[Open SoC Debug] CDM: Next Steps
shivam16195 at iiitd.ac.in
Wed Jun 27 22:52:54 CEST 2018
I have worked on the following features of the OSD-GDB server:
1. Connection with GDB over TCP.
2. Send/Receive data to the client.
3. Send/Receive packet-data from the obtained buffer ($packet-data#checksum)
4. Validate the packet-data using checksum
I have opened a PR and will keep updating it with the rest of the features.
I tested these features with GDB by compiling the code completely
the OSD framework. As a workaround, I used the below code for testing.
All the features are working successfully.
Is it possible to test the gdbserver using unit-tests (for example,
1. To implement following use-cases/commands of the RSP protocol:
`?' -- last signal
`g' -- read registers
`G'XX... -- write regs
`m'ADDR`,'LENGTH -- read memory
`M'ADDR,LENGTH`:'XX... -- write mem
'c' -- continue
`z'TYPE`,'ADDR`,'LENGTH -- remove breakpoint or watchpoint (only, SW
breakpoint (for now))
`Z'TYPE`,'ADDR`,'LENGTH -- insert breakpoint or watchpoint (only SW
breakpoint (for now))
's' -- step (optional)
2. To implement signal-handling. As per my understanding, we only need
and SIGTRAP (for minimum support) to set software breakpoints, etc.
Please provide some insight for this part.
On Sun, Jun 24, 2018 at 12:37 AM, Philipp Wagner <philipp.wagner at tum.de>
> Am 23.06.2018 um 20:08 schrieb SHIVAM AGGARWAL:
> > I believe I can first ensure the complete implementation of GDBserver on
> > GDB side.
> > Please see below for details.
> Yes, sounds like a plan.
> > So, as per my understanding, I should keep the things simple and smooth
> > in the
> > beginning. This would ensure a rough yet working solution in hand. As
> > per your
> > proposed implementation details, I believe we can divide the code as
> > osd_gdbserver.c: This file should contain all the API covering
> > connection(),
> > set_breakpoint(), get_breakpoint(), fetch_data(), RSP specific
> > functions, etc.
> > Also, I should making a or1k specific file and keep the things minimal.
> That sounds roughly right, but we need to discuss that based on the code
> itself once it's there.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the OpenSoCDebug