Subject: Re: Getting "TLB IPI rendezvous failed..."
To: Frank van der Linden <>
From: Stephan Uphoff <>
List: tech-kern
Date: 01/24/2005 11:56:23
On Sat, 2005-01-22 at 13:22, Frank van der Linden wrote:
> On Sat, Jan 22, 2005 at 09:21:59AM -0500, Stephan Uphoff wrote:
> > Could you try the attached patch?
> > Please make sure that all your com devices show up.
> Hmm.. I see your line of thinking here: you think that all the com
> devices, establishing interrupts, will overflow into the IDT entry
> for IPIs?

Yes - overflows into the same priority class as the IDT entry for the
IPI. With late interrupt acknowledge to the lapic this could block the

> That shouldn't happen, since there are only four ioapic pins involved
> here, and each of them will take one IDT entry, so they should end
> up as 0xd0-0xd3.

Unless something else is wrong (or COMPAT_SVR4 block 0xd2) I agree.

> The patch is good though, idt_vec_alloc should not go higher than
> entry 0xdf. 


> Zeroing out idt_allocmap isn't needed, since it's declared
> statically.

Mhhh OK - but who is zeroing BSS?