It is now possible to walk the vma tree using the rcu read locks and is beneficial to do so to reduce lock contention. Doing so while a MAP_FIXED mapping is executing means that a reader may see a gap in the vma tree that should never logically exist - and does not when using the mmap lock in read mode. The temporal gap exists because mmap_region() calls munmap() prior to installing the new mapping. This patch set stops rcu readers from seeing the temporal gap by splitting up the munmap() function into two parts. The first part prepares the vma tree for modifications by doing the necessary splits and tracks the vmas in a side tree. The second part completes the munmapping of the vmas after the vma tree has been overwritten (either by a MAP_FIXED replacement vma or by a NULL in the munmap() case). Please note that rcu walkers will still be able to see a temporary state of split vmas that may be in the process of being removed, but the temporal gap will not be exposed. vma_start_write() are called on both parts of the split vma, so this state is detectable. I am sending this as an RFC as Andrii Nakryiko [1] and Suren Baghdasaryan are both working on features that require the vma tree to avoid exposing this temporal gap to rcu readers. [1] https://lore.kernel.org/all/gkhzuurhqhtozk6u53ufkesbhtjse5ba6kovqm7mnzrqe3szma@3tpbspq7hxjl/ Liam R. Howlett (5): mm/mmap: Correctly position vma_iterator in __split_vma() mm/mmap: Split do_vmi_align_munmap() into a gather and complete operation mm/mmap: Introduce vma_munmap_struct for use in munmap operations mm/mmap: Change munmap to use vma_munmap_struct() for accounting and surrounding vmas mm/mmap: Use split munmap calls for MAP_FIXED mm/internal.h | 22 +++ mm/mmap.c | 382 +++++++++++++++++++++++++++++++------------------- 2 files changed, 258 insertions(+), 146 deletions(-) -- 2.43.0
