atca-hotswap-devel Mailing List for Advanced TCA Hotswap

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Yang, Harold wrote:

>=20
> =20
>
>>Harold,
>>
>>Currently we are working on supporting a pre-ATCA boards/chassis with
>>alot of ATCA like qualities (switch/hub blades, disk blades,=20
>>cpu blades,
>>all controlled via IPMI ). Until ATCA boards/chassis become available,
>>this is the best we can do. We have a real customer paying=20
>>for the work,
>>so it makes sense to start there.
>>
>>What I have done (some of which isn't available on the=20
>>project page yet) is
>>forced unmount
>>forcibly unmounts a filesystem while in use (developed by my=20
>>office mate
>>Mark Huth)
>>block device removal
>>If a block device is in use, by files, by raid, or by filesystems, all
>>access to the device will be removed so it can be safely hotswapped.
>>scsi/fc hotswap
>>removes a fc/scsi device from the scsi mid layer after all access is
>>removed by block device removal
>>ga mapper
>>adds/removes device links for hotswap added/removed devices
>>cluster manager
>>coordinates IPMI events into cluster events such as adding/removing ga
>>maps and disks, etc.
>>ipmi driver
>>used by the cluster manager to receive events, get informtion about
>>devices, etc
>>sys_reboot extensions
>>the syscall sys_reboot is extended to automatically set reset=20
>>asserted,
>>power down the board, and light the blue led of the board=20
>>after running
>>the command "poweroff". Since doing this via IPMI isn't quite=20
>>standard,
>>there is a generic framework to include several types of IPMI targets
>>into the extension.
>>   =20
>>
>
>I am quite interested in your work and would like to contribute somethin=
g
>to this project. But I am not very sure if your developing code is open
>sourced? Can I access them somewhere?
>
> =20
>
Most of my work is open source. It isn't clear if my company will open
source the cluster manager itself, but all of the kernel features will
be GPL of course. I will put the patches up when I have them all sorted
out, but I'm just not ready to put them up yet. I'll post to lkml and
this list when they are up

> =20
>
>>I would like to support more bus types beyond FibreChannel,=20
>>since it is
>>unclear if FibreChannel will become a standard transport over the ATCA
>>backplane. I assume my next target will be PCI Express, but it isn't
>>clear what the hardware developers are focusing on for perhipheral
>>boards at this point.
>>   =20
>>
>
>Yes, I think bus interface of ATCA is pending. I guess maybe there will
>be several kinds of bus interface (FC, PCI-Express...) in ATCA?
>
> =20
>
>>In any of the target perhiperhals we support, I want support for
>>surprise removal (extraction without pressing the removal=20
>>button). This
>>really is a requirement, even though there is no slot removed=20
>>sensor, it
>>can be extracted by other methods. CompactPCI doesn't have too much
>>problem with this because PCI reads 0xFF. SCSI/FC on the other hand,
>>have big problems because they have error correction routines=20
>>which are
>>unneccessary if the board has been removed.
>>   =20
>>
>
>cPCI definitely has not such kind of issue, because when you remove a
>board, you MUST open the eject handler, and an ENUM signal will be=20
>triggered on cPCI bus. Driver can handle it then.
>
> =20
>
Yes cpci ejectors make alot more sense then what I have seen of atca
ejectors.

> =20
>
>>If you have any knowledge of perhipheral targets, it might get me
>>thinking about how to support those types of target hotswap devices
>>(what kernel components need to be done, what features need=20
>>to be added
>>to the cluster manager).
>>
>>We certainly need help on other target types and what types of future
>>perhipherals beyond disk nodes we will support (and how to=20
>>support). Any
>>help you can provide there would be extremely useful.
>>   =20
>>
>
>It's my pleasure. :)
>
>
> =20
>
>>The end goal of this project is to provide all of the kernel=20
>>components
>>needed to support ATCA hotswap for any ATCA perhipheral board=20
>>available.
>>I know, it sounds like a tall order, but since we are=20
>>starting with the
>>hardware developers, atleast we are on even ground.
>>
>>Thanks
>>-steve
>>
>>Yang, Harold wrote:
>>
>>   =20
>>
>>>Hi, Steve:
>>>
>>>I have studied some of the source code developed by you,
>>>and I think most of your existing work focuses on hot-swap
>>>support of fibre channel scsi disks. Though I have no hand-on
>>>experience on ATCA platforms, I DO think SCSI disk hs might
>>>just be spcial case of ATCA, right? Do you have any plan to
>>>implement a generic ATCA hs framework to support ATCA
>>>peripheral boards? (It's my personal opnion, or my fault. I just
>>>have some analogous knowledge on cPCI platforms.)
>>>
>>>What's is the your plan for this project and what can I contribute
>>>to this project?
>>>
>>>
>>>Best Regards
>>>Harold Yang
>>>
>>>Intel China Software Lab
>>>Tel: 86-21-5257-4545-1279
>>>iNet: 8752-1279
>>>
>>>
>>>
>>>-----Original Message-----
>>>From: Steven Dake [mailto:sd...@mv...]
>>>Sent: 2002=C4=EA12=D4=C215=C8=D5 6:38
>>>To: Yang, Harold
>>>Subject: Re: [Atca-hotswap-devel] hello!
>>>
>>>
>>>Harold,
>>>
>>>Sorry for the long delay in adding you as a developer to sf.
>>>
>>>Your now added Thanks!
>>>-steve
>>>
>>>Yang, Harold wrote:
>>>
>>>=20
>>>
>>>     =20
>>>
>>>>I forgot one thing: my sf.net id is harold_yang. Thanks!
>>>>
>>>>- Harold
>>>>-----Original Message-----
>>>>From: 	Steven Dake [mailto:sd...@mv...]=20
>>>>Sent:	2002=C4=EA12=D4=C23=C8=D5 1:20
>>>>To:	Yang, Harold
>>>>Cc:	'atc...@li...'
>>>>Subject:	Re: [Atca-hotswap-devel] hello!
>>>>
>>>>Harold,
>>>>
>>>>This project is just beginning to pick up steam and there=20
>>>>       =20
>>>>
>>are not yet
>>   =20
>>
>>>>alot of participants (hey, ATCA is pretty new itself).
>>>>
>>>>I've been busily working on many features of an ATCA system=20
>>>>       =20
>>>>
>>to support
>>   =20
>>
>>>>hotswap in Linux that melds with the current Linux 2.4 and Linux 2.5
>>>>architectures. I should have complete patches for the 2.4 kernel
>>>>sometime in December. I put up interim patches so people=20
>>>>       =20
>>>>
>>could see what
>>   =20
>>
>>>>we are working on.
>>>>
>>>>We welcome any help you may want to contribute to get=20
>>>>       =20
>>>>
>>things rolling or
>>   =20
>>
>>>>discuss atca hotswap in general.
>>>>
>>>>Thanks !
>>>>-steve
>>>>
>>>>Yang, Harold wrote:
>>>>
>>>>
>>>>
>>>>  =20
>>>>
>>>>       =20
>>>>
>>>>>Is there anybody active in this mailing list?
>>>>>
>>>>>
>>>>>Regards,
>>>>>Harold
>>>>>
>>>>>
>>>>>-------------------------------------------------------
>>>>>This SF.net email is sponsored by: Get the new Palm Tungsten T=20
>>>>>handheld. Power & Color in a compact size!=20
>>>>>http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0002en
>>>>>_______________________________________________
>>>>>Atca-hotswap-devel mailing list
>>>>>Atc...@li...
>>>>>https://lists.sourceforge.net/lists/listinfo/atca-hotswap-devel
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> =20
>>>>>
>>>>>    =20
>>>>>
>>>>>         =20
>>>>>
>>>>
>>>>  =20
>>>>
>>>>       =20
>>>>
>>>
>>>=20
>>>
>>>     =20
>>>
>
>
>
> =20
>