NFSv4.1: try SECINFO_NO_NAME flavs until one works
Subsystems (2)
PATCH 1/1 NFSv4.1: try SECINFO_NO_NAME flavs until one works
Call nfs4_lookup_root_sec for each flavor returned by SECINFO_NO_NAME until one works.
One example of a situation this fixes:
- server configured for krb5
- server principal somehow gets deleted from KDC
- server still thinking krb is good, sends krb5 as first entry in SECINFO_NO_NAME response
- client tries krb5, but this fails without even sending an RPC because gssd's requests to the KDC can't find the server's principal
--- fs/nfs/nfs4proc.c | 30 +++++++++++++++++++++++++++--- 1 file changed, 27 insertions(+), 3 deletions(-)
This is version 2 of the patch.
fs/nfs/nfs4proc.c +27 -3
@@ -7566,6 +7566,8 @@ nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,struct page *page;rpc_authflavor_t flavor;struct nfs4_secinfo_flavors *flavors; +struct nfs4_secinfo4 *secinfo; +int i; page =alloc_page (GFP_KERNEL );if (!page) {
@@ -7587,9 +7589,31 @@ nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,if (err) goto out_freepage; - flavor =nfs_find_best_sec (flavors); -if (err ==0 ) - err =nfs4_lookup_root_sec (server, fhandle, info, flavor); +for (i =0 ; i < flavors->num_flavors; i++) { + secinfo = &flavors->flavors[i]; + +switch (secinfo->flavor) { +case RPC_AUTH_NULL : +case RPC_AUTH_UNIX : +case RPC_AUTH_GSS : + flavor =rpcauth_get_pseudoflavor (secinfo->flavor, + &secinfo->flavor_info); +break ; +default : + flavor =RPC_AUTH_MAXFLAVOR ; +break ; + } + +if (flavor !=RPC_AUTH_MAXFLAVOR ) { + err =nfs4_lookup_root_sec (server, fhandle, + info, flavor); +if (!err) +break ; + } + } + +if (flavor ==RPC_AUTH_MAXFLAVOR ) + err = -EPERM ; out_freepage:put_page (page);
Signature · 1 line
1.7.12.4 (Apple Git-37)
Hi Dros,
Quoted 50 lines · Weston Andros Adamson
Call
nfs4_lookup_root_secfor each flavor returned bySECINFO_NO_NAMEuntil one works.One example of a situation this fixes:
- server configured for krb5
- server principal somehow gets deleted from KDC
- server still thinking krb is good, sends krb5 as first entry inSECINFO_NO_NAMEresponse
- client tries krb5, but this fails without even sending an RPC because gssd's requests to the KDC can't find the server's principalSigned-off-by Weston Andros Adamson <[email protected]>--- fs/nfs/nfs4proc.c | 30 +++++++++++++++++++++++++++--- 1 file changed, 27 insertions(+), 3 deletions(-)This is version 2 of the patch.
fs/nfs/nfs4proc.c +12
-3@@ -7566,6 +7566,8 @@ nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,struct page *page;rpc_authflavor_t flavor;struct nfs4_secinfo_flavors *flavors; +struct nfs4_secinfo4 *secinfo; +int i; page =alloc_page (GFP_KERNEL );if (!page) {@@ -7587,9 +7589,31 @@ nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,if (err) goto out_freepage; - flavor =nfs_find_best_sec (flavors); -if (err ==0 ) - err =nfs4_lookup_root_sec (server, fhandle, info, flavor); +for (i =0 ; i < flavors->num_flavors; i++) { + secinfo = &flavors->flavors[i]; + +switch (secinfo->flavor) { +case RPC_AUTH_NULL : +case RPC_AUTH_UNIX : +case RPC_AUTH_GSS : + flavor =rpcauth_get_pseudoflavor (secinfo->flavor, + &secinfo->flavor_info); +break ;
fs/nfs/nfs4proc.c nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle, · line 7611
Quoted 10 lines
fs/nfs/nfs4proc.c +10
-0+default : + flavor =RPC_AUTH_MAXFLAVOR ; +break ; + } + +if (flavor !=RPC_AUTH_MAXFLAVOR ) { + err =nfs4_lookup_root_sec (server, fhandle, + info, flavor); +if (!err) +break ;
I think we should only try the next sec flavor if nfs4_lookup_root_sec() returns -EACCESS, since this error means there was a problem with rpcauth_create() for the given flavor. Other errors should be passed on to the user, since they happen farther on in the mount process.
Anna
Quoted 12 lines
fs/nfs/nfs4proc.c +5
-1+ } + } + +if (flavor ==RPC_AUTH_MAXFLAVOR ) + err = -EPERM ; out_freepage:put_page (page); --1.7.12.4 (Apple Git-37)
Quoted 72 lines · Anna Schumaker
Hi Dros,
Quoted 50 lines · Weston Andros Adamson
Call
nfs4_lookup_root_secfor each flavor returned bySECINFO_NO_NAMEuntil one works.One example of a situation this fixes:
- server configured for krb5
- server principal somehow gets deleted from KDC
- server still thinking krb is good, sends krb5 as first entry inSECINFO_NO_NAMEresponse
- client tries krb5, but this fails without even sending an RPC because gssd's requests to the KDC can't find the server's principalSigned-off-by Weston Andros Adamson <[email protected]>---
fs/nfs/nfs4proc.c| 30 +++++++++++++++++++++++++++---1 file changed, 27 insertions(+), 3 deletions(-)This is version 2 of the patch.
fs/nfs/nfs4proc.c +12
-3@@ -7566,6 +7566,8 @@ nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,struct page *page;rpc_authflavor_t flavor;struct nfs4_secinfo_flavors *flavors; +struct nfs4_secinfo4 *secinfo; +int i; page =alloc_page (GFP_KERNEL );if (!page) {@@ -7587,9 +7589,31 @@ nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,if (err) goto out_freepage; - flavor =nfs_find_best_sec (flavors); -if (err ==0 ) - err =nfs4_lookup_root_sec (server, fhandle, info, flavor); +for (i =0 ; i < flavors->num_flavors; i++) { + secinfo = &flavors->flavors[i]; + +switch (secinfo->flavor) { +case RPC_AUTH_NULL : +case RPC_AUTH_UNIX : +case RPC_AUTH_GSS : + flavor =rpcauth_get_pseudoflavor (secinfo->flavor, + &secinfo->flavor_info); +break ;fs/nfs/nfs4proc.c nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle, · line 7611
Quoted 10 lines
fs/nfs/nfs4proc.c +10
-0+default : + flavor =RPC_AUTH_MAXFLAVOR ; +break ; + } + +if (flavor !=RPC_AUTH_MAXFLAVOR ) { + err =nfs4_lookup_root_sec (server, fhandle, + info, flavor); +if (!err) +break ;I think we should only try the next sec flavor if
nfs4_lookup_root_sec()returns-EACCESS, since this error means there was a problem withrpcauth_create()for the given flavor. Other errors should be passed on to the user, since they happen farther on in the mount process.
Quoted 2 lines
Anna
Yeah, I thought about that, but it seemed to me that it'd be nice to try the next flavor no matter what the error was. The other option is the mount just failing. That said, I don't feel strongly either way.
Anyone else want to weigh in?
-dros
Quoted 16 lines
Quoted 8 lines
fs/nfs/nfs4proc.c +5
-0+ } + } + +if (flavor ==RPC_AUTH_MAXFLAVOR ) + err = -EPERM ;
out_freepage:Quoted 4 lines
put_page(page);--
1.7.12.4 (Apple Git-37)--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html