lkml.org 
[lkml]   [2011]   [Oct]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v5 3.1.0-rc4-tip 26/26] uprobes: queue signals while thread is singlestepping.
* Oleg Nesterov <oleg@redhat.com> [2011-10-05 20:01:39]:

> Srikar, I am still reading this series, need more time to read this
> patch, but:

Okay,

>
> On 09/27, Srikar Dronamraju wrote:
> >
> > I did a rethink and implemented this patch a little differently using
> > block_all_signals, unblock_all_signals. This wouldnt need the
> > #ifdeffery + no changes in kernel/signal.c
>
> No, Please don't. block_all_signals() must be killed. This interface
> simply do not work. At all. It is buggy as hell. I guess I should ping
> David Airlie again.
>

I could use sigprocmask instead of block_all_signals.

The patch (that I sent out as part of v5 patchset) uses per task
pending sigqueue and start queueing the signals when the task
singlesteps. After completion of singlestep, walks thro the pending
signals.

But I was thinking if I should block signals instead of queueing them in
a different sigqueue. So Idea is to block signals just before the task
enables singlestep and unblock after task disables singlestep.

Instead of using block_all_signals, I could use sigprocmask to achieve
the same.

Which approach do you suggest or do you have any other approach to look
at?

--
Thanks and Regards
Srikar




\
 
 \ /
  Last update: 2011-10-06 08:07    [from the cache]
©2003-2011 Jasper Spaans