Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

[Max Grabert]

[-- Attachment #1: Type: text/plain, Size: 1757 bytes --]

On 26/03/06, John David Anglin <dave@hiauly1.hia.nrc.ca> wrote:
>
> > But I did booted yesterday the latest CVS 32bit Kernel on my c3k and it
> worked fine, so I assume you shouldn't see any problems either...
>
> I tried the default 2.6.16-pa5 from www.parisc-linux.org this afternoon.
> I still get a lot of ip_tables messages on the console which unfortunately
> aren't captured in any log file.  I've attached the diff between
> 2.6.15-rc7-pa
> and 2.6.16-pa5 dmesg files.  Is the eth0 difference significant?  There's
> also a segv in dirmngr that occurs late in both boots.


I had a similar problem (regarding iptables). It seems they are implementing
a new unified interface (NFNETLINK) which might break the userspace
netfilter tools (aka. iptables/ip6tables), unless you use the CVS version of
them. This is was introduced somewhere between 2.6.15 and 2.6.16.
IIRC after disabling CONFIG_NETFILTER_NETLINK (or at least set it to compile
it as a module) and enabling CONFIG_NETFILTER_XTABLES I didn't have any
problems anymore.


Regarding the segfault, I can't help you there ... although I have seen many
applications not longer working on my Debian/testing system lately
(segfault, 'unresolvable relocation error 0x42' most of the time).
The affected applications/debian packages are:
- aide
- samba (libsmbclient, samba, samba-common, smbclient, smbfs etc.)
- autofs

Recompiling those applications via 'apt-get build-dep <package>' and
'apt-get -b source <package>' with gcc-3.4 produces working executables. At
least for samba I also tried recompiling with gcc-4.0, with also worked. It
seems that there is something wrong with the setup Debian's buildd uses ...
... I think I should rather report this bug to the debian-hppa m-l ...

   Max

[-- Attachment #2: Type: text/html, Size: 2183 bytes --]

Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

[John David Anglin]

> Regarding the segfault, I can't help you there ... although I have seen many
> applications not longer working on my Debian/testing system lately
> (segfault, 'unresolvable relocation error 0x42' most of the time).

This should be fixed in libc6 2.3.6-3 which has support for the above
relocation.  If it's still occurring, there needs to be a new bug report.
In the situations that I've seen the relocations are coming from
/usr/lib/Scrt1.o when pie code is being used.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

[Helge Deller]

On Sunday 26 March 2006 06:05, John David Anglin wrote:
> > Regarding the segfault, I can't help you there ... although I have seen many
> > applications not longer working on my Debian/testing system lately
> > (segfault, 'unresolvable relocation error 0x42' most of the time).
> 
> This should be fixed in libc6 2.3.6-3 which has support for the above
> relocation.  If it's still occurring, there needs to be a new bug report.
> In the situations that I've seen the relocations are coming from
> /usr/lib/Scrt1.o when pie code is being used.

Hi Dave,

I'm running libc6-2.3.6-4 and see some strange crashes. Might it be related, or does  2.3.6-4 doesn't includes your fix ?

Regards,
Helge

root@c3000:~# dpkg -l | grep libc
ii  libc6                       2.3.6-4                        GNU C Library: Shared libraries and Timezone data
ii  libc6-dev                   2.3.6-4                        GNU C Library: Development Libraries and Header Files
ii  libc6-pic                   2.3.6-4                        GNU C Library: PIC archive library

dbus-0.61-5.deb
nscd-2.3.6-4.deb

do_page_fault() pid=1854 command='dbus-daemon' type=15 address=0x41493520
vm_start = 0x4095d000, vm_end = 0x4095f000

     YZrvWESTHLNXBCVMcbcbcbcbOGFRQPDI
PSW: 00000000000001101111110100001111 Not tainted
r00-03  00000000 0062c6c3 409148df 40493818
r04-07  4095b714 404933b8 00000005 40493000
r08-11  0000000f 0007a960 00000005 fb556648
r12-15  fb55664c 4095c714 4095c714 fb55674c
r16-19  4095ed9c 00000001 00000000 4095b714
r20-23  01000168 00000000 000196a8 0007a960
r24-27  00000005 000000d3 a2d3a6be 00070eec
r28-31  00000000 404931e4 fb556880 409148f3
sr0-3   00000000 00000000 00000000 00001765
sr4-7   00001765 00001765 00001765 00001765

      VZOUICununcqcqcqcqcqcrmunTDVZOUI
FPSR: 00000000000000000000000000000000
FPER1: 00000000
fr00-03  0000000000000000 0000000000000000 0000000000000000 0000000000000000
fr04-07  11bb69e08f41d7fc 000000001017b2f8 1051e2201051e220 11ba1b0810436960
fr08-11  0000000300000000 11bb69e00098963a 0000000000000002 11ba1b0810436960
fr12-15  0007a8a810517010 0000000010125938 0007a8c81054d6a0 0000000000030000
fr16-19  0000000b1051e810 11ba409c11ba4148 1051e220103af867 00002000103d15a0
fr20-23  11b9400000000000 0000000010166ef8 0000000000000000 0000000010514010
fr24-27  0000000110436810 10125a9800000000 232aaf8049c5f780 0098963a11b94000
fr28-31  ffffff9c11ba1b08 11ba414810167050 000000001051e010 00000000f0400004

IASQ: 00001765 00001765 IAOQ: 40914917 4091491b
 IIR: 0e85001c    ISR: 00001765  IOR: 41493520
 CPU:        0   CR30: 8eb90000 CR31: 104bc000
 ORIG_R28: 00000000
 IAOQ[0]: 0x40914917
 IAOQ[1]: 0x4091491b
 RP(r2): 0x409148df


do_page_fault() pid=1830 command='nscd' type=15 address=0x41493534
vm_start = 0x41425000, vm_end = 0x41428000

     YZrvWESTHLNXBCVMcbcbcbcbOGFRQPDI
PSW: 00000000000001001111111100001111 Not tainted
r00-03  00000000 410d0fdc 410b80cf 404c8c80
r04-07  410d0ca0 00000000 4426c56b 404937e8
r08-11  41493520 00000079 404933b8 00000079
r12-15  404c8e89 000000d3 000000d2 410d0df0
r16-19  4426c56b 00000059 410d0fdc 410d0ca0
r20-23  00000000 01000168 00000000 00000008
r24-27  404c8cb8 404c8e48 404c8cb8 410d0ca0
r28-31  00000001 00000803 404c8f40 00000000
sr0-3   00001701 00000000 00000000 00001701
sr4-7   00001701 00001701 00001701 00001701

      VZOUICununcqcqcqcqcqcrmunTDVZOUI
FPSR: 00000000000000000000000000000000
FPER1: 00000000
fr00-03  0000000000000000 0000000000000000 0000000000000000 0000000000000000
fr04-07  11bb69e08f41d7fc 000000001017b2f8 1051e2201051e220 11ba1b0810436960
fr08-11  0000000300000000 11bb69e00098963a 0000000000000002 11ba1b0810436960
fr12-15  0000100010517010 404c8e4010125938 402ac4b41054d6a0 0000000000030000
fr16-19  0000000b1051e810 11ba409c11ba4148 1051e220103af867 00002000103d15a0
fr20-23  11b9400000000000 0000000010166ef8 40095a2bcccccccd 404c8cb810514010
fr24-27  0000000a10436810 10125a9800000000 232aaf8049c5f780 0098963a11b94000
fr28-31  ffffff9c11ba1b08 11ba414810167050 000000001051e010 00000000f0400004

IASQ: 00001701 00001701 IAOQ: 410b7f43 410b7f47
 IIR: 491c0028    ISR: 00001701  IOR: 41493534
 CPU:        0   CR30: 8e9cc000 CR31: 104bc000
 ORIG_R28: 00000001
 IAOQ[0]: 0x410b7f43
 IAOQ[1]: 0x410b7f47
 RP(r2): 0x410b80cf
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

[Joel Soete]

Hello Matthias,

Helge Deller wrote:
> On Sunday 26 March 2006 06:05, John David Anglin wrote:
> 
>>>Regarding the segfault, I can't help you there ... although I have seen many
>>>applications not longer working on my Debian/testing system lately
>>>(segfault, 'unresolvable relocation error 0x42' most of the time).
>>
>>This should be fixed in libc6 2.3.6-3 which has support for the above
>>relocation.  If it's still occurring, there needs to be a new bug report.
>>In the situations that I've seen the relocations are coming from
>>/usr/lib/Scrt1.o when pie code is being used.
> 
> 
> Hi Dave,
> 
> I'm running libc6-2.3.6-4 and see some strange crashes. Might it be related, or does  2.3.6-4 doesn't includes your fix ?
> 
Do you know in which gcc-4.0 dpkg release the jda patch 
(<http://lists.parisc-linux.org/pipermail/parisc-linux/2006-March/028548.html>) was in?
I can so check in buildd with which compiler this new glibc-2.3.6-4 was build ;-)

TIA,
	Joel

Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

[John David Anglin]

> > I'm running libc6-2.3.6-4 and see some strange crashes. Might it be related, or does  2.3.6-4 doesn't includes your fix ?

2.3.6-4 should have the glibc fix...

> Do you know in which gcc-4.0 dpkg release the jda patch 
> (<http://lists.parisc-linux.org/pipermail/parisc-linux/2006-March/028548.html>) was in?
> I can so check in buildd with which compiler this new glibc-2.3.6-4 was build ;-)

As far as GCC goes, the fix is now in all 4.x branches but not in 3.4.
I'm not sure about the debian status.  I believe that Matthias integrated
the change but there was a regression due to a missing glibc patch
needed to prevent a SIGSEGV in the unwind code.  Matthias said this
problem was fixed on March 20.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

[John David Anglin]

> I'm not sure about the debian status.  I believe that Matthias integrated
> the change but there was a regression due to a missing glibc patch
> needed to prevent a SIGSEGV in the unwind code.  Matthias said this
> problem was fixed on March 20.

2.3.6-3 does't have the fix:

dave@hiauly6:~$ objdump -d /lib/libpthread.so.0
/lib/libpthread.so.0:     file format elf32-hppa-linux

Disassembly of section .init:

000041ac <_init>:
    41ac:       6b c2 3f d9     stw rp,-14(sp)
    41b0:       6f c4 00 80     stw,ma r4,40(sp)
    41b4:       6b d3 3f c1     stw r19,-20(sp)
    41b8:       e8 42 02 a0     b,l 8310 <__pthread_initialize_minimal>,rp
    41bc:       08 13 02 44     copy r19,r4
    41c0:       08 04 02 53     copy r4,r19
    41c4:       e8 40 0c 70     b,l 4804 <frame_dummy>,rp
    41c8:       08 00 02 40     nop
    41cc:       e8 44 1d 2c     b,l e068 <__do_global_ctors_aux>,rp
    41d0:       08 00 02 40     nop
    41d4:       4b c2 3f 59     ldw -54(sp),rp
    41d8:       08 04 02 53     copy r4,r19
    41dc:       e8 40 c0 00     bv r0(rp)

000041e0 <_end_init>:
    41e0:       4f c4 3f 81     ldw,mb -40(sp),r4

This .init section isn't built correctly in shared libraries.  The
PIC register r19 needs to be retored after every call if it is
subsequently used.  As can be seen above, there is no restore after
the call to frame_dummy and r19 is needed for the call to
__do_global_ctors_aux.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

[John David Anglin]

> I'm running libc6-2.3.6-4 and see some strange crashes. Might it be related, or does  2.3.6-4 doesn't includes your fix ?

Don't know.  Note the r19 bug and the dynamic loader relocation bug
are completely different bugs.  It's easy to see the r19 bug by looking
at the .init section of shared libraries as shown in my previous
message.  A backtrace would also help to see where the SEGV is
occuring.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

[Joel Soete]

> > I'm not sure about the debian status.  I believe that Matthias integr=
ated
> > the change but there was a regression due to a missing glibc patch
> > needed to prevent a SIGSEGV in the unwind code.  Matthias said this
> > problem was fixed on March 20.
> 
> 2.3.6-3 does't have the fix:
> 
> dave@hiauly6:~$ objdump -d /lib/libpthread.so.0
> /lib/libpthread.so.0:     file format elf32-hppa-linux
> 
> Disassembly of section .init:
> 
> 000041ac <_init>:
>     41ac:       6b c2 3f d9     stw rp,-14(sp)
>     41b0:       6f c4 00 80     stw,ma r4,40(sp)
>     41b4:       6b d3 3f c1     stw r19,-20(sp)
>     41b8:       e8 42 02 a0     b,l 8310 <__pthread_initialize_minimal>=
,rp
>     41bc:       08 13 02 44     copy r19,r4
>     41c0:       08 04 02 53     copy r4,r19
>     41c4:       e8 40 0c 70     b,l 4804 <frame_dummy>,rp
>     41c8:       08 00 02 40     nop
>     41cc:       e8 44 1d 2c     b,l e068 <__do_global_ctors_aux>,rp
>     41d0:       08 00 02 40     nop
>     41d4:       4b c2 3f 59     ldw -54(sp),rp
>     41d8:       08 04 02 53     copy r4,r19
>     41dc:       e8 40 c0 00     bv r0(rp)
> 
> 000041e0 <_end_init>:
>     41e0:       4f c4 3f 81     ldw,mb -40(sp),r4
> 
> This .init section isn't built correctly in shared libraries.  The
> PIC register r19 needs to be retored after every call if it is
> subsequently used.  As can be seen above, there is no restore after
> the call to frame_dummy and r19 is needed for the call to
> __do_global_ctors_aux.
> 
mmm not yet in 2.3.6-4:
# objdump -d /lib/libpthread.so.0 | more

/lib/libpthread.so.0:     file format elf32-hppa-linux

Disassembly of section .init:

000041ac <_init>:
    41ac:       6b c2 3f d9     stw rp,-14(sp)
    41b0:       6f c4 00 80     stw,ma r4,40(sp)
    41b4:       6b d3 3f c1     stw r19,-20(sp)
    41b8:       e8 42 02 a0     b,l 8310 <__pthread_initialize_minimal>,r=
p
    41bc:       08 13 02 44     copy r19,r4
    41c0:       08 04 02 53     copy r4,r19
    41c4:       e8 40 0c 70     b,l 4804 <frame_dummy>,rp
    41c8:       08 00 02 40     nop
    41cc:       e8 44 1d 2c     b,l e068 <__do_global_ctors_aux>,rp
    41d0:       08 00 02 40     nop
    41d4:       4b c2 3f 59     ldw -54(sp),rp
    41d8:       08 04 02 53     copy r4,r19
    41dc:       e8 40 c0 00     bv r0(rp)

000041e0 <_end_init>:
    41e0:       4f c4 3f 81     ldw,mb -40(sp),r4
Disassembly of section .text:
=3D=3D=3D=3D<>=3D=3D=3D=3D

and the same with my testing 2.3.6-4.1 build with gcc dpkg:
dpkg -l gcc-4.0
Desired=3DUnknown/Install/Remove/Purge/Hold
| Status=3DNot/Installed/Config-files/Unpacked/Failed-config/Half-install=
ed
|/ Err?=3D(none)/Hold/Reinst-required/X=3Dboth-problems (Status,Err: uppe=
rcase=3Dbad)
||/ Name                        Version                     Description
+++-=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D-=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D-=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D
ii  gcc-4.0                     4.0.3-1                     The GNU C com=
piler

Thanks,
    Joel
=0A=0A---------------------------------------------------------------=0AA=
 free anti-spam and anti-virus filter on all Scarlet mailboxes=0AMore inf=
o on http://www.scarlet.be/

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

[Max Grabert]

[-- Attachment #1.1: Type: text/plain, Size: 1041 bytes --]

On Sunday 26 March 2006 06:05, John David Anglin wrote:
> > Regarding the segfault, I can't help you there ... although I have seen
many
> > applications not longer working on my Debian/testing system lately
> > (segfault, 'unresolvable relocation error 0x42' most of the time).
>
> This should be fixed in libc6 2.3.6-3 which has support for the above
> relocation.  If it's still occurring, there needs to be a new bug report.
> In the situations that I've seen the relocations are coming from
> /usr/lib/Scrt1.o when pie code is being used.

It seems with glibc-2.3.6-3 I no longer see the 'unresolved relocation 0x42'
error, but still quite a lot of (the same) apps are (still) not working
correctly:
- autofs (automount doesn't crash, but doesn't work as supposed either)
- samba (crashes)
- squid (just opens up an strange UDP port instead of the neccessary TCP
port)
- aide (crashes)

At least for autofs, samba and aide the old trick of recompiling with
gcc-3.4 seems to work, but I think squid seems to be genuinely broken ;-P

Max

[-- Attachment #1.2: Type: text/html, Size: 1164 bytes --]

[-- Attachment #2: Type: text/plain, Size: 169 bytes --]

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: linux-2.6 deller (ioremap-changes) and more

[Joel Soete]

John David Anglin wrote:
>>>I'm running libc6-2.3.6-4 and see some strange crashes. Might it be related, or does  2.3.6-4 doesn't includes your fix ?
> 
> 
> 2.3.6-4 should have the glibc fix...
> 
> 
>>Do you know in which gcc-4.0 dpkg release the jda patch 
>>(<http://lists.parisc-linux.org/pipermail/parisc-linux/2006-March/028548.html>) was in?
>>I can so check in buildd with which compiler this new glibc-2.3.6-4 was build ;-)
> 
> 
> As far as GCC goes, the fix is now in all 4.x branches but not in 3.4.
> I'm not sure about the debian status.  I believe that Matthias integrated
> the change but there was a regression due to a missing glibc patch
> needed to prevent a SIGSEGV in the unwind code.  Matthias said this
> problem was fixed on March 20.
Sorry for delay but I come back to you with this stuff because I finaly take some time today to make some free space to check:
	* this patch is included into gcc-snapshot dpkg (namely ref to 2006 03 25 ;-) )
	* but neither gcc-4.0 nor gcc-4.1 dpkg contained this patch ;-(
	  (should I ask to Matthias, aka doko, lol?)

Thanks,
	Joel