NFS: regression in stable kernel 4.9.78 from 4.9.75

NFS: regression in stable kernel 4.9.78 from 4.9.75

[Wolfgang Walter]

Hello,

after upgrading our nfs-server from  4.9.75 to 4.9.78 group permissions stop 
working (for clients). If you need group permissions to access a file or 
directory, sometimes access is granted, but rather often denied. Often access 
to the same object is denied within seconds after access was granted in an 
earlier access. user permissions work fine.

Downgrading to 4.9.75 fixes the issue.

We use kerberos.

Regards,
-- 
Wolfgang Walter
Studentenwerk M�nchen
Anstalt des �ffentlichen Rechts

Re: NFS: regression in stable kernel 4.9.78 from 4.9.75

[Greg KH]

On Mon, Jan 29, 2018 at 01:22:49PM +0100, Wolfgang Walter wrote:
> Hello,
> 
> after upgrading our nfs-server from  4.9.75 to 4.9.78 group permissions stop 
> working (for clients). If you need group permissions to access a file or 
> directory, sometimes access is granted, but rather often denied. Often access 
> to the same object is denied within seconds after access was granted in an 
> earlier access. user permissions work fine.
> 
> Downgrading to 4.9.75 fixes the issue.

Any chance you can use 'git bisect' to find the offending commit?

And also, does 4.14 work properly for you?  4.15?

thanks,

greg k-h

Re: NFS: regression in stable kernel 4.9.78 from 4.9.75

[Wolfgang Walter]

Hello!

Am Montag, 29. Januar 2018, 13:22:49 schrieb Wolfgang Walter:
> Hello,
> 
> after upgrading our nfs-server from  4.9.75 to 4.9.78 group permissions stop
> working (for clients). If you need group permissions to access a file or
> directory, sometimes access is granted, but rather often denied. Often
> access to the same object is denied within seconds after access was granted
> in an earlier access. user permissions work fine.
> 
> Downgrading to 4.9.75 fixes the issue.
> 
> We use kerberos.
> 
> Regards,

This seems to be fixed in 4.15 with commit 
1995266727fa8143897e89b55f5d3c79aa828420:


commit 1995266727fa8143897e89b55f5d3c79aa828420
Author: Ben Hutchings <ben.hutchings@codethink.co.uk>
Date:   Mon Jan 22 20:11:06 2018 +0000

    nfsd: auth: Fix gid sorting when rootsquash enabled
    
    Commit bdcf0a423ea1 ("kernel: make groups_sort calling a responsibility
    group_info allocators") appears to break nfsd rootsquash in a pretty
    major way.
    
    It adds a call to groups_sort() inside the loop that copies/squashes
    gids, which means the valid gids are sorted along with the following
    garbage.  The net result is that the highest numbered valid gids are
    replaced with any lower-valued garbage gids, possibly including 0.
    
    We should sort only once, after filling in all the gids.
    
    Fixes: bdcf0a423ea1 ("kernel: make groups_sort calling a responsibility 
...")
    Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
    Acked-by: J. Bruce Fields <bfields@redhat.com>
    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>


So this should be applied to stables 4.4, 4.9 and 4.14 (and others where 
bdcf0a423ea1 has been backported to).

Regards,
-- 
Wolfgang Walter
Studentenwerk München
Anstalt des öffentlichen Rechts

Re: NFS: regression in stable kernel 4.9.78 from 4.9.75

[Wolfgang Walter]

Hello!

Am Montag, 29. Januar 2018, 13:22:49 schrieb Wolfgang Walter:
> Hello,
> 
> after upgrading our nfs-server from  4.9.75 to 4.9.78 group permissions stop
> working (for clients). If you need group permissions to access a file or
> directory, sometimes access is granted, but rather often denied. Often
> access to the same object is denied within seconds after access was granted
> in an earlier access. user permissions work fine.
> 
> Downgrading to 4.9.75 fixes the issue.
> 
> We use kerberos.
> 
> Regards,

This seems to be fixed in 4.15 with commit 
1995266727fa8143897e89b55f5d3c79aa828420:


commit 1995266727fa8143897e89b55f5d3c79aa828420
Author: Ben Hutchings <ben.hutchings@codethink.co.uk>
Date:   Mon Jan 22 20:11:06 2018 +0000

    nfsd: auth: Fix gid sorting when rootsquash enabled
    
    Commit bdcf0a423ea1 ("kernel: make groups_sort calling a responsibility
    group_info allocators") appears to break nfsd rootsquash in a pretty
    major way.
    
    It adds a call to groups_sort() inside the loop that copies/squashes
    gids, which means the valid gids are sorted along with the following
    garbage.  The net result is that the highest numbered valid gids are
    replaced with any lower-valued garbage gids, possibly including 0.
    
    We should sort only once, after filling in all the gids.
    
    Fixes: bdcf0a423ea1 ("kernel: make groups_sort calling a responsibility 
...")
    Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
    Acked-by: J. Bruce Fields <bfields@redhat.com>
    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>


So this should be applied to stables 4.4, 4.9 and 4.14 (and others where 
bdcf0a423ea1 has been backported to).

Regards,
-- 
Wolfgang Walter
Studentenwerk M�nchen
Anstalt des �ffentlichen Rechts

Re: NFS: regression in stable kernel 4.9.78 from 4.9.75

[Greg KH]

On Mon, Jan 29, 2018 at 02:42:49PM +0100, Wolfgang Walter wrote:
> Hello!
> 
> Am Montag, 29. Januar 2018, 13:22:49 schrieb Wolfgang Walter:
> > Hello,
> > 
> > after upgrading our nfs-server from  4.9.75 to 4.9.78 group permissions stop
> > working (for clients). If you need group permissions to access a file or
> > directory, sometimes access is granted, but rather often denied. Often
> > access to the same object is denied within seconds after access was granted
> > in an earlier access. user permissions work fine.
> > 
> > Downgrading to 4.9.75 fixes the issue.
> > 
> > We use kerberos.
> > 
> > Regards,
> 
> This seems to be fixed in 4.15 with commit 
> 1995266727fa8143897e89b55f5d3c79aa828420:
> 
> 
> commit 1995266727fa8143897e89b55f5d3c79aa828420
> Author: Ben Hutchings <ben.hutchings@codethink.co.uk>
> Date:   Mon Jan 22 20:11:06 2018 +0000
> 
>     nfsd: auth: Fix gid sorting when rootsquash enabled
>     
>     Commit bdcf0a423ea1 ("kernel: make groups_sort calling a responsibility
>     group_info allocators") appears to break nfsd rootsquash in a pretty
>     major way.
>     
>     It adds a call to groups_sort() inside the loop that copies/squashes
>     gids, which means the valid gids are sorted along with the following
>     garbage.  The net result is that the highest numbered valid gids are
>     replaced with any lower-valued garbage gids, possibly including 0.
>     
>     We should sort only once, after filling in all the gids.
>     
>     Fixes: bdcf0a423ea1 ("kernel: make groups_sort calling a responsibility 
> ...")
>     Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
>     Acked-by: J. Bruce Fields <bfields@redhat.com>
>     Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
> 
> 
> So this should be applied to stables 4.4, 4.9 and 4.14 (and others where 
> bdcf0a423ea1 has been backported to).

Ah, good catch, I missed that this had been merged already.

Now queued up to all relevant stable trees, thanks so much for the
report.

greg k-h