selftests: forwarding: Import top-level lib.sh through $lib_dir
Subsystems (5)
PATCH 1/1 selftests: forwarding: Import top-level lib.sh through $lib_dir
The commit cited below moved code from net/forwarding/lib.sh to net/lib.sh, and had the former source the latter. This causes issues with selftests from directories other than net/forwarding/, in particular drivers/net/... For those files, the line "source ../lib.sh" would look for lib.sh in the directory above the script file's one, which is incorrect. The way these selftests source net/forwarding/lib.sh is like this:
lib_dir=$(dirname $0)/../../../net/forwarding source $lib_dir/lib.sh
Reuse the variable lib_dir, if set, in net/forwarding/lib.sh to source net/lib.sh as well.
--- tools/testing/selftests/net/forwarding/lib.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/testing/selftests/net/forwarding/lib.sh +1 -1
@@ -38,7 +38,7 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi -source ../lib.sh +source ${lib_dir-.}/../lib.sh ############################################################################## # Sanity checks
Signature · 1 line
2.41.0
Quoted 25 lines · Petr Machata
The commit cited below moved code from
net/forwarding/lib.shtonet/lib.sh, and had the former source the latter. This causes issues with selftests from directories other thannet/forwarding/, in particulardrivers/net/... For those files, the line "source../lib.sh" would look forlib.shin the directory above the script file's one, which is incorrect. The way these selftests sourcenet/forwarding/lib.shis like this:lib_dir=$(dirname $0)/../../../net/forwarding source $lib_dir/lib.shReuse the variable
lib_dir, if set, innet/forwarding/lib.shto sourcenet/lib.shas well.Fixes 25ae948b4478 ("selftests/net: add lib.sh")Cc Hangbin Liu <[email protected]>Signed-off-by Petr Machata <[email protected]>--- tools/testing/selftests/net/forwarding/lib.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)tools/testing/selftests/net/forwarding/lib.sh +0
-0
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 43
Quoted 7 lines
tools/testing/selftests/net/forwarding/lib.sh +1
-1@@ -38,7 +38,7 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi -source ../lib.sh +source ${lib_dir-.}/../lib.sh ############################################################################## # Sanity checks
Hi Petr,
Thanks for the report. However, this doesn't fix the soft link scenario. e.g. The bonding tests tools/testing/selftests/drivers/net/bonding add a soft link net_forwarding_lib.sh and source it directly in dev_addr_lists.sh.
So how about something like:
tools/testing/selftests/net/forwarding/lib.sh +2 -1
@@ -38,7 +38,8 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi -source ../lib.sh +forwarding_dir=$(dirname $(readlink -f $BASH_SOURCE)) +source ${forwarding_dir}/../lib.sh
Thanks
Hangbin
Quoted 13 lines · Hangbin Liu
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 43
Quoted 8 lines · Petr Machata
tools/testing/selftests/net/forwarding/lib.sh +1
-1@@ -38,7 +38,7 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi -source ../lib.sh +source ${lib_dir-.}/../lib.sh ############################################################################## # Sanity checksHi Petr,
Quoted 3 lines
Thanks for the report. However, this doesn't fix the soft link scenario. e.g. The bonding tests
tools/testing/selftests/drivers/net/bondingadd a soft linknet_forwarding_lib.shand source it directly indev_addr_lists.sh.
I see, I didn't realize those exist.
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 42
Quoted 13 lines
So how about something like:
tools/testing/selftests/net/forwarding/lib.sh +2
-1@@ -38,7 +38,8 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi -source ../lib.sh +forwarding_dir=$(dirname $(readlink -f $BASH_SOURCE)) +source ${forwarding_dir}/../lib.sh
Yep, that's gonna work.
I'll pass through our tests and send later this week.
Quoted 32 lines · Petr Machata
Quoted 13 lines · Hangbin Liu
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 43
Quoted 8 lines · Petr Machata
tools/testing/selftests/net/forwarding/lib.sh +1
-1@@ -38,7 +38,7 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi -source ../lib.sh +source ${lib_dir-.}/../lib.sh ############################################################################## # Sanity checksHi Petr,
Quoted 3 lines
Thanks for the report. However, this doesn't fix the soft link scenario. e.g. The bonding tests
tools/testing/selftests/drivers/net/bondingadd a soft linknet_forwarding_lib.shand source it directly indev_addr_lists.sh.I see, I didn't realize those exist.
Quoted 10 lines
So how about something like:
tools/testing/selftests/net/forwarding/lib.sh +0
-0@@ -38,7 +38,8 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi
Quoted 7 lines
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 42
Quoted 3 lines
tools/testing/selftests/net/forwarding/lib.sh +2
-1-source ../lib.sh +forwarding_dir=$(dirname $(readlink -f $BASH_SOURCE)) +source ${forwarding_dir}/../lib.shYep, that's gonna work.
I'll pass through our tests and send later this week.
There is also another related issue which is that generating a test archive using gen_tar for the tests under drivers/net/bonding does not include the new lib.sh. This is similar to the issue reported here:
https://lore.kernel.org/netdev/[email protected]/
/tmp/x# ./run_kselftest.sh
TAP version 13
[...]
# timeout set to 120 # selftests: drivers/net/bonding: dev_addr_lists.sh # ./net_forwarding_lib.sh: line 41: ../lib.sh: No such file or directory # TEST: bonding cleanup mode active-backup [ OK ] # TEST: bonding cleanup mode 802.3ad [ OK ] # TEST: bonding LACPDU multicast address to slave (from bond down) [ OK ] # TEST: bonding LACPDU multicast address to slave (from bond up) [ OK ] ok 4 selftests: drivers/net/bonding: dev_addr_lists.sh [...]
Quoted 49 lines · Benjamin Poirier
Quoted 32 lines · Petr Machata
Quoted 13 lines · Hangbin Liu
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 43
Quoted 8 lines · Petr Machata
tools/testing/selftests/net/forwarding/lib.sh +1
-1@@ -38,7 +38,7 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi -source ../lib.sh +source ${lib_dir-.}/../lib.sh ############################################################################## # Sanity checksHi Petr,
Quoted 3 lines
Thanks for the report. However, this doesn't fix the soft link scenario. e.g. The bonding tests
tools/testing/selftests/drivers/net/bondingadd a soft linknet_forwarding_lib.shand source it directly indev_addr_lists.sh.I see, I didn't realize those exist.
Quoted 10 lines
So how about something like:
tools/testing/selftests/net/forwarding/lib.sh +0
-0@@ -38,7 +38,8 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fiQuoted 7 lines
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 42
Quoted 3 lines
tools/testing/selftests/net/forwarding/lib.sh +2
-1-source ../lib.sh +forwarding_dir=$(dirname $(readlink -f $BASH_SOURCE)) +source ${forwarding_dir}/../lib.shYep, that's gonna work.
I'll pass through our tests and send later this week.There is also another related issue which is that generating a test archive using
gen_tarfor the tests underdrivers/net/bondingdoes not include the newlib.sh. This is similar to the issue reported here:
https://lore.kernel.org/netdev/[email protected]//tmp/x# ./run_kselftest.sh
TAP version 13
[...]
Quoted 9 lines
# timeout set to 120 # selftests: drivers/net/bonding: dev_addr_lists.sh # ./net_forwarding_lib.sh: line 41: ../lib.sh: No such file or directory # TEST: bonding cleanup mode active-backup [ OK ] # TEST: bonding cleanup mode 802.3ad [ OK ] # TEST: bonding LACPDU multicast address to slave (from bond down) [ OK ] # TEST: bonding LACPDU multicast address to slave (from bond up) [ OK ] ok 4 selftests: drivers/net/bonding: dev_addr_lists.sh [...]
Hmm.. Is it possible to write a rule in the Makefile to create the net/ and net/forwarding folder so we can source the relative path directly. e.g.
]# tree
.
├── drivers │ └── net │ └── bonding │ ├── bond-arp-interval-causes-panic.sh │ ├── ... │ └── settings ├── kselftest │ ├── module.sh │ ├── prefix.pl │ └── runner.sh ├── kselftest-list.txt ├── net │ ├── forwarding │ │ └── lib.sh │ └── lib.sh └── run_kselftest.sh
Thanks
Hangbin
Quoted 2 lines · Hangbin Liu
Quoted 1 line · Benjamin Poirier
On 2023-12-12 18:22 +0100, Petr Machata wrote:
[...]
Quoted 23 lines
Quoted 8 lines
There is also another related issue which is that generating a test archive using
gen_tarfor the tests underdrivers/net/bondingdoes not include the newlib.sh. This is similar to the issue reported here:
https://lore.kernel.org/netdev/[email protected]//tmp/x# ./run_kselftest.sh
TAP version 13
[...]Quoted 9 lines
# timeout set to 120 # selftests: drivers/net/bonding: dev_addr_lists.sh # ./net_forwarding_lib.sh: line 41: ../lib.sh: No such file or directory # TEST: bonding cleanup mode active-backup [ OK ] # TEST: bonding cleanup mode 802.3ad [ OK ] # TEST: bonding LACPDU multicast address to slave (from bond down) [ OK ] # TEST: bonding LACPDU multicast address to slave (from bond up) [ OK ] ok 4 selftests: drivers/net/bonding: dev_addr_lists.sh [...]Hmm.. Is it possible to write a rule in the
Makefileto create thenet/and net/forwarding folder so we can source the relative path directly. e.g.]# tree
.
Quoted 16 lines
├── drivers │ └── net │ └── bonding │ ├── bond-arp-interval-causes-panic.sh │ ├── ... │ └── settings ├── kselftest │ ├── module.sh │ ├── prefix.pl │ └── runner.sh ├── kselftest-list.txt ├── net │ ├── forwarding │ │ └── lib.sh │ └── lib.sh └── run_kselftest.sh
That sounds like a good idea. I started to work on that approach but I'm missing recursive inclusion. For instance
cd tools/testing/selftests
make install TARGETS="drivers/net/bonding" ./kselftest_install/run_kselftest.sh -t drivers/net/bonding:dev_addr_lists.sh
includes net/forwarding/lib.sh but is missing net/lib.sh. I feel that my 'make' skills are rusty but I guess that with enough make code, it could be done. A workaround is simply to manually list the transitive dependencies in TEST_SH_LIBS:
TEST_SH_LIBS := \
- net/forwarding/lib.sh
+ net/forwarding/lib.sh \
+ net/lib.sh
I only converted a few files to validate that the approach is viable. I used the following tests:drivers/net/bonding/dev_addr_lists.sh net/test_vxlan_vnifiltering.shnet/forwarding/pedit_ip.sh
Let me know what you think.
--- tools/testing/selftests/Makefile | 5 ++++- tools/testing/selftests/drivers/net/bonding/Makefile | 6 ++++-- .../testing/selftests/drivers/net/bonding/dev_addr_lists.sh | 2 +- .../selftests/drivers/net/bonding/net_forwarding_lib.sh | 1 - tools/testing/selftests/lib.mk | 3 +++ tools/testing/selftests/net/forwarding/Makefile | 3 +++ tools/testing/selftests/net/forwarding/lib.sh | 3 ++- 7 files changed, 17 insertions(+), 6 deletions(-) delete mode 120000 tools/testing/selftests/drivers/net/bonding/net_forwarding_lib.sh
tools/testing/selftests/Makefile +4 -1
@@ -256,7 +256,10 @@ ifdef INSTALL_PATH @ret=1; \ for TARGET in $(TARGETS); do \ BUILD_TARGET=$$BUILD/$$TARGET; \ - $(MAKE) OUTPUT=$$BUILD_TARGET -C $$TARGET INSTALL_PATH=$(INSTALL_PATH)/$$TARGET install \ + $(MAKE) install OUTPUT=$$BUILD_TARGET -C $$TARGET \ + INSTALL_PATH=$(INSTALL_PATH)/$$TARGET \ + SH_LIBS_PATH=$(INSTALL_PATH) \ + BASE_PATH=$(shell pwd) \ O=$(abs_objtree) \ $(if $(FORCE_TARGETS),|| exit); \ ret=$$((ret * $$?)); \
tools/testing/selftests/drivers/net/bonding/Makefile +4 -2
@@ -15,7 +15,9 @@ TEST_PROGS := \ TEST_FILES := \ lag_lib.sh \ bond_topo_2d1c.sh \ - bond_topo_3d1c.sh \ - net_forwarding_lib.sh + bond_topo_3d1c.sh + +TEST_SH_LIBS := \ + net/forwarding/lib.sh include ../../../lib.mk
tools/testing/selftests/drivers/net/bonding/dev_addr_lists.sh +1 -1
@@ -14,7 +14,7 @@ ALL_TESTS=" REQUIRE_MZ=no NUM_NETIFS=0 lib_dir=$(dirname "$0") -source "$lib_dir"/net_forwarding_lib.sh +source "$lib_dir"/../../../net/forwarding/lib.sh source "$lib_dir"/lag_lib.sh
tools/testing/selftests/drivers/net/bonding/net_forwarding_lib.sh deleted +0 -1
@@ -1 +0,0 @@ -../../../net/forwarding/lib.sh
tools/testing/selftests/lib.mk +3 -0
@@ -137,6 +137,9 @@ endef install: all ifdef INSTALL_PATH $(INSTALL_RULE) + $(if $(TEST_SH_LIBS), \ + cd $(BASE_PATH) && rsync -aR $(TEST_SH_LIBS) $(SH_LIBS_PATH)/ \ + ) else $(error Error: set INSTALL_PATH to use install) endif
tools/testing/selftests/net/forwarding/Makefile +3 -0
@@ -128,4 +128,7 @@ TEST_PROGS_EXTENDED := devlink_lib.sh \ sch_tbf_etsprio.sh \ tc_common.sh +TEST_SH_LIBS := \ + net/lib.sh + include ../../lib.mk
tools/testing/selftests/net/forwarding/lib.sh +2 -1
@@ -38,7 +38,8 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi -source ../lib.sh +libdir=$(dirname "$(readlink -f "$BASH_SOURCE")") +source "$libdir"/../lib.sh ############################################################################## # Sanity checks
Signature · 1 line
2.43.0
Hi Benjamin,
Quoted 34 lines · Benjamin Poirier
Quoted 5 lines
Hmm.. Is it possible to write a rule in the
Makefileto create thenet/and net/forwarding folder so we can source the relative path directly. e.g.]# tree
.Quoted 16 lines
├── drivers │ └── net │ └── bonding │ ├── bond-arp-interval-causes-panic.sh │ ├── ... │ └── settings ├── kselftest │ ├── module.sh │ ├── prefix.pl │ └── runner.sh ├── kselftest-list.txt ├── net │ ├── forwarding │ │ └── lib.sh │ └── lib.sh └── run_kselftest.shThat sounds like a good idea. I started to work on that approach but I'm missing recursive inclusion. For instance
cd
tools/testing/selftests
make install TARGETS="drivers/net/bonding"./kselftest_install/run_kselftest.sh-t drivers/net/bonding:dev_addr_lists.shincludes
net/forwarding/lib.shbut is missingnet/lib.sh. I feel that my 'make' skills are rusty but I guess that with enough make code, it could be done. A workaround is simply to manually list the transitive dependencies inTEST_SH_LIBS:
TEST_SH_LIBS:= \
Review
Quoted 3 lines
(quoted diff) +2
-1- net/forwarding/lib.sh + net/forwarding/lib.sh \ + net/lib.sh
Yes, this makes the user need to make sure all the recursive inclusions listed here. A little inconvenient. But I "make" skill is worse than you...
Quoted 8 lines
I only converted a few files to validate that the approach is viable. I used the following tests:
drivers/net/bonding/dev_addr_lists.shnet/test_vxlan_vnifiltering.shnet/forwarding/pedit_ip.shLet me know what you think.
Thanks! This works for me.
Cheers
Hangbin
Quoted 51 lines · Hangbin Liu
Hi Benjamin,
Quoted 34 lines · Benjamin Poirier
Quoted 5 lines
Hmm.. Is it possible to write a rule in the
Makefileto create thenet/and net/forwarding folder so we can source the relative path directly. e.g.]# tree
.Quoted 16 lines
├── drivers │ └── net │ └── bonding │ ├── bond-arp-interval-causes-panic.sh │ ├── ... │ └── settings ├── kselftest │ ├── module.sh │ ├── prefix.pl │ └── runner.sh ├── kselftest-list.txt ├── net │ ├── forwarding │ │ └── lib.sh │ └── lib.sh └── run_kselftest.shThat sounds like a good idea. I started to work on that approach but I'm missing recursive inclusion. For instance
cd
tools/testing/selftests
make install TARGETS="drivers/net/bonding"./kselftest_install/run_kselftest.sh-t drivers/net/bonding:dev_addr_lists.shincludes
net/forwarding/lib.shbut is missingnet/lib.sh. I feel that my 'make' skills are rusty but I guess that with enough make code, it could be done. A workaround is simply to manually list the transitive dependencies inTEST_SH_LIBS:
TEST_SH_LIBS:= \Review
Quoted 3 lines
(quoted diff) +2
-1- net/forwarding/lib.sh + net/forwarding/lib.sh \ + net/lib.shYes, this makes the user need to make sure all the recursive inclusions listed here. A little inconvenient. But I "make" skill is worse than you...
Quoted 7 lines
I only converted a few files to validate that the approach is viable. I used the following tests:
drivers/net/bonding/dev_addr_lists.shnet/test_vxlan_vnifiltering.shnet/forwarding/pedit_ip.sh
Quoted 3 lines
Quoted 1 line
Let me know what you think.
Thanks! This works for me.
I started to make the adjustments to all the tests but I got stuck on the dsa tests. The problem is that the tests which are symlinked (like bridge_locked_port.sh) expect to source lib.sh (net/forwarding/lib.sh) from the same directory. That lib.sh then expects to source net/lib.sh from the parent directory. Because rsync --copy-unsafe-links is used, all those links become regular files after export so we can't rely on readlink -f.
Honestly, given how the dsa tests are organized, I don't see a clean way to support these tests without error after commit 25ae948b4478 ("selftests/net: add lib.sh").
The only way that I see to run these tests via runner.sh is by using a
command like:
make -C tools/testing/selftests install TARGETS="drivers/net/dsa"
KSELFTEST_BRIDGE_LOCKED_PORT_SH_ARGS="swp1 swp2 swp3 swp4" KSELFTEST_BRIDGE_MDB_SH_ARGS="swp1 swp2 swp3 swp4" KSELFTEST_BRIDGE_MLD_SH_ARGS="swp1 swp2 swp3 swp4" KSELFTEST_BRIDGE_VLAN_AWARE_SH_ARGS="swp1 swp2 swp3 swp4" KSELFTEST_BRIDGE_VLAN_MCAST_SH_ARGS="swp1 swp2 swp3 swp4" KSELFTEST_BRIDGE_VLAN_UNAWARE_SH_ARGS="swp1 swp2 swp3 swp4" KSELFTEST_LOCAL_TERMINATION_SH_ARGS="swp1 swp2 swp3 swp4" KSELFTEST_NO_FORWARDING_SH_ARGS="swp1 swp2 swp3 swp4" KSELFTEST_TC_ACTIONS_SH_ARGS="swp1 swp2 swp3 swp4" KSELFTEST_TEST_BRIDGE_FDB_STRESS_SH_ARGS="swp1 swp2 swp3 swp4" tools/testing/selftests/kselftest_install/run_kselftest.sh
This is very cumbersome so it makes me question the value of drivers/net/dsa/Makefile.
Patrice, Vladimir, Martin, how do you run the dsa tests? Could we revert 6ecf206d602f ("selftests: net: dsa: Add a Makefile which installs the selftests")?
Do you have other suggestions to avoid the following error about lib.sh:
tools/testing/selftests# make install TARGETS="drivers/net/dsa" [...]
tools/testing/selftests# ./kselftest_install/run_kselftest.sh TAP version 13
1..10
# timeout set to 45 # selftests: drivers/net/dsa: bridge_locked_port.sh # lib.sh: line 41: ../lib.sh: No such file or directory [...]
Quoted 11 lines · Benjamin Poirier
I started to make the adjustments to all the tests but I got stuck on the dsa tests. The problem is that the tests which are symlinked (like
bridge_locked_port.sh) expect to sourcelib.sh(net/forwarding/lib.sh) from the same directory. Thatlib.shthen expects to sourcenet/lib.shfrom the parent directory. Becausersync --copy-unsafe-linksis used, all those links become regular files after export so we can't rely onreadlink -f.Honestly, given how the dsa tests are organized, I don't see a clean way to support these tests without error after commit
25ae948b4478("selftests/net: addlib.sh").
No worry, the last way is just make net/forwarding/lib.sh not source net/lib.sh :) Although this would make us copy a lot functions from net/forwarding/lib.sh to source net/lib.sh. So before that, let's try if we can
Move all the dsa tests back to net/forwarding (actually only needed for test_bridge_fdb_stress.sh). And add a forwarding.config.dsa.sample. Maybe a test list for dsa testing. But with this way the dsa testing will not able to run via run_kselftest.sh.
Or, we can remove the symlinks, and add the dsa tests with exec the relative forwarding path's tests directly. e.g.
### in das test folder
$ cat bridge_mld.sh #!/bin/bash ../../../net/forwarding/bridge_mld.sh
$ cat Makefile # SPDX-License-Identifier: GPL-2.0+ OR MIT
TEST_PROGS = \
bridge_mld.sh
TEST_SH_LIBS := \
net/lib.sh \
net/forwarding/lib.sh \
net/forwarding/bridge_mld.shTEST_FILES := forwarding.config
include ../../../lib.mk
### for net/forwarding/lib.sh looks we need source the ${PWD}/forwarding.config if run via run_kselftest.sh
if [[ -f ${PWD}/forwarding.config ]]; then
source "$PWD/forwarding.config"
elif [[ -f $relative_path/forwarding.config ]]; then
source "$relative_path/forwarding.config"
fi
What do you think?
Thanks
Hangbin
Quoted 48 lines · Hangbin Liu
Quoted 11 lines · Benjamin Poirier
I started to make the adjustments to all the tests but I got stuck on the dsa tests. The problem is that the tests which are symlinked (like
bridge_locked_port.sh) expect to sourcelib.sh(net/forwarding/lib.sh) from the same directory. Thatlib.shthen expects to sourcenet/lib.shfrom the parent directory. Becausersync --copy-unsafe-linksis used, all those links become regular files after export so we can't rely onreadlink -f.Honestly, given how the dsa tests are organized, I don't see a clean way to support these tests without error after commit
25ae948b4478("selftests/net: addlib.sh").No worry, the last way is just make
net/forwarding/lib.shnot sourcenet/lib.sh:) Although this would make us copy a lot functions fromnet/forwarding/lib.shto sourcenet/lib.sh. So before that, let's try if we canMove all the dsa tests back to net/forwarding (actually only needed for
test_bridge_fdb_stress.sh). And add a forwarding.config.dsa.sample. Maybe a test list for dsa testing. But with this way the dsa testing will not able to run viarun_kselftest.sh.Or, we can remove the symlinks, and add the dsa tests with exec the relative forwarding path's tests directly. e.g.
### in das test folder
$ cat bridge_mld.sh #!/bin/bash ../../../net/forwarding/bridge_mld.sh$ cat Makefile # SPDX-License-Identifier: GPL-2.0+ OR MIT
TEST_PROGS= \
bridge_mld.shTEST_SH_LIBS := \ net/lib.sh \ net/forwarding/lib.sh \ net/forwarding/bridge_mld.sh
TEST_FILES:= forwarding.configinclude
../../../lib.mk### for
net/forwarding/lib.shlooks we need source the${PWD}/forwarding.configif run viarun_kselftest.sh
Quoted 7 lines
if [[ -f
${PWD}/forwarding.config]]; then
source "$PWD/forwarding.config"
elif [[ -f$relative_path/forwarding.config]]; then
source "$relative_path/forwarding.config"
fiWhat do you think?
That's a good idea. Thank you for your suggestion. I made a few adjustment from what you show and the result seems to work so far.
I'm working on the full patchset. If there is no problem, I'll post it next week.
Quoted 1 line · Benjamin Poirier
Patrice, Vladimir, Martin, how do you run the dsa tests?
Not how they're supposed to, apparently.
I used to rsync the "selftests" folder to the device under test, go to the drivers/net/dsa directory, and ./ the test from there. That is absolutely sufficient without all the "make kselftest" / run_kselftest.sh overhead which I don't need, but apparently that broken now too.
I don't have a strong objection against eliminating the symlinks. They were just a handy way of filtering those tests from net/forwarding/ which were relevant to DSA, and just ignore the test.
What might turn out to be problematic is DSA's forwarding.config, where we actually rely on the STABLE_MAC_ADDRS option for some tests to pass. I'm not actually sure what is the "recommended" way of deploying a custom forwarding.config file anyway.
Quoted 10 lines · Vladimir Oltean
Quoted 1 line · Benjamin Poirier
Patrice, Vladimir, Martin, how do you run the dsa tests?
Not how they're supposed to, apparently.
I used to rsync the "selftests" folder to the device under test, go to the
drivers/net/dsadirectory, and ./ the test from there. That is absolutely sufficient without all the "make kselftest" /run_kselftest.shoverhead which I don't need, but apparently that broken now too.
Quoted 8 lines
I don't have a strong objection against eliminating the symlinks. They were just a handy way of filtering those tests from
net/forwarding/which were relevant to DSA, and just ignore the test.What might turn out to be problematic is DSA's forwarding.config, where we actually rely on the
STABLE_MAC_ADDRSoption for some tests to pass. I'm not actually sure what is the "recommended" way of deploying a custom forwarding.config file anyway.
Thank you for sharing that info. There are so many different ways to run the selftests, there is no "one true way". Thanks to Hangbin's suggestion earlier in the thread I found a way to make things work for the dsa tests after commit 25ae948b4478 ("selftests/net: add lib.sh") by adding a wrapper script that sources another test; see the RFC that I just posted:
https://lore.kernel.org/netdev/[email protected]/T/#t
drivers/net/dsa will still have links to the net/forwarding/ tests so today's functionality should be preserved, including forwarding.config.
Quoted 48 lines · Benjamin Poirier
Quoted 32 lines · Petr Machata
Quoted 13 lines · Hangbin Liu
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 43
Quoted 8 lines · Petr Machata
tools/testing/selftests/net/forwarding/lib.sh +1
-1@@ -38,7 +38,7 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi -source ../lib.sh +source ${lib_dir-.}/../lib.sh ############################################################################## # Sanity checksHi Petr,
Quoted 3 lines
Thanks for the report. However, this doesn't fix the soft link scenario. e.g. The bonding tests
tools/testing/selftests/drivers/net/bondingadd a soft linknet_forwarding_lib.shand source it directly indev_addr_lists.sh.I see, I didn't realize those exist.
Quoted 10 lines
So how about something like:
tools/testing/selftests/net/forwarding/lib.sh +0
-0@@ -38,7 +38,8 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fiQuoted 6 lines
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 42
Quoted 3 lines
tools/testing/selftests/net/forwarding/lib.sh +2
-1-source ../lib.sh +forwarding_dir=$(dirname $(readlink -f $BASH_SOURCE)) +source ${forwarding_dir}/../lib.shYep, that's gonna work.
I'll pass through our tests and send later this week.There is also another related issue which is that generating a test archive using
gen_tarfor the tests underdrivers/net/bondingdoes not include the newlib.sh. This is similar to the issue reported here:
https://lore.kernel.org/netdev/[email protected]//tmp/x# ./run_kselftest.sh
TAP version 13
[...]
Quoted 9 lines
# timeout set to 120 # selftests: drivers/net/bonding: dev_addr_lists.sh # ./net_forwarding_lib.sh: line 41: ../lib.sh: No such file or directory # TEST: bonding cleanup mode active-backup [ OK ] # TEST: bonding cleanup mode 802.3ad [ OK ] # TEST: bonding LACPDU multicast address to slave (from bond down) [ OK ] # TEST: bonding LACPDU multicast address to slave (from bond up) [ OK ] ok 4 selftests: drivers/net/bonding: dev_addr_lists.sh [...]
The issue is that the symlink becomes a text file during install, and readlink on that file then becomes a nop. Maybe the bonding tests should include net/forwarding/lib.sh through a relative path like other tests in drivers/, instead of this symlink business?
Quoted 60 lines · Petr Machata
Quoted 48 lines · Benjamin Poirier
Quoted 32 lines · Petr Machata
Quoted 13 lines · Hangbin Liu
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 43
Quoted 8 lines · Petr Machata
tools/testing/selftests/net/forwarding/lib.sh +1
-1@@ -38,7 +38,7 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fi -source ../lib.sh +source ${lib_dir-.}/../lib.sh ############################################################################## # Sanity checksHi Petr,
Quoted 3 lines
Thanks for the report. However, this doesn't fix the soft link scenario. e.g. The bonding tests
tools/testing/selftests/drivers/net/bondingadd a soft linknet_forwarding_lib.shand source it directly indev_addr_lists.sh.I see, I didn't realize those exist.
Quoted 10 lines
So how about something like:
tools/testing/selftests/net/forwarding/lib.sh +0
-0@@ -38,7 +38,8 @@ if [[ -f $relative_path/forwarding.config ]]; then source "$relative_path/forwarding.config" fiQuoted 6 lines
tools/testing/selftests/net/forwarding/lib.sh if [[ -f $relative_path/forwarding.config ]]; then · line 42
Quoted 3 lines
tools/testing/selftests/net/forwarding/lib.sh +2
-1-source ../lib.sh +forwarding_dir=$(dirname $(readlink -f $BASH_SOURCE)) +source ${forwarding_dir}/../lib.shYep, that's gonna work.
I'll pass through our tests and send later this week.There is also another related issue which is that generating a test archive using
gen_tarfor the tests underdrivers/net/bondingdoes not include the newlib.sh. This is similar to the issue reported here:
https://lore.kernel.org/netdev/[email protected]//tmp/x# ./run_kselftest.sh
TAP version 13
[...]Quoted 9 lines
# timeout set to 120 # selftests: drivers/net/bonding: dev_addr_lists.sh # ./net_forwarding_lib.sh: line 41: ../lib.sh: No such file or directory # TEST: bonding cleanup mode active-backup [ OK ] # TEST: bonding cleanup mode 802.3ad [ OK ] # TEST: bonding LACPDU multicast address to slave (from bond down) [ OK ] # TEST: bonding LACPDU multicast address to slave (from bond up) [ OK ] ok 4 selftests: drivers/net/bonding: dev_addr_lists.sh [...]
Quoted 4 lines
The issue is that the symlink becomes a text file during install, and readlink on that file then becomes a nop. Maybe the bonding tests should include
net/forwarding/lib.shthrough a relative path like other tests indrivers/, instead of this symlink business?
Or wait, the goal is for make install in the bonding directory to produce a working install, right?
Hmm.
This worked before exactly because the symlink became a text file. It does not work in general, but it did work for bonding.
I don't have ideas how to solve this elegantly honestly.
Maybe have bonding tests set $net_lib_dir and forwarding/lib.sh then source net/lib.sh through that path if set; have bonding symlink net/lib.sh in addition to forwarding/lib.sh, and set net_lib=.? It's ugly as sin, but... should work?
Hmm, maybe we could side-step the issue? I suspect that vast majority of what bonding uses are just generic helpers. log_test, check_err, that sort of stuff. Unless I missed something, all of them set NUM_NETIFS=0. Those things could all be in the generic net/lib.sh. So long-term it might be possible for bonding to do the trick with symlinking, except with just net/lib.sh, not both libs.
I think that most of forwarding/lib.sh actually belongs to net/lib.sh. We reinvent a lot of that functionality in various net/ tests, because, presumably, people find it odd to source forwarding/lib.sh. If it all lived in net/, we could reuse all these tools instead of cut'n'pasting them from one test to the other. Stuff like the mcast_packet_test, start/stop_traffic, etc., would probably stay in forwarding.
So that's long-term. And short-term we can live with the ugly-as-sin workaround that I propose?
Ideas? Comments?
Quoted 14 lines · Petr Machata
Hmm, maybe we could side-step the issue? I suspect that vast majority of what bonding uses are just generic helpers.
log_test,check_err, that sort of stuff. Unless I missed something, all of them set NUM_NETIFS=0. Those things could all be in the genericnet/lib.sh. So long-term it might be possible for bonding to do the trick with symlinking, except with justnet/lib.sh, not both libs.I think that most of
forwarding/lib.shactually belongs tonet/lib.sh. We reinvent a lot of that functionality in variousnet/tests, because, presumably, people find it odd to sourceforwarding/lib.sh. If it all lived innet/, we could reuse all these tools instead of cut'n'pasting them from one test to the other. Stuff like themcast_packet_test, start/stop_traffic, etc., would probably stay in forwarding.
Quoted 2 lines
So that's long-term. And short-term we can live with the ugly-as-sin workaround that I propose?
For bonding only I think this is a good resolution. For other driver tests that may still need to include forwarding/lib.sh. I think the way Benjamin suggested is a good choice.
Thanks
Hangbin