| Message ID | 20240806114118.17198-3-ansuelsmth@gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | mtd: improve block2mtd + airoha parser | expand |
On Tue, Aug 06, 2024 at 01:41:12PM +0200, Christian Marangi wrote: > Introduce support for a dedicated node for a nvme card. This will be a > subnode of the nvme controller node that will have the "nvme-card" > compatible. > > This follow a similar implementation done for mmc where the specific mmc > card have a dedicated of_node. > > This can be used for scenario where block2mtd module is used to declare > partition in DT and block2mtd is called on the root block of the nvme > card, permitting the usage of fixed-partition parser or alternative > ones. Err, hell no. Why would you wire up a purely PCIe device to OF? PCIe is self-discovering.
On Tue, Aug 06, 2024 at 02:43:12PM +0200, Christoph Hellwig wrote: > On Tue, Aug 06, 2024 at 01:41:12PM +0200, Christian Marangi wrote: > > Introduce support for a dedicated node for a nvme card. This will be a > > subnode of the nvme controller node that will have the "nvme-card" > > compatible. > > > > This follow a similar implementation done for mmc where the specific mmc > > card have a dedicated of_node. > > > > This can be used for scenario where block2mtd module is used to declare > > partition in DT and block2mtd is called on the root block of the nvme > > card, permitting the usage of fixed-partition parser or alternative > > ones. > > Err, hell no. Why would you wire up a purely PCIe device to OF? > PCIe is self-discovering. > Well on embedded pure PCIe card most of the time are not a thing... Unless it's an enterprise product, everything is integrated in the pcb and not detachable for cost saving measure or also if the thing use PCIe protocol but it tighlty coupled with the SoC. This implementation is already very common for all kind of pcie devices like wireless card, gpio expander that are integrated in the PCB and require property in DT like calibration data, quirks or GPIO pin definitions, i2c... In modern SoC we are seeing an influx of using cheap flash storage option instead of NAND or NOR as modern hw require more space and price increase is not that high... Almost any high tier device is switching to using emmc and even attached NVME and simulating MTD with them for easy usage. Please consider this well used scenario in emebedded where PCIe is just a comunication way and the concept of detachable doesn't exist at all and things can be described in DT as static. Also these storage are used for rootfs mount so userspace is not so viable.
On Tue, Aug 06, 2024 at 03:03:24PM +0200, Christian Marangi wrote: > Well on embedded pure PCIe card most of the time are not a thing... > Unless it's an enterprise product, everything is integrated in the pcb > and not detachable for cost saving measure or also if the thing use PCIe > protocol but it tighlty coupled with the SoC. Yes, PCIe has a bunch of form factors, including just soldered on BGA devices, but none of that matters at all for the logical protocol. > This implementation is already very common for all kind of pcie devices > like wireless card, gpio expander that are integrated in the PCB and > require property in DT like calibration data, quirks or GPIO pin > definitions, i2c... Do you have a document on that/
On Tue, Aug 06, 2024 at 03:09:38PM +0200, Christoph Hellwig wrote: > On Tue, Aug 06, 2024 at 03:03:24PM +0200, Christian Marangi wrote: > > Well on embedded pure PCIe card most of the time are not a thing... > > Unless it's an enterprise product, everything is integrated in the pcb > > and not detachable for cost saving measure or also if the thing use PCIe > > protocol but it tighlty coupled with the SoC. > > Yes, PCIe has a bunch of form factors, including just soldered on BGA > devices, but none of that matters at all for the logical protocol. > Correct, for the context of soldered stuff tho things can be malfunctioning (problem in the PCIe driver) or working so it's possibile to make assumption and attach OF node in DT. Consider that the thing is probed only if the card is correctly detected. Card not present -> nothing is done. > > This implementation is already very common for all kind of pcie devices > > like wireless card, gpio expander that are integrated in the PCB and > > require property in DT like calibration data, quirks or GPIO pin > > definitions, i2c... > > Do you have a document on that/ > You mean example of PCIe that makes use of OF? Pretty much ath10k-11k mediatek and in general all wireless card. Example ath11k-pci.yaml [1] Compatible is set to the ID. PCIe will attach an OF node if found in the pcie child nodes (already supported) On NVMe driver side, the NVMe controller will already have an OF node present. We are now just adding an additional subnode "nvme-card" and attach it on the root disk when it's created. [1] https://github.com/torvalds/linux/blob/master/Documentation/devicetree/bindings/net/wireless/qcom%2Cath11k-pci.yaml
Hello all, On 2024-08-06 15:03, Christian Marangi wrote: > On Tue, Aug 06, 2024 at 02:43:12PM +0200, Christoph Hellwig wrote: >> On Tue, Aug 06, 2024 at 01:41:12PM +0200, Christian Marangi wrote: >> > Introduce support for a dedicated node for a nvme card. This will be a >> > subnode of the nvme controller node that will have the "nvme-card" >> > compatible. >> > >> > This follow a similar implementation done for mmc where the specific mmc >> > card have a dedicated of_node. >> > >> > This can be used for scenario where block2mtd module is used to declare >> > partition in DT and block2mtd is called on the root block of the nvme >> > card, permitting the usage of fixed-partition parser or alternative >> > ones. >> >> Err, hell no. Why would you wire up a purely PCIe device to OF? >> PCIe is self-discovering. >> > > Well on embedded pure PCIe card most of the time are not a thing... > Unless it's an enterprise product, everything is integrated in the pcb > and not detachable for cost saving measure or also if the thing use > PCIe > protocol but it tighlty coupled with the SoC. > > This implementation is already very common for all kind of pcie devices > like wireless card, gpio expander that are integrated in the PCB and > require property in DT like calibration data, quirks or GPIO pin > definitions, i2c... > > In modern SoC we are seeing an influx of using cheap flash storage > option instead of NAND or NOR as modern hw require more space and price > increase is not that high... Almost any high tier device is switching > to > using emmc and even attached NVME and simulating MTD with them for easy > usage. > > Please consider this well used scenario in emebedded where PCIe is just > a comunication way and the concept of detachable doesn't exist at all > and things can be described in DT as static. Also these storage are > used > for rootfs mount so userspace is not so viable. As a note, perhaps this is another good example of a "fixed layout" PCIe device found on an SBC: https://lore.kernel.org/linux-rockchip/20240805073425.3492078-1-jacobe.zang@wesion.com/T/#u
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 053d5b4909cd..344523274d1b 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -14,6 +14,7 @@ #include <linux/kernel.h> #include <linux/module.h> #include <linux/backing-dev.h> +#include <linux/of.h> #include <linux/slab.h> #include <linux/types.h> #include <linux/pr.h> @@ -4651,6 +4652,7 @@ void nvme_uninit_ctrl(struct nvme_ctrl *ctrl) nvme_hwmon_exit(ctrl); nvme_fault_inject_fini(&ctrl->fault_inject); dev_pm_qos_hide_latency_tolerance(ctrl->device); + of_node_put(ctrl->device->of_node); cdev_device_del(&ctrl->cdev, ctrl->device); nvme_put_ctrl(ctrl); } @@ -4771,6 +4773,8 @@ int nvme_init_ctrl(struct nvme_ctrl *ctrl, struct device *dev, else ctrl->device->groups = nvme_dev_attr_groups; ctrl->device->release = nvme_free_ctrl; + ctrl->device->of_node = of_get_compatible_child(ctrl->dev->of_node, + "nvme-card"); dev_set_drvdata(ctrl->device, ctrl); return ret;
Introduce support for a dedicated node for a nvme card. This will be a subnode of the nvme controller node that will have the "nvme-card" compatible. This follow a similar implementation done for mmc where the specific mmc card have a dedicated of_node. This can be used for scenario where block2mtd module is used to declare partition in DT and block2mtd is called on the root block of the nvme card, permitting the usage of fixed-partition parser or alternative ones. Signed-off-by: Christian Marangi <ansuelsmth@gmail.com> --- drivers/nvme/host/core.c | 4 ++++ 1 file changed, 4 insertions(+)