Re: Is SNP broken?
- From: "Pavel A." <pavel_a@xxxxxxxxxxxxxxx>
- Date: Mon, 19 May 2008 21:55:08 +0300
I've asked this because several approaches are developing to accelerate
tcp/ip processing - varying from filtering and "pre-cooking" L2 packets before
they reach the OS (which uses standard drivers and IP stack), to
solutions that intercept sockets API in usermode and completely
bypass everything below it.
So if the 3rd party (or even MS) software can not cope with MS own IP
acceleration solution, are all these attempts doomed too?
Regards,
--PA
"PCAUSA" <pcausa@xxxxxxxxx> wrote in message news:b33b736e-3518-469c-ab9c-cec51006fb92@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Surely it doesn't make much sense. If the SNP is added and depends on.
NDIS 5.2 miniports, then surely it should be bright enough to detect
whether NDIS 5.2 support is present. IIRC, there are OIDs for this
purpose.
Thomas F. Divine
On May 15, 7:41 pm, "Pavel A." <pave...@xxxxxxxxxxxxxxx> wrote:Hmm. We could assume that there is a ndistest suite for TOE, and all
shipping 3rd party drivers
have passed the formal and functional tests, and also some interop testing.
I'd rather agree with Thomas, but the KB article does not say anything
about any other software, besides of MS own components (dhcp, rdp client...)
--PA
"GNR" <G...@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:32BE2BB2-3687-40CF-89E2-B0EDACEE1B56@xxxxxxxxxxxxxxxx> No, I think it is the other way.
> In the workarounds section, the "Method 1" says "The driver must meet
> Network Driver Interface Specification (NDIS) 5.2 or a later version of
> this
> specification."
> This means if the Network Card has NDIS 5.2 driver, it works fine and > you
> don't need to turn off SNP.
> Am I reading it right?
> --
> --GNR
> "PCAUSA" wrote:
>> On May 15, 5:58 pm, "Pavel A." <pave...@xxxxxxxxxxxxxxx> wrote:
>> > Just noticed this KB article -http://support.microsoft.com/kb/948496
>> > So what does this mean? are RSS and toe broken somewhere in NDIS or >> > IP
>> > stack, or the issues are caused by netcard drivers?
>> > Regards,
>> > --PA
>> Pavel,
>> I think that this is saying:
>> IF you have TCP/IP Offload-enabled network adapter, then there is a
>> lot of software that doesn't know how to deal with TOE adapters with
>> the SNP. In fact, some TOE operations are implemented using a TCP/Ip
>> engine in the adapter itself and may not even be observable by various
>> filters. Hence - they break.
>> If you want the features of the non-TOE-aware services, then turn off
>> SNP.
>> Your thoughts?
>> Thomas F. Divine
- References:
- Is SNP broken?
- From: Pavel A.
- Re: Is SNP broken?
- From: PCAUSA
- Re: Is SNP broken?
- From: GNR
- Re: Is SNP broken?
- From: Pavel A.
- Re: Is SNP broken?
- From: PCAUSA
- Is SNP broken?
- Prev by Date: Re: finding out application dependencies;
- Next by Date: Re: Where to read messages about VSS writer?
- Previous by thread: Re: Is SNP broken?
- Next by thread: USB Function application testing
- Index(es):
Relevant Pages
|
