Re: building current - signal 6

2016-05-06 Thread Christos Zoulas
In article <20160506091115.ga15...@britannica.bec.de>,
Joerg Sonnenberger   wrote:
>On Fri, May 06, 2016 at 03:03:12AM +, Christos Zoulas wrote:
>> It is probably because ASLR is making a vmcmd fail. It should be simple to
>> build a kernel with DEBUG_EXEC and see what's breaking.
>
>Setting PAX_ASLR_STACK stopped all random aborts I saw in my bulk build.

That would be my guess too :-)

christos



Re: building current - signal 6

2016-05-06 Thread Joerg Sonnenberger
On Fri, May 06, 2016 at 03:03:12AM +, Christos Zoulas wrote:
> It is probably because ASLR is making a vmcmd fail. It should be simple to
> build a kernel with DEBUG_EXEC and see what's breaking.

Setting PAX_ASLR_STACK stopped all random aborts I saw in my bulk build.

Joerg


Re: building current - signal 6

2016-05-05 Thread Christos Zoulas
In article <5729be9e.5030...@libero.it>,
Riccardo Mottola   wrote:
>Hi Erik,
>
>
>Erik Fair wrote:
>> #define SIGABRT 6   /* abort() */
>>
>> See also signal(7) and abort(3).
>>
>> It would appear that make(1) called abort(3) since rm(1) doesn’t
>have any direct calls to it that I could find by reading the base
>source. Cursory grep through/usr/src/usr.bin/make/* shows possibilities
>…
>>
>> Perhaps turning on some make debug options will reveal more.
>
>I just reissued the command and it worked... like a temporary error 
>then. Strange.

Yes, is is random because ASLR is randomizing.
Probably some addresses don't work. It should be easy to test by hacking
the random number generator and looking what happens when the value is
set at either ends of the range.

christoa




Re: building current - signal 6

2016-05-05 Thread Christos Zoulas
In article <33d21884-f50f-4933-a887-337520338...@netbsd.org>,
Erik Fair   wrote:
>In /usr/include/sys/signal.h find:
>
>#define SIGABRT 6   /* abort() */
>
>See also signal(7) and abort(3).
>
>It would appear that make(1) called abort(3) since rm(1) doesn’t have
>any direct calls to it that I could find by reading the base source.
>Cursory grep through /usr/src/usr.bin/make/* shows possibilities …
>
>Perhaps turning on some make debug options will reveal more.

That ain't it, if you ktrace the process, you'll see that there was no
signal posted.  Instead what's happening is:

kern_exec.c:exit1(l, error, SIGABRT);

It is probably because ASLR is making a vmcmd fail. It should be simple to
build a kernel with DEBUG_EXEC and see what's breaking.

christos




Re: building current - signal 6

2016-05-05 Thread Erik Fair
Transient errors always worry me.

Erik


> On May 4, 2016, at 02:19, Riccardo Mottola  wrote:
> 
> Hi Erik,
> 
> 
> Erik Fair wrote:
>> #define SIGABRT 6   /* abort() */
>> 
>> See also signal(7) and abort(3).
>> 
>> It would appear that make(1) called abort(3) since rm(1) doesn’t have any 
>> direct calls to it that I could find by reading the base source. Cursory 
>> grep through/usr/src/usr.bin/make/* shows possibilities …
>> 
>> Perhaps turning on some make debug options will reveal more.
> 
> I just reissued the command and it worked... like a temporary error then. 
> Strange.
> 
> Riccardo




Re: building current - signal 6

2016-05-05 Thread Riccardo Mottola

Hi Erik,


Erik Fair wrote:

#define SIGABRT 6   /* abort() */

See also signal(7) and abort(3).

It would appear that make(1) called abort(3) since rm(1) doesn’t have any 
direct calls to it that I could find by reading the base source. Cursory grep 
through/usr/src/usr.bin/make/* shows possibilities …

Perhaps turning on some make debug options will reveal more.


I just reissued the command and it worked... like a temporary error 
then. Strange.


Riccardo



Re: building current - signal 6

2016-05-05 Thread Erik Fair
In /usr/include/sys/signal.h find:

#define SIGABRT 6   /* abort() */

See also signal(7) and abort(3).

It would appear that make(1) called abort(3) since rm(1) doesn’t have any 
direct calls to it that I could find by reading the base source. Cursory grep 
through /usr/src/usr.bin/make/* shows possibilities …

Perhaps turning on some make debug options will reveal more.

Erik Fair

> On May 3, 2016, at 16:59, Riccardo Mottola  wrote:
> 
> Hi,
> 
> when I build current on amd64, my build fails with:
> 
> #create  font-misc-misc/6x12-ISO8859-4.bdf
> /usr/src/../obj/external/mit/xorg/tools/ucs2any/ucs2any 
> /usr/xsrc/external/mit/font-misc-misc/dist/6x12.bdf 
> /usr/src/../obj/destdir.amd64/usr/X11R7/lib/X11/fonts/util/map-ISO8859-4 
> ISO8859-4
> Writing 223 characters into file '6x12-ISO8859-4.bdf'.
> #create  font-misc-misc/6x12-ISO8859-4.pcf.gz
> rm -f 6x12-ISO8859-4.pcf.gz
> *** Signal 6
> 
> Stop.
> nbmake[11]: stopped in 
> /usr/src/external/mit/xorg/share/fonts/misc/font-misc-mis c
> 
> *** Failed target:  dependall
> *** Failed command: cd 
> "/usr/src/external/mit/xorg/share/fonts/misc/font-misc-misc"; 
> /usr/src/../tools/bin/nbmake realall
> *** Error code 1
> 
> Stop.
> nbmake[10]: stopped in 
> /usr/src/external/mit/xorg/share/fonts/misc/font-misc-misc
> 
> 
> is the issue "Signal 6", what is going wrong?
> 
> 
> Thanks, Riccardo




Re: building current - signal 6

2016-05-04 Thread Christos Zoulas
In article <572a7735.5040...@libero.it>,
Riccardo Mottola   wrote:
>Hi Erik,
>
>Erik Fair wrote:
>> Transient errors always worry me.
>
>me too. It means you can't reproduce what failed and are left with the 
>doubt that it could hunt you down again.
>
>I am running a current kernel

This is caused by ASLR (that is why it is random). I will add some
debugging to figure out the exact conditions that cause it to happen.

christos



Re: building current - signal 6

2016-05-04 Thread Riccardo Mottola

Hi Erik,

Erik Fair wrote:

Transient errors always worry me.


me too. It means you can't reproduce what failed and are left with the 
doubt that it could hunt you down again.


I am running a current kernel

Riccardo


building current - signal 6

2016-05-03 Thread Riccardo Mottola

Hi,

when I build current on amd64, my build fails with:

#create  font-misc-misc/6x12-ISO8859-4.bdf
/usr/src/../obj/external/mit/xorg/tools/ucs2any/ucs2any 
/usr/xsrc/external/mit/font-misc-misc/dist/6x12.bdf 
/usr/src/../obj/destdir.amd64/usr/X11R7/lib/X11/fonts/util/map-ISO8859-4 
ISO8859-4

Writing 223 characters into file '6x12-ISO8859-4.bdf'.
#create  font-misc-misc/6x12-ISO8859-4.pcf.gz
rm -f 6x12-ISO8859-4.pcf.gz
*** Signal 6

Stop.
nbmake[11]: stopped in 
/usr/src/external/mit/xorg/share/fonts/misc/font-misc-mis c


*** Failed target:  dependall
*** Failed command: cd 
"/usr/src/external/mit/xorg/share/fonts/misc/font-misc-misc"; 
/usr/src/../tools/bin/nbmake realall

*** Error code 1

Stop.
nbmake[10]: stopped in 
/usr/src/external/mit/xorg/share/fonts/misc/font-misc-misc



is the issue "Signal 6", what is going wrong?


Thanks, Riccardo