Repository navigation
Wrong enable_mask when sending core enable IPC #394
Description
Wrong enable_mask is set for core enable IPC.
Kernel should keep track of how many cores should be currently enabled in order to handle this correctly.
Let's consider scenario, where the first pipeline in topology is scheduled on core 1 and the rest of them on core 0. Current flow:
- Kernel sends IPC with enable_mask = 0x2 (core 1) for the first pipeline.
- Kernel sends IPC with enable_mask = 0x1 (core 0) for the second pipeline.
The second IPC disables core 1, because the appropriate bit is 0. It should have value 0x3 instead.
Metadata
Metadata
Assignees
Labels
No labels
Activity
plbossart commentedon Dec 16, 2018
@ranj063 can you comment and/or describe how we might solve this? Wondering if this is a quick fix which can go in the next upstream batch or more time needed?
tlauda commentedon Jan 9, 2019
@ranj063 Any updates?
tlauda commentedon Jan 16, 2019
@plbossart @ranj063 Can we at least get estimates when it can be fixed?
btrfs: wait on ordered extents on abort cleanup
plbossart commentedon Mar 7, 2019
@ranj063 @tlauda is this issue still valid, we've merged a PR a while ago for this.
ranj063 commentedon Mar 7, 2019
@plbossart yes, this is fixed.
bonding: fix lockdep warning in bond_get_stats()
bpf: improve duplicate source code line detection