I am probably missing something but I still don't see the difference given that buffer pool will otherwise be implemented in terms of malloc, and malloc is still implemented in terms of mmap so we're back to square one.
However, I could imagine that it is possible for contention in kernel-space mmap locks could be artificially relieved because of the user-space locks in malloc.
Currently I don't see how TLB shootdowns would be specific to mmap only but not to malloc but I may find some time to read the paper.
What we’re talking about is mmap-ing files vs doing some anonymous mapping and manually reading files. Anonymous mappings don’t incur all the TLB shootdowns that mmap’d files do.
However, I could imagine that it is possible for contention in kernel-space mmap locks could be artificially relieved because of the user-space locks in malloc.
Currently I don't see how TLB shootdowns would be specific to mmap only but not to malloc but I may find some time to read the paper.