Improving the kernel/i386 timecounter performance (GSoC proposal)

[Suleiman Souhlal]

On Mar 27, 2009, at 11:30 AM, Scott Long wrote:

> Robert Watson wrote:
>> On Fri, 27 Mar 2009, Scott Long wrote:
>>> I've been talking about this for years.  All I need is help with  
>>> the VM magic to create the page on fork.  I also want two pages,  
>>> one global for gettimeofday (and any other global data we can  
>>> think of) and one per-process for static data like getpid/getgid.
>> FWIW, there are some variations in schemes across OS's -- one  
>> extreme is the Linux approach, which actually exports a mini  
>> shared library in ELF format on the shared page, providing  
>> implementations of various services (such as entering system  
>> calls), time stuff, etc.  Less extreme are the shared pages  
>> offered on Mac OS X, etc.
>
> Yes, but I'd like to start somewhere, and considering that it's been
> impossible in _5_ years to get the 30 minutes of Peter or JeffR or JHB
> time to get the basic VM magic done, I'm keeping my expectations as
> modest as possible.
>

You can find a proof-of-concept implementation for amd64 of a global  
page mapped in every process at http://people.freebsd.org/~ssouhlal/ 
testing/syspage-20090328.diff .

It exports ticks to userland at VM_MIN_KERNEL_ADDRESS  
(0xfffffffe40000000).

In order for this to work on architectures without a direct map, the  
page will need to be mapped a second time as read/write (you might  
want to have a vm_offset_t pmap_map_syspage(vm_page_t m) function  
that does the right thing for each architecture).

Unfortunately, this trick probably won't work for per-process pages  
without more work, because we wouldn't be able to just insert the  
page in kernel_map.

-- Suleiman