[25.12] hev-socks5-tunnel: allow multiple instances

[Ser9ei]

Add support for running multiple service instances.

• switch UCI config from a single named section to "instance" sections
• update the procd init script to start multiple instances
• migrate existing single-instance configurations during package upgrade (uci-defaults script)
• bump PKG_RELEASE

(cherry picked from commit d04b36b)

📦 Package Details

Maintainer: @heiher
_{(You can find this by checking the history of the package Makefile.)}

Description:

---

🧪 Run Testing Details

• OpenWrt Version: 25.12.5
• OpenWrt Target/Subtarget: mediatek/filogic
• OpenWrt Device: Xiaomi Mi Router AX3000T

[openwrt-ai]

Backport review: the diff itself is a clean cherry-pick. net/hev-socks5-tunnel/ at this head is byte-identical to the same directory on master after the upstream commit (Makefile, 90-hev-socks5-tunnel, .config, .init all match), and the base branch already carries the 2.17.0 update the change sits on top of, so no prerequisite is missing. One issue with the trailer:

Commit checks

• 0a11982 "hev-socks5-tunnel: allow multiple instances" — the (cherry picked from commit f9fdfcf09a3a79251ccd88853cbda6331fe43722) trailer points at a commit that is not on master. f9fdfcf is the pre-merge commit from the contributor branch (committer date 2026-08-07, committed by the author). The commit that actually landed on master is d04b36bbba607e65072ecd0abfae5a89af468607 (committer date 2026-08-09, committed by the maintainer who merged it). The tree content of the two is identical, so nothing in the code is affected — but the trailer is meant to let someone follow the change back to master, and this SHA does not resolve there. Redoing the pick as git cherry-pick -x d04b36bbba607e65072ecd0abfae5a89af468607 off master writes the correct trailer automatically.

Nothing else to flag: per backport posture, any remaining style/design questions about the multi-instance rework belong on a fix-to-master PR, not here.

---

Generated by Claude Code

[Ser9ei]

The CI failure appears to be unrelated to this PR. One of the CI tests fails while downloading a package index from the OpenWrt download server.
The same failure is reproducible on other PRs, while the other CI jobs pass.
No changes to this PR are required.

[openwrt-ai]

Reviewed 1 new commit; no new issues found.

The trailer issue from my previous review is resolved: edee87c now carries (cherry picked from commit d04b36bbba607e65072ecd0abfae5a89af468607), and that SHA does resolve on master as the merged commit (d04b36b, committed by the maintainer who merged the upstream PR). The tree is unchanged from the previously reviewed head — git diff 0a11982..edee87c is empty — so this force-push only rewrote the commit message.

Re-confirmed on the new head:

• net/hev-socks5-tunnel/{Makefile,files/90-hev-socks5-tunnel,files/hev-socks5-tunnel.config,files/hev-socks5-tunnel.init} are byte-identical to the same files on master, so the pick has no divergence from upstream.
• The base branch already carries the 2.17.0 update this change sits on (PKG_VERSION:=2.17.0, PKG_RELEASE:=1 on openwrt-25.12, going to 2 here — same version/release pair as master), so no prerequisite commit is missing.

Per backport posture, any remaining style or design questions about the multi-instance rework itself belong on a fix-to-master PR rather than here.

---

Generated by Claude Code

[Ser9ei]

@commodo, could you please help merge the backport for 25.12 and take a look at the similar one for 24.10 - #30271?

[openwrt-ai]

Reviewed 1 new commit; no new issues found.

Since my last review at edee87c, the branch was rebased onto the current openwrt-25.12 tip (dda798f's parent is 64b6155). Nothing about the change itself moved:

• The patch this commit applies to net/hev-socks5-tunnel/ is byte-identical to the previously reviewed one (git show edee87c -- net/hev-socks5-tunnel vs the same on dda798f is empty), and the commit message is unchanged. The rebase re-committed it (committer is now the maintainer who re-pushed), which is why the SHA changed.
• All four files still match the upstream commit exactly — the blob hashes of Makefile, files/90-hev-socks5-tunnel, files/hev-socks5-tunnel.config and files/hev-socks5-tunnel.init at dda798f are identical to those at d04b36b on master, so the pick still has no divergence.
• The (cherry picked from commit d04b36bbba607e65072ecd0abfae5a89af468607) trailer is intact and still resolves: d04b36b is the commit on master that touched files/hev-socks5-tunnel.init (authored 2026-08-02 by Sergei S., committed 2026-08-09 by Alexandru Ardelean).
• No prerequisite is missing after the rebase — the new base still carries PKG_VERSION:=2.17.0 / PKG_RELEASE:=1, so the bump to 2 here lands on the same version/release pair as master.

CI on this head is clean: the three FormalityCheck runs and Wait for FormalityCheck all passed, and the ten Feeds Package Test Build jobs are queued rather than failed. The earlier download-server failure the author noted is not present on this head.

Per backport posture, any remaining style or design questions about the multi-instance rework itself belong on a fix-to-master PR rather than here.

---

Generated by Claude Code

[Ser9ei]

Maybe as a small CI improvement/workaround, apk update could be retried a few times with a short delay between attempts. The current failure looks like a transient/incomplete download of packages.adb (wget: exited with error 8, unexpected end of file), so retries might help avoid false-negative CI failures caused by temporary issues with the snapshot feed.