[iproute PATCH 0/2] Two minor testsuite fixes

[iproute PATCH 0/2] Two minor testsuite fixes

[Phil Sutter]

While playing around with testsuite, I noticed two minor nits which this
series attempts to fix.

Phil Sutter (2):
  testsuite: skip link show test on big endian systems
  testsuite: Search kernel config in modules dir also

 .gitignore                                    | 1 +
 testsuite/Makefile                            | 7 +++++++
 testsuite/tests/ip/link/show_dev_wo_vf_rate.t | 5 +++++
 testsuite/tools/Makefile                      | 2 ++
 testsuite/tools/is_big_endian.c               | 7 +++++++
 5 files changed, 22 insertions(+)
 create mode 100644 testsuite/tools/Makefile
 create mode 100644 testsuite/tools/is_big_endian.c

-- 
2.11.0

[iproute PATCH 1/2] testsuite: skip link show test on big endian systems

[Phil Sutter]

Netlink protocol is in host byte order, so the provided binary netlink
message buffer being in little endian format will cause the test to fail
on big endian systems.

Signed-off-by: Phil Sutter <phil@nwl.cc>
---
 .gitignore                                    | 1 +
 testsuite/Makefile                            | 1 +
 testsuite/tests/ip/link/show_dev_wo_vf_rate.t | 5 +++++
 testsuite/tools/Makefile                      | 2 ++
 testsuite/tools/is_big_endian.c               | 7 +++++++
 5 files changed, 16 insertions(+)
 create mode 100644 testsuite/tools/Makefile
 create mode 100644 testsuite/tools/is_big_endian.c

diff --git a/.gitignore b/.gitignore
index 74a5496ddf7aa..a1f295dac3dd2 100644
--- a/.gitignore
+++ b/.gitignore
@@ -36,6 +36,7 @@ series
 # tests
 testsuite/results
 testsuite/iproute2/iproute2-this
+testsuite/tools/is_big_endian
 
 # doc files generated at runtime
 doc/*.aux
diff --git a/testsuite/Makefile b/testsuite/Makefile
index 2027650051d48..17568881b8b66 100644
--- a/testsuite/Makefile
+++ b/testsuite/Makefile
@@ -24,6 +24,7 @@ configure:
 
 compile: configure
 	echo "Entering iproute2" && cd iproute2 && $(MAKE) && cd ..;
+	echo "Entering tools" && cd tools && $(MAKE) && cd ..;
 
 listtests:
 	@for t in $(TESTS); do \
diff --git a/testsuite/tests/ip/link/show_dev_wo_vf_rate.t b/testsuite/tests/ip/link/show_dev_wo_vf_rate.t
index a600ba65c5bec..ad90af5400271 100755
--- a/testsuite/tests/ip/link/show_dev_wo_vf_rate.t
+++ b/testsuite/tests/ip/link/show_dev_wo_vf_rate.t
@@ -2,5 +2,10 @@
 
 source lib/generic.sh
 
+if ./tools/is_big_endian; then
+	ts_log "won't work on big endian system"
+	ts_skip
+fi
+
 NL_FILE="tests/ip/link/dev_wo_vf_rate.nl"
 ts_ip "$0" "Show VF devices w/o VF rate info" -d monitor file $NL_FILE
diff --git a/testsuite/tools/Makefile b/testsuite/tools/Makefile
new file mode 100644
index 0000000000000..ecbea16c2c1cc
--- /dev/null
+++ b/testsuite/tools/Makefile
@@ -0,0 +1,2 @@
+is_big_endian: is_big_endian.c
+	$(CC) -o $@ $<
diff --git a/testsuite/tools/is_big_endian.c b/testsuite/tools/is_big_endian.c
new file mode 100644
index 0000000000000..303e91b4603e8
--- /dev/null
+++ b/testsuite/tools/is_big_endian.c
@@ -0,0 +1,7 @@
+//#include <stdio.h>
+#include <arpa/inet.h>
+
+int main(void)
+{
+	return 1 != ntohs(1);
+}
-- 
2.11.0

[iproute PATCH 2/2] testsuite: Search kernel config in modules dir also

[Phil Sutter]

At least in Fedora there is no /proc/config.gz but instead
/lib/modules/`uname -r`/config, so use that as a fallback.

Signed-off-by: Phil Sutter <phil@nwl.cc>
---
 testsuite/Makefile | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/testsuite/Makefile b/testsuite/Makefile
index 17568881b8b66..41a3afc92ae98 100644
--- a/testsuite/Makefile
+++ b/testsuite/Makefile
@@ -15,6 +15,12 @@ IPVERS := $(filter-out iproute2/Makefile,$(wildcard iproute2/*))
 
 ifneq (,$(wildcard /proc/config.gz))
 	KENV := $(shell cat /proc/config.gz | gunzip | grep ^CONFIG)
+else
+KVER := $(shell uname -r)
+KCPATH := /lib/modules/${KVER}/config
+ifneq (,$(wildcard ${KCPATH}))
+	KENV := $(shell cat ${KCPATH} | grep ^CONFIG)
+endif
 endif
 
 .PHONY: compile listtests alltests configure $(TESTS)
-- 
2.11.0

Re: [iproute PATCH 1/2] testsuite: skip link show test on big endian systems

[Stephen Hemminger]

On Wed,  8 Feb 2017 00:04:21 +0100
Phil Sutter <phil@nwl.cc> wrote:

> Netlink protocol is in host byte order, so the provided binary netlink
> message buffer being in little endian format will cause the test to fail
> on big endian systems.
> 
> Signed-off-by: Phil Sutter <phil@nwl.cc>

Maybe better to figure out how to generate the files in host order?

Re: [iproute PATCH 1/2] testsuite: skip link show test on big endian systems

[Phil Sutter]

On Tue, Feb 07, 2017 at 03:12:49PM -0800, Stephen Hemminger wrote:
> On Wed,  8 Feb 2017 00:04:21 +0100
> Phil Sutter <phil@nwl.cc> wrote:
> 
> > Netlink protocol is in host byte order, so the provided binary netlink
> > message buffer being in little endian format will cause the test to fail
> > on big endian systems.
> > 
> > Signed-off-by: Phil Sutter <phil@nwl.cc>
> 
> Maybe better to figure out how to generate the files in host order?

Yes, I had thought about that as well. Not sure whether it's worth the
effort though to write a program which constructs the messages and dumps
them into a file for 'ip monitor' to read. I think there's a certain
chance this will eventually test the message builder instead of iproute.
:)

Cheers, Phil