Client-side OFFLOAD_STATUS implementation
Subsystems (2)
[email protected]
maintainerv2 cover Client-side OFFLOAD_STATUS implementation
From: Chuck Lever <[email protected]>
SCSI implementation experience has shown that an interrupt-only COPY offload implementation is not reliable. There are too many common scenarios where the client can miss the completion interrupt (in our case, this is a CB_OFFLOAD callback).
Therefore, a polling mechanism is needed. The NFSv4.2 protocol provides one in the form of the new OFFLOAD_STATUS operation. Linux NFSD implements OFFLOAD_STATUS already. This series adds a Linux NFS client implementation of the OFFLOAD_STATUS operation that can query the state of a background COPY on the server.
These patches are also available here:
https://git.kernel.org/pub/scm/linux/kernel/git/cel/linux.git/log/?h=fix-async-copy
Changes since v1:
- nfs42_proc_offload_status() now uses a synchronous RPC
Chuck Lever (7):
NFS: CB_OFFLOAD can return NFS4ERR_DELAY
NFS: Fix typo in OFFLOAD_CANCEL comment
NFS: Rename struct nfs4_offloadcancel_data
NFS: Implement NFSv4.2's OFFLOAD_STATUS XDR
NFS: Implement NFSv4.2's OFFLOAD_STATUS operation
NFS: Use NFSv4.2's OFFLOAD_STATUS operation
NFS: Refactor trace_nfs4_offload_cancel
fs/nfs/callback_proc.c | 2 +- fs/nfs/nfs42proc.c | 188 ++++++++++++++++++++++++++++++++++---- fs/nfs/nfs42xdr.c | 88 +++++++++++++++++- fs/nfs/nfs4proc.c | 3 +- fs/nfs/nfs4trace.h | 11 ++- fs/nfs/nfs4xdr.c | 1 + include/linux/nfs4.h | 1 + include/linux/nfs_fs_sb.h | 1 + include/linux/nfs_xdr.h | 5 +- 9 files changed, 275 insertions(+), 25 deletions(-)
Signature · 1 line
2.47.0
[email protected]
maintainerPATCH v2 1/7 NFS: CB_OFFLOAD can return NFS4ERR_DELAY
From: Chuck Lever <[email protected]>
RFC 7862 permits the callback service to respond to a CB_OFFLOAD operation with NFS4ERR_DELAY. Use that instead of NFS4ERR_SERVERFAULT for temporary memory allocation failure, as that is more consistent with how other operations report memory allocation failure.
--- fs/nfs/callback_proc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fs/nfs/callback_proc.c +1 -1
@@ -718,7 +718,7 @@ __be32 nfs4_callback_offload(void *data, void *dummy, copy =kzalloc (sizeof (struct nfs4_copy_state ),GFP_KERNEL );if (!copy) -return htonl (NFS4ERR_SERVERFAULT ); +return cpu_to_be32 (NFS4ERR_DELAY );spin_lock (&cps->clp->cl_lock);rcu_read_lock ();
Signature · 1 line
2.47.0
[email protected]
maintainerPATCH v2 2/7 NFS: Fix typo in OFFLOAD_CANCEL comment
In reply to [email protected]From: Chuck Lever <[email protected]>
--- fs/nfs/nfs42xdr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fs/nfs/nfs42xdr.c +1 -1
@@ -549,7 +549,7 @@ static void nfs4_xdr_enc_copy(struct rpc_rqst *req, }/* -* Encode OFFLOAD_CANEL request +* Encode OFFLOAD_CANCEL request */ static void nfs4_xdr_enc_offload_cancel(struct rpc_rqst *req, struct xdr_stream *xdr,
Signature · 1 line
2.47.0
[email protected]
maintainerPATCH v2 3/7 NFS: Rename struct nfs4_offloadcancel_data
In reply to [email protected]From: Chuck Lever <[email protected]>
Refactor: This struct can be used unchanged for the new OFFLOAD_STATUS implementation, so give it a more generic name.
--- fs/nfs/nfs42proc.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)
fs/nfs/nfs42proc.c +9 -9
@@ -498,15 +498,15 @@ ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,return err; } -struct nfs42_offloadcancel_data { +struct nfs42_offload_data {struct nfs_server *seq_server;struct nfs42_offload_status_args args;struct nfs42_offload_status_res res; }; -static void nfs42_offload_cancel_prepare (struct rpc_task *task,void *calldata) +static void nfs42_offload_prepare (struct rpc_task *task,void *calldata) { -struct nfs42_offloadcancel_data *data = calldata; +struct nfs42_offload_data *data = calldata; nfs4_setup_sequence(data->seq_server->nfs_client, &data->args.osa_seq_args,
@@ -515,7 +515,7 @@ static void nfs42_offload_cancel_prepare(struct rpc_task *task, void *calldata)static void nfs42_offload_cancel_done (struct rpc_task *task,void *calldata) { -struct nfs42_offloadcancel_data *data = calldata; +struct nfs42_offload_data *data = calldata;trace_nfs4_offload_cancel (&data->args, task->tk_status);nfs41_sequence_done (task, &data->res.osr_seq_res);
@@ -525,22 +525,22 @@ static void nfs42_offload_cancel_done(struct rpc_task *task, void *calldata)rpc_restart_call_prepare (task); } -static void nfs42_free_offloadcancel_data (void *data) +static void nfs42_offload_release (void *data) {kfree (data); }static const struct rpc_call_ops nfs42_offload_cancel_ops = { - .rpc_call_prepare = nfs42_offload_cancel_prepare, + .rpc_call_prepare = nfs42_offload_prepare, .rpc_call_done = nfs42_offload_cancel_done, - .rpc_release = nfs42_free_offloadcancel_data, + .rpc_release = nfs42_offload_release, };static int nfs42_do_offload_cancel_async (struct file *dst,nfs4_stateid *stateid) {struct nfs_server *dst_server =NFS_SERVER (file_inode (dst)); -struct nfs42_offloadcancel_data *data =NULL ; +struct nfs42_offload_data *data =NULL ;struct nfs_open_context *ctx =nfs_file_open_context (dst);struct rpc_task *task;struct rpc_message msg = {
@@ -559,7 +559,7 @@ static int nfs42_do_offload_cancel_async(struct file *dst,if (!(dst_server->caps &NFS_CAP_OFFLOAD_CANCEL ))return -EOPNOTSUPP ; - data =kzalloc (sizeof (struct nfs42_offloadcancel_data ),GFP_KERNEL ); + data =kzalloc (sizeof (struct nfs42_offload_data ),GFP_KERNEL );if (data ==NULL )return -ENOMEM ;
Signature · 1 line
2.47.0
[email protected]
maintainerPATCH v2 4/7 NFS: Implement NFSv4.2's OFFLOAD_STATUS XDR
In reply to [email protected]From: Chuck Lever <[email protected]>
Add XDR encoding and decoding functions for the NFSv4.2 OFFLOAD_STATUS operation.
--- fs/nfs/nfs42xdr.c | 86 +++++++++++++++++++++++++++++++++++++++++ fs/nfs/nfs4xdr.c | 1 + include/linux/nfs4.h | 1 + include/linux/nfs_xdr.h | 5 ++- 4 files changed, 91 insertions(+), 2 deletions(-)
fs/nfs/nfs42xdr.c +86 -0
@@ -35,6 +35,11 @@#define encode_offload_cancel_maxsz (op_encode_hdr_maxsz + \ XDR_QUADLEN(NFS4_STATEID_SIZE))#define decode_offload_cancel_maxsz (op_decode_hdr_maxsz) +#define encode_offload_status_maxsz (op_encode_hdr_maxsz + \ + XDR_QUADLEN(NFS4_STATEID_SIZE)) +#define decode_offload_status_maxsz (op_decode_hdr_maxsz + \ + 2 /* osr_count */ + \ + 2 /* osr_complete */)#define encode_copy_notify_maxsz (op_encode_hdr_maxsz + \ XDR_QUADLEN(NFS4_STATEID_SIZE) + \ 1 +/* nl4_type */ \
@@ -143,6 +148,14 @@ decode_sequence_maxsz + \ decode_putfh_maxsz + \ decode_offload_cancel_maxsz) +#define NFS4_enc_offload_status_sz (compound_encode_hdr_maxsz + \ + encode_sequence_maxsz + \ + encode_putfh_maxsz + \ + encode_offload_status_maxsz) +#define NFS4_dec_offload_status_sz (compound_decode_hdr_maxsz + \ + decode_sequence_maxsz + \ + decode_putfh_maxsz + \ + decode_offload_status_maxsz)#define NFS4_enc_copy_notify_sz (compound_encode_hdr_maxsz + \ encode_putfh_maxsz + \ encode_copy_notify_maxsz)
@@ -343,6 +356,14 @@ static void encode_offload_cancel(struct xdr_stream *xdr,encode_nfs4_stateid (xdr, &args->osa_stateid); } +static void encode_offload_status (struct xdr_stream *xdr, +const struct nfs42_offload_status_args *args, +struct compound_hdr *hdr) +{ +encode_op_hdr (xdr,OP_OFFLOAD_STATUS , decode_offload_status_maxsz, hdr); +encode_nfs4_stateid (xdr, &args->osa_stateid); +} +static void encode_copy_notify (struct xdr_stream *xdr,const struct nfs42_copy_notify_args *args,struct compound_hdr *hdr)
@@ -567,6 +588,25 @@ static void nfs4_xdr_enc_offload_cancel(struct rpc_rqst *req,encode_nops (&hdr); } +/* +* Encode OFFLOAD_STATUS request +*/ +static void nfs4_xdr_enc_offload_status (struct rpc_rqst *req, +struct xdr_stream *xdr, +const void *data) +{ +const struct nfs42_offload_status_args *args = data; +struct compound_hdr hdr = { + .minorversion =nfs4_xdr_minorversion (&args->osa_seq_args), + }; + +encode_compound_hdr (xdr, req, &hdr); +encode_sequence (xdr, &args->osa_seq_args, &hdr); +encode_putfh (xdr, args->osa_src_fh, &hdr); +encode_offload_status (xdr, args, &hdr); +encode_nops (&hdr); +} +/* * Encode COPY_NOTIFY request */
@@ -919,6 +959,26 @@ static int decode_offload_cancel(struct xdr_stream *xdr,return decode_op_hdr (xdr,OP_OFFLOAD_CANCEL ); } +static int decode_offload_status (struct xdr_stream *xdr, +struct nfs42_offload_status_res *res) +{ +ssize_t result; +int status; + + status =decode_op_hdr (xdr,OP_OFFLOAD_STATUS ); +if (status) +return status; +/* osr_count */ +if (xdr_stream_decode_u64 (xdr, &res->osr_count) <0 ) +return -EIO ; +/* osr_complete<1> */ + result =xdr_stream_decode_uint32_array (xdr, &res->osr_complete,1 ); +if (result <0 ) +return -EIO ; + res->complete_count = result; +return 0 ; +} +static int decode_copy_notify (struct xdr_stream *xdr,struct nfs42_copy_notify_res *res) {
@@ -1368,6 +1428,32 @@ static int nfs4_xdr_dec_offload_cancel(struct rpc_rqst *rqstp,return status; } +/* +* Decode OFFLOAD_STATUS response +*/ +static int nfs4_xdr_dec_offload_status (struct rpc_rqst *rqstp, +struct xdr_stream *xdr, +void *data) +{ +struct nfs42_offload_status_res *res = data; +struct compound_hdr hdr; +int status; + + status =decode_compound_hdr (xdr, &hdr); +if (status) + goto out; + status =decode_sequence (xdr, &res->osr_seq_res, rqstp); +if (status) + goto out; + status =decode_putfh (xdr); +if (status) + goto out; + status =decode_offload_status (xdr, res); + +out: +return status; +} +/* * Decode COPY_NOTIFY response */
fs/nfs/nfs4xdr.c +1 -0
@@ -7702,6 +7702,7 @@ const struct rpc_procinfo nfs4_procedures[] = {PROC42 (CLONE , enc_clone, dec_clone),PROC42 (COPY , enc_copy, dec_copy),PROC42 (OFFLOAD_CANCEL , enc_offload_cancel, dec_offload_cancel), +PROC42 (OFFLOAD_STATUS , enc_offload_status, dec_offload_status),PROC42 (COPY_NOTIFY , enc_copy_notify, dec_copy_notify),PROC (LOOKUPP , enc_lookupp, dec_lookupp), PROC42(LAYOUTERROR, enc_layouterror, dec_layouterror),
include/linux/nfs4.h +1 -0
@@ -695,6 +695,7 @@ enum {NFSPROC4_CLNT_LISTXATTRS ,NFSPROC4_CLNT_REMOVEXATTR ,NFSPROC4_CLNT_READ_PLUS , + NFSPROC4_CLNT_OFFLOAD_STATUS, };/* nfs41 types */
include/linux/nfs_xdr.h +3 -2
@@ -1520,8 +1520,9 @@ struct nfs42_offload_status_args {struct nfs42_offload_status_res {struct nfs4_sequence_res osr_seq_res; -uint64_t osr_count; -int osr_status; +u64 osr_count; +int complete_count; +u32 osr_complete; };struct nfs42_copy_notify_args {
Signature · 1 line
2.47.0
[email protected]
maintainerPATCH v2 5/7 NFS: Implement NFSv4.2's OFFLOAD_STATUS operation
In reply to [email protected]From: Chuck Lever <[email protected]>
Enable the Linux NFS client to observe the progress of an offloaded asynchronous COPY operation. This new operation will be put to use in a subsequent patch.
--- fs/nfs/nfs42proc.c | 101 ++++++++++++++++++++++++++++++++++++++ fs/nfs/nfs4proc.c | 3 +- include/linux/nfs_fs_sb.h | 1 + 3 files changed, 104 insertions(+), 1 deletion(-)
fs/nfs/nfs42proc.c +101 -0
@@ -21,6 +21,8 @@#define NFSDBG_FACILITY NFSDBG_PROCstatic int nfs42_do_offload_cancel_async (struct file *dst,nfs4_stateid *std); +static int nfs42_proc_offload_status (struct file *file,nfs4_stateid *stateid, +u64 *copied);static void nfs42_set_netaddr (struct file *filep,struct nfs42_netaddr *naddr) {
@@ -582,6 +584,105 @@ static int nfs42_do_offload_cancel_async(struct file *dst,return status; } +static int +_nfs42_proc_offload_status (struct nfs_server *server,struct file *file, +struct nfs42_offload_data *data) +{ +struct nfs_open_context *ctx =nfs_file_open_context (file); +struct rpc_message msg = { + .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OFFLOAD_STATUS ], + .rpc_argp = &data->args, + .rpc_resp = &data->res, + .rpc_cred = ctx->cred, + }; +int status; + + status =nfs4_call_sync (server->client, server, &msg, + &data->args.osa_seq_args, + &data->res.osr_seq_res,1 ); +switch (status) { +case 0 : +break ; + +case -NFS4ERR_ADMIN_REVOKED : +case -NFS4ERR_BAD_STATEID : +case -NFS4ERR_OLD_STATEID : +/* +* Server does not recognize the COPY stateid. CB_OFFLOAD +* could have purged it, or server might have rebooted. +* Since COPY stateids don't have an associated inode, +* avoid triggering state recovery. +*/ + status = -EBADF ; +break ; +case -NFS4ERR_NOTSUPP : +case -ENOTSUPP : +case -EOPNOTSUPP : + server->caps &= ~NFS_CAP_OFFLOAD_STATUS ; + status = -EOPNOTSUPP ; +break ; + } + +return status; +} + +/** +* nfs42_proc_offload_status - Poll completion status of an async copy operation +* @dst: handle of file being copied into +* @stateid: copy stateid (from async COPY result) +* @copied: OUT: number of bytes copied so far +* +* Return values: +* %0: Server returned an NFS4_OK completion status +* %-EINPROGRESS: Server returned no completion status +* %-EREMOTEIO: Server returned an error completion status +* %-EBADF: Server did not recognize the copy stateid +* %-EOPNOTSUPP: Server does not support OFFLOAD_STATUS +* %-ERESTARTSYS: Wait interrupted by signal +* +* Other negative errnos indicate the client could not complete the +* request. +*/ +static int __maybe_unused +nfs42_proc_offload_status (struct file *dst,nfs4_stateid *stateid,u64 *copied) +{ +struct inode *inode =file_inode (dst); +struct nfs_server *server =NFS_SERVER (inode); +struct nfs4_exception exception = { + .inode = inode, + }; +struct nfs42_offload_data *data; +int status; + +if (!(server->caps &NFS_CAP_OFFLOAD_STATUS )) +return -EOPNOTSUPP ; + + data =kzalloc (sizeof (*data),GFP_KERNEL ); +if (!data) +return -ENOMEM ; + data->seq_server = server; + data->args.osa_src_fh =NFS_FH (inode); +memcpy (&data->args.osa_stateid, stateid, +sizeof (data->args.osa_stateid)); + exception.stateid = &data->args.osa_stateid; +do { + status =_nfs42_proc_offload_status (server, dst, data); +if (status == -EOPNOTSUPP ) + goto out; + status =nfs4_handle_exception (server, status, &exception); + }while (exception.retry); + + *copied = data->res.osr_count; +if (!data->res.complete_count) + status = -EINPROGRESS ; +else if (data->res.osr_complete !=NFS_OK ) + status = -EREMOTEIO ; + +out: +kfree (data); +return status; +} +static int _nfs42_proc_copy_notify (struct file *src,struct file *dst,struct nfs42_copy_notify_args *args,struct nfs42_copy_notify_res *res)
fs/nfs/nfs4proc.c +2 -1
@@ -10769,7 +10769,8 @@ static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = { | NFS_CAP_CLONE |NFS_CAP_LAYOUTERROR |NFS_CAP_READ_PLUS - | NFS_CAP_MOVEABLE, + |NFS_CAP_MOVEABLE + | NFS_CAP_OFFLOAD_STATUS, .init_client = nfs41_init_client, .shutdown_client = nfs41_shutdown_client, .match_stateid = nfs41_match_stateid,
include/linux/nfs_fs_sb.h +1 -0
@@ -290,6 +290,7 @@ struct nfs_server {#define NFS_CAP_CASE_INSENSITIVE (1U << 6)#define NFS_CAP_CASE_PRESERVING (1U << 7)#define NFS_CAP_REBOOT_LAYOUTRETURN (1U << 8) +#define NFS_CAP_OFFLOAD_STATUS (1U << 9)#define NFS_CAP_OPEN_XOR (1U << 12)#define NFS_CAP_DELEGTIME (1U << 13)#define NFS_CAP_POSIX_LOCK (1U << 14)
Signature · 1 line
2.47.0
[email protected]
maintainerPATCH v2 6/7 NFS: Use NFSv4.2's OFFLOAD_STATUS operation
In reply to [email protected]From: Chuck Lever <[email protected]>
We've found that there are cases where a transport disconnection results in the loss of callback RPCs. NFS servers typically do not retransmit callback operations after a disconnect.
This can be a problem for the Linux NFS client's current implementation of asynchronous COPY, which waits indefinitely for a CB_OFFLOAD callback. If a transport disconnect occurs while an async COPY is running, there's a good chance the client will never get the completing CB_OFFLOAD.
Fix this by implementing the OFFLOAD_STATUS operation so that the Linux NFS client can probe the NFS server if it doesn't see a CB_OFFLOAD in a reasonable amount of time.
This patch implements a simplistic check. As future work, the client might also be able to detect whether there is no forward progress on the request asynchronous COPY operation, and CANCEL it.
--- fs/nfs/nfs42proc.c | 70 ++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 59 insertions(+), 11 deletions(-)
fs/nfs/nfs42proc.c +59 -11
@@ -175,6 +175,25 @@ int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len)return err; } +/* Wait this long before checking progress on a COPY operation */ +enum { +NFS42_COPY_TIMEOUT =3 *HZ , +}; + +static void nfs4_copy_dequeue_callback (struct nfs_server *dst_server, +struct nfs_server *src_server, +struct nfs4_copy_state *copy) +{ +spin_lock (&dst_server->nfs_client->cl_lock); +list_del_init (©->copies); +spin_unlock (&dst_server->nfs_client->cl_lock); +if (dst_server != src_server) { +spin_lock (&src_server->nfs_client->cl_lock); +list_del_init (©->src_copies); +spin_unlock (&src_server->nfs_client->cl_lock); + } +} + static int handle_async_copy(struct nfs42_copy_res *res, struct nfs_server *dst_server, struct nfs_server *src_server,
@@ -184,9 +203,10 @@ static int handle_async_copy(struct nfs42_copy_res *res, bool *restart) {struct nfs4_copy_state *copy, *tmp_copy =NULL , *iter; -int status =NFS4_OK ;struct nfs_open_context *dst_ctx =nfs_file_open_context (dst);struct nfs_open_context *src_ctx =nfs_file_open_context (src); +int status =NFS4_OK ; +u64 copied; copy =kzalloc (sizeof (struct nfs4_copy_state ),GFP_KERNEL );if (!copy)
@@ -224,15 +244,12 @@ static int handle_async_copy(struct nfs42_copy_res *res,spin_unlock (&src_server->nfs_client->cl_lock); } - status =wait_for_completion_interruptible (©->completion); -spin_lock (&dst_server->nfs_client->cl_lock); -list_del_init (©->copies); -spin_unlock (&dst_server->nfs_client->cl_lock); -if (dst_server != src_server) { -spin_lock (&src_server->nfs_client->cl_lock); -list_del_init (©->src_copies); -spin_unlock (&src_server->nfs_client->cl_lock); - } +wait: + status =wait_for_completion_interruptible_timeout (©->completion, +NFS42_COPY_TIMEOUT ); +if (!status) + goto timeout; +nfs4_copy_dequeue_callback (dst_server, src_server, copy);if (status == -ERESTARTSYS ) { goto out_cancel; }else if (copy->flags || copy->error ==NFS4ERR_PARTNER_NO_AUTH ) {
@@ -242,6 +259,7 @@ static int handle_async_copy(struct nfs42_copy_res *res, } out: res->write_res.count = copy->count; +/* Copy out the updated write verifier provided by CB_OFFLOAD. */ memcpy (&res->write_res.verifier, ©->verf,sizeof (copy->verf)); status = -copy->error;
@@ -253,6 +271,36 @@ static int handle_async_copy(struct nfs42_copy_res *res,if (!nfs42_files_from_same_server (src, dst))nfs42_do_offload_cancel_async (src, src_stateid); goto out_free; +timeout: + status =nfs42_proc_offload_status (dst, ©->stateid, &copied); +if (status == -EINPROGRESS ) + goto wait; +nfs4_copy_dequeue_callback (dst_server, src_server, copy); +switch (status) { +case 0 : +/* The server recognized the copy stateid, so it hasn't +* rebooted. Don't overwrite the verifier returned in the +* COPY result. */ + res->write_res.count = copied; + goto out_free; +case -EREMOTEIO : +/* COPY operation failed on the server. */ + status = -EOPNOTSUPP ; + res->write_res.count = copied; + goto out_free; +case -EBADF : +/* Server did not recognize the copy stateid. It has +* probably restarted and lost the plot. */ + res->write_res.count =0 ; + status = -EOPNOTSUPP ; +break ; +case -EOPNOTSUPP : +/* RFC 7862 REQUIREs server to support OFFLOAD_STATUS when +* it has signed up for an async COPY, so server is not +* spec-compliant. */ + res->write_res.count =0 ; + } + goto out_free; } static int process_copy_commit(struct file *dst, loff_t pos_dst,
@@ -643,7 +691,7 @@ _nfs42_proc_offload_status(struct nfs_server *server, struct file *file,* Other negative errnos indicate the client could not complete the * request. */ -static int __maybe_unused +static int nfs42_proc_offload_status (struct file *dst,nfs4_stateid *stateid,u64 *copied) {struct inode *inode =file_inode (dst);
Signature · 1 line
2.47.0
[email protected]
maintainerPATCH v2 7/7 NFS: Refactor trace_nfs4_offload_cancel
In reply to [email protected]From: Chuck Lever <[email protected]>
Add a trace_nfs4_offload_status trace point that looks just like trace_nfs4_offload_cancel. Promote that event to an event class to avoid duplicating code.
An alternative approach would be to expand trace_nfs4_offload_status to report more of the actual OFFLOAD_STATUS result.
--- fs/nfs/nfs42proc.c | 1 + fs/nfs/nfs4trace.h | 11 ++++++++++- 2 files changed, 11 insertions(+), 1 deletion(-)
fs/nfs/nfs42proc.c +1 -0
@@ -648,6 +648,7 @@ _nfs42_proc_offload_status(struct nfs_server *server, struct file *file, status =nfs4_call_sync (server->client, server, &msg, &data->args.osa_seq_args, &data->res.osr_seq_res,1 ); +trace_nfs4_offload_status (&data->args, status);switch (status) {case 0 :break ;
fs/nfs/nfs4trace.h +10 -1
@@ -2608,7 +2608,7 @@ TRACE_EVENT(nfs4_copy_notify, ) ); -TRACE_EVENT(nfs4_offload_cancel, +DECLARE_EVENT_CLASS(nfs4_offload_class, TP_PROTO( const struct nfs42_offload_status_args *args, int error
@@ -2640,6 +2640,15 @@ TRACE_EVENT(nfs4_offload_cancel, __entry->stateid_seq, __entry->stateid_hash ) ); +#define DEFINE_NFS4_OFFLOAD_EVENT (name) \ + DEFINE_EVENT(nfs4_offload_class, name, \ + TP_PROTO( \ + const struct nfs42_offload_status_args *args, \ + int error \ + ), \ + TP_ARGS(args, error)) +DEFINE_NFS4_OFFLOAD_EVENT (nfs4_offload_cancel); +DEFINE_NFS4_OFFLOAD_EVENT (nfs4_offload_status); DECLARE_EVENT_CLASS(nfs4_xattr_event, TP_PROTO(
Signature · 1 line
2.47.0
Quoted 29 lines · On Fri, 2024-12-20 at 10:42 -0500, [email protected] wrote:
From: Chuck Lever <[email protected]>
SCSI implementation experience has shown that an interrupt-only COPY offload implementation is not reliable. There are too many common scenarios where the client can miss the completion interrupt (in our case, this is a
CB_OFFLOADcallback).Therefore, a polling mechanism is needed. The NFSv4.2 protocol provides one in the form of the new
OFFLOAD_STATUSoperation. Linux NFSD implementsOFFLOAD_STATUSalready. This series adds a Linux NFS client implementation of theOFFLOAD_STATUSoperation that can query the state of a background COPY on the server.These patches are also available here:
https://git.kernel.org/pub/scm/linux/kernel/git/cel/linux.git/log/?h=fix-async-copy
Changes since v1:
-nfs42_proc_offload_status()now uses a synchronous RPCChuck Lever (7):
NFS:CB_OFFLOADcan returnNFS4ERR_DELAY
NFS: Fix typo inOFFLOAD_CANCELcomment
NFS: Rename structnfs4_offloadcancel_data
NFS: Implement NFSv4.2'sOFFLOAD_STATUSXDR
NFS: Implement NFSv4.2'sOFFLOAD_STATUSoperation
NFS: Use NFSv4.2'sOFFLOAD_STATUSoperation
NFS: Refactortrace_nfs4_offload_cancel
Quoted 11 lines
fs/nfs/callback_proc.c | 2 +- fs/nfs/nfs42proc.c | 188 ++++++++++++++++++++++++++++++++++---- fs/nfs/nfs42xdr.c | 88 +++++++++++++++++- fs/nfs/nfs4proc.c | 3 +- fs/nfs/nfs4trace.h | 11 ++- fs/nfs/nfs4xdr.c | 1 + include/linux/nfs4.h | 1 + include/linux/nfs_fs_sb.h | 1 + include/linux/nfs_xdr.h | 5 +- 9 files changed, 275 insertions(+), 25 deletions(-)
Nice work, Chuck. This all looks good to me.
Re: [PATCH v2 6/7] NFS: Use NFSv4.2's OFFLOAD_STATUS operation
In reply to [email protected]Quoted 33 lines · On Fri, Dec 20, 2024 at 10:46 AM <[email protected]> wrote:
From: Chuck Lever <[email protected]>
We've found that there are cases where a transport disconnection results in the loss of callback RPCs. NFS servers typically do not retransmit callback operations after a disconnect.
This can be a problem for the Linux NFS client's current implementation of asynchronous COPY, which waits indefinitely for a
CB_OFFLOADcallback. If a transport disconnect occurs while an async COPY is running, there's a good chance the client will never get the completingCB_OFFLOAD.Fix this by implementing the
OFFLOAD_STATUSoperation so that the Linux NFS client can probe the NFS server if it doesn't see aCB_OFFLOADin a reasonable amount of time.This patch implements a simplistic check. As future work, the client might also be able to detect whether there is no forward progress on the request asynchronous COPY operation, and CANCEL it.
Suggested-by Olga Kornievskaia <[email protected]>Signed-off-by Chuck Lever <[email protected]>--- fs/nfs/nfs42proc.c | 70 ++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 59 insertions(+), 11 deletions(-)fs/nfs/nfs42proc.c +0
-0
fs/nfs/nfs42proc.c int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len) · line 180
Quoted 6 lines
fs/nfs/nfs42proc.c +3
-0@@ -175,6 +175,25 @@ int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len)return err; } +/* Wait this long before checking progress on a COPY operation */ +enum { +NFS42_COPY_TIMEOUT =3 *HZ ,
I'm really not a fan of such a short time out. This make the OFFLOAD_STATUS a more likely operation rather than a less likely operation to occur during a copy. OFFLOAD_STATUS and CB_OFFLOAD being concurrent operations introduce races which we have to try to account for.
Quoted 62 lines
fs/nfs/nfs42proc.c +25
-10+}; + +static void nfs4_copy_dequeue_callback (struct nfs_server *dst_server, +struct nfs_server *src_server, +struct nfs4_copy_state *copy) +{ +spin_lock (&dst_server->nfs_client->cl_lock); +list_del_init (©->copies); +spin_unlock (&dst_server->nfs_client->cl_lock); +if (dst_server != src_server) { +spin_lock (&src_server->nfs_client->cl_lock); +list_del_init (©->src_copies); +spin_unlock (&src_server->nfs_client->cl_lock); + } +} + static int handle_async_copy(struct nfs42_copy_res *res, struct nfs_server *dst_server, struct nfs_server *src_server,@@ -184,9 +203,10 @@ static int handle_async_copy(struct nfs42_copy_res *res, bool *restart) {struct nfs4_copy_state *copy, *tmp_copy =NULL , *iter; -int status =NFS4_OK ;struct nfs_open_context *dst_ctx =nfs_file_open_context (dst);struct nfs_open_context *src_ctx =nfs_file_open_context (src); +int status =NFS4_OK ; +u64 copied; copy =kzalloc (sizeof (struct nfs4_copy_state ),GFP_KERNEL );if (!copy)@@ -224,15 +244,12 @@ static int handle_async_copy(struct nfs42_copy_res *res,spin_unlock (&src_server->nfs_client->cl_lock); } - status =wait_for_completion_interruptible (©->completion); -spin_lock (&dst_server->nfs_client->cl_lock); -list_del_init (©->copies); -spin_unlock (&dst_server->nfs_client->cl_lock); -if (dst_server != src_server) { -spin_lock (&src_server->nfs_client->cl_lock); -list_del_init (©->src_copies); -spin_unlock (&src_server->nfs_client->cl_lock); - } +wait: + status =wait_for_completion_interruptible_timeout (©->completion, +NFS42_COPY_TIMEOUT ); +if (!status) + goto timeout; +nfs4_copy_dequeue_callback (dst_server, src_server, copy);if (status == -ERESTARTSYS ) { goto out_cancel; }else if (copy->flags || copy->error ==NFS4ERR_PARTNER_NO_AUTH ) {@@ -242,6 +259,7 @@ static int handle_async_copy(struct nfs42_copy_res *res, } out: res->write_res.count = copy->count; +/* Copy out the updated write verifier provided by CB_OFFLOAD. */ memcpy (&res->write_res.verifier, ©->verf,sizeof (copy->verf)); status = -copy->error;
fs/nfs/nfs42proc.c static int handle_async_copy(struct nfs42_copy_res *res, · line 275
Quoted 5 lines
fs/nfs/nfs42proc.c +2
-0@@ -253,6 +271,36 @@ static int handle_async_copy(struct nfs42_copy_res *res,if (!nfs42_files_from_same_server (src, dst))nfs42_do_offload_cancel_async (src, src_stateid); goto out_free; +timeout: + status =nfs42_proc_offload_status (dst, ©->stateid, &copied);
Regardless of what OFFLOAD_STATUS returned we have to check whether or not the CB_OFFLOAD had arrived while we were waiting for the reply to the OFFLOAD_STATUS.
fs/nfs/nfs42proc.c static int handle_async_copy(struct nfs42_copy_res *res, · line 285
Quoted 10 lines
fs/nfs/nfs42proc.c +10
-0+if (status == -EINPROGRESS ) + goto wait; +nfs4_copy_dequeue_callback (dst_server, src_server, copy); +switch (status) { +case 0 : +/* The server recognized the copy stateid, so it hasn't +* rebooted. Don't overwrite the verifier returned in the +* COPY result. */ + res->write_res.count = copied; + goto out_free;
In case OFFLOAD_STATUS was successful and CB_OFFLOAD was received we should take the verifier from the CB_OFFLOAD otherwise we are sending the unneeede and expensive COMMIT because OFFLOAD_STATUS carries the completion and value of copy it does not carry the "how committed" value and thus we are forced to use async copy's "how committed" value.
fs/nfs/nfs42proc.c static int handle_async_copy(struct nfs42_copy_res *res, · line 296
Quoted 11 lines
fs/nfs/nfs42proc.c +11
-0+case -EREMOTEIO : +/* COPY operation failed on the server. */ + status = -EOPNOTSUPP ; + res->write_res.count = copied; + goto out_free; +case -EBADF : +/* Server did not recognize the copy stateid. It has +* probably restarted and lost the plot. */ + res->write_res.count =0 ; + status = -EOPNOTSUPP ; +break ;
This is the case of receiving a BAD_STATEID from OFFLOAD_STATUS and this would lead to copy falling back to read/write operation. IF we don't check the existence of CB_OFFLOAD reply, then OFFLOAD_STATUS can and will carry BAD_STATEID as the server is free to delete copy status after it get CB_OFFLOAD reply (which as i said we have to check). And If we did then we need take the result of the CB_OFFLOAD and not act on OFFLOAD_STATUS's error.
Quoted 23 lines
fs/nfs/nfs42proc.c +8
-2+case -EOPNOTSUPP : +/* RFC 7862 REQUIREs server to support OFFLOAD_STATUS when +* it has signed up for an async COPY, so server is not +* spec-compliant. */ + res->write_res.count =0 ; + } + goto out_free; } static int process_copy_commit(struct file *dst, loff_t pos_dst,@@ -643,7 +691,7 @@ _nfs42_proc_offload_status(struct nfs_server *server, struct file *file,* Other negative errnos indicate the client could not complete the * request. */ -static int __maybe_unused +static int nfs42_proc_offload_status (struct file *dst,nfs4_stateid *stateid,u64 *copied) {struct inode *inode =file_inode (dst); --2.47.0
Re: [PATCH v2 6/7] NFS: Use NFSv4.2's OFFLOAD_STATUS operation
Quoted 37 lines · Olga Kornievskaia
Quoted 36 lines · On Fri, Dec 20, 2024 at 10:46 AM <[email protected]> wrote:
From: Chuck Lever <[email protected]>
We've found that there are cases where a transport disconnection results in the loss of callback RPCs. NFS servers typically do not retransmit callback operations after a disconnect.
This can be a problem for the Linux NFS client's current implementation of asynchronous COPY, which waits indefinitely for a
CB_OFFLOADcallback. If a transport disconnect occurs while an async COPY is running, there's a good chance the client will never get the completingCB_OFFLOAD.Fix this by implementing the
OFFLOAD_STATUSoperation so that the Linux NFS client can probe the NFS server if it doesn't see aCB_OFFLOADin a reasonable amount of time.This patch implements a simplistic check. As future work, the client might also be able to detect whether there is no forward progress on the request asynchronous COPY operation, and CANCEL it.
Suggested-by Olga Kornievskaia <[email protected]>Signed-off-by Chuck Lever <[email protected]>--- fs/nfs/nfs42proc.c | 70 ++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 59 insertions(+), 11 deletions(-)fs/nfs/nfs42proc.c +0
-0@@ -175,6 +175,25 @@ int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len)return err; }
Quoted 5 lines
fs/nfs/nfs42proc.c int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len) · line 180
Quoted 3 lines
fs/nfs/nfs42proc.c +3
-0+/* Wait this long before checking progress on a COPY operation */ +enum { +NFS42_COPY_TIMEOUT =3 *HZ ,I'm really not a fan of such a short time out.
I'm not wedded to 3 seconds.
For the purpose of a proof-of-concept, it can be valuable to make such timeouts short to exacerbate races. But perhaps we are about to step out of the realm of proof-of-concept. Suggestions are welcome.
Quoted 5 lines
This make the
OFFLOAD_STATUSa more likely operation rather than a less likely operation to occur during a copy.OFFLOAD_STATUSandCB_OFFLOADbeing concurrent operations introduce races which we have to try to account for.
However, we have to close these windows whether the timeout is short or long. Making a timeout longer simply to avoid races seems like a band aid to me.
Back to choosing a value.
For a very long running copy, every 3 seconds might indeed be overkill. We could use the advertised lease timeout (or a fraction of it) instead.
Neil suggested an exponential backoff here. That feels to me like an optimization that can be made later if we find there is evidence in favor of doing so.
Quoted 61 lines
Quoted 61 lines
fs/nfs/nfs42proc.c +25
-10+}; + +static void nfs4_copy_dequeue_callback (struct nfs_server *dst_server, +struct nfs_server *src_server, +struct nfs4_copy_state *copy) +{ +spin_lock (&dst_server->nfs_client->cl_lock); +list_del_init (©->copies); +spin_unlock (&dst_server->nfs_client->cl_lock); +if (dst_server != src_server) { +spin_lock (&src_server->nfs_client->cl_lock); +list_del_init (©->src_copies); +spin_unlock (&src_server->nfs_client->cl_lock); + } +} + static int handle_async_copy(struct nfs42_copy_res *res, struct nfs_server *dst_server, struct nfs_server *src_server,@@ -184,9 +203,10 @@ static int handle_async_copy(struct nfs42_copy_res *res, bool *restart) {struct nfs4_copy_state *copy, *tmp_copy =NULL , *iter; -int status =NFS4_OK ;struct nfs_open_context *dst_ctx =nfs_file_open_context (dst);struct nfs_open_context *src_ctx =nfs_file_open_context (src); +int status =NFS4_OK ; +u64 copied; copy =kzalloc (sizeof (struct nfs4_copy_state ),GFP_KERNEL );if (!copy)@@ -224,15 +244,12 @@ static int handle_async_copy(struct nfs42_copy_res *res,spin_unlock (&src_server->nfs_client->cl_lock); } - status =wait_for_completion_interruptible (©->completion); -spin_lock (&dst_server->nfs_client->cl_lock); -list_del_init (©->copies); -spin_unlock (&dst_server->nfs_client->cl_lock); -if (dst_server != src_server) { -spin_lock (&src_server->nfs_client->cl_lock); -list_del_init (©->src_copies); -spin_unlock (&src_server->nfs_client->cl_lock); - } +wait: + status =wait_for_completion_interruptible_timeout (©->completion, +NFS42_COPY_TIMEOUT ); +if (!status) + goto timeout; +nfs4_copy_dequeue_callback (dst_server, src_server, copy);if (status == -ERESTARTSYS ) { goto out_cancel; }else if (copy->flags || copy->error ==NFS4ERR_PARTNER_NO_AUTH ) {@@ -242,6 +259,7 @@ static int handle_async_copy(struct nfs42_copy_res *res, } out: res->write_res.count = copy->count; +/* Copy out the updated write verifier provided by CB_OFFLOAD. */ memcpy (&res->write_res.verifier, ©->verf,sizeof (copy->verf)); status = -copy->error;
Quoted 10 lines
fs/nfs/nfs42proc.c static int handle_async_copy(struct nfs42_copy_res *res, · line 275
Quoted 6 lines
fs/nfs/nfs42proc.c +2
-0@@ -253,6 +271,36 @@ static int handle_async_copy(struct nfs42_copy_res *res,if (!nfs42_files_from_same_server (src, dst))nfs42_do_offload_cancel_async (src, src_stateid); goto out_free; +timeout: + status =nfs42_proc_offload_status (dst, ©->stateid, &copied);Regardless of what
OFFLOAD_STATUSreturned we have to check whether or not theCB_OFFLOADhad arrived while we were waiting for the reply to theOFFLOAD_STATUS.
I think that could be done. The only reason to add that extra complexity is because CB_OFFLOAD carries a verifier and stable_how, and OFFLOAD_STATUS does not.
But it seems to me this race scenario should be pretty infrequent. And I'm not yet convinced that the COPY will misbehave if the client ignores a racing CB_OFFLOAD request. It might be a little slower, is all.
For me, the Minimum Viable Product for this series is that COPY should always be able to make forward progress. Are we at that point yet? Anything else is an optimization that should be made based on whether there is evidence that shows there is a substantial application-visible improvement.
Quoted 11 lines
fs/nfs/nfs42proc.c static int handle_async_copy(struct nfs42_copy_res *res, · line 285
Quoted 10 lines
fs/nfs/nfs42proc.c +10
-0+if (status == -EINPROGRESS ) + goto wait; +nfs4_copy_dequeue_callback (dst_server, src_server, copy); +switch (status) { +case 0 : +/* The server recognized the copy stateid, so it hasn't +* rebooted. Don't overwrite the verifier returned in the +* COPY result. */ + res->write_res.count = copied; + goto out_free;
Quoted 6 lines
In case
OFFLOAD_STATUSwas successful andCB_OFFLOADwas received we should take the verifier from theCB_OFFLOADotherwise we are sending the unneeede and expensive COMMIT becauseOFFLOAD_STATUScarries the completion and value of copy it does not carry the "how committed" value and thus we are forced to use async copy's "how committed" value.
I quibble with the description of COMMIT as "expensive".
- If the COPY was UNSTABLE, then a COMMIT will have to be done whether the client checks for an intervening CB_OFFLOAD or not.
- If the COPY was FILE_SYNC, then the COMMIT is a no-op, and the only expense here is an extra network round trip.
- This is not a performance path. If the client missed the CB_OFFLOAD in the first place, then it's already waited a few seconds longer than the server took to do the COPY.
So my feeling is that avoiding the COMMIT is an optimization. It isn't required to make COPY or OFFLOAD_STATUS work correctly AFAICT.
Quoted 12 lines
fs/nfs/nfs42proc.c static int handle_async_copy(struct nfs42_copy_res *res, · line 296
Quoted 11 lines
fs/nfs/nfs42proc.c +11
-0+case -EREMOTEIO : +/* COPY operation failed on the server. */ + status = -EOPNOTSUPP ; + res->write_res.count = copied; + goto out_free; +case -EBADF : +/* Server did not recognize the copy stateid. It has +* probably restarted and lost the plot. */ + res->write_res.count =0 ; + status = -EOPNOTSUPP ; +break ;
Quoted 5 lines
This is the case of receiving a
BAD_STATEIDfromOFFLOAD_STATUSand this would lead to copy falling back to read/write operation. IF we don't check the existence ofCB_OFFLOADreply, thenOFFLOAD_STATUScan and will carryBAD_STATEIDas the server is free to delete copy status after it getCB_OFFLOADreply (which as i said we have to check).
I don't agree that the client /has/ to check for it. I think it's an optimization for an infrequent race. As the meme goes: "Change my mind". ;-)
But back to changing the logic in this arm: what is the problem with falling back to a read/write copy in such cases? This seems to me like an infrequent case where we should prioritize recovering reliably.
Quoted 3 lines
And
If we did then we need take the result of theCB_OFFLOADand not act on OFFLOAD_STATUS's error.
Checking for CB_OFFLOAD after an OFFLOAD_STATUS operation seems racy to me. There is still a gap where the client has done OFFLOAD_STATUS, checks for the CB_OFFLOAD and doesn't see one, and then the CB_OFFLOAD actually shows up. So this does not eliminate a race, it merely narrows the window.
Quoted 23 lines
Quoted 23 lines
fs/nfs/nfs42proc.c +8
-2+case -EOPNOTSUPP : +/* RFC 7862 REQUIREs server to support OFFLOAD_STATUS when +* it has signed up for an async COPY, so server is not +* spec-compliant. */ + res->write_res.count =0 ; + } + goto out_free; } static int process_copy_commit(struct file *dst, loff_t pos_dst,@@ -643,7 +691,7 @@ _nfs42_proc_offload_status(struct nfs_server *server, struct file *file,* Other negative errnos indicate the client could not complete the * request. */ -static int __maybe_unused +static int nfs42_proc_offload_status (struct file *dst,nfs4_stateid *stateid,u64 *copied) {struct inode *inode =file_inode (dst); --2.47.0
Signature · 1 line
Chuck Lever