Quoting Nathan Lynch (ntl@xxxxxxxxx): > Hello- > > On Tue, 2010-04-06 at 11:54 +0900, Jiro SEKIBA wrote: > > I tried checkpointing/restarting with this binary, but still get SEGV at the same point, > > still get same gdb warning(Cannot find thread-local variables on this target) > > I believe I have been experiencing a similar problem: > > https://lists.linux-foundation.org/pipermail/containers/2010-April/024105.html > > I may have found a fix. Can you please test whether this kernel patch > makes the SEGVs after restart go away? Acked-by: Serge Hallyn <serue@xxxxxxxxxx> Tested-by: Serge Hallyn <serue@xxxxxxxxxx> Thanks, Nathan. So I could confirm that: 1. my regular ckpt-v20 worked fine. 2. setting CONFIG_CC_STACKPROTECTOR=y made cr_tests/sleeptest fail while cr_tests/simple succeeded 3. applying this patch lets sleeptest succeed. Thanks! -serge > > > arch/x86/kernel/checkpoint_32.c | 3 +-- > 1 files changed, 1 insertions(+), 2 deletions(-) > > diff --git a/arch/x86/kernel/checkpoint_32.c b/arch/x86/kernel/checkpoint_32.c > index 32cde34..d80c168 100644 > --- a/arch/x86/kernel/checkpoint_32.c > +++ b/arch/x86/kernel/checkpoint_32.c > @@ -86,7 +86,6 @@ static unsigned short decode_segment(__u16 seg) > > void save_cpu_regs(struct ckpt_hdr_cpu *h, struct task_struct *t) > { > - struct thread_struct *thread = &t->thread; > struct pt_regs *regs = task_pt_regs(t); > unsigned long _gs; > > @@ -116,7 +115,7 @@ void save_cpu_regs(struct ckpt_hdr_cpu *h, struct task_struct *t) > if (t == current) > _gs = get_user_gs(regs); > else > - _gs = thread->gs; > + _gs = task_user_gs(t); > > h->fsindex = encode_segment(regs->fs); > h->gsindex = encode_segment(_gs); > > > _______________________________________________ > Containers mailing list > Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx > https://lists.linux-foundation.org/mailman/listinfo/containers _______________________________________________ Containers mailing list Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/containers
