[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi

I'm unhappy you guys uses offensive word so much. Please cool down all you guys. :-/
In fact, _BOTH_ the behavior before and after Cristoph's patch doesn't have cleaner semantics.
And PeterZ proposed make new cleaner one rather than revert. No need to hassle.

I'm 100% sure -rt people need stronger-mlock api. Please join discussion to make better API.
In my humble opinion is: we should make mlock3(addr, len flags) new syscall (*) and support
-rt requirement directly. And current strange IB RLIMIT_MEMLOCK usage should gradually migrate
it.
(*) or, to enhance mbind() is an option because i expect apps need to pin pages nearby NUMA nodes
in many case.

As your know, current IB pinning implementation doesn't guarantee no minor fault when fork
is used. It's ok for IB. They uses madvise(MADV_NOFORK) too. But I'm not sure *all* of rt
application are satisfied this. We might need to implement copy-on-fork or might not. I'd
like hear other people's opinion.

Also, all developer should know this pinning breaks when memory hot-plug is happen likes
cpu bounding bysched_setaffinity() may break when cpu hot-remove.

--
To unsubscribe from this list: send the line "unsubscribe trinity" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux SCSI]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux