| Message ID | 20230613215440.2465708-10-vladimir.oltean@nxp.com |
|---|---|
| State | Handled Elsewhere |
| Headers | show |
| Series | Improve the taprio qdisc's relationship with its children | expand |
On 13/06/2023 18:54, Vladimir Oltean wrote: > The reason behind commit af7b29b1deaa ("Revert "net/sched: taprio: make > qdisc_leaf() see the per-netdev-queue pfifo child qdiscs"") was that the > patch it reverted caused a crash when attaching a CBS shaper to one of > the taprio classes. Prevent that from happening again by adding a test > case for it, which now passes correctly in both offload and software > modes. > > Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Other than the comment below, Reviewed-by: Pedro Tammela <pctammela@mojatatu.com> > --- > v1->v2: patch is new > > .../tc-testing/tc-tests/qdiscs/taprio.json | 50 +++++++++++++++++++ > 1 file changed, 50 insertions(+) > > diff --git a/tools/testing/selftests/tc-testing/tc-tests/qdiscs/taprio.json b/tools/testing/selftests/tc-testing/tc-tests/qdiscs/taprio.json > index 58d4d97f4499..47692335bcf1 100644 > --- a/tools/testing/selftests/tc-testing/tc-tests/qdiscs/taprio.json > +++ b/tools/testing/selftests/tc-testing/tc-tests/qdiscs/taprio.json > @@ -179,5 +179,55 @@ > "$TC qdisc del dev $ETH root", > "echo \"1\" > /sys/bus/netdevsim/del_device" > ] > + }, > + { > + "id": "a7bf", > + "name": "Graft cbs as child of software taprio", > + "category": [ > + "qdisc", > + "taprio", > + "cbs" > + ], > + "plugins": { > + "requires": "nsPlugin" > + }, > + "setup": [ > + "echo \"1 1 8\" > /sys/bus/netdevsim/new_device", > + "$TC qdisc replace dev $ETH handle 8001: parent root stab overhead 24 taprio num_tc 8 map 0 1 2 3 4 5 6 7 queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 base-time 0 sched-entry S ff 20000000 clockid CLOCK_TAI" > + ], > + "cmdUnderTest": "$TC qdisc replace dev $ETH handle 8002: parent 8001:8 cbs idleslope 20000 sendslope -980000 hicredit 30 locredit -1470", > + "expExitCode": "0", > + "verifyCmd": "$TC -d qdisc show dev $ETH", > + "matchPattern": "qdisc cbs 8002: parent 8001:8 hicredit 30 locredit -1470 sendslope -980000 idleslope 20000 offload 0", > + "matchCount": "1", > + "teardown": [ > + "$TC qdisc del dev $ETH root", > + "echo \"1\" > /sys/bus/netdevsim/del_device" > + ] > + }, > + { > + "id": "6a83", > + "name": "Graft cbs as child of offloaded taprio", > + "category": [ > + "qdisc", > + "taprio", > + "cbs" > + ], > + "plugins": { > + "requires": "nsPlugin" > + }, > + "setup": [ > + "echo \"1 1 8\" > /sys/bus/netdevsim/new_device", > + "$TC qdisc replace dev $ETH handle 8001: parent root stab overhead 24 taprio num_tc 8 map 0 1 2 3 4 5 6 7 queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 base-time 0 sched-entry S ff 20000000 flags 0x2" > + ], > + "cmdUnderTest": "$TC qdisc replace dev $ETH handle 8002: parent 8001:8 cbs idleslope 20000 sendslope -980000 hicredit 30 locredit -1470", > + "expExitCode": "0", > + "verifyCmd": "$TC -d qdisc show dev $ETH", > + "matchPattern": "qdisc cbs 8002: parent 8001:8 hicredit 30 locredit -1470 sendslope -980000 idleslope 20000 offload 0", Seems like this test is missing the 'refcnt 2' in the match pattern > + "matchCount": "1", > + "teardown": [ > + "$TC qdisc del dev $ETH root", > + "echo \"1\" > /sys/bus/netdevsim/del_device" > + ] > } > ]
On Wed, Jun 14, 2023 at 01:45:42PM -0300, Pedro Tammela wrote: > > + "cmdUnderTest": "$TC qdisc replace dev $ETH handle 8002: parent 8001:8 cbs idleslope 20000 sendslope -980000 hicredit 30 locredit -1470", > > + "expExitCode": "0", > > + "verifyCmd": "$TC -d qdisc show dev $ETH", > > + "matchPattern": "qdisc cbs 8002: parent 8001:8 hicredit 30 locredit -1470 sendslope -980000 idleslope 20000 offload 0", > > Seems like this test is missing the 'refcnt 2' in the match pattern Makes sense. This is consistent with the idea of my patch set, which is that in offloaded taprio mode, each child Qdisc has a refcount elevated by the fact that it's attached to a netdev TX queue (hence the 2 here). I had copied this expected output from the "Graft cbs as child of software taprio" test a7bf (not sure why I didn't catch the failure), but there, the expected refcount of child Qdiscs is 1 (and thus, not shown).
diff --git a/tools/testing/selftests/tc-testing/tc-tests/qdiscs/taprio.json b/tools/testing/selftests/tc-testing/tc-tests/qdiscs/taprio.json index 58d4d97f4499..47692335bcf1 100644 --- a/tools/testing/selftests/tc-testing/tc-tests/qdiscs/taprio.json +++ b/tools/testing/selftests/tc-testing/tc-tests/qdiscs/taprio.json @@ -179,5 +179,55 @@ "$TC qdisc del dev $ETH root", "echo \"1\" > /sys/bus/netdevsim/del_device" ] + }, + { + "id": "a7bf", + "name": "Graft cbs as child of software taprio", + "category": [ + "qdisc", + "taprio", + "cbs" + ], + "plugins": { + "requires": "nsPlugin" + }, + "setup": [ + "echo \"1 1 8\" > /sys/bus/netdevsim/new_device", + "$TC qdisc replace dev $ETH handle 8001: parent root stab overhead 24 taprio num_tc 8 map 0 1 2 3 4 5 6 7 queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 base-time 0 sched-entry S ff 20000000 clockid CLOCK_TAI" + ], + "cmdUnderTest": "$TC qdisc replace dev $ETH handle 8002: parent 8001:8 cbs idleslope 20000 sendslope -980000 hicredit 30 locredit -1470", + "expExitCode": "0", + "verifyCmd": "$TC -d qdisc show dev $ETH", + "matchPattern": "qdisc cbs 8002: parent 8001:8 hicredit 30 locredit -1470 sendslope -980000 idleslope 20000 offload 0", + "matchCount": "1", + "teardown": [ + "$TC qdisc del dev $ETH root", + "echo \"1\" > /sys/bus/netdevsim/del_device" + ] + }, + { + "id": "6a83", + "name": "Graft cbs as child of offloaded taprio", + "category": [ + "qdisc", + "taprio", + "cbs" + ], + "plugins": { + "requires": "nsPlugin" + }, + "setup": [ + "echo \"1 1 8\" > /sys/bus/netdevsim/new_device", + "$TC qdisc replace dev $ETH handle 8001: parent root stab overhead 24 taprio num_tc 8 map 0 1 2 3 4 5 6 7 queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 base-time 0 sched-entry S ff 20000000 flags 0x2" + ], + "cmdUnderTest": "$TC qdisc replace dev $ETH handle 8002: parent 8001:8 cbs idleslope 20000 sendslope -980000 hicredit 30 locredit -1470", + "expExitCode": "0", + "verifyCmd": "$TC -d qdisc show dev $ETH", + "matchPattern": "qdisc cbs 8002: parent 8001:8 hicredit 30 locredit -1470 sendslope -980000 idleslope 20000 offload 0", + "matchCount": "1", + "teardown": [ + "$TC qdisc del dev $ETH root", + "echo \"1\" > /sys/bus/netdevsim/del_device" + ] } ]
The reason behind commit af7b29b1deaa ("Revert "net/sched: taprio: make qdisc_leaf() see the per-netdev-queue pfifo child qdiscs"") was that the patch it reverted caused a crash when attaching a CBS shaper to one of the taprio classes. Prevent that from happening again by adding a test case for it, which now passes correctly in both offload and software modes. Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> --- v1->v2: patch is new .../tc-testing/tc-tests/qdiscs/taprio.json | 50 +++++++++++++++++++ 1 file changed, 50 insertions(+)