My Linux AIO Implementation Calling for Test

[Intron]

Divacky Roman wrote:

> On Thu, Jul 27, 2006 at 05:44:01PM +0800, Intron wrote:
>> Download: http://ftp.intron.ac/tmp/linux_aio-20060727.tar.bz2
>> 
>> Based on FreeBSD aio(4), my patch set has implemented 5 system calls
>> of Linux Asynchronous Input/Output: io_setup(2), io_destroy(2),
>> io_getevents(2), io_submit(2) and io_cancel(2). It only works with
>> 7.0-CURRENT.
> 
> havent studied details but here are my suggestions:
> 
> 1) please conform to style(9)

Do you refer to "if" directive and line width?
I will obey style(9) as possible, keeping most of legibility.

> 
> 2) please use LMSG, ARGS macros to be conform with the rest of the linuxolator

I will.

> 
> 3) why do you use uma directly? linuxolator uses malloc from within M_LINUX. I
> personally dont mind using uma zone directly but if there is a reason for it.

The function set zone(9) is good at managing (allocating/freeing) small memory block frequently. It obtains big memory block from VM in batches
and re-allocate small memory block (e.g. structure/union of C) to caller.
When it re-allocates small memory block frequently, in contrast
VM page mapping table (i.e. GDT and LDT on x86 and x86_64) will NOT be
frequently modified, which can relax VM.

My code manages many C structures (Linux AIO context and Linux AIO
request). So I believe zone(9) is more fit for my code than malloc(9).

If you have different opinion, I will respect you.

> 
> thnx for your contribution!
> 
> roman

------------------------------------------------------------------------
                                                From Beijing, China