Merge tag 'nfs-for-5.1-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client updates from Trond Myklebust:
 "Highlights include:

  Stable fixes:
   - Fixes for NFS I/O request leakages
   - Fix error handling paths in the NFS I/O recoalescing code
   - Reinitialise NFSv4.1 sequence results before retransmitting a
     request
   - Fix a soft lockup in the delegation recovery code
   - Bulk destroy of layouts needs to be safe w.r.t. umount
   - Prevent thundering herd issues when the SUNRPC socket is not
     connected
   - Respect RPC call timeouts when retrying transmission

  Features:
   - Convert rpc auth layer to use xdr_streams
   - Config option to disable insecure RPCSEC_GSS crypto types
   - Reduce size of RPC receive buffers
   - Readdirplus optimization by cache mechanism
   - Convert SUNRPC socket send code to use iov_iter()
   - SUNRPC micro-optimisations to avoid indirect calls
   - Add support for the pNFS LAYOUTERROR operation and use it with the
     pNFS/flexfiles driver
   - Add trace events to report non-zero NFS status codes
   - Various removals of unnecessary dprintks

  Bugfixes and cleanups:
   - Fix a number of sparse warnings and documentation format warnings
   - Fix nfs_parse_devname to not modify it's argument
   - Fix potential corruption of page being written through pNFS/blocks
   - fix xfstest generic/099 failures on nfsv3
   - Avoid NFSv4.1 "false retries" when RPC calls are interrupted
   - Abort I/O early if the pNFS/flexfiles layout segment was
     invalidated
   - Avoid unnecessary pNFS/flexfiles layout invalidations"

* tag 'nfs-for-5.1-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (90 commits)
  SUNRPC: Take the transport send lock before binding+connecting
  SUNRPC: Micro-optimise when the task is known not to be sleeping
  SUNRPC: Check whether the task was transmitted before rebind/reconnect
  SUNRPC: Remove redundant calls to RPC_IS_QUEUED()
  SUNRPC: Clean up
  SUNRPC: Respect RPC call timeouts when retrying transmission
  SUNRPC: Fix up RPC back channel transmission
  SUNRPC: Prevent thundering herd when the socket is not connected
  SUNRPC: Allow dynamic allocation of back channel slots
  NFSv4.1: Bump the default callback session slot count to 16
  SUNRPC: Convert remaining GFP_NOIO, and GFP_NOWAIT sites in sunrpc
  NFS/flexfiles: Clean up mirror DS initialisation
  NFS/flexfiles: Remove dead code in ff_layout_mirror_valid()
  NFS/flexfile: Simplify nfs4_ff_layout_select_ds_stateid()
  NFS/flexfile: Simplify nfs4_ff_layout_ds_version()
  NFS/flexfiles: Simplify ff_layout_get_ds_cred()
  NFS/flexfiles: Simplify nfs4_ff_find_or_create_ds_client()
  NFS/flexfiles: Simplify nfs4_ff_layout_select_ds_fh()
  NFS/flexfiles: Speed up read failover when DSes are down
  NFS/flexfiles: Don't invalidate DS deviceids for being unresponsive
  ...

1 files changed, 10 insertions, 7 deletions

diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
index adc3c40cc733..28956c70100a 100644
--- a/net/sunrpc/sched.c
+++ b/net/sunrpc/sched.c
@@ -19,6 +19,7 @@
 #include <linux/spinlock.h>
 #include <linux/mutex.h>
 #include <linux/freezer.h>
+#include <linux/sched/mm.h>
 
 #include <linux/sunrpc/clnt.h>
 
@@ -784,8 +785,7 @@ void rpc_exit(struct rpc_task *task, int status)
 {
 	task->tk_status = status;
 	task->tk_action = rpc_exit_task;
-	if (RPC_IS_QUEUED(task))
-		rpc_wake_up_queued_task(task->tk_waitqueue, task);
+	rpc_wake_up_queued_task(task->tk_waitqueue, task);
 }
 EXPORT_SYMBOL_GPL(rpc_exit);
 
@@ -902,7 +902,10 @@ void rpc_execute(struct rpc_task *task)
 
 static void rpc_async_schedule(struct work_struct *work)
 {
+	unsigned int pflags = memalloc_nofs_save();
+
 	__rpc_execute(container_of(work, struct rpc_task, u.tk_work));
+	memalloc_nofs_restore(pflags);
 }
 
 /**
@@ -921,16 +924,13 @@ static void rpc_async_schedule(struct work_struct *work)
  * Most requests are 'small' (under 2KiB) and can be serviced from a
  * mempool, ensuring that NFS reads and writes can always proceed,
  * and that there is good locality of reference for these buffers.
- *
- * In order to avoid memory starvation triggering more writebacks of
- * NFS requests, we avoid using GFP_KERNEL.
  */
 int rpc_malloc(struct rpc_task *task)
 {
 	struct rpc_rqst *rqst = task->tk_rqstp;
 	size_t size = rqst->rq_callsize + rqst->rq_rcvsize;
 	struct rpc_buffer *buf;
-	gfp_t gfp = GFP_NOIO | __GFP_NOWARN;
+	gfp_t gfp = GFP_NOFS;
 
 	if (RPC_IS_SWAPPER(task))
 		gfp = __GFP_MEMALLOC | GFP_NOWAIT | __GFP_NOWARN;
@@ -1011,7 +1011,7 @@ static void rpc_init_task(struct rpc_task *task, const struct rpc_task_setup *ta
 static struct rpc_task *
 rpc_alloc_task(void)
 {
-	return (struct rpc_task *)mempool_alloc(rpc_task_mempool, GFP_NOIO);
+	return (struct rpc_task *)mempool_alloc(rpc_task_mempool, GFP_NOFS);
 }
 
 /*
@@ -1067,7 +1067,10 @@ static void rpc_free_task(struct rpc_task *task)
 
 static void rpc_async_release(struct work_struct *work)
 {
+	unsigned int pflags = memalloc_nofs_save();
+
 	rpc_free_task(container_of(work, struct rpc_task, u.tk_work));
+	memalloc_nofs_restore(pflags);
 }
 
 static void rpc_release_resources_task(struct rpc_task *task)