Conversation
On arm, where FileLoad is a stub, kexec printed only "function not implemented", after an "invalid FDT magic" line from the universal payload probe. Neither says what failed or what to do about it. - Name kexec_file_load and the architecture in the ENOSYS stub error. - When kexec_file_load fails with ENOSYS or ENOEXEC, suggest --loadsyscall. ENOEXEC is what riscv returns for an Image before Linux 6.16. - Only log the universal payload probe failure with -d when the file is not an FDT at all, since that is the common case. - Drop "(default true)" from the -d and -L docs; both default to false. On arm the output goes from Failed to load universalpayload (failed to read fdt file: ... invalid FDT magic, got 0x0000a0e1, expected 0xd00dfeed), try legacy kernel.. function not implemented to SYS_kexec_file_load is not supported on arm: function not implemented; try --loadsyscall to use kexec_load instead Fixes u-root#3384 Signed-off-by: Hung-Chun Tseng <alan.tseng.cs@gmail.com>
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## main #3774 +/- ##
==========================================
- Coverage 61.46% 61.33% -0.14%
==========================================
Files 648 648
Lines 45716 45727 +11
==========================================
- Hits 28099 28045 -54
- Misses 17617 17682 +65
Flags with carried forward coverage won't be shown. Click here to find out more.
🚀 New features to boost your workflow:
|
rminnich
reviewed
Oct 3, 2026
| return err | ||
| } | ||
| if errors.Is(err, syscall.ENOSYS) || errors.Is(err, syscall.ENOEXEC) { | ||
| return fmt.Errorf("%w; try --loadsyscall to use kexec_load instead", err) |
Member
There was a problem hiding this comment.
I sometimes get complaints from tools that the %w should be last, not first; can you make it last?
| } | ||
| } else { | ||
| // universalpayload package suppresses warning message, we print messages here. | ||
| if warningMsg != nil { |
Member
There was a problem hiding this comment.
I am sad to see packages returning nil pointers, it's going to catch us sooner or later. Any chance of fixing that package?
| // case. Only mention it when debugging. | ||
| linux.Debug("%s is not a universal payload (%v), loading it as a kernel", opts.kernelpath, err) | ||
| } else { | ||
| log.Printf("Failed to load universalpayload (%v), try legacy kernel..", err) |
Member
There was a problem hiding this comment.
if this is an error, shouldn't it just return here? This thicket of if ... else might be reduced a bit.
rminnich
added
the
Awaiting author
Member
|
Awesome, thanks for looking into this! I just referenced this PR in the corresponding issue I had filed a good while ago. |
Address review on u-root#3774: - A file that is a universal payload but fails to load or execute now returns the error instead of falling back to loading it as a kernel, which could only fail again with a less useful message. The if/else becomes a switch. - Move %w to the end of the --loadsyscall hint. Signed-off-by: Hung-Chun Tseng <alan.tseng.cs@gmail.com>
Contributor
Author
|
Thanks. Pushed 1942ca9:
|
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #3384.
On arm, where
FileLoadis a stub,kexecprinted onlyfunction not implemented, after aninvalid FDT magicline from the universal payload probe. Neither says what failed or what to do about it.kexec_file_loadand the architecture in the ENOSYS stub error (still wrapssyscall.ENOSYS).kexec_file_loadfails with ENOSYS or ENOEXEC for a Linux image, suggest--loadsyscall. ENOEXEC is what riscv returns for anImagebefore Linux 6.16 (kexec_file_load error #3236).-dwhen the file is not an FDT at all (ErrFailToReadFdtFile), since that is the common case. A file that is a universal payload but fails to load or exec now returns that error instead of falling back to the legacy path.(default true)from the-dand-Ldocs; both default to false.Before (arm32 vmtest kernel,
kexec -l -c KEXEC=Y /kernel):After:
and on riscv64 with the 6.6 vmtest kernel:
Testing: added
TestLoadError;go test ./cmds/core/kexec ./pkg/boot ./pkg/boot/kexec ./pkg/boot/universalpayloadpasses, and the outputs above are from QEMU runs underrunvmtestwith VMTEST_ARCH=arm and riscv64.