[PATCH] NFS: Fix kbuild errors due to missing symbols

[PATCH] NFS: Fix kbuild errors due to missing symbols

[Anna Schumaker]

My recent patch 644aa303: "NFS: Enabling v4.2 should not recompile nfsd
and lockd" caused build errors when CONFIG_NFS_USE_LEGACY_DNS=y and when
CONFIG_NFS_V4=n.  This patch fixes those errors.

Signed-off-by: Anna Schumaker <bjschuma@netapp.com>
---
 fs/nfs/nfs4_fs.h | 19 +++++++++----------
 1 file changed, 9 insertions(+), 10 deletions(-)

diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index e916e86..dc1c915 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -9,21 +9,20 @@
 #ifndef __LINUX_FS_NFS_NFS4_FS_H
 #define __LINUX_FS_NFS_NFS4_FS_H
 
-#if IS_ENABLED(CONFIG_NFS_V4)
-
-#define NFS4_MAX_LOOP_ON_RECOVER (10)
-
-#include <linux/seqlock.h>
-
 #if defined(CONFIG_NFS_V4_2)
 #define NFS4_MAX_MINOR_VERSION 2
-#else
-#if defined(CONFIG_NFS_V4_1)
+#elif defined(CONFIG_NFS_V4_1)
 #define NFS4_MAX_MINOR_VERSION 1
 #else
 #define NFS4_MAX_MINOR_VERSION 0
-#endif /* CONFIG_NFS_V4_1 */
-#endif /* CONFIG_NFS_V4_2 */
+#endif
+
+#if IS_ENABLED(CONFIG_NFS_V4)
+
+#define NFS4_MAX_LOOP_ON_RECOVER (10)
+
+#include <linux/seqlock.h>
+#include <linux/nfs_fs.h>
 
 struct idmap;
 
-- 
1.8.4.2

Re: [PATCH] NFS: Fix kbuild errors due to missing symbols

[Trond Myklebust]

On Fri, 2013-11-15 at 17:30 -0500, Anna Schumaker wrote:
> My recent patch 644aa303: "NFS: Enabling v4.2 should not recompile nfsd
> and lockd" caused build errors when CONFIG_NFS_USE_LEGACY_DNS=y and when
> CONFIG_NFS_V4=n.  This patch fixes those errors.
> 

Wait... 

        config NFS_USE_LEGACY_DNS
                bool "Use the legacy NFS DNS resolver"
                depends on NFS_V4
                help
                  The kernel now provides a method for translating a host name into an
                  IP address.  Select Y here if you would rather use your own DNS
                  resolver script.
        
                  If unsure, say N
        
        config NFS_USE_KERNEL_DNS
                bool
                depends on NFS_V4 && !NFS_USE_LEGACY_DNS
                select DNS_RESOLVER
                default y

How can that result in NFS_V4=n && NFS_USE_LEGACY_DNS=y?

Re: [PATCH] NFS: Fix kbuild errors due to missing symbols

[Trond Myklebust]

On Nov 15, 2013, at 18:16, Trond Myklebust <trondmy@gmail.com> wrote:

> On Fri, 2013-11-15 at 17:30 -0500, Anna Schumaker wrote:
>> My recent patch 644aa303: "NFS: Enabling v4.2 should not recompile nfsd
>> and lockd" caused build errors when CONFIG_NFS_USE_LEGACY_DNS=y and when
>> CONFIG_NFS_V4=n.  This patch fixes those errors.
>> 
> 
> Wait... 
> 
>        config NFS_USE_LEGACY_DNS
>                bool "Use the legacy NFS DNS resolver"
>                depends on NFS_V4
>                help
>                  The kernel now provides a method for translating a host name into an
>                  IP address.  Select Y here if you would rather use your own DNS
>                  resolver script.
> 
>                  If unsure, say N
> 
>        config NFS_USE_KERNEL_DNS
>                bool
>                depends on NFS_V4 && !NFS_USE_LEGACY_DNS
>                select DNS_RESOLVER
>                default y
> 
> How can that result in NFS_V4=n && NFS_USE_LEGACY_DNS=y?

OK. The problem is resolved using an include of nfs_fs.h in the right spot. Will fix…

Re: [PATCH] NFS: Fix kbuild errors due to missing symbols

[Anna Schumaker]

On 11/15/2013 06:16 PM, Trond Myklebust wrote:
> On Fri, 2013-11-15 at 17:30 -0500, Anna Schumaker wrote:
>> My recent patch 644aa303: "NFS: Enabling v4.2 should not recompile nfsd
>> and lockd" caused build errors when CONFIG_NFS_USE_LEGACY_DNS=y and when
>> CONFIG_NFS_V4=n.  This patch fixes those errors.
>>
> 
> Wait... 
> 
>         config NFS_USE_LEGACY_DNS
>                 bool "Use the legacy NFS DNS resolver"
>                 depends on NFS_V4
>                 help
>                   The kernel now provides a method for translating a host name into an
>                   IP address.  Select Y here if you would rather use your own DNS
>                   resolver script.
>         
>                   If unsure, say N
>         
>         config NFS_USE_KERNEL_DNS
>                 bool
>                 depends on NFS_V4 && !NFS_USE_LEGACY_DNS
>                 select DNS_RESOLVER
>                 default y
> 
> How can that result in NFS_V4=n && NFS_USE_LEGACY_DNS=y?
> 

It's a poorly phrased commit message, actually.  I meant that just setting CONFIG_NFS_USE_LEGACY_DNS=y causes build problems, and then setting CONFIG_NFS_V4=n causes a different set of errors.

Re: [PATCH] NFS: Fix kbuild errors due to missing symbols

[Myklebust, Trond]

On Nov 18, 2013, at 12:50, Anna Schumaker <bjschuma@netapp.com> wrote:
> 
> It's a poorly phrased commit message, actually.  I meant that just setting CONFIG_NFS_USE_LEGACY_DNS=y causes build problems, and then setting CONFIG_NFS_V4=n causes a different set of errors.

Ah. OK, now I see why you need it. The code in fs/nfs/super.c relies on it when !defined CONFIG_NFS_V4.

You also need to move the CONFIG_NFS_V4_SECURITY_LABEL section outside the CONFIG_NFS_V4, since fs/nfs/dir.c and others need the dummy nfs4_label_alloc and nfs4_label_free inline functions (see Stephen Rothwell
’s email).


--
Trond Myklebust
Linux NFS client maintainer

NetApp
Trond.Myklebust@netapp.com
www.netapp.com

Re: [PATCH] NFS: Fix kbuild errors due to missing symbols

[Anna Schumaker]

On 11/18/2013 01:37 PM, Myklebust, Trond wrote:
> 
> On Nov 18, 2013, at 12:50, Anna Schumaker <bjschuma@netapp.com> wrote:
>>
>> It's a poorly phrased commit message, actually.  I meant that just setting CONFIG_NFS_USE_LEGACY_DNS=y causes build problems, and then setting CONFIG_NFS_V4=n causes a different set of errors.
> 
> Ah. OK, now I see why you need it. The code in fs/nfs/super.c relies on it when !defined CONFIG_NFS_V4.
> 
> You also need to move the CONFIG_NFS_V4_SECURITY_LABEL section outside the CONFIG_NFS_V4, since fs/nfs/dir.c and others need the dummy nfs4_label_alloc and nfs4_label_free inline functions (see Stephen Rothwell
> ’s email).

Right.  I have a patch that moves them to internal.h instead to get around the whole-file #ifdef that's in nfs4_fs.h.  Would you prefer me to submit that independently or as a v2 to this patch?

As a bonus, I've used Jenkins to compile as many combinations as possible without an error (I probably should have done that the first time, though...)

Anna

> 
> 
> --
> Trond Myklebust
> Linux NFS client maintainer
> 
> NetApp
> Trond.Myklebust@netapp.com
> www.netapp.com
> 

Re: [PATCH] NFS: Fix kbuild errors due to missing symbols

[Myklebust, Trond]

On Nov 18, 2013, at 13:41, Anna Schumaker <bjschuma@netapp.com> wrote:

> On 11/18/2013 01:37 PM, Myklebust, Trond wrote:
>> 
>> On Nov 18, 2013, at 12:50, Anna Schumaker <bjschuma@netapp.com> wrote:
>>> 
>>> It's a poorly phrased commit message, actually.  I meant that just setting CONFIG_NFS_USE_LEGACY_DNS=y causes build problems, and then setting CONFIG_NFS_V4=n causes a different set of errors.
>> 
>> Ah. OK, now I see why you need it. The code in fs/nfs/super.c relies on it when !defined CONFIG_NFS_V4.
>> 
>> You also need to move the CONFIG_NFS_V4_SECURITY_LABEL section outside the CONFIG_NFS_V4, since fs/nfs/dir.c and others need the dummy nfs4_label_alloc and nfs4_label_free inline functions (see Stephen Rothwell
>> ’s email).
> 
> Right.  I have a patch that moves them to internal.h instead to get around the whole-file #ifdef that's in nfs4_fs.h.  Would you prefer me to submit that independently or as a v2 to this patch?

Please do it as a v2.

> As a bonus, I've used Jenkins to compile as many combinations as possible without an error (I probably should have done that the first time, though...)

--
Trond Myklebust
Linux NFS client maintainer

NetApp
Trond.Myklebust@netapp.com
www.netapp.com