[PATCH v3] nfs: keep server info for remounts

[PATCH v3] nfs: keep server info for remounts

[Martin Kaiser]

With newer kernels that use fs_context for nfs mounts, remounts fail with
-EINVAL.

$ mount -t nfs -o nolock 10.0.0.1:/tmp/test /mnt/test/
$ mount -t nfs -o remount /mnt/test/
mount: mounting 10.0.0.1:/tmp/test on /mnt/test failed: Invalid argument

For remounts, the nfs server address and port are populated by
nfs_init_fs_context and later overwritten with 0x00 bytes by
nfs23_parse_monolithic. The remount then fails as the server address is
invalid.

Fix this by not overwriting nfs server info in nfs23_parse_monolithic if
we're doing a remount.

Fixes: f2aedb713c28 ("NFS: Add fs_context support.")
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
---
 v3:
 - rebased against linux-next from 12th April 2024

 v2:
 - rebased against linux-next from 26th February 2024

Dear all,
I'm resending this patch again. The problem that I'm trying to fix is still
present in linux-next. Thanks in advance for any reviews and comments.

I guess that we're taking this path for remounts

do_remount
    fs_context_for_reconfigure
        alloc_fs_context
            init_fs_context == nfs_init_fs_context
               fc->root is set for remounts
               ctx->nfs_server is populated
    parse_monolithic_mount_data
        nfs_fs_context_parse_monolithic
            nfs23_parse_monolithic
               ctx->nfs_server is overwritten with data from mount request

An alternative to checking for !is_remount_fc(fc) would be to check
if (ctx->nfs_server.addrlen == 0)

fs/nfs/fs_context.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/fs/nfs/fs_context.c b/fs/nfs/fs_context.c
index d0a0956f8a13..cac1157be2c2 100644
--- a/fs/nfs/fs_context.c
+++ b/fs/nfs/fs_context.c
@@ -1112,9 +1112,12 @@ static int nfs23_parse_monolithic(struct fs_context *fc,
 		ctx->acdirmax	= data->acdirmax;
 		ctx->need_mount	= false;
 
-		memcpy(sap, &data->addr, sizeof(data->addr));
-		ctx->nfs_server.addrlen = sizeof(data->addr);
-		ctx->nfs_server.port = ntohs(data->addr.sin_port);
+		if (!is_remount_fc(fc)) {
+			memcpy(sap, &data->addr, sizeof(data->addr));
+			ctx->nfs_server.addrlen = sizeof(data->addr);
+			ctx->nfs_server.port = ntohs(data->addr.sin_port);
+		}
+
 		if (sap->ss_family != AF_INET ||
 		    !nfs_verify_server_address(sap))
 			goto out_no_address;
-- 
2.39.2

Re: [PATCH v3] nfs: keep server info for remounts

[Jeff Layton]

On Sun, 2024-04-14 at 19:01 +0200, Martin Kaiser wrote:
> With newer kernels that use fs_context for nfs mounts, remounts fail with
> -EINVAL.
> 
> $ mount -t nfs -o nolock 10.0.0.1:/tmp/test /mnt/test/
> $ mount -t nfs -o remount /mnt/test/
> mount: mounting 10.0.0.1:/tmp/test on /mnt/test failed: Invalid argument
> 
> For remounts, the nfs server address and port are populated by
> nfs_init_fs_context and later overwritten with 0x00 bytes by
> nfs23_parse_monolithic. The remount then fails as the server address is
> invalid.
> 
> Fix this by not overwriting nfs server info in nfs23_parse_monolithic if
> we're doing a remount.
> 
> Fixes: f2aedb713c28 ("NFS: Add fs_context support.")
> Signed-off-by: Martin Kaiser <martin@kaiser.cx>
> ---
>  v3:
>  - rebased against linux-next from 12th April 2024
> 
>  v2:
>  - rebased against linux-next from 26th February 2024
> 
> Dear all,
> I'm resending this patch again. The problem that I'm trying to fix is still
> present in linux-next. Thanks in advance for any reviews and comments.
> 
> I guess that we're taking this path for remounts
> 
> do_remount
>     fs_context_for_reconfigure
>         alloc_fs_context
>             init_fs_context == nfs_init_fs_context
>                fc->root is set for remounts
>                ctx->nfs_server is populated
>     parse_monolithic_mount_data
>         nfs_fs_context_parse_monolithic
>             nfs23_parse_monolithic
>                ctx->nfs_server is overwritten with data from mount request
> 
> An alternative to checking for !is_remount_fc(fc) would be to check
> if (ctx->nfs_server.addrlen == 0)
> 
> fs/nfs/fs_context.c | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/nfs/fs_context.c b/fs/nfs/fs_context.c
> index d0a0956f8a13..cac1157be2c2 100644
> --- a/fs/nfs/fs_context.c
> +++ b/fs/nfs/fs_context.c
> @@ -1112,9 +1112,12 @@ static int nfs23_parse_monolithic(struct fs_context *fc,
>  		ctx->acdirmax	= data->acdirmax;
>  		ctx->need_mount	= false;
>  
> -		memcpy(sap, &data->addr, sizeof(data->addr));
> -		ctx->nfs_server.addrlen = sizeof(data->addr);
> -		ctx->nfs_server.port = ntohs(data->addr.sin_port);
> +		if (!is_remount_fc(fc)) {
> +			memcpy(sap, &data->addr, sizeof(data->addr));
> +			ctx->nfs_server.addrlen = sizeof(data->addr);
> +			ctx->nfs_server.port = ntohs(data->addr.sin_port);
> +		}
> +
>  		if (sap->ss_family != AF_INET ||
>  		    !nfs_verify_server_address(sap))
>  			goto out_no_address;

Doesn't nfs4_parse_monolithic need the same fix? 
-- 
Jeff Layton <jlayton@kernel.org>

Re: [PATCH v3] nfs: keep server info for remounts

[Martin Kaiser]

Thus wrote Jeff Layton (jlayton@kernel.org):

> On Sun, 2024-04-14 at 19:01 +0200, Martin Kaiser wrote:
> > With newer kernels that use fs_context for nfs mounts, remounts fail with
> > -EINVAL.

> > $ mount -t nfs -o nolock 10.0.0.1:/tmp/test /mnt/test/
> > $ mount -t nfs -o remount /mnt/test/
> > mount: mounting 10.0.0.1:/tmp/test on /mnt/test failed: Invalid argument

> > For remounts, the nfs server address and port are populated by
> > nfs_init_fs_context and later overwritten with 0x00 bytes by
> > nfs23_parse_monolithic. The remount then fails as the server address is
> > invalid.

> > Fix this by not overwriting nfs server info in nfs23_parse_monolithic if
> > we're doing a remount.

> > Fixes: f2aedb713c28 ("NFS: Add fs_context support.")
> > Signed-off-by: Martin Kaiser <martin@kaiser.cx>
> > ---
> >  v3:
> >  - rebased against linux-next from 12th April 2024

> >  v2:
> >  - rebased against linux-next from 26th February 2024

> > Dear all,
> > I'm resending this patch again. The problem that I'm trying to fix is still
> > present in linux-next. Thanks in advance for any reviews and comments.

> > I guess that we're taking this path for remounts

> > do_remount
> >     fs_context_for_reconfigure
> >         alloc_fs_context
> >             init_fs_context == nfs_init_fs_context
> >                fc->root is set for remounts
> >                ctx->nfs_server is populated
> >     parse_monolithic_mount_data
> >         nfs_fs_context_parse_monolithic
> >             nfs23_parse_monolithic
> >                ctx->nfs_server is overwritten with data from mount request

> > An alternative to checking for !is_remount_fc(fc) would be to check
> > if (ctx->nfs_server.addrlen == 0)

> > fs/nfs/fs_context.c | 9 ++++++---
> >  1 file changed, 6 insertions(+), 3 deletions(-)

> > diff --git a/fs/nfs/fs_context.c b/fs/nfs/fs_context.c
> > index d0a0956f8a13..cac1157be2c2 100644
> > --- a/fs/nfs/fs_context.c
> > +++ b/fs/nfs/fs_context.c
> > @@ -1112,9 +1112,12 @@ static int nfs23_parse_monolithic(struct fs_context *fc,
> >  		ctx->acdirmax	= data->acdirmax;
> >  		ctx->need_mount	= false;

> > -		memcpy(sap, &data->addr, sizeof(data->addr));
> > -		ctx->nfs_server.addrlen = sizeof(data->addr);
> > -		ctx->nfs_server.port = ntohs(data->addr.sin_port);
> > +		if (!is_remount_fc(fc)) {
> > +			memcpy(sap, &data->addr, sizeof(data->addr));
> > +			ctx->nfs_server.addrlen = sizeof(data->addr);
> > +			ctx->nfs_server.port = ntohs(data->addr.sin_port);
> > +		}
> > +
> >  		if (sap->ss_family != AF_INET ||
> >  		    !nfs_verify_server_address(sap))
> >  			goto out_no_address;

> Doesn't nfs4_parse_monolithic need the same fix? 

Sorry for the delayed response. It took me a moment to set up a test with nfs4
(busybox mount has no nfs4 support).

The nfs4 remounts do not fail for me. The mount syscall goes into
nfs4_parse_monolithic and 

   if (data->version != 1)                                                                                             
      return generic_parse_monolithic(fc, data);                                                                       

branches off into generic_parse_monolithic before the server address is
overwritten (this is what breaks nfs23).

Best regards,

   Martin