Hi Fabrizio, The Ethernet driver is starting to show its age. It should use the mux_sys subsystem that is passed in via the io_ops for initialising the pins. I will create a ticket for this to be fixed, but I am afraid that it will not be high priority. - AlexK On Wed, 2017-06-21 at 18:54 -0400, Fabrizio Bertocci wrote:

Hi all, I am facing an issue with our existing project (currently based on seL4 1.0). I am in the process of migrating our project to the latest version of seL4. After all the required changes I am facing a problem I need some help.

In our code, we use libplatsupport for timer & serial I/O as well as the libethdriver in order to use the network interface of the IMX6 board.

The problem appears that the IOMUX device registers are mapped twice. A first time from libplatsupport (and never released):

-> sel4platsupport_new_io_ops() -> calls: mux_sys_init() -> calls the macro: MAP_IF_NULL(io_ops, IMX6_IOMUXC, _mux.iomuxc);

This end up mapping the device region IMX6_IOMUXC: 0x20e0000 - 0x20e100 (size_bits=12).

So far so good... but then they are mapped again from the ethernet driver: ethif_new_lwip_driver_no_malloc() -> call my initialization fn -> call ethif_imx6_init() -> call setup_iomux_enet() that end up mapping again the region @ IOMUXC_PADDR = 0x20e1000

This second time, the mapping function _utspace_split_alloc fails because apparently the region 0x20e000 has been already mapped (and is not in the regions available).

If I comment out the call to mux_sys_init() in libsel4platsupport_new_io_ops() my ethernet driver successfully initialize (but clearly that's not the correct workaround).

Any suggestions? Regards, Fabrizio Bertocci Real-Time Innovations, Inc. Sunnyvale, CA

_______________________________________________ Devel mailing list Devel@sel4.systems https://sel4.systems/lists/listinfo/devel