Closer, but still not there yet... :-(
With the dcm_module version change I described earlier, the bof file
loads and I get a tinysh prompt instead of hanging (progress!). I
can read and write my registers as expected (so far) in tinysh
(progress!). My BORPH reads and writes via the /proc/<pid>/hw/ioreg
files no longer hang (progress!), but the results are not as expected
(problem!). For example, I have a "From processor" register named
"reset". In tinysh, I can read and write to this register as
expected, but via BORPH it reads back values that appear to come from
one of my other "To processor" registers.
On Jun 21, 2010, at 7:26 , Jason Manley wrote:
I think the mkbof header might have changed somewhere along the
line for ROACH/10.1. Have you tried the old mkbof from the 7.1 flow
on the 10.1 flow?
Yes, based on your suggestion, I tried the 7.1 mkbof, but I get
similar (maybe the same) invalid results via BORPH. I also tried
enabling borph debug with "echo 1 > /proc/borph/debug", but I only
get this when loading the FPGA...
Jun 21 22:33:42 b02 kernel: binfmt_bof: b_machine=0x2,
b_elfmachine=0x14, b_version=0x6
Jun 21 22:33:42 b02 kernel: binfmt_bof: queue this file to bkexecd...
Jun 21 22:33:42 b02 kernel:
Jun 21 22:33:42 b02 kernel: binfmt_bof: read elf header at 0x1000 [7f
45 4c 46]
...and this when killing the bof process...
Jun 21 22:41:32 b02 kernel: borph: kill fpga process at region 0x0
Jun 21 22:41:32 b02 kernel: unconfigure bee2 fpga 0
FWIW, I get the same b_* values as shown above regardless of which
mkbof version I used.
Judging from /proc/borph/version, it looks like the ctl fpga is
running BORPH version 1.0-rc4. I haven't yet looked for a BORPH
version compatibility table, but I suspect that I might need an
updated ctl fpga image. How would I go about building one of those?
Any other ideas?
Thanks,
Dave
