[openib-general] [PATCH] install.sh: Cause less pain to SRP users who didn't RTFM

chas williams - CONTRACTOR chas at cmf.nrl.navy.mil
Thu Dec 14 07:49:08 PST 2006


In message <ada64cfqkpg.fsf at cisco.com>,"Roland Dreier" writes:
>I'm not sure who declared it "unsupported" and I would really like to
>know what issue(s) led to that declaration.  Your report is the first
>I've heard of anything like this, and I have to say that it seems
>pretty implausible that running a 32-bit kernel on 64-bit-capable
>hardware would be the source of problems -- if there is an issue then
>I would expect it to be something to do with the 32-bit kernel.

we saw this "problem" last week actually.  a new instal dual core
duo machine was installed with a 32-bit version of suse 10.  srp ran,
but sometimes the scsi data buffers had minor single byte errors (they
didnt appear to be at page boundaries but i am not certain about 
that).  perhaps a kmap issue?  64-bit machines running 32-bit/PAE
with more than 4GB of memory?

this is anecdotal evidence of course.  we were (are) seeing symbol
errors on the cable but i should think these errors get as far as
the srp layer.




More information about the general mailing list