On Thu, 14 Mar 2024, Russell King (Oracle) wrote:
It is really surprising that this didn't blow up for anyone else so
far... This means that the $subject patch is fine.
Wow. I guess we've been lucky with that allocation hitting memory
containing zeros. Well done at tracking it down!
It would have blown up with slub_debug because that includes poisoning the
contents of all allocations via the slab allocator. Why did that not
occur? We should have seen a backtrace with data in registers etc showing
poisoning values for an unitialized object.
Note that this was indeed triggered by OFFSTACK because
(z)alloc_cpumask_var() only generates a kmalloc allocation if that option
is set.
The config option was never set before my patch was applied on ARM64.
