[haiku-development] Re: Virtual 8086 mode support
- From: Jan Klötzke <jan.kloetzke@xxxxxxxxxx>
- To: haiku-development@xxxxxxxxxxxxx
- Date: Sat, 26 Apr 2008 15:46:34 +0200
Hi Kernel Hackers,
Good news is that I have the in-kernel vm86 mode call working. :-) But at the
moment I'm a bit stuck on the error handling part. Let me first explain the
current solution:
There are now two functions:
status_t x86_vm86_enter(struct vm86_iframe *frame);
void x86_vm86_return(struct vm86_iframe *frame, status_t retval);
To enter vm86 mode the caller has to prepare an appropriate iframe and then
call x86_vm86_enter(). Exceptions handlers (such as for #GP) can call
x86_vm86_return() which will leave vm86 mode and cause the original
x86_vm86_enter() call to return. Nothing fancy so far.
So when the BIOS is called the invocation of vm86 mode should not be
interrupted, expect for fatal exceptions. The only problem here is the page
fault exception. If the page fault could not be handled then vm86 mode must
be left, otherwise not. Unfortunately the arch code has no clue what happened
because vm_page_fault() always returns B_HANDLED_INTERRUPT. I thought
checking for a pending SIGSEGV in the arch page fault exception handler could
also be an options but apparently the SIGSEGV signal is dispatched to the
main thread of the team and not to the thread who caused it. Therefore I have
two questions:
* May it be an option to change the return value of vm_page_fault() to know in
the arch code if the page fault could be handled or not?
* Why are all signals sent to the team (main thread) and not to the causing
thread?
/Jan
Other related posts:
