Re: TCP limitations (was Re: ISCSI: Urgent Flag requirement violates TCP.)

[julian_satran@il.ibm.com]

JP,

I think this assumption is incorrect even with many local (parallel bus)
attached drives.
If the drive is unbuffered or you have a parity error on the bus while data
is DMAed you will end up with
a dirty buffer.

Julo

Raghavendra Rao <jp.raghavendra@india.sun.com> on 27/11/2000 19:25:43

Please respond to Raghavendra Rao <jp.raghavendra@india.sun.com>

To:   ips@ece.cmu.edu
cc:
Subject:  Re: TCP limitations (was Re: ISCSI: Urgent Flag requirement
      violates  TCP.)





> > >
> > > Store the message in user memory where it
> > > belongs, no matter what. No NIC memory would
> > > ever be needed ...
> > >
> > >
> > > Much superior...
> > >

Is there even a remote possibility that some application would break if
it reuses the buffer with an assumption that its original contents of
the buffer are unaltered when an I/O error is reported ?

In other words,

     bzero(buffer, sizeof (buffer))
     error = disk_read(disk_handle, buffer, sizeof (buffer))
     if (error) {
          do something with an assumption that buffer is still zero'ed
     }

Regardless of this situation, one could simply document this side effect
for
applications.

-JP