If the client interface is down, or there is a network partition between
the client and server that prevents the callback request to reach the
client, TCP on the server will keep re-transmitting the callback for about
~9 minutes before giving up and closing the connection.
If the connection between the client and the server is re-established
before the connection is closed and after the callback timed out (9 secs)
then the re-transmitted callback request will arrive at the client. When
the server receives the reply of the callback, receive_cb_reply prints the
"Got unrecognized reply..." message in the system log since the callback
request was already removed from the server xprt's recv_queue.
Even though this scenario has no effect on the server operation, a
malfunctioning or malicious client can fill up the server's system log.
Signed-off-by: Dai Ngo <dai.ngo@oracle.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Having an nfsd thread waiting for an RDMA Read completion is
problematic if the Read responder (ie, the client) stops responding.
We need to go back to handling RDMA Reads by getting the svc scheduler
to call svc_rdma_recvfrom() a second time to finish building an RPC
message after a Read completion.
This is the final patch, and makes several changes that have to
happen concurrently:
1. svc_rdma_process_read_list no longer waits for a completion, but
simply builds and posts the Read WRs.
2. svc_rdma_read_done() now queues a completed Read on
sc_read_complete_q for later processing rather than calling
complete().
3. The completed RPC message is no longer built in the
svc_rdma_process_read_list() path. Finishing the message is now
done in svc_rdma_recvfrom() when it notices work on the
sc_read_complete_q. The "finish building this RPC message" code
is removed from the svc_rdma_process_read_list() path.
This arrangement avoids the need for an nfsd thread to wait for an
RDMA Read non-interruptibly without a timeout. It's basically the
same code structure that Tom Tucker used for Read chunks along with
some clean-up and modernization.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Once a set of RDMA Reads are complete, the Read completion handler
will poke the transport to trigger a second call to
svc_rdma_recvfrom(). recvfrom() will then merge the RDMA Read
payloads with the previously received RPC header to form a completed
RPC Call message.
The new code is copied from the svc_rdma_process_read_list() path.
A subsequent patch will make use of this code and remove the code
that this was copied from (svc_rdma_rw.c).
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Having an nfsd thread waiting for an RDMA Read completion is
problematic if the Read responder (ie, the client) stops responding.
We need to go back to handling RDMA Reads by allowing the nfsd
thread to return to the svc scheduler, then waking a second thread
finish the RPC message once the Read completion fires.
As a next step, add a list_head upon which completed Reads are queued.
A subsequent patch will make use of this queue.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Having an nfsd thread waiting for an RDMA Read completion is
problematic if the Read responder (the client) stops responding. We
need to go back to handling RDMA Reads by allowing the nfsd thread
to return to the svc scheduler, then waking a second thread finish
the RPC message once the Read completion fires.
To start with, restore the rc_pages field so that RDMA Read pages
can be managed across calls to svc_rdma_recvfrom().
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
The comment that starts "Qualify ..." applies to only some of the
following code paragraph. Re-arrange the lines so the comment makes
more sense.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
These won't have much diagnostic value for site administrators.
Since they can't be disabled, they become noise.
What's more, the subsequent rdma_create_qp() call adjusts the Send
Queue size (possibly downward) without warning, making the size
reported by these pr_warns inaccurate.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
There are a couple of dprintk() call sites in svc_rdma_accept()
that show pointer addresses. These days, displayed pointer addresses
are hashed and thus have little or no diagnostic value, especially
for site administrators.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
The atomic_inc_return() in svc_rdma_send_cid_init() is expensive.
Some svc_rdma_chunk_ctxt's now reside in long-lived container
structures. They don't need a fresh completion ID for every I/O
operation.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Now that the chunk_ctxt for Reads is no longer dynamically allocated
it can be initialized once for the life of the object that contains
it (struct svc_rdma_recv_ctxt).
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_read_special() can use that recv_ctxt to derive the
read_info rather than the other way around. This removes another
usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_read_call_chunk() can use that recv_ctxt to derive the
read_info rather than the other way around. This removes another
usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_read_multiple_chunks() can use that recv_ctxt to derive the
read_info rather than the other way around. This removes another
usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_copy_inline_range() can use that recv_ctxt to derive the
read_info rather than the other way around. This removes another
usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_build_read_data_item() can use that recv_ctxt to derive
that information rather than the other way around. This removes
another usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_build_read_chunk_range() can use that recv_ctxt to derive
that information rather than the other way around. This removes
another usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_build_read_chunk() can use that recv_ctxt to derive that
information rather than the other way around. This removes another
usage of the ri_readctxt field, enabling its removal in a
subsequent patch.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Since the RDMA Read I/O state is now contained in the recv_ctxt,
svc_rdma_build_read_segment() can use the recv_ctxt to derive that
information rather than the other way around. This removes one usage
of the ri_readctxt field, enabling its removal in a subsequent
patch.
At the same time, the use of ri_rqst can similarly be replaced with
a passed-in function parameter.
Start with build_read_segment() because it is a common utility
function at the bottom of the Read chunk path.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Since the request's svc_rdma_recv_ctxt will stay around for the
duration of the RDMA Read operation, the contents of struct
svc_rdma_read_info can reside in the request's svc_rdma_recv_ctxt
rather than being allocated separately. This will eventually save a
call to kmalloc() in a hot path.
Start this clean-up by moving the Read chunk's svc_rdma_chunk_ctxt.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Prepare for nestling these into the send and recv ctxts so they
no longer have to be allocated dynamically.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
SG_CHUNK_SIZE is 128, making struct svc_rdma_rw_ctxt + the first
SGL array more than 4200 bytes in length, pushing the memory
allocation well into order 1.
Even so, the RDMA rw core doesn't seem to use more than max_send_sge
entries in that array (typically 32 or less), so that is all wasted
space.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
A send/recv_ctxt already records transport-related information
in the cq.id, thus there is no need to record the IP addresses of
the transport endpoints.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Update the DMA error flow tracepoints to report the completion ID of
the failing context. This ties the wait/failure to a particular
operation or request, which is more useful than knowing only the
failing transport.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Update the Send Queue's error flow tracepoints to report the
completion ID of the waiting or failing context. This ties the
wait/failure to a particular operation or request, which is a little
more useful than knowing only the transport that is about to close.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Two svcrdma-related transport locks can become quite contended.
Collate their use and make them easy to find in /proc/lock_stat for
better observability.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
DMA unmapping can take quite some time, so it should not be handled
in a single-threaded completion handler. Defer releasing write_info
structs to the recently-added workqueue.
With this patch, DMA unmapping can be handled in parallel, and it
does not cause head-of-queue blocking of Write completions.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
DMA unmapping can take quite some time, so it should not be handled
in a single-threaded completion handler. Defer releasing send_ctxts
to the recently-added workqueue.
With this patch, DMA unmapping can be handled in parallel, and it
does not cause head-of-queue blocking of Send completions.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
To handle work in the background, set up an UNBOUND workqueue for
svcrdma. Subsequent patches will make use of it.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
The original reason for allocating svc_rdma_recv_ctxt objects during
Receive completion was to ensure the objects were allocated on the
NUMA node closest to the underlying IB device.
Since commit c5d68d25bd ("svcrdma: Clean up allocation of
svc_rdma_recv_ctxt"), however, the device's favored node is
explicitly passed to the memory allocator.
To enable switching Receive completion to soft IRQ context, move
memory allocation out of completion handling, since it can be
costly, and it can sleep.
A limited number of objects is now allocated at "accept" time.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
The svc_rdma_recv_ctxt free list uses a lockless list to avoid the
need for a spin lock in the fast path. llist_del_first(), which is
used by svc_rdma_recv_ctxt_get(), requires serialization, however,
when there are multiple list producers that are unserialized.
I mistakenly thought there was only one caller of
svc_rdma_recv_ctxt_get() (svc_rdma_refresh_recvs()), thus explicit
serialization would not be necessary. But there is another caller:
svc_rdma_bc_sendto(), and these two are not serialized against each
other. I haven't seen ill effects that I could directly ascribe to
a lack of serialization. It's just an observation based on code
audit.
When DMA-mapping before sending a Reply, the passed-in struct
svc_rdma_recv_ctxt is used only for its write and reply PCLs. These
are currently always empty in the backchannel case. So, instead of
passing a full svc_rdma_recv_ctxt object to
svc_rdma_map_reply_msg(), let's pass in just the Write and Reply
PCLs.
This change makes it unnecessary for the backchannel to acquire a
dummy svc_rdma_recv_ctxt object when sending an RPC Call. The need
for svc_rdma_recv_ctxt free list serialization is now completely
avoided.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Callback operations enum is defined in client and server, move it to
common header file.
Signed-off-by: ChenXiaoSong <chenxiaosong@kylinos.cn>
Acked-by: Anna Schumaker <Anna.Schumaker@netapp.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
We check "state" for NULL on the previous line so it can't be NULL here.
No need to check again.
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/r/202312031425.LffZTarR-lkp@intel.com/
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Avoid the use of an atomic bitop, and prepare for adding a run-time
switch for using splice reads.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
RQ_SPLICE_OK is a bit of a layering violation. Also, a subsequent
patch is going to provide a mechanism for always disabling splice
reads.
Splicing is an issue only for NFS READs, so refactor nfsd_read() to
check the auth type directly instead of relying on an rq_flag
setting.
The new helper will be added into the NFSv4 read path in a
subsequent patch.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
NFSD will use this new API to determine whether nfsd_splice_read is
safe to use. This avoids the need to add a dependency to NFSD for
CONFIG_SUNRPC_GSS.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Al Viro notes that normal system calls hold f_pos_lock when calling
->iterate_shared and ->llseek; however nfsd_readdir() does not take
that mutex when calling these methods.
It should be safe however because the struct file acquired by
nfsd_readdir() is not visible to other threads.
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
This trace point was for debugging the DRC's garbage collection. In
the field it's just noise.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
