Currently the source's export is mounted and unmounted on every
inter-server copy operation. This patch is an enhancement to delay
the unmount of the source export for a certain period of time to
eliminate the mount and unmount overhead on subsequent copy operations.
After a copy operation completes, a work entry is added to the
delayed unmount list with an expiration time. This list is serviced
by the laundromat thread to unmount the export of the expired entries.
Each time the export is being used again, its expiration time is
extended and the entry is re-inserted to the tail of the list.
The unmount task and the mount operation of the copy request are
synced to make sure the export is not unmounted while it's being
used.
Signed-off-by: Dai Ngo <dai.ngo@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
In order to ensure that knfsd threads don't linger once the nfsd
pseudofs is unmounted (e.g. when the container is killed) we let
nfsd_umount() shut down those threads and wait for them to exit.
This also should ensure that we don't need to do a kernel mount of
the pseudofs, since the thread lifetime is now limited by the
lifetime of the filesystem.
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
nfsd stats counters can be updated by concurrent nfsd threads without any
protection.
Convert some nfsd_stats and nfsd_net struct members to use percpu counters.
The longest_chain* members of struct nfsd_net remain unprotected.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Drop the repeated word "the" in a comment.
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: linux-nfs@vger.kernel.org
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Currently, nfsd4_encode_exchange_id() encodes the utsname nodename
string in the server_scope field. In a multi-host container
environemnt, if an nfsd container is restarted on a different host than
it was originally running on, clients will see a server_scope mismatch
and will not attempt to reclaim opens.
Instead, set the server_scope while we're in a process context during
service startup, so we get the utsname nodename of the current process
and store that in nfsd_net.
Signed-off-by: Scott Mayhew <smayhew@redhat.com>
[bfields: fix up major_id too]
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
A couple of time_t variables are only used to track the state of the
lease time and its expiration. The code correctly uses the 'time_after()'
macro to make this work on 32-bit architectures even beyond year 2038,
but the get_seconds() function and the time_t type itself are deprecated
as they behave inconsistently between 32-bit and 64-bit architectures
and often lead to code that is not y2038 safe.
As a minor issue, using get_seconds() leads to problems with concurrent
settimeofday() or clock_settime() calls, in the worst case timeout never
triggering after the time has been set backwards.
Change nfsd to use time64_t and ktime_get_boottime_seconds() here. This
is clearly excessive, as boottime by itself means we never go beyond 32
bits, but it does mean we handle this correctly and consistently without
having to worry about corner cases and should be no more expensive than
the previous implementation on 64-bit architectures.
The max_cb_time() function gets changed in order to avoid an expensive
64-bit division operation, but as the lease time is at most one hour,
there is no change in behavior.
Also do the same for server-to-server copy expiration time.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
[bfields@redhat.com: fix up copy expiration]
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The local boot time variable gets truncated to time_t at the moment,
which can lead to slightly odd behavior on 32-bit architectures.
Use ktime_get_real_seconds() instead of get_seconds() to always
get a 64-bit result, and keep it that way wherever possible.
It still gets truncated in a few places:
- When assigning to cl_clientid.cl_boot, this is already documented
and is only used as a unique identifier.
- In clients_still_reclaiming(), the truncation is to 'unsigned long'
in order to use the 'time_before() helper.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Add support to allow the server to reset the boot verifier in order to
force clients to resend I/O after a timeout failure.
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Lance Shelton <lance.shelton@hammerspace.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
We want clientid's on the wire to be randomized for reasons explained in
ebd7c72c63 "nfsd: randomize SETCLIENTID reply to help distinguish
servers". But I'd rather have mostly small integers for the clients/
directory.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Keep around one internal mount of the nfsd filesystem so that we can add
stuff to it when clients come and go, regardless of whether anyone has
it mounted.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
After 89a26b3d29 "nfsd: split DRC global spinlock into per-bucket
locks", there is no longer a single global spinlock to protect these
stats.
So, really we need to fix that. For now, at least fix the comment.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The nfsd duplicate reply cache should not be shared between network
namespaces.
The most straightforward way to fix this is just to move every global in
the code to per-net-namespace memory, so that's what we do.
Still todo: sort out which members of nfsd_stats should be global and
which per-net-namespace.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Based on 2 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version this program is distributed in the
hope that it will be useful but without any warranty without even
the implied warranty of merchantability or fitness for a particular
purpose see the gnu general public license for more details you
should have received a copy of the gnu general public license along
with this program if not write to the free software foundation inc
51 franklin street fifth floor boston ma 02110 1301 usa
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option [no]_[pad]_[ctrl] any later version this program is
distributed in the hope that it will be useful but without any
warranty without even the implied warranty of merchantability or
fitness for a particular purpose see the gnu general public license
for more details you should have received a copy of the gnu general
public license along with this program if not write to the free
software foundation inc 51 franklin street fifth floor boston ma
02110 1301 usa
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-or-later
has been chosen to replace the boilerplate/reference in 176 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Jilayne Lovejoy <opensource@jilayne.com>
Reviewed-by: Steve Winslow <swinslow@gmail.com>
Reviewed-by: Allison Randal <allison@lohutok.net>
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190519154040.652910950@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Support use of the --nfs-version/--no-nfs-version arguments to rpc.nfsd
in containers.
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
When using nfsdcld for NFSv4 client tracking, track the number of
RECLAIM_COMPLETE operations we receive from "known" clients to help in
deciding if we can lift the grace period early (or whether we need to
start a v4 grace period at all).
Signed-off-by: Scott Mayhew <smayhew@redhat.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Upon receiving a request for async copy, create a new kthread. If we
get asynchronous request, make sure to copy the needed arguments/state
from the stack before starting the copy. Then start the thread and reply
back to the client indicating copy is asynchronous.
nfsd_copy_file_range() will copy in a loop over the total number of
bytes is needed to copy. In case a failure happens in the middle, we
ignore the error and return how much we copied so far. Once done
creating a workitem for the callback workqueue and send CB_OFFLOAD with
the results.
The lifetime of the copy stateid is bound to the vfs copy. This way we
don't need to keep the nfsd_net structure for the callback. We could
keep it around longer so that an OFFLOAD_STATUS that came late would
still get results, but clients should be able to deal without that.
We handle OFFLOAD_CANCEL by sending a signal to the copy thread and
calling kthread_stop.
A client should cancel any ongoing copies before calling DESTROY_CLIENT;
if not, we return a CLIENT_BUSY error.
If the client is destroyed for some other reason (lease expiration, or
server shutdown), we must clean up any ongoing copies ourselves.
Signed-off-by: Olga Kornievskaia <kolga@netapp.com>
[colin.king@canonical.com: fix leak in error case]
[bfields@fieldses.org: remove signalling, merge patches]
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
If the client is only renewing state a little sooner than once a lease
period, then it might not discover the server has restarted till close
to the end of the grace period, and might run out of time to do the
actual reclaim.
Extend the grace period by a second each time we notice there are
clients still trying to reclaim, up to a limit of another whole lease
period.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
nfsd_inet[6]addr_event uses nn->nfsd_serv without taking nfsd_mutex,
which can be changed during execution of notifiers and crash the host.
Moreover if notifiers were enabled in one net namespace they are enabled
in all other net namespaces, from creation until destruction.
This patch allows notifiers to access nn->nfsd_serv only after the
pointer is correctly initialized and delays cleanup until notifiers are
no longer in use.
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Tested-by: Scott Mayhew <smayhew@redhat.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
do_gettimeofday() is deprecated and we should generally use time64_t
based functions instead.
In case of nfsd, all three users of nfssvc_boot only use the initial
time as a unique token, and are not affected by it overflowing, so they
are not affected by the y2038 overflow.
This converts the structure to timespec64 anyway and adds comments
to all uses, to document that we have thought about it and avoid
having to look at it again.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Make struct pernet_operations::id unsigned.
There are 2 reasons to do so:
1)
This field is really an index into an zero based array and
thus is unsigned entity. Using negative value is out-of-bound
access by definition.
2)
On x86_64 unsigned 32-bit data which are mixed with pointers
via array indexing or offsets added or subtracted to pointers
are preffered to signed 32-bit data.
"int" being used as an array index needs to be sign-extended
to 64-bit before being used.
void f(long *p, int i)
{
g(p[i]);
}
roughly translates to
movsx rsi, esi
mov rdi, [rsi+...]
call g
MOVSX is 3 byte instruction which isn't necessary if the variable is
unsigned because x86_64 is zero extending by default.
Now, there is net_generic() function which, you guessed it right, uses
"int" as an array index:
static inline void *net_generic(const struct net *net, int id)
{
...
ptr = ng->ptr[id - 1];
...
}
And this function is used a lot, so those sign extensions add up.
Patch snipes ~1730 bytes on allyesconfig kernel (without all junk
messing with code generation):
add/remove: 0/0 grow/shrink: 70/598 up/down: 396/-2126 (-1730)
Unfortunately some functions actually grow bigger.
This is a semmingly random artefact of code generation with register
allocator being used differently. gcc decides that some variable
needs to live in new r8+ registers and every access now requires REX
prefix. Or it is shifted into r12, so [r12+0] addressing mode has to be
used which is longer than [r8]
However, overall balance is in negative direction:
add/remove: 0/0 grow/shrink: 70/598 up/down: 396/-2126 (-1730)
function old new delta
nfsd4_lock 3886 3959 +73
tipc_link_build_proto_msg 1096 1140 +44
mac80211_hwsim_new_radio 2776 2808 +32
tipc_mon_rcv 1032 1058 +26
svcauth_gss_legacy_init 1413 1429 +16
tipc_bcbase_select_primary 379 392 +13
nfsd4_exchange_id 1247 1260 +13
nfsd4_setclientid_confirm 782 793 +11
...
put_client_renew_locked 494 480 -14
ip_set_sockfn_get 730 716 -14
geneve_sock_add 829 813 -16
nfsd4_sequence_done 721 703 -18
nlmclnt_lookup_host 708 686 -22
nfsd4_lockt 1085 1063 -22
nfs_get_client 1077 1050 -27
tcf_bpf_init 1106 1076 -30
nfsd4_encode_fattr 5997 5930 -67
Total: Before=154856051, After=154854321, chg -0.00%
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Bruce was hitting some lockdep warnings in testing, showing that we
could hit a deadlock with the new CB_NOTIFY_LOCK handling, involving a
rather complex situation involving four different spinlocks.
The crux of the matter is that we end up taking the nn->client_lock in
the lm_notify handler. The simplest fix is to just declare a new
per-nfsd_net spinlock to protect the new CB_NOTIFY_LOCK structures.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
It's possible for a client to call in on a lock that is blocked for a
long time, but discontinue polling for it. A malicious client could
even set a lock on a file, and then spam the server with failing lock
requests from different lockowners that pile up in a DoS attack.
Add the blocked lock structures to a per-net namespace LRU when hashing
them, and timestamp them. If the lock request is not revisited after a
lease period, we'll drop it under the assumption that the client is no
longer interested.
This also gives us a mechanism to clean up these objects at server
shutdown time as well.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The nfsd4_client_tracking_ops structures are never modified, so declare
them as const.
Done with the help of Coccinelle.
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Reviewed-by: Jeff Layton <jlayton@poochiereds.net>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
If using clientid_counter, it seems possible that gen_confirm could
generate the same verifier for the same client in some situations.
Add a new counter for client confirm verifier to make sure gen_confirm
generates a different verifier on each call for the same clientid.
Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Reviewed-by: Jeff Layton <jlayton@poochiereds.net>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Add some comments that describe what each of these objects is, and how
they related to one another.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The clid counter is a global counter currently. Move it to be a per-net
property so that it can be properly protected by the nn->client_lock
instead of relying on the client_mutex.
The verifier generator is also potentially racy if there are two
simultaneous callers. Generate the verifier when we generate the clid
value, so it's also created under the client_lock. With this, there's
no need to keep two counters as they'd always be in sync anyway, so
just use the clientid_counter for both.
As Trond points out, what would be best is to eventually move this
code to use IDR instead of the hash tables. That would also help ensure
uniqueness, but that's probably best done as a separate project.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Preparation for removing the client_mutex.
Convert the open owner hash table into a per-client table and protect it
using the nfs4_client->cl_lock spin lock.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Just like open-owners, lock-owners are associated with a name, a clientid
and, in the case of minor version 0, a sequence id. There is no association
to a file.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Currently, the maximum number of connections that nfsd will allow
is based on the number of threads spawned. While this is fine for a
default, there really isn't a clear relationship between the two.
The number of threads corresponds to the number of concurrent requests
that we want to allow the server to process at any given time. The
connection limit corresponds to the maximum number of clients that we
want to allow the server to handle. These are two entirely different
quantities.
Break the dependency on increasing threads in order to allow for more
connections, by adding a new per-net parameter that can be set to a
non-zero value. The default is still to base it on the number of threads,
so there should be no behavior change for anyone who doesn't use it.
Cc: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
When starting without nfsv2 and nfsv3, nfsd does not need to start
lockd (and certainly doesn't need to fail because lockd failed to
register with the portmapper).
Reported-by: Gareth Williams <gareth@garethwilliams.me.uk>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This patch makes main step in NFSd containerisation.
There could be different approaches to how to make NFSd able to handle
incoming RPC request from different network namespaces. The two main
options are:
1) Share NFSd kthreads betwween all network namespaces.
2) Create separated pool of threads for each namespace.
While first approach looks more flexible, second one is simpler and
non-racy. This patch implements the second option.
To make it possible to allocate separate pools of threads, we have to
make it possible to allocate separate NFSd service structures per net.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This is simple: an NFSd service can be started at different times in
different network environments. So, its "boot time" has to be assigned
per net.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This patch introduces introduces per-net "nfsd_net_up" boolean flag, which has
the same purpose as general "nfsd_up" flag - skip init or shutdown of per-net
resources in case of they are inited on shutted down respectively.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Pointer to client tracking operations - client_tracking_ops - have to be
containerized, because different environment can support different trackers
(for example, legacy tracker currently is not suported in container).
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
I added in a generic for-each loop that takes a pass over the client_lru
list for the current net namespace and calls some function. The next few
patches will update other operations to use this function as well. A value
of 0 still means "forget everything that is found".
Signed-off-by: Bryan Schumaker <bjschuma@netapp.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Grace time is a part of NFSv4 state engine, which is constructed per network
namespace.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Lease time is a part of NFSv4 state engine, which is constructed per network
namespace.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Flag in_grace is a part of client tracking state, which is network namesapce
aware. So let'a replace global static variable with per-net one.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Opening and closing of this file is done in client tracking init and exit
operations.
Client tracking is done in network namespace context already. So let's make
this file opened and closed per network context - this will simlify it's
management.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This lock protects the client lru list and session hash table, which are
allocated per network namespace already.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This patch moves laundromat_work to nfsd per-net context, thus allowing to run
multiple laundries.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This list holds nfs4 clients (open) stateowner queue for last close replay,
which are network namespace aware. So let's make this list per network
namespace too.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This list holds nfs4 clients queue for lease renewal, which are network
namespace aware. So let's make this list per network namespace too.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This hash holds established sessions state and closely associated with
nfs4_clients info, which are network namespace aware. So let's make it
allocated per network namespace too.
Note: this hash can be allocated in per-net operations. But it looks
better to allocate it on nfsd state start and thus don't waste resources
if server is not running.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This hash holds file lock owners and closely associated with nfs4_clients info,
which are network namespace aware. So let's make it allocated per network
namespace too.
Note: this hash can be allocated in per-net operations. But it looks
better to allocate it on nfsd state start and thus don't waste resources
if server is not running.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This hash holds open owner state and closely associated with nfs4_clients
info, which are network namespace aware. So let's make it allocated per
network namespace too.
Note: this hash can be allocated in per-net operations. But it looks
better to allocate it on nfsd state start and thus don't waste resources
if server is not running.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
This hash holds nfs4_clients info, which are network namespace aware.
So let's make it allocated per network namespace.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Dai Ngo
Trond Myklebust
Amir Goldstein
Randy Dunlap
J. Bruce Fields
Scott Mayhew
Arnd Bergmann
Trond Myklebust
Thomas Gleixner
Olga Kornievskaia
Vasily Averin
Alexey Dobriyan
Jeff Layton
Julia Lawall
Kinglong Mee
Jeff Layton
Trond Myklebust
Stanislav Kinsbursky
Bryan Schumaker