jtag: fix handling of 'tap enable' error

if a tap could not be _enabled_, the error message was
'failed to disable tap'. Fixed that. Also, display the failing
tap's name.

Signed-off-by: Thomas Koeller <thomas.koeller@baslerweb.com>

DM36x: Set OSCDIV divider

The ability to set up the OSCDIV divider was missing.

Signed-off-by: Thomas Koeller <thomas.koeller@baslerweb.com>

DM36x: Disable unused SYSCLKs

Clear the enable bits for all clocks that are not set explicitly.
This is done to increase robustness by removing pre-existing
state.

Signed-off-by: Thomas Koeller <thomas.koeller@baslerweb.com>

DM36x: Use enable bit for PLL pre-divider

The PLL pre- and postdividers seem to have enable bits, although
these are not mentioned in the chip documentation.

Signed-off-by: Thomas Koeller <thomas.koeller@baslerweb.com>

str9x: faster flash erase of entire chip

The patch improves flash erase for STR9x in case of a full bank erase.
Then the chip erase command is used instead which improves speed significantly.

Also I think it might help if e.g. STR912 enters some state where flash banks are locked, and a chip erase command is the key for unlocking the flash.

tcl: remove silly ocd_ prefix to array2mem and mem2array

ocd_ prefix is used internally in OpenOCD as a kludge more
or less to deal with the two kinds of commands that OpenOCD
has.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

board: added at91cap7a stk w/sdram config scripts

The strange thing here with this board is that 16MHz kinda
works, but only 2MHz is really stable.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

debug: use assert's when approperiate

error was returned instead of using assert.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm: add missing error reporting

when an unknown core mode is read from the target,
report error. Can be communication failure.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

config scripts: remove useless reference to OpenOCD docs

clutters config scripts.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

cfg: add omapl138 support and da850evm preliminary support

This patch adds support for the omapl138 target and preliminary support for the da850evm. The
target cfg file is based on the icepick routing done by the target/ti_dm6446.cfg file.

I have performed limited testing with this setup. I am posting this patch in the interest of
sharing cfg files and in the hopes that the experts on this list can correct errors I have made or
point out enhancements.

The testing I have performed is debugging uboot with gdb where I also use the following local.cfg
and gdbinit files. Debugging appears to work in so much as 'ni' works.

local.cfg:
gdb_memory_map disable

gdbinit:
target remote localhost:3333
set remote hardware-breakpoint-limit 2
set remote hardware-watchpoint-limit 2
monitor poll on

Comments welcome.

Best Regards,
Ben Gardiner

arm7/9: fix "reset run + halt"

if polling is off, then "reset run + halt" would fail
since halt incorrectly assumed the target was in the
reset state as it is the internal poll implementation
that moves the sw tracking of the target state out
of the reset state.

To reproduce:

> reset run; halt
JTAG tap: zy1000.cpu tap/device found: 0x1f0f0f0f (mfg: 0x787, part: 0xf0f0, ver: 0x1)
BUG: arm7/9 does not support halt during reset. This is handled in arm7_9_assert_reset()

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

target: if polling fails, back off

back-off algorithm for polling. Double polling
interval up to 5000ms when it fails.

when polling succeeds, reset backoff.

This avoids flooding logs(as much) when working
with conditions where the target polling will fail.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

Luminary-icdi comment update

Clarify that ICDI is the generic logic, but this config is
for the JTAG-only (no-SWD) mode.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>

xscale documentation: vector table handling

Hi everyone.  I noticed some incorrect information in the user manual
regarding how the vector table is handled on the xscale, so for your
consideration, here's a short patch that corrects it, and adds a
little more detail I thought might be helpful.

The documentation states that OpenOCD does not attempt to synchronize
the vector tables in memory with those stored in the "mini instruction
cache".  In fact, on each resume it does copy from memory to the cache
all entries in the high and low tables that were not previously
defined using the 'xscale vector_table' command. (In
src/target/xscale.c, see xscale_update_vectors(), which is invoked by
xscale_resume().)  I take advantage of this during Linux boot-up.  The
extra detail describes in general terms how I do this.

Corrections, comments are of course gratefully received.

Thanks,
Mike

Signed-off-by: Mike Dunn <mikedunn@newsguy.com>

verify_image: print out a statement that there are no further errors

It is useful to know that the printed errors are *all* the
errors there were.

Added missing error handling(found by inspection).

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

zy1000: use correct base clock when calculating speed divisor

revc uses 60MHz and revb 64MHz, use this in calculations.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

zy1000: print out khz correctly in response to setting JTAG speed

Calculate printout based on same core routines.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

lpc1768: even if rclk "works", it isn't necessarily the correct clk

rclk = 4MHz oon lpc1768, the correct JTAG clk is 666MHz(4MHz/6).

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

ioutil: fix internal fn names

prefix w/ioutil_

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

jtag: measure_clk debug proc

It can be useful to get an approximate measurement of
rtck frequency for debugging purposes.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

jtag: add jtag_flush_queue_sleep debug command

it can be useful to throttle performance: test
differences in behavior, test performance effect
of long roundtrips.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

util: ms command to calculate length of operations

This can be used to calculate approximate RTCK frequency
for instance.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

Support NGX Technologies product NGX ARM USB JTAG

This is a standard FT2232 device. More info at their web page:
http://shop.ngxtechnologies.com/product_info.php?cPath=26&products_id=30

Remove srst_pulls_trst from LPC2148 target

srst_pulls_trst is only true on some (broken) LPC2148 boards, a fact
which is already documented in doc/openocd.texi, so it shouldn't be
set unconditionally in the target tcl.

This patch was needed to reflash when an Abort exception occured very
early after reset, before OpenOCD tried to halt the CPU.

stellaris flash comment update

This flash driver works on more than just two chips.
(Though it does need work still, e.g. to protect more than 64K.
(On non-'3748-A0 chips where errata allow that.))

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>

lpc7168: make flash available upon reset init

set user mode to avoid ROM being mapped at address
0 rather than flash.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

"transport select" returns Jim value

Make it scriptable, so code can be conditionalized based on
what transport is in use for the session.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>

Buspirate: fix shadow

docs: removed unused cmd 'fast'

The cmd fast was removed in commit c0d14dc7f19d785702eee5f69de5b1a63902554b

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

flash: remove algorithm exit_point address for supported targets

For the above targets the exit_point is
optional when used with run_algorithm, so remove it.

This makes updating the algorithm less error prone.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

mips32: exit_point optional for mips32_run_algorithm

As the mips32 uses instruction breakpoints for algorithms we do not really
need to check the pc on exit.

This now matches the behaviour of the arm codebase.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

armv7m: exit_point optional for armv7m_run_algorithm

As the armv7m uses instruction breakpoints for algorithms we do not really
need to check the pc on exit.

This now matches the behaviour of the arm4_5 codebase.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

armv4_5: add algorithms instruction breakpoint support

Update the arm_checksum_memory and arm_blank_check_memory
algorithms to use a breakpoint instruction on v5 arch.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

arm11 error propagation fixes

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm_jtag_scann error propagation fixes

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm_dpm: error propagation fixes

found by inspection

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

adi_jtag_ error propagation

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm: error propagation of arm_jtag_set_instr

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm_adi_v5: dap_run() error propagation

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm_adi_v5: mem_ap_write error propagation

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm_adi_v5: error propagation of mem_ap_read_atomic_u32 failure

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

mem_ap_read_u32 error propagation

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm_adi_v5: add error propagation for dap_setup_accessport

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

debug: debug entry error propagation

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

cortex a8: added timeout handling

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm: add error propagation for enable/disable mmu caches

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm: add error propagation to generic get_ttb fn

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

more careful luminary init

Set up more of the Luminary-specific signals, and stop cloning
a few of the JTAG defaults.  More comments too.

Still leaves the "dap info 0" bugs unresolved (presumably coupled
to this particular adapter family) where TPIU, ITM, DWT, and other
debug modules wrongly display as extra NVICs.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>

cfg: add Amontec JTAGkey2p interface config (Issue #26)

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

flash: add nuc910 nand driver

This adds a nand driver support for the nuc910 target.
Note that ECC is not currently supported by this driver, although
it is supported by the peripheral.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

nand: nand probe also outputs manufacturer

The nand probe now outputs the manufacturer if found.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

cfg: update rsc-w910 script

 - Only enable the FMI (NAND) and DMA clocks.
 - Select NAND interface on the MFSEL.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

docs: missing parameter from nand check_bad_blocks

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

cortex a8: lots of error propagation fixes

found by code inspection

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

cortex a8: mem_ap_read_buf_u32() error handling

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

cortex a8: add missing error handling for cortex_a8_dap_write/read_coreregister_u32()

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

cortex a8: add missing error handling from cortex_a8_exec_opcode()

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

cortex a8: add missing error handling for mem_ap_atomic_write_u32()

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

cortex a8: add missing error handling for mem_ap_read_atomic_u32()

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

comments for Luminary ICDI layout

Provide $defines for more of the signals involved in the
Luminary ICDI hardware, and comment some of what's going on.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>

spelling fixes in comments

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

debug-feature: jtagtcpip, improve jtag performance

postpone callbacks until jtag execute queue time.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

lm3s811-ek uses generic stellaris target config

There's no point in an lm3s811-specific target file,
so remove it in favor of the generic "stellaris.cfg".

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>

ARM ADI-V5: cleanup CID/PID addressing

Use addition for offsetting, not masking.  Shorten some lines.
Make "component_start" print-only (unused otherwise; don't save).

Still doesn't resolve the issue where multiple components
are wrongly displaying as NVICs on some Cortex-M3 parts because
many PIDs appear to be zeroes ... maybe adapter related??

Signed-off-by: David Brownell <db@helium.(none)>

flash: use bkpt for str9 flash algo algorithm

STR9 is ARMv5 arch so use bkpt instruction rather than
waste a breakpoint.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

fix warnings

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

debug-feature: jtagtcpip, reduce performance impact of ping times

queue read back of data

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

debug feature: jtagtcpip, improve performance

waiting for ZY1000 fifo to idle is now queued as
an asynchronous command. This radically improves
performance when waitIdle() is interspersed with
writes as no readback is required over TCP/IP.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

debug feature: jtagtcpip, improved performance

only check error flag when rclk is actually enabled.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

ARM ADI-V5: PIDs and CIDs are 8 bits

Mask the upper bits after 32-bit reads.

Alsoo revert the ugly changes to use PRIx32; just cast to unsized
integers when printing (two chars not eight).

Signed-off-by: David Brownell <db@helium.(none)>

cfg: add Avalue RSC-W910 config

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

jtag: fix shadow issues in adapter_init

Use global jtag_only rather than local static.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

arm9: revert change arm9tdmi cmd group name to arm9

This reverts the incorrect change made to the arm9 cmd group in commit
d1eca9a74c7c57ba6a3210c51b2a10cc5adb22e1.

The code now matches the docs and the release notes.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

docs: fix transport typo

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

adapter.c: missing space in handle_interface_command

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

transport selection tweaks

* Bugfix and simplify legacy jtag-only defaulting
* Make "dummy" declare its jtag-only nature
* likewise update ft2232
* warn if selection is _required_ (multi-transport adapters),

fixes the "only ft2232 works" bug for at least dummy, with
other drivers going the "legacy" path (submit patches).

Signed-off-by: David Brownell <db@helium.(none)>

transport: fix bug/typo in transport_register cmd

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

transport: fix segfault in transport select

String compare against addresses in range 0 or so due
to not checking if there was an active session first.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

transport: fix bug/typo in interface_transports command

Classic manifestation of weakness of dangling {
formatting :-) Hard to spot these sort of mistakes visually.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

gdbserver: incorrect memory map for multiple targets (bug #24)

The gdb server incorrectly reports the memory map if we have
multiple targets with multiple flash banks.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

transport: fix segfault in setup_command_handler()

Commit 93f2afa45f4c dropped the sentinel off the end
of the command_registrants[] array.  The loop immediately
following the initialization will walk right off the end.

Signed-off-by: Stacey Sheldon <stac@solidgoldbomb.org>

Fix minor openocd.texi bug
::X
Signed-off-by: David Brownell <db@helium.(none)>

initial "transport" framework

This adds the guts of a transport framework with initialization,
which should work with current JTAG-only configurations (tested
with FT2232).

Each debug adapter can declare the transports it supports, and
exactly one transport is initialized.  (with its commands) in
any given OpenOCD session.

  * Define a new "struct transport with init hooks and a few
 "transport"  subcommands to support it:

     "list" ... list the transports configured (just "jtag" for now)
     "select" ... makes the debug session use that transport
     "init" ... initializes the selected transport (internal)

  * "interface_transports" ... declares transports the current interface
    can support.  (Some will do this from C code instead, when there are
    no hardware versioning (or other) issues to prevent it.

Plus some FT2232 tweaks, including a few to streamline upcoming
support for an SWD transport (initially for Luminary adapters).

Eventually src/jtag should probably become src/transport, moving
jtag-specific stuff  to transport/jtag.

Signed-off-by: David Brownell <db@helium.(none)>

ft2232: revert ft2232_read_scan changes

Revert change made in commit dd88b461da1cb8642200dd5c96fb1ff384ca9f7b.
Caused segfaults when using ftdi driver under win32.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

ft2232: bugfix from previous commit

fix build error with commit dd88b461da1cb8642200dd5c96fb1ff384ca9f7b

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>

ft2232: simplify ft2232_read_scan

ft2232: cleanup ft2232_write

* Include the size check into ft2232_write, so calling it is simpler.
* Use sizeof(buffer) when possible

at91sam3s* support

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

openocd: setup_command_handler() must not be static

when OpenOCD is linked with an app this fn can be used from
the outside.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

arm11: fix gaffe in no-ack transfers

The code did not transfer the last word in no-ack transfers.

The strange thing is that this did not lead to any
observable errors.

This gaffe was introduced in commit 1f5883ea56cb058221f

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

error number: review

Review allocation of error numbers in openocd
to avoid overlap.
Put brackets around negative numbers to avoid
issues during macro expansion.

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>

am3517 evm: use physical write to memory while target is running

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

cortex a8: only physical read/write's are available when target is running

Memory read/writes to virtual memory, requires that the CPU is
halted.

Use 'phys' option to write to memory while target is running.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

target: $_TARGET mdw now has a phys option

just like the mdw command

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

target: mwX on target object now supporst phys argument

$_TARGETNAME mww phys 0x10 0xdeadbeef

=> write 0xdeadbeef to physical address 0x10

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

xsvf: Fix shadow issues on Mac

wait is declared in /usr/include/sys/wait.h

Signed-off-by: Edgar Grimberg <edgar.grimberg@zylin.com>

target: Fix shadow issues on Mac

wait is declared in /usr/include/sys/wait.h

Signed-off-by: Edgar Grimberg <edgar.grimberg@zylin.com>

flash: fix shadow issues on Mac

Wait is declared in /usr/include/sys/wait.h

Signed-off-by: Edgar Grimberg <edgar.grimberg@zylin.com>

board: add alpha am3517evm ti board config file

Signs of life: reset(kinda), halt, resume and memory
display/modify.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>