Tested with SiFive HiFive1 development board.
Change-Id: I96a9a528057fcf9fc54d3da46a672d2cd54c3d5f
Signed-off-by: Marc Schink <openocd-dev@marcschink.de>
Reviewed-on: http://openocd.zylin.com/4885
Tested-by: jenkins
Reviewed-by: Tim Newsome <tim@sifive.com>
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
While initializing struct command_registration, the field's name "name"
is not specified, thus relying on the fact that it is the first field
declared in the struct and it's initialization value can be listed as
the first one.
Be coherent in the struct initialization and always use the field's
name.
Change-Id: Iefaeb15cc051db9f1e0f0140fe2f231b45f5bb12
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: http://openocd.zylin.com/5013
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Tested-by: jenkins
Reviewed-by: Tim Newsome <tim@sifive.com>
* Remove unnecessary 0.11 code.
Don't need need_strict_step anymore now that we have
riscv_hit_watchpoint().
Don't need 32-bit warning in riscv011_resume() now that address is a
target_address_t.
Change-Id: I375c023a7ec9f62d80b037ddb64d14526ba0a3dc
* WIP little refactor working towards hasel support.
Change-Id: Ie0b8dfd9e5ae2e36613fa00e14c3cd32749141bf
* More refactoring.
Change-Id: I083387c2ecff78ddfea3ed5078444732d77b909b
* More refactoring.
Change-Id: Icea1308499492da51354f89e1529353e8385f3a1
* Starting to work towards actual hasel changes.
Change-Id: If0df05ffa66cc58400b4855f9630a8b1bae3030e
* Implement simultaneous resume using hasel.
Change-Id: I97971d7564fdb159d2052393c8b82a2ffaa8833f
* Add support back for targets that don't have hasel.
Change-Id: I6d5439f0615d5d5333127d280e4f2642649a119a
* Make hasel work with >32 harts.
Change-Id: I3c55009d48bfc5dd62e3341df4e4bd21df2fe44f
This represents months of continuing RISC-V work, with too many changes
to list individually. Some improvements:
* Fixed memory leaks.
* Better handling of dbus timeouts.
* Add `riscv expose_custom` command.
* Somewhat deal with cache coherency.
* Deal with more timeouts during block memory accesses.
* Basic debug compliance test.
* Tell gdb which watchpoint hit.
* SMP support for use with -rtos hwthread
* Add `riscv set_ir`
Change-Id: Ica507ee2a57eaf51b578ab1d9b7de71512fdf47f
Signed-off-by: Tim Newsome <tim@sifive.com>
Reviewed-on: http://openocd.zylin.com/4922
Tested-by: jenkins
Reviewed-by: Philipp Guehring <pg@futureware.at>
Reviewed-by: Liviu Ionescu <ilg@livius.net>
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
Targets can use this to expose how many address bits there are.
gdb_server uses this to send gdb the appropriate upper limit in the
memory-map. (Before this change the upper limit would only be correct
for 32-bit targets.)
Change-Id: Idb0933255ed53951fcfb05e040674bcdf19441e1
Signed-off-by: Tim Newsome <tim@sifive.com>
Reviewed-on: http://openocd.zylin.com/4947
Tested-by: jenkins
Reviewed-by: Peter Mamonov <pmamonov@gmail.com>
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
* Implement riscv_get_thread_reg().
This is necessary because riscv_get_gdb_reg_list() now reads all
registers, which ended up causing `-rtos riscv` to read all registers
whenever one was requested (because the register cache is wiped every
time we switch to a different hart).
CustomRegisterTest went from 1329s to 106s.
Change-Id: I8e9918b7a532d44bca927f67aae5ac34954a8d32
* Also implement riscv_set_reg().
Now all the `-rtos riscv` tests pass again, at regular speed.
Change-Id: I55164224672d9dcc9eb4d1184b47258ff3c2cff1
* Better error messages.
Change-Id: I4125f9a54750d9d0ee22c4fa84b9dd3f5af203f5
* Add target_get_gdb_reg_list_noread().
Being explicit about what's expected gets `-rtos riscv` back to `-rtos
hwthread` time.
Change-Id: I6e57390c2fe79b5e6799bfda980d89697e2e29f7
* Revert a change I made that has no effect.
I don't understand exactly what all this test protects against, and I
shouldn't change it unless I do.
Change-Id: Ib329d4e34d65d2b38559b89b7afb3678f439ad2c
Without this change xxx_start_algorithm() writes all register
parameters no matter of their direction. It usually results
in writing of uninitialized reg_params[].value - possibly
reported by valgrind.
While on it fix the wrong parameter direction in
kinetis_disable_wdog_algo(). This bug did not have any
impact because of unconditional write of reg_params.
Change-Id: Ia9c6a7b37f77d5eb6e5f5463012dddd50471742b
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: http://openocd.zylin.com/4813
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
Previously the code made the assumption (which is valid for conventional
RTOSs) that special registers (e.g. CSRs) are the same across threads.
26/45 tests pass.
Change-Id: Ibb3398790d7354a995d506772375d869f608f1f0
* Changed logging level
* Added logging statement
* Removed halt event when attaching to target
* Extended some packet handling
* Extended handling of rtos_hart_id and clearing of register cache
* Extended execute_fence to handle all harts
* Removing logging statement again
* Updated according to review comments
* Forgot to re-add the return statement
* Was removing too much for the if statement to work
* This needs to >= 3 now to handle both a fence and a fence.i
* Added `riscv expose_custom` command.
Seems to work for reading. I need to do some more testing for writes, as
well as minor cleanup.
Change-Id: I85d5d00897d5da4add4a6643b538be37d31a016f
* Conform to OpenOCD style.
Change-Id: I40a316f06f418d2b63d9e11aea03ef51da8d8faf
* Free all the memory allocated by register init.
Change-Id: I04e35ab54613f99708cee85e41fef989079adefc
* Properly document `riscv expose_custom`.
Change-Id: Id78a02b7a00c161df80f11b521a306e0cf3d7478
* Add riscv_hit_watchpoint function for RV32I loads and stores
For GDB to fully support hardware watchpoints, OpenOCD needs to tell GDB
which data address has been hit. OpenOCD relies on a target-specific
hit_watchpoint function to do this. If GDB is not given the address, it
will not print the hit variable name or its old and new value.
There does not seem to be a way for the hardware to tell us which trigger
was hit (0.13 introduced the 'hit bit' but this is optional). Alternatively,
we can decode the instruction at dpc and find out which memory address
it accesses.
This commit adds support for RV32I load and store instructions
and could be extended for additional instructions in the future.
* 0.11: change debug reason for hw triggers to DBG_REASON_WATCHPOINT
This is to make sure riscv_hit_watchpoint is called to check for a data
address hit.
* Fix style issues
* Change %lx to PRIx64 to clear -m32 build errors
* Add clarifying comments/todos
* Fix types in format strings
This supports both 0.11 and 0.13 versions of the debug spec.
Support for `-rtos riscv` will come in a separate commit since it was
easy to separate out, and is likely to be more controversial.
Flash support for the SiFive boards will also come in a later commit.
Change-Id: I1d38fe669c2041b4e21a5c54a091594aac3e2190
Signed-off-by: Tim Newsome <tim@sifive.com>
Reviewed-on: http://openocd.zylin.com/4578
Tested-by: jenkins
Reviewed-by: Liviu Ionescu <ilg@livius.net>
Reviewed-by: Matthias Welwarsky <matthias@welwarsky.de>
This improves startup time, which is important when connecting to
simulators. One problem is that triggers that are set when the debugger
connects are not cleared until enumeration happens. Execution may halt
due to a trigger set by a previous debug session, which could confuse
the user. If this happens, triggers will be instantly enumerated, so it
will only happen once per session.
Change-Id: I3396f713f16980a8b74745a1672fe8b8a2d4abae
Just remove our nop implementation. The default behavior when this is
left NULL does the same thing.
Change-Id: I865976c694d24661941584cb0efc92fc26612316
It's not implemented for 0.11 because we don't need it. Returning error
caused 0.11 targets to not be debuggable since change
848062d0d1.
Change-Id: I8b04a1fcf3c3e8bf8340cbf39aaf475d2a213519
* Enforce OpenOCD style guide.
Change-Id: I579a9f54ed22a774bf52f6aa5bc13bcbd2e82cd8
* Fail if `git diff` fails
Change-Id: I57256b0a24247f6123cb0e25a89c1b59867cb3f9
* Maybe every line gets its own shell?
Change-Id: I1a6f83e9f3d7cfd39f8933f0dba13c3cf76f71f6
* Maybe this will error properly.
Change-Id: I50803cfc229e61158569fb6b609195f7191ecac9
* Take different approach than merge-base
Change-Id: I345cbc4eecc4755c7127e8e36e403f7b727010b1
* Fix style issues.
Change-Id: I90e71f710858524812d0ab160b25c486b7b099e7
This works around some side effects of the -rtos hack, namely that we
were unable to set hardware breakpoints on harts whose misa differed
from the first one. There may be other bugs like this one lurking
elsewhere. The only proper solution is for gdb to have a better user
interface when talking to a server that exposes multiple targets, but
that's a very big project.
This fixes#194.
Change-Id: I81aedddeaa922d220e936730e9c731545953ae21
This replaces the earlier mechanism which would propagate errors only
for targets that decided they wanted to. It was suggested by Matthias
Welwarsky from the OpenOCD team.
Change-Id: Ibe8e97644abb47aff26d74b8280377d42615a4d3
This allows a user to tell OpenOCD to prefer system bus access for
memory access, which can be useful for testing, or when there really is
a difference in behavior.
Change-Id: I8c2f15b89a2ccdae568c68ee743b75a74f9ad6bd
Mostly addresses #207.
Also changed dmi_read() to return an error, and fixed all the call sites
to propagate that error if possible.
Change-Id: Ie6fd1f9e7eb46ff92cdb5021a7311ea7334904f1
They can be used to authenticate to a Debug Module.
There's a bit of a chicken and egg problem here, because the RISCV
commands aren't available until the target is initialized, but
initialization involves examine(), which can't interact with the target
until authentication has happened. So to use this you run `init`, which
will print out an error, and then run the `riscv authdata_read` and
`riscv authdata_write` commands. When authdata_write() notices that the
authenticated bit went high, it will call examine() again.
Example usage (very simple challenge-response protocol):
```
init
set challenge [ocd_riscv authdata_read]
riscv authdata_write [expr $challenge + 1]
reset halt
```
Change-Id: Id9ead00a7eca111e5ec879c4af4586c30af51f4d
... by disabling all triggers, single stepping, enabling them, and then
resuming as usual. Without this change, you'd just be stuck on an
address trigger and would have to manually disable it.
Change-Id: I5834984671baa6b64f72e533c4aa94555c64617e
Fixed abstract register access for registers that aren't XLEN wide.
Avoided excessive errors cases where we attempted to execute a fence but
failed.
Don't mark all the CSRs as caller-save. gdb was saving/restoring
dscratch, which broke function calls as a side effect. dscratch is
accessible for people who really know what they're doing, but gdb should
never quietly access it. The same is probably true for other CSRs.
Change-Id: I7bcdbbcb7e3c22ad92cbc205bf537c1fe548b160
This avoids trying to read memory from the wrong hart, if the current
hart was changed by an earlier call (eg. to poll()).
Change-Id: I73da1e01c8d01d68f01ac7fdd6c548380a70cfd3
This lets users tell OpenOCD which non-standard CSRs exist on their
target, that will also be accessible and whose existence will be
communicated to gdb.
Change-Id: I56163a9fcb84ad7ebe815ae74fbd9fcc208f5a9d
(It's really only 2 bits, but something wonky happens between gdb and
OpenOCD if I make it that size.)
Change-Id: I562a65cb0ebe5aa0edcc54c251d0fea0e26f9cb1
Because there is no instruction that moves just half of a 64-bit FPR
to/from a GPR, we need to use scratch memory for this operation. This
code can theoretically use:
1. DMI_DATA, if it is memory mapped in the target.
2. DMI_PROGBUF, if it is writable in the target.
3. A user-configured address.
I have only tested this code very lightly. One reason is that gdb thinks
that on RV32 harts every register is 32 bits wide. Another is that this
is mostly proof-of-concept to satisfy the small program buffer code
review, which I don't want to drag out forever.
Existing tests don't realize that floating support was broken with
RV32D, and don't realize that it still doesn't work because of the gdb
problem mentioned above.
This change improves Issue #110 but there's more work to be done.
Change-Id: I99b8a36e5fea26f1d9e16e36cf99adc7be26b944
The actual implementation of triggers didn't change between those two
versions, so there's no need to duplicate the code.
In the process, I also fixed a minor multicore bug where tselect didn't
always get written on all harts.
When first connecting to a target, have the debugger disable any
hardware triggers that are set by a previously connected debugger.
The 0.11 code already did this, but 0.13 did not.
To achieve this I decided to share the code to enumerate triggers
between 0.11 and 0.13, which required me to implement get_register() and
set_register() for 0.11, which made the whole change a lot larger than
you might have guessed.
Hopefully this sets us up to in the future share the code to set/remove
triggers as well.
Rather than having a bunch of "if rtos" stuff, I now just check "if
hart_enabled". This makes some code paths cleaner, all of which were
buggy in the non-RTOS multi-hart mode.
When I disappeared the polls everywhere I forgot to sanitize the hartid
after halting. This is an invariant that GDB expects: when you return
from a halt whatever thread is marked as currently selected is the
thread that the next register accesses reference.
Main change is to make riscv_addr_t be unsigned. The rest is mechanical
fixing of types, print statements, and a few signed/unsigned compares.
Smoketest indicates everything is working more or less as before.
This means I don't know what hart to look at, so I might as well
invalidate the register cache. Without this, you might get stale
registers the first time you ask for them.
I thought OpenOCD did this, but it looks like that doesn't happen when
runningi in RTOS mode. With this I can get to the end of most of the
RTOS tests, but they SIGINT instead of exiting.
This is a major rewrite of the RISC-V v0.13 OpenOCD port. This
shouldn't have any meaningful effect on the v0.11 support, but it does
add generic versions of many functions that will allow me to later
refactor the v0.11 support so it's easier to maintain both ports. This
started as an emergency feature branch and went on for a long time, so
it's all been squashed down into one commit so there isn't a big set of
broken commits lying around. The changes are:
* You can pass "-rtos riscv" to the target in OpenOCD's configuration
file, which enables multi-hart mode. This uses OpenOCD's RTOS
support to control all the harts from the debug module using commands
like "info threads" in GDB. This support is still expermental.
* There is support for RV64I, but due to OpenOCD limitations we only
support 32-bit physical addresses. I hope to remedy this by rebasing
onto the latest OpenOCD release, which I've heard should fix this.
* This matches the latest draft version of the RISC-V debug spec, as of
April 26th. This version fixes a number of spec bugs and should be
close to the final debug spec.
If the working area is large enough, every fespi_write() results in just
a single algorithm execution.
Change-Id: I87a12e29f50ef6ea1f46fbd1edf440f9e54a2162
Halting didn't work right in slow targets, because some code assumed the
register cache is valid before it was guaranteed to be.
Also made dbus_busy_delay and interrupt_high_delay grow faster, so that
on slow targets it takes less time to learn the correct values.
Change-Id: I948a49d4e3cd0638f5449ab94994406319fd5f42
It's not a failure in the debugger or even a real problem if a user asks
to access memory that's not accessible.
Change-Id: I30b8424d5265d1996fe4826012ed160a83f0bc6c
Also only do work for debug RAM that actually exists on the target
(exposing the off-by-one error on 32-bit targets).
Change-Id: I37e0005b6a70e949286f1d6494716f3abea03c12
Read dtmcontrol's idle field to decide how many run-test/idle cycles are
required to communicate with the target.
In riscv_examine(), write and read all of Debug RAM to check the target
is at least somewhat sane.
Change-Id: Ieedb7a50418fa1f5e0d456cde53c52f7fc51662b
Old code would write 64 bytes of DRAM if the dbus was busy in
cache_write().
New code clears the dbus error condition when the bus is busy. (This
part is untested.)
Change-Id: Ia396fe819fa1828bb75726d85513b113cc9e13f0
Now logging is consistent and more readable.
I did remove most logging during riscv_poll() since it clutters up the
log/screen and is not generally helpful.
Users can use this register to inspect and change the privilege level of
the core. It doesn't make any assumptions about the actual underlying
debug mechanism (as opposed to having the user change DCSR directly,
which may not exist in all debug implementations).
Now you can attach with gdb, and it'll attempt to read a register. That
will fail because the core won't clear debug interrupt. Adding nops
doesn't help this time.
Tim Newsome
Greg Savin
Marc Schink
Antonio Borneo
Tomas Vanek
Darius Rad
Pavel S. Smirnov
Carsten Gosvig
Megan Wachs
craigblackmore
Dmitry Ryzhov
Liviu Ionescu
Dan Robertson
Ryan Macdonald
Palmer Dabbelt