I am reproducing a bit of Kevin Sullivan's bug report below:

>Description:
>APM used to work fine on my laptop (an IBM Thinkpad 701cs).  Lately
>though it has been having problems.  It goes to sleep okay, but when I
>...
>errors about filesystem corruption or I cannot run whichever program
>first accessed the disk.  When I reboot the machine the errors generally
>go away (though once it trashed about 10% of my filesystem).


IBM has been stupid for a long time.  On the 701c and 755 they steal
the last physical page before the gap (0xbffff) and use of as part of
their suspend/resume processing; they will store into the page!!!!
You must not use this physical page for netbsd.  In my case, this page
often turned out to be in the buffer pool and I could clobber pages or
inodes whenever I did a suspend/resume.  I'd recommend a patchable
flag or option that allowes this page to be ignored.