First steps with Sculpt on Dell XPS 13

[Hinnerk van Bruinehsen]

I made some progress on booting Genode on my XPS 15: adding you and Romans
changes (to limit the resolution to 1920 width), I am able to boot it until I
get a black screen and a usable cursor (at least after increasing the different
RAM quotas).
After that nothing really happens. I don't have easily access to a supported
device so sadly I am unable to perform the necessary configurations for sculpt
as a hardware-probing instrument.
Do you have any hints on what I may tweak to get the boot continuing?

WKR Hinnerk

On Thu, Sep 06, 2018 at 10:45:59AM +0200, Norman Feske wrote:
> Hi Roman,
>
> thanks for the interesting experience report. I'm particularly delighted
> about the success story of the report-dump mechanism!
>
> > [4]
> https://gist.github.com/rite/1e5602c9ed55a810fe9306475e3962c0#file-log-L501
>
> From the data, I can immediately spot two problems:
>
> * On your machine, both the AHCI and NVMe drivers are started within
>   the [drivers -> dynamic] subsystem. This ultimately exhausts the RAM
>   quota of this subsystem. Instead of the designated 8 MiB, the
>   NVMe driver only receives the remaining 2 MiB. Hence, the first
>   lifesign we see from the driver is the plea for more resources:
>
>     [drivers -> dynamic -> nvme_drv] resource_request: ram_quota=4M,
> cap_quota=0
>
>   I suspect that the blocking NVMe driver is the reason sculpt could
>   not reach the deploy stage (for executing the report-dump pkg).
>
> * The RAM quota assigned to the intel_fb driver does not suffice for
>   the allocation of the frame buffer for your big resolution. From
>   the log (and from the drivers/dynamic/state), we can see that the
>   driver requests an additional 11 MiB of RAM.
>
> Both problems can be addressed by increasing the respective RAM quotas.
>
> 1. Increase the RAM quota of the drivers subsystem [1] from 66 MiB to,
>    let's say, 100 MiB. This gives the drivers subsystem an additional
>    34 MiB to work with, which should already resolve the NVMe issue.
>
> 2. Tweak the driver manager's quota assignment for the intel_fb driver,
>    increasing the RAM quota by at least 12 MiB [2].
>
> 3. Make sure to integrate the tweaked version of the driver manager
>    in a new Sculpt image, e.g., by adding the following line to the
>    sculpt.run file [3] and specifying the 'driver_manager' as boot
>    module.
>
>      build { app/driver_manager }
>
>      append boot_modules { driver_manager }
>
> If these steps solve the issue for you, there is still the chance that
> other components (like [leitzentrale -> nit_fader]) exceed their
> assigned quotas. Please look out for resource requests in the log.
>
> The current quota assignments are quite conservative to facilitate the
> test-driving Sculpt in Qemu. Once you succeeded in bringing up the
> system, I'd appreciate you sharing your quota assignments so we can
> adjust Sculpt accordingly (within reasonable bounds).
>
> [1]
> https://github.com/genodelabs/genode/blob/master/repos/gems/run/sculpt.run#L234
> [2]
> https://github.com/genodelabs/genode/blob/master/repos/gems/src/app/driver_manager/main.cc#L129
> [2]
> https://github.com/genodelabs/genode/blob/master/repos/gems/run/sculpt.run#L624
>
> Good luck!
>
> Norman
>
> --
> Dr.-Ing. Norman Feske
> Genode Labs
>
> https://www.genode-labs.com · https://genode.org
>
> Genode Labs GmbH · Amtsgericht Dresden · HRB 28424 · Sitz Dresden
> Geschäftsführer: Dr.-Ing. Norman Feske, Christian Helmuth
>




> _______________________________________________
> Genode users mailing list
> users at lists.genode.org
> https://lists.genode.org/listinfo/users