iframe-proxy

ehds · 2026-10-13T11:34:46Z

What problem does this PR solve?

Issue Number: resolve

Problem Summary:

Collecting process metrics through commands such as ps and lsof may fail due to permission restrictions in MacOS sandboxed applications.

brpc/src/bvar/default_variables.cpp:127 read_proc_status] Fail to sscanf: Undefined error: 0
sh: /bin/ps: Operation not permitted

The previous implementation also leaves several process statistics unavailable, like process_faults_minor_second, process_faults_minor_second, process_faults_major.

What is changed and the side effects?

Changed:

This change uses native macOS APIs to collect these metrics:

getpid(), getppid(), getpgrp(), and getsid() for process and session IDs.
getpriority() for the process nice value.
proc_pidinfo() with PROC_PIDTBSDINFO for process flags, state, and terminal information.
proc_pidinfo() with PROC_PIDTASKINFO for priority and thread count.
getrusage() for process and waited-child page faults and CPU times, converted to clock ticks using sysconf(_SC_CLK_TCK).
task_info() with MACH_TASK_BASIC_INFO for virtual and resident memory sizes.
proc_pidinfo() with PROC_PIDLISTFDS for open file descriptor counts.

Side effects:

Performance effects:
Breaking backward compatibility:

Check List:

Please make sure your changes are compilable.
When providing us with a new feature, it is best to add related tests.
Please follow Contributor Covenant Code of Conduct.

Copilot

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

4 open findings

stat is no longer zero-initialized on macOS (the previous implementation did `memset(&stat, 0,… · New
On failure, the function now returns false without clearing m, whereas the previous code path… · New
The return value assumes bytes is always a clean multiple of sizeof(proc_fdinfo). If… · New
This log line prints only the numeric kern_return_t, which is difficult to interpret… · New

What changed in this PR

This PR replaces macOS process-metrics collection that relied on shelling out to ps/lsof with native macOS APIs to improve reliability under sandbox/permission restrictions and to expose additional metrics.

Changes:

Reworked macOS read_proc_status to use getpid/getppid/getpgrp/getsid, getpriority, proc_pidinfo, and getrusage.
Reworked macOS read_proc_memory to use task_info(MACH_TASK_BASIC_INFO) for RSS/VSS.
Implemented macOS FD counting via proc_pidinfo(PROC_PIDLISTFDS) and adjusted FD scan-limit condition.

🧠 Review effort: Lite

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

ehds · 2026-10-14T05:32:34Z

+    const pid_t pid = getpid();
+    stat.pid = pid;
+    stat.ppid = getppid();
+    stat.pgrp = getpgrp();
+    const pid_t session = getsid(0);
+    stat.session = session >= 0 ? session : 0;
+    errno = 0;
+    const int nice = getpriority(PRIO_PROCESS, 0);
+    if (errno == 0)
+      stat.nice = nice;
+    struct proc_bsdinfo bsd = {};


stat is already value-initialized at the beginning of read_proc_status() via stat = ProcStat().

ehds · 2026-10-14T05:32:35Z

+    mach_task_basic_info_data_t info = {};
+    mach_msg_type_number_t count = MACH_TASK_BASIC_INFO_COUNT;
+    const kern_return_t result = task_info(
+        mach_task_self(), MACH_TASK_BASIC_INFO,
+        reinterpret_cast<task_info_t>(&info), &count);
+    if (result != KERN_SUCCESS) {
+        LOG_ONCE(WARNING) << "Cannot read own task memory, Mach error=" << result;
        return false;
    }


m has been zero-initialized at the beginning of m = ProcMemory(), it won’t be partially assigned when return false here.

ehds · 2026-10-14T05:33:19Z

-    return std::min(count, limit);
-    */
-    return 0;
+    return bytes / sizeof(proc_fdinfo);


Added validation that the returned byte count is a multiple of sizeof(proc_fdinfo) and does not exceed the buffer size. Unexpected lengths are logged and return -1, allowing the reader to retain the last valid count.
Since the buffer holds exactly limit entries, a validated result cannot exceed limit, so an additional clamp is unnecessary.

ehds · 2026-10-14T05:32:39Z

+        mach_task_self(), MACH_TASK_BASIC_INFO,
+        reinterpret_cast<task_info_t>(&info), &count);
+    if (result != KERN_SUCCESS) {
+        LOG_ONCE(WARNING) << "Cannot read own task memory, Mach error=" << result;


Copilot

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

6 open findings

This lambda does not capture ticks_per_second, so it won’t compile (use of an enclosing-scope… · New
stat is no longer zero-initialized on macOS (the previous implementation did `memset(&stat, 0,…
This allocates a std::vector on every call, which can add avoidable CPU/memory churn if FD… · New
The return value assumes bytes is always a clean multiple of sizeof(proc_fdinfo). If…
On failure, the function now returns false without clearing m, whereas the previous code path…
This log line prints only the numeric kern_return_t, which is difficult to interpret…

🧠 Review effort: Lite

ehds · 2026-10-14T07:28:22Z

+    static const long ticks_per_second = sysconf(_SC_CLK_TCK);
+    const auto to_ticks = [](const timeval &time) -> unsigned long {
+      if (ticks_per_second <= 0) {
+        return 0;
+      }
+      return static_cast<uint64_t>(time.tv_sec) * ticks_per_second +
+             static_cast<uint64_t>(time.tv_usec) * ticks_per_second / 1000000;
+    };


ticks_per_second is a local static variable which caches the constant tick rate, avoiding repeated sysconf() calls, so the lambda can access it without capturing it.

ehds · 2026-10-14T07:30:49Z

+    std::vector<proc_fdinfo> fds(limit);
+    errno = 0;
+    // This query opens no additional FD.
+    const int bytes = proc_pidinfo(getpid(), PROC_PIDLISTFDS, 0,
+                                  fds.data(), fds.size() * sizeof(proc_fdinfo));


Changed to a thread_local std::vector to reuse the allocation while keeping each thread’s buffer separate.

ehds changed the title ~~Hds/native macos api~~ Use native APIs for macOS process metrics Oct 13, 2026

Use native APIs for macOS process metrics

44986d5

ehds force-pushed the hds/native_macos_api branch from 4f77516 to 44986d5 Compare October 13, 2026 11:41

chenBright requested a lite review from Copilot October 13, 2026 13:55

Copilot AI reviewed Oct 13, 2026

View reviewed changes

Copilot stopped reviewing on behalf of chenBright due to an error October 13, 2026 19:42
The job was not acquired by a runner even after multiple attempts.

Validate macOS FD results and improve Mach error logs

a8189cb

wwbmmm requested a lite review from Copilot October 14, 2026 06:05

Copilot AI reviewed Oct 14, 2026

View reviewed changes

Copilot stopped reviewing on behalf of wwbmmm due to an error October 14, 2026 06:50
The job was not acquired by a runner even after multiple attempts.

make fds thread_local avoid repeating allocate memory

5d26e83

Sunbelt Computer Software

PL/B Language Development and Support

Navigation Menu

Repository navigation

Use native APIs for macOS process metrics - #3602

Use native APIs for macOS process metrics#3602
ehds wants to merge 3 commits into
apache:masterfrom
ehds:hds/native_macos_api

ehds commented Oct 13, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

ehds Oct 14, 2026 •

edited

Loading

Uh oh!

ehds Oct 14, 2026

Uh oh!

ehds Oct 14, 2026

Uh oh!

ehds Oct 14, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

ehds Oct 14, 2026

Uh oh!

ehds Oct 14, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Sunbelt Computer Software

PL/B Language Development and Support

Conversation

ehds commented Oct 13, 2026

What problem does this PR solve?

What is changed and the side effects?

Check List:

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

Uh oh!

ehds Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ehds Oct 14, 2026

Choose a reason for hiding this comment

Uh oh!

ehds Oct 14, 2026

Choose a reason for hiding this comment

Uh oh!

ehds Oct 14, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

Uh oh!

ehds Oct 14, 2026

Choose a reason for hiding this comment

Uh oh!

ehds Oct 14, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

ehds Oct 14, 2026 •

edited

Loading