smb3 mount fix for Macs

smb3 mount fix for Macs

[Steve French]

pushed patch to workaround smb3 mount failure to Mac

(we had to relax our length checks on the response to smb2 ioctl for
validate negotiate since they pad in order to return the maximum
buffer allowed - 16K for what is normally a rather small ioctl
response).

smb3 mounts now work to Macs again (smb2.1 mounts to Mac had always
been find - it only failed to smb3 due to the validate negotiate
ioctl)

https://git.samba.org/?p=sfrench/cifs-2.6.git;a=commit;h=be60379eac1c023f9b77565060414e5b122de766



-- 
Thanks,

Steve

Re: smb3 mount fix for Macs

[Steve French]

Minor update and added cc:stable

https://git.samba.org/?p=sfrench/cifs-2.6.git;a=commit;h=7db0a6efdc3e990cdfd4b24820d010e9eb7890ad

On Wed, May 3, 2017 at 9:23 PM, Steve French <smfrench-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> pushed patch to workaround smb3 mount failure to Mac
>
> (we had to relax our length checks on the response to smb2 ioctl for
> validate negotiate since they pad in order to return the maximum
> buffer allowed - 16K for what is normally a rather small ioctl
> response).
>
> smb3 mounts now work to Macs again (smb2.1 mounts to Mac had always
> been find - it only failed to smb3 due to the validate negotiate
> ioctl)
>
> https://git.samba.org/?p=sfrench/cifs-2.6.git;a=commit;h=be60379eac1c023f9b77565060414e5b122de766
>
>
>
> --
> Thanks,
>
> Steve



-- 
Thanks,

Steve

Re: smb3 mount fix for Macs

[David Disseldorp]

On Wed, 3 May 2017 21:25:50 -0500, Steve French wrote:

> Minor update and added cc:stable
> 
> https://git.samba.org/?p=sfrench/cifs-2.6.git;a=commit;h=7db0a6efdc3e990cdfd4b24820d010e9eb7890ad

The relaxed validate negotiate rsplen check looks okay, but I think the
reduced MaxOutputResponse change should probably be split out into a
separate commit, as it affects all other ioctls.

Given that the response size is known in advance, couldn't we have the
validate_negotiate caller pass in the desired MaxOutputResponse value?

Cheers, David

Re: smb3 mount fix for Macs

[Steve French]

On Thu, May 4, 2017 at 5:24 AM, David Disseldorp <ddiss-l3A5Bk7waGM@public.gmane.org> wrote:
> On Wed, 3 May 2017 21:25:50 -0500, Steve French wrote:
>
>> Minor update and added cc:stable
>>
>> https://git.samba.org/?p=sfrench/cifs-2.6.git;a=commit;h=7db0a6efdc3e990cdfd4b24820d010e9eb7890ad
>
> The relaxed validate negotiate rsplen check looks okay, but I think the
> reduced MaxOutputResponse change should probably be split out into a
> separate commit, as it affects all other ioctls.

That is a good thing in a way - because we don't want any ioctls to
send more than maxbufsize until we can handle this.

> Given that the response size is known in advance, couldn't we have the
> validate_negotiate caller pass in the desired MaxOutputResponse value?

Perhaps - but I don't know if that makes sense until we have way of
handling more than max buf size (defaults to 16K payload) in smb2
ioctl/fsctl response.

My gut reaction is to change the code to allow passing in maxoutput
for response on ioctl when we have a way of handling 'very large'
ioctl responses in demultiplexing in smb2 transport layer.


-- 
Thanks,

Steve

Re: smb3 mount fix for Macs

[Pavel Shilovsky]

2017-05-04 3:30 GMT-07:00 Steve French <smfrench-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>:
> On Thu, May 4, 2017 at 5:24 AM, David Disseldorp <ddiss-l3A5Bk7waGM@public.gmane.org> wrote:
>> On Wed, 3 May 2017 21:25:50 -0500, Steve French wrote:
>>
>>> Minor update and added cc:stable
>>>
>>> https://git.samba.org/?p=sfrench/cifs-2.6.git;a=commit;h=7db0a6efdc3e990cdfd4b24820d010e9eb7890ad
>>
>> The relaxed validate negotiate rsplen check looks okay, but I think the
>> reduced MaxOutputResponse change should probably be split out into a
>> separate commit, as it affects all other ioctls.
>
> That is a good thing in a way - because we don't want any ioctls to
> send more than maxbufsize until we can handle this.
>

I agree that we need to split the patch into two because it is
targeted to stable. SMB2_ioctl() and smb3_validate_negotiate() were
introduced in different times, so the single patch won't be applied
properly on some stable kernel versions.

--
Best regards,
Pavel Shilovsky