[v2] NFSD: fix management of pending async copies

Commit Message

Olga Kornievskaia Dec. 17, 2024, 9:18 p.m. UTC

Currently the pending_async_copies count is decremented just
before a struct nfsd4_copy is destroyed. After commit aa0ebd21df9c
("NFSD: Add nfsd4_copy time-to-live") nfsd4_copy structures sticks
around for 10 lease periods after the COPY itself has completed,
the pending_async_copies count stays high for a long time. This
causes NFSD to avoid the use of background copy even though the
actual background copy workload might no longer be running.

In this patch, decrement pending_async_copies once async copy thread
is done processing the copy work.

Fixes: aa0ebd21df9c ("NFSD: Add nfsd4_copy time-to-live")
Signed-off-by: Olga Kornievskaia <[e-mail]>
---
 fs/nfsd/nfs4proc.c | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

Comments

Chuck Lever Dec. 17, 2024, 9:37 p.m. UTC | #1

From: Chuck Lever <[e-mail]>

On Tue, 17 Dec 2024 16:18:12 -0500, Olga Kornievskaia wrote:                                              
> Currently the pending_async_copies count is decremented just
> before a struct nfsd4_copy is destroyed. After commit aa0ebd21df9c
> ("NFSD: Add nfsd4_copy time-to-live") nfsd4_copy structures sticks
> around for 10 lease periods after the COPY itself has completed,
> the pending_async_copies count stays high for a long time. This
> causes NFSD to avoid the use of background copy even though the
> actual background copy workload might no longer be running.
> 
> [...]                                                                        

Applied to nfsd-fixes for v6.13, thanks!                                                                

[1/1] NFSD: fix management of pending async copies
      commit: 9048cf05a17a7bc26f0b8e2e53750b1237303970                                                                      

--                                                                              
Chuck Lever

Patch

diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index f8a10f90bc7a..ad44ad49274f 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -1347,7 +1347,6 @@  static void nfs4_put_copy(struct nfsd4_copy *copy)
 {
 	if (!refcount_dec_and_test(&copy->refcount))
 		return;
-	atomic_dec(&copy->cp_nn->pending_async_copies);
 	kfree(copy->cp_src);
 	kfree(copy);
 }
@@ -1870,6 +1869,7 @@  static int nfsd4_do_async_copy(void *data)
 	set_bit(NFSD4_COPY_F_COMPLETED, &copy->cp_flags);
 	trace_nfsd_copy_async_done(copy);
 	nfsd4_send_cb_offload(copy);
+	atomic_dec(&copy->cp_nn->pending_async_copies);
 	return 0;
 }
 
@@ -1927,19 +1927,19 @@  nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 		/* Arbitrary cap on number of pending async copy operations */
 		if (atomic_inc_return(&nn->pending_async_copies) >
 				(int)rqstp->rq_pool->sp_nrthreads)
-			goto out_err;
+			goto out_dec_async_copy_err;
 		async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
 		if (!async_copy->cp_src)
-			goto out_err;
+			goto out_dec_async_copy_err;
 		if (!nfs4_init_copy_state(nn, copy))
-			goto out_err;
+			goto out_dec_async_copy_err;
 		memcpy(&result->cb_stateid, &copy->cp_stateid.cs_stid,
 			sizeof(result->cb_stateid));
 		dup_copy_fields(copy, async_copy);
 		async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
 				async_copy, "%s", "copy thread");
 		if (IS_ERR(async_copy->copy_task))
-			goto out_err;
+			goto out_dec_async_copy_err;
 		spin_lock(&async_copy->cp_clp->async_lock);
 		list_add(&async_copy->copies,
 				&async_copy->cp_clp->async_copies);
@@ -1954,6 +1954,9 @@  nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 	trace_nfsd_copy_done(copy, status);
 	release_copy_files(copy);
 	return status;
+out_dec_async_copy_err:
+	if (async_copy)
+		atomic_dec(&nn->pending_async_copies);
 out_err:
 	if (nfsd4_ssc_is_inter(copy)) {
 		/*