Set required_irql in sockops and sockaddr program information #3159
Conversation
Signed-off-by: Alan Jowett <[email protected]>
Alan-Jowett
requested review from
dthaler,
poornagmsft,
saxena-anurag,
shankarseal,
dv-msft,
gtrevi,
shpalani,
matthewige,
mtfriesen and
rectified95
as code owners
docs/eBpfExtensions.md
Outdated
| @@ -73,7 +73,7 @@ functions that override the global helper functions provided by the eBPF runtime | |||
| structure from provided data and context buffers. | |||
| * `context_destroy`: Pointer to `ebpf_program_context_destroy_t` function that destroys a program type specific | |||
| context structure and populates the returned data and context buffers. | |||
| * `required_irql`: IRQL that the eBPF program runs at. | |||
| * `required_irql`: IRQL that the eBPF program runs at when being invoked via bpf_prog_test_run_opts. | |||
There was a problem hiding this comment.
| * `required_irql`: IRQL that the eBPF program runs at when being invoked via bpf_prog_test_run_opts. | |
| * `required_irql`: IRQL at which the eBPF program is invoked by bpf_prog_test_run_opts. |
There was a problem hiding this comment.
So does this mean the IRQL used by bpf_prog_test_run_opts, and the IRQL from a real invocation can be different? If so, the label "required_irql' seems wrong to me.
There was a problem hiding this comment.
Agreed. Longer term there is a need to have the bpf_prog_test_run_opts specify what IRQL to run the BPF program at and for the extension to be able to validate this, which will be addressed by #3157, but that is a larger change. The goal of this is to unblock the BPF performance testing so we can detect regressions.
Before #3157 can be implemented, there needs to be a decision on:
- Do we need to update the version of the interface for pre-release eBPF core or is a breaking changer acceptable.
- How can this be passed this via the libbpf API in a way that maintains compatability.
There was a problem hiding this comment.
Agreed that the label is wrong? If so, why not change it now?
There was a problem hiding this comment.
Ok, I take it changing the label would be a breaking change, yes?
Co-authored-by: Dave Thaler <[email protected]>
Description
The sockops and sockaddr program information are missing the required_irql fields. This causes bpf_prog_test_run_opts to invoke them at IRQL == PASSIVE_LEVEL instead of IRQL == DISPATCH_LEVEL.
Testing
CI/CD
Documentation
No.
Installation
No.