Re: [Patch BUGFIX] kcore: fix its wrong size on x86_64

[Tao Ma]

ebiederm@xxxxxxxxxxxx wrote:
> TaoMa <tao.ma@xxxxxxxxxx> writes:
>
> > ebiederm@xxxxxxxxxxxx wrote:
> > > Tao Ma <tao.ma@xxxxxxxxxx> writes:
> > >
> > >   
> > > > Hi Amerigo,
> > > >
> > > > The wrong number I mean is 131941393240064.
> > > >
> > > > So do you think
> > > > [root@test3 ~]# ls -l /proc/kcore
> > > > -r-------- 1 root root 131941393240064 Jun 15 13:39 /proc/kcore
> > > >
> > > > is better than
> > > >
> > > > [taoma@test2 ~]$ ll /proc/kcore
> > > > -r-------- 1 root root 281474974617600 Jun 15 15:20 /proc/kcore
> > > > ?
> > > >
> > > > I don't think so.
> > > >
> > > > Actually the right result should look like
> > > >
> > > > [root@test8 ~]# ls -l /proc/kcore
> > > > -r-------- 1 root root 5301604352 Jun 15 13:35 /proc/kcore
> > > >
> > > > And with your patch I can't get this number.
> > > >     
> > > Actually that value is the bug.  It has absolutely nothing
> > > to do with the offsets that are valid within /proc/kcore.
> > >
> > > Why do you prefer the smaller number?
> > >   
> > Amerigo said in the previous e-mail that " the man page for/proc/kcore is wrong,
> > its size can be more than the physical memory size, because it also contains
> > memory area of vmalloc(), vsyscall etc..."
> >
> > I have 4G memory, and 5301604352 is just a bit larger than 4G and looks sane. So
> > I misunderstand that this number is right.
>
> It should also include the 32 Tebibyte range we have for vmalloc.  So
> a completely dense encoding would be a bit larger than 35184372088832
> bytes.  You can see that range in your readelf -l output.
>
> Since the encoding is not dense the size actually comes to. 256TiB.
> Or roughly 281474976710656 bytes.
>
> > But if it is also a bug, I am willing to test any of the new patch. ;)
>
> Not in the sense that anything could go wrong.  Merely in the sense that
> we have a contradictory definition.  Which causes loads of confusion.
>
> I am wondering if this difference in definition has caused any
> problems applications to fail or if this just started out as an
> observation of an anomaly?
I first noticed it when my el5 box refused to start kdump service and 
kexec said something like "Can't find kernel text map area from kcore". 
And then I found this number which looked a bit strange.
I also just have another x86 box and "ls -l /proc/kcore" shows:
-r-------- 1 root root 939528192 Jun 16 10:01 /proc/kcore
So I thought this may be a bug and started this thread.

Anyway, later I found that kexec's problem isn't related to this issue.
So maybe we can leave as-is.

regards,
Tao
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/