Re: [PATCH v3 3/4] virt: tdx-guest: Use a variable to store the Quote buffer size

[Peter Fang]

On Thu, Sep 03, 2026 at 01:59:25PM +0800, Xiaoyao Li wrote:
> On 9/3/2026 7:15 AM, Peter Fang wrote:
> > You are also dropping PAGE_ALIGN() in alloc and free calls. You can
> > mention it in commit log.
> 
> I think it's better to keep the PAGE_ALIGN() when assign quote_data_len,
> although GET_QUOTE_BUF_SIZE is aligned already. This makes it super clear
> we want to page size align the buffer.
> 
> And in next patch, we can make it
> 
>   quote_data_len = PAGE_ALIGN(get_quote_buf_size());

Hmm... Looking at this again actually makes me think it could be
reworked a bit. There's a discrepancy in alloc_quote_buf() if for
whatever reason @len is not page aligned. alloc_pages_exact() rounds it
up internally but set_memory_decrypted() rounds it *down*, so we'd end
up with a useless page at the end...

What if alloc_quote_buf() took a @npages instead? Then the expectations
would be pretty clear:


  -static void *alloc_quote_buf(void)
  +static unsigned int get_quote_buf_npages(void)
  +{
  +     u32 quote_size = tdx_get_max_quote_size();
  +     size_t len;
  +
  +     if (quote_size)
  +             len = TDX_QUOTE_BUF_LEN(quote_size);
  +     else
  +             len = GET_QUOTE_DEFAULT_BUF_SIZE;
  +
  +     return PFN_UP(len);
  +}
  +
  +static void *alloc_quote_buf(unsigned int npages)
   {
  -     size_t len = PAGE_ALIGN(GET_QUOTE_BUF_SIZE);
  -     unsigned int count = len >> PAGE_SHIFT;
        void *addr;

  -     addr = alloc_pages_exact(len, GFP_KERNEL | __GFP_ZERO);
  +     addr = alloc_pages_exact(PAGE_SIZE * npages, GFP_KERNEL | __GFP_ZERO);
        if (!addr)
                return NULL;

  -     if (set_memory_decrypted((unsigned long)addr, count))
  +     if (set_memory_decrypted((unsigned long)addr, npages))
                return NULL;

        return addr;
   }

   static int __init tdx_guest_init(void)
   {
        ...
  -     quote_data = alloc_quote_buf();
  +     quote_data_npages = get_quote_buf_npages();
  +     quote_data = alloc_quote_buf(quote_data_npages);
        if (!quote_data) {
                ...
        }
   }


Does this look better to you?