Re: /proc/kcore has a unreasonable size(281474974617600) in x86_64 2.6.30-rc8.

[Andrew Morton <akpm@linux-foundation.org>]

On Fri, 5 Jun 2009 17:09:54 +0800 Am__rico Wang <xiyou.wangcong@gmail.com> wrote:

> On Fri, Jun 5, 2009 at 4:57 PM, Tao Ma<tao.ma@oracle.com> wrote:
> >
> >
> > Amerigo Wang wrote:
> >>
> >> On Fri, Jun 05, 2009 at 02:59:46PM +0800, Tao Ma wrote:
> >>>
> >>> Andrew Morton wrote:
> >>>>
> >>>> On Fri, 05 Jun 2009 12:03:52 +0800 Tao Ma <tao.ma@oracle.com> wrote:
> >>>>
> >>>>> Hi list,
> >>>>> __ __ __ In 2.6.30-rc8, /proc/kcore in x86_64's size is unreasonable large
> >>>>> to be 281474974617600.
> >>>>> While in a x86 box, it is 931131392 which looks sane.
> >>>>>
> >>>>> [root@test8 ~]# ll /proc/kcore
> >>>>> -r-------- 1 root root 281474974617600 Jun __5 11:15 /proc/kcore
> >>>>>
> >>>>> [root@ocfs2-test9 ~]$ ll /proc/kcore
> >>>>> -r-------- 1 root root 931131392 Jun __5 11:58 /proc/kcore
> >>>>>
> >>>>> I just noticed this when kexec fails in "Can't find kernel text map
> >>>>> area from kcore".
> >>>>>
> >>>>> Is there something wrong?
> >>>>>
> >>>> fs/proc/kcore.c hasn't changed since October last year. __Was 2.6.29 OK?
> >>>> Earlier kernels?
> >>>
> >>> with 2.6.29, ls shows the same output.
> >>> [root@test8 ~]# ll /proc/kcore
> >>> -r-------- 1 root root 281474974617600 Jun __5 14:35 /proc/kcore
> >>
> >>
> >> Thanks.
> >>
> >> It looks like the value of 'high_memory' is insane..
> >> Can you get its value on your machine? You can add a printk() or use
> >> systemtap etc..
> >
> > Just did that.
> > Also a strange number.
> > high memory 18446612137615818752.
> >

(top-posting repaired)

> Add some Cc: to x86 people. :)
> 
> Yinghai?
> 

Please send the boot logs: dmesg -s 1000000 > foo