NFSv4: Fix a potential state reclaim deadlock
Subsystems (2)
[email protected]
maintainerPATCH 1/3 NFSv4: Fix a potential state reclaim deadlock
From: Trond Myklebust <[email protected]>
If the server reboots while we are engaged in a delegation return, and there is a pNFS layout with return-on-close set, then the current code can end up deadlocking in pnfs_roc() when nfs_inode_set_delegation() tries to return the old delegation.
Now that delegreturn actually uses its own copy of the stateid, it should be safe to just always update the delegation stateid in place.
--- fs/nfs/delegation.c | 36 +++++++++++++++++------------------- 1 file changed, 17 insertions(+), 19 deletions(-)
fs/nfs/delegation.c +17 -19
@@ -228,8 +228,7 @@ static int nfs_delegation_claim_opens(struct inode *inode,* */ void nfs_inode_reclaim_delegation (struct inode *inode,const struct cred *cred, -fmode_t type, -const nfs4_stateid *stateid, +fmode_t type,const nfs4_stateid *stateid,unsigned long pagemod_limit) {struct nfs_delegation *delegation;
@@ -239,25 +238,24 @@ void nfs_inode_reclaim_delegation(struct inode *inode, const struct cred *cred, delegation =rcu_dereference (NFS_I (inode)->delegation);if (delegation !=NULL ) {spin_lock (&delegation->lock); -if (nfs4_is_valid_delegation (delegation,0 )) { -nfs4_stateid_copy (&delegation->stateid, stateid); - delegation->type = type; - delegation->pagemod_limit = pagemod_limit; - oldcred = delegation->cred; - delegation->cred =get_cred (cred); -clear_bit (NFS_DELEGATION_NEED_RECLAIM , - &delegation->flags); -spin_unlock (&delegation->lock); -rcu_read_unlock (); -put_cred (oldcred); -trace_nfs4_reclaim_delegation (inode, type); -return ; - } -/* We appear to have raced with a delegation return. */ +nfs4_stateid_copy (&delegation->stateid, stateid); + delegation->type = type; + delegation->pagemod_limit = pagemod_limit; + oldcred = delegation->cred; + delegation->cred =get_cred (cred); +clear_bit (NFS_DELEGATION_NEED_RECLAIM , &delegation->flags); +if (test_and_clear_bit (NFS_DELEGATION_REVOKED , + &delegation->flags)) +atomic_long_inc (&nfs_active_delegations);spin_unlock (&delegation->lock); +rcu_read_unlock (); +put_cred (oldcred); +trace_nfs4_reclaim_delegation (inode, type); + }else { +rcu_read_unlock (); +nfs_inode_set_delegation (inode, cred, type, stateid, + pagemod_limit); } -rcu_read_unlock (); -nfs_inode_set_delegation (inode, cred, type, stateid, pagemod_limit); }static int nfs_do_return_delegation (struct inode *inode,struct nfs_delegation *delegation,int issync)
Signature · 1 line
2.37.3
[email protected]
maintainerPATCH 2/3 NFSv4.1: Handle RECLAIM_COMPLETE trunking errors
From: Trond Myklebust <[email protected]>
If RECLAIM_COMPLETE sets the NFS4CLNT_BIND_CONN_TO_SESSION flag, then we need to loop back in order to handle it.
--- fs/nfs/nfs4state.c | 1 + 1 file changed, 1 insertion(+)
fs/nfs/nfs4state.c +1 -0
@@ -2671,6 +2671,7 @@ static void nfs4_state_manager(struct nfs_client *clp)if (status <0 ) goto out_error;nfs4_state_end_reclaim_reboot (clp); +continue ; }/* Detect expired delegations... */
Signature · 1 line
2.37.3
[email protected]
maintainerPATCH 3/3 NFSv4.1: We must always send RECLAIM_COMPLETE after a reboot
From: Trond Myklebust <[email protected]>
Currently, we are only guaranteed to send RECLAIM_COMPLETE if we have open state to recover. Fix the client to always send RECLAIM_COMPLETE after setting up the lease.
--- fs/nfs/nfs4state.c | 1 + 1 file changed, 1 insertion(+)
fs/nfs/nfs4state.c +1 -0
@@ -1787,6 +1787,7 @@ static void nfs4_state_mark_reclaim_helper(struct nfs_client *clp,static void nfs4_state_start_reclaim_reboot (struct nfs_client *clp) { +set_bit (NFS4CLNT_RECLAIM_REBOOT , &clp->cl_state);/* Mark all delegations for reclaim */ nfs_delegation_mark_reclaim (clp);nfs4_state_mark_reclaim_helper (clp, nfs4_state_mark_reclaim_reboot);
Signature · 1 line
2.37.3