@@ -211,10 +211,16 @@ static struct nfs4_ff_layout_mirror *ff_layout_alloc_mirror(gfp_t gfp_flags)
static void ff_layout_free_mirror(struct nfs4_ff_layout_mirror *mirror)
{
+ struct rpc_cred *cred;
+
ff_layout_remove_mirror(mirror);
kfree(mirror->fh_versions);
- if (mirror->cred)
- put_rpccred(mirror->cred);
+ cred = rcu_dereference(mirror->ro_cred);
+ if (cred)
+ put_rpccred(cred);
+ cred = rcu_dereference(mirror->rw_cred);
+ if (cred)
+ put_rpccred(cred);
nfs4_ff_layout_put_deviceid(mirror->mirror_ds);
kfree(mirror);
}
@@ -408,7 +414,7 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh,
struct nfs4_deviceid devid;
struct nfs4_deviceid_node *idnode;
struct auth_cred acred = {0};
- struct rpc_cred *cred;
+ struct rpc_cred __rcu *cred;
u32 ds_count, fh_count, id;
int j;
@@ -499,23 +505,33 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh,
acred.gid = make_kgid(&init_user_ns, id);
/* find the cred for it */
- cred = rpc_lookup_generic_cred(&acred, 0, gfp_flags);
+ rcu_assign_pointer(cred, rpc_lookup_generic_cred(&acred, 0, gfp_flags));
if (IS_ERR(cred)) {
rc = PTR_ERR(cred);
goto out_err_free;
}
- rcu_assign_pointer(fls->mirror_array[i]->cred, cred);
+ if (lgr->range.iomode == IOMODE_READ)
+ rcu_assign_pointer(fls->mirror_array[i]->ro_cred, cred);
+ else
+ rcu_assign_pointer(fls->mirror_array[i]->rw_cred, cred);
mirror = ff_layout_add_mirror(lh, fls->mirror_array[i]);
if (mirror != fls->mirror_array[i]) {
/* swap cred ptrs so free_mirror will clean up old */
- fls->mirror_array[i]->cred = xchg(&mirror->cred, cred);
+ if (lgr->range.iomode == IOMODE_READ) {
+ cred = xchg(&mirror->ro_cred, cred);
+ rcu_assign_pointer(fls->mirror_array[i]->ro_cred, cred);
+ } else {
+ cred = xchg(&mirror->rw_cred, cred);
+ rcu_assign_pointer(fls->mirror_array[i]->rw_cred, cred);
+ }
ff_layout_free_mirror(fls->mirror_array[i]);
fls->mirror_array[i] = mirror;
}
- dprintk("%s: uid %u gid %u\n", __func__,
+ dprintk("%s: iomode %s uid %u gid %u\n", __func__,
+ lgr->range.iomode == IOMODE_READ ? "READ" : "RW",
from_kuid(&init_user_ns, acred.uid),
from_kgid(&init_user_ns, acred.gid));
}
@@ -76,7 +76,8 @@ struct nfs4_ff_layout_mirror {
u32 fh_versions_cnt;
struct nfs_fh *fh_versions;
nfs4_stateid stateid;
- struct rpc_cred *cred;
+ struct rpc_cred __rcu *ro_cred;
+ struct rpc_cred __rcu *rw_cred;
atomic_t ref;
spinlock_t lock;
struct nfs4_ff_layoutstat read_stat;
@@ -305,9 +305,12 @@ int ff_layout_track_ds_error(struct nfs4_flexfile_layout *flo,
static struct rpc_cred *
ff_layout_get_mirror_cred(struct nfs4_ff_layout_mirror *mirror, u32 iomode)
{
- struct rpc_cred *cred, **pcred;
+ struct rpc_cred *cred, __rcu **pcred;
- pcred = &mirror->cred;
+ if (iomode == IOMODE_READ)
+ pcred = &mirror->ro_cred;
+ else
+ pcred = &mirror->rw_cred;
rcu_read_lock();
do {
A mirror can be shared between multiple layouts, even with different iomodes. That makes stats gathering simpler, but it causes a problem when we get different creds in READ vs. RW layouts. The current code drops the newer credentials onto the floor when this occurs. That's problematic when you fetch a READ layout first, and then a RW. If the READ layout doesn't have the correct creds to do a write, then writes will fail. We could just overwrite the READ credentials with the RW ones, but that would break the ability for the server to fence the layout for reads if things go awry. We need to be able to revert to the earlier READ creds if the RW layout is returned afterward. The simplest fix is to just keep two sets of creds per mirror. One for READ layouts and one for RW, and then use the appropriate set depending on the iomode of the layout segment. Also fix up some RCU nits that sparse found. Signed-off-by: Jeff Layton <jeff.layton@primarydata.com> --- fs/nfs/flexfilelayout/flexfilelayout.c | 30 +++++++++++++++++++++++------- fs/nfs/flexfilelayout/flexfilelayout.h | 3 ++- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 7 +++++-- 3 files changed, 30 insertions(+), 10 deletions(-)