Re: [PATCH v2 4/5] rds: Add runchecks.cfg for net/rds

From: santosh.shilimkar@xxxxxxxxxx
Date: Sat Dec 16 2017 - 15:00:35 EST


On 12/16/17 10:24 AM, Joe Perches wrote:
On Sat, 2017-12-16 at 09:45 -0800, Stephen Hemminger wrote:
On Sat, 16 Dec 2017 15:42:29 +0100 Knut Omang <knut.omang@xxxxxxxxxx> wrote:
+# Code simplification:
+#
+except ALLOC_WITH_MULTIPLY ib.c
+except PREFER_PR_LEVEL ib_cm.c ib_recv.c ib_send.c rdma_transport.c threads.c transport.c
+except UNNECESSARY_ELSE ib_fmr.c
+except UNNECESSARY_PARENTHESES ib_rdma.c rdma.c recv.c send.c
+except PRINTK_RATELIMITED ib_frmr.c
+except EMBEDDED_FUNCTION_NAME ib_rdma.c
+
+# Style and readability:
+#
+except BRACES ib_cm.c ib_rdma.c ib_recv.c send.c transport.c
+except OOM_MESSAGE ib.c tcp.c
+except LONG_LINE_STRING ib.c ib_recv.c ib_send.c
+except FUNCTION_ARGUMENTS ib.h ib_mr.h rds.h tcp.h
+except OPEN_ENDED_LINE recv.c ib_recv.c
+
+# Candidates to leave as exceptions (don't fix):
+except MULTIPLE_ASSIGNMENTS ib_send.c
+except LONG_LINE_STRING connection.c
+except OPEN_BRACE connection.c
+

Why start letting subsystems have a free-pass?
Also this would mean that new patches to IB would continue the bad habits.
And I don't need any free pass for RDS either.

I missed V1 of this series but Knut, please don't add
any exceptions for RDS and if there is something needs to
be fixed, we can address it. Once your infrastructure
gets merged, the subsequent fixes can be added.


I agree with this comment at least for net/rds.

Most of these existing messages from checkpatch should
probably be inspected and corrected where possible to
minimize the style differences between this subsystem
and the rest of the kernel.

For instance, here's a trivial patch to substitute
pr_<level> for printks and a couple braces next to
these substitutions.

Thanks Joe. I actually had a similar patch a while back but
since it was lot of churn, and code was already merged,
never submitted it and then later forgot about it.

Will look into it.

btw:

in ib_cm, why is one call to ib_modify_qp emitted
with a -ret and the other with a positive err?

Its oversight and will fix that.

Regards,
Santosh