On 2020-11-07 11:36, Qu Wenruo wrote:
On 2019/11/21 上午1:05, Lorenzo Pieralisi wrote:
On Sat, Nov 16, 2019 at 12:54:19PM +0000, Robin Murphy wrote:
The 0V9 and 1V8 supplies power the PCIe block in the SoC itself, and
are thus fundamental to PCIe being usable at all. As such, it makes
sense to treat them as non-optional and rely on dummy regulators if
not explicitly described.
Signed-off-by: Robin Murphy <robin.murphy@xxxxxxx>
---
drivers/pci/controller/pcie-rockchip-host.c | 69
++++++++-------------
1 file changed, 25 insertions(+), 44 deletions(-)
Applied to pci/rockchip, thanks.
Sorry, this commit is cause regression for RK3399 boards unable to
detect the controller anymore.
The 1v8 (and 0v9) is causing -517 and reject the controller
initialization.
That's -EPROBE_DEFER, which must mean that a regulator *is* described,
but you're missing the relevant driver - that's an issue with your
config/initrd. Being optional should only change the behaviour if the
supply is totally absent (i.e. you get -ENODEV instead of a dummy
regulator), so I don't see that it would make any difference in this
situation anyway :/
I'm not a PCI guy, but a quick google search shows these two voltages
are not related to PCIE core functionality, especially considering the
controller used in RK3399 are mostly to provide NVME support.
Unlike the 12V and 3V3 supplies to the slot, these supplies are to the
PCIE_AVDD_0V9 and PCIE_AVDD_1V8 pins on the SoC itself, which the
datasheet describe as "Supply voltage for PCIE". Having power is kind of
important for the I/O circuits on all the signal pins to work.
Now it's almost certainly true that these supplies technically belong to
the phy rather than the controller, but it's a bit late to change the
bindings for the sake of semantics.
This bug makes all RK3399 users who put root fs into NVME driver unable
to boot the device.
I really hope some one could test the patch before affecting the end
users or at least try to understand how most users would use the PCIE
interface for.
I *am* that end user in this case - I use an M.2 NVME on my board, which
prompted me to take a look at the regulator handling here in the first
place, to see if it might be possible to shut up the annoying message
about a 12V supply that is entirely irrelevant to a board without a
full-size PCIe slot. I use a mainline-based distro, so I've been running
this change for nearly a year since it landed in v5.5, and I'm sure many
others have too. I've not heard of any other complaints in that time...
