init script can fail when devices are unassigned
Bug #1558485 reported by
Christian Ehrhardt
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
dpdk (Ubuntu) |
Fix Released
|
Medium
|
Christian Ehrhardt |
Bug Description
While usually not the case when the init script runs devices could be unassigned when e.g. a service dpdk restart happens.
In that case the code already does the right things but has still some implementation details that fail afterwards.
Also in that path it still has a remaining tr change o - to _ which is wrong, see bug 1554214.
Both should be fixed, so that in case an admin or experiment unassigned devices and wants to get back to normal by restarting the script would not fail.
Changed in dpdk (Ubuntu): | |
status: | New → Triaged |
importance: | Undecided → Medium |
assignee: | nobody → ChristianEhrhardt (paelzer) |
milestone: | none → ubuntu-16.04 |
To post a comment you must log in.
This bug was fixed in the package dpdk - 2.2.0-0ubuntu6
---------------
dpdk (2.2.0-0ubuntu6) xenial; urgency=medium
* d/dpdk-init fix handling of multiple huge page sizes (LP: #1557532): hugepage- size mountpoints are fix-lpm- use-after- free-and- leak.patch lpm/lpm6 (LP: #1554009): backport- [01-26] backport for stability (LP: #1559981): backport- 27-virtio- fix-restart. patch for (LP: #1559981): dev_start/ stop
- dpdk-init now ensures non-default-
available as well.
- extra mountpoints are only created if requested in dpdk.conf and not yet
available (e.g. by the admin).
* d/dpdk-init fix issues with unassigned devices (LP: #1558485):
- dpdk-init no more checks /sys/.../driver of the device unconditionally
- removed the superfluous tr call in that path
* d/p/ubuntu-
- lpm/lpm6 fix use after free on lpm[6]_create
- lpm/lpm6 fix missing frees of rules_tbl substructure
- lpm/lpm6 fix missing free of lpm due to early exit
- make RTE_LOG messages of the failed allocation unique
* d/p/ubuntu-
- these will be in the following dpdk release and delta can then be dropped
- 26 fixes that do not change api/behaviour but fix serious issues
- 01 d3a274ce app/testpmd: handle SIGINT and SIGTERM
- 02 308df2bf Handle SIGINT and SIGTERM in l3fwd.
- 03 da82ee17 tools: fix unbinding failure handling
- 04 16c1814c tools: support Python 3 in bind script
- 05 bb9f4085 tools: support binding to built-in kernel modules
- 06 6e7caa1a eal/linux: support built-in kernel modules
- 07 86f36ff9 mempool: fix leak when creation fails
- 08 ca67ed28 vhost: fix leak of fds and mmaps
- 09 fa11a8a7 port: fix crash for ring writer nodrop
- 10 04f36690 port: fix crash for ethdev writer nodrop
- 11 c7a4ff80 i40e: fix overflow
- 12 097e920c i40e: fix inverted check for no refcount
- 13 330aa319 i40e: fix VLAN filtering
- 14 9f44dd3d i40e/base: fix missing check for stopped admin queue
- 15 8a880736 i40e/base: fix driver load failure
- 16 7656a546 fm10k: fix VLAN flag in scattered Rx
- 17 c6fb0e55 pcap: fix captured frame length
- 18 6e027237 bonding: fix detach of bonded device
- 19 df3e8ad7 bonding: fix detach of slave devices
- 20 786c990a bonding: copy entire config structure in mode 4
- 21 6698820b bonding: do not ignore multicast in mode 4
- 22 8997a10b bonding: fix active slaves with no primary
- 23 7a7122ed bonding: do not activate slave twice
- 24 2186fff3 bonding: fix crash when no slave device
- 25 c680a4a8 virtio: fix crash in statistics functions
- 26 3b1e3e4e virtio: fix descriptors pointing to the same buffer
* d/p/ubuntu-
- fixing re-initializing the ethdev as openvswitch-dpdk does in the
virtio pmd driver by moving the detection of already being initialized
from virtio_dev_close to virtio_
- this will be in the following dpdk release and delta can then be dropped
* d/rules build with debuginfo (LP: #1560839):
- exporting CFLAGS for all dpdk build processes
- dh_strip...