| Message ID | 20200126220508.7.I66a7bcb7db931e196e8319d2dbca4e418274f39b@changeid |
|---|---|
| State | RFC |
| Delegated to: | Bin Meng |
| Headers | show |
| Series | RFC: dm: Add programatic generation of ACPI tables | expand |
Hi Simon, On Mon, Jan 27, 2020 at 1:08 PM Simon Glass <sjg@chromium.org> wrote: > > Add a means to avoid configuring a device when needed. Add an explanation > of why this is useful to the binding file. > Does disabling CONFIG_PCI_PNP not help? > Signed-off-by: Simon Glass <sjg@chromium.org> > --- > > doc/device-tree-bindings/pci/x86-pci.txt | 24 ++++++++++++++++++++++++ > drivers/pci/pci-uclass.c | 2 ++ > 2 files changed, 26 insertions(+) > Regards, Bin
Hi Bin, On Sat, 8 Feb 2020 at 08:15, Bin Meng <bmeng.cn@gmail.com> wrote: > > Hi Simon, > > On Mon, Jan 27, 2020 at 1:08 PM Simon Glass <sjg@chromium.org> wrote: > > > > Add a means to avoid configuring a device when needed. Add an explanation > > of why this is useful to the binding file. > > > > Does disabling CONFIG_PCI_PNP not help? That is for all devices. Here we just want the UART to stay where it is so that we can debug serial drivers, FSP-S, etc. If the UART moves then any output after that hangs the machine until we reinit the serial driver. > > > Signed-off-by: Simon Glass <sjg@chromium.org> > > --- > > > > doc/device-tree-bindings/pci/x86-pci.txt | 24 ++++++++++++++++++++++++ > > drivers/pci/pci-uclass.c | 2 ++ > > 2 files changed, 26 insertions(+) > > Regards, SImon
Hi Simon, On Sat, Feb 8, 2020 at 11:51 PM Simon Glass <sjg@chromium.org> wrote: > > Hi Bin, > > On Sat, 8 Feb 2020 at 08:15, Bin Meng <bmeng.cn@gmail.com> wrote: > > > > Hi Simon, > > > > On Mon, Jan 27, 2020 at 1:08 PM Simon Glass <sjg@chromium.org> wrote: > > > > > > Add a means to avoid configuring a device when needed. Add an explanation > > > of why this is useful to the binding file. > > > > > > > Does disabling CONFIG_PCI_PNP not help? > > That is for all devices. Here we just want the UART to stay where it > is so that we can debug serial drivers, FSP-S, etc. If the UART moves > then any output after that hangs the machine until we reinit the > serial driver. OK, I think this is for debugging purpose, right? As you must have get Coral boot for some time, and this is only for retaining the console debug output? Reviewed-by: Bin Meng <bmeng.cn@gmail.com> Regards, Bin
Hi Bin, On Sat, 8 Feb 2020 at 08:54, Bin Meng <bmeng.cn@gmail.com> wrote: > > Hi Simon, > > On Sat, Feb 8, 2020 at 11:51 PM Simon Glass <sjg@chromium.org> wrote: > > > > Hi Bin, > > > > On Sat, 8 Feb 2020 at 08:15, Bin Meng <bmeng.cn@gmail.com> wrote: > > > > > > Hi Simon, > > > > > > On Mon, Jan 27, 2020 at 1:08 PM Simon Glass <sjg@chromium.org> wrote: > > > > > > > > Add a means to avoid configuring a device when needed. Add an explanation > > > > of why this is useful to the binding file. > > > > > > > > > > Does disabling CONFIG_PCI_PNP not help? > > > > That is for all devices. Here we just want the UART to stay where it > > is so that we can debug serial drivers, FSP-S, etc. If the UART moves > > then any output after that hangs the machine until we reinit the > > serial driver. > > OK, I think this is for debugging purpose, right? As you must have get > Coral boot for some time, and this is only for retaining the console > debug output? Sort-of, but I've actually left it enabled. The problem is that if you get an error or any debug output very early after relocation in U-Boot, it hangs. I feel it is quite confusing for people to add a bit of debugging and have it crash. Admittedly there is a very small window where this happens (roughly between arch_fsp_init_r() returning and initr_serial() being called. But i found it pretty annoying. What do you think? Also I'd like to change it to suppress console output when it is not allowed (and print a warning later), but haven't tried that yet. > > Reviewed-by: Bin Meng <bmeng.cn@gmail.com> > > Regards, > Bin
Hi Simon, On Sun, Feb 9, 2020 at 12:07 AM Simon Glass <sjg@chromium.org> wrote: > > Hi Bin, > > On Sat, 8 Feb 2020 at 08:54, Bin Meng <bmeng.cn@gmail.com> wrote: > > > > Hi Simon, > > > > On Sat, Feb 8, 2020 at 11:51 PM Simon Glass <sjg@chromium.org> wrote: > > > > > > Hi Bin, > > > > > > On Sat, 8 Feb 2020 at 08:15, Bin Meng <bmeng.cn@gmail.com> wrote: > > > > > > > > Hi Simon, > > > > > > > > On Mon, Jan 27, 2020 at 1:08 PM Simon Glass <sjg@chromium.org> wrote: > > > > > > > > > > Add a means to avoid configuring a device when needed. Add an explanation > > > > > of why this is useful to the binding file. > > > > > > > > > > > > > Does disabling CONFIG_PCI_PNP not help? > > > > > > That is for all devices. Here we just want the UART to stay where it > > > is so that we can debug serial drivers, FSP-S, etc. If the UART moves > > > then any output after that hangs the machine until we reinit the > > > serial driver. > > > > OK, I think this is for debugging purpose, right? As you must have get > > Coral boot for some time, and this is only for retaining the console > > debug output? > > Sort-of, but I've actually left it enabled. The problem is that if you > get an error or any debug output very early after relocation in > U-Boot, it hangs. I feel it is quite confusing for people to add a bit > of debugging and have it crash. > > Admittedly there is a very small window where this happens (roughly > between arch_fsp_init_r() returning and initr_serial() being called. > But i found it pretty annoying. > > What do you think? > Sounds good to me for leaving it enabled by default. > Also I'd like to change it to suppress console output when it is not > allowed (and print a warning later), but haven't tried that yet. > Regards, Bin
diff --git a/doc/device-tree-bindings/pci/x86-pci.txt b/doc/device-tree-bindings/pci/x86-pci.txt index 3aa5bd9a46..62b29a4e36 100644 --- a/doc/device-tree-bindings/pci/x86-pci.txt +++ b/doc/device-tree-bindings/pci/x86-pci.txt @@ -10,6 +10,17 @@ Optional properties: configuration in TPL/SPL to reduce code size and boot time, since these phases only know about a small subset of PCI devices. +For PCI devices the following optional property is available: + +- pci,no-autoconfig : Don't automatically configure this PCI device at all. + This is used when the device is statically configured and must maintain + this same config throughout the boot process. An example is a serial + UART being used to debug PCI configuration, since reconfiguring it stops + the UART from working until the driver is re-probed, and this can cause + output to be lost. This should not generally be used in production code, + although it is often harmless. + + Example: pci { @@ -21,4 +32,17 @@ pci { 0x42000000 0x0 0xb0000000 0xb0000000 0 0x10000000 0x01000000 0x0 0x1000 0x1000 0 0xefff>; u-boot,skip-auto-config-until-reloc; + + + serial: serial@18,2 { + reg = <0x0200c210 0 0 0 0>; + u-boot,dm-pre-reloc; + compatible = "intel,apl-ns16550"; + early-regs = <0xde000000 0x20>; + reg-shift = <2>; + clock-frequency = <1843200>; + current-speed = <115200>; + acpi,name = "URT3"; + pci,no-autoconfig; + }; }; diff --git a/drivers/pci/pci-uclass.c b/drivers/pci/pci-uclass.c index bb272b67a1..f5f713fdf4 100644 --- a/drivers/pci/pci-uclass.c +++ b/drivers/pci/pci-uclass.c @@ -535,6 +535,8 @@ int pci_auto_config_devices(struct udevice *bus) int ret; debug("%s: device %s\n", __func__, dev->name); + if (dev_read_bool(dev, "pci,no-autoconfig")) + continue; ret = dm_pciauto_config_device(dev); if (ret < 0) return ret;
Add a means to avoid configuring a device when needed. Add an explanation of why this is useful to the binding file. Signed-off-by: Simon Glass <sjg@chromium.org> --- doc/device-tree-bindings/pci/x86-pci.txt | 24 ++++++++++++++++++++++++ drivers/pci/pci-uclass.c | 2 ++ 2 files changed, 26 insertions(+)