Status:
Closed June 16, 2004
Participants:
John Jacobsen working on the DOR driver
Symptoms:
Occasionally a block of data is corrupted. A 149 byte long fragment of
the data block is placed in the buffer 149 bytes from where it
should be. See related emails in the archive for samples of
problem data.
Approach:
driver debugging (anything else will get into far more detail than
anyone will read. If they want to know, they can ask me or look
at my voluminous notes)
Solution:
fixed a race condition in the read function of the driver
Summary:
A bug was found in some real-time part of the driver code, and
fixed.
<closing date> LBNL for IceCube
Hi folks,
I have fixed a bug in the DOR driver where data received from the
DOM got corrupted under extremely flow-controlled situations. It
turns out that the problems we had with this particular test involved
both hardware and firmware. The firmware bug was fixed a few days
ago by Kalle and Thorsten and the driver bug (a subtle race between the
read function and the interrupt handler) was fixed today.
I plan to deliver a new version of the driver shortly, since some
new-pkt use is already occurring at PSL. I then plan to run the
comms test I described a week or so ago at both warm and cold
temperatures, modulo freezer availability at LBNL. Currently the
only problem I know of with the new packet format protocol is that DMA
is still broken (I think). This is a performance issue only,
though, and shouldn't interfere with testing at PSL.
Cheers,
John