SELFTEST FAILURE MODES FOR FAILSAFE MIGRATION:(9)

💅 Selftest failure modes for failsafe migration:(9) 💅

okay so real talk, this page is from the Kernel API Manual (section 9, bestie, not a drill) and it's basically the chaos menu for making migration fail on purpose so the failsafe paths get their main character moment ✨

🦋 the failure modes, explained 🦋

• fail_gpu_migration — when this is on, the gpu blit that gets scheduled is always a clear blit and not a copy blit 🙈 and then, bestie, we force the failure paths as if the blit fence came back with an error 💀 no cap, we're just out here pretending the fence flopped.
• fail_work_allocation — this one fails the kmalloc of the async worker, and then we sync the gpu blit 🫶 and if that sync'd blit then fails, or if fail_gpu_migration is set to true, then a memcpy operation is performed sync 🌸 like, the fallback ate and left no crumbs.

🌈 SEE ALSO 🌈

• Kernel file ./drivers/gpu/drm/i915/gem/i915_gem_ttm_move.c — the literal source file, go read it iykyk 📖
• i915_ttm_adjust_domains_after_move(9)
• i915_ttm_adjust_gem_after_move(9)
• i915_ttm_move_notify(9)
• struct i915_ttm_memcpy_arg(9)
• struct i915_ttm_memcpy_work(9)
• i915_ttm_move(9)
• i915_gem_obj_copy_ttm(9)

bestie reminder: these are deliberate failure modes for testing, not bugs — do not come for the kernel on this one 😭

that's the whole page, you're so welcome 💜

sign-off szn 🎀 — this one comes from ./drivers/gpu/drm/i915/gem, dated October 2026, and the page itself is Selftest fai...e migration:(9). slay, go test something 💅