GigE NIC (Was: Who is RobertF?)
Till Straumann
strauman at slac.stanford.edu
Tue Nov 25 21:03:02 UTC 2008
Joel Sherrill wrote:
> Till Straumann wrote:
>> Robert Fu wrote:
>>
>>> Hi Joel,
>>>
>>>
>>>> Which GigE NIC is it?
>>>>
>>>
>>> Realtek RTL8111C is on the kit I'm interested most, and Intel 82574L
>>> is on
>>> other kits I'm also interested. I really wish they were supported in
>>> RTEMS.
>>>
>> I have a port of intel's 'em-6.5.3' covering many intel GigE chips
>> but the
>> 82574L is not among them. Shouldn't be very hard to port their latest
>> FreeBSD driver. em-6.9.5 seems to support the 82574L.
>>
> Till.. we missed getting your ported drivers into 4.9. Can we
> take a stab at merging them now?
> It will likely be easier for someone to at least update to the
> latest from there.
The show-stopper was (and AFAIK still is) that the pc386 interrupt
dispatcher still doesn't
support shared interrupts. Someone would have to upgrade the pc386
code first (can't be that hard; maybe ppc and 386 can be merged into
a single more generic piece implementing most of <rtems/irq.h>
and <rtems/irq-extensions.h>
T.
>
> --joel
>> FWIW
>> -- Till
>>
>>> As I mentioned before, for new hardware kits with this kind of
>>> gigabit ethernet cards
>>> and without GPIO, it seems that serial port becomes the only
>>> supported bi-directional
>>> link that can be used to interact with external systems. I think
>>> that for most cases,
>>> a RTOS must be able to retrieve information from external systems,
>>> and present
>>> processing result back to external systems, otherwise there is no
>>> need for such
>>> RTOS+Hardware. Due to limited capability of serial ports, coping
>>> with only serial
>>> communication is not a good situation.
>>>
>>> Thanks a lot,
>>> Robert Fu
>>>
>>>
>>>
>>>> Date: Mon, 24 Nov 2008 09:52:30 -0600
>>>> From: joel.sherrill at oarcorp.com
>>>> To: wendell.silva at compsisnet.com.br
>>>> CC: robert.fu at live.com; rtems-users at rtems.org
>>>> Subject: Re: RES: RES: Who is RobertF?
>>>>
>>>> Wendell Pereira da Silva wrote:
>>>>
>>>>> *De:* Robert Fu [mailto:robert.fu at live.com]
>>>>> *Enviada em:* sábado, 22 de novembro de 2008 17:19
>>>>> *Para:* Wendell Pereira da Silva
>>>>> *Cc:* rtems-users at rtems.org
>>>>> *Assunto:* RE: RES: Who is RobertF?
>>>>>
>>>>>
>>>>>
>>>>> Hi Wendell,
>>>>>
>>>>> Thanks a lot for your information about VMWare. It increases my
>>>>> confidence about VMWare.
>>>>>
>>>>>
>>>>>
>>>> Although they haven't spoke up, I know of another project that is
>>>>
>>> doing all
>>>
>>>> development inside VMWare. This lets all team members have exactly
>>>> the same host environment. When there are tool upgrades, a new OS
>>>> image is released and everyone switches to it.
>>>>
>>>>> In your short message, you actually mentioned something interested to
>>>>> me. Using Google translator, I briefly browsed COMPSIS website. With
>>>>> 19 years of history, I guess COMPSIS is a big company with a lot of
>>>>> resource and is involved in activities important to your country. I'm
>>>>> new to RTEMS, and have very limited resource: just some spare time
>>>>> outside working hours and limited personal funding, so I'll need a
>>>>>
>>> lot
>>>
>>>>> of help from RTEMS community on various RTEMS issues down the road.
>>>>>
>>>>>
>>>>>
>>>>> Not at all. J
>>>>>
>>>>>
>>>>>
>>>>> Limited available resource is one of the reasons I'm using VMWare
>>>>>
>>> now.
>>>
>>>>> In the past I abandoned some hardware (a waste to already limited
>>>>> funding), either after finding out it's unsuitable or outdated due to
>>>>> the limited time I can work on it. This time I hope using VMWare
>>>>>
>>> first
>>>
>>>>> can help me avoid some of such costs. For example, like you mentioned
>>>>> about 2 UARTs, since I already use one UART for remote debugging, I
>>>>> better select hardware kit with more than 2 UARTs (for sensor
>>>>> interface). Now with Eclipse plug-in and GDB working in my
>>>>> environment, I plan to shift focus to hardware part.
>>>>>
>>>>>
>>>>>
>>>>> Although RTEMS currently does not support the gigabit ethernet on the
>>>>> hardware kit that I have in mind, I'm still interested in how you
>>>>>
>>> made
>>>
>>>>> GDB server interface via TCP/IP work in RTEMS. To my understanding,
>>>>> RTEMS is a system with single process, and GDB server is another
>>>>> process, so how do you launch GDB server in RTEMS?
>>>>>
>>>>>
>>>> It runs as a thread. There is another way of doing things using CEXP
>>>> and dynamic loading
>>>> of your application. This ends up working a lot like people are
>>>> used to
>>>> with VxWorks.
>>>>
>>>> Network drivers are often ported over from NetBSD or FreeBSD since the
>>>> RTEMS TCP/IP
>>>> stack is an old FreeBSD one. Till Straumann has mentioned having a
>>>> kit/procedure for
>>>> making this moderately painless.
>>>>
>>>> Which GigE NIC is it?
>>>>
>>>>> Robert, what we did was a custom implementation of a GDB Server
>>>>> interface inside SIS (SPARC Instruction Simulator) through a TCP/IP
>>>>> socket. So the RTEMS program (running on the SIS) is not aware about
>>>>> this interface. RTEMS already has a glue code to implement the GDB
>>>>> interface for some targets, including i386, but not for SPARC
>>> processors.
>>>
>>>>> You also mentioned GPIO simulation with PC's parallel port. For the
>>>>> hardware kits I have in mind, some have 4-bit In/4-bit OUT GPIO, some
>>>>> don't. I checked Discrete Driver chapter in RTEMS BSP and Device
>>>>> Driver Development Guide. There are descriptions about the driver,
>>>>>
>>> but
>>>
>>>>> no concrete example code is given. Do you know where I can find
>>>>>
>>> such a
>>>
>>>>> concrete example? Also in
>>> c\src\lib\libbsp\i386\shared\comm\i386_io.h,
>>>
>>>>> there some macros like outp(port, val) and inp(port) calling assembly
>>>>> code directly. May such macros be used to access GPIO pins directly
>>>>> without using the discrete driver?
>>>>>
>>>>>
>>>>>
>>>>>
>>>> The RTEMS powerpc gdb has psim which can map UNIX shared memory into
>>>> simulated PowerPC
>>>> processor address space. When using this, you can do all IO into
>>>> shared
>>>> memory and let a UNIX
>>>> process mirror IO in and out.
>>>>
>>>> --joel
>>>>
>>>>> In the same way, we implemented the GPIO code inside SIS. That is,
>>>>> when the RTEMS program access the address of the GPIO registers it
>>>>> redirects such access to the PC’s parallel port mapped through
>>>>> VMWare.
>>>>>
>>>>> Thanks a lot for your help, and have a nice weekend!
>>>>>
>>>>>
>>>>>
>>>>> No problem.
>>>>>
>>>>>
>>>>> Robert Fu
>>>>>
>>>>> Wendell.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>> ------------------------------------------------------------------------
>>>
>>>
>>>>> Subject: RES: Who is RobertF?
>>>>> Date: Fri, 21 Nov 2008 11:01:15 -0300
>>>>> From: wendell.silva at compsisnet.com.br
>>>>> To: robert.fu at live.com
>>>>> CC: rtems-users at rtems.org
>>>>>
>>>>> Hi Robert Fu,
>>>>>
>>>>>
>>>>>
>>>>> This is just to complement your information.
>>>>>
>>>>> Here, at COMPSIS (Brazil), our development environment includes
>>>>>
>>> VMWare
>>>
>>>>> 6.0 playing Fedora 7, RTEMS 4.8.0, Eclipse 3.3 CDT and a customized
>>>>> (home made) version of SIS (with GPIO emulation with the PC’s
>>> parallel
>>>
>>>>> port and a GDB Server interface via TCP/IP – which allow us to use
>>>>> both UARTs to run our protocols). Everything has worked very well
>>>>> since its implementation (6 month ago). VMWare have worked very well
>>>>> in Windows Vista 32 in our case. None of our problems (in the
>>>>> beginning) wasn’t regarding to the VMWare.
>>>>>
>>>>>
>>>>>
>>>>> Att.
>>>>>
>>>>> Wendell P. Silva.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>> ------------------------------------------------------------------------
>>>
>>>
>>>>> *De:* rtems-users-bounces at rtems.org
>>>>> [mailto:rtems-users-bounces at rtems.org] *Em nome de *Robert Fu
>>>>> *Enviada em:* sexta-feira, 21 de novembro de 2008 04:33
>>>>> *Para:* daronchabot at gmail.com
>>>>> *Cc:* rtems-users at rtems.org
>>>>> *Assunto:* RE: Who is RobertF?
>>>>>
>>>>>
>>>>>
>>>>> Hi Daron,
>>>>>
>>>>> Thanks for clarify Option --enable-rdbg and --enable-rtems-debug.
>>>>> The RTEMS version I used is 4.9.0. I'm not sure if the problem has
>>>>> anything to do with VMWare, or some of my heap configurations
>>>>> were not right. I'll keep eye on it in the future.
>>>>>
>>>>> Thanks,
>>>>> Robert Fu
>>>>>
>>>>>
>>>>>> From: daronchabot at gmail.com
>>>>>> Subject: Re: Who is RobertF?
>>>>>> Date: Thu, 20 Nov 2008 07:10:53 -0600
>>>>>> To: robert.fu at live.com
>>>>>> CC: rtems-users at rtems.org
>>>>>>
>>>>>>
>>>>>> On 20-Nov-08, at 4:24 AM, Robert Fu wrote:
>>>>>>
>>>>>>
>>>>>>> Hi Jeol,
>>>>>>>
>>>>>>> It's me. Last night I uploaded screen shots while
>>> writing/previewing
>>>
>>>>>>> "RTEMS Eclipse Plug-In" wiki page. Although some files were
>>>>>>>
>>> uploaded,
>>>
>>>>>>> error page (HTTP 500 Internal Server Error or Could not create
>>>>>>> directory)
>>>>>>> came up for almost every uploading, and later I could not access
>>>>>>> the wiki site.
>>>>>>>
>>>>>>> I developed an Eclipse plug-in for RTEMS
>>> (http://www.rtems.com/wiki/
>>>
>>>>>>> index.php/RTEMS_Eclipse_Plug-in).
>>>>>>> Please take a look. I'm new to RTEMS, and also new to Eclipse
>>>>>>>
>>> plug-in
>>>
>>>>>>> development. This is the 1st plug-in that I ever developed. So
>>>>>>>
>>> please
>>>
>>>>>>> be patient if you encounter problems while using it.
>>>>>>>
>>>>>> Well done! Thanks for your contribution.
>>>>>>
>>>>>> However, as Sebastian pointed out the plugin is not accessible
>>>>>>
>>> on the
>>>
>>>>>> FTP site. I'd really like to check out your work.
>>>>>>
>>>>>>
>>>>>>> While working on the GDB part of the plug-in, I initially used --
>>>>>>> enable-rdbg in
>>>>>>> configure. Later I found --enable-rtems-debug was mentioned in a
>>>>>>> wiki page.
>>>>>>> When I used both --enable-rdbg and --enable-rtems-debug in
>>>>>>> configure, during startup
>>>>>>> my application hang at "_HAssert(_Heap_Is_prev_used
>>>>>>> ( next_block ));" in
>>>>>>> cpukit\score\src\heapsizeofuserarea.c. I went back to use
>>>>>>>
>>> --enable-
>>>
>>>>>>> rdbg. So my
>>>>>>> question is: which configure option should be used to enable
>>>>>>>
>>> remote
>>>
>>>>>>> debugging?
>>>>>>>
>>>>>> Option --enable-rdbg is the one to use to enable remote debugging.
>>>>>>
>>>>>> Option --enable-rtems-debug activates some instrumentation within
>>>>>> RTEMS itself (malloc statistics at least...).
>>>>>>
>>>>>> I use both switches simultaneously and have not seen the problem you
>>>>>> reported. What version of RTEMS are you using ??
>>>>>>
>>>>>>
>>>>>> -- dc
>>>>>> _______________________________________________
>>>>>> rtems-users mailing list
>>>>>> rtems-users at rtems.com
>>>>>> http://rtems.rtems.org/mailman/listinfo/rtems-users
>>>>>>
>>>>>
>>> ------------------------------------------------------------------------
>>>
>>>
>>>>> Get more done, have more fun, and stay more connected with Windows
>>>>> Mobile®. See how. <http://clk.atdmt.com/MRT/go/119642556/direct/01/>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>> ------------------------------------------------------------------------
>>>
>>>
>>>>> Color coding for safety: Windows Live Hotmail alerts you to
>>> suspicious
>>>
>>>>> email. Sign up today.
>>>>>
>>>>>
>>> <http://windowslive.com/Explore/Hotmail?ocid=TXT_TAGLM_WL_hotmail_acq_safety_112008%20>
>>>
>>>
>>>> --
>>>> Joel Sherrill, Ph.D. Director of Research & Development
>>>> joel.sherrill at OARcorp.com On-Line Applications Research
>>>> Ask me about RTEMS: a free RTOS Huntsville AL 35805
>>>> Support Available (256) 722-9985
>>>>
>>>>
>>>>
>>> ------------------------------------------------------------------------
>>>
>>> Access your email online and on the go with Windows Live Hotmail.
>>> Sign up today.
>>> <http://windowslive.com/Explore/Hotmail?ocid=TXT_TAGLM_WL_hotmail_acq_access_112008>
>>>
>>> ------------------------------------------------------------------------
>>>
>>>
>>> _______________________________________________
>>> rtems-users mailing list
>>> rtems-users at rtems.com
>>> http://rtems.rtems.org/mailman/listinfo/rtems-users
>>>
>>
>>
>
>
More information about the users
mailing list