> > > +/* > > > + * Walks over all memblock memory regions that are intended to be > > > + * converted to TDX memory. Essentially, it is all memblock memory > > > + * regions excluding the low memory below 1MB. > > > + * > > > + * This is because on some TDX platforms the low memory below 1MB is > > > + * not included in CMRs. Excluding the low 1MB can still guarantee > > > + * that the pages managed by the page allocator are always TDX memory, > > > + * as the low 1MB is reserved during kernel boot and won't end up to > > > + * the ZONE_DMA (see reserve_real_mode()). > > > + */ > > > +#define memblock_for_each_tdx_mem_pfn_range(i, p_start, p_end, p_nid) \ > > > + for_each_mem_pfn_range(i, MAX_NUMNODES, p_start, p_end, p_nid) \ > > > + if (!pfn_range_skip_lowmem(p_start, p_end)) > > > > Let's summarize where we are at this point: > > > > 1. All RAM is described in memblocks > > 2. Some memblocks are reserved and some are free > > 3. The lower 1MB is marked reserved > > 4. for_each_mem_pfn_range() walks all reserved and free memblocks, so we > > have to exclude the lower 1MB as a special case. > > > > That seems superficially rather ridiculous. Shouldn't we just pick a > > memblock iterator that skips the 1MB? Surely there is such a thing. > > Perhaps you are suggesting we should always loop the _free_ ranges so we don't > need to care about the first 1MB which is reserved? > > The problem is some reserved memory regions are actually later freed to the page > allocator, for example, initrd. So to cover all those 'late-freed-reserved- > regions', I used for_each_mem_pfn_range(), instead of for_each_free_mem_range(). > > Btw, I do have a checkpatch warning around this code: > > ERROR: Macros with complex values should be enclosed in parentheses > #109: FILE: arch/x86/virt/vmx/tdx/tdx.c:377: > +#define memblock_for_each_tdx_mem_pfn_range(i, p_start, p_end, p_nid) \ > + for_each_mem_pfn_range(i, MAX_NUMNODES, p_start, p_end, p_nid) \ > + if (!pfn_range_skip_lowmem(p_start, p_end)) > > But it looks like a false positive to me. Hi Dave, Sorry to ping. Just double check, any comments around here, .. > > > Or, should we be doing something different with the 1MB in the memblock > > structure? > > memblock APIs are used by other kernel components. I don't think we should > modify memblock code behaviour for TDX. Do you have any specific suggestion? > > One possible option I can think is explicitly "register" memory regions as TDX > memory when they are firstly freed to the page allocator. [...] > > This will require new data structures to represent TDX memblock and the code to > create, insert and merge contiguous TDX memblocks, etc. The advantage is we can > just iterate those TDX memblocks when constructing TDMRs. > > And here?
