]> git.sur5r.net Git - openocd/log
openocd
15 years agoarm_adi,armv7[am]: use COMMAND_HELPER for helpers
Zachary T Welch [Tue, 10 Nov 2009 13:32:04 +0000 (05:32 -0800)]
arm_adi,armv7[am]: use COMMAND_HELPER for helpers

Rewrites the dap_* command helpers to use the COMMAND_HELPER paradigm.
Uses CALL_COMMAND_HELPER to hide inherited calling conventions.

15 years agouse COMMAND_HELPER for command helper functions
Zachary T Welch [Tue, 10 Nov 2009 11:12:21 +0000 (03:12 -0800)]
use COMMAND_HELPER for command helper functions

Define the numerous helpers that inherit command handler parameters
using the COMMAND_HELPER macro.

15 years agouse COMMAND_HANDLER macro to define all commands
Zachary T Welch [Tue, 10 Nov 2009 07:56:52 +0000 (23:56 -0800)]
use COMMAND_HANDLER macro to define all commands

15 years agoadd command_handler_t type
Zachary T Welch [Wed, 11 Nov 2009 03:00:01 +0000 (19:00 -0800)]
add command_handler_t type

This patch adds new typedefs for command handler callback functions.
Users of this type signature were updated to use these new types.
It uses the new __COMMAND_HANDLER macro to prevent duplication.

15 years agoadd COMMAND_HANDLER and COMMAND_HELPER macros
Zachary T Welch [Wed, 11 Nov 2009 02:51:32 +0000 (18:51 -0800)]
add COMMAND_HANDLER and COMMAND_HELPER macros

The COMMAND_HANDLER and COMMAND_HELPER macros allow commands to be
defined in a manner that decouples them from the exact order and type of
their parameters.  Once converted, incremental changes to the command
handler type can be addressed in incremental patches that do not need to
touch the entire tree.

These macros' implementation, __COMMAND_HANDLER, is used to define the
new command_handler_t type, and additional patches will use it to derive
new macros to define extended command types (e.g. flash, nand, pld).
The CALL_COMMAND_HANDLER provides a means of calling helpers or nested
handlers from withing a command handler.

This patch uses C99 varadic macro expansion.  Please report compilers
that cannot handle this code.

15 years agoadd command_output_handler_t
Zachary T Welch [Wed, 11 Nov 2009 02:42:45 +0000 (18:42 -0800)]
add command_output_handler_t

Add a typedef for command output handler function type, simplifying
the appearance of functions that use it and eliminating duplicate code.

15 years agoCortex-A8: fix indent
David Brownell [Fri, 13 Nov 2009 16:45:20 +0000 (08:45 -0800)]
Cortex-A8: fix indent

The "remove (forward) declarations" patch goofed indentation on the
"cortexa8_target" struct; fix.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agonand: rename device to nand
Zachary T Welch [Fri, 13 Nov 2009 05:19:41 +0000 (21:19 -0800)]
nand: rename device to nand

To be more informative (and consistent with flash and pld trees), change
'device' parameter name to 'nand' in NAND source files.  This change
eliminates confusing 'device->device->' instance from the code, and
it simplifies the forthcoming command handler patches.

15 years agoETM: start support for ETMv2+
David Brownell [Fri, 13 Nov 2009 04:24:41 +0000 (20:24 -0800)]
ETM: start support for ETMv2+

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>
15 years agoWinXP-x64: find right D2XX libraries
Liam Redmond [Thu, 12 Nov 2009 20:50:41 +0000 (12:50 -0800)]
WinXP-x64: find right D2XX libraries

This trivial patch allows the config script to find the correct
FTDI2xx libraries under Windows XP x64 using msys.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoparport: add support for the jtag_khz command.
Jonas Horberg [Thu, 12 Nov 2009 20:39:37 +0000 (12:39 -0800)]
parport: add support for the jtag_khz command.

Add the khz and speed_div functions to the parport interface driver.
Add the parport_toggling_time function that tells the parport driver
how long (in nanoseconds) it takes for the hardware to toggle TCK.

[dbrownell@users.sourceforge.net: tweak doc for clarity, mention
multimeter, and whitespace fixes]

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoETM: remove old mid-level ETM handle
David Brownell [Thu, 12 Nov 2009 05:57:44 +0000 (21:57 -0800)]
ETM: remove old mid-level ETM handle

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>
15 years agoETM: use new toplevel ETM handle
David Brownell [Thu, 12 Nov 2009 05:55:19 +0000 (21:55 -0800)]
ETM: use new toplevel ETM handle

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>
15 years agoETM: update port drivers
David Brownell [Thu, 12 Nov 2009 05:52:02 +0000 (21:52 -0800)]
ETM: update port drivers

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>
15 years agoETM: update arm[79]tdmi_examine()
David Brownell [Thu, 12 Nov 2009 05:50:10 +0000 (21:50 -0800)]
ETM: update arm[79]tdmi_examine()

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>
15 years agoARM: start generalized base type
David Brownell [Thu, 12 Nov 2009 05:49:14 +0000 (21:49 -0800)]
ARM: start generalized base type

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>
15 years agofix 'jtag interface' behavior
Zachary T Welch [Wed, 11 Nov 2009 07:10:26 +0000 (23:10 -0800)]
fix 'jtag interface' behavior

Without this patch, running "openocd -c 'jtag interface'" segfaults.
Now, it returns the string "undefined" when the interface is unset.

15 years agoimprove command registration
Zachary T Welch [Wed, 11 Nov 2009 18:20:55 +0000 (10:20 -0800)]
improve command registration

Eliminate duplicate code for linking commands into a list.

Adds a check to ensure the command does not already exist;
if it does, return that one instead of creating a duplicate.

15 years agoadd help regardless of callback
Zachary T Welch [Wed, 11 Nov 2009 17:40:28 +0000 (09:40 -0800)]
add help regardless of callback

Add help for commands regardless of whether a handler is involved.
With this, all sorts of new commands can be found in 'help' text.
Hopefully, all of them have been documented....

Sadly, the lsort function appears to handle nested lists poorly, such
that sub-commands do not group with their parents.

15 years agoeliminate duplicate helptext management
Zachary T Welch [Wed, 11 Nov 2009 16:20:09 +0000 (08:20 -0800)]
eliminate duplicate helptext management

Add helpers to manage adding entries to the helptext list.
Adds support for arbitrarily nested commands.

15 years agoadd command_name helper
Zachary T Welch [Wed, 11 Nov 2009 13:26:17 +0000 (05:26 -0800)]
add command_name helper

The command_name function returns a malloced string for a given
command and its parents.  This can be used to display a message
to the user, but it is used internally to handle registration
and syntax errors.  This helps permit arbitrary command nesting.

15 years agowrap help command
Zachary T Welch [Wed, 11 Nov 2009 15:19:13 +0000 (07:19 -0800)]
wrap help command

Use continuation characters to wrap the help command into 80 columns.

15 years agoremove obsolete doxygen comments
Zachary T Welch [Wed, 11 Nov 2009 11:19:29 +0000 (03:19 -0800)]
remove obsolete doxygen comments

Documenting command handler parameters is redundant and pointless.

15 years agoremove more useless declarations
Zachary T Welch [Tue, 10 Nov 2009 07:20:34 +0000 (23:20 -0800)]
remove more useless declarations

Removes forward declarations by moving command registration after
defintion of the command handlers.

15 years agocortex_a8: remove declarations, use static keyword
Zachary T Welch [Wed, 11 Nov 2009 02:26:13 +0000 (18:26 -0800)]
cortex_a8: remove declarations, use static keyword

15 years agoadd const keyword to some APIs
Zachary T Welch [Tue, 10 Nov 2009 12:27:15 +0000 (04:27 -0800)]
add const keyword to some APIs

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.

15 years agochange argv to args in command handlers
Zachary T Welch [Tue, 10 Nov 2009 10:43:11 +0000 (02:43 -0800)]
change argv to args in command handlers

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.

15 years agoscript_debug(): improve types
Zachary T Welch [Mon, 9 Nov 2009 14:07:21 +0000 (06:07 -0800)]
script_debug(): improve types

Use unsigned type for number of arguments.

15 years agocommand.c: make private routines static
Zachary T Welch [Mon, 9 Nov 2009 14:09:30 +0000 (06:09 -0800)]
command.c: make private routines static

This patch also improves the signature of run_command function.

15 years agolog: improve log_callback_fn signature
Zachary T Welch [Mon, 9 Nov 2009 14:12:02 +0000 (06:12 -0800)]
log: improve log_callback_fn signature

Use unsigned type for line number in log_callback_fn signature.

15 years agotime_support: improve use of types
Zachary T Welch [Mon, 9 Nov 2009 13:45:57 +0000 (05:45 -0800)]
time_support: improve use of types

Update timeval_add_time to use long int; implement timeval_add with it.
Update timeval_ms to check gettimeofday return value, return int64_t.

15 years agomakefiles: improve build order
Zachary T Welch [Tue, 10 Nov 2009 12:27:02 +0000 (04:27 -0800)]
makefiles: improve build order

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.

15 years agoETM cleanup
David Brownell [Wed, 11 Nov 2009 12:42:50 +0000 (04:42 -0800)]
ETM cleanup

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>
15 years agofix bug in ARM720: bugfix
Zachary T Welch [Wed, 11 Nov 2009 05:02:30 +0000 (21:02 -0800)]
fix bug in ARM720: bugfix

15 years agoARM720: bugfix
David Brownell [Wed, 11 Nov 2009 03:06:10 +0000 (19:06 -0800)]
ARM720: bugfix

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>
15 years agotarget: MMU-aware init for memory read/write
David Brownell [Tue, 10 Nov 2009 19:58:31 +0000 (11:58 -0800)]
target: MMU-aware init for memory read/write

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>
15 years agotarget.cfg: (re)move some bogus reset_config lines
David Brownell [Tue, 10 Nov 2009 17:25:26 +0000 (09:25 -0800)]
target.cfg: (re)move some bogus reset_config lines

General rule, this is all board-specific and doesn't belong
in target config files.  Some of these were just cosmetic.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agostm32.cfg: remove reset_config
Thomas Kindler [Tue, 10 Nov 2009 17:16:12 +0000 (09:16 -0800)]
stm32.cfg: remove reset_config

Here's a patch for the double-reset problem on STM32.  I've tested
downloading and debugging with GDB and Eclipse, and everything seems
to work fine.

This effectively sets reset_config to none. trst_only would also
be ok, but that's better left to a board configuration file since
not all boards wire it up.

The NVIC is used to trigger reset, which at least on this chip also
pulses nSRST so the whole system does get rest -- exactly once.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoarm11: add etmr/etmw registers to access ETM via DBGTAP scan chain
Michael Bruck [Tue, 27 Oct 2009 21:41:00 +0000 (22:41 +0100)]
arm11: add etmr/etmw registers to access ETM via DBGTAP scan chain

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>
15 years agoARM11: remove old mrc/mcr commands
Øyvind Harboe [Mon, 26 Oct 2009 13:39:32 +0000 (14:39 +0100)]
ARM11: remove old mrc/mcr commands

Switch to new commands in config scripts

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
15 years agotelo.cfg: fix search paths
Øyvind Harboe [Tue, 10 Nov 2009 12:09:15 +0000 (13:09 +0100)]
telo.cfg: fix search paths

Add the missing "target/" prefix for scripts in the
target folder.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
15 years agoTarget: minor cleanup
David Brownell [Tue, 10 Nov 2009 10:01:20 +0000 (02:01 -0800)]
Target: minor cleanup

 - 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>
15 years agoARM920: implement basic MMU ops
David Brownell [Tue, 10 Nov 2009 09:36:59 +0000 (01:36 -0800)]
ARM920: implement basic MMU ops

mmu() works; virt2phys() fails and logs an error.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoARM720: implement basic MMU ops
David Brownell [Tue, 10 Nov 2009 09:35:50 +0000 (01:35 -0800)]
ARM720: implement basic MMU ops

mmu() works; virt2phys() fails and logs an error.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoTarget: fix bad error messages
David Brownell [Tue, 10 Nov 2009 09:33:59 +0000 (01:33 -0800)]
Target: fix bad error messages

And shrink a few too-long lines.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agohttpd: fix warnings, more robust error handling, improved MIME handling
Øyvind Harboe [Sun, 8 Nov 2009 18:50:35 +0000 (19:50 +0100)]
httpd: fix warnings, more robust error handling, improved MIME handling

The httpd is work in progress...

No mime type set by default. Let the browser guess.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
15 years agotcl: HostOs now picks up eCos as well during compile time
Øyvind Harboe [Tue, 10 Nov 2009 07:36:19 +0000 (08:36 +0100)]
tcl: HostOs now picks up eCos as well during compile time

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
15 years agosvf,xsvf,arm7_9_common: trim forard declarations
Zachary T Welch [Tue, 10 Nov 2009 05:24:45 +0000 (21:24 -0800)]
svf,xsvf,arm7_9_common: trim forard declarations

Remove forward declarations by reordering command registration.

15 years agocommand.c: make commands static
Zachary T Welch [Tue, 10 Nov 2009 05:23:38 +0000 (21:23 -0800)]
command.c: make commands static

Removes useless declarations, moving the handler functions to appear
before their use in the (much bigger) command registriation function.

15 years agoserver: remove useless declarations
Zachary T Welch [Tue, 10 Nov 2009 05:22:48 +0000 (21:22 -0800)]
server: remove useless declarations

Remove server command declarations, make handler routines static.

15 years agojtag: remove useless declarations
Zachary T Welch [Tue, 10 Nov 2009 05:21:06 +0000 (21:21 -0800)]
jtag: remove useless declarations

Contrary to my previous assessment, some opportunities to remove forward
declarations were overlooked.  Remove them by moving the definitions
of the command registration and interface structure to the end of files.

15 years agotarget.c: remove useless declarations
Zachary T Welch [Tue, 10 Nov 2009 03:24:06 +0000 (19:24 -0800)]
target.c: remove useless declarations

This patch removes the last batch of forward references from the tree,
moving the target command registration routines to the end of the file.

15 years ago{pic32m,stm32}x.c: remove useless declarations
Zachary T Welch [Mon, 9 Nov 2009 17:20:34 +0000 (09:20 -0800)]
{pic32m,stm32}x.c: remove useless declarations

Remove useless forward declarations.
Moves command registrations to end of files.
Moves flash structure definitions to end of files.

Signed-off-by: Zachary T Welch <zw@superlucidity.net>
15 years agostr{7,9}x*.c: remove useless forward declarations
Zachary T Welch [Mon, 9 Nov 2009 17:20:34 +0000 (09:20 -0800)]
str{7,9}x*.c: remove useless forward declarations

Remove useless forward declarations.
Moves command registrations to end of files.
Moves flash structure definitions to end of files.

Signed-off-by: Zachary T Welch <zw@superlucidity.net>
15 years agoflash/<various>.c: remove useless declarations
Zachary T Welch [Mon, 9 Nov 2009 17:20:33 +0000 (09:20 -0800)]
flash/<various>.c: remove useless declarations

Remove useless forward declarations.
Moves flash structure definitions to end of files.

Signed-off-by: Zachary T Welch <zw@superlucidity.net>
15 years agoflash/<assorted>.c: remove useless declarations
Zachary T Welch [Mon, 9 Nov 2009 17:20:33 +0000 (09:20 -0800)]
flash/<assorted>.c: remove useless declarations

Remove useless forward declarations.
Moves command registrations to end of files.
Moves flash structure definitions to end of files.

Signed-off-by: Zachary T Welch <zw@superlucidity.net>
15 years agoflash/at91sam[37].c: remove useless declarations
Zachary T Welch [Mon, 9 Nov 2009 17:20:33 +0000 (09:20 -0800)]
flash/at91sam[37].c: remove useless declarations

Remove useless forward declarations.
Moves command registration to end of file.
Moves flash structure definitions to end of files.

Changes a few references to global flash structure to local refs.

Signed-off-by: Zachary T Welch <zw@superlucidity.net>
15 years agoflash/lpc2???.c: remove useless declarations
Zachary T Welch [Mon, 9 Nov 2009 17:20:33 +0000 (09:20 -0800)]
flash/lpc2???.c: remove useless declarations

Remove useless forward declarations.
Moves command registrations to end of files.
Moves flash structure definitions to end of files.

Signed-off-by: Zachary T Welch <zw@superlucidity.net>
15 years agoflash/*nand*.c: remove useless declarations
Zachary T Welch [Mon, 9 Nov 2009 17:20:33 +0000 (09:20 -0800)]
flash/*nand*.c: remove useless declarations

Remove useless forward declarations.
Moves command registration to end of files.
Moves flash structure definition to end of files.

Signed-off-by: Zachary T Welch <zw@superlucidity.net>
15 years agoflash/flash.c: remove forward declarations
Zachary T Welch [Mon, 9 Nov 2009 17:20:33 +0000 (09:20 -0800)]
flash/flash.c: remove forward declarations

Remove useless forward declarations.
Moves command registration to end of file.
Moves flash structure definition to end of file.

Signed-off-by: Zachary T Welch <zw@superlucidity.net>
15 years agoRevert "target: add target->type->has_mmu fn"
David Brownell [Mon, 9 Nov 2009 22:46:23 +0000 (14:46 -0800)]
Revert "target: add target->type->has_mmu fn"

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>
15 years agofinish removing deprecated/obsolete commands
David Brownell [Mon, 9 Nov 2009 21:16:32 +0000 (13:16 -0800)]
finish removing deprecated/obsolete commands

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>
15 years agoUser's Guide: Flash/NAND doc tweaks
David Brownell [Mon, 9 Nov 2009 20:02:23 +0000 (12:02 -0800)]
User's Guide: Flash/NAND doc tweaks

Rename the "Drivers, Options, and Commands" sections to be
just "Driver List" matching the earlier reference.  Add an
example of parallel CFI flash.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agosrc/target: remove 'extern' and wrap headers
Zachary T Welch [Mon, 9 Nov 2009 12:22:23 +0000 (04:22 -0800)]
src/target: remove 'extern' and wrap headers

Remove extern keywords from function prototypes and wrap long lines.

15 years agotarget.h: remove extern keyword and wrap
Zachary T Welch [Mon, 9 Nov 2009 12:21:36 +0000 (04:21 -0800)]
target.h: remove extern keyword and wrap

Removes 'extern' keyword from function prototypes and wraps long lines.

15 years agosrc/{server,pld,svf,xsvf}: remove 'extern' keyword
Zachary T Welch [Mon, 9 Nov 2009 11:38:34 +0000 (03:38 -0800)]
src/{server,pld,svf,xsvf}: remove 'extern' keyword

Removes 'extern' keyword from function declarations in header filess.

15 years agosrc/flash: remove 'extern' and wrap headers
Zachary T Welch [Mon, 9 Nov 2009 11:10:11 +0000 (03:10 -0800)]
src/flash: remove 'extern' and wrap headers

Removes 'extern' keywords from function prototypes in the flash headers.
Wraps long lines to fit into 80 columns.

Adds multiple inclusion protection for s3c2xx_nand.h.

15 years agosrc/jtag: remove 'extern' and wrap headers.
Zachary T Welch [Mon, 9 Nov 2009 10:53:22 +0000 (02:53 -0800)]
src/jtag: remove 'extern' and wrap headers.

Removes the 'extern' keyword from function declarations.
Wraps long prototypes to fit into 80 columns.

Fixes documentation for jtag_tap_s::{,has}idcode fields.

15 years agosrc/helper: wrap and clean headers.
Zachary T Welch [Mon, 9 Nov 2009 10:47:00 +0000 (02:47 -0800)]
src/helper: wrap and clean headers.

Remove all useless 'extern' keywords from function prototypes.
Wraps long lines for readability.

15 years agoUser's Guide: bugfix global state info
David Brownell [Mon, 9 Nov 2009 17:40:55 +0000 (09:40 -0800)]
User's Guide: bugfix global state info

The "$ocd_HOSTOS" variable was wrongly documented.  Fix its
documentation, and its value on Linux.

Shrink a few of the too-long lines.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoEmbeddedICE: minor cleanups
David Brownell [Mon, 9 Nov 2009 17:34:51 +0000 (09:34 -0800)]
EmbeddedICE: minor cleanups

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>
15 years agoOverhaul time support API
Zachary T Welch [Sun, 8 Nov 2009 07:20:33 +0000 (23:20 -0800)]
Overhaul time support API

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.

15 years agoAdd private header for ARM11 internals.
Zachary T Welch [Fri, 6 Nov 2009 11:50:26 +0000 (03:50 -0800)]
Add private header for ARM11 internals.

Reduces confusion about location of associated routines and
reduces clutter in the arm11 header.

Removes extra whitespace around the lines touched by these changes.

15 years agoARM11: remove exports and forward decls
Zachary T Welch [Fri, 6 Nov 2009 11:36:46 +0000 (03:36 -0800)]
ARM11: remove exports and forward decls

Unneeded exports cause confusion about the module interfaces.
Make almost everything static in the arm11.c module.

15 years agoARM: minor simulator cleanup
David Brownell [Sun, 8 Nov 2009 20:44:28 +0000 (12:44 -0800)]
ARM: minor simulator cleanup

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>
15 years agoCortex-M3: fix (debug) message priority
David Brownell [Sun, 8 Nov 2009 20:23:17 +0000 (12:23 -0800)]
Cortex-M3: fix (debug) message priority

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agotarget.cfg: remove "-work-area-virt 0"
David Brownell [Sun, 8 Nov 2009 16:52:40 +0000 (08:52 -0800)]
target.cfg: remove "-work-area-virt 0"

The semantics of "-work-area-virt 0" (or phys) changed with
the patch to require specifying physical or virtrual work
area addresses.  Specifying zero was previously a NOP.  Now
it means that address zero is valid.

This patch addresses three related issues:

 - MMU-less processors should never specify work-area-virt;
   remove those specifications.  Such processors include
   ARM7TDMI, Cortex-M3, and ARM966.

 - MMU-equipped processors *can* specify work-area-virt...
   but zero won't be appropriate, except in mischievous
   contexts (which hide null pointer exceptions).

   Remove those specs from those processors too.  If any of
   those mappings is valid, someone will need to submit a
   patch adding it ... along with a comment saying what OS
   provides the mapping, and in which context.  Example,
   say "works with Linux 2.6.30+, in kernel mode".  (Note
   that ARM Linux doesn't map kernel memory to zero ...)

 - Clarify docs on that "-virt" and other work area stuff.

Seems to me work-area-virt is quite problematic; not every
operating system provides such static mappings; if they do,
they're not in every MMU context...

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoFix vsllink bulk out endpoint parsing.
Zachary T Welch [Sat, 7 Nov 2009 21:36:47 +0000 (13:36 -0800)]
Fix vsllink bulk out endpoint parsing.

15 years agoMerge branch 'master' of ssh://dbrownell@openocd.git.sourceforge.net/gitroot/openocd...
David Brownell [Fri, 6 Nov 2009 23:01:46 +0000 (15:01 -0800)]
Merge branch 'master' of ssh://dbrownell@openocd.git.sourceforge.net/gitroot/openocd/openocd

15 years agotarget: don't swap MMU/no-MMU work areas
David Brownell [Fri, 6 Nov 2009 22:57:21 +0000 (14:57 -0800)]
target: don't swap MMU/no-MMU work areas

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>
15 years agoSimplify nand indentation.
Zachary T Welch [Fri, 6 Nov 2009 02:40:52 +0000 (18:40 -0800)]
Simplify nand indentation.

Removes check covered by new nand_command_get_device_by_num helper.
Reverses logic of probe check to further reduce indentation.

15 years agoImprove flash indentation.
Zachary T Welch [Fri, 6 Nov 2009 07:52:03 +0000 (23:52 -0800)]
Improve flash indentation.

Removes redundant tests and reverses backwards logic to reduce the
indentation level in flash.c.

15 years agoFix arm11 vcr command parsing.
Zachary T Welch [Fri, 6 Nov 2009 08:16:39 +0000 (00:16 -0800)]
Fix arm11 vcr command parsing.

15 years agoARM: shrink offsets
David Brownell [Fri, 6 Nov 2009 06:04:25 +0000 (22:04 -0800)]
ARM: shrink offsets

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>
15 years agoARM: other code uses the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 06:04:22 +0000 (22:04 -0800)]
ARM: other code uses the new inheritance/nesting scheme

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>
15 years agoCortex-A8: use the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 06:04:13 +0000 (22:04 -0800)]
Cortex-A8: use the new inheritance/nesting scheme

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>
15 years agoARM7TDMI uses the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 06:04:04 +0000 (22:04 -0800)]
ARM7TDMI uses the new inheritance/nesting scheme

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>
15 years agoARM720 uses the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 06:03:56 +0000 (22:03 -0800)]
ARM720 uses the new inheritance/nesting scheme

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>
15 years agoXScale uses the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 06:03:45 +0000 (22:03 -0800)]
XScale uses the new inheritance/nesting scheme

Use target_to_xscale(), replacing needless pointer traversals
and simplifying a bunch of code.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoARM9TDMI uses the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 06:03:40 +0000 (22:03 -0800)]
ARM9TDMI uses the new inheritance/nesting scheme

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>
15 years agoARM966 uses the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 06:03:33 +0000 (22:03 -0800)]
ARM966 uses the new inheritance/nesting scheme

Use target_to_arm966(), replacing needless pointer traversals.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoARM926 uses the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 06:03:30 +0000 (22:03 -0800)]
ARM926 uses the new inheritance/nesting scheme

Use target_to_arm926(), replacing needless pointer traversals
and simplifying a bunch of code.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoFA526 uses the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 06:03:24 +0000 (22:03 -0800)]
FA526 uses the new inheritance/nesting scheme

Replace needless pointer traversals, simplify.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
15 years agoARM920 uses the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 06:03:13 +0000 (22:03 -0800)]
ARM920 uses the new inheritance/nesting scheme

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>
15 years agoCortex-M3: use the new inheritance/nesting scheme
David Brownell [Fri, 6 Nov 2009 05:59:45 +0000 (21:59 -0800)]
Cortex-M3: use the new inheritance/nesting scheme

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>
15 years agotarget: provide container_of()
David Brownell [Fri, 6 Nov 2009 05:59:39 +0000 (21:59 -0800)]
target: provide container_of()

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>
15 years agoARM7TDMI: remove forward decls
David Brownell [Fri, 6 Nov 2009 04:36:42 +0000 (20:36 -0800)]
ARM7TDMI: remove forward decls

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>
15 years agoARM720: remove exports and forward decls
David Brownell [Fri, 6 Nov 2009 04:36:27 +0000 (20:36 -0800)]
ARM720: remove exports and forward decls

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>
15 years agoFA526: remove exports and forward decls
David Brownell [Fri, 6 Nov 2009 04:36:18 +0000 (20:36 -0800)]
FA526: remove exports and forward decls

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>
15 years agoARM9TDMI: remove forward decls
David Brownell [Fri, 6 Nov 2009 04:36:09 +0000 (20:36 -0800)]
ARM9TDMI: remove forward decls

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>