FreeBSD: add an I/O screen with per-process I/O rates

[samm-git]

What

Adds a second default screen, I/O, on FreeBSD with three new process columns:

• IO_RATE (IO R/W)
• IO_READ_RATE (IO READ)
• IO_WRITE_RATE (IO WRITE)

Linux and PCP already ship such a screen; FreeBSD only had Main.

How

The rates are computed from the per-process rusage block-operation counters
(ki_rusage.ru_inblock / ru_oublock) that are already returned by the existing
kvm_getprocs() call, i.e. the same data source the base system top(1) uses for
its -m io mode. No new syscalls and no privileges required.

Because these counters count block operations rather than bytes, the columns
report operations per second, unlike their Linux counterparts (FreeBSD exposes
no unprivileged per-process byte counters). The manual page notes this difference.

Counters are refreshed on every scan, so switching to the I/O screen shows valid
rates immediately rather than a lifetime-sized first sample.

Testing

• ./configure --enable-unicode --enable-werror && make on FreeBSD 14.5 (clang) - clean.
• Ran the binary: the I/O tab appears and shows live rates.
• Fresh configs generate both screens; existing configs keep their own screens
  (default screens are only created when no screens section is present).

Notes

• IO_PRIORITY, PERCENT_IO_DELAY and PERCENT_SWAP_DELAY are intentionally not
  added: FreeBSD has no equivalent per-process I/O priority or delay accounting.

Assisted-by: DeepSeek V4.1 Flash

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

• Configuration used: Repository UI
• Review profile: ASSERTIVE
• Plan: Advanced
• Run ID: 281d001c-b4c1-4d28-b5c2-d0ef1044d7d7

📥 Commits

Reviewing files that changed from the base of the PR and between 629a82b and d1474e7.

📒 Files selected for processing (8)

• Row.c
• Row.h
• freebsd/FreeBSDProcess.c
• freebsd/FreeBSDProcess.h
• freebsd/FreeBSDProcessTable.c
• freebsd/Platform.c
• freebsd/ProcessField.h
• htop.1.in

Included review availability: This review used your included allowance. Your plan provides up to 8 included reviews per hour; 7 remain after this review.

---

📝 Walkthrough

Walkthrough

FreeBSD now calculates per-process block read and write operation rates from sampled counters. It provides read, write, and combined rate fields with display and sorting support. The default screens include an I/O screen, and the manual documents the FreeBSD rate units.

Suggested reviewers: benbe

Priority: ⬇️ Low

Change: Feature

Merge Risk: ⚪ Minimal · up to d1474

The previously identified PID-reuse and column-alignment issues appear corrected. No actionable merge-blocking risk remains after normal checks.

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Counters mark the passing time
Read and write rates take their place
A total joins the measured flow
Columns sort and values show
FreeBSD’s I/O screen comes to life}

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 2

---

ℹ️ Review info

⚙️ Run configuration

• Configuration used: Repository UI
• Review profile: ASSERTIVE
• Plan: Advanced
• Run ID: 0c40b2e7-0c6e-4b05-8a82-628e7878f360

📥 Commits

Reviewing files that changed from the base of the PR and between 44a59cf and 62eb8b0.

📒 Files selected for processing (6)

• freebsd/FreeBSDProcess.c
• freebsd/FreeBSDProcess.h
• freebsd/FreeBSDProcessTable.c
• freebsd/Platform.c
• freebsd/ProcessField.h
• htop.1.in

Included review availability: This review used your included allowance. Your plan provides up to 8 included reviews per hour; 7 remain after this review.

[BenBE]

Are the normal IO rate values (B/s) available too or could that be added?

[samm-git]

Are the normal IO rate values (B/s) available too or could that be added?

Not really

[BenBE]

@samm-git Full review of the PR will take some time. I noted the things that I noticed while skimming over in my initial review.

AFAICT, FreeBSD's RACCT framework potentiall exposes readbps/writebps in addition to readiops/writeiops. Using rctl_get_racct(2) could be used to return byte-based I/O rates when those counters are available at runtime, as that is usually more informative than raw operation counts alone. Given this has to be done per process, an extra column flag for readbps/writebps would be needed. If you decide to add this (a separate PR is entirely okay for this), please do this as a separate commit.

[samm-git]

@samm-git Full review of the PR will take some time. I noted the things that I noticed while skimming over in my initial review.

AFAICT, FreeBSD's RACCT framework potentiall exposes readbps/writebps in addition to readiops/writeiops. Using rctl_get_racct(2) could be used to return byte-based I/O rates when those counters are available at runtime, as that is usually more informative than raw operation counts alone. Given this has to be done per process, an extra column flag for readbps/writebps would be needed. If you decide to add this (a separate PR is entirely okay for this), please do this as a separate commit.

Thank you for your reply. Somehow I was wrongly assuming that RACCT would not work for ZFS.

However, the real problem is that rctl_get_racct is a root-only call, even for own processes. E.g., nobody gets EPERM for both process:<own> and user:<own>. So this could be a root-only feature, but I'm not sure it's a good idea. This is likely why top -mio also uses ops/s on BSD (I used it as inspiration).