Jorgen,

Can you provide some details about the stack trace when you hit this
failure. All of the ms_freetrees should be empty by the time you can
range_tree_destroy(). So to debug this we need to understand the calling
stack to determine why that isn't happening.

Thanks,
George

On Sat, Aug 22, 2015 at 8:13 PM, Jorgen Lundman <lund...@lundman.net> wrote:

>
> Hello list;
>
> So running the test environment, we can regularly trigger the following
> VERIFY:
>
> panic(cpu 0 caller 0xffffff7f8a065e08): "VERIFY3(" "0" " " "==" " "
> "rt->rt_space" ") " "failed (" "0" " " "==" " " "14336"
> ")\n"@range_tree.c:153
>
> I threw in some code to determine the allocator;
>
> ```
> range_tree_destroy(range_tree_t *rt)
> {
>     if (rt->rt_space) {
>         printf("ZFS: rt->rt_space is !0 : space %llu from %s:%u \n",
>                rt->rt_space, rt->rt_debug_allocator, rt->rt_debug_line);
>         range_tree_walk(rt, dumpdump, NULL);
>     }
>     //VERIFY0(rt->rt_space);
> ```
>
> and output is similar to:
>
> 22/08/2015 4:27:35.000 PM kernel[0]: ZFS: rt->rt_space is !0 : space 9728
> from metaslab.c:1975
> 22/08/2015 4:27:35.000 PM kernel[0]: zfs: 4194304 (1536)
> 22/08/2015 4:27:35.000 PM kernel[0]: zfs: 4197888 (1536)
> 22/08/2015 4:27:35.000 PM kernel[0]: zfs: 4202496 (1024)
> 22/08/2015 4:27:35.000 PM kernel[0]: zfs: 4208640 (512)
> 22/08/2015 4:27:35.000 PM kernel[0]: zfs: 4212224 (5120)
>
> In fact, all of them are from metaslab.c:1975, which corresponds to
>
>             msp->ms_freetree[t] = range_tree_create(NULL, msp,
>                 &msp->ms_lock);
>
> The tests run at the time are the "pool upgrade tests", so possibly
> related to the upgrade code.
>
>
> Has anyone come across this already, is it known anywhere? ZOL #3390 and
> #2947 at least mention it.
>
> Lund
>
> https://github.com/openzfsonosx/zfs/issues/361
>
>
>
> _______________________________________________
> developer mailing list
> developer@open-zfs.org
> http://lists.open-zfs.org/mailman/listinfo/developer
>
_______________________________________________
developer mailing list
developer@open-zfs.org
http://lists.open-zfs.org/mailman/listinfo/developer

Reply via email to