On 12/27/2017 09:12 PM, Dmitry Vyukov wrote: >> >> Not really a problem caused by the patch above, it's merily exposing a code >> path which relies blindly on "enough memory available" assumptions. >> >> Throwing more memory at the VM makes the problem go away... > > Hi Thomas, > > We just need a check inside of early_alloc() to properly diagnose such > situation, right? > KASAN allocates 1/8 of physical memory, so the code has _some_ logic > behind it. But I guess it runs into a corner case where 1/8 is somehow > larger than 1 :) >
We may simply don't have 1/8 available. And we actually need slightly more than 1/8. Kernel image is present in 2 locations (direct mapping/kernel mapping) so at least ram/8 + kernel_size/8. The same for cpu_entry_area. Plus some memory to build page tables. And note that kernel_size could be hundreds of megabytes.
