Re: Regression in 5.1.20: Reading long directory fails

From: Jason L Tibbitts III
Date: Fri Sep 06 2019 - 16:47:34 EST

>>>>> "JBF" == J Bruce Fields <bfields@xxxxxxxxxxxx> writes:

JBF> Those readdir changes were client-side, right? Based on that I'd
JBF> been assuming a client bug, but maybe it'd be worth getting a full
JBF> packet capture of the readdir reply to make sure it's legit.

I have been working with bcodding on IRC for the past couple of days on
this. Fortunately I was able to come up with way to fill up a directory
in such a way that it will fail with certainty and as a bonus doesn't
include any user data so I can feel OK about sharing packet captures. I
have a capture alongside a kernel trace of the problematic operation in Not that I can particularly tell
anything useful from that, but bcodding says that it seems to point to
some issue in sunrpc.

And because I can easily reproduce this and I was able to do a bisect:

2c94b8eca1a26cd46010d6e73a23da5f2e93a19d is the first bad commit
commit 2c94b8eca1a26cd46010d6e73a23da5f2e93a19d
Author: Chuck Lever <chuck.lever@xxxxxxxxxx>
Date: Mon Feb 11 11:25:41 2019 -0500

SUNRPC: Use au_rslack when computing reply buffer size

au_rslack is significantly smaller than (au_cslack << 2). Using
that value results in smaller receive buffers. In some cases this
eliminates an extra segment in Reply chunks (RPC/RDMA).

Signed-off-by: Chuck Lever <chuck.lever@xxxxxxxxxx>
Signed-off-by: Anna Schumaker <Anna.Schumaker@xxxxxxxxxx>

:040000 040000 d4d1ce2fbe0035c5bd9df976b8c448df85dcb505 7011a792dfe72ff9cd70d66e45d353f3d7817e3e M net

But of course, I can't say whether this is the actual bad commit or
whether it just introduced a behavior change which alters the conditions
under which the problem appears.

And just to make sure that the blame doesn't lie with the old RHEL7
kernel, I rsynced over the problematic directory to a machine running
something slightly more modern (5.1.11, which I know I need to update,
but it's already set up to do kerberised NFS) and the same problem
exists, though the directory listing does fail at a different place.

- J<