tech-toolchain archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Aprilis LLDB/NetBSD



On 05.04.2017 02:21, Christos Zoulas wrote:
> On Apr 5,  1:10am, n54%gmx.com@localhost (Kamil Rytarowski) wrote:
> -- Subject: Re: Aprilis LLDB/NetBSD
> 
> | Currently there is nothing planned on the base system side. All the
> | enhancements are to be done on the LLDB side.
> | 
> | Our core-dump files contain notes like: "NetBSD-CORE@1". This "1" is LWP
> | and it must be properly handled in LLDB to define a thread. Currently
> | LLDB just recognizes NetBSD-CORE files and "NetBSD-CORE" notes. It needs
> | now to instantiate appropriate threads.
> | 
> | On the other hand I've studied inspiring articles like:
> | "FreeBSD Userspace Coredumps"
> | https://backtrace.io/blog/blog/2015/10/03/whats-a-coredump/index.html
> | 
> | Presenting a list of improvements for the NetBSD system is currently
> | premature for me as I need to study up more on the topic. It appears
> | also beyond the tasks related to LLDB.
> | 
> | In other words we seem to be good enough to get all the needed things done.=
> 
> Ok, so nothing so far to do.
> 

Well, we might want to build some interfaces like one to pass
siginfo_t.. however I don't think this is that important for core(5)
files as of now to get them just functional. There is already
infrastructure in the kernel for this (as far as I remember we pass it
in p_sigctx) that could be reused in future.

There is certainly room for improvement, but nothing stops us.

> | > Sounds good, do you have the PT_{G,S}ETSTEP functional description somewh=
> | ere?
> | >=20
> | 
> | PT_SETSTEP This request will turn on single stepping of the specified
> | process.
> | 
> | PT_CLEARSTEP This request will turn off single stepping of the specified
> | process.
> | 
> | In NetBSD case:
> |  ptrace(PT_SETSTEP, pid, NULL, lwp);
> |  ptrace(PT_CLEARSTEP, pid, NULL, lwp);
> | 
> | I'm going to cleanup it and send for review.
> 
> Yes, but as PT_STEP is things going to be an MD thing? How will it be
> implemented?
> 

Yes, I planned to add it as MD.

In the source code protect it with
#ifdef PT_STEP
PT_STEP:
PT_SETSTEP:
PT_CLEARSTEP:
#endif

I will show a patch tomorrow.

> christos
> 


Attachment: signature.asc
Description: OpenPGP digital signature



Home | Main Index | Thread Index | Old Index