radeon_mn_invalidate(9)

πŸ“¦ version 7.3.1, in nixpkgs unstable

πŸ’… radeon_mn_invalidate β€” like, the manual page, bestie

okay so this is radeon_mn_invalidate(9), straight out of the Kernel API Manual, and it's literally the callback that goes "hey bestie, the mm changed πŸ‘€". Let's get into it ✨


πŸŽ€ NAME πŸŽ€

radeon_mn_invalidate β€” the callback that notifies you about an mm change. no cap, that's the whole vibe.


🌸 SYNOPSIS 🌸

bool radeon_mn_invalidate (struct mmu_interval_notifier *mn , const
       struct mmu_notifier_range *range , unsigned long cur_seq );

copy it exactly like that, don't @ me πŸ’…


πŸ¦‹ ARGUMENTS πŸ¦‹

  • mn β€” our notifier πŸ’– like, the one that belongs to us.
  • range β€” the VMA that's under invalidation rn 🫢
  • cur_seq β€” the value you pass to mmu_interval_set_seq, periodt.

✨ DESCRIPTION ✨

So here's the tea: we block until all the BOs between start and end are idle 😴 β€” yes, all of them, we wait β€” and then we unmap them by moving them back into the system domain again. iconic behaviour honestly πŸ‘‘


πŸ’œ SEE ALSO πŸ’œ


ngl, if you're poking around in here you should probably also say hi to its register/unregister siblings up there πŸ‘† iykyk

and that's the whole page, bestie πŸ“πŸ’– sign-off time: this one came from ./drivers/gpu/drm/radeon and it's dated October 2026 β€” slay, go forth and invalidate responsibly 🌈⭐

πŸ“¦ where this page comes from