Dev/lineage 22.2

[awkoo]

No description provided.

[chatgpt-codex-connector]

💡 Codex Review

kernel_mi845_lineage/fs/exec.c

Line 1866 in 82b068a

return do_execveat_common(fd, filename, argv, envp, flags);

Restore KSU hook for execveat entrypoints

The KSU exec hook was moved out of do_execveat_common() and is now only called in do_execve()/compat_do_execve(), so direct execveat and compat_execveat calls bypass ksu_handle_execveat. This is a behavioral regression from the previous version (which hooked in the common path) and causes sucompat/ksud path rewriting to be skipped for callers that use execveat, leading to inconsistent root behavior depending on which syscall userspace chooses.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Re-apply KSU stat hook in shared vfs_fstatat path

ksu_handle_stat was removed from vfs_fstatat() and re-added only in specific syscall wrappers (newfstatat/fstatat64), which means other stat-family syscalls that route through vfs_stat()/vfs_lstat() no longer get rewritten. Previously, all those call paths were covered via vfs_fstatat; now stat/lstat (and compat variants that do not go through the wrapped syscalls) can bypass sucompat path translation, producing inconsistent behavior for /system/bin/su checks.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Avoid double-adjusting fstat size in syscall-table mode

This adds a direct ksu_handle_newfstat_ret call inside sys_newfstat, but syscall-table hooks already invoke the same helper after calling the original syscall (hook_aarch64_newfstat_ret/hook_armeabi_fstat64_ret). When CONFIG_KSU_TAMPER_SYSCALL_TABLE=y and CONFIG_KSU_KPROBES_KSUD=n (the non-kprobe path), the size patching runs twice and inflates st_size by 2 * ksu_rc_len, which corrupts the fake init.rc sizing KSU depends on.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Guard reboot hook to prevent duplicate KSU supercall handling

Calling ksu_handle_sys_reboot from sys_reboot under !CONFIG_KSU_KPROBES_KSUD overlaps with syscall-table hijack mode, where hook_aarch64_reboot/hook_armeabi_reboot already run the same handler before dispatching to the original syscall. In the CONFIG_KSU_TAMPER_SYSCALL_TABLE=y + CONFIG_KSU_KPROBES_KSUD=n configuration, KSU reboot supercalls are processed twice (for example, KSU fd install runs twice and creates duplicate FDs), so this should also exclude tamper mode.

Useful? React with 👍 / 👎.