This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] linux: spawni.c: simplify error reporting to parent


On 09/28/2016 08:12 PM, Adhemerval Zanella wrote:
On 28/09/2016 08:22, Rich Felker wrote:
On Wed, Sep 28, 2016 at 05:03:41PM +0200, Andreas Schwab wrote:
On Sep 28 2016, Rich Felker <dalias@libc.org> wrote:

This is a bad idea for at least one reason: running under strace seems
to cause vfork _not_ to wait in the parent, leading to stack
clobbering and runaway wrong code execution.

This is not true.  With modern kernels implementing ptrace events a
tracer can correctly handle all clone variants including vfork.

The problem might very well be limited to older (but still supported,
I think) kernels or older versions of strace; I'm not sure. That's
what I meant by saying I don't have a formula to reproduce the issue
at the moment.

Rich, I value your comments, I really do, but this isn't actionable information, I'm afraid.

I am not convinced this specific issue should be a blocker for this
patch.  It seems quite limited in scope (running the process in
strace) on older/not supported kernel. And the behaviour is clearly
a kernel issue.

Also I checked with CentOS 6 (2.6.32-642.el6.x86_64, strace 4.8) and
it seems to follow the expected behaviour.  I would also consider
any deviation in more recent Linux releases as a kernel regression
and I see no point and trying to support it.

Agreed, seems reasonable.

Thanks,
Florian


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]