Odd Crash today with new build (built about an hour ago)

Matthew Dillon dillon at apollo.backplane.com
Thu Aug 26 09:29:47 PDT 2004

    If this can be repeated a trace probably isn't enough.  I'd need
    a kernel core dump.  This is a safety panic... in DragonFly the FPU
    is *always* supposed to be in a useable state when in kernel mode,
    requiring only that the saved-state vector be modified, but if it hit
    this trap then the FPU was marked as being in-use.  This points to a
    coding error somewhere but it is most likely to be related to a 
    preemptive interrupt thread switch, in which case the backtrace won't
    have all the information needed to track down the problem.

					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>

:On Wed, Aug 25, 2004 at 10:46:26PM -0400, Marina Brown wrote:
:> Here is what i could salvage from my dmesg.
:> Switching keyboard from one box to another
:> had DDB scrolling and unusable.
:> Fatal trap 22: FPU device not available while in kernel mode
:> instruction pointer     = 0x8:0xc03e26c0
:> stack pointer           = 0x10:0xda18f74c
:> frame pointer           = 0x10:0xda18f77c
:> code segment            = base 0x0, limit 0xfffff, type 0x1b
:>                        = DPL 0, pres 1, def32 1, gran 1
:> processor eflags        = interrupt enabled, resume, IOPL = 0
:> current process         = 688 (vi)
:> current thread          = pri 10
:> interrupt mask          = none
:Give us a "trace", please.
:> ...I was editing a makefile ;-)
:> Marina Brown

More information about the Submit mailing list