On Fri, 4 Feb 2022 at 11:14, Peter Maydell <peter.mayd...@linaro.org> wrote: > > On Wed, 2 Feb 2022 at 15:23, Jimmy Brisson <jimmy.bris...@linaro.org> wrote: > > > > Turns out that this manifests in being unable to configure > > the ethernet access permissions, as the IotKitPPC looks > > these up by name. > > > > With this fix, eth is configurable > > > > Signed-off-by: Jimmy Brisson <jimmy.bris...@linaro.org> > > Can you explain the issue here in more detail, and maybe > provide a repro case ?
Sure. I documented how I found this on my blog here: https://theotherjimmy.github.io/blog/blog/2022-01-31-debug-qemu-eth.html It's long, so I'll provide a quick summary here. My reproducer is the following zephyr example: ``` west build -p auto -b mps3_an547_ns zephyr/samples/net/dhcpv4_client \ -d build/mps3_an547_ns/net/dhcpv4_client -- \ -DOVERLAY_CONFIG=overlay-smsc911x.conf \ -DCONFIG_NET_QEMU_USER=y \ -DCONFIG_BUILD_WITH_TFM=y \ -DCOONFIG_TFM_IPC=y ``` As part of my debugging, I was using gdb on both the guest (tfm & zephyr and host qemu). I was able to reproduce this quickly by dumping the associated smsc911x struct in guest gdb, and watching the breakpoints in the `tz_ppc_read` function. Digging through, it seems that when setting up the peripheral access control in `iotkit_secctl_update_ppc_ap`, it wrote the lowest 3 bits of the ap and ignored the rest. This matches the expected behavior from the APB PPC EXT0. Let me know if you have further questions, Jimmy > The AN547 document definitely thinks > that APB PPC EXP 0 has the Memory Protection Controllers and > AHB PPC EXP 0 has the GPIO, USB and Ethernet devices: > https://developer.arm.com/-/media/Arm%20Developer%20Community/PDF/DAI0547B_SSE300_PLUS_U55_FPGA_for_mps3.pdf > (tables 6-2 to 6-4 on pages 35, 36). > > thanks > -- PMM
