Vista and NDIS 5.0, Do they play nicely together?

We have an NDIS 5.0 compliant WLAN driver sitting on top of a proprietary bus driver. Vista does
load the bus driver and when the bus driver enumerates the WLAN driver, Vista does install and load
the WLAN driver. The only problem is that the installed WLAN driver is not recognized as a network
driver. The install package works fine in Win2k and XP. We are actually "" that close to Logo on XP
so the inf files and Installshield at this point should not be in question. Does Vista install and
run NDIS 5.0 compliant WLAN drivers?



Thanks

-Ron-

AntonBassov
Mon Mar 05 17:10:00 CST 2007

> Does Vista install and
> run NDIS 5.0 compliant WLAN drivers?

AFAIK, Vista supports all network drivers that XP does, apart from FDDI,
ARCNET and serialized intermediate drivers. However, this does not come for
free. Once Vista does not provide native support for NDIS versions below 6,
NDIS has to do quite a lot to make your NDIS 5 driver work. Just to give you
an idea, once parameters to NDIS 6 functions are naturally different from
NDIS 5 ones, NDIS has to translate all parameters that it passes to your
driver from NDIS 6, which is native for Vista, to NDIS 5. The higher the link
speed, the more noticeable the resulting overhead is. Why don't you want just
to port your driver to NDIS 6???


There is one more thing worth mentioning - some NDIS 5 functions are
obsolete, but still usable under Vista, and some are not usable at all

Anton Bassov

"RonM" wrote:

> We have an NDIS 5.0 compliant WLAN driver sitting on top of a proprietary bus driver. Vista does
> load the bus driver and when the bus driver enumerates the WLAN driver, Vista does install and load
> the WLAN driver. The only problem is that the installed WLAN driver is not recognized as a network
> driver. The install package works fine in Win2k and XP. We are actually "" that close to Logo on XP
> so the inf files and Installshield at this point should not be in question. Does Vista install and
> run NDIS 5.0 compliant WLAN drivers?
>
>
>
> Thanks
>
> -Ron-
>
>
>
>
>

RonM
Mon Mar 05 17:45:34 CST 2007


Thanks for your reply.
We are going to port to NDIS 6.0. We are a small company about to release product so right
now we are going for "works on Vista." Once we release this version I get to do the NDIS 6.0 port
and move everything from WDM to WDF etc. Oh, I answered my own question. I think there is something
missing in the 5.0 to 6.0 translation. I just built our WLAN driver for NDIS 5.1 and Vista does
recognise it as a wireless LAN driver and we see the equivalent of WZC showing local access points
etc. So I have empirically discovered that our driver claiming NDIS 5.0 compliance does not get
recognised as a WLAN driver but changing the version in NdisMRegisterMiniport() makes Vista
recognise our driver as a WLAN device. Hummmm Oh well, next week NDIS 6.0

Thanks
-Ron-

"Anton Bassov" <AntonBassov@discussions.microsoft.com> wrote in message
news:<1FEE41BD-F87C-4396-9AF7-CA0C32DADBAF@microsoft.com>...
> > Does Vista install and
> > run NDIS 5.0 compliant WLAN drivers?
>
> AFAIK, Vista supports all network drivers that XP does, apart from FDDI,
> ARCNET and serialized intermediate drivers. However, this does not come for
> free. Once Vista does not provide native support for NDIS versions below 6,
> NDIS has to do quite a lot to make your NDIS 5 driver work. Just to give you
> an idea, once parameters to NDIS 6 functions are naturally different from
> NDIS 5 ones, NDIS has to translate all parameters that it passes to your
> driver from NDIS 6, which is native for Vista, to NDIS 5. The higher the link
> speed, the more noticeable the resulting overhead is. Why don't you want just
> to port your driver to NDIS 6???
>
>
> There is one more thing worth mentioning - some NDIS 5 functions are
> obsolete, but still usable under Vista, and some are not usable at all
>
> Anton Bassov
>
> "RonM" wrote:
>
> > We have an NDIS 5.0 compliant WLAN driver sitting on top of a proprietary bus driver. Vista does
> > load the bus driver and when the bus driver enumerates the WLAN driver, Vista does install and
load
> > the WLAN driver. The only problem is that the installed WLAN driver is not recognized as a
network
> > driver. The install package works fine in Win2k and XP. We are actually "" that close to Logo on
XP
> > so the inf files and Installshield at this point should not be in question. Does Vista install
and
> > run NDIS 5.0 compliant WLAN drivers?
> >
> >
> >
> > Thanks
> >
> > -Ron-
> >
> >
> >
> >
> >

AntonBassov
Mon Mar 05 19:25:54 CST 2007

> Oh well, next week NDIS 6.0

I hope your are going to like it. Up to this point, I just love NDIS 6 - it
is so much more logical and simple than earlier NDIS versions ( only
packet-based sends, receipts and indications are allowed - once no
media-specific indications are allowed, you don't have to worry about
abomination known as partial indication;
the very idea of NDIS LWF is so much more logical, compared to that of NDIS
5 IM filter; etc). Judging from my experience with NDIS 6, it seems to be
trouble-free - it looks like Alireza and his team did a good job......

Anton Bassov

"RonM" wrote:

>
> Thanks for your reply.
> We are going to port to NDIS 6.0. We are a small company about to release product so right
> now we are going for "works on Vista." Once we release this version I get to do the NDIS 6.0 port
> and move everything from WDM to WDF etc. Oh, I answered my own question. I think there is something
> missing in the 5.0 to 6.0 translation. I just built our WLAN driver for NDIS 5.1 and Vista does
> recognise it as a wireless LAN driver and we see the equivalent of WZC showing local access points
> etc. So I have empirically discovered that our driver claiming NDIS 5.0 compliance does not get
> recognised as a WLAN driver but changing the version in NdisMRegisterMiniport() makes Vista
> recognise our driver as a WLAN device. Hummmm Oh well, next week NDIS 6.0
>
> Thanks
> -Ron-
>
> "Anton Bassov" <AntonBassov@discussions.microsoft.com> wrote in message
> news:<1FEE41BD-F87C-4396-9AF7-CA0C32DADBAF@microsoft.com>...
> > > Does Vista install and
> > > run NDIS 5.0 compliant WLAN drivers?
> >
> > AFAIK, Vista supports all network drivers that XP does, apart from FDDI,
> > ARCNET and serialized intermediate drivers. However, this does not come for
> > free. Once Vista does not provide native support for NDIS versions below 6,
> > NDIS has to do quite a lot to make your NDIS 5 driver work. Just to give you
> > an idea, once parameters to NDIS 6 functions are naturally different from
> > NDIS 5 ones, NDIS has to translate all parameters that it passes to your
> > driver from NDIS 6, which is native for Vista, to NDIS 5. The higher the link
> > speed, the more noticeable the resulting overhead is. Why don't you want just
> > to port your driver to NDIS 6???
> >
> >
> > There is one more thing worth mentioning - some NDIS 5 functions are
> > obsolete, but still usable under Vista, and some are not usable at all
> >
> > Anton Bassov
> >
> > "RonM" wrote:
> >
> > > We have an NDIS 5.0 compliant WLAN driver sitting on top of a proprietary bus driver. Vista does
> > > load the bus driver and when the bus driver enumerates the WLAN driver, Vista does install and
> load
> > > the WLAN driver. The only problem is that the installed WLAN driver is not recognized as a
> network
> > > driver. The install package works fine in Win2k and XP. We are actually "" that close to Logo on
> XP
> > > so the inf files and Installshield at this point should not be in question. Does Vista install
> and
> > > run NDIS 5.0 compliant WLAN drivers?
> > >
> > >
> > >
> > > Thanks
> > >
> > > -Ron-
> > >
> > >
> > >
> > >
> > >
>
>
>

Calvin
Mon Mar 05 20:42:20 CST 2007

Anton,

I make a living of writing nd4x, nd5x and nd6x for the best selling gbe. My
experience is: for IM or filters, maybe ndis6 is simpler but if you are
working on a ndis6x miniport, you are going to hate it. I personally like
ndis5x better. Yes, networking performance can be better with NDIS6x if the
everyone in the stack play nicely.

However, it pays my bills so I can't complain too much.

--
Calvin Guan (expiring DDK MVP)
Sr. Staff Engineer
NetXtreme NTX Miniport
Broadcom Corporation
Connecting Everything(r)

"Anton Bassov" <AntonBassov@discussions.microsoft.com> wrote in message
news:DEAD99AD-62C2-4BF8-84C5-CFC52AB513C7@microsoft.com...
>> Oh well, next week NDIS 6.0
>
> I hope your are going to like it. Up to this point, I just love NDIS 6 -
> it
> is so much more logical and simple than earlier NDIS versions ( only
> packet-based sends, receipts and indications are allowed - once no
> media-specific indications are allowed, you don't have to worry about
> abomination known as partial indication;
> the very idea of NDIS LWF is so much more logical, compared to that of
> NDIS
> 5 IM filter; etc). Judging from my experience with NDIS 6, it seems to be
> trouble-free - it looks like Alireza and his team did a good job......
>
> Anton Bassov
>

pavel_a
Tue Mar 06 00:53:03 CST 2007

"Calvin Guan" wrote:
>My > experience is: for IM or filters, maybe ndis6 is simpler but if you are
> working on a ndis6x miniport, you are going to hate it. I personally like
> ndis5x better. Yes, networking performance can be better with NDIS6x if the
> everyone in the stack play nicely.

Why, Calvin? NDIS6 it so complex, or there are some showstopper bugs?


Regards,
--PA

Stephan
Tue Mar 06 04:11:13 CST 2007

On Mar 6, 3:42 am, "Calvin Guan" <h...@nospam.broadcom.com> wrote:

> I make a living of writing nd4x, nd5x and nd6x for the best selling gbe.

Best selling? Umm, how about the Marvell "Yukon" GigE chipsets?

[http://www.marvell.com/products/pcconn/yukon/]

Stephan

AntonBassov
Tue Mar 06 07:29:03 CST 2007

Pavel,

> >My > experience is: for IM or filters, maybe ndis6 is simpler but if you are
> > working on a ndis6x miniport, you are going to hate it. I personally like
> > ndis5x better. Yes, networking performance can be better with NDIS6x if the
> > everyone in the stack play nicely.
>
> Why, Calvin? NDIS6 it so complex, or there are some showstopper bugs?

I think that Calvin and myself speak about exactly the same thing, but from
the different perspectives - various "possibilities" in NDIS versions < 6
that may be convenient for miniport developers (due to the specifics of
network cards that they
interface) are quite annoying for everyone higher on the stack. For example,
partial
data indication that may be suitable for some particular card is, from NDIS IM
developer, just abominable. NDIS 6 is more "standard" than earlier NDIS
versions, which is good news for filter/protocol developers and bad news for
miniport ones

In addition to that, NDIS 6 does not support serialized miniports, you have to
handle the possibility of your miniport being paused and restarted, etc.....


Anton Bassov


"Pavel A." wrote:

> "Calvin Guan" wrote:
> >My > experience is: for IM or filters, maybe ndis6 is simpler but if you are
> > working on a ndis6x miniport, you are going to hate it. I personally like
> > ndis5x better. Yes, networking performance can be better with NDIS6x if the
> > everyone in the stack play nicely.
>
> Why, Calvin? NDIS6 it so complex, or there are some showstopper bugs?
>
>
> Regards,
> --PA

Calvin
Tue Mar 06 11:25:28 CST 2007

Oh, yes. We also know who wins/loses which tire one, two accounts every
year:)
I knew what Marvell does. Their GPHY is pretty good.


"Stephan Wolf [MVP]" <stewo68@hotmail.com> wrote in message
news:1173175872.883613.290580@t69g2000cwt.googlegroups.com...
> On Mar 6, 3:42 am, "Calvin Guan" <h...@nospam.broadcom.com> wrote:
>
>> I make a living of writing nd4x, nd5x and nd6x for the best selling gbe.
>
> Best selling? Umm, how about the Marvell "Yukon" GigE chipsets?
>
> [http://www.marvell.com/products/pcconn/yukon/]
>
> Stephan
>

Calvin
Tue Mar 06 12:24:14 CST 2007

to add what Anton said a little bit...

NDIS6x introduces some performance features such as RSS, GSO, TCP chimney
offload, RDMA (not quite standardized yet) and some other cool stuff you are
going to see shortly (can't say much about that because of the NDA). Let
alone a complexity in the silicon implementations and the miniport driver,
it would make writing smart IM drivers more challenging than ever, i.e. load
balancing and/or fail-overing a TCP connection on a TCP chimney+RSS capable
NIC. In the past, a server would be completely busy while servicing multiple
gb port or one 10Gb port, but now with smart nic, people can put more in the
box. load balancing becomes very important for server. If I was a
consultant, it'd be a good chance to make money.

--
Calvin Guan (expiring DDK MVP)
Sr. Staff Engineer
NetXtreme NTX Miniport
Broadcom Corporation
Connecting Everything(r)

"Anton Bassov" <AntonBassov@discussions.microsoft.com> wrote in message
news:5BDAEC88-02E0-4B9B-AB44-F724F08BABB5@microsoft.com...
> Pavel,
>
>> >My > experience is: for IM or filters, maybe ndis6 is simpler but if you
>> >are
>> > working on a ndis6x miniport, you are going to hate it. I personally
>> > like
>> > ndis5x better. Yes, networking performance can be better with NDIS6x if
>> > the
>> > everyone in the stack play nicely.
>>
>> Why, Calvin? NDIS6 it so complex, or there are some showstopper bugs?
>
> I think that Calvin and myself speak about exactly the same thing, but
> from
> the different perspectives - various "possibilities" in NDIS versions < 6
> that may be convenient for miniport developers (due to the specifics of
> network cards that they
> interface) are quite annoying for everyone higher on the stack. For
> example,
> partial
> data indication that may be suitable for some particular card is, from
> NDIS IM
> developer, just abominable. NDIS 6 is more "standard" than earlier NDIS
> versions, which is good news for filter/protocol developers and bad news
> for
> miniport ones
>
> In addition to that, NDIS 6 does not support serialized miniports, you
> have to
> handle the possibility of your miniport being paused and restarted,
> etc.....
>
>
> Anton Bassov
>
>
> "Pavel A." wrote:
>
>> "Calvin Guan" wrote:
>> >My > experience is: for IM or filters, maybe ndis6 is simpler but if you
>> >are
>> > working on a ndis6x miniport, you are going to hate it. I personally
>> > like
>> > ndis5x better. Yes, networking performance can be better with NDIS6x if
>> > the
>> > everyone in the stack play nicely.
>>
>> Why, Calvin? NDIS6 it so complex, or there are some showstopper bugs?
>>
>>
>> Regards,
>> --PA

Maxim
Tue Mar 06 14:39:10 CST 2007

What about Intel's GigE chip?

--
Maxim Shatskih, Windows DDK MVP
StorageCraft Corporation
maxim@storagecraft.com
http://www.storagecraft.com

"Calvin Guan" <hguan@nospam.broadcom.com> wrote in message
news:e5GB4RBYHHA.992@TK2MSFTNGP04.phx.gbl...
> Oh, yes. We also know who wins/loses which tire one, two accounts every
> year:)
> I knew what Marvell does. Their GPHY is pretty good.
>
>
> "Stephan Wolf [MVP]" <stewo68@hotmail.com> wrote in message
> news:1173175872.883613.290580@t69g2000cwt.googlegroups.com...
> > On Mar 6, 3:42 am, "Calvin Guan" <h...@nospam.broadcom.com> wrote:
> >
> >> I make a living of writing nd4x, nd5x and nd6x for the best selling gbe.
> >
> > Best selling? Umm, how about the Marvell "Yukon" GigE chipsets?
> >
> > [http://www.marvell.com/products/pcconn/yukon/]
> >
> > Stephan
> >
>
>

Calvin
Tue Mar 06 15:25:45 CST 2007

I personally keep track of the pros and cons of the US based major players
in case I was letting go one day-:(. I know quite a bit of each others but I
can't make comments in public forum like this.

--
Calvin Guan (expiring DDK MVP)
Sr. Staff Engineer
NetXtreme NTX Miniport
Broadcom Corporation
Connecting Everything(r)

"Maxim S. Shatskih" <maxim@storagecraft.com> wrote in message
news:eZB0r9CYHHA.3772@TK2MSFTNGP06.phx.gbl...
> What about Intel's GigE chip?
>
> --
> Maxim Shatskih, Windows DDK MVP
> StorageCraft Corporation
> maxim@storagecraft.com
> http://www.storagecraft.com
>
> "Calvin Guan" <hguan@nospam.broadcom.com> wrote in message
> news:e5GB4RBYHHA.992@TK2MSFTNGP04.phx.gbl...
>> Oh, yes. We also know who wins/loses which tire one, two accounts every
>> year:)
>> I knew what Marvell does. Their GPHY is pretty good.
>>
>>
>> "Stephan Wolf [MVP]" <stewo68@hotmail.com> wrote in message
>> news:1173175872.883613.290580@t69g2000cwt.googlegroups.com...
>> > On Mar 6, 3:42 am, "Calvin Guan" <h...@nospam.broadcom.com> wrote:
>> >
>> >> I make a living of writing nd4x, nd5x and nd6x for the best selling
>> >> gbe.
>> >
>> > Best selling? Umm, how about the Marvell "Yukon" GigE chipsets?
>> >
>> > [http://www.marvell.com/products/pcconn/yukon/]
>> >
>> > Stephan
>> >
>>
>>
>

Stephan
Wed Mar 07 07:14:51 CST 2007

On Mar 6, 9:39 pm, "Maxim S. Shatskih" <m...@storagecraft.com> wrote:
> What about Intel's GigE chip?

AFAIK, even various Intel motherboards have the Marvell "Yukon"
Gigabit Ethernet chipset onboard, see e.g.

http://mysearch.intel.com/corporate/default.aspx?culture=en-US&q=yukon

Stephan

Maxim
Wed Mar 07 14:03:01 CST 2007

> AFAIK, even various Intel motherboards have the Marvell "Yukon"
> Gigabit Ethernet chipset onboard, see e.g.

I have an Intel card of VEN_8086&DEV_100E&SUBSYS_002E8086&REV_02

I don't think it is Marvell.

--
Maxim Shatskih, Windows DDK MVP
StorageCraft Corporation
maxim@storagecraft.com
http://www.storagecraft.com

Calvin
Wed Mar 07 14:59:05 CST 2007


INTC did put MRVL's chips on some of their mobo (for good reasons). If one's
in this area, s/he would know more funny deals(like the GPHY core - the key
technology in Gbe) between INTC and MRVL.

I think INTC's been doing a great job overall although they've made 2 big
mistakes in the Gbe business but they are picking up and coming back strong.
On the other hand, the relation between INTC and BRCM has been traditionally
extremely hostile since BRCM turned into teenage. I believe INTC would
rather have no solution than asking anything from BRCM:)

well, this is a funny game as new technologies advance. Who will do better
in the next couple years? only time can tell...

--
Calvin Guan (expiring DDK MVP)
Sr. Staff Engineer
NetXtreme NTX Miniport
Broadcom Corporation
Connecting Everything(r)

"Maxim S. Shatskih" <maxim@storagecraft.com> wrote in message
news:%23moWKOPYHHA.3928@TK2MSFTNGP03.phx.gbl...
>> AFAIK, even various Intel motherboards have the Marvell "Yukon"
>> Gigabit Ethernet chipset onboard, see e.g.
>
> I have an Intel card of VEN_8086&DEV_100E&SUBSYS_002E8086&REV_02
>
> I don't think it is Marvell.
>
> --
> Maxim Shatskih, Windows DDK MVP
> StorageCraft Corporation
> maxim@storagecraft.com
> http://www.storagecraft.com
>

Stephan
Thu Mar 08 04:23:17 CST 2007

On Mar 7, 9:03 pm, "Maxim S. Shatskih" <m...@storagecraft.com> wrote:
> I have an Intel card of VEN_8086&DEV_100E&SUBSYS_002E8086&REV_02
>
> I don't think it is Marvell.

0x8086 is Intel's PCI vendor code (see e.g. http://www.pcidatabase.com/vendors.php?sort=id).

0x1148 is Marvell Germany (former SysKonnect, see e.g.
http://www.pcidatabase.com/vendor_details.php?id=861).

Stephan
All speaking for myself.

Stephan
Thu Mar 08 04:26:13 CST 2007

Couldn't agree more on what you say, Calvin.

Cheers, Stephan
---
On Mar 7, 9:59 pm, "Calvin Guan" <h...@nospam.broadcom.com> wrote:
> INTC did put MRVL's chips on some of their mobo (for good reasons). If one's
> in this area, s/he would know more funny deals(like the GPHY core - the key
> technology in Gbe) between INTC and MRVL.
>
> I think INTC's been doing a great job overall although they've made 2 big
> mistakes in the Gbe business but they are picking up and coming back strong.
> On the other hand, the relation between INTC and BRCM has been traditionally
> extremely hostile since BRCM turned into teenage. I believe INTC would
> rather have no solution than asking anything from BRCM:)
>
> well, this is a funny game as new technologies advance. Who will do better
> in the next couple years? only time can tell...
>
> --
> Calvin Guan (expiring DDK MVP)
> Sr. Staff Engineer
> NetXtreme NTX Miniport
> Broadcom Corporation
> Connecting Everything(r)
>
> "Maxim S. Shatskih" <m...@storagecraft.com> wrote in messagenews:%23moWKOPYHHA.3928@TK2MSFTNGP03.phx.gbl...
>
>
>
> >> AFAIK, even various Intel motherboards have the Marvell "Yukon"
> >> Gigabit Ethernet chipset onboard, see e.g.
>
> > I have an Intel card of VEN_8086&DEV_100E&SUBSYS_002E8086&REV_02
>
> > I don't think it is Marvell.
>
> > --
> > Maxim Shatskih, Windows DDK MVP
> > StorageCraft Corporation
> > m...@storagecraft.com
> >http://www.storagecraft.com- Hide quoted text -
>
> - Show quoted text -

Alireza
Tue Mar 13 21:04:21 CDT 2007

Calvin, your comments on LBFO IM drivers notwithstanding, I still don't
understand why somebody writing an NDIS 6 miniport driver would hate it! or
in this day and age would prefer NDIS 5 model.

-ali

--
This posting is provided "AS IS" with no warranties, and confers no rights.

"Calvin Guan" <hguan@nospam.broadcom.com> wrote in message
news:%23%23TPryBYHHA.2316@TK2MSFTNGP04.phx.gbl...
> to add what Anton said a little bit...
>
> NDIS6x introduces some performance features such as RSS, GSO, TCP chimney
> offload, RDMA (not quite standardized yet) and some other cool stuff you
> are going to see shortly (can't say much about that because of the NDA).
> Let alone a complexity in the silicon implementations and the miniport
> driver, it would make writing smart IM drivers more challenging than ever,
> i.e. load balancing and/or fail-overing a TCP connection on a TCP
> chimney+RSS capable NIC. In the past, a server would be completely busy
> while servicing multiple gb port or one 10Gb port, but now with smart nic,
> people can put more in the box. load balancing becomes very important for
> server. If I was a consultant, it'd be a good chance to make money.
>
> --
> Calvin Guan (expiring DDK MVP)
> Sr. Staff Engineer
> NetXtreme NTX Miniport
> Broadcom Corporation
> Connecting Everything(r)
>
> "Anton Bassov" <AntonBassov@discussions.microsoft.com> wrote in message
> news:5BDAEC88-02E0-4B9B-AB44-F724F08BABB5@microsoft.com...
>> Pavel,
>>
>>> >My > experience is: for IM or filters, maybe ndis6 is simpler but if
>>> >you are
>>> > working on a ndis6x miniport, you are going to hate it. I personally
>>> > like
>>> > ndis5x better. Yes, networking performance can be better with NDIS6x
>>> > if the
>>> > everyone in the stack play nicely.
>>>
>>> Why, Calvin? NDIS6 it so complex, or there are some showstopper bugs?
>>
>> I think that Calvin and myself speak about exactly the same thing, but
>> from
>> the different perspectives - various "possibilities" in NDIS versions < 6
>> that may be convenient for miniport developers (due to the specifics of
>> network cards that they
>> interface) are quite annoying for everyone higher on the stack. For
>> example,
>> partial
>> data indication that may be suitable for some particular card is, from
>> NDIS IM
>> developer, just abominable. NDIS 6 is more "standard" than earlier NDIS
>> versions, which is good news for filter/protocol developers and bad news
>> for
>> miniport ones
>>
>> In addition to that, NDIS 6 does not support serialized miniports, you
>> have to
>> handle the possibility of your miniport being paused and restarted,
>> etc.....
>>
>>
>> Anton Bassov
>>
>>
>> "Pavel A." wrote:
>>
>>> "Calvin Guan" wrote:
>>> >My > experience is: for IM or filters, maybe ndis6 is simpler but if
>>> >you are
>>> > working on a ndis6x miniport, you are going to hate it. I personally
>>> > like
>>> > ndis5x better. Yes, networking performance can be better with NDIS6x
>>> > if the
>>> > everyone in the stack play nicely.
>>>
>>> Why, Calvin? NDIS6 it so complex, or there are some showstopper bugs?
>>>
>>>
>>> Regards,
>>> --PA
>
>

Calvin
Tue Mar 13 22:47:09 CDT 2007

Ali, in ndis6 miniport, *rigorous* synchronization between driver states and
traffic path *without* hurting MP performance requires significant amount of
hard work and experiences, especially in 1 and 10Gbe. We do see the
performance boost with the 6x stack in point to multipoint Chariot test.
Good work, guys!

Speaking for myself only..
--
Calvin Guan (expiring DDK MVP)
Sr. Staff Engineer
NetXtreme NTX Miniport
Broadcom Corporation
Connecting Everything(r)

"Alireza Dabagh [MS]" <alid@online.microsoft.com> wrote in message
news:45f75825$1@news.microsoft.com...
> Calvin, your comments on LBFO IM drivers notwithstanding, I still don't
> understand why somebody writing an NDIS 6 miniport driver would hate it!
> or in this day and age would prefer NDIS 5 model.
>
> -ali
>
> --
> This posting is provided "AS IS" with no warranties, and confers no
> rights.
>
> "Calvin Guan" <hguan@nospam.broadcom.com> wrote in message
> news:%23%23TPryBYHHA.2316@TK2MSFTNGP04.phx.gbl...
>> to add what Anton said a little bit...
>>
>> NDIS6x introduces some performance features such as RSS, GSO, TCP chimney
>> offload, RDMA (not quite standardized yet) and some other cool stuff you
>> are going to see shortly (can't say much about that because of the NDA).
>> Let alone a complexity in the silicon implementations and the miniport
>> driver, it would make writing smart IM drivers more challenging than
>> ever, i.e. load balancing and/or fail-overing a TCP connection on a TCP
>> chimney+RSS capable NIC. In the past, a server would be completely busy
>> while servicing multiple gb port or one 10Gb port, but now with smart
>> nic, people can put more in the box. load balancing becomes very
>> important for server. If I was a consultant, it'd be a good chance to
>> make money.
>>
>> --
>> Calvin Guan (expiring DDK MVP)
>> Sr. Staff Engineer
>> NetXtreme NTX Miniport
>> Broadcom Corporation
>> Connecting Everything(r)
>>
>> "Anton Bassov" <AntonBassov@discussions.microsoft.com> wrote in message
>> news:5BDAEC88-02E0-4B9B-AB44-F724F08BABB5@microsoft.com...
>>> Pavel,
>>>
>>>> >My > experience is: for IM or filters, maybe ndis6 is simpler but if
>>>> >you are
>>>> > working on a ndis6x miniport, you are going to hate it. I personally
>>>> > like
>>>> > ndis5x better. Yes, networking performance can be better with NDIS6x
>>>> > if the
>>>> > everyone in the stack play nicely.
>>>>
>>>> Why, Calvin? NDIS6 it so complex, or there are some showstopper bugs?
>>>
>>> I think that Calvin and myself speak about exactly the same thing, but
>>> from
>>> the different perspectives - various "possibilities" in NDIS versions <
>>> 6
>>> that may be convenient for miniport developers (due to the specifics of
>>> network cards that they
>>> interface) are quite annoying for everyone higher on the stack. For
>>> example,
>>> partial
>>> data indication that may be suitable for some particular card is, from
>>> NDIS IM
>>> developer, just abominable. NDIS 6 is more "standard" than earlier NDIS
>>> versions, which is good news for filter/protocol developers and bad news
>>> for
>>> miniport ones
>>>
>>> In addition to that, NDIS 6 does not support serialized miniports, you
>>> have to
>>> handle the possibility of your miniport being paused and restarted,
>>> etc.....
>>>
>>>
>>> Anton Bassov
>>>
>>>
>>> "Pavel A." wrote:
>>>
>>>> "Calvin Guan" wrote:
>>>> >My > experience is: for IM or filters, maybe ndis6 is simpler but if
>>>> >you are
>>>> > working on a ndis6x miniport, you are going to hate it. I personally
>>>> > like
>>>> > ndis5x better. Yes, networking performance can be better with NDIS6x
>>>> > if the
>>>> > everyone in the stack play nicely.
>>>>
>>>> Why, Calvin? NDIS6 it so complex, or there are some showstopper bugs?
>>>>
>>>>
>>>> Regards,
>>>> --PA
>>
>>
>