On Jun 10, 2011 13:51, Dawe wrote:
> Hi,
>
> I can reliably trigger an uvm_fault on two current amd64 systems with the
> following steps:
>
> - start an xterm
> - maximize it
> - start ghci
> - type ":m + <TAB>"
> - answer the question with "y"
> -> uvm_fault (no keyboard response)
>
> uvm_fault(0xfffffe803cb76820, 0x0, 0, 4) -> e
> kernel: page fault trap, code=0
> Stopped at 0: uvm_fault(0xfffffe803cb76820, 0x0, 0, 1) -> e
> kernel: page fault trap, code=0
> Stopped at db_read_bytes+0x20: movzbl 0(%rdi,%rcx,1),%eax
> ddb{1}>
>
> (handtyped)
> You can trigger the fault remotely over ssh, too.
> But not as reliable as with the above steps.
>
> I'm getting the fault in cwm and fvwm but never on the local console outside
> of
> X.
> Using rxvt as a terminal doesn't trigger it so far.
> Keeping the xterm in its original size also doesn't trigger the bug, resizing
> it
> only slightly is sometimes enough for the fault to happen.
> I can reproduce it with the vesa driver, too.
> Setting kern.bufcachepercent=5 has no effect.
> If someone has further ideas to analyze this, I'd be happy to help.
> Thanks.
>
Sorry, I should have mentioned that ghci is part of the ghc port.
It's the interactive haskell environment of ghc.
