Add crashdump example and include snapshot/scratch in core dumps#1264
Add crashdump example and include snapshot/scratch in core dumps#1264jsturtevant wants to merge 8 commits intohyperlight-dev:mainfrom
Conversation
jsturtevant
requested review from
danbugs,
dblnz,
devigned,
jprendes,
ludfjig,
marosset,
simongdavies and
syntactically
as code owners
|
@jsturtevant I believe you were working on another version of this that explicitly walks through the whole guest virtual address space? |
Yes, been working with @dblnz on getting it working. Should have something today |
jsturtevant
force-pushed
the
crashdump
branch
6 times, most recently
from
21b7183 to
137b016
Compare
dblnz
left a comment
dblnz
left a comment
There was a problem hiding this comment.
LGTM! Having participated in these changes, I'd feel more comfortable if other people had a look/approved this 😄
ludfjig
left a comment
ludfjig
left a comment
There was a problem hiding this comment.
this is great, i like the added tests!
syntactically
left a comment
syntactically
left a comment
There was a problem hiding this comment.
The direction of this looks good! I have a few comments, most of which are minor nits or just signposting things I thought were interesting points to discuss; the only thing that I'm actually particularly concerned about with the present state of the code is the mmap regions being present in the dump at VAs where they are not actually present in the sandbox.
At a higher level, it occurs to me that an alternative for crashdumps might be to just take a snapshot, and then write a separate utility that prepares a crashdump from a snapshot. That's probably worse, since in the future we might not be able to take a snapshot at arbitrary times, but only when I/O to the guest is quiet? However, I wanted to mention the idea to see what anyone else thought.
Co-authored-by: Doru Blânzeanu <dblnz@pm.me> Signed-off-by: James Sturtevant <jsturtevant@gmail.com>
Signed-off-by: James Sturtevant <jsturtevant@gmail.com>
The runtime config was passed by reference into set_up_hypervisor_partition then immediately cloned, but no caller needs it afterward so it is now passed by value. The entry point field uses Option<u64> instead of a bare zero default so a missing value is detectable rather than silently producing a bogus AT_ENTRY. Signed-off-by: James Sturtevant <jsturtevant@gmail.com>
generate_crashdump_to_dir accepts the output directory as a parameter instead of requiring callers to set the HYPERLIGHT_CORE_DUMP_DIR environment variable. This removes the need for unsafe std::env::set_var in tests while preserving the existing env-var fallback path for the automatic dump and the no-argument generate_crashdump method. Signed-off-by: James Sturtevant <jsturtevant@gmail.com>
Signed-off-by: James Sturtevant <jsturtevant@gmail.com>
Signed-off-by: James Sturtevant <jsturtevant@gmail.com>
Signed-off-by: James Sturtevant <jsturtevant@gmail.com>
Signed-off-by: James Sturtevant <jsturtevant@gmail.com>
5 participants
Core dumps generated by Hyperlight were missing the snapshot and scratch memory regions, making post-mortem debugging with GDB incomplete — register state was present but the guest's code, stack, heap, and page tables were absent. This adds the snapshot and scratch regions to the ELF core dump alongside any dynamically mapped regions so that GDB can show full backtraces, disassemble at the crash site, and inspect guest memory. A new runnable crashdump example demonstrates automatic dumps (VM-level faults), on-demand dumps (guest-caught exceptions), and per-sandbox opt-out, with GDB-based integration tests that validate register and memory content in the generated ELF files. The debugging docs are also updated with practical GDB commands for inspecting crash dumps.