pcim_p2pdma_provider(9)

📦 version 7.3.1, in nixpkgs unstable

💅 pcim_p2pdma_provider 💅

okay so this is a page from the Kernel API Manual (section 9, kernel-space girlies rise up) and it's literally all about grabbing a peer-to-peer DMA provider for a PCI device. stay with me bestie, it's giving niche-but-important 🫶

✨ NAME ✨

pcim_p2pdma_provider — get peer-to-peer DMA provider 🌈

✨ SYNOPSIS ✨

struct p2pdma_provider * pcim_p2pdma_provider (struct pci_dev *pdev ,
int bar );

✨ ARGUMENTS ✨

  • pdev — the PCI device you wanna enable P2PDMA for 🎀
  • bar — the BAR index you're getting the provider from, iykyk

✨ DESCRIPTION ✨

so this function gets you the peer-to-peer DMA provider for a PCI device. cute! 💖

but here's the tea: the lifetime of the provider (and obviously the MMIO too, like, they're bonded fr fr) is bound to the lifetime of the driver. so if you're a driver calling this function, you must make sure that every single reference to the provider — and any DMA mappings you created for any MMIO — are all cleaned up before your driver remove completes. not optional. not a vibe check you can skip. 👑

and because P2P is almost always shared with a second driver (the way it's always shared 😭), that means you need some kind of system to notify, invalidate and revoke the MMIO's DMA in place to actually use this function. for example, a revoke can be built using DMABUF. slay, that's your escape hatch. 🦋

bestie please don't leak the provider or the mappings — that's the whole warning, and it's real. 🔥

✨ SEE ALSO ✨

okay that's the whole page, no cap 💅 this icon comes from ./drivers/pci, dated October 2026, and it's filed under pcim_p2pdma_provider(9) — periodt. 🌸✨

📦 where this page comes from