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.
Tim Newsome
Megan Wachs
Palmer Dabbelt