search

riscv-non-isa / riscv-iommu

RISC-V IOMMU Specification
https://jira.riscv.org/browse/RVG-55
Creative Commons Attribution 4.0 International
76 stars 15 forks source link

Questions about ATS translation request and ATS Page Request handling #321

Closed wangyongzhen0322 closed 2 months ago

wangyongzhen0322 commented 2 months ago

  1. Whether the DTF affects the following logged in the fault queue

    ATS [1] translation requests that encounter a configuration error results in a Completer Abort (CA) response to the requester. The following cause codes belong to this category: • Instruction access fault (cause = 1) • Read access fault (cause = 5) • Write/AMO access fault (cause = 7) • MSI PTE load access fault (cause = 261) • MSI PTE misconfigured (cause = 263) • PDT entry load access fault (cause = 265) • PDT entry misconfigured (cause = 267)

  2. Whether the following status cause a fault to log in the fault queue? The conditons where no faults are logged in the fault queue for PCIe ATS "Page Request" messages don't include both the following status

    The status is set to Invalid Request if the following faults are encountered: • ddtp.iommu_mode is Bare • EN_PRI is set to 0

ved-rivos commented 2 months ago

Whether the DTF affects the following logged in the fault queue

Yes it does. Please also see table 11.

Whether the following status cause a fault to log in the fault queue? The conditons where no faults are logged in the fault queue for PCIe ATS "Page Request" messages don't include both the following status

Yes, they will. Please also see section 4.2 that states "If a fault condition prevents locating a valid device context then the DTF value assumed for reporting such faults is 0."

wangyongzhen0322 commented 2 months ago

Yes, they will. Please also see section 4.2 that states "If a fault condition prevents locating a valid device context then the DTF value assumed for reporting such faults is 0."

So what is the cause of each of them?I think that when ddtp.iommu_mode is Bare,report"Transaction type disallowed" (cause = 260);and when EN_PRI is set to 0, report "DDT entry misconfigured" (cause = 259).And the ttyp should be "PCIe Message Request"(ttyp ==9).Is that right?

ved-rivos commented 2 months ago

The cause is 260 for both of these and the transaction type is message request.

wangyongzhen0322 commented 2 months ago

I also have another question. For the PRI fault report, the software can get the relevant information through the interruption caused by fip or pip, except when pqen=0 or pqon=0. I can understand why a fault report is not generated to log in fault queue when page-request queue encountered a memory access fault (pqcsr.pqmf == 1) or "Page Request" could not be queued due to the page-request queue being full (pqt == pqh - 1) or had a overflow (pqcsr.pqof == 1). That's because pqof and pqmf themselves generate an interrupt. Why there is no fault report logged in fault queue when pqen==0 or pqon==0?

ved-rivos commented 2 months ago

Why there is no fault report logged in fault queue when pqen==0 or pqon==0?

When software has just disabled the page requests queue then it is not useful to remind software that it has disabled the page request queue.

wangyongzhen0322 commented 2 months ago

Thanks for your answer.