xe_ttm_stol...s_needs_ggtt(9)

💅 xe_ttm_stolen_cpu_access_needs_ggtt(9) — from the Kernel API Manual, bestie 🎀

okay so real talk, this one's from the Kernel API Manual and it's a whole function called xe_ttm_stolen_cpu_access_needs_ggtt ✨ lowkey love that for us.

---

🌸 NAME 🌸

xe_ttm_stolen_cpu_access_needs_ggtt — like, if we literally can't CPU access stolen directly, can we fallback to mapping through the GGTT? that's the whole question 💭

---

🦋 SYNOPSIS 🦋

bool xe_ttm_stolen_cpu_access_needs_ggtt (struct xe_device *xe );

she returns a bool, she takes one arg, she's iconic, periodt 👑

---

🍓 ARGUMENTS 🍓

• xe — the xe device. that's it. that's the arg. the main character of this entire function fr fr 🫶

---

🌈 DESCRIPTION 🌈

okay so picture this: some older integrated platforms are NOT it when it comes to reliable CPU access for stolen 😭 like they just can't do it, no cap.

but here's the slay — on that kind of hardware we can always use the mappable part of the GGTT for CPU access 💅 so it's giving "we found a way."

so this function just checks if that's the case for this device 👀 vibe check on the hardware, basically. xe walks in, gets checked, and we find out whether the GGTT fallback situation applies ✨

lowkey important: this is about older integrated platforms not supporting reliable CPU access for stolen — the GGTT fallback is the whole point of the check, so don't skip it 🎀

---

🧚 SEE ALSO 🧚

• Kernel file ./drivers/gpu/drm/xe/xe_ttm_stolen_mgr.c — go read the source bestie, it's right there 🔥

---

👑 sign-off 👑

so that's the tea ☕ this page lives in ./drivers/gpu/drm/xe, dated October 2026, and it's giving extremely niche kernel docs energy and we're obsessed 💖 bye bestie, go check your GGTT 🦋✨