From: Allen Curtis <hidden> Date: 2002-08-22 03:47:49
Perhaps this will jog a couple brain cells.
I retrieved 2.4.19 final from BitKeeper and have successfully gotten PCI
working with the SYM53C8xx version 2 driver. (No modifications were made to
the PCI setup)
However it only works if you select the "normal I/O" option with the
SYM53C8xx version 2 driver. SYM53C8xx version 1 does not work. Version 2
does not work if the I/O option is not selected.
It sounds like a byte-swap issue with the memory interface. Has anyone else
run into this? I am using an 8260 not a Mac/PowerPC. Perhaps there is
something there?
TIA
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
I retrieved 2.4.19 final from BitKeeper and have successfully gotten PCI
working with the SYM53C8xx version 2 driver. (No modifications were made to
the PCI setup)
However it only works if you select the "normal I/O" option with the
SYM53C8xx version 2 driver. SYM53C8xx version 1 does not work. Version 2
does not work if the I/O option is not selected.
It might help if you described "does not work" in more detail... could
you paste error messages?
It sounds like a byte-swap issue with the memory interface. Has anyone else
run into this? I am using an 8260 not a Mac/PowerPC. Perhaps there is
something there?
I've seen a problem on PReP where although the SCSI card is addressed at
0xc0000000, the BAR is 0x0 (because of the host bridge offset). From
memory, the sym2 driver checks the BAR and if it's 0 errors out.
Removing that check or enabling "normal IO" worked fine. Of course, this
might not be your problem, but without actual error messages...
-Hollis
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
From: Allen Curtis <hidden> Date: 2002-08-23 14:35:19
quoted
However it only works if you select the "normal I/O" option with the
SYM53C8xx version 2 driver. SYM53C8xx version 1 does not work. Version 2
does not work if the I/O option is not selected.
It might help if you described "does not work" in more detail... could
you paste error messages?
I've seen a problem on PReP where although the SCSI card is addressed at
0xc0000000, the BAR is 0x0 (because of the host bridge offset). From
memory, the sym2 driver checks the BAR and if it's 0 errors out.
Removing that check or enabling "normal IO" worked fine. Of course, this
might not be your problem, but without actual error messages...
This sounds promising. Is this check in both versions of the driver? I don't
remember a BAR check. I thought it just ioremapped() the region. Could you
provide a reference? (line#)
THX
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
However it only works if you select the "normal I/O" option with the
SYM53C8xx version 2 driver. SYM53C8xx version 1 does not work. Version 2
does not work if the I/O option is not selected.
It might help if you described "does not work" in more detail... could
you paste error messages?
That's a lot more than I want to think about. :) Of course I'm assuming
you have your PCI config set up correctly - in that case the below is an
error I've seen.
quoted
I've seen a problem on PReP where although the SCSI card is addressed at
0xc0000000, the BAR is 0x0 (because of the host bridge offset). From
memory, the sym2 driver checks the BAR and if it's 0 errors out.
Removing that check or enabling "normal IO" worked fine. Of course, this
might not be your problem, but without actual error messages...
This sounds promising. Is this check in both versions of the driver? I don't
remember a BAR check. I thought it just ioremapped() the region. Could you
provide a reference? (line#)
I didn't notice you were still having problems with version 1 of the
driver; I thought you had resolved that. IME that almost always means
your PCI config is not correct.
The BAR check in sym2 is around line 2531 of sym53c8xx_2/sym_misc.c .
-Hollis
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
From: Allen Curtis <hidden> Date: 2002-08-23 15:08:53
I didn't notice you were still having problems with version 1 of the
driver; I thought you had resolved that. IME that almost always means
your PCI config is not correct.
I have used a peek/poke utility to access the SYM53C8xx via PCI memory and
I/O. Both seem to be working fine. Obviously since the sym_2 driver works in
at least one configuration, the translation from the PCI bus back to CPU
memory is also correct. Still feels like an endian issue to me.
Thoughts?
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
I didn't notice you were still having problems with version 1 of the
driver; I thought you had resolved that. IME that almost always means
your PCI config is not correct.
I have used a peek/poke utility to access the SYM53C8xx via PCI memory and
I/O. Both seem to be working fine. Obviously since the sym_2 driver works in
at least one configuration, the translation from the PCI bus back to CPU
memory is also correct.
Well, I would start by finding the error ("CACHE TEST FAILED: DMA error
(dstat=0x81)."), see under what conditions it occurs, and work backwards
from there.
It may really be that "normal IO" uses inb/outb. Do other drivers that
use ioremap/readb/writeb work on your system?
Still feels like an endian issue to me.
I'm not sure it is an endian problem because as I mentioned I've used
the sym2 driver on PPC without problems.
-Hollis
** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/