Driver for talking to PLX PEX8xxx PCIe switch over I2C

Most of the PLX PCI Express switches (and also switches from other vendors such IDT) provide an I2C based secondary interface to access and program the switch. This can be used to program the switch in situations where the PCIe interface may not be suitable, or even available. (For instance, we may need to program the device in a particular fashion before we even begin enumerating on the PCI, or to even enable the PCIe interface of the device). More information on usecases and other details has been provided in the documentation being added as part of this patchset.

This is a driver that allows talking to the PLX PEX8xxx family of PCIe switches over its I2C interface. Here are the patch descriptions:

[Patch 1/4] - Adds the PEX8xxx driver that exports API calls to read / write. [Patch 2/4] - Adds a sysfs interface, to allow access to the switch. [Patch 3/4] - Adds the MAINTAINERS entry.
[Patch 4/4] - Adds the documentation in Documentation/PCI/pex8xxx_i2c.txt

Since this is a driver for a PCIe switch, I currently put it under /driver/pci/. I'm very open to suggestions for moving this around (Does introducing drivers/pci/switch/ seem any better?). Especially I can see that the drivers/pci/Kconfig appears under "Bus Options" in the main menu which does not look like the right place for this driver. I am looking for suggestions here.

Rajat Jain (4):
pci: Add I2C driver for the PLX PEX8xxx PCIe switch pci/pex8xxx: Add sysfs interface for userspace access.
MAINTAINERS: Add new device driver entry for PEX pex8xxx
Documentation: Add documentation for the PCI switch PEX8xxx I2C
driver

 Documentation/PCI/pex8xxx_i2c.txt |  134 +++++++++
 MAINTAINERS                       |    7 +
 drivers/pci/Kconfig               |   13 +
 drivers/pci/Makefile              |    2 +
 drivers/pci/pex8xxx_i2c.c         |  538 +++++++++++++++++++++++++++++++++++++
 include/linux/i2c/pex8xxx_i2c.h   |   36 +++
 6 files changed, 730 insertions(+)
 create mode 100644 Documentation/PCI/pex8xxx_i2c.txt
 create mode 100644 drivers/pci/pex8xxx_i2c.c
 create mode 100644 include/linux/i2c/pex8xxx_i2c.h

--
1.7.9.5

Quoted 6 lines

Since this is a driver for a PCIe switch, I currently put it under /driver/pci/. I'm very open to suggestions for moving this around (Does introducing drivers/pci/switch/ seem any better?). Especially I can see that the drivers/pci/Kconfig appears under "Bus Options" in the main menu which does not look like the right place for this driver. I am looking for suggestions here.

Definitely interesting work. I have no idea where to put it. It doesn't have anything to do with PCI core services or structures, so drivers/pci doesn't seem like the ideal place for it, but I don't have any better ideas. If it did end up under drivers/pci, I agree that a drivers/pci/switch or something similar would probably be better than having it directly in drivers/pci.

Maybe Wolfram or other I2C folks will have an idea.

Quoted 14 lines

Quoted 6 lines

Since this is a driver for a PCIe switch, I currently put it under /driver/pci/. I'm very open to suggestions for moving this around (Does introducing drivers/pci/switch/ seem any better?). Especially I can see that the drivers/pci/Kconfig appears under "Bus Options" in the main menu which does not look like the right place for this driver. I am looking for suggestions here.

Definitely interesting work. I have no idea where to put it. It doesn't have anything to do with PCI core services or structures, so drivers/pci doesn't seem like the ideal place for it, but I don't have any better ideas. If it did end up under drivers/pci, I agree that a drivers/pci/switch or something similar would probably be better than having it directly in drivers/pci.

Quoted 1 line

Maybe Wolfram or other I2C folks will have an idea.

Hmm, is this a common interface for those switches? Then a directory like drivers/pci/switch could make sense. Otherwise I'd suggest drivers/misc?

Quoted 16 lines · Wolfram Sang

Quoted 15 lines

Quoted 6 lines

Since this is a driver for a PCIe switch, I currently put it under /driver/pci/. I'm very open to suggestions for moving this around (Does introducing drivers/pci/switch/ seem any better?). Especially I can see that the drivers/pci/Kconfig appears under "Bus Options" in the main menu which does not look like the right place for this driver. I am looking for suggestions here.

Definitely interesting work. I have no idea where to put it. It doesn't have anything to do with PCI core services or structures, so drivers/pci doesn't seem like the ideal place for it, but I don't have any better ideas. If it did end up under drivers/pci, I agree that a drivers/pci/switch or something similar would probably be better than having it directly in drivers/pci.

Maybe Wolfram or other I2C folks will have an idea.

Quoted 4 lines

Hmm, is this a common interface for those switches? Then a directory like drivers/pci/switch could make sense. Otherwise I'd suggest drivers/misc?

We had discussed drivers/misc internally, but it didn't seem to be the right location to me. After all, it is not a misc driver, but a driver to configure a PCIe switch. Not that I have a better idea, though.

Quoted 14 lines

Quoted 6 lines

Since this is a driver for a PCIe switch, I currently put it under /driver/pci/. I'm very open to suggestions for moving this around (Does introducing drivers/pci/switch/ seem any better?). Especially I can see that the drivers/pci/Kconfig appears under "Bus Options" in the main menu which does not look like the right place for this driver. I am looking for suggestions here.

Definitely interesting work. I have no idea where to put it. It doesn't have anything to do with PCI core services or structures, so drivers/pci doesn't seem like the ideal place for it, but I don't have any better ideas. If it did end up under drivers/pci, I agree that a drivers/pci/switch or something similar would probably be better than having it directly in drivers/pci.

Quoted 1 line

Maybe Wolfram or other I2C folks will have an idea.

Hmm, is this a common interface for those switches? Then a directory like drivers/pci/switch could make sense. Otherwise I'd suggest drivers/misc?

===========================================================
...<snip>...

+0. Why have an I2C interface to a PCIe switch?
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

(quoted diff) +18 -0

+Other than the regular PCI express interface, most modern PCIe switches (e.g.
+from IDT and PLX) have an I2C based secondary interface. This interface allows
+access to all the registers of the switch. Some of these registers may not even
+be accessible over the regular PCI interface. Also, there are certain registers
+that can be written to, using only the I2C interface and may only be read-only
+using the PCI interface.
+
+This I2C interface is often used in designs involving these switches, and can
+be used for a variety of use cases where the switch needs to be configured
+independent of the PCI subsystem (and likely before PCI enumeration). Some
+examples:
+
+* Dividing a PCIe switch into multiple "virtual" switches. Using this feature,
+  a switch could be connected to 2 root ports for instance, each managing its
+  own PCI hierarchy, and the traffic from one virtual switch does not leak into
+  another.
+
+* Managing Transparent / Non-transparent bridging, and changing them

on-the-fly.

(quoted diff) +17 -0

+  There are ports that can be converted into "Non-transparent" bridge ports.
+  Essentially this is used to create different domains (not visible to
+  software). In a dynamic distributed system, it may be desirable to change a
+  transparent bridge to non-transparent or vice versa, for example, to handle a
+  failover situation.
+
+* Buggy hardware / Bad EEPROM configuration. There may be cases where an errata
+  involving register writes need to be applied before enumerating over PCI.
+  Also these switches are typically attached to an EEPROM that is supposed to
+  initialize the switch. If that EEPROM is not present, or contains bad
+  initialization data, this I2C interface can be used to fix that.
+
+* Changing switch configuration on the fly. In a multi-homed or complex
+  distributed systemsystem, there may be a need to change the switch
+  configuration (eg. change the upstream port, or the port or lane
+  configuration etc) to address run time scenarios (CPU plug out etc).
+

...<snip>...
===========================================================

I am myself leaning towards drivers/pci/switch/. And am wanting to hear what other think.

Quoted 43 lines

===========================================================
...<snip>...

+0. Why have an I2C interface to a PCIe switch?
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

(quoted diff) +18 -0

+Other than the regular PCI express interface, most modern PCIe switches (e.g.
+from IDT and PLX) have an I2C based secondary interface. This interface allows
+access to all the registers of the switch. Some of these registers may not even
+be accessible over the regular PCI interface. Also, there are certain registers
+that can be written to, using only the I2C interface and may only be read-only
+using the PCI interface.
+
+This I2C interface is often used in designs involving these switches, and can
+be used for a variety of use cases where the switch needs to be configured
+independent of the PCI subsystem (and likely before PCI enumeration). Some
+examples:
+
+* Dividing a PCIe switch into multiple "virtual" switches. Using this feature,
+  a switch could be connected to 2 root ports for instance, each managing its
+  own PCI hierarchy, and the traffic from one virtual switch does not leak into
+  another.
+
+* Managing Transparent / Non-transparent bridging, and changing them

on-the-fly.

(quoted diff) +17 -0

+  There are ports that can be converted into "Non-transparent" bridge ports.
+  Essentially this is used to create different domains (not visible to
+  software). In a dynamic distributed system, it may be desirable to change a
+  transparent bridge to non-transparent or vice versa, for example, to handle a
+  failover situation.
+
+* Buggy hardware / Bad EEPROM configuration. There may be cases where an errata
+  involving register writes need to be applied before enumerating over PCI.
+  Also these switches are typically attached to an EEPROM that is supposed to
+  initialize the switch. If that EEPROM is not present, or contains bad
+  initialization data, this I2C interface can be used to fix that.
+
+* Changing switch configuration on the fly. In a multi-homed or complex
+  distributed systemsystem, there may be a need to change the switch
+  configuration (eg. change the upstream port, or the port or lane
+  configuration etc) to address run time scenarios (CPU plug out etc).
+

Quoted 5 lines

...<snip>...
===========================================================

I am myself leaning towards drivers/pci/switch/. And am wanting to hear what other think.

I had a closer look and my first question is now if we really need a kernel driver for this?

The sysfs interface is basically to PEEK/POKE registers of the switch. How registers get accessed is part of userspace. Well, then it might even be easier to construct the proper i2c messages in userspace, too, and send them via i2c-dev (instead of writing to various files in sysfs).

The Kernel API is also PEEK/POKE register. I am not an PCI expert, but I can't see a user in this form. What I'd expect would be more like

pci_switch_set_mode(some_device, PCI_BRIDGE_MODE_TRANSPARENT);

and then a shim layer would make sure the proper driver for the switch sends the proper commands. Again, I don't know if the decision of setting a mode/creating virtual switches... is even desirable in kernel space. If so, then this is probably a sub-subsystem drivers/pci/switches, but needs way more abstraction. Otherwise, userspace will currently do IMO.

I understand and agree. In fact in the internal version of this driver (that I have not yet sent out for review), we do have APIs added similar to what you mention above. Currently I have APIs that:
- Enable / Disable PCIe links.
- Change the upstream port.
- Enable / Disable Non-transparent mode etc.

However, I did not send them all out for review because I don't have the hardware to try and test them out on ALL the supported devices (also would require considerable amount of time). I have tested those APIs on PEX8713 only, because for e.g.I only have PEX8713 in a HW that connects to 2 CPUs at the same time.

the switch. Yes, I agree that we can have another layer of abstraction so that we have:

- The Core logic code (that knows "How do we want the switch to behave")
- A PEX8xxx driver (that knows "which registers to program")
- A PEX8xxx I2C driver ("How to program those registers") - this driver.

I do understand that your suggestion is to include and bundle the latter two into this same driver.

However since the possibilities
(about which APIs to provide) are too much and not enough hardware to test it, would it be acceptable if I include those APIs, but support them for only 1 device (and return error for others)?

with those APIs, I feel exposing the Read/Write APIs will be useful - because what core logic wants to achieve can be highly device and platform specific.

Also, a sysfs interface for this switch is proving
to be a very helpful development aid :-) (personal experience :-))

Quoted 7 lines

I understand and agree. In fact in the internal version of this driver (that I have not yet sent out for review), we do have APIs added similar to what you mention above. Currently I have APIs that:
- Enable / Disable PCIe links.
- Change the upstream port.
- Enable / Disable Non-transparent mode etc.

Now, that sounds better to me...

Quoted 5 lines

However, I did not send them all out for review because I don't have the hardware to try and test them out on ALL the supported devices (also would require considerable amount of time). I have tested those APIs on PEX8713 only, because for e.g.I only have PEX8713 in a HW that connects to 2 CPUs at the same time.

That is a common problem to not have enough hardware to test. You could ask on the PCI mailing list for testers. The solution usually lies in showing the code rather than not showing the code.

Quoted 9 lines

the switch. Yes, I agree that we can have another layer of abstraction so that we have:

- The Core logic code (that knows "How do we want the switch to behave")
- A PEX8xxx driver (that knows "which registers to program")
- A PEX8xxx I2C driver ("How to program those registers") - this driver.

I do understand that your suggestion is to include and bundle the latter two into this same driver.

It definately should be this way. Nobody else than the PEX8xxx driver should be able to send I2C messeges to the device! And this is absolutely standard, the logic how to talk to a device knows also how to prepare the I2C messages. One reason where it could be factored out is if there are multiple ways of transportation possible, like I2C or SPI.

Quoted 4 lines

However since the possibilities
(about which APIs to provide) are too much and not enough hardware to test it, would it be acceptable if I include those APIs, but support them for only 1 device (and return error for others)?

Start with what YOU need and show us (all of it). From there, we can
decide: do we start with one driver and factor out later, do we start with a sub-subsystem right away, etc... And there is still the question what APIs you provide, how they are implemented and if we really should have them in-kernel. I think that question will be more interesting for Bjorn because I don't really know much about switches in the PCI world.

Quoted 3 lines

with those APIs, I feel exposing the Read/Write APIs will be useful - because what core logic wants to achieve can be highly device and platform specific.

That could also be solved by fixup-callbacks, but we'd need to see the core to tell, really.

Quoted 2 lines

Also, a sysfs interface for this switch is proving
to be a very helpful development aid :-) (personal experience :-))

Sure, but such development aids don't need to be upstream. Especially if they create ABI such as sysfs-entries.

Thanks and regards,

Wolfram