regression (3beb971 -> a4a1fbe): fails to load fwupd

[julian-klode]

In Ubuntu, we recently updated our shim snapshot from 3beb971 to a4a1fbe - and we realized a bit after that that we cannot load fwupd anymore.

fwupdx64.efi is located in \EFI\ubuntu\ alongside shimx64.efi, it's bootloader entry is

Boot0001* Linux-Firmware-Updater        HD(1,GPT,a155bb6c-2d31-4a90-8678-f0f921e11e9b,0x800,0x100000)/File(\EFI\ubuntu\shimx64.efi)\.f.w.u.p.d.x.6.4...e.f.i...

This worked fine in commit 3beb971, but fails to load in commit a4a1fbe - shim just loads grub instead.

I tried launching fwupdx64.efi from an EFI shell as well, but shimx64.efi fwupdx64.efi also loaded grub (though that might be caused by #181).

I tried copying fwupdx64.efi to the root of the ESP, because there was a slash in front of it, but that did not seem to help either.

My next step would be to bisect this, but I'm going to need time to set this up first (wanna bisect in a VM, not on my actual machine), so I'm writing the bug first.

[superm1]

Before bothering to bisect, maybe try updating to a newer snapshot of shim?

[superm1]

That being said, a625fa5 and e563bc3 would be the most suspicious commits to me which haven't changed in any significant way in master.

[martinezjavier]

@julian-klode as @superm1 correctly pointed out, commit e563bc3 is the one to blame for this bug.

In the is_our_path() function, two UCS-2 strings are compared but I wrongly used strlen() to calculate the string length instead of using StrLen(). Sorry about that...

@vathpela fixed this bug some time ago with commit 1870bae ("Fix a use of strlen() instead of Strlen()").

[superm1]

@martinezjavier thanks for your confirmation and pointer on the fixed commit. I suppose this can be closed now then.

[martinezjavier]

@superm1 you are welcome. Yes, I'll close it now. Sorry for the inconvenience with this bug.

[julian-klode]

@superm1 @martinezjavier I'm trying to see if the mentioned commit fixes the bug, but it does not seem to help - fwupd is still not being loaded. To investigate this further, I'd like to look at this in a VM with a serial console so I can actually see the messages printed and don't have them fly past me. Any pointers? Any IRC channel to join?

[julian-klode]

Uh, um, this should have gone to the invalid fwupd bug I suppose, as I need to see if I can load fwupd from shim and hence need fwupd to install itself to the ESP and the boot order, which it does not do if there are no update-able devices.

[superm1]

@julian-klode Assuming you have an updatable device, you should be able to use the reinstall command in fwupd to try to reinstall the current firmware.

[julian-klode]

Oh I was dumb, and never actually updated the shim on the ESP, hence why it still failed, sorry for the noise. Time to finish the update and start signing request and stuff.