oom_adj

[Matthew Woehlke]

Maksim Orlovich wrote:
>> Lubos Lunak wrote:
>>> It is a question if we should and can protect Xorg, since
>>> the kernel can't tell which X client is the one creating too many X
>>> resources
>>> and causing Xorg to use a lot of memory.
>> I've seen X eat 7 GB of address space (no, I am not kidding), apparently
>> related to plasma, but killing plasma didn't help; I had to kill X. I'd
>> rather see Qt learn to reconnect apps when X dies,
> 
> That's basically impossible for a properly written X application w/o
> wasting tons of memory.

To reconnect, or to make X bloat like that?

>> and not protect X, since X is sometimes the problem.
>>
>> (FTR, I believe this is a bug in nvidia's cra^H^H^Hdrivers, not plasma.)
> 
> xrestop should tell what's leaking.

Please re-read the post; xrestop reported plasma as having the lion's 
share of the memory. However I suspect that this is because plasma 
tickles a bug in The Blob; stuff is allocated by plasma during normal 
operation, but something done by The Blob prevents it from ever being 
freed as it should. Especially since killing plasma didn't help. (IOW I 
would guess that The Blob is wrongly holding handles.)

-- 
Matthew
Please do not quote my e-mail address unobfuscated in message bodies.
-- 
"Nobody expects the traditional Bourne shell!"