Project

General

Profile

Bug #3874

Packagemanager core dumps with hipster

Added by Gary Gendel about 7 years ago. Updated almost 7 years ago.

Status:
Closed
Priority:
Urgent
Assignee:
Category:
OI-Userland
Target version:
-
Start date:
2013-07-08
Due date:
2013-12-10
% Done:

100%

Estimated time:
8.00 h
Difficulty:
Medium
Tags:
python

Description

After an update from hipster, packagemanager core dumps. The trace-back is:

Core was generated by `/usr/bin/python2.6 /usr/bin/packagemanager'.
Program terminated with signal 11, Segmentation fault.
[New process 67755 ]
#0 0xfecfb660 in PyString_Format (format=0x84e14c0, args=0x80aeb3c)
at /data/jenkins/jobs/oi-userland/workspace/components/python/python26/Python-2.6.8/Objects/stringobject.c:5051
5051 /data/jenkins/jobs/oi-userland/workspace/components/python/python26/Python-2.6.8/Objects/stringobject.c: No such file or directory.
in /data/jenkins/jobs/oi-userland/workspace/components/python/python26/Python-2.6.8/Objects/stringobject.c

pkg works fine, so it seems to be something specific to the GUI wrapper. The left-hand pane is displayed but it crashes before populating the right-hand pane.


Related issues

Related to OpenIndiana Distribution - Bug #3876: hipster: libpython2.6_db.so.1.0 is invalidResolved2013-07-08

Actions
Related to illumos gate - Bug #3912: crti needs to make sure _init and _fini are 16-byte stack alignedResolved2013-07-26

Actions
Related to OpenIndiana Distribution - Bug #4038: GConf Error: Failed to contact configuration server; some possible causes are that you need to enable TCP/IP networking for ORBit, or you have stale NFS locks due to a system crash. See http://projects.gnome.org/gconf/ for information.Resolved2013-08-12

Actions
Related to OpenIndiana Distribution - Bug #1431: packagemanager a noop?Closed2011-08-312013-12-10

Actions
Related to OpenIndiana Distribution - Bug #879: Update Manager tells I am running an Live ImageClosed2011-04-042013-12-15

Actions
Related to OpenIndiana Distribution - Bug #4258: Pkg fails updating in Hipster, when /opt (without onbld) is mounted on separate datasetClosed2013-10-222013-12-10

Actions

History

#1

Updated by Marcel Telka about 7 years ago

It looks like there is unaligned movdqa read (it should be aligned to 16-byte boundary, but it isn't):

> ::status
debugging core file of packagemanager (32-bit) from telcontar
file: /usr/bin/i86/isapython2.6
initial argv: /usr/bin/python2.6 /usr/bin/packagemanager
threading model: native threads
status: process terminated by SIGSEGV (Segmentation Fault), addr=fecfb660
> ::stack!head
libpython2.6.so.1.0`PyString_Format+0x650(849a580, 80aeb3c, 80441dc, fedd5000, fed08a3b)
libpython2.6.so.1.0`string_mod+0x3f(849a580, 80aeb3c, 80441f4, fecf7394, 55, ffffffff)
libpython2.6.so.1.0`binary_op1+0xa0(10, afcfd18, 804424c, fecabb78, 8e27e90, 840f040)
libpython2.6.so.1.0`PyNumber_Remainder+0x24(849a580, 80aeb3c, 0, 0, fedd5000, ef4de60)
libpython2.6.so.1.0`PyEval_EvalFrameEx+0x28aa(afcfbb4, 0, 843aa44, 0, 0, 849b218)
libpython2.6.so.1.0`PyEval_EvalCodeEx+0x83b(849b770, 843aa44, 0, afcf73c, 1, afcf740)
libpython2.6.so.1.0`PyEval_EvalFrameEx+0x518c(afcf5c4, 0, 843aa44, 0, 0, 8e719b0)
libpython2.6.so.1.0`PyEval_EvalCodeEx+0x83b(8441bf0, 843aa44, 0, afca094, 2, afca09c)
libpython2.6.so.1.0`PyEval_EvalFrameEx+0x518c(afc9f4c, 0, 8e5957c, 0, 7ff, 2e7c3488)
libpython2.6.so.1.0`PyEval_EvalFrameEx+0x5330(afc2fa4, 0, 8e5957c, 0, 804547b, 2)
> PyString_Format+0x650::dis
libpython2.6.so.1.0`PyString_Format+0x617:      subl   0xfffffee4(%ebp),%eax
libpython2.6.so.1.0`PyString_Format+0x61d:      leal   -0x1(%edx,%eax),%edx
libpython2.6.so.1.0`PyString_Format+0x621:      movl   %edx,0xfffffee8(%ebp)
libpython2.6.so.1.0`PyString_Format+0x627:      movl   0xfffffee4(%ebp),%edx
libpython2.6.so.1.0`PyString_Format+0x62d:      leal   -0x1(%edx,%eax),%edx
libpython2.6.so.1.0`PyString_Format+0x631:      movl   %edx,0xfffffee4(%ebp)
libpython2.6.so.1.0`PyString_Format+0x637:      cmpl   $0x20,0xfffffed8(%ebp)
libpython2.6.so.1.0`PyString_Format+0x63e:      je     +0xb04   <libpython2.6.so.1.0`PyString_Format+0x1148>
libpython2.6.so.1.0`PyString_Format+0x644:      movl   -0xf8(%ebp),%eax
libpython2.6.so.1.0`PyString_Format+0x64a:      movl   0xfffffef4(%ebp),%edx
libpython2.6.so.1.0`PyString_Format+0x650:      movdqa %xmm0,0xfffffe28(%ebp)   <-----------------------
libpython2.6.so.1.0`PyString_Format+0x658:      movl   %eax,0x4(%esp)
libpython2.6.so.1.0`PyString_Format+0x65c:      movl   0xfffffef0(%ebp),%eax
libpython2.6.so.1.0`PyString_Format+0x662:      movl   %edx,0x8(%esp)
libpython2.6.so.1.0`PyString_Format+0x666:      movl   %eax,(%esp)
libpython2.6.so.1.0`PyString_Format+0x669:      call   -0x5ea46 <PLT=libc_hwcap1.so.1`memcpy>
libpython2.6.so.1.0`PyString_Format+0x66e:      movl   0xfffffef4(%ebp),%edx
libpython2.6.so.1.0`PyString_Format+0x674:      movdqa 0xfffffe28(%ebp),%xmm0
libpython2.6.so.1.0`PyString_Format+0x67c:      addl   %edx,0xfffffef0(%ebp)
libpython2.6.so.1.0`PyString_Format+0x682:      subl   %edx,0xfffffee8(%ebp)
libpython2.6.so.1.0`PyString_Format+0x688:      movl   0xfffffee4(%ebp),%edx
> <ebp+0xfffffe28=X
                8043fd4         
>
#2

Updated by Ken Mays about 7 years ago

  • Assignee set to OI Userland
#3

Updated by Ken Mays about 7 years ago

  • Assignee changed from OI Userland to OI PKG
#4

Updated by Ken Mays about 7 years ago

  • Category set to PKG (Image Packaging System)
  • Tags changed from needs-triage to pkg
#5

Updated by r a about 7 years ago

Updated to the lastest version of Hipster yesterday, the Package Manager still crashes.

#6

Updated by Adam Števko about 7 years ago

Could you try updating python package and run package manager? Segfaults should be fixed.

#7

Updated by Alexander Pyhalov about 7 years ago

No, this issue is still present.

#8

Updated by Alex Viskovatoff almost 7 years ago

  • Priority changed from Normal to Urgent
#9

Updated by Alexander Pyhalov almost 7 years ago

Marcel Telka wrote:

It looks like there is unaligned movdqa read (it should be aligned to 16-byte boundary, but it isn't):

[...]

Isn't it aligned? 0xfffffe28 is just -472. Address at 472 bytes is 32-bit aligned. So, it's something else.
And if it were not aligned, it would be SIGBUS, not SIGSEGV...

#10

Updated by Marcel Telka almost 7 years ago

Alexander Pyhalov wrote:

Isn't it aligned? 0xfffffe28 is just -472. Address at 472 bytes is 32-bit aligned. So, it's something else.
And if it were not aligned, it would be SIGBUS, not SIGSEGV...

The misaligned address is 8043fd4, not 0xfffffe28 (which is misaligned too; it should be 16-byte/128-bit aligned). And I just tested that misaligned movdqa generates SIGSEGV, not SIGBUS.

#11

Updated by Milan Jurik almost 7 years ago

  • Category changed from PKG (Image Packaging System) to 10
  • Assignee changed from OI PKG to OI Userland
  • Tags changed from pkg to python
#12

Updated by Predrag Zečević almost 7 years ago

Hi,

still happens.
Regards.

#13

Updated by Gary Gendel almost 7 years ago

Still crashes after hipster update on 12/4/2013, same stack in core dump.

What is new is this preamble when launching:
  • Message: pyobject_register_sinkfunc is deprecated (GtkWindow)
  • Message: pyobject_register_sinkfunc is deprecated (GtkInvisible)
  • Message: pyobject_register_sinkfunc is deprecated (GtkObject)

However, I'm not sure if hipster has really gotten the illumos updates:

$ uname -a
SunOS phoenix 5.11 illumos-40ccc6a i86pc i386 i86pc Solaris

#14

Updated by Ken Mays almost 7 years ago

  • Due date set to 2013-12-10
  • Status changed from New to Closed
  • % Done changed from 0 to 100
  • Estimated time set to 8.00 h

pkg was updated in hipster, decommissioning packagemanager. Resolved. Use pkg tools in CLI for now.

Also available in: Atom PDF