[lvm-devel] [PATCH 0/6] cmirrord memory handling fixes

[Brassow Jonathan]

> On Aug 10, 2015, at 2:06 AM, Ferenc Wagner <wferi at niif.hu> wrote:
> 
> Ferenc Wágner <wferi at niif.hu> writes:
> 
>> During clustered pvmove tests cmirrord kept on dying.  To ease
>> debugging, I added a --foreground option to it [1-3], which made the
>> GCC stack protector messages stand out at once.  The backtraces made
>> the fix straightforward [4].  As log_checkpoint seems a constant false
>> condition, preparing the message is generally wasted effort, but I
>> didn't want to change the logic, and now it does not crash the process
>> at least.  Looking for other unsafe sprintf() usages provided plentiful
>> results, I fixed the two easy ones [5,6].  Please use as you see fit.
>> 
>> Ferenc Wágner (6):
>>  cmirrord: move generic setup from daemonize() to init_all()
>>  cmirrord: introduce the --foreground option to avoid daemonization
>>  cmirrord: document the --foreground and --help options
>>  cmirrord: fix stack smashing
>>  cmirrord: avoid debugging buffer overflow in LOG_SPRINT
>>  cmirrord: avoid resync buffer overflow in LOG_SPRINT
>> 
>> daemons/cmirrord/clogd.c     |   62 +++++++++++++++++++++++++++++++++---------
>> daemons/cmirrord/cluster.c   |   23 ++++++++++++----
>> daemons/cmirrord/functions.c |    5 ++--
>> man/cmirrord.8.in            |    8 +++++-
>> 4 files changed, 76 insertions(+), 22 deletions(-)
> 
> Hi,
> 
> Are there any objections to these patches?  Shall I rework them to have
> any hope for inclusion?

No objections.  They look helpful, thanks!

I’ve looked them over, but I haven’t had a chance to test them.  I’ll try to get them pulled in soon.

 brassow