On Wed, Feb 19, 2025 at 08:42:01AM +0100, Johannes Berg wrote:
On Tue, 2025-02-18 at 15:27 -0500, Sasha Levin wrote:
From: Johannes Berg <johannes.b...@intel.com>
[ Upstream commit 5b166b782d327f4b66190cc43afd3be36f2b3b7a ]
This code can be called deep in the IRQ handling, for
example, and then cannot normally use kmalloc(). Have
its own pre-allocated memory and use from there instead
so this doesn't occur. Only in the (very rare) case of
memcpy_toio() we'd still need to allocate memory.
I don't believe this patch, "um: convert irq_lock to raw spinlock" and
"um: virtio_uml: use raw spinlock", are relevant to anything older than
6.12. I don't see how applying them would _hurt_, but I didn't have them
before 6.12 and had no lockdep complaints about it; I believe some other
internal IRQ rework caused the issues to pop up.
Never mind that we (Intel WiFi stuff) are probably the only ones ever
running this virtio_uml/virt-pci with lockdep :)
I'll drop it, thanks!
--
Thanks,
Sasha
