STRUCT XE_DEVCOREDUMP_SNAPSHOT(9)

🎀 struct xe_devcoredump_snapshot(9) 🎀

okay so this is a Kernel API Manual page, and it's literally the crash-snapshot struct for Xe — main character energy for when the GPU has a bad day 😭. it's giving "we captured the crash tea and saved it for later," and we're gonna walk through every single member, no cap. 🫶

✨ NAME ✨

struct xe_devcoredump_snapshot — Crash snapshot. that's the whole vibe: the struct that holds the crash snapshot, bestie 💅

🌈 SYNOPSIS 🌈

here's the full struct, character-for-character, because we do not play about code:

struct xe_devcoredump_snapshot {
    ktime_t snapshot_time;
    ktime_t boot_time;
    char process_name[TASK_COMM_LEN];
    pid_t pid;
    char *reason;
    struct xe_gt *gt;
    struct work_struct work;
    struct {
      struct xe_guc_ct_snapshot *ct;
      struct xe_guc_log_snapshot *log;
    } guc;
    struct xe_guc_submit_exec_queue_snapshot *ge;
    struct xe_hw_engine_snapshot *hwe[XE_NUM_HW_ENGINES];
    struct xe_sched_job_snapshot *job;
    struct __guc_capture_parsed_output *matched_node;
    struct xe_vm_snapshot *vm;
    struct {
      ssize_t size;
      ssize_t chunk_position;
      char *buffer;
    } read;
};

🧚 Members 🧚

• snapshot_time — Time of this capture. ⏱️
• boot_time — Relative boot time so the uptime can be calculated. 🥾
• process_name — Name of process that triggered this gpu hang. 💻
• pid — Process id of process that triggered this gpu hang. 🔢
• reason — The reason the coredump was triggered. 🚨
• gt — Affected GT, used by forcewake for delayed capture. 🎯
• work — Workqueue for deferred capture outside of signaling context. 🛠️
• guc — GuC snapshots. 🎮
  • guc.ct — GuC CT snapshot. 📸
  • guc.log — GuC log snapshot. 📝
• ge — GuC Submission Engine snapshot. 🚀
• hwe — HW Engine snapshot array. 🧩
• job — Snapshot of job state. 💼
• matched_node — The matched capture node for timedout job — this single-node tracker works because devcoredump will always only produce one hw-engine capture per devcoredump event. 👀
• vm — Snapshot of VM state. 🧠
• read — devcoredump in human readable format. 📖
  • read.size — size of devcoredump in human readable format. 📏
  • read.chunk_position — position of devcoredump chunk. 📍
  • read.buffer — buffer of devcoredump in human readable format. 🧺

💖 Description 💖

okay so this struct contains all the useful information that gets quickly captured at the time of the crash. so, any subsequent reads of the coredump points to a data that shows the state of the GPU of when the issue has happened. literally a time capsule for the GPU's bad moment, periodt. 💀✨

🔥 SEE ALSO 🔥

• Kernel file ./drivers/gpu/drm/xe/xe_devcoredump_types.h
• struct xe_devcoredump(9) — don't sleep on it, fr fr 👑

and that's the tea 🍵 — source/version: ./drivers/gpu/drm/xe · date: October 2026. mwah 💋