.
+
+ And since the entries are a ring, this may return an entry that
+ entirely preceeds OFFSET. This is the correct behavior: depending
+ on the sizes involved, we could still overlap such an area, with
+ wrap-around. */
+static struct area_entry *
+find_entry (struct pv_area *area, CORE_ADDR offset)
+{
+ struct area_entry *e = area->entry;
+
+ if (! e)
+ return 0;
+
+ /* If the next entry would be better than the current one, then scan
+ forward. Since we use '<' in this loop, it always terminates.
+
+ Note that, even setting aside the addr_mask stuff, we must not
+ simplify this, in high school algebra fashion, to
+ (e->next->offset < e->offset), because of the way < interacts
+ with wrap-around. We have to subtract offset from both sides to
+ make sure both things we're comparing are on the same side of the
+ discontinuity. */
+ while (((e->next->offset - offset) & area->addr_mask)
+ < ((e->offset - offset) & area->addr_mask))
+ e = e->next;
+
+ /* If the previous entry would be better than the current one, then
+ scan backwards. */
+ while (((e->prev->offset - offset) & area->addr_mask)
+ < ((e->offset - offset) & area->addr_mask))
+ e = e->prev;
+
+ /* In case there's some locality to the searches, set the area's
+ pointer to the entry we've found. */
+ area->entry = e;
+
+ return e;
+}
+
+
+/* Return non-zero if the SIZE bytes at OFFSET would overlap ENTRY;
+ return zero otherwise. AREA is the area to which ENTRY belongs. */
+static int
+overlaps (struct pv_area *area,
+ struct area_entry *entry,
+ CORE_ADDR offset,
+ CORE_ADDR size)
+{
+ /* Think carefully about wrap-around before simplifying this. */
+ return (((entry->offset - offset) & area->addr_mask) < size
+ || ((offset - entry->offset) & area->addr_mask) < entry->size);
+}
+
+
+void
+pv_area_store (struct pv_area *area,
+ pv_t addr,
+ CORE_ADDR size,
+ pv_t value)
+{
+ /* Remove any (potentially) overlapping entries. */
+ if (pv_area_store_would_trash (area, addr))
+ clear_entries (area);
+ else
+ {
+ CORE_ADDR offset = addr.k;
+ struct area_entry *e = find_entry (area, offset);
+
+ /* Delete all entries that we would overlap. */
+ while (e && overlaps (area, e, offset, size))
+ {
+ struct area_entry *next = (e->next == e) ? 0 : e->next;
+ e->prev->next = e->next;
+ e->next->prev = e->prev;
+
+ xfree (e);
+ e = next;
+ }
+
+ /* Move the area's pointer to the next remaining entry. This
+ will also zero the pointer if we've deleted all the entries. */
+ area->entry = e;
+ }
+
+ /* Now, there are no entries overlapping us, and area->entry is
+ either zero or pointing at the closest entry after us. We can
+ just insert ourselves before that.
+
+ But if we're storing an unknown value, don't bother --- that's
+ the default. */
+ if (value.kind == pvk_unknown)
+ return;
+ else
+ {
+ CORE_ADDR offset = addr.k;
+ struct area_entry *e = (struct area_entry *) xmalloc (sizeof (*e));
+ e->offset = offset;
+ e->size = size;
+ e->value = value;
+
+ if (area->entry)
+ {
+ e->prev = area->entry->prev;
+ e->next = area->entry;
+ e->prev->next = e->next->prev = e;
+ }
+ else
+ {
+ e->prev = e->next = e;
+ area->entry = e;
+ }
+ }
+}
+
+
+pv_t
+pv_area_fetch (struct pv_area *area, pv_t addr, CORE_ADDR size)
+{
+ /* If we have no entries, or we can't decide how ADDR relates to the
+ entries we do have, then the value is unknown. */
+ if (! area->entry
+ || pv_area_store_would_trash (area, addr))
+ return pv_unknown ();
+ else
+ {
+ CORE_ADDR offset = addr.k;
+ struct area_entry *e = find_entry (area, offset);
+
+ /* If this entry exactly matches what we're looking for, then
+ we're set. Otherwise, say it's unknown. */
+ if (e->offset == offset && e->size == size)
+ return e->value;
+ else
+ return pv_unknown ();
+ }
+}
+
+
+int
+pv_area_find_reg (struct pv_area *area,
+ struct gdbarch *gdbarch,
+ int reg,
+ CORE_ADDR *offset_p)
+{
+ struct area_entry *e = area->entry;
+
+ if (e)
+ do
+ {
+ if (e->value.kind == pvk_register
+ && e->value.reg == reg
+ && e->value.k == 0
+ && e->size == register_size (gdbarch, reg))
+ {
+ if (offset_p)
+ *offset_p = e->offset;
+ return 1;
+ }
+
+ e = e->next;
+ }
+ while (e != area->entry);
+
+ return 0;
+}
+
+
+void
+pv_area_scan (struct pv_area *area,
+ void (*func) (void *closure,
+ pv_t addr,
+ CORE_ADDR size,
+ pv_t value),
+ void *closure)
+{
+ struct area_entry *e = area->entry;
+ pv_t addr;
+
+ addr.kind = pvk_register;
+ addr.reg = area->base_reg;
+
+ if (e)
+ do
+ {
+ addr.k = e->offset;
+ func (closure, addr, e->size, e->value);
+ e = e->next;
+ }
+ while (e != area->entry);
+}
--- /dev/null
+/* Interface to prologue value handling for GDB.
+ Copyright 2003, 2004, 2005 Free Software Foundation, Inc.
+
+ This file is part of GDB.
+
+ This program 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 2 of the License, or
+ (at your option) any later version.
+
+ This program 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:
+
+ Free Software Foundation, Inc.
+ 51 Franklin St - Fifth Floor
+ Boston, MA 02110-1301
+ USA */
+
+#ifndef PROLOGUE_VALUE_H
+#define PROLOGUE_VALUE_H
+
+/* When we analyze a prologue, we're really doing 'abstract
+ interpretation' or 'pseudo-evaluation': running the function's code
+ in simulation, but using conservative approximations of the values
+ it would have when it actually runs. For example, if our function
+ starts with the instruction:
+
+ addi r1, 42 # add 42 to r1
+
+ we don't know exactly what value will be in r1 after executing this
+ instruction, but we do know it'll be 42 greater than its original
+ value.
+
+ If we then see an instruction like:
+
+ addi r1, 22 # add 22 to r1
+
+ we still don't know what r1's value is, but again, we can say it is
+ now 64 greater than its original value.
+
+ If the next instruction were:
+
+ mov r2, r1 # set r2 to r1's value
+
+ then we can say that r2's value is now the original value of r1
+ plus 64.
+
+ It's common for prologues to save registers on the stack, so we'll
+ need to track the values of stack frame slots, as well as the
+ registers. So after an instruction like this:
+
+ mov (fp+4), r2
+
+ then we'd know that the stack slot four bytes above the frame
+ pointer holds the original value of r1 plus 64.
+
+ And so on.
+
+ Of course, this can only go so far before it gets unreasonable. If
+ we wanted to be able to say anything about the value of r1 after
+ the instruction:
+
+ xor r1, r3 # exclusive-or r1 and r3, place result in r1
+
+ then things would get pretty complex. But remember, we're just
+ doing a conservative approximation; if exclusive-or instructions
+ aren't relevant to prologues, we can just say r1's value is now
+ 'unknown'. We can ignore things that are too complex, if that loss
+ of information is acceptable for our application.
+
+ So when I say "conservative approximation" here, what I mean is an
+ approximation that is either accurate, or marked "unknown", but
+ never inaccurate.
+
+ Once you've reached the current PC, or an instruction that you
+ don't know how to simulate, you stop. Now you can examine the
+ state of the registers and stack slots you've kept track of.
+
+ - To see how large your stack frame is, just check the value of the
+ stack pointer register; if it's the original value of the SP
+ minus a constant, then that constant is the stack frame's size.
+ If the SP's value has been marked as 'unknown', then that means
+ the prologue has done something too complex for us to track, and
+ we don't know the frame size.
+
+ - To see where we've saved the previous frame's registers, we just
+ search the values we've tracked --- stack slots, usually, but
+ registers, too, if you want --- for something equal to the
+ register's original value. If the ABI suggests a standard place
+ to save a given register, then we can check there first, but
+ really, anything that will get us back the original value will
+ probably work.
+
+ Sure, this takes some work. But prologue analyzers aren't
+ quick-and-simple pattern patching to recognize a few fixed prologue
+ forms any more; they're big, hairy functions. Along with inferior
+ function calls, prologue analysis accounts for a substantial
+ portion of the time needed to stabilize a GDB port. So I think
+ it's worthwhile to look for an approach that will be easier to
+ understand and maintain. In the approach used here:
+
+ - It's easier to see that the analyzer is correct: you just see
+ whether the analyzer properly (albiet conservatively) simulates
+ the effect of each instruction.
+
+ - It's easier to extend the analyzer: you can add support for new
+ instructions, and know that you haven't broken anything that
+ wasn't already broken before.
+
+ - It's orthogonal: to gather new information, you don't need to
+ complicate the code for each instruction. As long as your domain
+ of conservative values is already detailed enough to tell you
+ what you need, then all the existing instruction simulations are
+ already gathering the right data for you.
+
+ A 'struct prologue_value' is a conservative approximation of the
+ real value the register or stack slot will have. */
+
+struct prologue_value {
+
+ /* What sort of value is this? This determines the interpretation
+ of subsequent fields. */
+ enum {
+
+ /* We don't know anything about the value. This is also used for
+ values we could have kept track of, when doing so would have
+ been too complex and we don't want to bother. The bottom of
+ our lattice. */
+ pvk_unknown,
+
+ /* A known constant. K is its value. */
+ pvk_constant,
+
+ /* The value that register REG originally had *UPON ENTRY TO THE
+ FUNCTION*, plus K. If K is zero, this means, obviously, just
+ the value REG had upon entry to the function. REG is a GDB
+ register number. Before we start interpreting, we initialize
+ every register R to { pvk_register, R, 0 }. */
+ pvk_register,
+
+ } kind;
+
+ /* The meanings of the following fields depend on 'kind'; see the
+ comments for the specific 'kind' values. */
+ int reg;
+ CORE_ADDR k;
+};
+
+typedef struct prologue_value pv_t;
+
+
+/* Return the unknown prologue value --- { pvk_unknown, ?, ? }. */
+pv_t pv_unknown (void);
+
+/* Return the prologue value representing the constant K. */
+pv_t pv_constant (CORE_ADDR k);
+
+/* Return the prologue value representing the original value of
+ register REG, plus the constant K. */
+pv_t pv_register (int reg, CORE_ADDR k);
+
+
+/* Return conservative approximations of the results of the following
+ operations. */
+pv_t pv_add (pv_t a, pv_t b); /* a + b */
+pv_t pv_add_constant (pv_t v, CORE_ADDR k); /* a + k */
+pv_t pv_subtract (pv_t a, pv_t b); /* a - b */
+pv_t pv_logical_and (pv_t a, pv_t b); /* a & b */
+
+
+/* Return non-zero iff A and B are identical expressions.
+
+ This is not the same as asking if the two values are equal; the
+ result of such a comparison would have to be a pv_boolean, and
+ asking whether two 'unknown' values were equal would give you
+ pv_maybe. Same for comparing, say, { pvk_register, R1, 0 } and {
+ pvk_register, R2, 0}.
+
+ Instead, this function asks whether the two representations are the
+ same. */
+int pv_is_identical (pv_t a, pv_t b);
+
+
+/* Return non-zero if A is known to be a constant. */
+int pv_is_constant (pv_t a);
+
+/* Return non-zero if A is the original value of register number R
+ plus some constant, zero otherwise. */
+int pv_is_register (pv_t a, int r);
+
+
+/* Return non-zero if A is the original value of register R plus the
+ constant K. */
+int pv_is_register_k (pv_t a, int r, CORE_ADDR k);
+
+/* A conservative boolean type, including "maybe", when we can't
+ figure out whether something is true or not. */
+enum pv_boolean {
+ pv_maybe,
+ pv_definite_yes,
+ pv_definite_no,
+};
+
+
+/* Decide whether a reference to SIZE bytes at ADDR refers exactly to
+ an element of an array. The array starts at ARRAY_ADDR, and has
+ ARRAY_LEN values of ELT_SIZE bytes each. If ADDR definitely does
+ refer to an array element, set *I to the index of the referenced
+ element in the array, and return pv_definite_yes. If it definitely
+ doesn't, return pv_definite_no. If we can't tell, return pv_maybe.
+
+ If the reference does touch the array, but doesn't fall exactly on
+ an element boundary, or doesn't refer to the whole element, return
+ pv_maybe. */
+enum pv_boolean pv_is_array_ref (pv_t addr, CORE_ADDR size,
+ pv_t array_addr, CORE_ADDR array_len,
+ CORE_ADDR elt_size,
+ int *i);
+
+
+/* A 'struct pv_area' keeps track of values stored in a particular
+ region of memory. */
+struct pv_area;
+
+/* Create a new area, tracking stores relative to the original value
+ of BASE_REG. If BASE_REG is SP, then this effectively records the
+ contents of the stack frame: the original value of the SP is the
+ frame's CFA, or some constant offset from it.
+
+ Stores to constant addresses, unknown addresses, or to addresses
+ relative to registers other than BASE_REG will trash this area; see
+ pv_area_store_would_trash. */
+struct pv_area *make_pv_area (int base_reg);
+
+/* Free AREA. */
+void free_pv_area (struct pv_area *area);
+
+
+/* Register a cleanup to free AREA. */
+struct cleanup *make_cleanup_free_pv_area (struct pv_area *area);
+
+
+/* Store the SIZE-byte value VALUE at ADDR in AREA.
+
+ If ADDR is not relative to the same base register we used in
+ creating AREA, then we can't tell which values here the stored
+ value might overlap, and we'll have to mark everything as
+ unknown. */
+void pv_area_store (struct pv_area *area,
+ pv_t addr,
+ CORE_ADDR size,
+ pv_t value);
+
+/* Return the SIZE-byte value at ADDR in AREA. This may return
+ pv_unknown (). */
+pv_t pv_area_fetch (struct pv_area *area, pv_t addr, CORE_ADDR size);
+
+/* Return true if storing to address ADDR in AREA would force us to
+ mark the contents of the entire area as unknown. This could happen
+ if, say, ADDR is unknown, since we could be storing anywhere. Or,
+ it could happen if ADDR is relative to a different register than
+ the other stores base register, since we don't know the relative
+ values of the two registers.
+
+ If you've reached such a store, it may be better to simply stop the
+ prologue analysis, and return the information you've gathered,
+ instead of losing all that information, most of which is probably
+ okay. */
+int pv_area_store_would_trash (struct pv_area *area, pv_t addr);
+
+
+/* Search AREA for the original value of REGISTER. If we can't find
+ it, return zero; if we can find it, return a non-zero value, and if
+ OFFSET_P is non-zero, set *OFFSET_P to the register's offset within
+ AREA. GDBARCH is the architecture of which REGISTER is a member.
+
+ In the worst case, this takes time proportional to the number of
+ items stored in AREA. If you plan to gather a lot of information
+ about registers saved in AREA, consider calling pv_area_scan
+ instead, and collecting all your information in one pass. */
+int pv_area_find_reg (struct pv_area *area,
+ struct gdbarch *gdbarch,
+ int register,
+ CORE_ADDR *offset_p);
+
+
+/* For every part of AREA whose value we know, apply FUNC to CLOSURE,
+ the value's address, its size, and the value itself. */
+void pv_area_scan (struct pv_area *area,
+ void (*func) (void *closure,
+ pv_t addr,
+ CORE_ADDR size,
+ pv_t value),
+ void *closure);
+
+
+#endif /* PROLOGUE_VALUE_H */
projects / binutils-gdb.git / commitdiff