|PCIIDE(4)||Device Drivers Manual||PCIIDE(4)|
pciide — PCI IDE
pciide* at pci? flags 0x0000
pciide* at jmb? flags 0x0000
atapiscsi* at pciide? flags 0x0000
wd* at pciide? flags 0x0000
pciide driver supports the PCI IDE
controllers as specified in the "PCI IDE controller specification,
revision 1.0" draft, as well as the Serial ATA controllers working
either in parallel ATA compatibility mode or in native serial mode. It
provides the core functions for the wd(4) and
atapiscsi(4) drivers. This driver
includes specific, enhanced support for the CMD Tech PCI0640 controller and
IDE DMA/UltraDMA support for the following PCI IDE controllers:
Some of these controllers are only available in multifunction PCI chips.
0x0001 flag forces the
pciide driver to use DMA when there is no explicit
DMA mode setting support for the controller but DMA is present. If the BIOS
didn't configure the controller properly, this can cause a machine hang.
There's no way to know reliably if a CMD064x controller is enabled
or not. If the driver finds a PCI064x, it will assume it is enabled unless
PCIIDE_CMD064x_DISABLE is specified in the kernel
config file. This will be a problem only if the controller has been disabled
in the BIOS and another controller has been installed which uses the ISA
legacy I/O ports and interrupts.
The CMD0646U is reportedly unable to operate properly in UltraDMA
mode. UltraDMA will not be enabled on this controller unless
PCIIDE_CMD0646U_ENABLEUDMA is specified in the
kernel config file.
For proper operation of UltraDMA 3 and higher modes a 40-pin,
80-conductor cable must be used. On some controllers, the
pciide driver can probe the chip and see if it
detects that an 80-conductor or 40-conductor cable was used. If not, it will
down-grade to UltraDMA 2 mode. On other controllers, no such capability
exists, and the driver assumes you are using an 80-conductor cable. If you
are getting errors about corrupted data, check to make sure you are using
the correct cable. An 80-conductor cable is recommended for any IDE
installation, not just ones using faster UltraDMA modes, because it will
decrease electrical noise and increase data reliability.
The AMD756 chip revision D2 has a bug affecting DMA (but not Ultra-DMA) modes. The workaround documented by AMD is to not use DMA on any drive which does not support Ultra-DMA modes. This does not appear to be necessary on all drives, the PCIIDE_AMD756_ENABLEDMA option can be used to force multiword DMA on the buggy revisions. Multiword DMA can eventually be disabled on a per-drive basis with config flags, see wd(4). The bug, if triggered, will cause a total system hang.
The timings used for the PIO and DMA modes for controllers listed above are for a PCI bus running at 30 or 33 MHz. These timings will work on a slower bus, but they may cause errors on an over-clocked bus.
|July 13, 2014||OpenBSD-current|