- 16 Oct, 2023 40 commits
-
-
Chuck Lever authored
Refactor the encoder for FATTR4_CHANGE into a helper. In a subsequent patch, this helper will be called from a bitmask loop. The code is restructured a bit to use the modern xdr_stream flow, and the encoded cinfo value is made const so that callers of the encoders can be passed a const cinfo. Reviewed-by:
Jeff Layton <jlayton@kernel.org> Signed-off-by:
Chuck Lever <chuck.lever@oracle.com>
-
Chuck Lever authored
Refactor the encoder for FATTR4_FH_EXPIRE_TYPE into a helper. In a subsequent patch, this helper will be called from a bitmask loop. Reviewed-by:
-
Chuck Lever authored
Refactor the encoder for FATTR4_TYPE into a helper. In a subsequent patch, this helper will be called from a bitmask loop. In addition, restructure the code so that byte-swapping is done on constant values rather than at run time. Run-time swapping can be costly on some platforms, and "type" is a frequently-requested attribute. Reviewed-by:
-
Chuck Lever authored
Refactor the encoder for FATTR4_SUPPORTED_ATTRS into a helper. In a subsequent patch, this helper will be called from a bitmask loop. Reviewed-by:
-
Chuck Lever authored
Add an encoding helper that encodes a single boolean "false" value. Attributes that always return "false" can use this helper. In a subsequent patch, this helper will be called from a bitmask loop, so it is given a standardized synopsis. Reviewed-by:
-
Chuck Lever authored
Add an encoding helper that encodes a single boolean "true" value. Attributes that always return "true" can use this helper. In a subsequent patch, this helper will be called from a bitmask loop, so it is given a standardized synopsis. Reviewed-by:
-
Chuck Lever authored
I'm about to split nfsd4_encode_fattr() into a number of smaller functions. Instead of passing a large number of arguments to each of the smaller functions, create a struct that can gather the common argument variables into something with a convenient handle on it. Reviewed-by:
-
Chuck Lever authored
De-duplicate the encoding of bitmap4 results in nfsd4_encode_setattr(). Reviewed-by:
-
Chuck Lever authored
For alignment with the specification, the name of NFSD's encoder function should match the name of the XDR type. I've also replaced a few "naked integers" with symbolic constants that better reflect the usage of these values. Reviewed-by:
-
Chuck Lever authored
The generic XDR encoders return a length or a negative errno. NFSv4 encoders want to know simply whether the encode ran out of stream buffer space. The return values for server-side encoding are either nfs_ok or nfserr_resource. So far I've found it adds a lot of duplicate code to try to use the generic XDR encoder utilities when encoding the simple data types in the NFSv4 operation encoders. Add a set of NFSv4-specific utilities that handle the basic XDR data types. These are added in xdr4.h so they might eventually be used by the callback server and pNFS driver encoders too. Reviewed-by:
-
Chuck Lever authored
There is no need to take down the whole system for these assertions. I'd rather not attempt a heroic save here, as some bug has occurred that has left the transport data structures in an unknown state. Just warn and then leak the left-over resources. Acked-by:
Christian Brauner <brauner@kernel.org> Reviewed-by:
NeilBrown <neilb@suse.de> Reviewed-by:
-
Kinglong Mee authored
Signed-off-by:
Kinglong Mee <kinglongmee@gmail.com> Signed-off-by:
-
Chuck Lever authored
Signed-off-by: -
Lorenzo Bianconi authored
Introduce rpc_status netlink support for NFSD in order to dump pending RPC requests debugging information from userspace. Closes: https://bugzilla.linux-nfs.org/show_bug.cgi?id=366Tested-by:
Lorenzo Bianconi <lorenzo@kernel.org> Signed-off-by:
-
Lorenzo Bianconi authored
Generate stubs and uAPI for nfsd netlink protocol. For the moment, the new protocol has one operation: rpc_status. The generated header and source files are created by running: tools/net/ynl/ynl-regen.sh Tested-by:
-
Dai Ngo authored
If the GETATTR request on a file that has write delegation in effect and the request attributes include the change info and size attribute then the request is handled as below: Server sends CB_GETATTR to client to get the latest change info and file size. If these values are the same as the server's cached values then the GETATTR proceeds as normal. If either the change info or file size is different from the server's cached values, or the file was already marked as modified, then: . update time_modify and time_metadata into file's metadata with current time . encode GETATTR as normal except the file size is encoded with the value returned from CB_GETATTR . mark the file as modified If the CB_GETATTR fails for any reasons, the delegation is recalled and NFS4ERR_DELAY is returned for the GETATTR. Signed-off-by:Dai Ngo <dai.ngo@oracle.com> Reviewed-by:
-
Dai Ngo authored
Includes: . CB_GETATTR proc for nfs4_cb_procedures[] . XDR encoding and decoding function for CB_GETATTR request/reply . add nfs4_cb_fattr to nfs4_delegation for sending CB_GETATTR and store file attributes from client's reply. Signed-off-by: -
NeilBrown authored
This removes the need to store and update back-links in the list. It also remove the need for the _bh version of spin_lock(). Signed-off-by:
-
NeilBrown authored
sp_lock is now only used to protect sp_all_threads. This isn't needed as sp_all_threads is only manipulated through svc_set_num_threads(), which is already serialized. Read-acccess only requires rcu_read_lock(). So no more locking is needed. Signed-off-by:
-
NeilBrown authored
Using an atomic_t avoids the need to take a spinlock (which can soon be removed). Choosing a thread to kill needs to be careful as we cannot set the "die now" bit atomically with the test on the count. Instead we temporarily increase the count. Signed-off-by:
-
NeilBrown authored
lwq avoids using back pointers in lists, and uses less locking. This introduces a new spinlock, but the other one will be removed in a future patch. For svc_clean_up_xprts(), we now dequeue the entire queue, walk it to remove and process the xprts that need cleaning up, then re-enqueue the remaining queue. Signed-off-by:
-
NeilBrown authored
Currently if several items of work become available in quick succession, that number of threads (if available) will be woken. By the time some of them wake up another thread that was already cache-warm might have come along and completed the work. Anecdotal evidence suggests as many as 15% of wakes find nothing to do once they get to the point of looking. This patch changes svc_pool_wake_idle_thread() to wake the first thread on the queue but NOT remove it. Subsequent calls will wake the same thread. Once that thread starts it will dequeue itself and after dequeueing some work to do, it will wake the next thread if there is more work ready. This results in a more orderly increase in the number of busy threads. As a bonus, this allows us to reduce locking around the idle queue. svc_pool_wake_idle_thread() no longer needs to take a lock (beyond rcu_read_lock()) as it doesn't manipulate the queue, it just looks at the first item. The thread itself can avoid locking by using the new llist_del_first_this() interface. This will safely remove the thread itself if it is the head. If it isn't the head, it will do nothing. If multiple threads call this concurrently only one will succeed. The others will do nothing, so no corruption can result. If a thread wakes up and finds that it cannot dequeue itself that means either - that it wasn't woken because it was the head of the queue. Maybe the freezer woke it. In that case it can go back to sleep (after trying to freeze of course). - some other thread found there was nothing to do very recently, and placed itself on the head of the queue in front of this thread. It must check again after placing itself there, so it can be deemed to be responsible for any pending work, and this thread can go back to sleep until woken. No code ever tests for busy threads any more. Only each thread itself cares if it is busy. So svc_thread_busy() is no longer needed. Signed-off-by:
-
NeilBrown authored
Functions which directly manipulate a 'struct rqst', such as svc_rqst_alloc() or svc_rqst_release_pages(), can reasonably have "rqst" in there name. However functions that act on the running thread, such as XX_should_sleep() or XX_wait_for_work() should seem more natural with a "svc_thread_" prefix. So make those changes. Signed-off-by:
-
NeilBrown authored
lwq is a FIFO single-linked queue that only requires a spinlock for dequeueing, which happens in process context. Enqueueing is atomic with no spinlock and can happen in any context. This is particularly useful when work items are queued from BH or IRQ context, and when they are handled one at a time by dedicated threads. Avoiding any locking when enqueueing means there is no need to disable BH or interrupts, which is generally best avoided (particularly when there are any RT tasks on the machine). This solution is superior to using "list_head" links because we need half as many pointers in the data structures, and because list_head lists would need locking to add items to the queue. This solution is superior to a bespoke solution as all locking and container_of casting is integrated, so the interface is simple. Despite the similar name, this solution meets a distinctly different need to kfifo. kfifo provides a fixed sized circular buffer to which data can be added at one end and removed at the other, and does not provide any locking. lwq does not have any size limit and works with data structures (objects?) rather than data (bytes). A unit test for basic functionality, which runs at boot time, is included. Signed-off-by:
-
NeilBrown authored
llist_del_first_this() deletes a specific entry from an llist, providing it is at the head of the list. Multiple threads can call this concurrently providing they each offer a different entry. This can be uses for a set of worker threads which are on the llist when they are idle. The head can always be woken, and when it is woken it can remove itself, and possibly wake the next if there is an excess of work to do. Signed-off-by:
-
NeilBrown authored
With an llist we don't need to take a lock to add a thread to the list, though we still need a lock to remove it. That will go in the next patch. Unlike double-linked lists, a thread cannot reliably remove itself from the list. Only the first thread can be removed, and that can change asynchronously. So some care is needed. We already check if there is pending work to do, so we are unlikely to add ourselves to the idle list and then want to remove ourselves again. If we DO find something needs to be done after adding ourselves to the list, we simply wake up the first thread on the list. If that was us, we successfully removed ourselves and can continue. If it was some other thread, they will do the work that needs to be done. We can safely sleep until woken. We also remove the test on freezing() from rqst_should_sleep(). Instead we set TASK_FREEZABLE before scheduling. This makes is safe to schedule() when a freeze is pending. As we now loop waiting to be removed from the idle queue, this is a cleaner way to handle freezing. Signed-off-by:
-
NeilBrown authored
With list.h lists, it is easy to test if a node is on a list, providing it was initialised and that it is removed with list_del_init(). This patch provides similar functionality for llist.h lists. init_llist_node() marks a node as being not-on-any-list be setting the ->next pointer to the node itself. llist_on_list() tests if the node is on any list. llist_del_first_init() remove the first element from a llist, and marks it as being off-list. Signed-off-by:
-
NeilBrown authored
We can tell if a pool is congested by checking if the idle list is empty. We don't need a separate flag. Signed-off-by:
-
NeilBrown authored
Rather than searching a list of threads to find an idle one, having a list of idle threads allows an idle thread to be found immediately. This adds some spin_lock calls which is not ideal, but as the hold-time is tiny it is still faster than searching a list. A future patch will remove them using llist.h. This involves some subtlety and so is left to a separate patch. This removes the need for the RQ_BUSY flag. The rqst is "busy" precisely when it is not on the "idle" list. Signed-off-by:
-
NeilBrown authored
svc threads are currently stopped using kthread_stop(). This requires identifying a specific thread. However we don't care which thread stops, just as long as one does. So instead, set a flag in the svc_pool to say that a thread needs to die, and have each thread check this flag instead of calling kthread_should_stop(). The first thread to find and clear this flag then moves towards exiting. This removes an explicit dependency on sp_all_threads which will make a future patch simpler. Signed-off-by:
-
NeilBrown authored
Both nfsd and nfsv4-callback take a temporary reference to the svc_serv while calling svc_set_num_threads() to stop the last thread. lockd does not. This extra reference prevents the scv_serv from being freed when the last thread drops its reference count. This is not currently needed for lockd as the svc_serv is not accessed after the last thread is told to exit. However a future patch will require svc_exit_thread() to access the svc_serv after the svc_put() so it will need the code that calls svc_set_num_threads() to keep a reference and keep the svc_serv active. So copy the pattern from nfsd and nfsv4-cb to lockd, and take a reference around svc_set_num_threads(.., 0) Reviewed-by:
-
NeilBrown authored
Using svc_recv() for (NFSv4.1) back-channel handling means we have just one mechanism for waking threads. Also change kthread_freezable_should_stop() in nfs4_callback_svc() to kthread_should_stop() as used elsewhere. kthread_freezable_should_stop() effectively adds a try_to_freeze() call, and svc_recv() already contains that at an appropriate place. Signed-off-by:
-
Chuck Lever authored
The test robot complained that, in some build configurations, the @error variable in bc_svc_process's only caller is set but never used. This happens because dprintk() is the only consumer of that value. - Remove the dprintk() call sites in favor of the svc_process tracepoint - The @error variable and the return value of bc_svc_process() are now unused, so get rid of them. - The @serv parameter is set to rqstp->rq_serv by the only caller, and bc_svc_process() then uses it only to set rqstp->rq_serv. It can be removed. - Rename bc_svc_process() according to the convention that globally-visible RPC server functions have names that begin with "svc_"; and because it is globally-visible, give it a proper kdoc comment. Reported-by:
kernel test robot <lkp@intel.com> Closes: https://lore.kernel.org/oe-kbuild-all/202308121314.HA8Rq2XG-lkp@intel.com/Signed-off-by:
-
NeilBrown authored
svc_get_next_xprt() does a lot more than just get an xprt. It also decides if it needs to sleep, depending not only on the availability of xprts but also on the need to exit or handle external work. So rename it to svc_rqst_wait_for_work() and only do the testing and waiting. Move all the waiting-related code out of svc_recv() into the new svc_rqst_wait_for_work(). Move the dequeueing code out of svc_get_next_xprt() into svc_recv(). Previously svc_xprt_dequeue() would be called twice, once before waiting and possibly once after. Now instead rqst_should_sleep() is called twice. Once to decide if waiting is needed, and once to check against after setting the task state do see if we might have missed a wakeup. Signed-off-by:
-
NeilBrown authored
svc_xprt_handle() does lots of things itself, but leaves some to the caller - svc_recv(). This isn't elegant. Move that code out of svc_recv() into svc_xprt_handle() Move the calls to svc_xprt_release() from svc_send() and svc_drop() (the two possible final steps in svc_process()) and from svc_recv() (in the case where svc_process() wasn't called) into svc_xprt_handle(). Signed-off-by:
-
Alexander Aring authored
This patch adds a note to enable EXPORT_OP_ASYNC_LOCK for asynchronous lock request handling. Reviewed-by:
Alexander Aring <aahringo@redhat.com> Signed-off-by:
-
Alexander Aring authored
This patch fixes a race in async lock request handling between adding the relevant struct nlm_block to nlm_blocked list after the request was sent by vfs_lock_file() and nlmsvc_grant_deferred() does a lookup of the nlm_block in the nlm_blocked list. It could be that the async request is completed before the nlm_block was added to the list. This would end in a -ENOENT and a kernel log message of "lockd: grant for unknown block". To solve this issue we add the nlm_block before the vfs_lock_file() call to be sure it has been added when a possible nlmsvc_grant_deferred() is called. If the vfs_lock_file() results in an case when it wouldn't be added to nlm_blocked list, the nlm_block struct will be removed from this list again. Reviewed-by:
-
Alexander Aring authored
This patch returns nlm_lck_blocked in nlmsvc_lock() when an asynchronous lock request is pending. During testing I ran into the case with the side-effects that lockd is waiting for only one lm_grant() callback because it's already part of the nlm_blocked list. If another asynchronous for the same nlm_block is triggered two lm_grant() callbacks will occur but lockd was only waiting for one. To avoid any change of existing users this handling will only being made when export_op_support_safe_async_lock() returns true. Reviewed-by:
-
Alexander Aring authored
This patch reverts mostly commit 40595cdc ("nfs: block notification on fs with its own ->lock") and introduces an EXPORT_OP_ASYNC_LOCK export flag to signal that the "own ->lock" implementation supports async lock requests. The only main user is DLM that is used by GFS2 and OCFS2 filesystem. Those implement their own lock() implementation and return FILE_LOCK_DEFERRED as return value. Since commit 40595cdc ("nfs: block notification on fs with its own ->lock") the DLM implementation were never updated. This patch should prepare for DLM to set the EXPORT_OP_ASYNC_LOCK export flag and update the DLM plock implementation regarding to it. Acked-by:
-
Trond Myklebust authored
If fsync() is returning EAGAIN, then we can assume that the filesystem being exported is something like NFS with the 'softerr' mount option enabled, and that it is just asking us to replay the fsync() operation at a later date. If we see an ESTALE, then ditto: the file is gone, so there is no danger of losing the error. For those cases, do not reset the write verifier. A write verifier change has a global effect, causing retransmission by all clients of all uncommitted unstable writes for all files, so it is worth mitigating where possible. Link: https://lore.kernel.org/linux-nfs/20230911184357.11739-1-trond.myklebust@hammerspace.com/Signed-off-by:
Trond Myklebust <trond.myklebust@hammerspace.com> Reviewed-by:
-
