On Tue, 20 Feb 2024 04:05:21 -0500 rulinhuang <rulin.huang@xxxxxxxxx> wrote: > When allocating a new memory area where the mapping address range is > known, it is observed that the vmap_area lock is acquired twice. > The first acquisition occurs in the alloc_vmap_area() function when > inserting the vm area into the vm mapping red-black tree. The second > acquisition occurs in the setup_vmalloc_vm() function when updating the > properties of the vm, such as flags and address, etc. > Combine these two operations together in alloc_vmap_area(), which > improves scalability when the vmap_area lock is contended. By doing so, > the need to acquire the lock twice can also be eliminated. > With the above change, tested on intel icelake platform(160 vcpu, kernel > v6.7), a 6% performance improvement and a 7% reduction in overall > spinlock hotspot are gained on > stress-ng/pthread(https://github.com/ColinIanKing/stress-ng), which is > the stress test of thread creations. vmalloc.c has changed a lot lately. Please can you work against linux-next or against the mm-unstable branch of //git.kernel.org/pub/scm/linux/kernel/git/akpm/mm Thanks.
