Want to know the policy that Tekton request resource for the Task with multi containers

[NEUerYZY]

I have done an experiment with the Tekton version v0.25.0 but meet a problem .

I try to make a TaskRun with the following description in a namespace with limit range 20C 20Gi.

spec:
  steps:
    - name: step1
      resources:
        requests:
          memory: 4Gi
          cpu: 2C
        limits:
          memory: 4Gi
          cpu: 2C
    - name: step2
      resources:
        requests:
          memory: 8Gi
          cpu: 4C
        limits:
          memory: 8Gi
          cpu: 4C
    - name: step3
      resources:
        requests:
          memory: 16Gi
          cpu: 8C
        limits:
          memory: 16Gi
          cpu: 8C

Then I found Tekton request the POD from K8S cluster with request resource 8C 16GI and limit resource 2+4+8 = 14C , 4+8+16=28Gi . The POD doesn't execute because of the namespace level limitation 20GI < limit resource 28Gi.

From the description of Pull request #723 and the document [https://github.com/tektoncd/pipeline/blob/v0.25.0/docs/tasks.md#defining-steps] . I can know that Tetkon reserve the max Request resource for multi containers rather than the sum of request . But there is no related description with Limit resource

So , all of above I want to know

1. The principle that Tekton request limit resource from K8S cluster when a Task has multi containers .
2. Is there any solution for this scene that can request the limit resource using the max limit resource among containers ?

[sibelius]

did you find the fix ?