dma_fence_get_rcu_safe(9)

📦 version 7.3.1, in nixpkgs unstable

✨ dma_fence_get_rcu_safe 💅

okay so bestie, this is a page from the Kernel API Manual and it's about dma_fence_get_rcu_safe — literally the function that grabs you a reference to an RCU tracked fence. no cap. 🦋

🌸 NAME 🌸

dma_fence_get_rcu_safe — acquire a reference to an RCU tracked fence 🔥

🎀 SYNOPSIS 🎀

struct dma_fence * dma_fence_get_rcu_safe (struct dma_fence __rcu
**fencep );

🫶 ARGUMENTS 🫶

  • fencep — pointer to fence to increase refcount of. that's it, that's the whole argument, she's simple and she knows it 💅

💖 DESCRIPTION 💖

So the function returns NULL if no refcount could be obtained, or the fence. that's the vibe check: you either get the fence or you get nothing. ✨

Here's the iconic part: this function handles acquiring a reference to a fence that may be reallocated within the RCU grace period (such as with SLAB_TYPESAFE_BY_RCU), so long as the caller is using RCU on the pointer to the fence. yes bestie, reallocated, mid-grace-period, it's giving chaos but the function is unbothered. 👑

An alternative mechanism is to employ a seqlock to protect a bunch of fences, such as used by struct dma_resv. When using a seqlock, the seqlock must be taken before and checked after a reference to the fence is acquired (as shown here). before AND after, okay? both. do not skip the second one, i'm serious 🙈

The caller is required to hold the RCU read lock. not optional, not a suggestion, this is the whole premise. hold it. 🔒

🌈 SEE ALSO 🌈

so much lore to catch up on, drink some water and read these:

and that's the page!! 🍓 she lives in ./include/linux, dated October 2026, and she's page dma_fence_get_rcu_safe(9) of the Kernel API Manual. go forth and hold that RCU read lock, ily, mwah 💜🧚

📦 where this page comes from