Then every device will need do scan_static_bus.
Also the debug device need to find his parent....
YH
-----Original Message----- From: Li-Ta Lo [mailto:ollie@lanl.gov] Sent: Thursday, November 04, 2004 3:26 PM To: YhLu Cc: LinuxBIOS Subject: RE: Removing device driver code in mainboard.c for Tyan boards?
On Thu, 2004-11-04 at 15:44, YhLu wrote:
debug device added
chip drivers/generic/debug device pnp 0.0 on end end
debug device will be sibling to PCI_DOMAIN and APIC_CLUSTER...
Cool, that is exactly what I want. Actually by defining different debug devices, we can construct different 'probes' into the device space. We can either probe software data structures or the real hardware and we can do it at different phases of the device enumeration as we want. For example, the current debug device (pnp 0.0) dumps the PCI CS registers at the end of device enumeration. We can define another debug device to show the topology of the device tree or just print the famous .name field ;-).
The best thing about the debug device is it totally extensible and plug/unplugable. We don't have to add any code to the existing device driver nor the hardwaremain(). It is decoupled from the rest of LinuxBIOS. The operations we need to act on the (normal) devices are encapsulated in the debug device.
Ollie
On Thu, 2004-11-04 at 17:13, YhLu wrote:
Then every device will need do scan_static_bus.
Also the debug device need to find his parent....
Why ??
Ollie
Li-Ta Lo ollie@lanl.gov writes:
On Thu, 2004-11-04 at 17:13, YhLu wrote:
Then every device will need do scan_static_bus.
Also the debug device need to find his parent....
Why ??
Ollie
YhLu I think Ollie was just thinking about different kinds of debug devices for different kinds of debugging information.
One to print the out pci space. Another one dump the cpu msrs.
Etc.
Basically just configurable hooks that we can collect and everyone can share.
It should not be necessary to place the debug devices in different places in the tree.
Eric