mount options not propagating to NFSACL and NSM RPC clients
Subsystems (3)
Hi,
On Linux v6.6-14500-g1c41041124bd, I added a sysfs file for debugging /sys/kernel/debug/sunrpc/rpc_clnt/*/info, and noticed that when passing the following mount options: soft,timeo=50,retrans=16,vers=3, NFSACL and NSM seem to take the defaults from somewhere else (xprt). Specifically, locking operation behave as if in a hard mount with these mount options.
Is it intentional?
cl_prog: 100000
cl_softrtry: 1
cl_timeout: to_initval=10000, to_maxval=10000, to_increment=0, to_retries=2, to_exponential=0
cl_prog: 100003
cl_softrtry: 1
cl_timeout: to_initval=5000, to_maxval=85000, to_increment=5000, to_retries=16, to_exponential=0
cl_prog: 100000
cl_softrtry: 1
cl_timeout: to_initval=10000, to_maxval=10000, to_increment=0, to_retries=2, to_exponential=0
cl_prog: 100021
cl_softrtry: 0
cl_timeout: to_initval=10000, to_maxval=60000, to_increment=10000, to_retries=5, to_exponential=0
cl_prog: 100003
cl_softrtry: 1
cl_timeout: to_initval=5000, to_maxval=85000, to_increment=5000, to_retries=16, to_exponential=0
cl_prog: 100227
cl_softrtry: 1
cl_timeout: to_initval=60000, to_maxval=60000, to_increment=0, to_retries=2, to_exponential=0Also, here's the patch I used to expose this information:
-- >8 --
Subject: [PATCH] sunrpc: add per-client 'info' node in debugfs
--- net/sunrpc/debugfs.c | 75 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 75 insertions(+)
net/sunrpc/debugfs.c +75 -0
@@ -142,6 +142,77 @@ static int do_xprt_debugfs(struct rpc_clnt *clnt, struct rpc_xprt *xprt, void *nreturn 0 ; } +static int +clnt_info_show (struct seq_file *seq,void *v) +{ +struct rpc_clnt *clnt = seq->private; +const struct rpc_timeout *timeout = clnt->cl_timeout; + +seq_printf (seq,"cl_count: %d\n" ,refcount_read (&clnt->cl_count)); +seq_printf (seq,"cl_pid: %d\n" ,atomic_read (&clnt->cl_pid)); +seq_printf (seq,"cl_prog: %d\n" , clnt->cl_prog); +seq_printf (seq,"cl_vers: %d\n" , clnt->cl_vers); +seq_printf (seq,"cl_maxproc: %d\n" , clnt->cl_maxproc); + +seq_printf (seq,"cl_softrtry: %d\n" , clnt->cl_softrtry); +seq_printf (seq,"cl_softerr: %d\n" , clnt->cl_softerr); +seq_printf (seq,"cl_discrtry: %d\n" , clnt->cl_discrtry); +seq_printf (seq,"cl_noretranstimeo: %d\n" , clnt->cl_noretranstimeo); +seq_printf (seq,"cl_autobind: %d\n" , clnt->cl_autobind); +seq_printf (seq,"cl_chatty: %d\n" , clnt->cl_chatty); +seq_printf (seq,"cl_shutdown: %d\n" , clnt->cl_shutdown); + +seq_printf (seq,"cl_xprtsec: %d\n" , clnt->cl_xprtsec.policy); +seq_printf (seq,"cl_rtt: timeo=%ld\n" , clnt->cl_rtt->timeo); +seq_printf (seq,"cl_timeout: " +"to_initval=%ld, to_maxval=%ld, to_increment=%ld, " +"to_retries=%d, to_exponential=%d\n" , + timeout->to_initval, timeout->to_maxval, + timeout->to_increment, + timeout->to_retries, timeout->to_exponential); + +seq_printf (seq,"cl_max_connect: %d\n" , clnt->cl_max_connect); +seq_printf (seq,"cl_nodename: " ); +seq_write (seq, clnt->cl_nodename, clnt->cl_nodelen); +seq_printf (seq,"\n" ); + +return 0 ; +} + +static int +clnt_info_open (struct inode *inode,struct file *filp) +{ +int ret; +struct rpc_clnt *clnt = inode->i_private; + + ret =single_open (filp, clnt_info_show, clnt); + +if (!ret && !refcount_inc_not_zero (&clnt->cl_count)) { +single_release (inode, filp); +return -EINVAL ; + } + +return ret; +} + +static int +clnt_info_release (struct inode *inode,struct file *filp) +{ +struct seq_file *seq = filp->private_data; +struct rpc_clnt *clnt = seq->private; + +rpc_release_client (clnt); +return seq_release (inode, filp); +} + +static const struct file_operations info_fops = { + .owner =THIS_MODULE , + .open = clnt_info_open, + .read = seq_read, + .llseek = seq_lseek, + .release = clnt_info_release, +}; +void rpc_clnt_debugfs_register (struct rpc_clnt *clnt) {
@@ -160,6 +231,10 @@ rpc_clnt_debugfs_register(struct rpc_clnt *clnt)debugfs_create_file ("tasks" ,S_IFREG |0400 , clnt->cl_debugfs, clnt, &tasks_fops); +/* make info file */ +debugfs_create_file ("info" ,S_IFREG |0400 , clnt->cl_debugfs, clnt, + &info_fops); +rpc_clnt_iterate_for_each_xprt (clnt, do_xprt_debugfs, &xprtnum); }
Signature · 5 lines
2.39.3 -- Dan Aloni
Hi Dan,
Quoted 10 lines · Dan Aloni
Hi,
On Linux v6.6-14500-g1c41041124bd, I added a sysfs file for debugging
/sys/kernel/debug/sunrpc/rpc_clnt/*/info, and noticed that when passing the following mount options:soft,timeo=50,retrans=16,vers=3, NFSACL and NSM seem to take the defaults from somewhere else (xprt). Specifically, locking operation behave as if in a hard mount with these mount options.Is it intentional?
Yes, it usually is intentional. The various rpc clients that make parts of NFS work don't all inherit the mount flags due to reasons about how the system should behave as a whole. I think that you can find usually find the reasoning the git logs around "struct rpc_create_args".
Are you getting a system hung up in a lock operation?
Ben
Quoted 13 lines · Benjamin Coddington
Hi Dan,
Quoted 9 lines · Dan Aloni
Hi,
On Linux v6.6-14500-g1c41041124bd, I added a sysfs file for debugging
/sys/kernel/debug/sunrpc/rpc_clnt/*/info, and noticed that when passing the following mount options:soft,timeo=50,retrans=16,vers=3, NFSACL and NSM seem to take the defaults from somewhere else (xprt). Specifically, locking operation behave as if in a hard mount with these mount options.
Quoted 8 lines
Quoted 1 line
Is it intentional?
Yes, it usually is intentional. The various rpc clients that make parts of NFS work don't all inherit the mount flags due to reasons about how the system should behave as a whole. I think that you can find usually find the reasoning the git logs around "struct
rpc_create_args".Are you getting a system hung up in a lock operation?
Actually my concern is the NFSACL prog. With cl_softrtrt == 1 and to_initval == to_maxval, does it mean retires will not happen regardless of to_retries and to_increment?
I encountered a situation where the NFSACL program did not retry but could have had, whereas NFS3 did successfully. Not sure regarding NSM, but it seems to me that it would make sense at least for NFSACL to behave the same as NFS3.
Signature · 1 line
Dan Aloni
Quoted 21 lines · Dan Aloni
Quoted 20 lines · Benjamin Coddington
Hi Dan,
Quoted 10 lines · Dan Aloni
Hi,
On Linux v6.6-14500-g1c41041124bd, I added a sysfs file for debugging
/sys/kernel/debug/sunrpc/rpc_clnt/*/info, and noticed that when passing the following mount options:soft,timeo=50,retrans=16,vers=3, NFSACL and NSM seem to take the defaults from somewhere else (xprt). Specifically, locking operation behave as if in a hard mount with these mount options.Is it intentional?
Yes, it usually is intentional. The various rpc clients that make parts of NFS work don't all inherit the mount flags due to reasons about how the system should behave as a whole. I think that you can find usually find the reasoning the git logs around "struct
rpc_create_args".
Quoted 5 lines
Quoted 1 line
Are you getting a system hung up in a lock operation?
Actually my concern is the NFSACL prog. With
cl_softrtrt == 1andto_initval == to_maxval, does it mean retires will not happen regardless ofto_retriesandto_increment?
Possibly? I'm not exactly certain of what should happen in that case.
Quoted 4 lines
I encountered a situation where the NFSACL program did not retry but could have had, whereas NFS3 did successfully. Not sure regarding NSM, but it seems to me that it would make sense at least for NFSACL to behave the same as NFS3.
I agree, but I could be missing something -- maybe its a bug. There's the sunrpc:rpc_timeout_status tracepoint that might be helpful. If you turn that up can you see rpc_check_timeout() getting called from call_transmit_status()?
Sorry, I know this isn't a lot of help so far.
Ben
Quoted 11 lines · Benjamin Coddington
Quoted 3 lines
Actually my concern is the NFSACL prog. With
cl_softrtrt == 1andto_initval == to_maxval, does it mean retires will not happen regardless ofto_retriesandto_increment?Possibly? I'm not exactly certain of what should happen in that case.
Quoted 4 lines
I encountered a situation where the NFSACL program did not retry but could have had, whereas NFS3 did successfully. Not sure regarding NSM, but it seems to me that it would make sense at least for NFSACL to behave the same as NFS3.
Quoted 4 lines
I agree, but I could be missing something -- maybe its a bug. There's the sunrpc:rpc_timeout_status tracepoint that might be helpful. If you turn that up can you see
rpc_check_timeout()getting called fromcall_transmit_status()?
Sorry, took awhile to get a test working while busy on other stuff.
So it looks really like a bug, here are the details.
Server: nfsd with extra fault injection code that calls svc_drop() only once on a single NFS GETACL request.
Client: Linux v6.8, NFS mount with soft,timeo=50,retrans=16,vers=3.
I trace client execution with the following:
sudo perf trace -e sunrpc:rpc_task_timeout -e sunrpc:xprt_retransmit
A simple ls -l gets stuck and shows an IO failure:
[root@client export]# ls -l
ls: file: Input/output error
total 0
-rw-r--r-- 1 root root 0 Apr 10 10:02 file
I get a single event out of the tracing above:
``` kthreadd/7926 sunrpc:rpc_task_timeout(task_id: 203, client_id: 6, xprt_id: 3, action: 0xffffffffc0accc60, runstate: 22, flags: 35456) ```
So looks like the request is not being retransmitted. Just to be sure, if I cause the nfsd to drop the regular NFS3 prog I/Os like ACCESS and LOOKUP, I only get the expected 5 seconds delay following a successful retry.
Seems we only have an issue with the NFS3ACL prog.
Signature · 1 line
Dan Aloni
Quoted 42 lines · Dan Aloni
Quoted 11 lines · Benjamin Coddington
Quoted 3 lines
Actually my concern is the NFSACL prog. With
cl_softrtrt == 1andto_initval == to_maxval, does it mean retires will not happen regardless ofto_retriesandto_increment?Possibly? I'm not exactly certain of what should happen in that case.
Quoted 4 lines
I encountered a situation where the NFSACL program did not retry but could have had, whereas NFS3 did successfully. Not sure regarding NSM, but it seems to me that it would make sense at least for NFSACL to behave the same as NFS3.
Quoted 4 lines
I agree, but I could be missing something -- maybe its a bug. There's the sunrpc:rpc_timeout_status tracepoint that might be helpful. If you turn that up can you see
rpc_check_timeout()getting called fromcall_transmit_status()?Sorry, took awhile to get a test working while busy on other stuff.
So it looks really like a bug, here are the details.
Server: nfsd with extra fault injection code that calls
svc_drop()only once on a single NFS GETACL request.
Client: Linux v6.8, NFS mount withsoft,timeo=50,retrans=16,vers=3.I trace client execution with the following:
sudo perf trace -e sunrpc:rpc_task_timeout -e sunrpc:xprt_retransmit
A simple
ls -lgets stuck and shows an IO failure:[root@client export]# ls -l
ls: file: Input/output error
total 0
-rw-r--r-- 1 root root 0 Apr 10 10:02 fileI get a single event out of the tracing above:
``` kthreadd/7926 sunrpc:rpc_task_timeout(task_id: 203, client_id: 6, xprt_id: 3, action: 0xffffffffc0accc60, runstate: 22, flags: 35456) ```
Quoted 6 lines
So looks like the request is not being retransmitted. Just to be sure, if I cause the nfsd to drop the regular NFS3 prog I/Os like ACCESS and LOOKUP, I only get the expected 5 seconds delay following a successful retry.
Seems we only have an issue with the NFS3ACL prog.
It looks like the client_acl program gets created with rpc_bind_new_program() which doesn't setup the timeouts/retry strategy, and there's nothing after the setup to do it either.
I think the problem has existed since 331702337f2b2.. I think this should fix it up, would you like to test it?
Ben
---
From 54a35f530d78a8042dc6fb8ff522f5a4a33fa50b Mon Sep 17 00:00:00 2001
Message-ID: <54a35f530d78a8042dc6fb8ff522f5a4a33fa50b.1712848680.git.bcodding@redhat.com>
From: Benjamin Coddington <[email protected]>
Date: Thu, 11 Apr 2024 11:03:06 -0400
Subject: [PATCH] NFS: Set v3 ACL client default timeout values
It appears the client_acl rpc_clnt doesn't take any timeout values, so retries for the ACL procedures are never attempted. Set them based on the mount's default timeout values.
--- fs/nfs/nfs3client.c | 1 + 1 file changed, 1 insertion(+)
fs/nfs/nfs3client.c +1 -0
@@ -33,6 +33,7 @@ static void nfs_init_server_aclclient(struct nfs_server *server)if (IS_ERR (server->client_acl)) goto out_noacl; + server->client_acl->cl_timeout = &server->client->cl_timeout_default;nfs_sysfs_link_rpc_client (server, server->client_acl,NULL );/* No errors! Assume that Sun nfsacls are supported */
Signature · 1 line
2.44.0
Quoted 5 lines · Benjamin Coddington
Quoted 5 lines
So looks like the request is not being retransmitted. Just to be sure, if I cause the nfsd to drop the regular NFS3 prog I/Os like ACCESS and LOOKUP, I only get the expected 5 seconds delay following a successful retry.
Quoted 8 lines
Quoted 1 line
Seems we only have an issue with the NFS3ACL prog.
It looks like the
client_aclprogram gets created withrpc_bind_new_program()which doesn't setup the timeouts/retry strategy, and there's nothing after the setup to do it either.I think the problem has existed since
331702337f2b2.. I think this should fix it up, would you like to test it?
Please allow me to propose a different change, which I already tested.
Looks to me that the 2012 change that I mentioned below is actually when the problem started happening, when the kmemdup call was removed, so since then we are simply just missing a field copy, and we are taking the timeout from transport-based default timeout globals instead.
Also, I have a hunch that we need to do something different, because of the following code in rpc_new_client:
clnt->cl_rtt = &clnt->cl_rtt_default;
rpc_init_rtt(&clnt->cl_rtt_default, clnt->cl_timeout->to_initval);Only setting clnt->cl_timeout _after_ client clone does not seem to be right if there are cl_rtt_default calculations that depend on it. So may as well need to call rpc_init_rtt too.
--
From 55737f82a9bb3e490836d10491995c8082ebcf11 Mon Sep 17 00:00:00 2001
From: Dan Aloni <[email protected]>
Date: Thu, 11 Apr 2024 18:30:56 +0300
Subject: [PATCH] sunrpc: fix NFSACL RPC retry on soft mount
It used to be quite awhile ago since 1b63a75180c6 ('SUNRPC: Refactor rpc_clone_client()'), in 2012, that cl_timeout was copied in so that all mount parameters propagate to NFSACL clients. However since that change, if mount options as follows are given:
soft,timeo=50,retrans=16,vers=3
The resultant NFSACL client receives:
cl_softrtry: 1
cl_timeout: to_initval=60000, to_maxval=60000, to_increment=0, to_retries=2, to_exponential=0These values lead to NFSACL operations not being retried under the condition of transient network outages with soft mount. Instead, getacl call fails after 60 seconds with EIO.
The simple fix is to copy cl_timeout and make sure cl_rtt_default is initialized from it.
--- net/sunrpc/clnt.c | 3 +++ 1 file changed, 3 insertions(+)
net/sunrpc/clnt.c +3 -0
@@ -669,6 +669,9 @@ static struct rpc_clnt *__rpc_clone_client(struct rpc_create_args *args, new->cl_chatty = clnt->cl_chatty; new->cl_principal = clnt->cl_principal; new->cl_max_connect = clnt->cl_max_connect; + new->cl_timeout = clnt->cl_timeout; +rpc_init_rtt (&clnt->cl_rtt_default, clnt->cl_timeout->to_initval); +return new; out_err:
Signature · 5 lines
2.39.3 -- Dan Aloni
Quoted 51 lines · Dan Aloni
Quoted 5 lines · Benjamin Coddington
Quoted 5 lines
So looks like the request is not being retransmitted. Just to be sure, if I cause the nfsd to drop the regular NFS3 prog I/Os like ACCESS and LOOKUP, I only get the expected 5 seconds delay following a successful retry.
Quoted 8 lines
Quoted 1 line
Seems we only have an issue with the NFS3ACL prog.
It looks like the
client_aclprogram gets created withrpc_bind_new_program()which doesn't setup the timeouts/retry strategy, and there's nothing after the setup to do it either.I think the problem has existed since
331702337f2b2.. I think this should fix it up, would you like to test it?Please allow me to propose a different change, which I already tested.
Looks to me that the 2012 change that I mentioned below is actually when the problem started happening, when the
kmemdupcall was removed, so since then we are simply just missing a field copy, and we are taking the timeout from transport-based default timeout globals instead.Also, I have a hunch that we need to do something different, because of the following code in
rpc_new_client:clnt->cl_rtt = &clnt->cl_rtt_default; rpc_init_rtt(&clnt->cl_rtt_default, clnt->cl_timeout->to_initval);Only setting
clnt->cl_timeout_after_ client clone does not seem to be right if there arecl_rtt_defaultcalculations that depend on it. So may as well need to callrpc_init_rtttoo.--
From
55737f82a9bb3e490836d10491995c8082ebcf11Mon Sep 17 00:00:00 2001
From: Dan Aloni <[email protected]>
Date: Thu, 11 Apr 2024 18:30:56 +0300
Subject: [PATCH] sunrpc: fix NFSACL RPC retry on soft mountIt used to be quite awhile ago since
1b63a75180c6('SUNRPC: Refactorrpc_clone_client()'), in 2012, thatcl_timeoutwas copied in so that all mount parameters propagate to NFSACL clients. However since that change, if mount options as follows are given:soft,timeo=50,retrans=16,vers=3The resultant NFSACL client receives:
cl_softrtry: 1 cl_timeout: to_initval=60000, to_maxval=60000, to_increment=0, to_retries=2, to_exponential=0
Quoted 6 lines
These values lead to NFSACL operations not being retried under the condition of transient network outages with soft mount. Instead, getacl call fails after 60 seconds with EIO.
The simple fix is to copy
cl_timeoutand make surecl_rtt_defaultis initialized from it.
Ah, good catch, and I think this means that the normal IO client (server->client) hasn't been properly setting up its RTT estimator as well?
Quoted 14 lines
Cc Chuck Lever <[email protected]>Cc Benjamin Coddington <[email protected]>Fixes 1b63a75180c6 ('SUNRPC: Refactor rpc_clone_client()')Signed-off-by Dan Aloni <[email protected]>--- net/sunrpc/clnt.c | 3 +++ 1 file changed, 3 insertions(+)net/sunrpc/clnt.c +0
-0
net/sunrpc/clnt.c static struct rpc_clnt *__rpc_clone_client(struct rpc_create_args *args, · line 674
Quoted 6 lines
net/sunrpc/clnt.c +3
-0@@ -669,6 +669,9 @@ static struct rpc_clnt *__rpc_clone_client(struct rpc_create_args *args, new->cl_chatty = clnt->cl_chatty; new->cl_principal = clnt->cl_principal; new->cl_max_connect = clnt->cl_max_connect; + new->cl_timeout = clnt->cl_timeout; +rpc_init_rtt (&clnt->cl_rtt_default, clnt->cl_timeout->to_initval); +
So we'll end up doing rpc_init_rtt twice, first in rpc_new_client then here. Maybe we should be passing cl_timeout on .timeout from the parent in the rpc_clone_client()'s rpc_create_args?
.. anyway - I'll stop "helping", you've clearly got this in hand. :)
Ben
Quoted 90 lines · Dan Aloni
Quoted 10 lines · Benjamin Coddington
Quoted 9 lines
So looks like the request is not being retransmitted. Just to be sure,
if I cause the nfsd to drop the regular NFS3 prog I/Os like ACCESS and
LOOKUP, I only get the expected 5 seconds delay following a successful
retry.Seems we only have an issue with the NFS3ACL prog.
Quoted 8 lines
It looks like the
client_aclprogram gets created withrpc_bind_new_program()which doesn't setup the timeouts/retry strategy, and
there's nothing after the setup to do it either.I think the problem has existed since 331702337f2b2.. I think this should
fix it up, would you like to test it?Please allow me to propose a different change, which I already tested.
Looks to me that the 2012 change that I mentioned below is actually when
the problem started happening, when thekmemdupcall was removed, so
since then we are simply just missing a field copy, and we are taking the timeout from transport-based default timeout globals instead.Also, I have a hunch that we need to do something different, because of
the following code inrpc_new_client:
clnt->cl_rtt=&clnt->cl_rtt_default; rpc_init_rtt(&clnt->cl_rtt_default, clnt->cl_timeout-Quoted 1 line
to_initval);Only setting
clnt->cl_timeout_after_ client clone does not seem to be
right if there arecl_rtt_defaultcalculations that depend on it. So
may as well need to callrpc_init_rtttoo.--
From
55737f82a9bb3e490836d10491995c8082ebcf11Mon Sep 17 00:00:00 2001
From: Dan Aloni <[email protected]>
Date: Thu, 11 Apr 2024 18:30:56 +0300
Subject: [PATCH] sunrpc: fix NFSACL RPC retry on soft mountIt used to be quite awhile ago since
1b63a75180c6('SUNRPC: Refactorrpc_clone_client()'), in 2012, thatcl_timeoutwas copied in so that
all mount parameters propagate to NFSACL clients. However since that change, if mount options as follows are given:soft,timeo=50,retrans=16,vers=3
The resultant NFSACL client receives:
cl_softrtry: 1
cl_timeout: to_initval=60000, to_maxval=60000, to_increment=0, to_retries=2, to_exponential=0These values lead to NFSACL operations not being retried under the condition of transient network outages with soft mount. Instead, getacl
call fails after 60 seconds with EIO.The simple fix is to copy
cl_timeoutand make surecl_rtt_defaultis initialized from it.Cc: Chuck Lever <[email protected]>
Cc: Benjamin Coddington <[email protected]>
Link:
https://lore.kernel.org/all/[email protected]/T/Fixes 1b63a75180c6 ('SUNRPC: Refactor rpc_clone_client()')Signed-off-by Dan Aloni <[email protected]>---
net/sunrpc/clnt.c | 3 +++1 file changed, 3 insertions(+)net/sunrpc/clnt.c +0
-0
net/sunrpc/clnt.c static struct rpc_clnt *__rpc_clone_client(struct · line 677
Quoted 9 lines
net/sunrpc/clnt.c +3
-0@@ -669,6 +669,9 @@ static struct rpc_clnt *__rpc_clone_client(structrpc_create_args *args, new->cl_chatty = clnt->cl_chatty; new->cl_principal = clnt->cl_principal; new->cl_max_connect = clnt->cl_max_connect; + new->cl_timeout = clnt->cl_timeout; +rpc_init_rtt (&clnt->cl_rtt_default, clnt->cl_timeout- >to_initval); + return new;
That ends up clobbering any timeout value that gets set in the rpc_create_args, and open codes something that is already being done in the call to rpc_new_client().
IOW: I'd much rather see us default the value of args->timeout to that of clnt->cl_timeout.
Signature · 3 lines
Trond Myklebust Linux NFS client maintainer, Hammerspace [email protected]
Quoted 16 lines · Trond Myklebust
Quoted 6 lines
net/sunrpc/clnt.c +0
-0net/sunrpc/clnt.c static struct rpc_clnt *__rpc_clone_client(struct · line 677
Quoted 9 lines
net/sunrpc/clnt.c +3
-0@@ -669,6 +669,9 @@ static struct rpc_clnt *__rpc_clone_client(structrpc_create_args *args, new->cl_chatty = clnt->cl_chatty; new->cl_principal = clnt->cl_principal; new->cl_max_connect = clnt->cl_max_connect; + new->cl_timeout = clnt->cl_timeout; +rpc_init_rtt (&clnt->cl_rtt_default, clnt->cl_timeout- >to_initval); + return new;
Quoted 6 lines
That ends up clobbering any timeout value that gets set in the
rpc_create_args, and open codes something that is already being done in the call torpc_new_client().IOW: I'd much rather see us default the value of
args->timeoutto that ofclnt->cl_timeout.
Makes sense. I confirm the following fix works for NFSACL and preserves the current hard mount behavior for NSM. Other callers to rpc_bind_new_program are unaffected by this.
However I am not sure if restoring this behavior is desired for the other call sites of __rpc_clone_client, which seems to be NFSv4.x users, so I left that part out. Should that be done in separate patches?
From 8715ef1d574970176e9ac87a7e826ad74f8b910d Mon Sep 17 00:00:00 2001
From: Dan Aloni <[email protected]>
Date: Thu, 11 Apr 2024 18:30:56 +0300
Subject: [PATCH] sunrpc: fix NFSACL RPC retry on soft mount
It used to be quite awhile ago since 1b63a75180c6 ('SUNRPC: Refactor rpc_clone_client()'), in 2012, that cl_timeout was copied in so that all mount parameters propagate to NFSACL clients. However since that change, if mount options as follows are given:
soft,timeo=50,retrans=16,vers=3
The resultant NFSACL client receives:
cl_softrtry: 1
cl_timeout: to_initval=60000, to_maxval=60000, to_increment=0, to_retries=2, to_exponential=0These values lead to NFSACL operations not being retried under the condition of transient network outages with soft mount. Instead, getacl call fails after 60 seconds with EIO.
The simple fix is to pass the existing client's cl_timeout as the new client timeout.
--- net/sunrpc/clnt.c | 1 + 1 file changed, 1 insertion(+)
net/sunrpc/clnt.c +1 -0
@@ -1068,6 +1068,7 @@ struct rpc_clnt *rpc_bind_new_program(struct rpc_clnt *old, .version = vers, .authflavor = old->cl_auth->au_flavor, .cred = old->cl_cred, + .timeout = old->cl_timeout, };struct rpc_clnt *clnt;int err;
Signature · 6 lines
2.39.3 -- Dan Aloni