> On May 19, 2020, at 6:25 PM, Chuck Lever <chuck.lever@xxxxxxxxxx> wrote: > > Hi Bruce- > >> On May 19, 2020, at 5:29 PM, Bruce Fields <bfields@xxxxxxxxxxxx> wrote: >> >> On Tue, May 19, 2020 at 12:14:22PM -0400, Chuck Lever wrote: >>> >>> >>>> On May 19, 2020, at 12:11 PM, J. Bruce Fields <bfields@xxxxxxxxxxxx> wrote: >>>> >>>> I'm getting a repeatable timeout failure on python 4.0 test WRT15. In >>>> pynfs, run: >>>> >>>> ./nfs4.0/testserver.py server:/export/path --rundeps --maketree WRT15 >>>> >>>> Looks like it sends WRITE+GETATTR(FATTR4_SIZE) compounds with write >>>> offset 0 and write length taking on every value from 0 to 8192. >>>> >>>> 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. >> >> It's SUNRPC: Refactor svc_recvfrom(). >> >> That was just from a quick automated bisect. I haven't tried to figure >> out where the bug is.... > > Your reproducer isn't working for me on EL7. > > [root@manet ~]# yum install krb5-devel python3-devel swig python3-gssapi python3-ply > Loaded plugins: ulninfo > Package krb5-devel-1.15.1-46.el7.x86_64 already installed and latest version > Package python3-devel-3.6.8-13.0.1.el7.x86_64 already installed and latest version > Package swig-2.0.10-5.el7.x86_64 already installed and latest version > No package python3-gssapi available. > No package python3-ply available. > Nothing to do > [root@manet ~]# logout > [cel@manet pynfs]$ ./nfs4.0/testserver.py server:/export/path --rundeps --maketree WRT15 > Traceback (most recent call last): > File "./nfs4.0/testserver.py", line 388, in <module> > main() > File "./nfs4.0/testserver.py", line 242, in main > opt.machinename = os.fsencode(opt.machinename) > AttributeError: 'module' object has no attribute 'fsencode' > [cel@manet pynfs]$ I've reproduced your original test failure. I can take a closer look tomorrow. Looks like python3 is now a requirement for pynfs, despite the comments and code in nfs4.0/testserver.py. Also, the README should explain that the server under test has to permit access from insecure source ports (this still might not be the default for some NFS servers). -- Chuck Lever
