There was a lot of needless handshaking overhead in the current
Cortex-A8 DCC/ITR operations, since the status read by each step
was discarded rather than letting the next step know it.
This shrinks the handshaking by: (a) passing status along from
previous steps, avoiding re-fetching; which enables the big win
(b) relying on a useful invariant: that the DSCR_INSTR_COMP bit
is set after every call to a DPM method.
A "reg sp_usr" call previously took 17 flushes; now it takes just 9.
This visibly speeds common operations like entry to debug state and
stepping, as well as "arm reg" and so on.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
This replaces two versions of register access functions. One
was commented out, and seemed to have uncertain intent. The
other was fairly new, and helped motivate the DPM framework
once I observed that the ARM11 was doing the very same ops.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
This implements the DPM interface for Cortex-A8 cores. It
also adds a synchronization operation to the DPM framework,
which is needed by the Cortex-A8 after CPSR writes.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Make various functions static, add some comments, report
vector catch as a flavor of DBG_REASON_BREAKPOINT, get
rid of needless/undesirable ARMV4_5_CORE_REG_MODE, etc.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
This finishes the basic switchover to the new register code,
for everything except the debug registers. (And maybe we
shouldn't have a cache for *those* which works this way...)
The context save/restore code now uses the new code, but
it's in a slightly different sequence. That should be fine
since the R0/PC/CPSR stuff is all that really matters (and
if we can update those, we can update the rest).
Now there's no longer a way any code can be confused about
which copy of "r1" (etc) to use.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
As with single stepping, the previous stuff was needed because
the ARM11 code wasn't using the standard ARM base type and
register access ... but now those mechanisms work, so we can
switch out that special-purpose glue, in favor of the more
thoroughly tested/capable "standard" code.
Fixes a bug in the resume() implementation: it wasn't handling
two of its arguments correctly, preventing the "flash erase_check"
algorithm from working. (This code needs a *subsequent* update
for correct register handling, though... removing the confusion
about which "r2", for example, to use.)
This should resolve some "FIXME" comments too, for Thumb and
processor mode support. It also gets rid of a nasty exit()
call; servers should only have *clean* shutdown paths.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The previous stuff was needed because the ARM11 code wasn't using
the standard ARM base type and register access ... but now those
mechanisms work, so we can switch out that special-purpose glue.
This should resolve all the "FIXME -- handle Thumb single stepping"
comments too, and properly handle the processor's mode. (Modulo
the issue that this code doesn't yet handle two-byte breakpoints.)
Clarify the comments about the the hardware single stepping. When
we eventually share breakpoint code with Cortex-A8, we can just make
that be the default on cores which support it. We may still want an
override command, not just to facilitate testing but to cope with
"instruction address mismatch" not quite being true single-step.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
This provides "standard" ARM register support -- with twenty or
more shadow registers on top of what this code now handles, but
properly associated with the various core modes -- parallel to
the current register code. That is, the current code is stilil
managing the "current" registers; the new code shadows them.
You can see all the registers with "arm reg", modify the shadows
like "r8_fiq" or "sp_abt" with "reg", and see them get properly
written back when you step. (Just don't do that with any of the
registers managed by the "old" code ...)
It also switches to using more standard code, relying on those
standard registers, in two places: (a) the poll status display,
which now shows core state (ARM/Thumb/...) and mode (Supervisor,
IRQ, etc); and (b) GDB register access.
So it's not a full migration, there are warts -- every place that
touches the old register cache is a potential bug -- but it's a
small more-or-less-comprehensible step that's even somewhat useful.
Later patches complete the migration.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
This was a private mechanism to snapshot registers before leaving
debug state, and then on reentry to optionally display what changed.
It was coupled to the private register cache, which won't be sticking
around in that form for much longer. Remove (instead of teaching
it how to handle *all* the registers).
(The idea is interesting, but we ought to be able to implement
this in a generic way. Ideally through Tcl scripts that can
automatically be invoked following debug entry...)
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
This is a very thin layer over some of the current ARM11
debug TAP utilities. The layer isn't yet hooked up.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
First version of interface for sharing code between ARMv6 and ARMv7a
debug modules ... now the architecture includes debug support. (Not
the same as for the trimmed-down v7m or v6m though!) This is a first
version of an interface that will let the ARM11 and Cortex-A8 support
share code, features, and bugfixes. Based on existing code from both
of those cores.
The ARM v7-AR architecture specification calls this commonality the
"Debug Programmer's Model (DPM)", which seemed to be an appropriate
acronym -- a TLA even! -- for use in our code. Made it so. :)
The initial scope of this just supports register access, and is geared
towards supporting top level "struct arm" mechanisms. Later, things
like breakpoint and watchpoint support should be included.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
For now there's no point in saving this stuff after examine()
checks it out as OK. Ditto exporting symbols that aren't
used outside of the module which defines them. In fact, those
two things needlessly complicate the code...
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Make this code look more like the rest of the OpenOCD code.
- Use calloc() directly, not NEW() ... and fix some potential
memory leaks while we're at it.
- Remove FNC_INFO ... it's a NOP that just clutters things,
and it's trivial for developers to add tracing as needed.
- Replace FNC_INFO_NOTIMPLEMENTED with LOG_WARNING calls;
ditto. And stop having those call sites wrongly succeed!
- Waste less space with the CHECK_RETVAL() macro.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
This will allow data to be allocated in read only
memory instead of on the stack. Speeds things up
and reduces stack usage.
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
buf_set_u32() operated on an uninitialized stack
variable with non-byte boundaries, which led to
warnings about reading uninitialized stack.
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
Start using the arm_reg_current() call. This shrinks and speeds
the affected code. It can also prevent some coredumps coming from
invalid CPSR values ... the ARMV4_5_CORE_REG_MODE() macro returns
bogus registers if e.g. "Secure Monitor" mode isn't supported by
the current CPU.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
We don't need to use size_t in these places; so it's easy
to be rid of the need for this #ifdef and its MS-derived
portability problems.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
With -O3 when inlining aggressively the total stack usage will
be the sum of many fn's, which can easily get out of hand.
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
Have arm_set_cpsr() handle the two core state flags, updating
the CPU state. This eliminates code in various debug_entry()
paths, and marginally improves handling of the J bit.
Catch and comment a few holes in the handling of the J bit on
ARM926ejs cores ... it's unlikely our users will care about
Jazelle mode, but we can at least warn of Impending Doom. If
anyone does use it, these breadcrumbs may help them to find
the right path through the code.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Define arm_reg_current() ... returning handle to a given register,
and encapsulating the current mode's register shadowing. It's got
one current use, for reporting the current register set to GDB.
This will let later patches clean up much ARMV4_5_CORE_REG_MODE()
nastiness, saving a bit of code.
Define and use arm_set_cpsr() ... initially it updates the cached
CPSR and sets up state used by arm_reg_current(), plus any SPSR
handle. (Later: can also set up for T and J bits.)
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Stash a pointer to the CPSR in the "struct arm", to help get rid
of the (common) references to its index in the register cache.
This removes almost all references to CPSR offsets outside of the
toplevel ARM code ... except a pair related to the current ARM11
"simulator" logic (which should be removable soonish).
This is a net minor code shrink of a few hundred bytes of object
code, and also makes the code more readable.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Minor cleanup of ARM11 register handling: remove disabled
register hooks. This should all be handled by shared code,
and this stuff is just clutter.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Implementations need to access the register struct they modify;
make it easier and less error-prone to identify the instance.
(This removes over 10% of the ARMV4_5_CORE_REG_MODE nastiness...)
Plus some minor fixes noted when making these updates: ARM7/ARM9
accessor methods should be static; don't leave CPSR wrongly marked
"dirty"; note significant XScale omissions in register handling;
and have armv4_5_build_reg_cache() record its result.
Rename "struct armv4_5_core_reg" as "struct arm_reg"; it's used
for more than those older architecture generations.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Remove two commands that were documented as "debug commands"
and where "you probably don't want to use this". We never
intended to support them, and at least one problem report
boiled down to using this when it shouldn't have been used.
Update the docs on the existing register commands to talk a
bit more about register access and cache behavior. (Those
debug commands existed largely to *bypass* the cache.) And
fix some minor doc goofs that snuck in with recent changes,
renaming "armv4_5" as "arm" and "arm9tdmi" as "arm9".
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The previous version never wrote dirty registers
for non-current CPU modes ... fix that.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
We *should* be able to read and write registers in any core mode,
instead of being stuck with whatever mode the core was when we
entered debug state. This patch makes them work.
Note that the current restore_context() only handles the current
mode; writing to other-mode registers is a NOP without a followup
patch fixing that. Also, that SPSR access needed some bugfixes;
it was confused with CPSR.
Secure monitor mode also seems dubious; there's probably more to
be done before that's sufficiently understood by the debugger.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Create a generic register_cache_invalidate(), and use it to
replace three all-but-identical core-specific routines:
- armv4_5_invalidate_core_regs()
- armv7m_invalidate_core_regs
- mips32_invalidate_core_regs() too.
Make cache->num_regs be unsigned, avoiding various errors.
Net code shrink and simplification.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Just pre-allocate memory for the cached register value.
Shrinks heap overhead; increases locality-of-reference.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Combine register names with other per-register data into a
single template structure. This saves space, and makes it
easier to change how registers get handled (by shrinking
the number of places that care about cache indices).
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The examine() method has some conceptual breakage. Cope
with it by manually splitting out the run-once parts from
the after-each-reset parts ... this gets rid of memory
leaks and speeds up resets after the first one.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
We don't want an ARMv7-specific core state enumeration just to
add ThumbEE state. Update the generic stuff to handle that,
and replace the V7-specific bits with it.
For Cortex-A8: on debug entry, check both the T and J bits
instead of just the T bit. When the J bit is set, set the
right state and warn appropriately.
(And while we're at it, move the generic arm struct to the front
of the v7a structure, for somewhat better code generation.)
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The only way ARMv7-A modes differ from ARMv4/ARMv5 flavors
is that v7-A is allowed to include "Secure monitor" support.
That's now handled by our standard top-level ARM code ... so
phase out the stuff that's specific to ARMv7-A.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Removing the fast command eliminates the fast_and_dangerous global,
which was used only by arm7_9_common as an initializer. The command
is not called in the tree; instead, more explicit commands are used.
The jim_global_long function was not used anywhere in the tree.
Rewrite arm11_handle_bool to provide a generic on/off command helper.
Refactors COMMAND_PARSE_BOOL to use new command_parse_bool helper,
which gets reused by the new command_parse_bool_any helper.
This later helper is called by the new command helper function to
accepts any on/off, enable/disable, true/false, yes/no, or 0/1 parameter.
Updates all command parsing of simple "enable" and "disable" arguments.
A few case in the tree use a tri-state or extended arguments, which
cannot use this simple macro.
Simlifies the xscale icache/dcache command handler logic.
Change the layout to show the "Secure Monitor" registers too,
when they're present.
Instead of lining registers for each of six (or seven) modes up
in adjacent vertical columns, display each mode's registers (or
shadows) in a single block, avoiding duplicate value displays.
This also lets us shrink the line length to fits in standard 80
character lines ... six or seven 18-character columns can't fit.
Relabel "r13" as "sp", so it's more meaningful.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
When we read the CPSR on debug entry, update the CPSR cache in all
cases, not just when the current processor state is User or System.
Plus minor cleanup of how the (too-many) other registers' cache
entries get updated.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
If the core doesn't provide an optimized version of this
method, provide one without core-specific optimizations.
Use this to make Cortex-A8 support the "arm reg" command.
Related: make the two register access methods properly static,
have the "set" log a "not halted" error too, and make sure
that the "valid" flag is set on successful reads.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
ARMv7-A doesn't need to duplicate all the standard ARM code
for register handling.
- Switch Cortex-A8 to use the standard register code
- Remove duplicated infrastructure from ARMv7-A
- Have ARMv7-A arch_state() show CPSR, like other ARMs
Add comments to show where the Cortex-A8 isn't actually doing
the right thing for register reads/writes, unless core happens
to be in the right mode to start with. (Looks like maybe there
may be generic confusion between saved/current PSR values in all
the ARM code ...)
Make related ARMv7-A and Cortex-A8 symbols properly static.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Teach the "armv4_5" register code to understand about the
secure monitor mode:
- Add the other three shadowed registers to the arrays
- Support another internal mode number (sigh) in mappings
- Catch malloc/calloc failures building that register cache
This should kick in for Cortex-A8 and ARM1176.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
It's used to flag cores with the "TrustZone" extension,
and is used in subsequent patches to set up support for
the registers shadowed by its new secure monitor mode.
The ARM1176 and Cortex-A8 both support this new mode.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Moves definitions for each layer into their own file, eliminating
layering violations in the built-in TCL code. Updates src/Makefile.am
rules to include all files in the final startup.tcl input file, and
others Makefile.am rules to distribute the new files in our packages.
The recent migration broke them, the fixes broken them in a new way,
but this should restore them to working order. Eliminates the
temporary variable, as the CMD_NAME macro can once again be use
in routines that increment CMD_ARGV without nasty side-effects.
Also, this is on the path to increasing the word size for
bit vectors from 8 to something wider(32? natural host machine
width?)
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
Add a new is_arm_mode() predicate, and use it to replace almost
all calls to current armv4_5_mode_to_number().
Eventually those internal mode numbers should vanish... along
with their siblings in the armv7a.c file.
Remove a handful of superfluous checks ... e.g. the mode number
was just initialized, or (debug entry methods) already validated.
Move one of the macros using internal mode numbers into the only
file which uses that macro. Make the tables manipulated with
those numbers be read-only and, where possible, static so they're
not confused with part of the generic ARM interface.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Add and use arm_mode_name() to map from PSR bits to user
meaningful names. It uses a new table which, later, can
be used to hold other mode-coupled data.
Add definitions for the "Secure Monitor" mode, as seen on
some ARM11 cores (like ARM1176) and on Cortex-A8. The
previous mode name scheme didn't understand that mode.
Remove the old mechanism ... there were two copies, caused
by Cortex-A8 needing to add "Secure Monitor" mode support.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
All ARM cores need to provide obsolete FPA registers in their
GDB register dumps. (Even though cores with floating point
support now generally use some version of VFP...)
Clean up that support a bit by sharing the same dummy registers,
and removing the duplicate copies. Eventually we shouldn't need
to export those dummies.
(This makes the ARMv7-M support include the armv4_5 header, and
cleans up related #includes, but doesn't yet use anything from
there except those dummies.)
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
No need to indirect from registered integers to pointers.
Just stash the pointers directly in the register struct,
and don't even bother registering.
This is a small code shrink, speeds register access just
a smidgeon, and gets rid of another rude exit() path.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
And move the rest of the vector_catch stuff into the C file;
it's not part of the module interface.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Also, switch integrity check over to the correct magic number,
and remove duplicate v4/v5 #define.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Have ARM11 register the "standard" ARM commands. For now, only
disassembly really works.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Rename the "armv4_5" command prefix to straight "arm" so it makes
more sense for newer cores. Add a simple compatibility script.
Make sure all the commands give the same "not an ARM" diagnostic
message (and fail properly) when called against non-ARM targets.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Use size_t instead of uint32_t when specifying file sizes. Update all
consumers up through the layers to use size_t when required. These
changes should be safe, but the higher-levels will need to be updated
further to receive the intended benefits (i.e. large file support).
Add error checking for fileio_read and file_write. Previously, all
errors were being silently ignored, so this change might cause some
problems for some people in some cases. However, it gives us the chance
to handle any errors that do occur at higher-levels, rather than burying
our heads in the sand.
Tweak "standard" ARM disassembler diagnostics to fail if the target
is not "an ARM" (vs. not "an ARMV4/5"), so it makes more sense for
cores inheriting this as the "generic" disassembler.
Also, to use the Thumb2 entry instead of the original Thumb entry.
This makes it work better for both newer cores (which support those
added instructions) and for BL and BLX instructions on older cores.
(Those instructions are 32-bits, which requires curious state-aware
code to go through a 16-bit decode interface...)
Plus minor cleanups, notably to have fewer exit paths and to make
sure they all return failure codes.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
They're really too big to inline, at least for code that's
not in any performance-critical loops.
Also move the associated string table to the rodata section.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Don't include it in more headers than necessary; just
use it in the few files that actually need it.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The ARRAY_SIZE macro was defined in several target files, so move it
to types.h.
This patch also removes two other identical macros: DIM (from jtag.h)
and asizeof (from arm11.h).
Eliminate redundant check that gets covered by using unsigned type.
Created to eliminate noise from subsequent patches, but this kind of
conversion will be beneficial in similar ways throughout the tree.
The container_of macro is useful as a general solution. It belongs
in types.h, rather than target.h where it was introduced. Requires
the offsetof macro, which comes from <stddef.h> (moved as well).
Don't include "target.h" from more headers than necessary. This
avoids needless interdependencies and duplicated include paths.
Don't needlessly include it in source files, either.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Same deal: "register.h" got needlessly included all over the
place because of being in a few widely included headers.
So take it out of the header files which included it, and put
it in files which use it ... reduce needless interdependencies.
Also, don't need that extra "types.h" inclusion.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Lots of files still include it, often through needless
duplicate inclusion of "log.h"; sigh.
This cleans up the inclusion graph a bunch, so there are
fewer inclusion paths, but it doesn't change much otherwise.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Most files in the tree seem to have ended up including this,
and *quite* needlessly ... only code implementing or using
downloadable algorithms actually needs these declarations.
So take it out of the header files which included it, and put
it in files which use it ... reduce needless interdependencies.
Also: "algorithm.h" doesn't need to include "types.h" again;
it already comes from a different header.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Most files in the tree seem to have ended up including this,
and *quite* needlessly ... only code implementing or using
breakpoints actually needs these declarations.
So take it out of the header files which included it, and put
it in files which use it ... reduce needless interdependencies.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Now the ARM11 cores can use the renamed arm_checksum_memory()
and arm_blank_check_memory() routines ... do so.
Sanity checked with "flash erase_check" of both NOR banks on an
OMAP2420 ... the algorithm code dumped four lines of of "poll"
status after each of almost 520 blocks (yes, *very* annoying) but
gave plausible results after producing that spam.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Three changes: remove ARM11_HANDLER() in favor of normal structure
initialization syntax; fix goofy indentation in that structure; and
don't needlessly export arm11_register_commands(), it's only called
through that method table.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The arm7_9_checksum_memory() and arm7_9_blank_check_memory()
routines are not actually specific to the ARM7 and ARM9 core
generations ... they can work for any core which can run
algorithms using basic ARM (not Thumb) instructions.
Rename them; move the declarations to a more generic site;
likewise move the code (and tidy it a bit in the process).
NOTE: the blank_check() method falsely returned a success
status (0) on one error path, when the algorithm failed.
Fixed this bug.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Previously this flag was stored in "target_type", so that for example
if there were two ARM7TDMI targets in a scan chain, both would claim
to have been examined although only the first one actually had its
examine() method called.
Move this state to where it should have been in the first place, and
hide a method that didn't need exposure ... the flag is write-once.
Provide some doxygen. The examine() method is confusing, since it
isn't separating one-time setup from the after-each-reset stuff. And
the ARM7/ARM9 version is, somewhat undesirably, not leaving the debug
state alone after reset ... probably more of an issue for trace setup
than for watchpoints and breakpoints.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Removes unused arm_jtag_buf_* helpers from arm_jtag.[ch]. These could
reappear if patches are provided to conver the tree to use them, but
this code should not be in the master tree until that series is ready.
This fixes an issue due to the new command handler syntax caused by the mw handler playing with the args pointer before
using the CMD_NAME macro. Fix is to move this call above the lines changing args.
Changed some printf format strings..
[dbrownell@users.sourceforge.net: shrink lines, fix indents]
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Kick in ETM (and ETB) support for ARM11. Tested on OMAP 2420,
so update that configuration. (That's an ARM1136ejs, ETB,
OpenGL ES1.1, C55x DSP, etc.)
Also update the other ARM11 ETM + ETB targets in the tree
to set up these modules. (Not tested.)
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
These aren't desirable, given "standard" ETM support.
Also remove the now-unused arm11_find_target().
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
No point in having two identical examine methods for the
ARM7TDMI and ARM9TDMI drivers; move, rename, shrink, share.
Add a bit of doxygen; stop needlessly exporting a method.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Return NULL from etm_build_reg_cache() not ERROR_OK; and share
code on that fault path.
Let ETM code handle any tracking of its cache -- not callers.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
This will enable reusing many common ARM utilities, in
particular the ETM and ETB support. The ARM11 support
can still be much simplified after this patch, though.
Note: none of those common utilities kick in yet...
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
We have too many different registers, and too many version and
context dependent interpretations, for this type of bitfield
management to be scalable.
(Anyone who really wants bitfield interpretation *can* do that
in Tcl code...)
There are ... quite a few copies of the same ARM dummy registers.
There should eventually be one copy; this many is craziness.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Remove the last external user of arm7_9_get_arch_pointers(), and
that annoying downcast utility. Add an is_arm7_9() predicate.
Stop returning specious success codes on various failure paths
in the ARM7/ARM9 commands which used that downcast utility.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The number of command arguments will always be 0 or more, so use
the right type in handlers. This has a cascading effect up through
the layers, but the new COMMAND_HANDLER macros prevented total chaos.
By using CALL_COMMAND_HANDLER, parameters can be reordered, added, or
even removed in inherited signatures, without requiring revisiting
all of the various call sites.
The "remove (forward) declarations" patch goofed indentation on the
"cortexa8_target" struct; fix.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
ARM11 and newer cores include updated ETM modules. Recognize
their version codes and some key config differences. Sanity
checked on an OMAP2, with an ETM11RV r0p1 (ETMv3.1).
This still handles only scan chain 6, with at most 128 registers.
Newer cores (mostly, Cortex) will need to use the DAP instead.
Note that the newer ETM modules don't quite fit the quirky config
model of the older ones ... having more port widths is easy, but
the modes aren't the same. That still needs to change.
Fix a curious bug ... how did the register cache NOT get saved??
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Now that nothing uses the old ETM handle any more, remove it.
Add minimal header tweaks, letting non-ARM7 and non-ARM9 cores
access ETM facilities.
Now ARM11 could support standard ETM (and ETB) access as soon as
it derives from "struct arm" ... its scanchain 6 is used access
the ETM, just like ARM7 and ARM9.
The Cortex parts (both M3 and A8) will need modified access methods
(via ETM init parameters), so they use the DAP. Our first A8 target
(OMAP3) needs that for both ETM and ETB, but the M3 ETM isn't very
useful without SWO trace support (it's painfully stripped down), so
that support won't be worth adding for a while.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Make ETM itself use the new toplevel ETM handle, instead
of the to-be-removed lower level one. As of this patch,
nothing should be using the old ARM7/ARM9-specific handle.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Make both useful ETM port drivers (etb, etm_dummy) use the new
toplevel ETM handle, instead of the to-be-removed lower level one.
Do the same for the "oocd-trace" prototype too; and fix its
error reporting paths: return failure codes, don't exit(), etc
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Make ARM7 and ARM9 cores use the new toplevel ETM handle to
trigger ETM setup, not the to-be-removed lower level one.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Rename "struct armv4_5_common_s" as "struct arm". It needs
a bit more work to be properly generic, and to move out of
this header, but it's the best start we have on that today.
Add and initialize an optional ETM pointer, since that will
be the first thing that gets generalized.
The intent being: all ARMs should eventually derive from
this "struct arm", so they can reuse the current ETM logic.
(And later, more.) Currently the ARM cores that *don't* so
derive are only ARMv7-M (and thus Cortex-M3) and ARM11.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Add 'const' keyword to 'char *' parameters to allow command handlers to
pass constant string arguments. These changes allow the 'args' command
handler to be changed to 'const' in a subsequent patch.
Subsequent patches expect all command handlers to use a uniform
parameter naming scheme. In the entire tree, these two files used
standard 'argv' instead of our non-standard 'args'. This patch opts
to reduces the noise required to unify the command handlers, using
dominant 'args' form.
A future patch may be used to convert us back to the standard argv, but
that requires coordination with all developers to minimize disruptions.
Separates various groups of files to be built in logical succession.
In each layer, the core module (target.c, nand.c, etc.) is built _after_
their helper modules (e.g. image.c, nand_ecc.c) but _before_ any of
their drivers (e.g. arm966e.c, mx3_nand.c).
This allows problems introduced at the bottom of the stack to result
in build failures as soon as possible, as the helpers and core should
wrap portions of them.
Various cleanups of ETM related code.
- Saner error return paths
- Simplify arm7_9 init ... no need for extra zeroing!
- Shrink some lines
- Tweak some diagnostics
- Use shorter name for ETM struct type.
- Don't exit()
and similar. The diagnostics look forward to having
this ETM code work with more than just ARM7/ARM9.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The "ARM720 uses the new inheritance/nesting scheme" patch
wrongly scrubbed a calloc() from arm720t_target_create().
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Start switching MMU handling over to a more sensible scheme.
Having an mmu() method enables MMU-aware behaviors. Not having
one kicks in simpler ones, with no distinction between virtual
and physical addresses.
Currently only a handful of targets have methods to read/write
physical memory: just arm720, arm920, and arm926. They should
all initialize OK now, but the arm*20 parts don't do the "extra"
stuff arm926 does (which should arguably be target-generic).
Also simplify how target_init() loops over all targets by making
it be a normal "for" loop, instead of scattering its three parts
to the four winds.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
First cut of these commands. Øyvind tinkered a bit with
the number parsing to bring it up to speed + rebased it.
Ready for testing.
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
- improve some names -- a "default" prefix is not descriptive
- add doxygen @todo entries for some issues
- avr8 isn't ever going to need those MMU hooks
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
This patch introduced a bug preventing flash writes from working
on Cortex-M3 targets like the STM32. Moreover, it's the wrong
approach for handling no-MMU targets.
The right way to handle no-MMU targets is to provide accessors
for physical addresses, and use them everywhere; and any code
which tries to work with virtual-to-physical mappings should use
a identity mapping (which can be defaulted).
And ... we can tell if a target has an MMU by seeing if it's
got an mmu() method. No such methood means no MMU.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
It's been about a year since these were deprecated and, in most
cases, removed. There's no point in carrying that documentation,
or backwards compatibility for "jtag_device" and "jtag_speed",
around forever. (Or a few remnants of obsolete code...)
Removed a few obsolete uses of "jtag_speed":
- The Calao stuff hasn't worked since July 2008. (Those Atmel
targets need to work with a 32KHz core clock after reset until
board-specific init-reset code sets up the PLL and enables a
faster JTAg clock.)
- Parport speed controls don't actually work (tops out at about
1 MHz on typical HW).
- In general, speed controls need to live in board.cfg files (or
sometimes target.cfg files), not interface.cfg ...
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Add comments (Doxygen and normal), remove unused code,
shrink some overlong lines. Get rid of a forward decl.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
This patch changes the duration_* API in several ways. First, it
updates the API to use better names. Second, string formatting has
been removed from the API (with its associated malloc). Finally, a
new function added to convert the time into seconds, which can be
used (or formatted) by the caller. This eliminates hidden calls to
malloc that require associated calls to free().
This patch also removes the useless extern keyword from prototypes,
and it eliminates the duration_t typedef (use 'struct duration').
These API also allows proper error checking, as it is possible for
gettimeofday to fail in certain circumstances.
The consumers have all been chased to use this new API as well, as
there were relatively few cases doing this type of measurement.
In most cases, the code performs additional checks for errors, but
the calling code looks much cleaner in every case.
Reduces confusion about location of associated routines and
reduces clutter in the arm11 header.
Removes extra whitespace around the lines touched by these changes.
Make several functions be static. Shrink some of the overlong
lines. Use pure tab indents in some places that mixed in spaces.
This gives a minor object code shrink (about 2% on amd64).
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Resolve serious bug inserted by the "target: require working
area for physical/virtual addresses to be specified" patch.
It forced use of (invalid) virtual addresses when the MMU
was disabled, and vice versa.
Observed to break at least Cortex-M3, ARM926, ARM7TDMI whenever
work areas are used, such as during bulk writes to flash, DDR2,
SRAM, and so on.
Also, fix overlong lines and whitespace goofs.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Move various embedded target structs to the beginnings of
their containers ... pretty much the way C++ or Obj-C
would for single inheritance.
This shrinks code that accesses those embedded structs by
letting common offsets use smaller instructions. Sample
before/after sizes (on amd64):
17181 312 0 17493 4455 arm920t.o
16810 312 0 17122 42e2 arm920t.o
Where the "after" is the smaller number, with this patch
over the ones leveraging that embedding knowledge.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Remove most remaining uses of target->arch_info from ARM
infrastructure, where it hasn't already been updated.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Use target_to_armv7a() etc, replacing needless pointer traversals.
Stop using X->arch_info scheme in most ARMv7-A and Cortex-A8 code.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Use target_to_arm7_9(), replacing needless pointer traversals.
Also: remove now-useless contents of arm7tdmi struct; it's
almost ready to be removed.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Use target_to_arm720(), replacing needless pointer traversals
and simplifying a bunch of nasty code. Stop setting arch_info
for arm720 type parts, it's not used any longer.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Use target_to_xscale(), replacing needless pointer traversals
and simplifying a bunch of code.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Replace needless pointer traversals and simplify. Also remove most
remaining contents from arm9tdmi struct; it's almost removable.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Use target_to_arm926(), replacing needless pointer traversals
and simplifying a bunch of code.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Use target_to_arm920(), replacing needless pointer traversals
and simplifying. Stop setting arm9tdmi->arch_info for arm920
type parts, it's not used any longer.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Use new target_to_cm3() and target_to_armv7m() inlines,
instead of a series of x->arch_info conversions. Remove
arch_info, since nothing uses it.
Also fix an omission: the Cortex-M3 commands didn't verify
that they were operating on that kind of target. Add comment
about the ARMv7M version of that omission.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Provide a cleaner way to handle single inheritance of targets
in C, using the same model Linux does: structs containing other
structs, un-nested via calls to a "container_of()" macro that
are packaged in typesafe inline functions.
Targets already use this containment idiom, but make it much
more complicated because they un-nest using embedded "void *"
pointers ... in chains of up to five per target, which is all
pure needless complication. (Example: arm92x core, arm9tdmi,
arm7_9, armv4_5 ... on top of the base "target" class.)
Applying this scheme consistently simplifies things, and gets
rid of many error-prone untyped pointers. It won't change any
part of the type model though -- it just simplifies things.
(And facilitates more cleanup later on.)
Rule of thumb: where there's an X->arch_info void* pointer,
access to that pointer can and should be removed. It may be
convenient to set up pointers to some of the embedded structs;
and shrink their current "*_common" names (annoyingly long).
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The forward decls are just code clutter; remove them, by moving
their references after definitions. This is another file which
never needed even one internal forward declaration.
Also shrink a few overly-long lines with function declarations
or definitions; get rid of arm7tdmi_register_commands(), it's
not needed (just delegated); minor whitespace declutter.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Unneeded exports cause confusion about the module interfaces.
Make all functions static. Add a short header comment.
The forward decls are just code clutter; remove them, by moving
their references after definitions. This is another file which
never needed even one internal forward declaration.
Remove unneeded indirection for the write_memory() method. Make
a table static, remove a can't-happen case with nasty exit().
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Unneeded exports cause confusion about the module interfaces.
Make most functions static.
The forward decls are just code clutter; remove them, by moving
their references after definitions. This is another file which
never needed even one internal forward declaration.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The forward decls are just code clutter; remove them, by moving
their references after definitions. This is another file which
never needed even one internal forward declaration.
Also shrink a few overly-long lines with function declarations
or definitions.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Unneeded exports cause confusion about the module interfaces.
Make most functions static.
The forward decls are just code clutter; remove them, by moving
their references after definitions. This is another file which
never needed even one internal forward declaration.
Also remove needless arm966e_init_target(), in favor of the
arm9tdmi routine to which it delegates its work.
This saved over 100 bytes of code on x86_32.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
An init_target() wrapper isn't needed, and target_create()
can shrink a bit. Add a header comment and some doxygen.
Remove arm926ejs_catch_broken_irscan() which has been a NOP
for quite a few months now, and in any case duplicates logic
in the JTAG core to validate IR capture data. But force the
capture mask to 0x0f, so those tests are most effective.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Unneeded exports cause confusion about the module interfaces.
Make most functions static. Add a short header comment.
The forward decls are just code clutter; remove them, by moving
their references after definitions. This is another file which
never needed even one internal forward declaration.
This saved almost 900 bytes of code on x86_32; it seems the
compiler can leverage its knowledge that these functions are
not called from the outside world...
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The arm920t has a concept of read modify write cycles
that may have to be represented in the mrcmcr interface
eventually.
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
Fail watchpoint_add() if it's the same address but the
parameters are different ... don't just assume having
the same address means the same watchpoint! (Note that
overlapping watchpoints aren't detected...)
Handle unrecognized return codes more sanely; don't exit()!
And describe command params right.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Expose most DWT registers via Tcl; there are a few more, but
those are mostly for profiling along with the ITM. Having
this set available enables operations which aren't possible
with just the standard watchpoint operations.
The cycle counter may be interesting. Turn it on after reset
by setting the LSB of the dwt_ctrl register, and it counts
CPU clocks. You can program the comparator 0 watchpoint to
trigger on a given cycle count, rather than a data address.
Likewise, comparator 1 may be able to match data values given
address matches from one or two other comparators. (Not all
hardware supports this capability though; try it. That is
something the standard watchpoint methods should eventually
handle, for the single address case.)
Minor cleanup: remove needless functional indirection for
exposing the v7m architctural registers.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
There's no reason to read which interrupts are enabled from
the NVIC; that state isn't used. Plus, it's highly dynamic
since firmware can change it at any time; remove the support
for those state records.
Remove duplicate definition of DWT_CTRL address; shrink a line.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Fix the watchpoint error checks, and do them in add(), not later
in set() when it's mostly too late. Support the full range of
watchpoint sizes (1 to 32K bytes each), and check alignments.
Minor cleanup of DWT access: shrink lines, use "+" for address
calculations, comment a few issues. Add debug message reporting
DWT capabilities, matching the message for FBP, and some minor
code and spec review comments.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Add Doxygen for the exported ARMv7-M interfaces.
Make the non-exported stuff static. Remove functions and
data which are now observably unused.
Add comment about a small speedup that the run_algorithm()
logic could use. Shrink a few too-long lines.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
improve default target->read/write_phys_memory, produce
more sensible error messages if the mmu interface
functions have not been implemented yet vs. will
not be implemented(e.g. cortex m3).
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
The quit entry point was not being invoked. Just a source
of confusion at this point. XScale ran 100x reset upon
quit, but that code made no sense, wasn't commented
and never invoke.
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
To support breakpoints, flush data cache line and invalidate
instruction cache when 4 and 2 byte words are written.
The previous code was trying to write directly to the physical
memory, which was buggy and had a number of other situations
that were not handled.
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
Fixed bug: if virtual address for working memory was not specified
and MMU was enabled, then address 0 would be used.
Require working address to be specified for both MMU enabled
and disabled case.
For some completely inexplicable reason this fixes the regression
in svn 2646 for flash write in arm926ejs target. The logs showed
that MMU was disabled in the case below:
https://lists.berlios.de/pipermail/openocd-development/2009-November/011882.html
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
Just use the array of names we're given, ignoring indices.
The "reserved means don't use" patch missed that change.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
I'm suspecting this code can never have worked, since the
original commit (svn #335) in early 2008.
Fix is just copy/paste from another (working) function.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Unneeded exports cause confusion about the module interfaces.
Make most functions static, and fix some line-too-long issues.
Delete some now-obviously-unused code.
The forward decls are just code clutter; move their references
later, after the normal declarations. (Or vice versa.)
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Unneeded exports cause confusion about the module interfaces.
Only the Feroceon code builds on this, so only routines it
reuses should be public.. Make most remaining functions
static, and fix some of the line-too-long issues.
The forward decls are just code clutter; move their references
later, after the normal declarations. Turns out we don't need
even one forward declaration in this file.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The Hex parser uses a fixed number of sections. When the
number of sections in the file is greater than that, the
stack get corrupted and a CHECKSUM ERROR is detected
which is very confusing.
This checks the number of sections read, and increases
IMAGE_MAX_SECTIONS so it works on my file.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Only type 1 branch instruction has a condition code, not type 2.
Currently they're both tagged with ARM_B which doesn't allow for the
distinction.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
A Thumb BLX instruction is branching to ARM code, and therefore the
first 2 bits of the target address must be cleared.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Currently, OpenOCD is always caching the PC value without the T bit.
This means that assignment to the PC register must clear that bit and set
the processor state to Thumb when it is set. And when the PC register
value is transferred to another register or stored into memory then
the T bit must be restored.
Discussion: It is arguable if OpenOCd should have preserved the original
PC value which would have greatly simplified this code. The processor
state could then be obtained simply by getting at bit 0 of the PC. This
however would require special handling elsewhere instead since the T bit
is not always relevant (like when PC is used with ALU insns or as an index
with some addressing modes). It is unclear which way would be simpler in
the end.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Whenever an unconditional branch with the H bits set to 0b10 is met, the
offset must be combined with the offset from the following opcode and not
ignored like it is now.
A comment in evaluate_b_bl_blx_thumb() suggests that the Thumb2 decoder
would be a simpler solution. That might be true when single-stepping of
Thumb2 code is implemented. But for now this appears to be the simplest
solution to fix Thumb1 support.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Calling it first with every opcodes and then testing if the opcode
was indeed a branch instruction is wasteful and rather strange.
If ever thumb_pass_branch_condition() has side effects (say, like
printing a debugging traces) then the result would be garbage for most
Thumb instructions which have no condition code.
While at it, let's make the nearby code more readable by reducing some of
the redundant brace noise and reworking the error handling construct.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Make the "dap info" output more comprehensible:
- Don't show CIDs unless they're incorrect (only four bits matter)
- For CoreSight parts, interpret the part type
- Interpret the part number
- Show all five PID bytes together
- Other minor cleanups
Also some whitespace fixes, and shrink a few overlong source lines.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Remove needless debug handler state.
- "handler_installed" became wrong as soon as the second TRST+SRST
reset was issued ... so the handler was never reloaded after the
reset removed it from the mini-icache.
This fixes the bug where subsequent resets fail on PXA255 (if the
first one even worked, which is uncommon). Other XScale chips
would have problems too; PXA270 seems to have, IXP425 maybe not.
- "handler_running" was never tested; it's pointless.
Plus a related bugfix: invalidate OpenOCD's ARM register cache on reset.
It was no more valid than the XScale's mini-icache. (Though ... such
invalidations might be better done in "SRST asserted" callbacks.)
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Bit 5 shouldn't be used. Remove all support for modifying it.
Matches the exception vector table, of course ... more than one
bootloader uses that non-vector to help distinguish valid boot
images from random garbage in flash.
The wrong variable (pc instead of r0) was used. Furthermore, someone
did cover this error by stupidly silencing the compiler warning that
occurred before a dummy void reference to r0 was added to the code.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
When dumping over 100 registers (as on most ARM9 + ETM cores),
aid readability by splitting them into logical groups.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The register names are perversely not documented as zero-indexed,
so rename them to match that convention. Also switch to lowercase
suffixes and infix numbering, matching ETB and EmbeddedICE usage.
Update docs to be a bit more accurate, especially regarding what
the "trigger" event can cause; and to split the issues into a few
more paragraphs, for clarity.
Make "configure" helptext point out that "oocd_trace" is prototype
hardware, not anything "real".
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Passing "--std=gun99" is unfortunately not sufficient to make current
MinGW compilers conform with respect to checking printf format strings.
(The C runtime seems not to have problems.)
Fix by using a "gnu_printf" format specifier not "printf".
Generate a C struct with the data, and use that, instead of an
assembly language file. The assembly language causes issues on
Darwin and MS-Windows, which don't necessarily use GNU AS; or
if they do, don't necessarily use its ELF syntax.
It's also better in two other ways: fewer global symbols; and
the init-time size check gets optimized away at compile time.
(Unless it fails, in which case bigger chunks of the file vanish.)
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Based on some patches from <redirect.slash.nil@gmail.com>
for preliminary Win64 compilation. More such updates are
needed, but they need work. Compile tested on 64 and 32 bit
Linuxes, and Cygwin.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
The LE check is obviously buggy (as easily triggered during some
testing), but I didn't audit the rest of the cases.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Resolve a "FIX" comment; yes that was superfluous given that the
JTAG core does that check by default. It was also buggy since it
wrote to a stack frame that went away before the write happened!!
Other fixes: remove pointless malloc(); zero-init scan_field_t
values wherever they appear; whitespace scrub; spelling fix.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Load the XScale debug handler from the read-only data section
instead of from a separate file that can get lost or garbaged.
This eliminates installation and versioning issues, and also
speeds up reset handling a bit.
Plus some minor bits of cleanup related to loading that handler:
comments about just what this handler does, and check fault codes
while writing it into the mini-icache.
The only behavioral changes should be cleaner failure modes after
errors during handler loading, and being a bit faster.
NOTE: presumes GNU assembly syntax, with ".incbin"; and ELF,
because of the syntax of the ".size" directive.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Streamline/shrink some needless JTAG stuff:
- Use #defines for the JTAG instructions; they can't ever change
- Remove an unused (!) shadow of tap->ir_length
- Stop using a copy of target->tap
- Don't bother saving the variant after sanity checking ir_length
Also, make target_create() work as on other targets: build the
register cache later, making init_target() no longer be a NOP.
Handle malloc failure; remove a comment that was obsoleted by the
not-so-new target syntax.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Remove unused and deprecated (in the arch spec) mode for loading
code into the *main* icache (vs the "mini" icache). Disable some
extremely noisy (and rarely useful) low-level debug messages
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Declare almost everything as static.
Move stuff to remove most forward references.
Remove most forward declarations.
Warn if the unimplemented register functions get called.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Just fill out the rest of the cache line with NOPs; don't change
the record of how much data we consumed. Otherwise the count of
how much data is left can roll over from positive to negative
("VERY positive") and skip the loop termination of zero.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Add a header comment referencing useful XScale specs.
Make most data static, and the tables readonly.
Scrub extra blank lines.
Return fault codes from one routine.
Remove a needless NOP methood.
(BUGFIX) When we update R0, mark R0 as dirty/valid ... not R15/PC!
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Observed:
openocd: core.c:318: jtag_checks: Assertion `jtag_trst == 0' failed.
The issue was that nothing disabled background polling during calls
from the TCL shell to "jtag_reset 1 1". Fix by moving the existing
poll-disable mechanism to the JTAG layer where it belongs, and then
augmenting it to always pay attention to TRST and SRST.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
- revert to previous default: don't talk JTAG during SRST
- add "srst_nogates" flag, the converse of "srst_gates_jtag"
- with no args, display the current configuration
And update the User's Guide text with bullet lists to be a bit more clear.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2818 b42882b7-edfa-0310-969c-e2dbd0fdcd60
- don't needlessly export this function
- handle "case 0" debug method-of-entry better (silent by default)
The "case 0" is a valid debug entry mode so it doesn't deserve the
warning int now gets. But it probably means that OpenOCD confused
itself somehow; or that it confused the ARM9EJS target.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2799 b42882b7-edfa-0310-969c-e2dbd0fdcd60
- ETB
* report _actual_ hardware status, not just expected status
* add a missing diagnostic on a potential ETB setup error
* prefix any diagnostics with "ETB"
- ETM
* make "etm status" show ETM hardware status too, instead of
just traceport status (which previously was fake, sigh)
- Docs
* flesh out "etm tracemode" docs a bit
* clarify "etm status" ... previously it was traceport status
* explain "etm trigger_percent" as a *traceport* option
ETM+ETB tracing still isn't behaving, but now I can see that part of
the reason is that the ETB turns itself off almost immediately after
being enabled, and before collecting any data.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2790 b42882b7-edfa-0310-969c-e2dbd0fdcd60
- Commands were supposed to have been "arm11 memwrite ..."
not "memwrite ..."
- Get rid of obfuscatory macros
- Re-alphabetize
- Add docs for "arm11 vcr"
git-svn-id: svn://svn.berlios.de/openocd/trunk@2776 b42882b7-edfa-0310-969c-e2dbd0fdcd60
only expose the registers which are actually present. They
could be missing for two basic reasons:
- This version might not support them at all; e.g. ETMv1.1
doesn't have some control/status registers. (My sample of
ARM9 boards shows all with ETMv1.3 support, FWIW.)
- The configuration on this chip may not populate as many
registers as possible; e.g. only two data value comparators
instead of eight.
Includes a bugfix in the "etm info" command: only one of the
two registers is missing on older silicon, so show the first
one before bailing.
Update ETM usage docs to explain that those registers need to be
written to configure what is traced, and that some ETM configs
are not yet handled. Also, give some examples of the kinds of
constrained trace which could be arranged.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2752 b42882b7-edfa-0310-969c-e2dbd0fdcd60
system, removes 20 non-existent registers ... but still includes
over 45 (!) ETM registers which don't even exist there ...
- Integrate the various tables to get one struct per register
- Get rid of needless per-register dynamic allocation
- Double check list of registers:
* Remove sixteen (!) non-registers for data comparators
* Remove four registers that imply newer ETM than we support
* Change some names to match current architecture specs
- Handle more register info
* some are write-only
* some are read-only
* record which versions have them, just in case
- Reorganize the registers to facilitate removing the extras
* group e.g. comparator/counter #N registers together
* add and use lookup-by-ID
git-svn-id: svn://svn.berlios.de/openocd/trunk@2751 b42882b7-edfa-0310-969c-e2dbd0fdcd60
- Add a header comment
- Line up the ETM context struct, pack it a bit
- Remove unused context_id (this doesn't support ETMv2 yet)
- Make most functions static
- Remove unused string table and other needless lines of code
- Correct "tracemode" helptext
Also provide and use an etm_reg_lookup() to find entries in the ETM
register cache. This will help cope with corrected contents of that
cache, which doesn't include entires for non-existent registers.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2750 b42882b7-edfa-0310-969c-e2dbd0fdcd60
- Shrink messaging during resets, primarily by getting rid of
"nothing happened" noise that hides *useful* information.
- Improve: the "no IDCODE" message by identifying which tap only
supports BYPASS; and the TAP event strings.
Related minor code updates:
- Remove two needless tests when examining the chain: we know
we have a TAP, and that all TAPs have names.
- Clean up two loops, turning "while"s into "for"s which better
show what's actually being done.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2736 b42882b7-edfa-0310-969c-e2dbd0fdcd60
and Tcl/external):
- Reorder so *both* paths (TCK/TMS or TRST) can enable TAPs with
ICEpick ... first C code flags TAPs that got disabled, then call
any Tcl code that might want to re-enable them.
- Always call the C/internal handlers when JTAG operations can be
issued; previously that wasn't done when TRST was used.
Plus some small cleanups (whitespace, strings, better messaging
during debug and on some errors) to reset-related code.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2730 b42882b7-edfa-0310-969c-e2dbd0fdcd60
- update comments to say so.
- update docs to clarify that the "arm9tdmi" command prefix
is a misnomer.
- bugfix some messages that wrongly assume only ARM9TDMI
based processors use this code.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2719 b42882b7-edfa-0310-969c-e2dbd0fdcd60
Cleanup some the downloaded ARM target algorithm code:
- Provide more complete disassembly of the DCC bulk write code
- Make code blocks "static const", in case GCC doesn't
- Fix some tabbing/layout issues
- Make some arm7_9_common.h flags be "bool" not "int"; and compact
the layout a bit (group most bools together)
git-svn-id: svn://svn.berlios.de/openocd/trunk@2698 b42882b7-edfa-0310-969c-e2dbd0fdcd60
Optionally shave time off the armv4_5 run_algorithm() code: let
them terminate using software breakpoints, avoiding roundtrips
to manage hardware ones.
Enable this by using BKPT to terminate execution instead of "branch
to here" loops. Then pass zero as the exit address, except when
running on an ARMv4 core. ARM7TDMI, ARM9TDMI, and derived cores
now set a flag saying they're ARMv4.
Use that mechanism in arm_nandwrite(), for about 3% speedup on a
DaVinci ARM926 core; not huge, but it helps. Some other algorithms
could use this too (mostly flavors of flash operation).
git-svn-id: svn://svn.berlios.de/openocd/trunk@2680 b42882b7-edfa-0310-969c-e2dbd0fdcd60
Provide an "armv7a disassemble" command. Current omissions include
VFP (except as coprocessor instructions), Neon, and various Thumb2
opcodes that are not available in ARMv7-M processors.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2676 b42882b7-edfa-0310-969c-e2dbd0fdcd60
lean up some loose ends with the ARM disassembler
- Add a header comment describing its current state and uses
and referencing the now-generally-available V7 arch spec
- Support some mode switch instructions:
* Thumb to Jazelle (BXJ)
* Thumb to ThumbEE (ENTERX)
* ThumbEE to Thumb (LEAVEX)
- Improve that recent warning fix (and associated whitespace goof)
- Declare the rest of the internal code and data "static". A
compiler may use this, and it helps clarify the scope of these
routines (e.g. what changes to them could affect).
git-svn-id: svn://svn.berlios.de/openocd/trunk@2675 b42882b7-edfa-0310-969c-e2dbd0fdcd60
By enabling this bit, the processor halts when a debug event
such as breakpoint occurs.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2668 b42882b7-edfa-0310-969c-e2dbd0fdcd60
that were added after ARMv5TE was defined:
- ARMv5J "BXJ" (for Java/Jazelle)
- ARMv6 "media" instructions (for OMAP2420, i.MX31, etc)
Compile-tested. This might not set up the simulator right for the
ARMv6 single step support; only BXJ branches though, and docs to
support Jazelle branching are non-public (still, sigh).
ARMv6 instructions known to be mis-handled by this disassembler
include: UMAAL, LDREX, STREX, CPS, SETEND, RFE, SRS, MCRR2, MRRC2
git-svn-id: svn://svn.berlios.de/openocd/trunk@2644 b42882b7-edfa-0310-969c-e2dbd0fdcd60
With DCCR we are asking the CPU to halt, we should wait until
the CPU has halted before proceeding with the operation.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2638 b42882b7-edfa-0310-969c-e2dbd0fdcd60
the ITR register but it will only be executed when the DSCR[13]
bit is set. The documentation is a bit weird as it classifies
the DSCR as read-only but the pseudo code is writing to it as
well. This is working on a beagleboard.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2634 b42882b7-edfa-0310-969c-e2dbd0fdcd60
instruction to be finished. This comes from the pseudo code
of the cortex a8 trm.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2632 b42882b7-edfa-0310-969c-e2dbd0fdcd60
For ARMv4/ARMv5:
- better command parameter error checking
- don't require an instruction count; default to one
- recognize thumb function addresses
- make function static
- shorten some too-long lines
For Cortex-M3:
- don't require an instruction count; default to one
With the relevant doc updates.
---
Nyet done: invoke the thumb2 disassembler on v4/v5,
to better handle branch instructions.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2624 b42882b7-edfa-0310-969c-e2dbd0fdcd60
reset operations. Maybe they can't; or it's a "not yet" thing.
Note that the assert/deassert operations can't yet trigger for
OMAP3 because resets currently include JTAG reset in all cases,
resetting the ICEpick and thus disabling the TAP for Cortex-A8.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2620 b42882b7-edfa-0310-969c-e2dbd0fdcd60
the values that are written in the mini-IC (plus documentation updates that
describe why this is needed).
git-svn-id: svn://svn.berlios.de/openocd/trunk@2613 b42882b7-edfa-0310-969c-e2dbd0fdcd60
nonfunctional cortex_a8 code with something that at least basically
works (for halt/step/resume, without MMU) even if it is incomplete.
(With tweaks from Øyvind, and cleanup from Dave.)
This code has mainly been developed and tested against R1606, it has
been built and tested against R2294 where it runs but step and resume
commands are broken due to regression (which should be fixed now).
This code is really written for OMAP3530. It doesn't identify debug
resources using generic DAP calls to scan the ROM table, or perform
topology detection. The OMAP3530 DAP exposes two memory access ports:
- Port #0 is connected to L3 interconnect (the main bus) with
passthrough to the L4 EMU bus ... so it will be used for most
memory accesses.
- Port #1 is connected to a dedicated debug bus (L4 EMU), with
access to L4 Wakeup, and holds the ROM table ... so it must
be used for most debug and control operations.
The are some defines to handle this in cortex_a8.c, which should be
replaced with more general code. Having access to another Cortex-A8
implementation would help get that right.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2609 b42882b7-edfa-0310-969c-e2dbd0fdcd60
and seed it with DAP access support using the current ADIv5 code.
(With tweaks and cleanup from Øyvind and Dave.)
The ARMv7-AR architecture manual is not publicly available (even
in subset form like the ARMv7-M spec), so it's hard to distinguish
between the Cortex-A8 implementation and the ARMv7-A architecture.
The register set presumably is architectural, and so it's stored
here; it's like earlier ARMs, with small additions. Ditto the
instruction set, though Thumb2 support is used (extending Thumb
support from ARMv6 with more 32-bit instructions) and there's this
ThumbEE thing too. There is a new "debug monitor" mode, not yet
fully addressed here, to support debugging in environments (like
motor control) where halting debug mode is inadvisable.
git-svn-id: svn://svn.berlios.de/openocd/trunk@2608 b42882b7-edfa-0310-969c-e2dbd0fdcd60
Zachary T Welch
David Brownell
Jerry Ling
Øyvind Harboe
Krzysztof Dziuba
Dean Glazeski
Ferdinand Postema
Michael Bruck
Franck HÉRÉSON
Nicolas Pitre
Redirect 'Slash' NIL
David Brownell
Lennert Buytenhek
Yauheni Kaliuta
dbrownell
oharboe
mlu
ntfreak
duane