Hello,
As I hoped, by copying the files from the Sculpt USB stick to my NVMe Ext2 partition, I can now boot into Sculpt on the ThinkPad L380 Yoga (using Norman's USB HID-disabling patches).
The results are basically the same as booting from from USB, but hopefully this will allow me to load local packages and save log files to the NVMe partition.
Does anyone have any ideas on what I can try, in order to generate some useful output?
To recap, the most important limitations at the moment are 1) the the Inspect window doesn't open, and 2) I can't connect to a network.
Thanks!
On 10/2/19 9:15 PM, John J. Karcher wrote:
Hello,
As I hoped, by copying the files from the Sculpt USB stick to my NVMe Ext2 partition, I can now boot into Sculpt on the ThinkPad L380 Yoga (using Norman's USB HID-disabling patches).
The results are basically the same as booting from from USB, but hopefully this will allow me to load local packages and save log files to the NVMe partition.
Does anyone have any ideas on what I can try, in order to generate some useful output?
To recap, the most important limitations at the moment are 1) the the Inspect window doesn't open, and 2) I can't connect to a network.
Now I'm really confused. Just a few minutes after writing that, I tried it again, and the Inspect window started working. I've booted several more times since, and the Inspect window now works reliably. The WiFi network scanning also started finding the hotspots, but it doesn't seem to be able to connect. (It has no trouble in Linux.)
That's at least a big improvement over the previous situation. Now it should be a lot easier to copy files around.
Let me know if anyone has any thoughts.
Thanks again! -- John J. Karcher devuser@alternateapproach.com
Hi John,
Now I'm really confused. Just a few minutes after writing that, I tried it again, and the Inspect window started working. I've booted several more times since, and the Inspect window now works reliably. The WiFi network scanning also started finding the hotspots, but it doesn't seem to be able to connect. (It has no trouble in Linux.)
That's at least a big improvement over the previous situation. Now it should be a lot easier to copy files around.
Let me know if anyone has any thoughts.
may the difference (working/non-working) correlate with booting without/with USB stick plugged in? If yes, I can offer a vague explanation:
On boot up, the Sculpt manager looks through all connected storage devices to determine the Genode partition according to the policy described at [1]. In this phase, the Sculpt manager spawns a USB block device driver for each connected USB-storage device. As we noticed, USB storage seems to get stuck on your machine. Consequently, the probing phase won't ever finish, which prevents the Sculpt manager from entering its regular state of operation. The corresponding logic inside the Sculpt manager is implemented at [2] (should you wish to tweak it).
[1] https://genode.org/documentation/articles/sculpt-19-07#Making_customizations... [2] https://github.com/genodelabs/genode/blob/master/repos/gems/src/app/sculpt_m...
It just crossed my mind that it might be interesting to try out our recent componentized USB driver ('usb_host') on your machine. Sculpt still uses the old (monolithic) one currently, which is based on an older Linux kernel than the new (componentized) one. For reference, you can find an example run script demonstrating the new driver at [3]. Since you don't have any serial output (I guess your machine does not feature AMT, or does it?), I'm afraid that you might need to extend this run script with a framebuffer driver, nitpicker, nit_fb, terminal, and terminal_log to get meaningful information out of it, though.
[3] https://github.com/genodelabs/genode/blob/master/repos/dde_linux/run/usb_hid...
Cheers Norman
On 10/4/19 2:32 AM, Norman Feske wrote:
Hi John,
It just crossed my mind that it might be interesting to try out our recent componentized USB driver ('usb_host') on your machine. Sculpt still uses the old (monolithic) one currently, which is based on an older Linux kernel than the new (componentized) one. For reference, you can find an example run script demonstrating the new driver at [3]. Since you don't have any serial output (I guess your machine does not feature AMT, or does it?), I'm afraid that you might need to extend this run script with a framebuffer driver, nitpicker, nit_fb, terminal, and terminal_log to get meaningful information out of it, though.
[3] https://github.com/genodelabs/genode/blob/master/repos/dde_linux/run/usb_hid...
Yes, I would like to try that. I wonder if the addition of USB C requires a driver update.
In any case, I suspect a generic "log to display" scenario would be useful in other circumstances also.
Just off the top of your head, do you think the new USB driver could be dropped into Sculpt without too much headache?
I might try attacking this from both ends (i.e. building up the test scenarios vs. modifying Sculpt), and see which turns out to be easier.
Thanks again! -- John J. Karcher devuser@alternateapproach.com