[Bug 1934192] Re: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64
After switching to the Mellanox NIC, I'm no longer seeing any issues. I therefore suspect the issue is with the Intel UEFI driver - or some other firmware - and not with the bootloader. ** Changed in: grub2-unsigned (Ubuntu) Status: New => Invalid ** Changed in: grub2-signed (Ubuntu) Status: New => Invalid -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1934192 Title: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/grub2-signed/+bug/1934192/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 1934192] Re: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64
One hypothesis is that the network stack/UEFI driver dies after loading the initial GRUB image. To test that, I started a ping at the server until it dropped to the GRUB prompt. Now, I never saw a single ping response - even while the server was successfully TFTP'ing GRUB from MAAS. Maybe it just has ping responses disabled. But what I did notice is that the GRUB process, which is sitting at a `grub>` prompt, becomes unresponsive while the system is being pinged. The failing system happens to have 2 models of NICs in it - Intel and Mellanox. I happen to be seeing this with the Intel NIC. I'll try to get a Mellanox one wired up to see if the issue follows the NIC model. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1934192 Title: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/grub2-signed/+bug/1934192/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 1934192] Re: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64
I bisected binutils and found the significant change that avoids the problem to be: commit d1bcae833b32f1408485ce69f844dcd7ded093a8 (HEAD, refs/bisect/bad) Author: H.J. Lu Date: Thu Jan 7 06:42:00 2021 -0800 ELF: Don't generate unused section symbols Now, this doesn't appear to be a bug fix per se, more of an optimization. I'm guessing it is just avoiding the underlying issue as a side-effect. I found this commit[*] in GRUB which purports to fix an arm64-specific issue with a patch we are carrying, and describes issues with section sizes: util/mkimage: Fix wrong PE32+ section sizes for some arches The commit f60ba9e5945 (util/mkimage: Refactor section setup to use a helper) added a helper function to setup PE sections. But it also changed how the raw data offsets were calculated since all the section sizes are aligned. However, for some platforms, i.e ia64-efi and arm64-efi, the kernel image size is not aligned using the section alignment. This leads to the situation in which the mods section offset in its PE section header does not match its real placement in the PE file. So, finally the GRUB is not able to locate and load built-in modules. While I'm not setting those exact symptoms (memdisk modules *do* load), I tested it to see if it addresses this issue. Unfortunately it does not. [*] https://git.savannah.gnu.org/cgit/grub.git/commit/?id=c0e647eb0e2bd09315612446cb4d90f7f75cb44c -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1934192 Title: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/grub2-signed/+bug/1934192/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 1934192] Re: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64
binutils 2.35.50.20210106-1ubuntu2 = BAD binutils 2.35.90.20210120-1ubuntu1 = GOOD So possibly a change between 2.35.50 and 2.35.90 -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1934192 Title: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/grub2-signed/+bug/1934192/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 1934192] Re: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64
Rebuilding in focal, but with hirsute's binutils makes the issue go away. Now to dig into binutils changes -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1934192 Title: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/grub2-signed/+bug/1934192/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 1934192] Re: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64
Another datapoint - the system that sees this issue is a Gigabyte G242-P30-JG. But I can not reproduce on an Ampere Mt. Jade system. Both use the Ampere Altra SoC. Gigabyte firmware: F07 03/22/2021 14:48:30 Mt. Jade firmware: 2021/05/26 17:35:33 Ver: 0ACOI 1.6.20210526 -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1934192 Title: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/grub2-signed/+bug/1934192/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 1934192] Re: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64
@juliank pointed out that 2.04-1ubuntu44 was built w/ the hirsute toolchain, 2.04-1ubuntu44.2 with focal's toolchain. I'll run some experiments to see if the issue follows the toolchain. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1934192 Title: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/grub2-signed/+bug/1934192/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
[Bug 1934192] Re: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64
Turns out the *unsigned* 2.04-1ubuntu44.2 binary boots fine - the issue only seems to occur w/ the signed one. I also tried the 2.04ubuntu44.2 grubnetaa64.efi.signed from -updates, which has a different checksum (different sig?), and the failure also occurs there. To be clear, this system does not have SecureBoot enabled, and does not appear to be passing through shim. ** Also affects: grub2-signed (Ubuntu) Importance: Undecided Status: New -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1934192 Title: regression between 2.04-1ubuntu44 and 2.04-1ubuntu44.2 breaks netboot on arm64 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/grub2-signed/+bug/1934192/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs