what's going on after upgrade to svn-latest 9.

[Andriy Gapon]

on 01/10/2013 22:29 Alexey Dokuchaev said the following:
> On Fri, Jun 21, 2013 at 01:17:31PM +0200, Wojciech Puchar wrote:
>> i'm getting this on my computer.
>> disk seems OK, smart shows nothing, dd reads whole partition without
>> problem.
>>
>> no other errors from disk itself (AHCI timeout or so).
>>
>> started exactly after i rebooted with new kernel. everything
>> unchanged. userland is in sync
>>
>> thanks for help
>>
>> g_vfs_done():ada0s3d.eli[WRITE(offset=30653186048,
>> length=1048576)]error = 11
> 
> i've just tried recent 10.0-alpha4 on my laptop (normally running 8-stable)
> and start to see this shit as well.  smart does not report anything wrong
> with my disks, and 8-stable never yielded anything like these messages, so
> it looks like some kind of false alarm.  any clues?
> 

It looks like the error comes from unmapped I/O code in GEOM and is caused by
KVA shortage:
        if (vmem_alloc(transient_arena, size, M_BESTFIT | M_NOWAIT, &addr)) {
                if (transient_map_retries != 0 &&
                    retried >= transient_map_retries) {
                        g_io_deliver(bp, EDEADLK/* XXXKIB */);

You could try to experiment with kern.geom.transient_map_retries

-- 
Andriy Gapon