| Message ID | 20230606123119.57499-1-manivannan.sadhasivam@linaro.org (mailing list archive) |
|---|---|
| Headers | show |
| Series | Add MHI Endpoint network driver | expand |
On Tue, Jun 06, 2023 at 06:01:16PM +0530, Manivannan Sadhasivam wrote: > Hi, > > This series adds a network driver for the Modem Host Interface (MHI) endpoint > devices that provides network interfaces to the PCIe based Qualcomm endpoint > devices supporting MHI bus (like Modems). This driver allows the MHI endpoint > devices to establish IP communication with the host machines (x86, ARM64) over > MHI bus. > > On the host side, the existing mhi_net driver provides the network connectivity > to the host. > > - Mani > > Manivannan Sadhasivam (3): > net: Add MHI Endpoint network driver > MAINTAINERS: Add entry for MHI networking drivers under MHI bus > net: mhi: Increase the default MTU from 16K to 32K > > MAINTAINERS | 1 + > drivers/net/Kconfig | 9 ++ > drivers/net/Makefile | 1 + > drivers/net/mhi_ep_net.c | 331 +++++++++++++++++++++++++++++++++++++++ > drivers/net/mhi_net.c | 2 +- Should we add a drivers/net/modem directory? Maybe modem is too generic, we want something which represents GSM, LTE, UMTS, 3G, 4G, 5G, ... XG etc. Andrew
On Tue, 6 Jun 2023 18:01:16 +0530 Manivannan Sadhasivam wrote: > This series adds a network driver for the Modem Host Interface (MHI) endpoint > devices that provides network interfaces to the PCIe based Qualcomm endpoint > devices supporting MHI bus (like Modems). This driver allows the MHI endpoint > devices to establish IP communication with the host machines (x86, ARM64) over > MHI bus. > > On the host side, the existing mhi_net driver provides the network connectivity > to the host. So the host can talk to the firmware over IP?
On Tue, Jun 06, 2023 at 02:59:00PM +0200, Andrew Lunn wrote: > On Tue, Jun 06, 2023 at 06:01:16PM +0530, Manivannan Sadhasivam wrote: > > Hi, > > > > This series adds a network driver for the Modem Host Interface (MHI) endpoint > > devices that provides network interfaces to the PCIe based Qualcomm endpoint > > devices supporting MHI bus (like Modems). This driver allows the MHI endpoint > > devices to establish IP communication with the host machines (x86, ARM64) over > > MHI bus. > > > > On the host side, the existing mhi_net driver provides the network connectivity > > to the host. > > > > - Mani > > > > Manivannan Sadhasivam (3): > > net: Add MHI Endpoint network driver > > MAINTAINERS: Add entry for MHI networking drivers under MHI bus > > net: mhi: Increase the default MTU from 16K to 32K > > > > MAINTAINERS | 1 + > > drivers/net/Kconfig | 9 ++ > > drivers/net/Makefile | 1 + > > drivers/net/mhi_ep_net.c | 331 +++++++++++++++++++++++++++++++++++++++ > > drivers/net/mhi_net.c | 2 +- > > Should we add a drivers/net/modem directory? Maybe modem is too > generic, we want something which represents GSM, LTE, UMTS, 3G, 4G, > 5G, ... XG etc. > The generic modem hierarchy sounds good to me because most of the times a single driver handles multiple technologies. The existing drivers supporting modems are already under different hierarchy like usb, wwan etc... So unifying them makes sense. But someone from networking community should take a call. - Mani > Andrew
On Tue, Jun 06, 2023 at 02:22:27PM -0700, Jakub Kicinski wrote: > On Tue, 6 Jun 2023 18:01:16 +0530 Manivannan Sadhasivam wrote: > > This series adds a network driver for the Modem Host Interface (MHI) endpoint > > devices that provides network interfaces to the PCIe based Qualcomm endpoint > > devices supporting MHI bus (like Modems). This driver allows the MHI endpoint > > devices to establish IP communication with the host machines (x86, ARM64) over > > MHI bus. > > > > On the host side, the existing mhi_net driver provides the network connectivity > > to the host. > > So the host can talk to the firmware over IP? That's the typical usecase of these PCIe based modems. On the host, mhi_net driver creates the network interface that communicates with the endpoint over MHI host stack. On the endpoint, mhi_ep_net driver creates the network interface that communicates with the host over MHI endpoint stack. These drivers work on top of MHI channels like IP_SW0, IP_HW0 etc... IP_SW0 channel represents the IP communication between host and modem CPUs while IP_HW0 represents IP communication between host and modem DSP. - Mani
On Wed, 7 Jun 2023 at 08:56, Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> wrote: > > On Tue, Jun 06, 2023 at 02:59:00PM +0200, Andrew Lunn wrote: > > On Tue, Jun 06, 2023 at 06:01:16PM +0530, Manivannan Sadhasivam wrote: > > > Hi, > > > > > > This series adds a network driver for the Modem Host Interface (MHI) endpoint > > > devices that provides network interfaces to the PCIe based Qualcomm endpoint > > > devices supporting MHI bus (like Modems). This driver allows the MHI endpoint > > > devices to establish IP communication with the host machines (x86, ARM64) over > > > MHI bus. > > > > > > On the host side, the existing mhi_net driver provides the network connectivity > > > to the host. > > > > > > - Mani > > > > > > Manivannan Sadhasivam (3): > > > net: Add MHI Endpoint network driver > > > MAINTAINERS: Add entry for MHI networking drivers under MHI bus > > > net: mhi: Increase the default MTU from 16K to 32K > > > > > > MAINTAINERS | 1 + > > > drivers/net/Kconfig | 9 ++ > > > drivers/net/Makefile | 1 + > > > drivers/net/mhi_ep_net.c | 331 +++++++++++++++++++++++++++++++++++++++ > > > drivers/net/mhi_net.c | 2 +- > > > > Should we add a drivers/net/modem directory? Maybe modem is too > > generic, we want something which represents GSM, LTE, UMTS, 3G, 4G, > > 5G, ... XG etc. > > > > The generic modem hierarchy sounds good to me because most of the times a > single driver handles multiple technologies. The existing drivers supporting > modems are already under different hierarchy like usb, wwan etc... So unifying > them makes sense. But someone from networking community should take a call. Yes, so there is already a drivers/net/wwan directory for this, in which there are drivers for control and data path, that together represent a given 'wwan' (modem) entity. So the generic mhi_net could be moved there, but the point is AFAIU, that MHI, despite his name, is not (more) used only for modem, but as a generic memory sharing based transport protocol, such as virtio. It would then not be necessarily true that a peripheral exposing MHI net channel is actually a modem? Regards, Loic
On Wed, Jun 07, 2023 at 09:12:00AM +0200, Loic Poulain wrote: > On Wed, 7 Jun 2023 at 08:56, Manivannan Sadhasivam > <manivannan.sadhasivam@linaro.org> wrote: > > > > On Tue, Jun 06, 2023 at 02:59:00PM +0200, Andrew Lunn wrote: > > > On Tue, Jun 06, 2023 at 06:01:16PM +0530, Manivannan Sadhasivam wrote: > > > > Hi, > > > > > > > > This series adds a network driver for the Modem Host Interface (MHI) endpoint > > > > devices that provides network interfaces to the PCIe based Qualcomm endpoint > > > > devices supporting MHI bus (like Modems). This driver allows the MHI endpoint > > > > devices to establish IP communication with the host machines (x86, ARM64) over > > > > MHI bus. > > > > > > > > On the host side, the existing mhi_net driver provides the network connectivity > > > > to the host. > > > > > > > > - Mani > > > > > > > > Manivannan Sadhasivam (3): > > > > net: Add MHI Endpoint network driver > > > > MAINTAINERS: Add entry for MHI networking drivers under MHI bus > > > > net: mhi: Increase the default MTU from 16K to 32K > > > > > > > > MAINTAINERS | 1 + > > > > drivers/net/Kconfig | 9 ++ > > > > drivers/net/Makefile | 1 + > > > > drivers/net/mhi_ep_net.c | 331 +++++++++++++++++++++++++++++++++++++++ > > > > drivers/net/mhi_net.c | 2 +- > > > > > > Should we add a drivers/net/modem directory? Maybe modem is too > > > generic, we want something which represents GSM, LTE, UMTS, 3G, 4G, > > > 5G, ... XG etc. > > > > > > > The generic modem hierarchy sounds good to me because most of the times a > > single driver handles multiple technologies. The existing drivers supporting > > modems are already under different hierarchy like usb, wwan etc... So unifying > > them makes sense. But someone from networking community should take a call. > > > Yes, so there is already a drivers/net/wwan directory for this, in > which there are drivers for control and data path, that together > represent a given 'wwan' (modem) entity. So the generic mhi_net could > be moved there, but the point is AFAIU, that MHI, despite his name, is > not (more) used only for modem, but as a generic memory sharing based > transport protocol, such as virtio. It would then not be necessarily > true that a peripheral exposing MHI net channel is actually a modem? > Agree, mhi_*_net drivers can be used by non-modem devices too as long as they support MHI protocol. - Mani > Regards, > Loic
> > Yes, so there is already a drivers/net/wwan directory for this, in > > which there are drivers for control and data path, that together > > represent a given 'wwan' (modem) entity. So the generic mhi_net could > > be moved there, but the point is AFAIU, that MHI, despite his name, is > > not (more) used only for modem, but as a generic memory sharing based > > transport protocol, such as virtio. It would then not be necessarily > > true that a peripheral exposing MHI net channel is actually a modem? > > > > Agree, mhi_*_net drivers can be used by non-modem devices too as long as they > support MHI protocol. O.K. I was just trying to avoid cluttering up the directory. But if this is shared code, not actual drivers, this is fine. Are there more features yet to be implemented? Would it make sense to create a mhi directory? Andrew
On 6/7/2023 1:41 AM, Manivannan Sadhasivam wrote: > On Wed, Jun 07, 2023 at 09:12:00AM +0200, Loic Poulain wrote: >> On Wed, 7 Jun 2023 at 08:56, Manivannan Sadhasivam >> <manivannan.sadhasivam@linaro.org> wrote: >>> >>> On Tue, Jun 06, 2023 at 02:59:00PM +0200, Andrew Lunn wrote: >>>> On Tue, Jun 06, 2023 at 06:01:16PM +0530, Manivannan Sadhasivam wrote: >>>>> Hi, >>>>> >>>>> This series adds a network driver for the Modem Host Interface (MHI) endpoint >>>>> devices that provides network interfaces to the PCIe based Qualcomm endpoint >>>>> devices supporting MHI bus (like Modems). This driver allows the MHI endpoint >>>>> devices to establish IP communication with the host machines (x86, ARM64) over >>>>> MHI bus. >>>>> >>>>> On the host side, the existing mhi_net driver provides the network connectivity >>>>> to the host. >>>>> >>>>> - Mani >>>>> >>>>> Manivannan Sadhasivam (3): >>>>> net: Add MHI Endpoint network driver >>>>> MAINTAINERS: Add entry for MHI networking drivers under MHI bus >>>>> net: mhi: Increase the default MTU from 16K to 32K >>>>> >>>>> MAINTAINERS | 1 + >>>>> drivers/net/Kconfig | 9 ++ >>>>> drivers/net/Makefile | 1 + >>>>> drivers/net/mhi_ep_net.c | 331 +++++++++++++++++++++++++++++++++++++++ >>>>> drivers/net/mhi_net.c | 2 +- >>>> >>>> Should we add a drivers/net/modem directory? Maybe modem is too >>>> generic, we want something which represents GSM, LTE, UMTS, 3G, 4G, >>>> 5G, ... XG etc. >>>> >>> >>> The generic modem hierarchy sounds good to me because most of the times a >>> single driver handles multiple technologies. The existing drivers supporting >>> modems are already under different hierarchy like usb, wwan etc... So unifying >>> them makes sense. But someone from networking community should take a call. >> >> >> Yes, so there is already a drivers/net/wwan directory for this, in >> which there are drivers for control and data path, that together >> represent a given 'wwan' (modem) entity. So the generic mhi_net could >> be moved there, but the point is AFAIU, that MHI, despite his name, is >> not (more) used only for modem, but as a generic memory sharing based >> transport protocol, such as virtio. It would then not be necessarily >> true that a peripheral exposing MHI net channel is actually a modem? >> > > Agree, mhi_*_net drivers can be used by non-modem devices too as long as they > support MHI protocol. I know of at-least 1 non-modem product in development that would benefit from these drivers.