Re: Moving from smbfs to cifs

On Mon, Jun 9, 2008 at 1:19 PM, Peter Wemm <peter_at_wemm.org> wrote:
> On Mon, Jun 9, 2008 at 11:02 AM, Garrett Cooper <yanefbsd_at_gmail.com> wrote:
>> Hello all,
>>        I was wondering if there's been any serious thought put into
>> migrating from smbfs (unmaintained project in kernel / userland since
>> 2001) to cifs (currently supported Samba project). This is the
>> mount_smbfs user tool that's available in userland.
>>        There are some related questions about this and observations
>> that I've made:
>> Pros:
>>        1. cifs is the successor to smbfs, which is good from a
>> performance and feature enhancement end.
>>        2. It's supported, which means that any bugs in the code can
>> be filed upstream and we'll be helped. This is an important point as I
>> appear to be hard locking up my system with some kind of non-MPSAFE
>> issue at kernel level on a very fresh copy of -CURRENT.
>>
>> Cons:
>>        1. cifs is currently Linux centric (it currently uses quite a
>> few Linux calls and references the Linux module code base); that will
>> need to be fixed.
>>        2. It's GPL v2 licensed, which means that more GPL code will
>> "infect" the kernel, whereas smbfs was in a more BSD-like license
>> format.
>>
>>        So, my question would be "do the pros outweigh the cons for
>> attempting to migrate from smbfs to cifs in the kernel?"
>> Thanks,
>> -Garrett
>
> I was surprised to discover that smbfs works as well as it does.  I
> really was expecting a whole pile of panics, lockups etc, but for my
> usage level, it seems to just work.  On my smp desktop:
>
> peter_at_overcee[1:16pm]~/fbp4/hammer/sys/dev/twa-193> mount | grep smbfs | wc -l
>       5
> peter_at_overcee[1:17pm]~/fbp4/hammer/sys/dev/twa-194> uptime
>  1:17PM  up 49 days, 21:46, 6 users, load averages: 0.29, 0.26, 0.17
>
> Maybe it'll all catch fire tomorrow..

Well, I wish I had your luck Peter. Every time I try and mount my
admin share on my Vista box (for testing purposes only now), the
mount_smbfs proc locks up the kernel with a strange recursive lock
error (I think that's what causes it at least; I can't attach kgdb or
core dump the machine... yet... I'm going to go ahead and force a
panic today to see what data's hanging out in the kernel). This is all
with a kernel / userland newly synced and rebuilt as of last Thursday
with a clean /usr/obj :(...

That's part of the reason why I was asking about migrating to cifs.
Besides, we'd probably see a performance boost moving to cifs at some
degree because it's more supported by the microsoft folks now and they
were having issues with performance and SMB (at least prior to Vista
SP1).

Just a few thoughts...
-Garrett