Hi Bhupesh, On Fri, Nov 9, 2018 at 8:30 AM Bhupesh Sharma <bhsharma@xxxxxxxxxx> wrote: > > Hi Vicenç, > > Sorry I was on holidays and came back to office just yesterday. > > On Sat, Oct 27, 2018 at 3:56 PM Vicente Bergas <vicencb@xxxxxxxxx> wrote: > > > > On Sat, Oct 27, 2018 at 12:55 AM Bhupesh Sharma <bhsharma@xxxxxxxxxx> wrote: > > > > > > Hi Vicenç, > > > > > > On Fri, Oct 26, 2018 at 6:42 PM Vicente Bergas <vicencb@xxxxxxxxx> wrote: > > > > > > > > Hello, > > > > when executing > > > > kexec -d --dtb dtb_without_chosen_node.dtb --append 'cmdline' --load Image > > > > it reports > > > > dtb_set_property: fdt_add_subnode failed: <valid offset/length> > > > > kexec: Set device tree bootargs failed. > > > > > > > > It has been tested on the arm64 architecture with version v2.0.17 and > > > > v2.0.18-rc1 > > > > > > Can you share some details on the underlying platform and kernel > > > version you are using? > > Again, would request you to share the above details, which would help > me answer your queries in a better way. It has been tested on the arm64 architecture and the kernel version was 4.19, now moved to 4.20-rc. > > > Looking at the logs, I am assuming you are running 'kexec' on a arm64 > > > platform (as it depends on creating a dtb to be passed to the 2nd > > > kernel). > > > > > > My general advice is to avoid the --dtb option as it is known to cause > > > issues with kexec in the past (see [0] for details) > > > Ideally, if you don't use the --dtb option, then kexec will read the existing > > > kernel's dtb from /proc/device-tree, and that device tree will include > > > all the changes that the boot loader has done, including adding details > > > of the available memory in the system. > > > > > > [0]. https://www.spinics.net/lists/arm-kernel/msg618236.html > > > > > > Hope this helps. > > > > > > Regards, > > > Bhupesh > > > > Hi Bhupesh, thanks for answering and yes, this helps! > > > > As you say, when not using the --dtb option it works fine. But this > > way is not applicable in my use case. > > > > I am trying to use Linux as a bootloader. This bootloader will reside > > in a difficult-to-write memory (SPI-NOR Flash) and, because of that, > > will rarely be updated. > > Over time, new kernel versions will provide updated DTB files, so, > > the DT from both kernels will diverge over time. > > This is the reason to use the --dtb option. > > > > Thanks for the information you provided, now I see a bug and also a > > missing feature. > > > > > > Focusing on the bug, I have traced it down to here: > > https://git.kernel.org/pub/scm/utils/kernel/kexec/kexec-tools.git/tree/kexec/dt-ops.c#n87 > > the result is tested for non-zero as error condition, but, as seen here > > https://git.kernel.org/pub/scm/utils/kernel/kexec/kexec-tools.git/tree/kexec/libfdt/fdt_rw.c#n345 > > it looks like positive values are not errors. > > Hmmm.. looking hard at it, I am not sure if its indeed a bug, but more > on the same below ... > > > Now, focusing on the missing feature. > > kexec-tools already has the (buggy) ability to add the "/chosen/bootargs" node. > > I assume that it also adds or updates the "/chosen/linux,initrd-start" > > and "/chosen/linux,initrd-end" nodes because it has the --initrd option. > > Again, I assume it also adds/updates all "/memreserve" nodes with > > the location where the DTB and initrd reside in memory. > > > So, just adding the "/memory" node will make it feature complete. > > This node can be populated with data provided from the command line as > > parameters or with data extracted from /proc/device-tree. > > > > Adding "/ethernetX/local-mac-address" would be a bonus. > > ... Not really, as again this is very board specific. For e.g. I can > have a 'usb' (or even multiple Ethernet > cards on the board) as the transport media, which means that the > 'kexec-tools' must be able to handle all of those > via the --dtb option and prepare the dtb to be sent to the 2nd kernel > accordingly. > > I hope you understand my point here that we have a minimalistic > implementation intentionally as catering to all > dtb nodes in kexec-tools is probably equivalent to adding > functionality similar to how a bootloader (like u-boot) allows > dtb properties to be changed. And even if we add some basic skeleton > for the same, we cannot be sure that it will be suitable for all arm64 > hardware available. Ok, I see the point, so, forget about new features. But I still think the situation is improvable in one or other way for DTBs without a chosen node: a.- In dt-ops.c:85 add the chosen node and proceed without error. The current code already adds it, but then it aborts. b.- Report an error message saying that this use case is not supported. Otherwise the current message looks like the program is broken. > > As a side node, can I suggest updating libfdt with latest upstream? > > from: > > https://git.kernel.org/pub/scm/utils/kernel/kexec/kexec-tools.git/tree/kexec/libfdt > > to: > > https://git.kernel.org/pub/scm/utils/dtc/dtc.git/tree/libfdt > > This is a good suggestion. I have this already on my to-do queue. I > will find some time to complete it over the weekend and share a patch > on the same probably by next week. > > Thanks, > Bhupesh Regards, Vicenç. _______________________________________________ kexec mailing list kexec@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/kexec
