ata_eh_decide_disposition(9)

💅 ata_eh_decide_disposition ✨

okay so bestie, we're deep in the Kernel API Manual, section 9, and this one lives in ./drivers/ata because of course it does 🫶

---

🌸 NAME 🌸

ata_eh_decide_disposition — literally decides the disposition of a qc based on sense data 💅

---

✨ SYNOPSIS ✨

enum scsi_disposition ata_eh_decide_disposition (struct ata_queued_cmd
*qc );

---

🎀 ARGUMENTS 🎀

• qc — the qc we're examining 👀 that's it, that's the whole argument list

---

🦋 DESCRIPTION 🦋

okay so picture this, bestie 🧚 for a regular SCSI command, the SCSI completion callback (scsi_done) calls scsi_complete, which calls scsi_decide_disposition, which calls scsi_check_sense. and then scsi_complete finally calls scsi_finish_command. it's giving a whole relay race 🏃‍♀️💨

and that's totally fine for SCSI, because any eventual sense data is usually returned in the completion itself — without ever invoking SCSI EH. no cap ✋

BUT 💀 for a QC, we always need to fetch the sense data explicitly using SCSI EH. always always. no shortcuts.

meanwhile, a command that's completed via SCSI EH gets completed a different way — using scsi_eh_flush_done_q, which calls scsi_finish_command directly, without ever calling scsi_check_sense. so that check literally never happens on that path 😭

so for a command that went through SCSI EH, it is the responsibility of the SCSI EH strategy handler to call scsi_decide_disposition — see e.g. how scsi_eh_get_sense calls scsi_decide_disposition for SCSI LLDDs that do not get the sense data as part of the completion. main character energy, honestly 👑

so yeah — for QC commands that went via SCSI EH, we need to call scsi_check_sense ourselves, similar to how scsi_eh_get_sense calls scsi_decide_disposition, which calls scsi_check_sense, in order to set the correct SCSI ML byte (if any). that's the whole vibe check right there ✅

LOCKING: EH context. periodt 🔒

---

🌈 RETURN 🌈

one of these four, bestie:

• SUCCESS
• FAILED
• NEEDS_RETRY
• ADD_TO_MLQUEUE

---

🍓 SEE ALSO 🍓

Kernel file ./drivers/ata/libata-eh.c, plus the whole squad 🔥:

__ata_ehi_push_desc(9), ata_ehi_push_desc(9), ata_ehi_clear_desc(9), ata_port_desc(9), ata_port_pbar_desc(9), ata_internal_cmd_timeout(9), ata_internal_cmd_timed_out(9), ata_eh_acquire(9), ata_eh_release(9), ata_scsi_error(9), ata_scsi_cmd_error_handler(9), ata_scsi_port_error_handler(9), ata_port_wait_eh(9), ata_eh_set_pending(9), ata_qc_schedule_eh(9), ata_std_sched_eh(9), ata_std_end_eh(9), ata_port_schedule_eh(9), ata_link_abort(9), ata_port_abort(9), __ata_port_freeze(9), ata_port_freeze(9), ata_eh_freeze_port(9), ata_eh_thaw_port(9), ata_eh_qc_complete(9), ata_eh_qc_retry(9), ata_dev_disable(9), ata_eh_detach_dev(9), ata_eh_about_to_do(9), ata_eh_done(9), ata_err_string(9), atapi_eh_tur(9), ata_eh_request_sense(9), atapi_eh_request_sense(9), ata_eh_analyze_serror(9), ata_eh_analyze_tf(9), ata_eh_speed_down_verdict(9), ata_eh_speed_down(9), ata_eh_worth_retry(9), ata_eh_quiet(9), ata_eh_link_set_lpm(9), ata_eh_link_autopsy(9), ata_eh_autopsy(9), ata_get_cmd_name(9), ata_eh_link_report(9), ata_eh_report(9), ata_eh_set_mode(9), atapi_eh_clear_ua(9), ata_eh_maybe_retry_flush(9), ata_eh_recover(9), ata_eh_finish(9), ata_std_error_handler(9), ata_eh_handle_port_suspend(9), ata_eh_handle_port_resume(9)

---

and that's the whole page, bestie 🎀 tysm for reading, this came from ./drivers/ata and is dated October 2026 — catch you in the next manpage, mwah 💖💅✨