It seems to create more issues than it fixes. Try to defend against
lost/late GPEs by using a callout to kick the state machine regulary.
Add some notes about the purpose and working of the driver.
online. It runs into issues in the pmap code and will handled
differently. This allows sysctl -w machdep.sleep_state=3 to at least
recover into a working system again.
is active. Reinitialize the watchdog timer on resume. If the
device is not powered, neither tickle the watchdog nor set the
watchdog's mode, but return EBUSY, instead.
man8/man8.i386/dosboot.8, and man8/man8.i386/pxeboot.8.
* In all:
- First few lines of each man page should be .Dd, .Os, .Dt, .Nm;
- Use Nx; new sentence, new line; serial comma;
- Update comments about which files to keep in sync.
* In MI boot(8):
- add reference to architecture-specific boot(8);
- add commented-out reference to boot.cfg(5);
- reinstate BUGS heading that was lost when the MI boot.8 was
created from i386 boot.8.
* In i386 boot.8:
- mention -x and -z flags;
- we can boot from media other than floppy, or from the network;
- remove first paragraph in BUGS section now that an MI boot(8)
man page exists.
* In i386 dosboot.8 and pxeboot.8:
- mention -c, -x, and -z flags.
cats was never able to dump a kernel core dump because reading from
VGA addresses (0xb8000) was causing the system to hang.
To workaround this reprogram the footbridge to map the memory to appear on
the PCI bus at 0x20000000, rather than at 0x0. Also configure the pci bus
to have a DMA range so that data is mapped correctly.
Note that -current kernels seem to hang when unmounting the fs. This
is a seperate issue, and appears to be because interrupts need to be
enabled to unmount filesystems.
So using reboot 0x104 does work, as it does a sync without unmounting the
filesystems.
Also arm savecore doesn't do anything with the memroy dump, as on arm we
currently just dump the raw memory, there's no header block to indicate
memory sizes or other useful information.
mjf