DragonFly kernel List (threaded) for 2007-06
DragonFly BSD
DragonFly kernel List (threaded) for 2007-06
[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]

Re: Decision time.... should NATA become the default for this release?


From: Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>
Date: Fri, 1 Jun 2007 10:40:04 -0700 (PDT)

:I gave NATA a try on my notebook PC, and now it boots into single-user mode!
:This is a progress, as it used to just get stuck somewhere in init(8) (BTW
:I tried to add debug print's in init, but somehow it didn't work), and
:all I could do then was to drop in to DDB or reboot.
:
:However, core dumping of random processes still remains, so it's still not
:usable in multi-user mode.  Sometimes, it panics on umount, after some
:read-only disk activities:
    
    Could you post your *entire* dmesg output booted with NATA?  In
    particular I want to know what controller the NATA code detected.

:  # mount -r /home
:  # find /home >/dev/null
:  # umount /home
:  panic: unmount: dangling vnode

    The same I/O issues causing the core dumps could be preventing some
    of the vnodes from being cleaned up.  Lets try to track down the
    first problem first, then see if the dangling vnode error is still
    present.

:Another weird thing with NATA is that I'm seeing odd interrupt counts
:from `vmstat -i'; with old-ATA, it looks like this:
:  # vmstat -iv
:  interrupt                   total       rate
:  irq0: clk                  402493        277
:  irq1: atkbd0                    4          0
:  irq9: acpi0                    54          0
:  irq11: fxp0                  2649          1
:  irq12: psm0                     0          0
:  irq14: ata0                  1634          1
:  irq15: ata1                     0          0
:  irq68: swi_vm                   0          0
:  irq69: swi_taskq                0          0
:  Total                      406834        280
:
:with NATA, it looks like this(in single-user mode):
:  # vmstat -iv
:  interrupt                   total       rate
:  irq0: clk                    1236        176
:  irq1: atkbd0                   24          3
:  irq9: acpi0                     1          0
:  irq11: fxp0                     0          1
:  irq12: psm0                     0          0
:  irq14: ata0                    99         14
:  irq15: ata1                     0          0
:  irq21                          98         14
:  irq68: swi_vm                   0          0
:  irq69: swi_taskq                0          0
:  Total                        1458        208
:
:irq21 has almost the same number as ata0 does and the number raises
:as the disk activity occurs.

    Insofar as I know these sorts of duplicated IRQs occur when
    interrupts are misrouted in hardware.  We need to bring in the
    FreeBSD ACPI interrupt routing code to whack that problem (or most
    of that problem, since it doesn't work perfectly even in FreeBSD).

    Could you post your *entire* dmesg output booted with NATA and
    also the entire dmesg output booted with ATA ?

:Cheers.
:
:--rwEMma7ioTxnRzrJ
:Content-Type: text/plain; charset=us-ascii
:Content-Disposition: attachment; filename="vmstat.c.diff"

    Commit that :-)

					-Matt
					Matthew Dillon 
					<dillon@backplane.com>



[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]