NFS: Return a delegation the client fails to record
Subsystems (2)
Chuck Lever
maintainerPATCH 1/1 NFS: Return a delegation the client fails to record
When an NFS server grants a delegation in an OPEN reply, nfs_inode_set_delegation() records it on the client. However, three of its error flows return without sending DELEGRETURN.
A delegation can be relinquished only by DELEGRETURN (RFC 8881 Section 20.2.4), so dropping one silently leaves the server believing the client still holds it. If the server happens to recall that delegation, the client answers CB_RECALL with NFS4ERR_BADHANDLE because it has no record of the stateid. The server revokes the delegation and moves it onto its cl_revoked list, because the client never sends the FREE_STATEID that would drain it. Every subsequent SEQUENCE reply then carries SEQ4_STATUS_RECALLABLE_STATE_REVOKED, and the client's state manager loops issuing TEST_STATEID across its delegations without ever clearing the condition.
The window is easy to reach now that a server offers a write delegation on any write OPEN: a delegation recalled for one opener races a re-open that the server answers with a fresh write delegation.
Instead of dropping it, hand the delegation back during these error flows.
--- fs/nfs/delegation.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-)
fs/nfs/delegation.c +14 -4
@@ -440,11 +440,14 @@ int nfs_inode_set_delegation(struct inode *inode, const struct cred *cred,struct nfs_inode *nfsi =NFS_I (inode);struct nfs_delegation *delegation, *old_delegation;struct nfs_delegation *freeme =NULL ; +bool orphaned = false;int status =0 ; delegation =kmalloc_obj (*delegation,GFP_KERNEL_ACCOUNT ); -if (delegation ==NULL ) +if (delegation ==NULL ) { +nfs4_proc_delegreturn (inode, cred, stateid,NULL ,0 );return -ENOMEM ; + }nfs4_stateid_copy (&delegation->stateid, stateid);refcount_set (&delegation->refcount,1 ); delegation->type = type;
@@ -493,11 +496,15 @@ int nfs_inode_set_delegation(struct inode *inode, const struct cred *cred, goto out; }if (test_and_set_bit (NFS_DELEGATION_RETURNING , - &old_delegation->flags)) + &old_delegation->flags)) { + orphaned = true; goto out; + } } -if (!nfs_detach_delegations_locked (nfsi, old_delegation, clp)) +if (!nfs_detach_delegations_locked (nfsi, old_delegation, clp)) { + orphaned = true; goto out; + } freeme = old_delegation; add_new:/*
@@ -532,8 +539,11 @@ int nfs_inode_set_delegation(struct inode *inode, const struct cred *cred,nfs_update_delegated_mtime (inode); out:spin_unlock (&clp->cl_lock); -if (delegation !=NULL ) +if (delegation !=NULL ) { +if (orphaned) +nfs_do_return_delegation (inode, delegation,0 );__nfs_free_delegation (delegation); + }if (freeme !=NULL ) {nfs_do_return_delegation (inode, freeme,0 );nfs_mark_delegation_revoked (server, freeme);
Signature · 1 line
2.54.0
Chuck Lever
maintainerQuoted 83 lines · Chuck Lever
When an NFS server grants a delegation in an OPEN reply,
nfs_inode_set_delegation()records it on the client. However, three of its error flows return without sending DELEGRETURN.A delegation can be relinquished only by DELEGRETURN (RFC 8881 Section 20.2.4), so dropping one silently leaves the server believing the client still holds it. If the server happens to recall that delegation, the client answers
CB_RECALLwithNFS4ERR_BADHANDLEbecause it has no record of the stateid. The server revokes the delegation and moves it onto itscl_revokedlist, because the client never sends theFREE_STATEIDthat would drain it. Every subsequent SEQUENCE reply then carriesSEQ4_STATUS_RECALLABLE_STATE_REVOKED, and the client's state manager loops issuingTEST_STATEIDacross its delegations without ever clearing the condition.The window is easy to reach now that a server offers a write delegation on any write OPEN: a delegation recalled for one opener races a re-open that the server answers with a fresh write delegation.
Instead of dropping it, hand the delegation back during these error flows.
Fixes ade04647dd56 ("NFSv4: Ensure we honour NFS_DELEGATION_RETURNINGin
nfs_inode_set_delegation()")Signed-off-by Chuck Lever <[email protected]>--- fs/nfs/delegation.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-)fs/nfs/delegation.c +4
-1@@ -440,11 +440,14 @@ int nfs_inode_set_delegation(struct inode *inode,const struct cred *cred, struct nfs_inode *nfsi = NFS_I(inode);struct nfs_delegation *delegation, *old_delegation;struct nfs_delegation *freeme =NULL ; +bool orphaned = false;int status =0 ; delegation =kmalloc_obj (*delegation,GFP_KERNEL_ACCOUNT ); -if (delegation ==NULL ) +if (delegation ==NULL ) { +nfs4_proc_delegreturn (inode, cred, stateid,NULL ,0 );return -ENOMEM ; + }nfs4_stateid_copy (&delegation->stateid, stateid);refcount_set (&delegation->refcount,1 );fs/nfs/delegation.c +6
-2delegation->type = type;@@ -493,11 +496,15 @@ int nfs_inode_set_delegation(struct inode *inode,const struct cred *cred, goto out; }if (test_and_set_bit (NFS_DELEGATION_RETURNING , - &old_delegation->flags)) + &old_delegation->flags)) { + orphaned = true; goto out; + } } -if (!nfs_detach_delegations_locked (nfsi, old_delegation, clp)) +if (!nfs_detach_delegations_locked (nfsi, old_delegation, clp)) { + orphaned = true; goto out; + } freeme = old_delegation; add_new:fs/nfs/delegation.c +4
-1/* @@ -532,8 +539,11 @@ int nfs_inode_set_delegation(struct inode *inode,const struct cred *cred,nfs_update_delegated_mtime (inode ); out:spin_unlock (&clp->cl_lock); -if (delegation !=NULL ) +if (delegation !=NULL ) { +if (orphaned) +nfs_do_return_delegation (inode, delegation,0 );__nfs_free_delegation (delegation); + }if (freeme !=NULL ) {nfs_do_return_delegation (inode, freeme,0 );
Quoted 3 lines
nfs_mark_delegation_revoked(server, freeme);Signature · 1 line
2.54.0
Friendly bump! What is the disposition of this submission?
Signature · 1 line
Chuck Lever
Anna Schumaker
maintainerHi Chuck,
Quoted 81 lines · Chuck Lever
Quoted 80 lines · Chuck Lever
When an NFS server grants a delegation in an OPEN reply,
nfs_inode_set_delegation()records it on the client. However, three of its error flows return without sending DELEGRETURN.A delegation can be relinquished only by DELEGRETURN (RFC 8881 Section 20.2.4), so dropping one silently leaves the server believing the client still holds it. If the server happens to recall that delegation, the client answers
CB_RECALLwithNFS4ERR_BADHANDLEbecause it has no record of the stateid. The server revokes the delegation and moves it onto itscl_revokedlist, because the client never sends theFREE_STATEIDthat would drain it. Every subsequent SEQUENCE reply then carriesSEQ4_STATUS_RECALLABLE_STATE_REVOKED, and the client's state manager loops issuingTEST_STATEIDacross its delegations without ever clearing the condition.The window is easy to reach now that a server offers a write delegation on any write OPEN: a delegation recalled for one opener races a re-open that the server answers with a fresh write delegation.
Instead of dropping it, hand the delegation back during these error flows.
Fixes ade04647dd56 ("NFSv4: Ensure we honour NFS_DELEGATION_RETURNINGin
nfs_inode_set_delegation()")Signed-off-by Chuck Lever <[email protected]>--- fs/nfs/delegation.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-)fs/nfs/delegation.c +4
-1@@ -440,11 +440,14 @@ int nfs_inode_set_delegation(struct inode *inode,const struct cred *cred, struct nfs_inode *nfsi = NFS_I(inode);struct nfs_delegation *delegation, *old_delegation;struct nfs_delegation *freeme =NULL ; +bool orphaned = false;int status =0 ; delegation =kmalloc_obj (*delegation,GFP_KERNEL_ACCOUNT ); -if (delegation ==NULL ) +if (delegation ==NULL ) { +nfs4_proc_delegreturn (inode, cred, stateid,NULL ,0 );return -ENOMEM ; + }nfs4_stateid_copy (&delegation->stateid, stateid);refcount_set (&delegation->refcount,1 );fs/nfs/delegation.c +6
-2delegation->type = type;@@ -493,11 +496,15 @@ int nfs_inode_set_delegation(struct inode *inode,const struct cred *cred, goto out; }if (test_and_set_bit (NFS_DELEGATION_RETURNING , - &old_delegation->flags)) + &old_delegation->flags)) { + orphaned = true; goto out; + } } -if (!nfs_detach_delegations_locked (nfsi, old_delegation, clp)) +if (!nfs_detach_delegations_locked (nfsi, old_delegation, clp)) { + orphaned = true; goto out; + } freeme = old_delegation; add_new:fs/nfs/delegation.c +3
-1/* @@ -532,8 +539,11 @@ int nfs_inode_set_delegation(struct inode *inode,const struct cred *cred,nfs_update_delegated_mtime (inode ); out:spin_unlock (&clp->cl_lock); -if (delegation !=NULL ) +if (delegation !=NULL ) { +if (orphaned) +nfs_do_return_delegation (inode, delegation,0 );__nfs_free_delegation (delegation);
Quoted 8 lines
Quoted 6 lines
fs/nfs/delegation.c +1
-0+ }if (freeme !=NULL ) {nfs_do_return_delegation (inode, freeme,0 );
nfs_mark_delegation_revoked(server, freeme);Signature · 1 line
2.54.0Friendly bump! What is the disposition of this submission?
I have it in a testing branch that I haven't pushed out yet. I'm planning to include it in the next round of bugfixes that I send out.
Anna
Quoted 4 lines
Signature · 2 lines
-- Chuck Lever