Re: [ANNOUNCE] crash gcore command, version 1.1 available

[HATAYAMA Daisuke <d.hatayama@xxxxxxxxxxxxxx>]

Hello Lei,

From: Lei Wen <leiwen@xxxxxxxxxxx>
Subject: RE:  [ANNOUNCE] crash gcore command, version 1.1 available
Date: Fri, 23 Mar 2012 07:15:31 -0700

> Hi HATAYAMA,
> 
>> -----Original Message-----
>> From: HATAYAMA Daisuke [mailto:d.hatayama@xxxxxxxxxxxxxx]
>> Sent: Friday, March 23, 2012 8:38 AM
>> To: crash-utility@xxxxxxxxxx; Lei Wen
>> Subject: Re:  [ANNOUNCE] crash gcore command, version 1.1
>> available
>> 
>> From: Dave Anderson <anderson@xxxxxxxxxx>
>> Subject: Re:  [ANNOUNCE] crash gcore command, version 1.1
>> available
>> Date: Thu, 22 Mar 2012 11:23:36 -0400 (EDT)
>> 
>> > Can you please format the crash-gcore-command-1.1.tar.gz file in
>> > the same manner as crash-gcore-command-1.0.tar.gz?  Here is the
>> > current version:
>> 
>> Thanks for pointing out that, Dave. I attached new one.
>> 
>> And for Lei, I made bigger fault that I missed arm source
>> file... Could you confirm that the new one can work on arm
>> environment?
> 
> Yes, this version works for me.
> 

I see.

> However, I think there is a point may need to be enhanced:
> diff --git a/extensions/libgcore/gcore_coredump.c b/extensions/libgcore/gcore_coredump.c
> index e2d9941..fec8ffc 100644
> --- a/extensions/libgcore/gcore_coredump.c
> +++ b/extensions/libgcore/gcore_coredump.c
> @@ -200,7 +200,8 @@ void gcore_coredump(void)
>                                         "readmem vma list",
>                                         gcore_verbose_error_handle());
>                         } else {
> -                               pagefaultf("page fault at %lx\n", addr);
> +                               if (paddr == PADDR_NOT_AVAILABLE)
> +                                       pagefaultf("page fault at %lx\n", addr);
>                                 BZERO(buffer, PAGE_SIZE);
>                         }
> 
> Without this, I would see many " page fault at XXX" for the lazy mapping as we talked in the previous thread:
> http://www.redhat.com/archives/crash-utility/2012-February/msg00112.html
> 

As Dave explains, gcore needs to inform users of the fact that some
part of memory is lost. However, it is cerntain that currnet
implementation displays ``page fault'' even for the pages that should
be displayed differently. I want to change the message for such pages
to proper one.

Also, to be honest, I didn't know PADDR_NOT_AVAILABLE. Looking at
source code, it appears to be used in Xen path. I have yet to know
whether or not to use the macro. I'll consider it more.

BTW, if you feel the page-fault message noisy, please use -v 0. Then,
gcore doesn't say almost anything.

Thanks.
HATAYAMA, Daisuke

--
Crash-utility mailing list
Crash-utility@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/crash-utility