Edgar Toernig wrote: [...]
I wasn't aware of that (and I would definitely prefer a different behaviour). But anyway, correct me if I'm wrong, revoke (V2) not simply removes the pages from the mmaped area as truncating does (the vma stays); revoke seems to completely remove the vma which is clearly a security bug. Future mappings may silently get mapped into the area of the revoked file without the app noticing it. It may then hand out data of the new file still thinking it's sending the old one.
One could remap to /dev/null - the file would then be free to be umounted, but the app could get confused. Or map inaccessible pages, so the app segfault on the next access. Helge Hafting - To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
