question about /proc/kcore

[Jamie Levy]

Hello,

I know this list has been rather quiet for a while...  so I might as
well ask a question :-)

I've been trying to find some information on /proc/kcore and have not
been entirely successful.  I know is that it is an alias to memory,
that is in ELF format, and that can be used with gdb to debug the
kernel.  The `k' seems to imply that it is the kernel's memory, but
all documentation I've found states that it is the size of physical
memory plus 4KB (example below):

https://www.redhat.com/docs/manuals/enterprise/RHEL-4-Manual/en-US/Reference...

This does not seem to be the case, however, when I do a listing in
/proc/kcore on my machine (F8) which has 2 GB of memory:

ls -lh /proc/kcore
-r-------- 1 root root 897M 2008-08-01 19:05 /proc/kcore

I have confirmed the same type of results on other machines to which I
have access.  I also do not seem to have the kcore.h file on my system
(and I have kernel-devel installed), since it is a zero byte file:

-rw-r--r-- 1 root root 0 2008-07-21 02:05
/usr/src/kernels/2.6.25.11-60.fc8-i686/include/config/proc/kcore.h

I am also aware of other tools that can be used to extract information
about kcore like the redhat crash utility:

http://people.redhat.com/anderson/

I am wondering why there is a discrepancy between the documentation
and the actual implementation of the /proc/kcore file?  I would also
like to know where I can find more detailed information about the
kcore file besides the usual description of what it is and how to use
it for debugging.

All the best,

-Jamie Levy

[Astarta]

Jamie Levy wrote:
> Hello,
>
> I know this list has been rather quiet for a while...  so I might as
> well ask a question :-)

Hello,

I really thought that this list is completely dead :-)
> This does not seem to be the case, however, when I do a listing in
> /proc/kcore on my machine (F8) which has 2 GB of memory:
>
> ls -lh /proc/kcore
> -r-------- 1 root root 897M 2008-08-01 19:05 /proc/kcore
>
> I have confirmed the same type of results on other machines to which I
> have access.  I also do not seem to have the kcore.h file on my system
> (and I have kernel-devel installed), since it is a zero byte file:
>
> -rw-r--r-- 1 root root 0 2008-07-21 02:05
> /usr/src/kernels/2.6.25.11-60.fc8-i686/include/config/proc/kcore.h

As far as I know, the size of kcore is equal not to the size of _all_ 
physical memory installed on the machine, but just to the amount of 
memory accessible by kernel.
Your kernel seems to be compiled with HIGHMEM support, and if you do `ls 
-l /proc/kcore` you see your LOWMEM size, the real memory that the 
kernel uses. That the reason why the difference you mentioned comes.

Someone will correct me, if I'm wrong :)
> All the best,
>
> -Jamie Levy
>
> Fedora-women-list mailing list
> [e-mail]
> https://www.redhat.com/mailman/listinfo/fedora-women-list

[Gayathri Swaminathan]

Think Jamie is right.

I have observed behaviour were this file grows based on accessed memory
(would this be kernel shared mem params?)

But it is definitely not just size of physical RAM+4KB on a machine.

gayathri

On Sat, Aug 2, 2008 at 3:54 PM, Astarta  wrote:
> Jamie Levy wrote:
> > Hello,
> >
> > I know this list has been rather quiet for a while...  so I might as
> > well ask a question :-)
>
> Hello,
>
> I really thought that this list is completely dead :-)
> > This does not seem to be the case, however, when I do a listing in
> > /proc/kcore on my machine (F8) which has 2 GB of memory:
> >
> > ls -lh /proc/kcore
> > -r-------- 1 root root 897M 2008-08-01 19:05 /proc/kcore
> >
> > I have confirmed the same type of results on other machines to which I
> > have access.  I also do not seem to have the kcore.h file on my system
> > (and I have kernel-devel installed), since it is a zero byte file:
> >
> > -rw-r--r-- 1 root root 0 2008-07-21 02:05
> > /usr/src/kernels/2.6.25.11-60.fc8-i686/include/config/proc/kcore.h
>
> As far as I know, the size of kcore is equal not to the size of _all_
> physical memory installed on the machine, but just to the amount of memory
> accessible by kernel.
> Your kernel seems to be compiled with HIGHMEM support, and if you do `ls -l
> /proc/kcore` you see your LOWMEM size, the real memory that the kernel uses.
> That the reason why the difference you mentioned comes.
>
> Someone will correct me, if I'm wrong :)
> > All the best,
> >
> > -Jamie Levy
> >
> > Fedora-women-list mailing list
> > [e-mail]
> > https://www.redhat.com/mailman/listinfo/fedora-women-list
>
> --
> with best regards,
> Fadeeva Marina.
>
> Fedora-women-list mailing list
> [e-mail]
> https://www.redhat.com/mailman/listinfo/fedora-women-list

[Jamie Levy]

Thank you both for your responses.  I still have not figured this out,
but I appreciate your replies.  I wasn't sure if this list was still
around, but I thought I would try to liven it up with a question ;-)

All the best,

-Jamie

On Sat, Aug 2, 2008 at 5:58 PM, Gayathri Swaminathan
 wrote:
> Think Jamie is right.
>
> I have observed behaviour were this file grows based on accessed memory
> (would this be kernel shared mem params?)
>
> But it is definitely not just size of physical RAM+4KB on a machine.
>
> gayathri
>
> On Sat, Aug 2, 2008 at 3:54 PM, Astarta  wrote:
> > Jamie Levy wrote:
> > > Hello,
> > >
> > > I know this list has been rather quiet for a while...  so I might as
> > > well ask a question :-)
> >
> > Hello,
> >
> > I really thought that this list is completely dead :-)
> > > This does not seem to be the case, however, when I do a listing in
> > > /proc/kcore on my machine (F8) which has 2 GB of memory:
> > >
> > > ls -lh /proc/kcore
> > > -r-------- 1 root root 897M 2008-08-01 19:05 /proc/kcore
> > >
> > > I have confirmed the same type of results on other machines to which I
> > > have access.  I also do not seem to have the kcore.h file on my system
> > > (and I have kernel-devel installed), since it is a zero byte file:
> > >
> > > -rw-r--r-- 1 root root 0 2008-07-21 02:05
> > > /usr/src/kernels/2.6.25.11-60.fc8-i686/include/config/proc/kcore.h
> >
> > As far as I know, the size of kcore is equal not to the size of _all_
> > physical memory installed on the machine, but just to the amount of memory
> > accessible by kernel.
> > Your kernel seems to be compiled with HIGHMEM support, and if you do `ls
> > -l /proc/kcore` you see your LOWMEM size, the real memory that the kernel
> > uses. That the reason why the difference you mentioned comes.
> >
> > Someone will correct me, if I'm wrong :)
> > > All the best,
> > >
> > > -Jamie Levy
> > >
> > > Fedora-women-list mailing list
> > > [e-mail]
> > > https://www.redhat.com/mailman/listinfo/fedora-women-list
> >
> > --
> > with best regards,
> > Fadeeva Marina.
> >
> > Fedora-women-list mailing list
> > [e-mail]
> > https://www.redhat.com/mailman/listinfo/fedora-women-list
>
> --
> Gayathri Swaminathan
> gpgkey: 3EFB3D39
> Volunteer, FDP

[Gayathri Swaminathan]

And, thanks for bringing it to life!

There are many RHEL-ish issues/resolutions that I encounter as well.

Will start sharing them here on :)

On Sun, Aug 3, 2008 at 9:31 PM, Jamie Levy  wrote:
> Thank you both for your responses.  I still have not figured this out,
> but I appreciate your replies.  I wasn't sure if this list was still
> around, but I thought I would try to liven it up with a question ;-)
>
> All the best,
>
> -Jamie