On Tue, 4 Apr 2017 14:47:58 -0400 Sinan Kaya <okaya@xxxxxxxxxxxxxx> wrote: > I'm looking for a guideline on how Access Control Services (ACS) need to be > implemented in HW for cases where peer to peer is and isn't supported. > > I see conflicting information in the code. > > iommu.c calls pci_acs_enabled() from pci_device_group() to determine if > ACS path is enabled for the following flags. > > This paragraph lays out the requirements. > > /* > * To consider a PCI device isolated, we require ACS to support Source > * Validation, Request Redirection, Completer Redirection, and Upstream > * Forwarding. This effectively means that devices cannot spoof their > * requester ID, requests and completions cannot be redirected, and all > * transactions are forwarded upstream, even as it passes through a > * bridge where the target device is downstream. > */ > #define REQ_ACS_FLAGS (PCI_ACS_SV | PCI_ACS_RR | PCI_ACS_CR | PCI_ACS_UF) > > I also see the following comment in pci_acs_flags_enabled() which gets > called from pci_acs_enabled() > > /* > * Except for egress control, capabilities are either required > * or only required if controllable. Features missing from the > * capability field can therefore be assumed as hard-wired enabled. > */ > > This comment in pci_acs_flags_enabled() is confusing. A capability that reads > 0 usually means it is not enabled. This code is assuming enabled. > Are we trying to say that we'll use the capability bits as a mask while > verifying the requested ACS flags? See the language in the PCIe spec: 6.12.1.1. ACS Downstream Ports - ACS Source Validation: must be implemented. - ACS Translation Blocking: must be implemented. - ACS P2P Request Redirect: must be implemented by Root Ports that support peer-to-peer traffic with other Root Ports; must be implemented by Switch Downstream Ports. - ACS P2P Completion Redirect: must be implemented by Root Ports that implement ACS P2P Request Redirect; must be implemented by Switch Downstream Ports. - ACS Upstream Forwarding: must be implemented by Root Ports if the RC supports Redirected Request Validation; must be implemented by Switch Downstream Ports. - ACS P2P Egress Control: implementation is optional. Thus for a downstream port, SV and TB must be implemented, period. RR, CR, and UF must be implemented by switch downstream ports, period. On a root port, if RR is not implemented, we can assume it does not support P2P with other root ports. If the root port supports RR, it must also support CR and UF. The kernel code takes a bit of a lax approach to this with the items you've identified above, we assume that if a capability isn't present then the device is compliant to the spec and not trying to circumvent it. It's a generalization trying to also encompass other component types. > For systems that don't support P2P, the code will check PCI_ACS_SV > only. So, if I summarize this correctly; For a downstream port or downstream switch port, this would not be spec compliant. TB would need to be implemented for either (not that the kernel cares for isolation) and all except EC would need to be implemented on a downstream switch port. > The requirement is to have an > 1. ACS capability with PCI_ACS_SV if p2p is not supported > 2. ACS capability with PCI_ACS_SV|PCI_ACS_RR | PCI_ACS_CR | > PCI_ACS_UF if p2p is supported. > > Did I get this right? I'd suggest following the spec, not the code, that way you always have a case for how you interpret the spec and the behavior of your hardware. The code is an attempt to validate the device against the spec, but more thorough implementations may follow. REQ_ACS_FLAGS defines the set of ACS capabilities we think are relevant to device isolation. In particular, UF seems like a key feature and our current test for it may not be fully correct. Note how RR and CR only specify p2p with other root ports while UF requires transaction towards to the RC. Section 6.12.2 further defines Redirected Request Validation as a feature within the context of RR and CR. So rather than look at the code, discuss section 6.12 with the hardware engineers and understand how it behaves relative to each device and transaction type. Implement the capabilities that best match. Attempting a minimal implementation based on the current software interpretation may bite later, for instance if we re-interpret how UF works. Thanks, Alex
