Le 29/05/2024 à 10:02, Oscar Salvador a écrit :

[Vous ne recevez pas souvent de courriers de osalvador@suse.com. D?couvrez pourquoi ceci est important ? https://aka.ms/LearnAboutSenderIdentification ]

On Mon, May 27, 2024 at 03:30:06PM +0200, Christophe Leroy wrote:

quoted

In order to fit better with standard Linux page tables layout, add
support for 8M pages using contiguous PTE entries in a standard
page table. Page tables will then be populated with 1024 similar
entries and two PMD entries will point to that page table.

The PMD entries also get a flag to tell it is addressing an 8M page,
this is required for the HW tablewalk assistance.

Signed-off-by: Christophe Leroy <redacted>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
---

...

quoted

+#define __HAVE_ARCH_HUGE_PTEP_GET
+static inline pte_t huge_ptep_get(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+{
+     if (ptep_is_8m_pmdp(mm, addr, ptep))
+             ptep = pte_offset_kernel((pmd_t *)ptep, 0);

Yes, you are right that this should have had the addr aligned down.

I can speak for others, but for me it is more clear to think of it this way:

1) check if ptep points to the first PMD entry for address
2) if it does, we know that the PMD describes a 8MB hugepage
3) return the PMD

But it is not exactly the way I see it, maybe I'm wrong but from my 
point of view:
1) check if ptep points to the first PMD entry for address
2) if it does, we know that the PMD relates to a 8MB hugepage
3) Return the first PTE in the page table pointed by the said PMD entry.

That is why I thought that directly calling pmd_page_vaddr() gave a more clear
overview.

Now, feel free to ignore this if you think this is not clear or adds confusion,
I just wanted to give my insight reflecting on what I considered more
logical.

For me using pte_offset_kernel() make it more clear that we want a PTE 
in a page table, and not some raw information at a given address pointed 
to by a huge-PMD.

Am I wrong ?

--
Oscar Salvador
SUSE Labs