STRUCT XE_PAGEFAULT_QUEUE(9)

🎀 NAME 🎀

struct xe_pagefault_queue — the Xe pagefault queue, but make it consumer 💅 this is the section 9 kernel API manual page bestie, we're in the big leagues now

✨ SYNOPSIS ✨

okay so picture this struct, absolutely iconic, here's the whole thing verbatim:

struct xe_pagefault_queue {
    struct xe_device *xe;
    void *data;
    u32 size;
    u32 head;
    u32 tail;
    spinlock_t lock;
    struct work_struct worker;
 };

🫶 Members 🫶

• xe — the back-pointer to the Xe device, she knows where she came from 💖
• data — the actual data in the queue, containing struct xe_pagefault, and bestie it is protected by lock 🔒
• size — size of the queue, in bytes, no cap
• head — head pointer, also in bytes, and it gets moved by the producer — protected by lock too 🫶
• tail — tail pointer, in bytes, moved by the consumer this time — yep, protected by lock 🌸
• lock — literally protects the page fault queue, that's her whole job 👑
• worker — the worker that processes page faults, main character energy fr fr ⚡

💜 Description 💜

So what's the vibe? This queue is used to capture all device page faults for deferred processing — like, every single one, she's collecting them 🦋

And the sizing advice, listen up: size this queue to absorb the device's worst-case number of outstanding faults. Not the average case. Not the "probably fine" case. The worst-case one 😭 do not come crying to me when your queue overflows because you sized it optimistically.

ngl the sizing thing is the actual important part, don't skip it 💀

🌈 SEE ALSO 🌈

• Kernel file ./drivers/gpu/drm/xe/xe_pagefault_types.h
• struct xe_pagefault(9)

k bye bestie, this page came to you from ./drivers/gpu/drm/xe, dated October 2026 — go read xe_pagefault(9) next, iykyk 🧚‍♀️✨