[linux-lvm] More Segfaults with LVM 1.0.1rc3

[linux-lvm] More Segfaults with LVM 1.0.1rc3

[Markus Dobel]

Hi,

vgextend, vgreduce and lvextend segfaulted for me today, too. I'm using 

  gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-95)

which I guess is the reason for those problems.

For vgextend/vgreduce gdb told me, that in vg_extend_reduce "vg_name"
pointed to 0x1. So i did a little debugging and therefore also
inserting the line:

int vg_extend ( char *vg_name, pv_t *pv, vg_t *vg) {
   printf("VG Name is %s\n",vg_name ); // this one
   return vg_extend_reduce ( vg_name, pv, vg, VG_EXTEND);
}

in vg_extend_reduce.c, the same in vg_reduce(). Without changing
anything else, vgextend and vgreduce work for me now.


Similar in lv_extend_reduce.c:

int lv_extend ( vg_t *vg, lv_t *lv, char *lv_name) {
   printf("Extending %s\n", lv_name); // added this line.
   return lv_extend_reduce ( vg, lv, lv_name, LV_EXTEND);
}

I don't speak C too fluently, and I don't know if printf() does some
implicit casting or something or if it's gcc 2.96, but maybe this
information helps a bit in finding the mistake.

Regards, Markus

-- 
Spiegel-Leser wissen mehr. Fuer ein Semester-Abo!
http://www.kawo2.rwth-aachen.de/~mdobel/semesterabo.html

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[Heinz J . Mauelshagen]

Markus,

go without optimization "-O0" rather than "-O2" in LVMs configure stript
*or* with gcc 2.95.2.

On Thu, Oct 04, 2001 at 10:33:43PM +0200, Markus Dobel wrote:
> Hi,
> 
> vgextend, vgreduce and lvextend segfaulted for me today, too. I'm using 
> 
>   gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-95)
> 
> which I guess is the reason for those problems.
> 
> For vgextend/vgreduce gdb told me, that in vg_extend_reduce "vg_name"
> pointed to 0x1. So i did a little debugging and therefore also
> inserting the line:
> 
> int vg_extend ( char *vg_name, pv_t *pv, vg_t *vg) {
>    printf("VG Name is %s\n",vg_name ); // this one
>    return vg_extend_reduce ( vg_name, pv, vg, VG_EXTEND);
> }
> 
> in vg_extend_reduce.c, the same in vg_reduce(). Without changing
> anything else, vgextend and vgreduce work for me now.
> 
> 
> Similar in lv_extend_reduce.c:
> 
> int lv_extend ( vg_t *vg, lv_t *lv, char *lv_name) {
>    printf("Extending %s\n", lv_name); // added this line.
>    return lv_extend_reduce ( vg, lv, lv_name, LV_EXTEND);
> }
> 
> I don't speak C too fluently, and I don't know if printf() does some
> implicit casting or something or if it's gcc 2.96, but maybe this
> information helps a bit in finding the mistake.
> 
> Regards, Markus
> 
> -- 
> Spiegel-Leser wissen mehr. Fuer ein Semester-Abo!
> http://www.kawo2.rwth-aachen.de/~mdobel/semesterabo.html
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html

-- 

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
Mauelshagen@Sistina.com                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[svetljo]

is gcc-3.0.1 safe, or i should also use "-o0"


Heinz J . Mauelshagen wrote:

>Markus,
>
>go without optimization "-O0" rather than "-O2" in LVMs configure stript
>*or* with gcc 2.95.2.
>
>On Thu, Oct 04, 2001 at 10:33:43PM +0200, Markus Dobel wrote:
>
>>Hi,
>>
>>vgextend, vgreduce and lvextend segfaulted for me today, too. I'm using 
>>
>>  gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-95)
>>
>>which I guess is the reason for those problems.
>>
>>For vgextend/vgreduce gdb told me, that in vg_extend_reduce "vg_name"
>>pointed to 0x1. So i did a little debugging and therefore also
>>inserting the line:
>>
>>int vg_extend ( char *vg_name, pv_t *pv, vg_t *vg) {
>>   printf("VG Name is %s\n",vg_name ); // this one
>>   return vg_extend_reduce ( vg_name, pv, vg, VG_EXTEND);
>>}
>>
>>in vg_extend_reduce.c, the same in vg_reduce(). Without changing
>>anything else, vgextend and vgreduce work for me now.
>>
>>
>>Similar in lv_extend_reduce.c:
>>
>>int lv_extend ( vg_t *vg, lv_t *lv, char *lv_name) {
>>   printf("Extending %s\n", lv_name); // added this line.
>>   return lv_extend_reduce ( vg, lv, lv_name, LV_EXTEND);
>>}
>>
>>I don't speak C too fluently, and I don't know if printf() does some
>>implicit casting or something or if it's gcc 2.96, but maybe this
>>information helps a bit in finding the mistake.
>>
>>Regards, Markus
>>
>>-- 
>>Spiegel-Leser wissen mehr. Fuer ein Semester-Abo!
>>http://www.kawo2.rwth-aachen.de/~mdobel/semesterabo.html
>>
>>_______________________________________________
>>linux-lvm mailing list
>>linux-lvm@sistina.com
>>http://lists.sistina.com/mailman/listinfo/linux-lvm
>>read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html
>>
>

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[Heinz J . Mauelshagen]

On Fri, Oct 05, 2001 at 09:25:56AM +0200, svetljo wrote:
> is gcc-3.0.1 safe, or i should also use "-o0"

Don't have a report for this :-(
Please try it.

> 
> 
> Heinz J . Mauelshagen wrote:
> 
> >Markus,
> >
> >go without optimization "-O0" rather than "-O2" in LVMs configure stript
> >*or* with gcc 2.95.2.
> >
> >On Thu, Oct 04, 2001 at 10:33:43PM +0200, Markus Dobel wrote:
> >
> >>Hi,
> >>
> >>vgextend, vgreduce and lvextend segfaulted for me today, too. I'm using 
> >>
> >>  gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-95)
> >>
> >>which I guess is the reason for those problems.
> >>
> >>For vgextend/vgreduce gdb told me, that in vg_extend_reduce "vg_name"
> >>pointed to 0x1. So i did a little debugging and therefore also
> >>inserting the line:
> >>
> >>int vg_extend ( char *vg_name, pv_t *pv, vg_t *vg) {
> >>   printf("VG Name is %s\n",vg_name ); // this one
> >>   return vg_extend_reduce ( vg_name, pv, vg, VG_EXTEND);
> >>}
> >>
> >>in vg_extend_reduce.c, the same in vg_reduce(). Without changing
> >>anything else, vgextend and vgreduce work for me now.
> >>
> >>
> >>Similar in lv_extend_reduce.c:
> >>
> >>int lv_extend ( vg_t *vg, lv_t *lv, char *lv_name) {
> >>   printf("Extending %s\n", lv_name); // added this line.
> >>   return lv_extend_reduce ( vg, lv, lv_name, LV_EXTEND);
> >>}
> >>
> >>I don't speak C too fluently, and I don't know if printf() does some
> >>implicit casting or something or if it's gcc 2.96, but maybe this
> >>information helps a bit in finding the mistake.
> >>
> >>Regards, Markus
> >>
> >>-- 
> >>Spiegel-Leser wissen mehr. Fuer ein Semester-Abo!
> >>http://www.kawo2.rwth-aachen.de/~mdobel/semesterabo.html
> >>
> >>_______________________________________________
> >>linux-lvm mailing list
> >>linux-lvm@sistina.com
> >>http://lists.sistina.com/mailman/listinfo/linux-lvm
> >>read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html
> >>
> >
> 
> 
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html

-- 

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
Mauelshagen@Sistina.com                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[svetljo]

Heinz J . Mauelshagen wrote:

>On Fri, Oct 05, 2001 at 09:25:56AM +0200, svetljo wrote:
>
>>is gcc-3.0.1 safe, or i should also use "-o0"
>>
>
>Don't have a report for this :-(
>Please try it.
>
it's seems to work

xxscan xxdiplay lvcreate lvremove lvrename lvextend lvreduce work without pb
but i tried yesterday to install 2.4.11-pre2-xfs-cvs lvm-cvs
and vgscan from the initrd segfaults , i can not boot to my LV's with it
probbably it's some other trouble, i'll try with 2.4.11-pre3-xfs
i think it's not relevant but all my VG's are over softRAIDs

>>
>>Heinz J . Mauelshagen wrote:
>>
>>>Markus,
>>>
>>>go without optimization "-O0" rather than "-O2" in LVMs configure stript
>>>*or* with gcc 2.95.2.
>>>
>>>On Thu, Oct 04, 2001 at 10:33:43PM +0200, Markus Dobel wrote:
>>>
>>>>Hi,
>>>>
>>>>vgextend, vgreduce and lvextend segfaulted for me today, too. I'm using 
>>>>
>>>> gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-95)
>>>>
>>>>which I guess is the reason for those problems.
>>>>
>>>>For vgextend/vgreduce gdb told me, that in vg_extend_reduce "vg_name"
>>>>pointed to 0x1. So i did a little debugging and therefore also
>>>>inserting the line:
>>>>
>>>>int vg_extend ( char *vg_name, pv_t *pv, vg_t *vg) {
>>>>  printf("VG Name is %s\n",vg_name ); // this one
>>>>  return vg_extend_reduce ( vg_name, pv, vg, VG_EXTEND);
>>>>}
>>>>
>>>>in vg_extend_reduce.c, the same in vg_reduce(). Without changing
>>>>anything else, vgextend and vgreduce work for me now.
>>>>
>>>>
>>>>Similar in lv_extend_reduce.c:
>>>>
>>>>int lv_extend ( vg_t *vg, lv_t *lv, char *lv_name) {
>>>>  printf("Extending %s\n", lv_name); // added this line.
>>>>  return lv_extend_reduce ( vg, lv, lv_name, LV_EXTEND);
>>>>}
>>>>
>>>>I don't speak C too fluently, and I don't know if printf() does some
>>>>implicit casting or something or if it's gcc 2.96, but maybe this
>>>>information helps a bit in finding the mistake.
>>>>
>>>>Regards, Markus
>>>>
>>>>-- 
>>>>Spiegel-Leser wissen mehr. Fuer ein Semester-Abo!
>>>>http://www.kawo2.rwth-aachen.de/~mdobel/semesterabo.html
>>>>
>>>>_______________________________________________
>>>>linux-lvm mailing list
>>>>linux-lvm@sistina.com
>>>>http://lists.sistina.com/mailman/listinfo/linux-lvm
>>>>read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html
>>>>
>>
>>
>>_______________________________________________
>>linux-lvm mailing list
>>linux-lvm@sistina.com
>>http://lists.sistina.com/mailman/listinfo/linux-lvm
>>read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html
>>
>

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[Nils Juergens]

On Fri, 05.10.01, svetljo <galia@st-peter.stw.uni-erlangen.de> wrote:
> 
> Heinz J . Mauelshagen wrote:
> 
> >On Fri, Oct 05, 2001 at 09:25:56AM +0200, svetljo wrote:
> >
> >>is gcc-3.0.1 safe, or i should also use "-o0"
> >>
> >
> >Don't have a report for this :-(
> >Please try it.
> >
> it's seems to work
> 

I tried gcc-3.0.1 with LVM on linux-2.4.10 and had a lot of problems,
but it was no problem in LVM, it works with 2.95.4.

What i was seeing was funny stuff happening to the stack after
a function call causing (in my case) lvremove to segfault.
I have seen this with gcc-3.0.1 before, on another program.


Nils

-- 
 >   Nils Juergens  | nils@muon.de | icq 7090774   <
 >  If we all work hard, we can turn this company  <
 >      around 360 degrees -- gruntbert(LOTD)      <

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[svetljo]

>
> Heinz J . Mauelshagen wrote:
>
>> On Fri, Oct 05, 2001 at 09:25:56AM +0200, svetljo wrote:
>>
>>> is gcc-3.0.1 safe, or i should also use "-o0"
>>>
>>
>> Don't have a report for this :-(
>> Please try it.
>>
> it's seems to work
>
> xxscan xxdiplay lvcreate lvremove lvrename lvextend lvreduce work 
> without pb
> but i tried yesterday to install 2.4.11-pre2-xfs-cvs lvm-cvs
> and vgscan from the initrd segfaults , i can not boot to my LV's with it
> probbably it's some other trouble, i'll try with 2.4.11-pre3-xfs
> i think it's not relevant but all my VG's are over softRAIDs
>
it works with 2.4.11-pre4-xfs-cvs and lvm-cvs from today
i couldn't test vgcreate and vgremove .. ( because i have no spare PV's)
but the other tools work without complaining

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[Heinz J . Mauelshagen]

Nile,
thanks for the info.

Does LVM compiled under gcc-3.0.1 *without* optimization work?

On Sat, Oct 06, 2001 at 11:18:26PM +0000, Nils Juergens wrote:
> On Fri, 05.10.01, svetljo <galia@st-peter.stw.uni-erlangen.de> wrote:
> > 
> > Heinz J . Mauelshagen wrote:
> > 
> > >On Fri, Oct 05, 2001 at 09:25:56AM +0200, svetljo wrote:
> > >
> > >>is gcc-3.0.1 safe, or i should also use "-o0"
> > >>
> > >
> > >Don't have a report for this :-(
> > >Please try it.
> > >
> > it's seems to work
> > 
> 
> I tried gcc-3.0.1 with LVM on linux-2.4.10 and had a lot of problems,
> but it was no problem in LVM, it works with 2.95.4.
> 
> What i was seeing was funny stuff happening to the stack after
> a function call causing (in my case) lvremove to segfault.
> I have seen this with gcc-3.0.1 before, on another program.
> 
> 
> Nils
> 
> -- 
>  >   Nils Juergens  | nils@muon.de | icq 7090774   <
>  >  If we all work hard, we can turn this company  <
>  >      around 360 degrees -- gruntbert(LOTD)      <
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html

-- 

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
Mauelshagen@Sistina.com                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[Heinz J . Mauelshagen]

On Sun, Oct 07, 2001 at 02:30:17PM +0200, svetljo wrote:
> >
> > Heinz J . Mauelshagen wrote:
> >
> >> On Fri, Oct 05, 2001 at 09:25:56AM +0200, svetljo wrote:
> >>
> >>> is gcc-3.0.1 safe, or i should also use "-o0"
> >>>
> >>
> >> Don't have a report for this :-(
> >> Please try it.
> >>
> > it's seems to work

Well, our impression is that optimizing under gcc > 2.95.2 doesn't seem
to be a good idea.

Have switched optimization off in configure for now.
Change is in CVS and next release.


> >
> > xxscan xxdiplay lvcreate lvremove lvrename lvextend lvreduce work 
> > without pb
> > but i tried yesterday to install 2.4.11-pre2-xfs-cvs lvm-cvs
> > and vgscan from the initrd segfaults , i can not boot to my LV's with it
> > probbably it's some other trouble, i'll try with 2.4.11-pre3-xfs
> > i think it's not relevant but all my VG's are over softRAIDs
> >
> it works with 2.4.11-pre4-xfs-cvs and lvm-cvs from today
> i couldn't test vgcreate and vgremove .. ( because i have no spare PV's)
> but the other tools work without complaining
> 
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://www.sistina.com/lvm/Pages/howto.html

-- 

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
Mauelshagen@Sistina.com                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[Paul Jakma]

On Mon, 8 Oct 2001, Heinz J . Mauelshagen wrote:

> Well, our impression is that optimizing under gcc > 2.95.2 doesn't seem
> to be a good idea.

have you pinned it down to a specific bug in gcc, or is it based on 
"doesn't happen with -O0"?

curious... the segv i had showed a stack trace with no obvious 
correspondent path in the lvcreate.c code. FWIW, after a glibc 
upgrade and reboot lvcreate started working again.

??? weird..

regards,
-- 
Paul Jakma	paul@clubi.ie	paul@jakma.org	Key ID: 64A2FF6A
Fortune:
What is algebra, exactly?  Is it one of those three-cornered things?
		-- J.M. Barrie

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[Luca Berra]

On Tue, Oct 09, 2001 at 01:25:47AM +0100, Paul Jakma wrote:
> On Mon, 8 Oct 2001, Heinz J . Mauelshagen wrote:
> 
> > Well, our impression is that optimizing under gcc > 2.95.2 doesn't seem
> > to be a good idea.
> 
> have you pinned it down to a specific bug in gcc, or is it based on 
> "doesn't happen with -O0"?
> 
i rebuilt latest cvs wih mandrake compiler:
gcc version 2.96 20000731 (Mandrake Linux 8.1 2.96-0.62mdk)

using -O2 -march=athlon and it looks like vgscan is ok

L.

-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[Joe Thornber]

On Tue, Oct 09, 2001 at 01:25:47AM +0100, Paul Jakma wrote:
> On Mon, 8 Oct 2001, Heinz J . Mauelshagen wrote:
> 
> > Well, our impression is that optimizing under gcc > 2.95.2 doesn't seem
> > to be a good idea.
> 
> have you pinned it down to a specific bug in gcc, or is it based on 
> "doesn't happen with -O0"?

It appears to be a bug in just the redhat 7.1 version of gcc
(2.96-85), which I'm told was not an official release of gcc.  There
are a couple of places where LVM has one line wrapper functions; these
are being optimised incorrectly.

- Joe

Re: [linux-lvm] More Segfaults with LVM 1.0.1rc3

[Nils Juergens]

On Mon, 08.10.01, "Heinz J . Mauelshagen" <mauelshagen@sistina.com> wrote:
> 
> Does LVM compiled under gcc-3.0.1 *without* optimization work?
> 
No, with -O0 i get the same seg'fault.

Nils
-- 
 >   Nils Juergens  | nils@muon.de | icq 7090774   <
 >  If we all work hard, we can turn this company  <
 >      around 360 degrees -- gruntbert(LOTD)      <