Skip to contentSkip to conversation

eal: ctrl thread calculation assumes 1:1 mapping between cpu and lcore id

dpdk-dev [email protected] 1 participant 2 messages Jul 16, 2019 – Jul 29, 2019

lore
Share & tools
lore
#0 lore raw

https://bugs.dpdk.org/show_bug.cgi?id=323

Bug ID: 323
Summary: eal: ctrl thread calculation assumes 1:1 mapping
between cpu and lcore id
Product: DPDK
Version: 18.11
Hardware: All
OS: All
Status: UNCONFIRMED
Severity: normal
Priority: Normal
Component: core
Assignee: [email protected]
Reporter: [email protected]
Target Milestone: ---

ctrl thread calculation assumes 1:1 mapping between cpu and lcore id. This is not the case when --lcores and the optional cpuset mask is used. The calculation introduced in c3568ea376700 (and 18.11.2 23a9f69ed3d9) assumes a 1:1 mapping between cpu and lcore id. This is not the case when --lcores and the optional cpuset mask is used.

The calculation fails when --lcores and cpuset mask is set so cpu and lcore no longer maps 1:1.

The code silently started on the wrong physical thread with following eal arguments(if not cpusets are used to limit the thread affinity):

--master-lcore 0 --lcores (0,19)@(19,1,2,3)

If cpusets are used(and the calculated thread is outside the cpuset) eal fails to start with the following output:

EAL: Failed to create thread for interrupt handling
EAL: FATAL: Cannot init interrupt-handling thread
EAL: Cannot init interrupt-handling thread
EAL: Error - exiting with code: 1
Cause: Invalid EAL arguments: -1 - Unknown error -1

----------
NOTE: A resubmit of 322 as suggested by Thomas Monjalon. Bug report 322 suffered by a UTF8 error in the bug tracker.

Signature · 2 lines
You are receiving this mail because:
You are the assignee for the bug.
#1 lore raw

https://bugs.dpdk.org/show_bug.cgi?id=323

Thomas Monjalon ([email protected]) changed:

What |Removed |Added
----------------------------------------------------------------------------

                 CC|                            |[email protected]
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |DUPLICATE

--- Comment #1 from Thomas Monjalon ([email protected]) --- Thanks for trying to open the bug a second time. The first bug can be accessed now, so I am closing this one.

*** This bug has been marked as a duplicate of bug 322 ***

Signature · 2 lines
You are receiving this mail because:
You are the assignee for the bug.