vmem: Add support for lazily zero-initialized memory regions by syntactically · Pull Request #1873 · hyperlight-dev/hyperlight

syntactically · 2026-10-02T10:50:04Z

This is useful for hyperlight-wasm, although it can't fully provide for all use cases in a uniform way until we have better support for hugepages.

Copilot

Copilot review overview

🟡 Changes recommended

Snapshot compatibility, upper-half address decoding, and AArch64 instruction-fault handling need correction.

Review effort: Balanced
Findings: 1 · 2

Open (3)

Bump snapshot ABI for new ZeroInit page-table encoding · New
Restore sign extension for upper-half ZeroInit addresses · New
Handle instruction faults on executable ZeroInit pages · New

What changed in this PR

Adds lazy zero-initialized memory regions to Hyperlight’s virtual-memory system for hyperlight-wasm.

Changes:

Defines ZeroInit mappings and their page-table encodings.
Extends guest fault handlers to allocate pages on demand.
Preserves untouched lazy regions in snapshots without storing page contents.

File	Description
`src/hyperlight_host/src/sandbox/snapshot/mod.rs`	Preserves untouched ZeroInit mappings.
`src/hyperlight_host/src/mem/mgr.rs`	Converts ZeroInit permissions to region flags.
`src/hyperlight_guest_bin/src/arch/amd64/exception/handle.rs`	Handles lazy-page allocation on AMD64.
`src/hyperlight_guest_bin/src/arch/aarch64/exception/handle.rs`	Handles lazy-page allocation on AArch64.
`src/hyperlight_common/src/vmem.rs`	Defines ZeroInit and makes mappings copyable.
`src/hyperlight_common/src/arch/amd64/vmem.rs`	Encodes and decodes AMD64 lazy mappings.
`src/hyperlight_common/src/arch/aarch64/vmem.rs`	Encodes and decodes AArch64 lazy mappings.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

jprendes

LGTM

This is useful for hyperlight-wasm, although it can't fully provide for all use cases in a uniform way until we have better support for hugepages. Signed-off-by: Lucy Menon <168595099+syntactically@users.noreply.github.com>

hyperlight-gh-bot · 2026-10-02T12:59:27Z

Benchmark Results

Measured commit: 187f419dc1fb
Baseline commit: 43cf3539bd10

kvm / amd (Linux) (➖ stable)

No benchmark improved or regressed.

Benchmark Results

function_call_codec

	`encode_control`	`decode_vec_bytes_copy`
`byte_chunks`	`773.08 ns` (➖ 1.00x faster)
`vec_bytes`	`583.54 ns` (➖ 1.00x slower)
		`374.41 µs` (➖ 1.00x faster)

payload_allocation

	`slot_pool_segmented`
`262144`	`523.95 ns` (➖ 1.01x faster)
`65536`	`142.06 ns` (➖ 1.01x faster)

sandboxes

	`create_initialized_and_drop`
`medium`	`79.03 ms` (➖ 1.01x faster)

slot_pool

	`alloc_dealloc_1500`	`alloc_dealloc_4096`	`alloc_dealloc_128`
	`8.07 ns` (➖ 1.04x slower)	`7.72 ns` (➖ 1.01x slower)	`7.72 ns` (➖ 1.00x faster)

snapshot_files

	`load_snapshot_unverified`
`small`	`92.78 µs` (➖ 1.00x slower)

virtq_readonly

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`262144`	`7.23 µs` (➖ 1.00x faster)	`7.19 µs` (➖ 1.02x faster)
`65536`	`2.08 µs` (➖ 1.01x slower)	`1.93 µs` (➖ 1.00x slower)

virtq_readwrite

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`65536`	`6.25 µs` (➖ 1.01x faster)	`7.20 µs` (➖ 1.06x slower)
`8192`	`1.11 µs` (➖ 1.05x slower)	`1.07 µs` (➖ 1.01x slower)
`262144`		`26.73 µs` (➖ 1.02x faster)

kvm / intel (Linux) (➖ stable)

No benchmark improved or regressed.

Benchmark Results

function_call_codec

	`encode_control`	`decode_vec_bytes_copy`
`byte_chunks`	`691.38 ns` (➖ 1.09x faster)
`vec_bytes`	`512.73 ns` (➖ 1.06x faster)
		`659.40 µs` (➖ 1.05x faster)

payload_allocation

	`slot_pool_segmented`
`262144`	`507.87 ns` (➖ 1.13x faster)
`65536`	`137.10 ns` (➖ 1.12x faster)

sandboxes

	`create_initialized_and_drop`
`medium`	`75.79 ms` (➖ 1.05x faster)

slot_pool

	`alloc_dealloc_1500`	`alloc_dealloc_4096`	`alloc_dealloc_128`
	`6.95 ns` (➖ 1.08x faster)	`6.94 ns` (➖ 1.08x faster)	`6.95 ns` (➖ 1.08x faster)

snapshot_files

	`load_snapshot_unverified`
`small`	`46.42 µs` (➖ 1.07x faster)

virtq_readonly

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`262144`	`7.64 µs` (➖ 1.11x faster)	`7.71 µs` (➖ 1.11x faster)
`65536`	`2.16 µs` (➖ 1.11x faster)	`2.16 µs` (➖ 1.10x faster)

virtq_readwrite

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`65536`	`7.22 µs` (➖ 1.10x faster)	`7.23 µs` (➖ 1.09x faster)
`8192`	`741.69 ns` (➖ 1.13x faster)	`736.27 ns` (➖ 1.07x faster)
`262144`		`30.42 µs` (➖ 1.10x faster)

mshv3 / amd (Linux) (➖ stable)

No benchmark improved or regressed.

Benchmark Results

function_call_codec

	`encode_control`	`decode_vec_bytes_copy`
`byte_chunks`	`942.94 ns` (➖ 1.01x faster)
`vec_bytes`	`710.25 ns` (➖ 1.00x faster)
		`271.62 µs` (➖ 1.30x faster)

payload_allocation

	`slot_pool_segmented`
`262144`	`769.88 ns` (➖ 1.04x slower)
`65536`	`201.51 ns` (➖ 1.03x slower)

sandboxes

	`create_initialized_and_drop`
`medium`	`56.56 ms` (➖ 1.02x slower)

slot_pool

	`alloc_dealloc_1500`	`alloc_dealloc_4096`	`alloc_dealloc_128`
	`9.86 ns` (➖ 1.01x slower)	`10.54 ns` (➖ 1.08x slower)	`9.68 ns` (➖ 1.06x faster)

snapshot_files

	`load_snapshot_unverified`
`small`	`83.41 µs` (➖ 1.00x faster)

virtq_readonly

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`262144`	`9.80 µs` (➖ 1.00x slower)	`9.64 µs` (➖ 1.08x slower)
`65536`	`2.43 µs` (➖ 1.08x slower)	`2.24 µs` (➖ 1.10x faster)

virtq_readwrite

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`65536`	`8.26 µs` (➖ 1.01x slower)	`8.33 µs` (➖ 1.08x slower)
`8192`	`1.35 µs` (➖ 1.06x slower)	`1.30 µs` (➖ 1.02x faster)
`262144`		`36.25 µs` (➖ 1.01x faster)

mshv3 / intel (Linux) (➖ stable)

No benchmark improved or regressed.

Benchmark Results

function_call_codec

	`encode_control`	`decode_vec_bytes_copy`
`byte_chunks`	`934.96 ns` (➖ 1.00x faster)
`vec_bytes`	`681.88 ns` (➖ 1.02x slower)
		`676.65 µs` (➖ 1.01x slower)

payload_allocation

	`slot_pool_segmented`
`262144`	`631.31 ns` (➖ 1.01x faster)
`65536`	`166.21 ns` (➖ 1.01x slower)

sandboxes

	`create_initialized_and_drop`
`medium`	`66.72 ms` (➖ 1.07x slower)

slot_pool

	`alloc_dealloc_1500`	`alloc_dealloc_4096`	`alloc_dealloc_128`
	`8.81 ns` (➖ 1.01x slower)	`8.75 ns` (➖ 1.00x faster)	`11.24 ns` (➖ 1.13x slower)

snapshot_files

	`load_snapshot_unverified`
`small`	`45.01 µs` (➖ 1.01x slower)

virtq_readonly

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`262144`	`7.89 µs` (➖ 1.01x slower)	`8.00 µs` (➖ 1.00x faster)
`65536`	`2.28 µs` (➖ 1.02x slower)	`2.25 µs` (➖ 1.01x faster)

virtq_readwrite

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`65536`	`7.50 µs` (➖ 1.00x slower)	`7.58 µs` (➖ 1.00x slower)
`8192`	`863.65 ns` (➖ 1.00x slower)	`895.25 ns` (➖ 1.01x slower)
`262144`		`38.72 µs` (➖ 1.02x slower)

hyperv-ws2025 / amd (Windows) (➖ stable)

No benchmark improved or regressed.

Benchmark Results

function_call_codec

	`encode_control`	`decode_vec_bytes_copy`
`byte_chunks`	`1.13 µs` (➖ 1.00x slower)
`vec_bytes`	`751.22 ns` (➖ 1.27x faster)
		`2.03 ms` (➖ 1.23x faster)

payload_allocation

	`slot_pool_segmented`
`262144`	`804.65 ns` (➖ 1.01x faster)
`65536`	`235.53 ns` (➖ 1.02x slower)

sandboxes

	`create_initialized_and_drop`
`medium`	`81.46 ms` (➖ 1.35x faster)

slot_pool

	`alloc_dealloc_1500`	`alloc_dealloc_4096`	`alloc_dealloc_128`
	`10.54 ns` (➖ 1.31x faster)	`10.75 ns` (➖ 1.01x slower)	`10.53 ns` (➖ 1.00x faster)

snapshot_files

	`load_snapshot_unverified`
`small`	`612.68 µs` (➖ 1.30x faster)

virtq_readonly

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`262144`	`8.68 µs` (➖ 1.00x slower)	`9.89 µs` (➖ 1.06x slower)
`65536`	`2.52 µs` (➖ 1.00x faster)	`2.55 µs` (➖ 1.06x faster)

virtq_readwrite

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`65536`	`8.87 µs` (➖ 1.01x slower)	`9.20 µs` (➖ 1.04x slower)
`8192`	`1.39 µs` (➖ 1.01x faster)	`1.44 µs` (➖ 1.02x slower)
`262144`		`44.58 µs` (➖ 1.14x slower)

hyperv-ws2025 / intel (Windows) (➖ stable)

No benchmark improved or regressed.

Benchmark Results

function_call_codec

	`encode_control`	`decode_vec_bytes_copy`
`byte_chunks`	`1.27 µs` (➖ 1.24x slower)
`vec_bytes`	`896.76 ns` (➖ 1.10x slower)
		`3.39 ms` (➖ 1.09x slower)

payload_allocation

	`slot_pool_segmented`
`262144`	`741.21 ns` (➖ 1.01x slower)
`65536`	`211.63 ns` (➖ 1.01x faster)

sandboxes

	`create_initialized_and_drop`
`medium`	`111.49 ms` (➖ 1.08x faster)

slot_pool

	`alloc_dealloc_1500`	`alloc_dealloc_4096`	`alloc_dealloc_128`
	`10.08 ns` (➖ 1.00x slower)	`10.19 ns` (➖ 1.00x slower)	`10.60 ns` (➖ 1.00x faster)

snapshot_files

	`load_snapshot_unverified`
`small`	`625.57 µs` (➖ 1.04x slower)

virtq_readonly

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`262144`	`7.76 µs` (➖ 1.03x slower)	`7.72 µs` (➖ 1.01x faster)
`65536`	`2.35 µs` (➖ 1.03x slower)	`2.33 µs` (➖ 1.00x faster)

virtq_readwrite

	`slot_pool_segmented_fragmented`	`slot_pool_segmented`
`65536`	`8.65 µs` (➖ 1.04x slower)	`8.66 µs` (➖ 1.01x slower)
`8192`	`1.14 µs` (➖ 1.10x faster)	`1.22 µs` (➖ 1.09x slower)
`262144`		`56.97 µs` (➖ 1.35x slower)

_{Reported by cargo ci bench-report --candidate run:37007006067 --baseline run:36945564806 --config-file bench_report.toml.}

Copilot AI balanced review requested due to automatic review settings October 2, 2026 10:50

syntactically requested review from andreiltd, danbugs, dblnz, devigned, jprendes, jsturtevant, ludfjig, simongdavies and squillace as code owners October 2, 2026 10:50

syntactically added C-enhancement For PRs adding features, improving functionality, docs, tests, etc. S-needs-review PR is ready for (re-)review labels Oct 2, 2026

syntactically added this pull request to stack #1875 October 2, 2026 10:51

Copilot started reviewing on behalf of syntactically October 2, 2026 10:53

View session

Copilot AI reviewed Oct 2, 2026

View reviewed changes

Comment thread src/hyperlight_host/src/sandbox/snapshot/mod.rs Resolved

Comment thread src/hyperlight_common/src/arch/amd64/vmem.rs Resolved

Comment thread src/hyperlight_guest_bin/src/arch/aarch64/exception/handle.rs Resolved

jprendes previously approved these changes Oct 2, 2026

View reviewed changes

This comment has been minimized.

Sign in to view

vmem: Add support for lazily zero-initialized memory regions

187f419

This is useful for hyperlight-wasm, although it can't fully provide for all use cases in a uniform way until we have better support for hugepages. Signed-off-by: Lucy Menon <168595099+syntactically@users.noreply.github.com>

syntactically dismissed jprendes’s stale review via 187f419 October 2, 2026 12:29

syntactically force-pushed the lm/zero-init-rgns branch from bf7521b to 187f419 Compare October 2, 2026 12:29

jprendes approved these changes Oct 2, 2026

View reviewed changes

syntactically merged commit 187f419 into main Oct 2, 2026
62 checks passed

syntactically deleted the lm/zero-init-rgns branch October 2, 2026 14:55

github-actions Bot removed the S-needs-review PR is ready for (re-)review label Oct 2, 2026

Conversation

syntactically commented Oct 2, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jprendes left a comment

Choose a reason for hiding this comment

Uh oh!

This comment has been minimized.

hyperlight-gh-bot Bot commented Oct 2, 2026

Benchmark Results

Benchmark Results

function_call_codec

payload_allocation

sandboxes

slot_pool

snapshot_files

virtq_readonly

virtq_readwrite

Benchmark Results

function_call_codec

payload_allocation

sandboxes

slot_pool

snapshot_files

virtq_readonly

virtq_readwrite

Benchmark Results

function_call_codec

payload_allocation

sandboxes

slot_pool

snapshot_files

virtq_readonly

virtq_readwrite

Benchmark Results

function_call_codec

payload_allocation

sandboxes

slot_pool

snapshot_files

virtq_readonly

virtq_readwrite

Benchmark Results

function_call_codec

payload_allocation

sandboxes

slot_pool

snapshot_files

virtq_readonly

virtq_readwrite

Benchmark Results

function_call_codec

payload_allocation

sandboxes

slot_pool

snapshot_files

virtq_readonly

virtq_readwrite

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants