comp.os.linux.development.system Google Group

Re: How to study debugger internals

2010-03-13T16:49:28Z

Have you looked at the information at
<[link]> ?
Please be warned that GDB is a pretty large and
complex program,

How to study debugger internals

2010-03-13T15:41:22Z

Excuse me if the post is NOT in scope of the community.
Want to study debuggers internals. Though I understand that they are
very much platform dependent, knowledge of internals for gdb and dbx
(for Intel/ARM) I believe should give me good understanding.
Please help me by telling the best way and if possible few links for

Re: semaphore, task control block, and kernel

2010-03-12T21:04:46Z

Such information does exist but only the kernel needs to know about it.
So it's all stored in the kernel data structures, not in user space.
[...]
Hey, I've been looking into this!
The answer is: it depends. Linux, for example, uses a low-level
lightweight synchronisation primitive called a futex. The glibc pthreads

Re: message passing: blocking non blocking

2010-03-12T20:57:42Z

That is likely incorrect. The correct definition of a blocking send is
"the sending process blocks until the message is 'sent', whatever that
means in context".
No idea, that seems confusing.
The point is this: either the implementation can take the data
immediately (such that the application no longer needs to do anything

message passing: blocking non blocking

2010-03-12T20:07:27Z

Hi,
i am confused about the definition of blocking, non blocking send and
receive operations with message passing. I have the definitions of the
textbook but it not that clear for me. let's exclude the case of
mailbox,. what do we really mean by blocking send? onereference says:
the sending process is blocked until the message is received . is that

semaphore, task control block, and kernel

2010-03-12T20:06:37Z

Hi,
It seems each semaphore has an internal queue of task control blocks
which correspond to the tasks waiting for the semaphore, obviously.
However, I thought since a multithreaded process might have many
semaphores, their locations / IDs should be stored in the main process
control block. I have not found any (see, [link]).

Re: Futex documentation

2010-03-12T00:32:37Z

[...]
For some bizarre reason that had never occurred to me, so, er, thanks.
But I *still* can't find anything resembling an actual spec for the
syscall API, but now I'm pretty much convinced no such thing exists.
Mailing list postings and tutorial articles are all very well, but
they're neither timely nor comprehensive --- I have no way of knowing

Re: objdump with detached symbols

2010-03-12T00:11:23Z

[...]
I'm trying to debug a third-party ELF loader that is somehow getting
something wrong in such a way that it's confusing ld-linux.so. I've
tried building it myself, but the glibc build system is hairy at the
best of times, and the rtld specific build rules are even hairier ---
I'm not confident of my ability to produce a binary with debug

Re: objdump with detached symbols

2010-03-11T23:08:24Z

In article <gk2m67-ujt....@marcumbill.bel lsouth.net>,
[...]
I think you've missed the point. Maybe you were temporarily blinded by that
obnoxious signature.
Anyway, the question was referring to the .gnu_debuglink feature, which
seems to be not quite working with objdump. The debugging symbols are
present in /usr/lib/debug

Re: root can't ls(1) a directory: Permission denied

2010-03-11T11:56:18Z

People who have used NFS have known that this can happen for the past
two decades and more. (-:
Go and read the kernel documentation for the FUSE filesystem, paying
particular attention to its discussion of non-privileged mounts.

Re: Futex documentation

2010-03-11T00:12:06Z

Google-fu on the blink? Try searching for these names. They are, after
all, the things that you want to find out about. (-:

Re: Futex documentation

2010-03-11T13:22:15Z

Something I found without much searching:
[link]
If you are really interested in FUTEX_WAIT_BITSET, why not search for
that?
[link]

Re: how linux maps system RAM to physical address

2010-03-11T07:08:31Z

Do you have memory-mapped peripherals which you want to drive
from user space?
IIRC, there are instructions in the Linux Device Drivers book.

Re: objdump with detached symbols

2010-03-11T02:54:40Z

Get the source code and compile it with debugging symbols. Is this just
for reference, or have you found an actual bug in the loader?

objdump with detached symbols

2010-03-11T00:54:55Z

Does anyone know how to persuade objdump to disassemble a file using
debugging information stored separately?
I have a binary --- /lib/ld-2.10.1.so, to be precise, the ELF dynamic
loader. I have the debugging symbols, stored in a different file ---
/usr/lib/debug/lib/ld-2.10.1.s o. I'm trying to produce an annotated