]> git.sur5r.net Git - openocd/log
openocd
14 years agoARMV4_5_MMU: review unused symbols
Antonio Borneo [Sat, 10 Apr 2010 08:46:59 +0000 (16:46 +0800)]
ARMV4_5_MMU: review unused symbols

Remove unused data:
- armv4_5_mmu_page_type_names
Remove prototype of not existing function:
- armv4mmu_translate_va

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agoARMV4_5: review scope of data
Antonio Borneo [Sat, 10 Apr 2010 08:46:01 +0000 (16:46 +0800)]
ARMV4_5: review scope of data

Add "static" qualifier to private data.

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agoEMBEDDEDICE: review scope of functions
Antonio Borneo [Sat, 10 Apr 2010 08:44:54 +0000 (16:44 +0800)]
EMBEDDEDICE: review scope of functions

Add "static" qualifier to private functions.

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agoTARGET: review unused symbols
Antonio Borneo [Fri, 9 Apr 2010 16:25:35 +0000 (00:25 +0800)]
TARGET: review unused symbols

Remove unused functions:
- target_all_handle_event

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agoTARGET: review scope of functions
Antonio Borneo [Fri, 9 Apr 2010 16:22:46 +0000 (00:22 +0800)]
TARGET: review scope of functions

Add "static" qualifier to private functions.
Remove unused "extern" in src/ecosboard.c

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agoARMv7M: review scope of functions
Antonio Borneo [Fri, 9 Apr 2010 16:07:40 +0000 (00:07 +0800)]
ARMv7M: review scope of functions

Add "static" qualifier to private functions.

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agoRestore deleted '!' character
David Brownell [Sun, 4 Apr 2010 08:28:50 +0000 (01:28 -0700)]
Restore deleted '!' character

I'm not sure what caused this significant character to get deleted.
it may be related to intermittent Editor or terminal flakes  I've
been seeing lately (sigh).  This fix is trivial.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agotarget: are we running algorithm code?
David Brownell [Sun, 4 Apr 2010 07:42:05 +0000 (00:42 -0700)]
target: are we running algorithm code?

Fixing one bug can easily uncover another  .... in this case,
making sure that we properly invalidate some cached NOR state when
resuming arbitrary target code turned up an issue when the code
wasn't quite arbitrary (and we couldn't know that, but some parts
of OpenOCD assumed the cache would not be invalidated.

Specifically:  some flash drivers (like CFI) update that state in loops
with downloaded algorithms, thus invalidating the state as it's probed.

 + Add a new target state flag, to record whether the target is
  running downloaded algorithm code.

 + Use that flag to add a special case:  "trust" downloaded algorithms
   not to corrupt that cached state, bypassing cache invalidation.

Also update some of the documentation to stipulate that this flavor of
trustworthiness is now *required* ... not just a fortuitous acident.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agosimplify and unconfuse target_run_algorithm()
David Brownell [Sun, 4 Apr 2010 07:38:39 +0000 (00:38 -0700)]
simplify and unconfuse target_run_algorithm()

For some reason there are *two* schemes for interposing logic into
the run_algorithm() code path...  One is a standard procedural wapper
around the target method invocation.

the other (superfluous) one hacked the method table by splicing
a second procedural wrapper into the method table.  Remove it:

* Rename its  slightly-more-featureful wrapper so it becomes
  the standard procedural wrapper, leaving its added logic
  (where it should have been in the first place.

          Also add a paranoia check, to report targets that don't
  support algorithms without traversing a NULL pointer, and
  tweak its code structure a bit so it's easier to modify.

* Get rid of the superfluous/conusing method table hacks.

This is a net simplification, making it simpler to analyse what's
going on, and then interpose logic . ... by ensuring there's only one
natural place for it to live.

------------

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agobuildfix
David Brownell [Sun, 4 Apr 2010 01:36:07 +0000 (18:36 -0700)]
buildfix

Without this, a system using gcc (GCC) 4.2.4 (Ubuntu 4.2.4-1ubuntu4)
aborts builds after reporting:

tcl.c: In function ‘handle_irscan_command’:
tcl.c:1168: warning: passing argument 1 of ‘buf_set_u32’ discards qualifiers from pointer target type

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoxscale: fix trace buffer functionality when resuming from a breakpoint
Mike Dunn [Sun, 28 Mar 2010 19:48:32 +0000 (15:48 -0400)]
xscale: fix trace buffer functionality when resuming from a breakpoint

Problem: halt at a breakpoint, enable trace buffer ('xscale trace_buffer enable
fill'), then resume.  Wait for debug exception when trace buffer fills (if not
sooner due to another breakpoint, vector catch, etc).  Instead, never halts.
When halted explicitly from OpenOCD and trace buffer dumped, it contains only
one entry; a branch to the address of the original breakpoint.  If the above
steps are repeated, except that the breakpoint is removed before resuming, the
trace buffer fills and the debug exception is generated, as expected.

Cause: related to how a breakpoint is stepped over on resume.  The breakpoint is
temporarily removed, and a hardware breakpoint is set on the next instruction
that will execute.  xscale_debug_entry() is called when that breakpoint hits.
This function checks if the trace buffer is enabled, and if so reads the trace
buffer from the target and then disables the trace (unless multiple trace
buffers are specified by the user when trace is enabled).  Thus you only trace
one instruction before it is disabled.

Solution: kind of a hack on top of a hack, but it's simple.  Anything better
would involve some refactoring.  This has been tested and trace now works as
intended, except that the very first instruction is not part of the trace when
resuming from a breakpoint.

TODO: still many issues with trace: doesn't work during single-stepping (trace
buffer is flushed each step), 'xscale analyze_trace' works only marginally for
a trace captured in 'fill' mode, and not at all for a trace captured in 'wrap'
mode.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag/tcl.c cleanup -- split out "adapter.c"
David Brownell [Sat, 27 Mar 2010 17:07:13 +0000 (10:07 -0700)]
jtag/tcl.c cleanup -- split out "adapter.c"

Clean up the jtag/tcl.c file, which was one of the biggest and
messiest ones in that directory.  Do it by splitting out all the
generic adapter commands to a separate "adapter.c" file (leaving
the "tcl.c" file holding only JTAG utilities).

Also rename the little-used "jtag interface" to "adapter_name", which
should have been at least re-categorized earlier (it's not jtag-only).

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoNOR TCL: fix usage message
Antonio Borneo [Fri, 26 Mar 2010 07:54:33 +0000 (15:54 +0800)]
NOR TCL: fix usage message

The command "flash bank" has updated syntax.
Add the mandatory parameter <target> to the usage message
that prints in case of error.

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agoTCL scripts: update to current "flash bank" syntax
Antonio Borneo [Fri, 26 Mar 2010 07:17:46 +0000 (15:17 +0800)]
TCL scripts: update to current "flash bank" syntax

While "flash bank" syntax has been changed long ago,
several tcl script are still not fully update.

Fix following cases related with "cfi" driver:
- syntax error: the mandatory <name> parameter is missing
- warning: the <target> parameter is a number, instead of
  the target name
- the comment line above the command does not report
  actual syntax

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agozy1000: dev tool
Øyvind Harboe [Thu, 4 Mar 2010 18:58:42 +0000 (19:58 +0100)]
zy1000: dev tool

first cut peek/poke over tcp/ip, used for debug/research
purposes only. Long term JTAG over TCP/IP might be an
offshoot. The performance is usable for development/testing
purposes.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agotarget: fix poll off
Øyvind Harboe [Thu, 25 Mar 2010 12:06:23 +0000 (13:06 +0100)]
target: fix poll off

I don't know when "poll off" broke, but "poll off" didn't
stop background polling of target. The polling status flag
simply wasn't checked in the handle_target timer callback.

All target polling(including power/reset state) is now stopped
upon "poll off".

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agochange %x and %d to PRIx32 and PRId32 where needed for cygwin
Daniel Bäder [Thu, 25 Mar 2010 10:24:42 +0000 (11:24 +0100)]
change %x and %d to PRIx32 and PRId32 where needed for cygwin

14 years agozy1000: allow it to build on linux host for testing purposes
Øyvind Harboe [Thu, 4 Mar 2010 18:03:26 +0000 (19:03 +0100)]
zy1000: allow it to build on linux host for testing purposes

For testing and checking the build this can be useful,
it doesn't have any practical application outside development.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agotelnet_server: review unused symbols
Antonio Borneo [Wed, 24 Mar 2010 05:45:29 +0000 (13:45 +0800)]
telnet_server: review unused symbols

Remove unused function

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agotelnet_server: review scope of functions
Antonio Borneo [Wed, 24 Mar 2010 05:44:40 +0000 (13:44 +0800)]
telnet_server: review scope of functions

Add "static" qualifier to private functions.

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agoFT2232 Messaaging fix
David Brownell [Wed, 24 Mar 2010 23:04:26 +0000 (16:04 -0700)]
FT2232 Messaaging fix

The init cleanup patch overlooked a message which was
wrongly specific to the "usbjtag" layout.  Fix.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoserver: review unused symbols
Antonio Borneo [Wed, 24 Mar 2010 04:08:45 +0000 (12:08 +0800)]
server: review unused symbols

Remove unused function

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agoserver: review scope of functions and data
Antonio Borneo [Wed, 24 Mar 2010 04:06:45 +0000 (12:06 +0800)]
server: review scope of functions and data

Add "static" qualifier to private functions and data.

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
14 years agozy1000: fix optimisaion bug in dcc writes
Øyvind Harboe [Mon, 22 Mar 2010 07:27:03 +0000 (08:27 +0100)]
zy1000: fix optimisaion bug in dcc writes

Introduced & corrected since 0.4.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agofix software breakpoints on xscale
Mike Dunn [Sat, 20 Mar 2010 14:53:47 +0000 (10:53 -0400)]
fix software breakpoints on xscale

This patch fixes xscale software breakpoints by cleaning the dcache and
invalidating the icache after the bkpt instruction is inserted or removed.  The
icache operation is necessary in order to flush the fetch buffers, even if the
icache is disabled (see section 4.2.7 of the xscale core developer's manual).
The dcache is presumed to be enabled; no harm done if not.  The dcache is also
invalidated after cleaning in order to safeguard against a future load of
invalid data, in the event that cache_clean_address points to memory that is
valid and in use.

Also corrected a confusing typo I noticed in a comment.

TODO (or not TODO...?): the xscale's 2K "mini dcache" is not cleaned.  This
cache is not used unless the 'X' bit in the page table entry is set.  This is a
proprietary xscale extension to the ARM architecture.  If a target's OS or
executive makes use of this for memory regions holding code, the breakpoint
problem will persist.  Flushing the mini dcache requires that 2K of valid
cacheable memory (mapped with 'X' bit set) be designated by the user for this
purpose.  The debug handler that gets downloaded to the target will also need to
be extended.

14 years agobitq: fix warning now that out_value is const
Øyvind Harboe [Mon, 22 Mar 2010 06:21:09 +0000 (07:21 +0100)]
bitq: fix warning now that out_value is const

This was an easy one. Just add the missing "const" to a
local variable definition.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agoft2232 init mess cleanup
David Brownell [Mon, 22 Mar 2010 05:49:23 +0000 (22:49 -0700)]
ft2232 init mess cleanup

In the ft2232 driver, initialization for many layouts punts to a routine
called usbjtag_init(), instead of a routine specific to each layout.

That routine is  a mess  built around a "what type layout am I" core.
That's a bad design ... in this case, especially so, since it bypasses
the layout-specific dispatch which was just done, and obfuscates the
initialization which is at least somewhat generic, instead of being
specific to the "usbjtag" layout.

Split and document out the generic parts of usbjtag_init(), and make
the rest of those layouts have layout-specific init methods.  Also,
rename usbjtag_reset() ... that also was not specific to the "usbjtag"
layout, and thus contributed to the previous code structure confusion.

(Eventually, all layout-specific code (and method tables) should probably
live in files specific to each layout.  These changes will facilitate
those and other cleanups to this driver.)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoarm breakpoints: amended fix comment
Øyvind Harboe [Sun, 21 Mar 2010 18:20:26 +0000 (19:20 +0100)]
arm breakpoints: amended fix comment

the handling of caches, should be moved into the breakpoint
specific callbacks rather than being plonked into generic
memory write fn's.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag: make out_value const
Øyvind Harboe [Fri, 19 Mar 2010 21:06:01 +0000 (22:06 +0100)]
jtag: make out_value const

Tightens up the jtag_add_xxx_scan() API

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag: move towards making out_value const
Øyvind Harboe [Fri, 19 Mar 2010 21:04:45 +0000 (22:04 +0100)]
jtag: move towards making out_value const

These were relatively straightforward fixes which are
backwards compatible.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agoFT2232 comment tweaks
David Brownell [Fri, 19 Mar 2010 17:31:44 +0000 (10:31 -0700)]
FT2232 comment tweaks

Note that the FT4232 chips have four channels not two, and
Elaborate on uses of the additional channels.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agozy1000: clean up jtag_add_xx_scan fn's
Øyvind Harboe [Fri, 19 Mar 2010 13:47:01 +0000 (14:47 +0100)]
zy1000: clean up jtag_add_xx_scan fn's

The implementation is now more straightforward as the
scan_fields have been greatly simplified over time.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agozy1000: fix bug in end state of DCC writes
Øyvind Harboe [Fri, 19 Mar 2010 13:27:19 +0000 (14:27 +0100)]
zy1000: fix bug in end state of DCC writes

Introduced in latest commits, found by code inspection &
GCC warning.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag: retire jtag_get/set_end_state()
Øyvind Harboe [Thu, 18 Mar 2010 17:53:39 +0000 (18:53 +0100)]
jtag: retire jtag_get/set_end_state()

Voila! This get rids of mysteries about what what
state the TAP is in.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag: remove jtag_get_end_state() usage
Øyvind Harboe [Thu, 18 Mar 2010 17:41:43 +0000 (18:41 +0100)]
jtag: remove jtag_get_end_state() usage

Code inspection indicated what constant end states to
use.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag: get rid of unecessary jtag_get_end_state()
Øyvind Harboe [Thu, 18 Mar 2010 17:37:23 +0000 (18:37 +0100)]
jtag: get rid of unecessary jtag_get_end_state()

By code inspection.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag: remove unecessary usage of jtag_get_end_state().
Øyvind Harboe [Thu, 18 Mar 2010 17:34:34 +0000 (18:34 +0100)]
jtag: remove unecessary usage of jtag_get_end_state().

By code inspection.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag: remove jtag_get_end_state()'s that should be unecessary
Øyvind Harboe [Thu, 18 Mar 2010 17:31:58 +0000 (18:31 +0100)]
jtag: remove jtag_get_end_state()'s that should be unecessary

By a bit of code inspection it seems like all of these
instances of jtag_get_end_state() can be unambigously
replaced by constants.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agoFix underlying problem with xscale icache and dcache commands
Mike Dunn [Fri, 19 Mar 2010 04:34:13 +0000 (21:34 -0700)]
Fix underlying problem with xscale icache and dcache commands

Fix problem with the xscale icache and dcache commands.  Both commands were
enabling or disabling the mmu, not the caches

I didn't look any further after my earlier patch fixed the trivial problem
with command argument parsing.  Turns out the underlying code was broken.

The resolution is straightforward when you look at the arguments to
xscale_enable_mmu_caches() and xscale_disable_mmu_caches().  I finally
took a deeper look after dumping the cp15 control register (XSCALE_CTRL)
and seeing that the cache bits weren't changing, but the mmu bit was
(which caused all manner of grief, as you can imagine).  This has been
tested and works OK now.

 src/target/xscale.c |   17 +++++++++++------
 1 files changed, 11 insertions(+), 6 deletions(-)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoADI_v5 - it's not always an "SWJ-DP"
David Brownell [Thu, 18 Mar 2010 19:32:35 +0000 (12:32 -0700)]
ADI_v5 - it's not always an "SWJ-DP"

So don't use the name "swjdp" for all DAPs; rename to
plain old "dap", which *is* always correct.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoMerge branch 'master' of ssh://dbrownell@openocd.git.sourceforge.net/gitroot/openocd...
David Brownell [Thu, 18 Mar 2010 19:11:58 +0000 (12:11 -0700)]
Merge branch 'master' of ssh://dbrownell@openocd.git.sourceforge.net/gitroot/openocd/openocd

14 years agoremove more duplication
David Brownell [Thu, 18 Mar 2010 18:56:17 +0000 (11:56 -0700)]
remove more duplication

Not sure how the original "move code to adi_v5_swd.c" patch left
some code in the "arm_adi_v5.c" file, but a recent patch was only
a partial fix -- it didn't remove all the duplication.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agojtag: retire one instance of jtag_get_end_state() usage
Øyvind Harboe [Thu, 18 Mar 2010 11:07:41 +0000 (12:07 +0100)]
jtag: retire one instance of jtag_get_end_state() usage

Less global variables....

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agooops: committed and pushed two temp files....
Øyvind Harboe [Thu, 18 Mar 2010 11:06:07 +0000 (12:06 +0100)]
oops: committed and pushed two temp files....

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agoDOCS: update flash bank examples
Spencer Oliver [Thu, 18 Mar 2010 09:18:53 +0000 (09:18 +0000)]
DOCS: update flash bank examples

 - include the $_FLASHNAME in all flash bank examples.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoPIC32MX: add unlock cmd
Spencer Oliver [Thu, 18 Mar 2010 09:35:45 +0000 (09:35 +0000)]
PIC32MX: add unlock cmd

'unlock' performs a full unlock/erase of the device, removing any
code protection.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoMIPS: remove unused arg from mips_ejtag_set_instr
Spencer Oliver [Wed, 17 Mar 2010 17:24:22 +0000 (17:24 +0000)]
MIPS: remove unused arg from mips_ejtag_set_instr

This arg was never used and was just taken from the arm jtag code.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agojtag: cut down on usage of unintended modification of global end state
Øyvind Harboe [Tue, 16 Mar 2010 13:13:03 +0000 (14:13 +0100)]
jtag: cut down on usage of unintended modification of global end state

jtag_get/set_end_state() is now deprecated.

There were lots of places in the code where the end state was
unintentionally modified.

The big Q is whether there were any places where the intention
was to modify the end state. 0.5 is a long way off, so we'll
get a fair amount of testing.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agolinker error: fix problem with duplicate fn
Øyvind Harboe [Wed, 17 Mar 2010 20:50:23 +0000 (21:50 +0100)]
linker error: fix problem with duplicate fn

A fn was copied instead of moved to a new file. The linker
can discard exact copies of fn's without warning.

This is a C++'ism.

However on my Ubuntu 9.10 machine, it fails.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agomips: fix warning
Øyvind Harboe [Wed, 17 Mar 2010 20:34:43 +0000 (21:34 +0100)]
mips: fix warning

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agogdb: long running "monitor mww" now works w/gdb
Øyvind Harboe [Wed, 17 Mar 2010 11:59:33 +0000 (12:59 +0100)]
gdb: long running "monitor mww" now works w/gdb

invoke keep_alive() to make sure that the default 2000ms
timeout does not trigger.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agotarget: faster mww operations
Øyvind Harboe [Tue, 16 Mar 2010 17:02:58 +0000 (18:02 +0100)]
target: faster mww operations

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agotarget: mdX/mwX on target were badly broken
Øyvind Harboe [Wed, 17 Mar 2010 11:43:47 +0000 (12:43 +0100)]
target: mdX/mwX on target were badly broken

- incorrect parsing of arguments
- mdX didn't display arguments correctly

I don't think anyone ever used that code path :-)

Did you know that "target mdw" and mdw are very different?

for {set i 0} {$i < 256} {set i [expr $i+1]} {mwb [expr 0x2000000+$i] $i}

 mdw 0x2000000 0x10
0x02000000: 03020100 07060504 0b0a0908 0f0e0d0c 13121110 17161514 1b1a1918 1f1e1d1c
0x02000020: 23222120 27262524 2b2a2928 2f2e2d2c 33323130 37363534 3b3a3938 3f3e3d3c

> zy1000.cpu mdb 0x2000000 0x20
0x02000000 00 01 02 03 04 05 06 07 08 09 0a 0b 0c 0d 0e 0f ................
0x02000010 10 11 12 13 14 15 16 17 18 19 1a 1b 1c 1d 1e 1f ................
> zy1000.cpu mdh 0x2000000 0x20
0x02000000 0100 0302 0504 0706 0908 0b0a 0d0c 0f0e ................
0x02000010 1110 1312 1514 1716 1918 1b1a 1d1c 1f1e ................
0x02000020 2120 2322 2524 2726 2928 2b2a 2d2c 2f2e  !"#$%&'()*+,-./
0x02000030 3130 3332 3534 3736 3938 3b3a 3d3c 3f3e 0123456789:;<=>?
> zy1000.cpu mdw 0x2000000 0x20
0x02000000 03020100 07060504 0b0a0908 0f0e0d0c ................
0x02000010 13121110 17161514 1b1a1918 1f1e1d1c ................
0x02000020 23222120 27262524 2b2a2928 2f2e2d2c  !"#$%&'()*+,-./
0x02000030 33323130 37363534 3b3a3938 3f3e3d3c 0123456789:;<=>?
0x02000040 43424140 47464544 4b4a4948 4f4e4d4c @ABCDEFGHIJKLMNO
0x02000050 53525150 57565554 5b5a5958 5f5e5d5c PQRSTUVWXYZ[\]^_
0x02000060 63626160 67666564 6b6a6968 6f6e6d6c `abcdefghijklmno
0x02000070 73727170 77767574 7b7a7978 7f7e7d7c pqrstuvwxyz{|}~.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agoSCRIPT: add add_script_search_dir cmd
Spencer Oliver [Wed, 17 Mar 2010 09:57:44 +0000 (09:57 +0000)]
SCRIPT: add add_script_search_dir cmd

Add a add_script_search_dir cmd so that adding search
dir's can be added to cfg scripts.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoPIC32: add software reset support
Spencer Oliver [Tue, 16 Mar 2010 12:54:08 +0000 (12:54 +0000)]
PIC32: add software reset support

The PIC32MX does not support the ejtag software reset - it is
optional in the ejtag spec.

We perform the equivalent using the microchip specific MTAP cmd's.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoMIPS: remove ejtag_srst variant
Spencer Oliver [Tue, 16 Mar 2010 12:48:53 +0000 (12:48 +0000)]
MIPS: remove ejtag_srst variant

The mips_m4k_assert_reset has now been restructured
so the variant ejtag_srst is not required anymore.
The ejtag software reset will be used if the target does not
have srst connected.

Remove ejtag_srst from docs.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agogdb_server: improved gdb load performance
Øyvind Harboe [Tue, 16 Mar 2010 13:45:07 +0000 (14:45 +0100)]
gdb_server: improved gdb load performance

by ack'ing memory writes immediately and reporting either
at next memory write or stepi/continue time. GDB will then
send off a new packet that is ready by the time the previous
packet has been written to target memory.

On faster adapters this can be as much as 10% improvement.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agozy1000: tweak the DCC inner loop a tiny bit
Øyvind Harboe [Tue, 16 Mar 2010 10:40:52 +0000 (11:40 +0100)]
zy1000: tweak the DCC inner loop a tiny bit

Uses FIFO a bit more efficiently now.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agoarm7/9: remove unused post_restore_context
Øyvind Harboe [Tue, 16 Mar 2010 09:58:21 +0000 (10:58 +0100)]
arm7/9: remove unused post_restore_context

Unused. If something should happen after context restore, then the
calling code can just do it afterwards.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agoADIv5 transport support moves to separate files
David Brownell [Tue, 16 Mar 2010 21:12:00 +0000 (14:12 -0700)]
ADIv5 transport support moves to separate files

Unclutter arm_adi_v5.c by moving most transport-specific code
to a transport-specific files adi_v5_{jtag,swd}.c ... it's not
a full cleanup, because of some issues which need to be addressed
as part of SWD support (along with implementing the DAP operations
on top of SWD transport):

 - The mess where mem_ap_read_buf_u32() is currently coded to
   know about JTAG scan chains, and thus needs rewriting before
   it will work with SWD;

 - Initialization is still JTAG-specific

Also  move JTAG_{DP,ACK}_* constants from adi_v5.h to the JTAG
file; no other code should care about those values.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agobitbang: add jtag_add_tms_seq support
Øyvind Harboe [Mon, 1 Mar 2010 07:25:18 +0000 (08:25 +0100)]
bitbang: add jtag_add_tms_seq support

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agoPARPORT: add PARPORTADDR tcl variable
Spencer Oliver [Tue, 16 Mar 2010 09:59:05 +0000 (09:59 +0000)]
PARPORT: add PARPORTADDR tcl variable

Add PARPORTADDR tcl variable making it easier to
change parallel port address in scripts.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoPIC32: add Microchip Explorer16 cfg
Spencer Oliver [Tue, 16 Mar 2010 09:55:20 +0000 (09:55 +0000)]
PIC32: add Microchip Explorer16 cfg

- add Microchip Explorer16 cfg using PIC32MX360F512L PIM.
 - remove reset config from PIC32 target cfg.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agolpc3180: LPC3180(LPC3250) SLC driver implemented
richard vegh [Tue, 16 Mar 2010 09:46:41 +0000 (10:46 +0100)]
lpc3180: LPC3180(LPC3250) SLC driver implemented

Until this time only basic  SLC functionality exists when you want to use SLC to access external nand flash.
Basic functionality can be selected with command:
     lpc3180 select 0 slc
It is anyway very slow to write/read to/from nand flash.

With the new command, SLC speed improved about 20 times, and hardware ECC info also read/written from/to nand flash OOB area:
     lpc3180 select 0 slc bulk
Speed improvement achieved by using working are in SRAM of the LPC3250 chip and controlling DMA controller to interact between SRAM and SLC peripheral.

Here are the patches, and if they are ok than take them.
Tested with hitex LPC3250 usb stick.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agoimage loading: fix problem with offsets > 0x80000000
Bradey Honsinger [Mon, 15 Mar 2010 07:43:41 +0000 (08:43 +0100)]
image loading: fix problem with offsets > 0x80000000

Fixes bug that prevented users from specifying a base address of
0x80000000 or higher in image commands (flash write_image, etm image,
xscale trace_image).

image.base_address is an offset from the start address contained in
the image file (if there is one), or from 0 (for binary files). As a
signed 32-bit int, it couldn't be greater than 0x7fffffff, which is a
problem when trying to write a binary file to flash above that
address. Changing it to a 64-bit long long keeps it as a signed
offset, but allows it to cover the entire 32-bit address space.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agomove "reset_config" out of JTAG command group
David Brownell [Mon, 15 Mar 2010 15:43:16 +0000 (08:43 -0700)]
move "reset_config" out of JTAG command group

The SRST configuration options are not specific to JTAG, so this
command may be needed with non-JTAG debug sessions.  Just move
the command to a different group.

(The TRST options are, however, clearly JTAG-specific, but for
compatibility, they're now left alone.  The flags they control
could later be disabled in non-JTAG sessions.)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agorename jtag_nsrst_assert_width as adapter_nsrst_assert_width
David Brownell [Mon, 15 Mar 2010 15:42:26 +0000 (08:42 -0700)]
rename jtag_nsrst_assert_width as adapter_nsrst_assert_width

Globally rename "jtag_nsrst_assert_width" as "adapter_nsrst_assert_width",
and move it out of the "jtag" command group ...  it needs to be used with
non-JTAG transports

Includes a migration aid (in jtag/startup.tcl) so that old user scripts
won't break.  That aid should Sunset in about a year.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agorename jtag_nsrst_delay as adapter_nsrst_delay
David Brownell [Mon, 15 Mar 2010 15:41:30 +0000 (08:41 -0700)]
rename jtag_nsrst_delay as adapter_nsrst_delay

Globally rename "jtag_nsrst_delay" as "adapter_nsrst_delay", and move it
out of the "jtag" command group ...  it needs to be used with non-JTAG
transports

Includes a migration aid (in jtag/startup.tcl) so that old user scripts
won't break.  That aid should Sunset in about a year.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agorename jtag_khz as adapter_khz
David Brownell [Mon, 15 Mar 2010 15:37:43 +0000 (08:37 -0700)]
rename jtag_khz as adapter_khz

Globally rename "jtag_khz" as "adapter_khz", and move it out of the "jtag"
command group ...  it needs to be used with non-JTAG transports

Includes a migration aid (in jtag/startup.tcl) so that old user scripts
won't break.  That aid should Sunset in about a year.  (We may want to
update it to include a nag message too.)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoFT2232: bulidfix
Spencer Oliver [Mon, 15 Mar 2010 10:06:47 +0000 (10:06 +0000)]
FT2232: bulidfix

Fix build issue with commit c23d4596d2239bdbba080499de837f53e0c89e59

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoPIC32MX: update cfg script
Spencer Oliver [Mon, 15 Mar 2010 09:36:46 +0000 (09:36 +0000)]
PIC32MX: update cfg script

The default config script will now dynamically setup the BMX registers
in the reset init script.
This will also work if the user overrides the default working area.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agorename jtag_interface_{init,quit}()
David Brownell [Sun, 14 Mar 2010 20:13:39 +0000 (13:13 -0700)]
rename jtag_interface_{init,quit}()

These routines apply to non-JTAG debug adapters too.  To
reduce confusion, give them better (non-misleading) names.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoFT2232: lookup and save layout just once
David Brownell [Sun, 14 Mar 2010 20:10:26 +0000 (13:10 -0700)]
FT2232: lookup and save layout just once

Streamline use of the layout:  have the "ft2232_layout" command
look it up and save the result, instead of having a few different
chunks of code looking it up later, and saving just its name (which
is already part of the layout).  This

   - is cleaner
   - reports errors sooner
   - facilitates earlier adapter-specific setup
   - removes unused "default to "usbjtag" logic

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agofix xscale icache and dcache commands
Mike Dunn [Fri, 12 Mar 2010 00:53:05 +0000 (16:53 -0800)]
fix xscale icache and dcache commands

Simple patch that fixes the broken xscale icache and dcache commands.
This broke when the helper functions and macros were changed.

[ dbrownell@users.sourceforge.net: don't use strcasecmp ]

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoversaloon cleanup patch
David Brownell [Thu, 11 Mar 2010 22:21:36 +0000 (14:21 -0800)]
versaloon cleanup patch

Remove undesirable
 - backslashes at end-of-line;
 - initializations of BSS data to zero/NULL;
 - overlong lines (80+ characters)
 - whitespace issues
 - brackets around single-line statements

And other minor issues reported by the Linux "checkpatch" utility

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoNew JTAG driver for Versaloon
simon qian [Thu, 11 Mar 2010 22:11:30 +0000 (14:11 -0800)]
New JTAG driver for Versaloon

This patch greatly simplifies the Versaloon driver:

 - reducing the code size from more than 50K to less than 28K
 - adding support for IR/DR scan with unlimited size
 - using tap_get_tms_path and tap_get_tms_path_len.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoAdd support for Bus Pirate as a JTAG adapter.
Michal Demin [Mon, 8 Mar 2010 12:45:14 +0000 (13:45 +0100)]
Add support for Bus Pirate as a JTAG adapter.

This includes a driver and matching config file.  This support needs to be
enabled through the initial "configure" (use "--enable-buspirate").

Signed-off-by: Michal Demin <michaldemin@gmail.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agosplit "interface" commands from "jtag" ones
David Brownell [Thu, 11 Mar 2010 17:47:47 +0000 (09:47 -0800)]
split "interface" commands from "jtag" ones

We'll need to be able to work with debug adapter interfaces (drivers)
even when they're not used for JTAG ... for example, while there are
multi-transport drivers which support JTAG *and* several other
transports (or just one more, like SWD) there are also adapters
with more limited goals (and no JTAG support at all).

Start decoupling the two concepts ("debug adapter driver", "jtag")
by having two command groups, which initialize separately.

This will help us support OpenOCD sessions using only non-JTAG
transports, in which JTAG commands should not be registered.
Update docs to mention that the JTAG, SVF, and XSVF commands
won't work without a JTAG transport.

Note that at least commands working with SRST are still inappropriately
coupled  to JTAG ... inappropriate because (a) SRST is not part of the
JTAG standard, for all that many platforms (like ARM) expect it; and also
(b) because they're used with non-JTAG debug and programming interfaces,
too.  They should perhaps become generic "interface" operations at some
point.  (Similarly with the clock rate to be used by a given adapter.)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoMIPS: make fixed code arrays static const
Spencer Oliver [Wed, 10 Mar 2010 22:23:01 +0000 (22:23 +0000)]
MIPS: make fixed code arrays static const

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoJLINK: user info message cleanup
Spencer Oliver [Wed, 10 Mar 2010 21:54:21 +0000 (21:54 +0000)]
JLINK: user info message cleanup

 - remove trailing LF's from user info messages.
 - split long lines.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoPIC32: add flash algorithm support
Spencer Oliver [Mon, 8 Mar 2010 22:54:18 +0000 (22:54 +0000)]
PIC32: add flash algorithm support

Add flash algorithm support for the PIC32MX.
Still a few things todo but this dramatically decreases
the programing time, eg. approx programming for 2.5k test file.
 - without fastload: 60secs
 - with fastload: 45secs
 - with fastload and algorithm: 2secs.

Add new devices to supported list.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoSTM32: flash loader cleanup
Spencer Oliver [Mon, 8 Mar 2010 20:32:11 +0000 (20:32 +0000)]
STM32: flash loader cleanup

- make algorithm array static const.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoSTR7: flash loader cleanup
Spencer Oliver [Mon, 8 Mar 2010 20:31:50 +0000 (20:31 +0000)]
STR7: flash loader cleanup

- make algorithm array static const.
 - increase algorithm buffer size to 32k.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoSTR9: flash loader cleanup
Spencer Oliver [Mon, 8 Mar 2010 20:31:25 +0000 (20:31 +0000)]
STR9: flash loader cleanup

- make algorithm array static const.
 - increase algorithm buffer size to 32k.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoADUC702x: flash loader cleanup
Spencer Oliver [Mon, 8 Mar 2010 20:30:53 +0000 (20:30 +0000)]
ADUC702x: flash loader cleanup

- make algorithm array static const.

Signed-off-by: Spencer Oliver <ntfreak@users.sourceforge.net>
14 years agoCFI CORE: bug-fix protect single sector
Antonio Borneo [Mon, 8 Mar 2010 16:43:59 +0000 (00:43 +0800)]
CFI CORE: bug-fix protect single sector

Cannot protect or unprotect single sector in cfi flash.
When first==last the procedure fails.

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agomove a constant table to .rodata section
David Brownell [Tue, 9 Mar 2010 03:10:05 +0000 (19:10 -0800)]
move a constant table to .rodata section

The table of command registration functions shouldn't be
in writable memory, where stray pointers can clobber it.
Also, it shouldn't be initialized at runtime; that just
consumes needless code space.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agocfi: simplify and clearify code
Antonio Borneo [Mon, 8 Mar 2010 17:31:27 +0000 (18:31 +0100)]
cfi: simplify and clearify code

At the end I have added comments /* FIXME: to be removed */
There are 3 lines in which my simplification is not complete due to
data dependency with LOG_DEBUG() messages visible in the patch.
Such log_debug has been introduced on Jan 22, 2007 with commit
4fc97d3f2726efa147cfdb0c456eace51550e1e3 during development activity
in this file/procedure.

From my point of view, these logs can be removed, since not part of a
consistent flow of information.
Alternatively, could be borrowed in the new cfi_send_command(), but
this will increase verbosity.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agozy1000: embedded ice dcc tweak
Øyvind Harboe [Mon, 8 Mar 2010 07:32:45 +0000 (08:32 +0100)]
zy1000: embedded ice dcc tweak

How many bits to shift out before/after enabled tap not
in bypass is calculated outside the loop. This is more of
a demonstration of principle and to clarify code than
a performance optimisation as such. Follows up a bit
on the simplification work in jtag interface.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag: simplify jtag_add_plain_ir/dr_scan
Øyvind Harboe [Sat, 6 Mar 2010 10:29:59 +0000 (11:29 +0100)]
jtag: simplify jtag_add_plain_ir/dr_scan

These fn's now clearly just clock out/in bits. No mystical
fields are involved.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag: jtag_add_ir_scan() now takes a single field
Øyvind Harboe [Thu, 4 Mar 2010 13:38:19 +0000 (14:38 +0100)]
jtag: jtag_add_ir_scan() now takes a single field

In the code a single field was all that was ever used. Makes
jtag_add_ir_scan() simpler and leaves more complicated stuff
to jtag_add_plain_ir_scan().

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agojtag: retire tap field
Øyvind Harboe [Mon, 1 Mar 2010 19:00:59 +0000 (20:00 +0100)]
jtag: retire tap field

jtag_add_dr/ir_scan() now takes the tap as the first
argument, rather than for each of the fields passed
in.

The code never exercised the path where there was
more than one tap being scanned, who knows if it even
worked.

This simplifies the implementation and reduces clutter
in the calling code.

use jtag_add_ir/dr_plain_scan() for more fancy situations.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agoCFI: review print of Voltage values
Antonio Borneo [Fri, 5 Mar 2010 16:56:36 +0000 (00:56 +0800)]
CFI: review print of Voltage values

JEDEC standard reports Vpp integer part encoded as 4 bit HEX value.
To print it using decimal digits, %u is required.
Other voltage values are coded as BCD, so %x is appropriate.

Code already prints one nibble at a time, so no need for field width
and precision in format string.

Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agodoc: not all debug adapters are "dongles"
David Brownell [Sat, 6 Mar 2010 05:09:03 +0000 (21:09 -0800)]
doc: not all debug adapters are "dongles"

Talk more about "debug adapters" instead of only "dongles".  Not all
adapters are discrete widgets; some are integrated onto boards.  If
we only talk about "dongles" we rule out many valid setups, and help
confuse some users (who may be using Dongle-free environments).

Also start bringing out the point that JTAG isn't the only transport
protocol, even though OpenOCD historically presumes "all is JTAG".
(Not all debug adapters are JTAG adapters, or JTAG-only adapters.)

Plus a few minor fixes (spelling etc) in the vicinity of those changes,
and updates about FT2232H clocking issues (they can go faster than the
older chips, and can support adaptive clocking).

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoREADME: update libftdi version
David Brownell [Fri, 5 Mar 2010 21:08:11 +0000 (13:08 -0800)]
README: update libftdi version

The FT2232H really wants libftdi 0.17 or newer; some notable
bugs got fixed in that version.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoADIv5 share DAP command support
David Brownell [Fri, 5 Mar 2010 18:39:25 +0000 (10:39 -0800)]
ADIv5 share DAP command support

Get rid of needless and undesirable code duplication for
all the DAP commands (resolving a FIXME) ... there's no
need for coreas to have private copies of that stuff.
Stick a pointer to the DAP in "struct arm", letting common
code get to it.

Also rename the "swjdp_info" symbol; just call it "dap".

This is an overall code shrink.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agominidriver: fix arm11 compilation problem
Øyvind Harboe [Thu, 4 Mar 2010 17:30:03 +0000 (18:30 +0100)]
minidriver: fix arm11 compilation problem

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
14 years agorename "swjdp_common" as "adiv5_dap"
David Brownell [Fri, 5 Mar 2010 05:51:58 +0000 (21:51 -0800)]
rename "swjdp_common" as "adiv5_dap"

This partially corrects an inappropriate name choice (and its
associated FIXME).

There are still too many variables named "swjdp", bug little
current code actually relies on them referencing an SWJ-DP instead
of some other flavor of DAP.  Only the two new dap_to{swd,jtag}()
calls could behave differently on an SWJ-DP instead of a SW-DP or
a JTAG-DP.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoNOR: trim range in flash_driver_protect()
David Brownell [Thu, 4 Mar 2010 05:01:16 +0000 (21:01 -0800)]
NOR: trim range in flash_driver_protect()

When the beginning or end of the specified range of sectors
already has the requested protection status, don't ask the
flash driver to change those sectors.

This will among other things turn command sequences like
this into the NOPs one would expect:

flash protect_check 0
flash info 0
... reports everything as unprotected ...
flash protect 0 0 1 off

That speeds things up (by whatever work was just avoided).

Also, with Stellaris (which can't unprotect flash at  page level)
this can eliminate some undesirable/false error reports.  (And
finishes fixing a bug currently listed in our bug database...)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
14 years agoNOR: invalidate cached state on target resume
David Brownell [Thu, 4 Mar 2010 04:57:49 +0000 (20:57 -0800)]
NOR: invalidate cached state on target resume

The NOR infrastructure caches some per-sector state, but
it's not used much ... because the cache is not trustworthy.

This patch addresses one part of that problem, by ensuring
that state cached by NOR drivers gets invalidated once we
resume the target -- since targets may then modify sectors.

Now if we see sector protection or erase status marked as
anything other than "unknown", we should be able to rely
on that as being accurate.  (That is ... if we assume the
drivers initialize and update this state correctly.)

Another part of that problem is that the cached state isn't
much used (being unreliable, it would have been unsafe).
Those issues can be addressed in later patches.

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