Crashdumps on ATA RAID controllers?

[Matthew Dillon]

:Hi,
:
:I've got some panics I think are reproducible, but I can't get it
:to crash dump to the ar0 device (the only disk available.)  The
:machine is in a remote location, so I can't get at the console
:(don't think I can set up a serial one, either.)  Anyone have
:a similar experience?  Do I have other options?
:
:-Paul.
 
    Lets start out with the panic message and DDB>  backtrace
    ('trace' command from the ddb> prompt).  

    There are some known issues with the twe driver, they could
    be the same with the ar driver.  Basically the twe driver is
    using malloc(... M_NOWAIT) but assuming that the result will
    always be non-NULL, causing a crash.  DragonFly will return
    NULL for malloc(... M_NOWAIT) if it cannot get hold of the kernel_map
    lock, which happens if an interrupt is interrupting mainline code
    that is holding the kernel_map lock.

					-Matt
					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>