fix(install): warn for known affected ublk kernels

[catyans]

Summary

• warn from scripts/install.sh when the running kernel is a known affected ublk release
• emit the same warning before Docker host setup loads and persists ublk_drv
• keep installation non-blocking while explaining the hang/panic risk and fixed-kernel guidance

Scope

This revision follows the maintainer guidance on #20 and replaces the earlier
runtime guard with an installation-only warning. The PR now changes only:

• scripts/install.sh
• scripts/docker-setup.sh

The delimiter-aware shell patterns cover the known upstream 6.18.4 / 6.18.5
families and Ubuntu Azure 6.17.0-1015 / 6.17.0-1018, including - and +
local-version suffixes without overmatching releases such as 6.18.40.

Validation

• git diff --check
• GitHub AENV CI, including installer verification

Fixes #20

[github-actions]

🔍 OpenCodeReview found 1 issue(s) in this PR.

• ✅ Successfully posted inline: 1 comment(s)

[catyans]

Addressed the latest suffix-matching review and completed an independent Linux validation on head e65b73f:

• cargo fmt --all -- --check ✅
• cargo test -p agentenv setup::ublk::tests -- --nocapture ✅ (3 passed, 0 failed)

The targeted cases include Azure kernel releases with -custom / +debug suffixes as affected, while near-prefix strings without a valid delimiter remain unaffected.

[catyans]

@yingdi-shan This is ready for maintainer review when convenient. The current head is mergeable; AENV CI, unit tests, clippy, both E2E jobs, agentenv tests, and OpenCodeReview are all green, and both review threads are resolved.

[yingdi-shan]

Thank you for your contribution! I left a comment on issue #20. I suggest updating only the installation script instead of modifying the entire script, as this keeps the change more focused.

[catyans]

Thanks, I narrowed the PR to that scope in 6d1ae9e. The previous Rust runtime guard, override, and documentation changes have been removed; the diff now only adds a non-blocking warning to scripts/install.sh and scripts/docker-setup.sh. The warning covers the known upstream and Azure affected release families, including delimiter-valid local-version suffixes. A refreshed CI run is starting against current main.

[github-actions]

[bug · high]
The matched-kernel path only emits a warning and then continues. With the default SKIP_SETUP=0, this script subsequently invokes server --setup-host, which provisions ublk and may trigger the exact host hang/panic described here. Abort before setup on affected kernels (optionally allowing an explicit unsafe override), rather than proceeding automatically.

Suggestion:

Suggested change

	cat >&2 <<EOF
	warning: Linux kernel ${KERNEL_RELEASE} is known to contain a ublk initialization regression.
	Creating the first ublk device may hang or panic the host.
	Upgrade to a fixed vendor kernel (upstream 6.18.6+ for 6.18.y) before running AgentENV.
	See: https://github.com/kvcache-ai/AgentENV/issues/20
	EOF
	;;
	cat >&2 <<EOF
	error: Linux kernel ${KERNEL_RELEASE} is known to contain a ublk initialization regression.
	Creating the first ublk device may hang or panic the host.
	Upgrade to a fixed vendor kernel (upstream 6.18.6+ for 6.18.y) before running AgentENV.
	See: https://github.com/kvcache-ai/AgentENV/issues/20
	EOF
	exit 1
	;;

[yingdi-shan]

We can add 6.17.0-1020-azure. Other kernels like 6.17.0-1018-aws also contain the vulnerability.

[yingdi-shan]

This code is duplicated with install.sh. Maybe we can add a file with affeted kernel version and read kernel versions from that file?