关于PAE页地址扩展的技术问询：条目大小与地址映射原理

Great question—PAE can feel pretty opaque at first, especially when you're just staring at that classic diagram without context. Let's break this down clearly:

Dir.Pointer entries are 64 bits (8 bytes) long—they are not 32 bits. This is a critical distinction from standard 32-bit paging. Each entry stores two core pieces of data:

• The physical base address of a Page Directory (shifted right by 12 bits, since all page structures are 4KB-aligned)
• Control flags (like present, read/write, user/supervisor, and others that manage page permissions and behavior)

PAE keeps the 32-bit virtual address space intact (so each process still sees a 4GB virtual range) but extends the physical address space to 36 bits (supporting up to 64GB of RAM). Here's the step-by-step mapping process:

1. Split the 32-bit virtual address into 4 components:

   • Bits 31-30: Index into the Page Directory Pointer Table (PDPT) — selects which Dir.Pointer entry to use
   • Bits 29-21: Index into the selected Page Directory — picks a Page Directory Entry (PDE)
   • Bits 20-12: Index into the selected Page Table — picks a Page Table Entry (PTE)
   • Bits 11-0: Offset within the final physical page

2. Traverse the table hierarchy:

   • Start with the PDPT's physical base address, stored in the CR3 register (in PAE mode, CR3 points to the start of the PDPT, which is 32 bytes aligned since it holds 4 x 8-byte Dir.Pointer entries)
   • Use the top 2 virtual address bits to fetch the 64-bit Dir.Pointer entry. This entry provides the physical base address of the Page Directory (the upper 24 bits of the entry correspond to physical address bits 35-12, thanks to 4KB alignment)
   • Next, use bits 29-21 to index into the Page Directory (each PDE is also 64 bits). This entry gives the physical base address of the Page Table
   • Then, use bits 20-12 to index into the Page Table (each PTE is 64 bits). This entry holds the physical base address of the target page
   • Finally, add the 12-bit page offset to the physical page base address to get the full 36-bit physical address

3. Key reason this works:

   • The extra PDPT layer (with 64-bit Dir.Pointer entries) is what unlocks the extended physical address space—standard 32-bit paging only has two levels, which can't store enough bits for a 36-bit physical address
   • All table entries (Dir.Pointer, PDE, PTE) are 64 bits, which gives enough space to store the 24 bits needed for the page base (since 36-12=24) plus all required control flags

内容的提问来源于stack exchange，提问作者Steven