Skip to contentSkip to conversation

NFSv4.2: Fix missing removal of SLAB_ACCOUNT on kmem_cache allocation

PATCH In review [email protected] 3 participants 10 messages Apr 7, 2022

Subsystems (3)
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

PATCH v2 1/8 NFSv4.2: Fix missing removal of SLAB_ACCOUNT on kmem_cache allocation

From: Muchun Song <[email protected]>

The commit 5c60e89e71f8 ("NFSv4.2: Fix up an invalid combination of memory allocation flags") has stripped GFP_KERNEL_ACCOUNT down to GFP_KERNEL, however, it forgot to remove SLAB_ACCOUNT from kmem_cache allocation. It means that memory is still limited by kmemcg. This patch also fix a NULL pointer reference issue [1] reported by NeilBrown.

Fixes 5c60e89e71f8 ("NFSv4.2: Fix up an invalid combination of memory allocation flags")
Fixes 5abc1e37afa0 ("mm: list_lru: allocate list_lru_one only when needed")
---
 fs/nfs/nfs42xattr.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/nfs/nfs42xattr.c +1 -1
index ad3405c64b9e..e7b34f7e0614 100644
@@ -997,7 +997,7 @@ int __init nfs4_xattr_cache_init(void)
 
 	nfs4_xattr_cache_cachep = kmem_cache_create("nfs4_xattr_cache_cache",
 	    sizeof(struct nfs4_xattr_cache), 0,
-	    (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|SLAB_ACCOUNT),
+	    (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
 	    nfs4_xattr_cache_init_once);
 	if (nfs4_xattr_cache_cachep == NULL)
 		return -ENOMEM;
Signature · 1 line
2.35.1
author maintainer
#1 lore raw

PATCH v2 2/8 SUNRPC: Handle ENOMEM in call_transmit_status()

From: Trond Myklebust <[email protected]>

Both call_transmit() and call_bc_transmit() can now return ENOMEM, so let's make sure that we handle the errors gracefully.

Fixes 0472e4766049 ("SUNRPC: Convert socket page send code to use iov_iter()")
---
 net/sunrpc/clnt.c | 2 ++
 1 file changed, 2 insertions(+)
net/sunrpc/clnt.c +2 -0
index 3c7407104d54..07328f1d3885 100644
@@ -2200,6 +2200,7 @@ call_transmit_status(struct rpc_task *task)
 		 * socket just returned a connection error,
 		 * then hold onto the transport lock.
 		 */
+	case -ENOMEM:
 	case -ENOBUFS:
 		rpc_delay(task, HZ>>2);
 		fallthrough;
@@ -2283,6 +2284,7 @@ call_bc_transmit_status(struct rpc_task *task)
 	case -ENOTCONN:
 	case -EPIPE:
 		break;
+	case -ENOMEM:
 	case -ENOBUFS:
 		rpc_delay(task, HZ>>2);
 		fallthrough;
Signature · 1 line
2.35.1
author maintainer
#2 lore raw

PATCH v2 3/8 SUNRPC: Handle low memory situations in call_status()

From: Trond Myklebust <[email protected]>

We need to handle ENFILE, ENOBUFS, and ENOMEM, because xprt_wake_pending_tasks() can be called with any one of these due to socket creation failures.

Fixes b61d59fffd3e ("SUNRPC: xs_tcp_connect_worker{4,6}: merge common code")
---
 net/sunrpc/clnt.c | 5 +++++
 1 file changed, 5 insertions(+)
net/sunrpc/clnt.c +5 -0
index 07328f1d3885..6757b0fa5367 100644
@@ -2367,6 +2367,11 @@ call_status(struct rpc_task *task)
 	case -EPIPE:
 	case -EAGAIN:
 		break;
+	case -ENFILE:
+	case -ENOBUFS:
+	case -ENOMEM:
+		rpc_delay(task, HZ>>2);
+		break;
 	case -EIO:
 		/* shutdown or soft timeout */
 		goto out_exit;
Signature · 1 line
2.35.1
author maintainer
#3 lore raw

PATCH v2 4/8 NFSv4/pnfs: Handle RPC allocation errors in nfs4_proc_layoutget

From: Trond Myklebust <[email protected]>

If rpc_run_task() fails due to an allocation error, then bail out early.

Fixes 910ad38697d9 ("NFS: Fix memory allocation in rpc_alloc_task()")
---
 fs/nfs/nfs4proc.c | 2 ++
 1 file changed, 2 insertions(+)
fs/nfs/nfs4proc.c +2 -0
index e3f5b380cefe..16106f805ffa 100644
@@ -9615,6 +9615,8 @@ nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
 	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
 
 	task = rpc_run_task(&task_setup_data);
+	if (IS_ERR(task))
+		return ERR_CAST(task);
 
 	status = rpc_wait_for_completion_task(task);
 	if (status != 0)
Signature · 1 line
2.35.1
author maintainer
#4 lore raw

PATCH v2 5/8 NFS: Ensure rpc_run_task() cannot fail in nfs_async_rename()

From: Trond Myklebust <[email protected]>

Ensure the call to rpc_run_task() cannot fail by preallocating the rpc_task.

Fixes 910ad38697d9 ("NFS: Fix memory allocation in rpc_alloc_task()")
---
 fs/nfs/unlink.c         | 1 +
 include/linux/nfs_xdr.h | 1 +
 2 files changed, 2 insertions(+)
fs/nfs/unlink.c +1 -0
index 5fa11e1aca4c..6f325e10056c 100644
@@ -347,6 +347,7 @@ nfs_async_rename(struct inode *old_dir, struct inode *new_dir,
 	data = kzalloc(sizeof(*data), GFP_KERNEL);
 	if (data == NULL)
 		return ERR_PTR(-ENOMEM);
+	task_setup_data.task = &data->task;
 	task_setup_data.callback_data = data;
 
 	data->cred = get_current_cred();
include/linux/nfs_xdr.h +1 -0
index 49ba486aea5f..2863e5a69c6a 100644
@@ -1694,6 +1694,7 @@ struct nfs_unlinkdata {
 struct nfs_renamedata {
 	struct nfs_renameargs	args;
 	struct nfs_renameres	res;
+	struct rpc_task		task;
 	const struct cred	*cred;
 	struct inode		*old_dir;
 	struct dentry		*old_dentry;
Signature · 1 line
2.35.1
author maintainer
#5 lore raw

PATCH v2 6/8 SUNRPC: Handle allocation failure in rpc_new_task()

From: Trond Myklebust <[email protected]>

If the call to rpc_alloc_task() fails, then ensure that the calldata is released, and that rpc_run_task() and rpc_run_bc_task() bail out early.

Fixes 910ad38697d9 ("NFS: Fix memory allocation in rpc_alloc_task()")
---
 net/sunrpc/clnt.c  | 7 +++++++
 net/sunrpc/sched.c | 5 +++++
 2 files changed, 12 insertions(+)
net/sunrpc/clnt.c +7 -0
index 6757b0fa5367..af0174d7ce5a 100644
@@ -1127,6 +1127,8 @@ struct rpc_task *rpc_run_task(const struct rpc_task_setup *task_setup_data)
 	struct rpc_task *task;
 
 	task = rpc_new_task(task_setup_data);
+	if (IS_ERR(task))
+		return task;
 
 	if (!RPC_IS_ASYNC(task))
 		task->tk_flags |= RPC_TASK_CRED_NOREF;
@@ -1227,6 +1229,11 @@ struct rpc_task *rpc_run_bc_task(struct rpc_rqst *req)
 	 * Create an rpc_task to send the data
 	 */
 	task = rpc_new_task(&task_setup_data);
+	if (IS_ERR(task)) {
+		xprt_free_bc_request(req);
+		return task;
+	}
+
 	xprt_init_bc_request(req, task);
 
 	task->tk_action = call_bc_encode;
net/sunrpc/sched.c +5 -0
index b258b87a3ec2..7f70c1e608b7 100644
@@ -1128,6 +1128,11 @@ struct rpc_task *rpc_new_task(const struct rpc_task_setup *setup_data)
 
 	if (task == NULL) {
 		task = rpc_alloc_task();
+		if (task == NULL) {
+			rpc_release_calldata(setup_data->callback_ops,
+					     setup_data->callback_data);
+			return ERR_PTR(-ENOMEM);
+		}
 		flags = RPC_TASK_DYNAMIC;
 	}
 
Signature · 1 line
2.35.1
author maintainer
#6 lore raw

PATCH v2 7/8 SUNRPC: svc_tcp_sendmsg() should handle errors from xdr_alloc_bvec()

From: Trond Myklebust <[email protected]>

The allocation is done with GFP_KERNEL, but it could still fail in a low memory situation.

Fixes 4a85a6a3320b ("SUNRPC: Handle TCP socket sends with kernel_sendpage() again")
---
 net/sunrpc/svcsock.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
net/sunrpc/svcsock.c +3 -1
index 478f857cdaed..6ea3d87e1147 100644
@@ -1096,7 +1096,9 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
 	int ret;
 
 	*sentp = 0;
-	xdr_alloc_bvec(xdr, GFP_KERNEL);
+	ret = xdr_alloc_bvec(xdr, GFP_KERNEL);
+	if (ret < 0)
+		return ret;
 
 	ret = kernel_sendmsg(sock, &msg, &rm, 1, rm.iov_len);
 	if (ret < 0)
Signature · 1 line
2.35.1
author maintainer
#7 lore raw

PATCH v2 8/8 SUNRPC: Move the call to xprt_send_pagedata() out of xprt_sock_sendmsg()

From: Trond Myklebust <[email protected]>

The client and server have different requirements for their memory allocation, so move the allocation of the send buffer out of the socket send code that is common to both.

Fixes b2648015d452 ("SUNRPC: Make the rpciod and xprtiod slab allocation modes consistent")
---
 net/sunrpc/socklib.c  |  6 ------
 net/sunrpc/svcsock.c  |  9 ++++++---
 net/sunrpc/xprtsock.c | 15 +++++++++++++--
 3 files changed, 19 insertions(+), 11 deletions(-)
net/sunrpc/socklib.c +0 -6
index 05b38bf68316..71ba4cf513bc 100644
@@ -221,12 +221,6 @@ static int xprt_send_kvec(struct socket *sock, struct msghdr *msg,
 static int xprt_send_pagedata(struct socket *sock, struct msghdr *msg,
 			      struct xdr_buf *xdr, size_t base)
 {
-	int err;
-
-	err = xdr_alloc_bvec(xdr, rpc_task_gfp_mask());
-	if (err < 0)
-		return err;
-
 	iov_iter_bvec(&msg->msg_iter, WRITE, xdr->bvec, xdr_buf_pagecount(xdr),
 		      xdr->page_len + xdr->page_base);
 	return xprt_sendmsg(sock, msg, base + xdr->page_base);
net/sunrpc/svcsock.c +6 -3
index 6ea3d87e1147..cc35ec433400 100644
@@ -579,15 +579,18 @@ static int svc_udp_sendto(struct svc_rqst *rqstp)
 	if (svc_xprt_is_dead(xprt))
 		goto out_notconn;
 
+	err = xdr_alloc_bvec(xdr, GFP_KERNEL);
+	if (err < 0)
+		goto out_unlock;
+
 	err = xprt_sock_sendmsg(svsk->sk_sock, &msg, xdr, 0, 0, &sent);
-	xdr_free_bvec(xdr);
 	if (err == -ECONNREFUSED) {
 		/* ICMP error on earlier request. */
 		err = xprt_sock_sendmsg(svsk->sk_sock, &msg, xdr, 0, 0, &sent);
-		xdr_free_bvec(xdr);
 	}
+	xdr_free_bvec(xdr);
 	trace_svcsock_udp_send(xprt, err);
-
+out_unlock:
 	mutex_unlock(&xprt->xpt_mutex);
 	if (err < 0)
 		return err;
net/sunrpc/xprtsock.c +13 -2
index a34a15750122..e16568f9a82d 100644
@@ -825,9 +825,14 @@ static int xs_stream_nospace(struct rpc_rqst *req, bool vm_wait)
 static int
 xs_stream_prepare_request(struct rpc_rqst *req)
 {
+	gfp_t gfp = rpc_task_gfp_mask();
+	int ret;
+
+	ret = xdr_alloc_bvec(&req->rq_snd_buf, gfp);
+	if (ret < 0)
+		return ret;
 	xdr_free_bvec(&req->rq_rcv_buf);
-	return xdr_alloc_bvec(
-		&req->rq_rcv_buf, GFP_KERNEL | __GFP_NORETRY | __GFP_NOWARN);
+	return xdr_alloc_bvec(&req->rq_rcv_buf, gfp);
 }
 
 /*
@@ -956,6 +961,9 @@ static int xs_udp_send_request(struct rpc_rqst *req)
 	if (!xprt_request_get_cong(xprt, req))
 		return -EBADSLT;
 
+	status = xdr_alloc_bvec(xdr, rpc_task_gfp_mask());
+	if (status < 0)
+		return status;
 	req->rq_xtime = ktime_get();
 	status = xprt_sock_sendmsg(transport->sock, &msg, xdr, 0, 0, &sent);
 
@@ -2557,6 +2565,9 @@ static int bc_sendto(struct rpc_rqst *req)
 	int err;
 
 	req->rq_xtime = ktime_get();
+	err = xdr_alloc_bvec(xdr, rpc_task_gfp_mask());
+	if (err < 0)
+		return err;
 	err = xprt_sock_sendmsg(transport->sock, &msg, xdr, 0, marker, &sent);
 	xdr_free_bvec(xdr);
 	if (err < 0 || sent != (xdr->len + sizeof(marker)))
Signature · 1 line
2.35.1
#8 lore raw

Re: [PATCH v2 7/8] SUNRPC: svc_tcp_sendmsg() should handle errors from xdr_alloc_bvec()

In reply to [email protected]
Quoted 21 lines · On Thu, 2022-04-07 at 14:46 -0400, [email protected] wrote:

From: Trond Myklebust <[email protected]>

The allocation is done with GFP_KERNEL, but it could still fail in a low
memory situation.

Fixes 4a85a6a3320b ("SUNRPC: Handle TCP socket sends with

kernel_sendpage() again")

---
 net/sunrpc/svcsock.c | 4 +++-

 1 file changed, 3 insertions(+), 1 deletion(-)
net/sunrpc/svcsock.c +0 -0
index 478f857cdaed..6ea3d87e1147 100644
@@ -1096,7 +1096,9 @@ static int svc_tcp_sendmsg(struct socket *sock,
 struct xdr_buf *xdr,
         int ret;

Quoted 8 lines

        *sentp = 0;
-       xdr_alloc_bvec(xdr, GFP_KERNEL);
+       ret = xdr_alloc_bvec(xdr, GFP_KERNEL);
+       if (ret < 0)
+               return ret;

        ret = kernel_sendmsg(sock, &msg, &rm, 1, rm.iov_len); if (ret < 0)

Chuck,

Do you mind if I send this and the 8/8 as part of the client pull request? I saw this while I was digging through the code and separating out the client and server uses of xdr_alloc_bvec().

Signature · 3 lines
Trond Myklebust
Linux NFS client maintainer, Hammerspace
[email protected]
#9 lore raw

Re: [PATCH v2 7/8] SUNRPC: svc_tcp_sendmsg() should handle errors from xdr_alloc_bvec()

Quoted 38 lines · Trond Myklebust
Quoted 28 lines · On Thu, 2022-04-07 at 14:46 -0400, [email protected] wrote:

From: Trond Myklebust <[email protected]>

The allocation is done with GFP_KERNEL, but it could still fail in a low
memory situation.

Fixes 4a85a6a3320b ("SUNRPC: Handle TCP socket sends with

kernel_sendpage() again")

---
 net/sunrpc/svcsock.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
net/sunrpc/svcsock.c +3 -1
index 478f857cdaed..6ea3d87e1147 100644
@@ -1096,7 +1096,9 @@ static int svc_tcp_sendmsg(struct socket *sock,
 struct xdr_buf *xdr,
        int ret;
 
        *sentp = 0;
-       xdr_alloc_bvec(xdr, GFP_KERNEL);
+       ret = xdr_alloc_bvec(xdr, GFP_KERNEL);
+       if (ret < 0)
+               return ret;
 
        ret = kernel_sendmsg(sock, &msg, &rm, 1, rm.iov_len);
Quoted 1 line
        if (ret < 0)

Chuck,

Quoted 3 lines

Do you mind if I send this and the 8/8 as part of the client pull request? I saw this while I was digging through the code and separating out the client and server uses of xdr_alloc_bvec().

I browsed through these a few minutes ago. I don't see any technical issues. But as you're listed as a maintainer of the SUNRPC code, I didn't think I needed to give explicit permission.

--
Chuck Lever