Failed to establish shared memory mapping

[Martin van Es]

Hi Duncan,

I just discovered that a newly created account doesn't suffer from this
warning. So what option in my carefully crafted (kde) configuration could
be responsible for plasma/kde not being able to initialize shared memory?

On Wed, Nov 18, 2015 at 8:43 AM, Martin van Es <mrvanes at gmail.com> wrote:

> Hi Duncan,
>
> Thx for your lengthy answer.
> I'm using Kubuntu Wily with custom built kernel 4.3.0 (but the same
> happens when booting the ubuntu kernel).
> I don't use apparmor and selinux is disabled according to getenforce.
>
> I use kubuntu-backports repository, so as for versions, I'm at plasma
> 5.4.3, framework 5.15 and apps 15.08.2
>
> Best regards,
> Martin
>
> On Wed, Nov 18, 2015 at 2:31 AM, Duncan <1i5t5.duncan at cox.net> wrote:
>
>> Martin van Es posted on Tue, 17 Nov 2015 12:50:35 +0100 as excerpted:
>>
>> > Hi,
>> >
>> > I've searched high and low but couldn't find a satisfactory answer to
>> > the question: why do all KDE/Plasma apps warn "Failed to establish
>> > shared memory mapping, will fallback to private memory -- memory usage
>> > will increase"?
>> >
>> > I have /dev/shm mounted (tmpfs) and /dev/shm has drwxrwxrwt mode bits,
>> > /run/shm is a symbolic link to /dev/shm
>> >
>> > What I find curious is that when I start the same app as root it doesn't
>> > show the warning, but I can't for the life of me understand why I can't
>> > establish shared memory as a user and can as root?
>> >
>> > /dev/shm contains file that have pulse-shm in their name, and ipcs -a
>> > shows me many shared memory segments?
>> > So why can't KDE apps establish shared memory?
>>
>> This /may/ be a result of some sort of security measures your distro has
>> setup.  Unfortunately you didn't mention the distro and version you're
>> running, but I've not seen such warnings here, on Gentoo/~amd64.
>>
>> Also, you didn't mention what version of kde/plasma.  With many distros
>> switching to kde-frameworks-5 and plasma-5, while others are still on
>> kde4, and even where the switch to 5 has taken place, individual apps may
>> remain kde4 based for the time being, this could be important, as well.
>>
>> FWIW, kde4 here, tho I have most of a minimal kde-frameworks-5/plasma-5
>> installed for easier testing, lacking only the few bits that can't be
>> installed along with the kde4 versions, so I can keep what's configured
>> and working, while quickly switching to the new versions for testing when
>> I have time.
>>
>> As for the warning itself, I don't believe it actually refers to shm.
>>
>> KDE (at least kde4, and kde3 before it, and I'd guess plasma5 does
>> something similar) normally starts up using a special initialization
>> sequence that starts one process and then forks several others off it,
>> doing it in such a way that they can share the same library (elf *.so
>> shared objects) address space, etc, thus allowing shared libraries with
>> faster launching and lower memory usage, even where security measures
>> such as memory-space randomization would normally force separate
>> applications to use their own separately randomized library addresses for
>> the same libraries, making it impossible to share the same library
>> address space and increasing memory usage.
>>
>> I'd guess that this isn't working in your case for whatever reason, very
>> possibly due to additional security measures taken by your distro.  If
>> so, it really doesn't have anything to do with tmpfs or /dev/shm and its
>> permissions, but rather, with whatever additional security measures your
>> distro is enforcing.
>>
>> --
>> Duncan - List replies preferred.   No HTML msgs.
>> "Every nonfree program has a lord, a master --
>> and if you use the program, he is your master."  Richard Stallman
>>
>> ___________________________________________________
>> This message is from the kde mailing list.
>> Account management:  https://mail.kde.org/mailman/listinfo/kde.
>> Archives: http://lists.kde.org/.
>> More info: http://www.kde.org/faq.html.
>
>
>
>
> --
> If 'but' was any useful, it would be a logic operator
>



-- 
If 'but' was any useful, it would be a logic operator
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde/attachments/20151119/cc7dbdaf/attachment.htm>
-------------- next part --------------
___________________________________________________
This message is from the kde mailing list.
Account management:  https://mail.kde.org/mailman/listinfo/kde.
Archives: http://lists.kde.org/.
More info: http://www.kde.org/faq.html.