On 02/28/2018 09:18 PM, Liam Mark wrote:
The issue:
Currently in ION if you allocate uncached memory it is possible that there are still dirty lines in the cache. And often these dirty lines in the cache are the zeros which were meant to clear out any sensitive kernel data.
What this means is that if you allocate uncached memory from ION, and then subsequently write to that buffer (using the uncached mapping you are provided by ION) then the data you have written could be corrupted at some point in the future if a dirty line is evicted from the cache.
Also this means there is a potential security issue. If an un-privileged userspace user allocated uncached memory (for example from the system heap) and then if they were to read from that buffer (through the un-cached mapping they are provided by ION), and if some of the zeros which were written to that memory are still in the cache then this un-privileged userspace user could read potentially sensitive kernel data.
For the use case you are describing we don't actually need the memory to be non-cached until it comes time to do the dma mapping. Here's a proposal to shoot holes in:
- Before any dma_buf attach happens, all mmap mappings are cached - At the time attach happens, we shoot down any existing userspace mappings, do the dma_map with appropriate flags to clean the pages and then allow remapping to userspace as uncached. Really this looks like a variation on the old Ion faulting code which I removed except it's for uncached buffers instead of cached buffers.
Potential problems: - I'm not 100% about the behavior here if the attaching device is already dma_coherent. I also consider uncached mappings enough of a device specific optimization that you shouldn't do them unless you know it's needed. - The locking/sequencing with userspace could be tricky since userspace may not like us ripping mappings out from underneath if it's trying to access.
Thoughts?
Thanks, Laura