[PATCH] nfs client: handle long symlinks properly

[PATCH] nfs client: handle long symlinks properly

[Marcelo Tosatti]

commit 87e03738fc15dc3ea4acde3a5dcb5f84b6b6152b
tree be323c0a65d7e380ad04cad1c3a80015a82056dd
parent bb52ef60b5caa8f973523eda15d3c3941f298e63
author Assar <assar@permabit.com> Wed, 14 Sep 2005 16:59:25 -0400
committer Marcelo Tosatti <marcelo@dmt.cnet> Thu, 22 Sep 2005 13:11:18 -0300

    [PATCH] nfs client: handle long symlinks properly

    In 2.4.31, the v2/3 nfs readlink accepts too long symlinks.
    I have tested this by having a server return long symlinks.

diff --git a/fs/nfs/nfs2xdr.c b/fs/nfs/nfs2xdr.c
--- a/fs/nfs/nfs2xdr.c
+++ b/fs/nfs/nfs2xdr.c
@@ -571,8 +571,11 @@ nfs_xdr_readlinkres(struct rpc_rqst *req
        strlen = (u32*)kmap(rcvbuf->pages[0]);
        /* Convert length of symlink */
        len = ntohl(*strlen);
-       if (len > rcvbuf->page_len)
-               len = rcvbuf->page_len;
+       if (len >= rcvbuf->page_len - sizeof(u32) || len > NFS2_MAXPATHLEN) {
+               printk(KERN_WARNING "NFS: server returned giant symlink!\n");
+               kunmap(rcvbuf->pages[0]);
+               return -ENAMETOOLONG;
+        }
        *strlen = len;
        /* NULL terminate the string we got */
        string = (char *)(strlen + 1);
diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c
--- a/fs/nfs/nfs3xdr.c
+++ b/fs/nfs/nfs3xdr.c
@@ -759,8 +759,11 @@ nfs3_xdr_readlinkres(struct rpc_rqst *re
        strlen = (u32*)kmap(rcvbuf->pages[0]);
        /* Convert length of symlink */
        len = ntohl(*strlen);
-       if (len > rcvbuf->page_len)
-               len = rcvbuf->page_len;
+       if (len >= rcvbuf->page_len - sizeof(u32)) {
+               printk(KERN_WARNING "NFS: server returned giant symlink!\n");
+               kunmap(rcvbuf->pages[0]);
+               return -ENAMETOOLONG;
+        }
        *strlen = len;
        /* NULL terminate the string we got */
        string = (char *)(strlen + 1);

Re: [PATCH] nfs client: handle long symlinks properly

[Trond Myklebust]

to den 22.09.2005 Klokka 13:14 (-0300) skreiv Marcelo Tosatti:
> commit 87e03738fc15dc3ea4acde3a5dcb5f84b6b6152b
> tree be323c0a65d7e380ad04cad1c3a80015a82056dd
> parent bb52ef60b5caa8f973523eda15d3c3941f298e63
> author Assar <assar@permabit.com> Wed, 14 Sep 2005 16:59:25 -0400
> committer Marcelo Tosatti <marcelo@dmt.cnet> Thu, 22 Sep 2005 13:11:18 -0300
> 
>     [PATCH] nfs client: handle long symlinks properly
> 
>     In 2.4.31, the v2/3 nfs readlink accepts too long symlinks.
>     I have tested this by having a server return long symlinks.
> 
> diff --git a/fs/nfs/nfs2xdr.c b/fs/nfs/nfs2xdr.c
> --- a/fs/nfs/nfs2xdr.c
> +++ b/fs/nfs/nfs2xdr.c
> @@ -571,8 +571,11 @@ nfs_xdr_readlinkres(struct rpc_rqst *req
>         strlen = (u32*)kmap(rcvbuf->pages[0]);
>         /* Convert length of symlink */
>         len = ntohl(*strlen);
> -       if (len > rcvbuf->page_len)
> -               len = rcvbuf->page_len;
> +       if (len >= rcvbuf->page_len - sizeof(u32) || len > NFS2_MAXPATHLEN) {

Shouldn't that be

	if (len > rcvbuf->page_len - sizeof(u32) || len > NFS2_MAXPATHLEN)

? As long as we use page_len == PAGE_SIZE, we probably don't care, but
if someone some day decides to set a different value for page_len, then
we want to make sure that we don't end up overflowing the buffer when we
NUL-terminate.

> +               printk(KERN_WARNING "NFS: server returned giant symlink!\n");

Please make this a dprintk().

> +               kunmap(rcvbuf->pages[0]);
> +               return -ENAMETOOLONG;
> +        }
>         *strlen = len;
>         /* NULL terminate the string we got */
>         string = (char *)(strlen + 1);
> diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c
> --- a/fs/nfs/nfs3xdr.c
> +++ b/fs/nfs/nfs3xdr.c
> @@ -759,8 +759,11 @@ nfs3_xdr_readlinkres(struct rpc_rqst *re
>         strlen = (u32*)kmap(rcvbuf->pages[0]);
>         /* Convert length of symlink */
>         len = ntohl(*strlen);
> -       if (len > rcvbuf->page_len)
> -               len = rcvbuf->page_len;
> +       if (len >= rcvbuf->page_len - sizeof(u32)) {

Ditto.

> +               printk(KERN_WARNING "NFS: server returned giant symlink!\n");

...and ditto.

> +               kunmap(rcvbuf->pages[0]);
> +               return -ENAMETOOLONG;
> +        }
>         *strlen = len;
>         /* NULL terminate the string we got */
>         string = (char *)(strlen + 1);

Cheers,
  Trond

Re: [PATCH] nfs client: handle long symlinks properly

[Assar]

Trond Myklebust <trond.myklebust@fys.uio.no> writes:
> > diff --git a/fs/nfs/nfs2xdr.c b/fs/nfs/nfs2xdr.c
> > --- a/fs/nfs/nfs2xdr.c
> > +++ b/fs/nfs/nfs2xdr.c
> > @@ -571,8 +571,11 @@ nfs_xdr_readlinkres(struct rpc_rqst *req
> >         strlen = (u32*)kmap(rcvbuf->pages[0]);
> >         /* Convert length of symlink */
> >         len = ntohl(*strlen);
> > -       if (len > rcvbuf->page_len)
> > -               len = rcvbuf->page_len;
> > +       if (len >= rcvbuf->page_len - sizeof(u32) || len > NFS2_MAXPATHLEN) {
> 
> Shouldn't that be
> 
> 	if (len > rcvbuf->page_len - sizeof(u32) || len > NFS2_MAXPATHLEN)
> 
> ? As long as we use page_len == PAGE_SIZE, we probably don't care, but
> if someone some day decides to set a different value for page_len, then
> we want to make sure that we don't end up overflowing the buffer when we
> NUL-terminate.

Wouldn't len == rcvbuf->page_len - sizeof(u32) mean that there isn't
room for writing the terminating NUL?

Re: [PATCH] nfs client: handle long symlinks properly

[Trond Myklebust]

to den 22.09.2005 Klokka 12:38 (-0400) skreiv Assar:
> Wouldn't len == rcvbuf->page_len - sizeof(u32) mean that there isn't
> room for writing the terminating NUL?

Yep. My bad... I was screwing up the test in my mind.

Cheers,
  Trond