On 06/06/2024 16:08, Christoph Hellwig wrote: > On Thu, Jun 06, 2024 at 03:57:25PM +0300, Ofir Gal wrote: >> The slab pages aren't allocated by the md-bitmap, they are pages that >> happens to be after the allocated pages. I'm applying a patch to the md >> subsystem asap. > Similar cases could happen by other means as well. E.g. if you write > to the last blocks in an XFS allocation group while also writing something > to superblock copy at the beginnig of the next one and the two writes get > merged. Probably not easy to reproduce but entirely possible. Just as > as lot of other scenarious could happen due to merges. I agree. Do we want to proceed with my patch or would we prefer a more efficient solution?
