kernel-packages team mailing list archive

[Anton Blanchard <anton@xxxxxxxxx>]

Based on the fail, I took a look at how gccgo handles stacks. It relies
on the split stack feature in gold, which doesn't appear to be
implemented for ppc64.

Running one of the go recursion testcases (attached) shows what happens
when we run out of stack and don't have the split stack feature to save
us:

#gccgo -g -O2 -o peano peano.go
# ./peano 
Segmentation fault

And we get the setup_rt_frame error in dmesg:

peano[4538]: bad frame in setup_rt_frame: 000000c20ff7f000 nip
0000000010001018 lr 0000000010001024

As expected, we are just continually recurse without checking out stack
pointer for overflow:

   0x0000000010001008 <+8>:     cmpdi   r3,0
   0x000000001000100c <+12>:    beq     0x10001040 <main.count+64>
   0x0000000010001010 <+16>:    mflr    r0
   0x0000000010001014 <+20>:    std     r0,16(r1)
   0x0000000010001018 <+24>:    stdu    r1,-32(r1)
   0x000000001000101c <+28>:    ld      r3,0(r3)
   0x0000000010001020 <+32>:    bl      0x10001008 <main.count+8>


** Attachment added: "peano.go"
   https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1304754/+attachment/4079310/+files/peano.go

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1304754

Title:
  gccgo compiled binaries are killed by SEGV on 64k ppc64el kernels

Status in “linux” package in Ubuntu:
  Incomplete

Bug description:
  On kernels 3.13-18 and 3.13-23 (there may be others) the kernel is
  killing gccgo compiled binaries

  [18519.444748] jujud[19277]: bad frame in setup_rt_frame:
  0000000000000000 nip 0000000000000000 lr 0000000000000000
  [18519.673632] init: juju-agent-ubuntu-local main process (19220)
  killed by SEGV signal
  [18519.673651] init: juju-agent-ubuntu-local main process ended, respawning

  In powerpc/kernel/signal_64.c:

  sys_rt_sigreturn is jumping to the badframe: label and executing an
  unconditional force_sigsegv which is delivered to the userland
  process. Like C++, gccgo tries to decode SIGSEGV as a nil pointer
  access and blame some random function that happened to be the top
  stack frame.

  Reverting to the 3.13-08 kernel appears to resolve the issue which
  (weakly) points the finger at the recent switch to 64k pages.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1304754/+subscriptions