CVE-2024-50182: secretmem: disable memfd_secret() if arch cannot set direct map

In the Linux kernel, the following vulnerability has been resolved:

secretmem: disable memfdsecret() if arch cannot set direct map

Return -ENOSYS from memfdsecret() syscall if !cansetdirectmap(). This is the case for example on some arm64 configurations, where marking 4k PTEs in the direct map not present can only be done if the direct map is set up at 4k granularity in the first place (as ARM's break-before-make semantics do not easily allow breaking apart large/gigantic pages).

More precisely, on arm64 systems with !cansetdirectmap(), setdirectmapinvalidnoflush() is a no-op, however it returns success (0) instead of an error. This means that memfdsecret will seemingly "work" (e.g. syscall succeeds, you can mmap the fd and fault in pages), but it does not actually achieve its goal of removing its memory from the direct map.

Note that with this patch, memfdsecret() will start erroring on systems where cansetdirectmap() returns false (arm64 with CONFIGRODATAFULLDEFAULTENABLED=n, CONFIGDEBUGPAGEALLOC=n and CONFIGKFENCE=n), but that still seems better than the current silent failure. Since CONFIGRODATAFULLDEFAULTENABLED defaults to 'y', most arm64 systems actually have a working memfdsecret() and aren't be affected.

From going through the iterations of the original memfdsecret patch series, it seems that disabling the syscall in these scenarios was the intended behavior [1] (preferred over having setdirectmapinvalidnoflush return an error as that would result in SIGBUSes at page-fault time), however the check for it got dropped between v16 [2] and v17 [3], when secretmem moved away from CMA allocations.

[1]: https://lore.kernel.org/lkml/20201124164930.GK8537@kernel.org/ [2]: https://lore.kernel.org/lkml/20210121122723.3446-11-rppt@kernel.org/#t [3]: https://lore.kernel.org/lkml/20201125092208.12544-10-rppt@kernel.org/