[debug] Support mcontrol and mcontrol6 based on HwbpManual #541
Conversation
lz-bro
force-pushed
the
enable_HwbpManual
branch
from
e880b5b to
769f08a
Compare
debug/gdbserver.py
Outdated
| self.gdb.p("$pc") | ||
| assertRegex(output, r"[bB]reakpoint") | ||
| assertIn("rot13 ", output) | ||
| self.gdb.stepi() |
There was a problem hiding this comment.
riscv-tests/debug/programs/debug.c
Lines 65 to 68 in 7dfa49b
According to the program, I think that the breakpoint should be hit 2 times at rot13, and then hit at crc32a and _exit sequentially. So stepi avoids hitting the rot13 breakpoint at the same position as main
There was a problem hiding this comment.
So stepi avoids hitting the rot13 breakpoint at the same position as main
I don't get it. What do you mean "at the same position as main" ? main and rot13 should have different addresses. When gdb issues a continue request it issues a single-step and then issues a real continue. The breakpoint at main should not be hit.
debug/gdbserver.py
Outdated
| if value == tdata1: | ||
| value = self.gdb.p("$tdata1") | ||
| if value == tdata1 or \ | ||
| (value >> (xlen-4) == 0x6 and \ |
There was a problem hiding this comment.
value >> (xlen-4) == 0x6 What is the purpose of this condition? If I read this right we explicitly ask for TYPE=2 trigger. How come we end up with TYPE=6?
There was a problem hiding this comment.
The purpose is to be compatible with spec 1.0 and 0.13. On our hardware that supports spec 0.13, setting the TYPE to 2 trigger actually we end up with TYPE=6.
There was a problem hiding this comment.
can we instead just re-request type==6 if the resulting value of tdata1 is different? I believe this would enable us to support wide range of platforms in a more cleaner way.
lz-bro
force-pushed
the
enable_HwbpManual
branch
2 times, most recently
from
4f5c95e to
785eee0
Compare
lz-bro
force-pushed
the
enable_HwbpManual
branch
from
785eee0 to
0f3a832
Compare
|
@aap-sc hello，I have made some modifications to support wide range of platforms. Could you please review it. |
Thanks, I agree that perform 2 separate calls (for type2 and type6) |
lz-bro
force-pushed
the
enable_HwbpManual
branch
from
0f3a832 to
c7d1719
Compare
debug/gdbserver.py
Outdated
| type_none = set_field(0, TDATA1_TYPE(self.hart.xlen), | ||
| TDATA1_TYPE_NONE) | ||
| if type_rb == type_match: | ||
| maskmax_rb = tdata1_rb & MCONTROL_MASKMAX(self.hart.xlen) |
There was a problem hiding this comment.
I'd sugest passing the mask to use before the comparison of tdata1 and tdata1_rb as an argument of set_manual_trigger()
lz-bro
force-pushed
the
enable_HwbpManual
branch
6 times, most recently
from
49f4f2e to
1cd0669
Compare
|
Changes to resolve the following issues:
@en-sc could you please take a look? |
lz-bro
force-pushed
the
enable_HwbpManual
branch
from
1cd0669 to
e7c4576
Compare
lz-bro
changed the title
* support mcontrol6 in HwbpManual * test non-zero maskmax is supported for mcontrol
lz-bro
force-pushed
the
enable_HwbpManual
branch
from
e7c4576 to
72af2b2
Compare
|
I want to ignore the [Line too long (81/80) ](gdbserver.py:854:0: C0301: Line too long (81/80) (line-too-long)) |
3 participants
Change-Id: I2ee8718f03d6e381369b7d3bb0222563789ed525