Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Content-Type: text/plain;
        charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 12.4 \(3445.104.11\))
Subject: Re: [PATCH v2 00/29] Possible NFSD patches for v5.8
From:   Chuck Lever <chuck.lever@oracle.com>
In-Reply-To: <20200519161108.GD25858@fieldses.org>
Date:   Tue, 19 May 2020 12:14:22 -0400
Cc:     Linux NFS Mailing List <linux-nfs@vger.kernel.org>,
        linux-rdma@vger.kernel.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <81E97D7E-7B8D-4C64-844A-18EF0346C49C@oracle.com>
References: <20200512211640.5826.77139.stgit@klimt.1015granger.net>
 <20200519161108.GD25858@fieldses.org>
To:     Bruce Fields <bfields@fieldses.org>
Sender: linux-nfs-owner@vger.kernel.org
Precedence: bulk


> On May 19, 2020, at 12:11 PM, J. Bruce Fields <bfields@fieldses.org> =
wrote:
>=20
> I'm getting a repeatable timeout failure on python 4.0 test WRT15.  In
> pynfs, run:=20
>=20
> 	./nfs4.0/testserver.py server:/export/path --rundeps --maketree =
WRT15
>=20
> Looks like it sends WRITE+GETATTR(FATTR4_SIZE) compounds with write
> offset 0 and write length taking on every value from 0 to 8192.
>=20
> Probably an xdr decoding bug of some kind?

My first thought is to bisect, but I don't see a particular change in my
v5.8 series that would plausibly introduce this class of problem.


> I don't see anything in the server logs.
>=20
> --b.
>=20
> On Tue, May 12, 2020 at 05:22:04PM -0400, Chuck Lever wrote:
>> Available to view:
>>  =
https://git.linux-nfs.org/?p=3Dcel/cel-2.6.git;a=3Dshortlog;h=3Drefs/heads=
/nfsd-5.8
>>=20
>> Pull from:
>>  git://git.linux-nfs.org/projects/cel/cel-2.6.git nfsd-5.8
>>=20
>> Highlights of this series:
>> * Remove serialization of sending RPC/RDMA Replies
>> * Convert the TCP socket send path to use xdr_buf::bvecs =
(pre-requisite for RPC-on-TLS)
>> * Fix svcrdma backchannel sendto return code
>> * Convert a number of dprintk call sites to use tracepoints
>> * Fix the "suggest braces around empty body in an =E2=80=98else=E2=80=99=
 statement" warning
>>=20
>> Changes since v1:
>> * Rebased on v5.7-rc5+
>> * Re-organized the series so changes interesting to linux-rdma appear =
together
>> * Addressed sparse warnings found by the kbuild test robot
>> * Included an additional minor clean-up: removal of the unused =
SVCRDMA_DEBUG macro
>> * Clarified several patch descriptions
>>=20
>> ---
>>=20
>> Chuck Lever (29):
>>      SUNRPC: Move xpt_mutex into socket xpo_sendto methods
>>      svcrdma: Clean up the tracing for rw_ctx_init errors
>>      svcrdma: Clean up handling of get_rw_ctx errors
>>      svcrdma: Trace page overruns when constructing RDMA Reads
>>      svcrdma: trace undersized Write chunks
>>      svcrdma: Fix backchannel return code
>>      svcrdma: Remove backchannel dprintk call sites
>>      svcrdma: Rename tracepoints that record header decoding errors
>>      svcrdma: Remove the SVCRDMA_DEBUG macro
>>      svcrdma: Displayed remote IP address should match stored address
>>      svcrdma: Add tracepoints to report ->xpo_accept failures
>>      SUNRPC: Remove kernel memory address from svc_xprt tracepoints
>>      SUNRPC: Tracepoint to record errors in svc_xpo_create()
>>      SUNRPC: Trace a few more generic svc_xprt events
>>      SUNRPC: Remove "#include <trace/events/skb.h>"
>>      SUNRPC: Add more svcsock tracepoints
>>      SUNRPC: Replace dprintk call sites in TCP state change callouts
>>      SUNRPC: Trace server-side rpcbind registration events
>>      SUNRPC: Rename svc_sock::sk_reclen
>>      SUNRPC: Restructure svc_tcp_recv_record()
>>      SUNRPC: Refactor svc_recvfrom()
>>      SUNRPC: Restructure svc_udp_recvfrom()
>>      SUNRPC: svc_show_status() macro should have enum definitions
>>      NFSD: Add tracepoints to NFSD's duplicate reply cache
>>      NFSD: Add tracepoints to the NFSD state management code
>>      NFSD: Add tracepoints for monitoring NFSD callbacks
>>      SUNRPC: Clean up request deferral tracepoints
>>      NFSD: Squash an annoying compiler warning
>>      NFSD: Fix improperly-formatted Doxygen comments
>>=20
>>=20
>> fs/nfsd/nfs4callback.c                     |  37 +-
>> fs/nfsd/nfs4proc.c                         |   7 +-
>> fs/nfsd/nfs4state.c                        |  63 ++--
>> fs/nfsd/nfscache.c                         |  57 +--
>> fs/nfsd/nfsctl.c                           |  26 +-
>> fs/nfsd/state.h                            |   7 -
>> fs/nfsd/trace.h                            | 345 ++++++++++++++++++
>> include/linux/sunrpc/svc.h                 |   1 +
>> include/linux/sunrpc/svc_rdma.h            |   6 +-
>> include/linux/sunrpc/svcsock.h             |   6 +-
>> include/trace/events/rpcrdma.h             | 142 ++++++--
>> include/trace/events/sunrpc.h              | 387 ++++++++++++++++++--
>> net/sunrpc/svc.c                           |  19 +-
>> net/sunrpc/svc_xprt.c                      |  41 +--
>> net/sunrpc/svcsock.c                       | 393 =
++++++++++-----------
>> net/sunrpc/xprtrdma/svc_rdma_backchannel.c |  86 +----
>> net/sunrpc/xprtrdma/svc_rdma_recvfrom.c    |  21 +-
>> net/sunrpc/xprtrdma/svc_rdma_rw.c          |  92 ++---
>> net/sunrpc/xprtrdma/svc_rdma_transport.c   |  55 ++-
>> 19 files changed, 1221 insertions(+), 570 deletions(-)
>>=20
>> --
>> Chuck Lever

--
Chuck Lever