Re: [PATCH v2] x86: update comment about pgd_list

Re: [PATCH v2] x86: update comment about pgd_list

[Brendan Jackman]

On Mon Mar 23, 2026 at 2:25 PM UTC, Brendan Jackman wrote:
> This venerable comment got detached from its context when the code moved
> in commit 394158559d4c ("x86: move all the pgd_list handling to one
> place"). Put it back next to its context. It was originally on
> pgd_list_add() but it actually describes pgd_list so put it there.
>
> While moving it, update it to strip away stale and superfluous info.
> pageattr.c doesn't exist any more. pgd_list is now required
> for all x86 architectures. Also be slightly more precise about what PGDs
> are in this list.
>
> Signed-off-by: Brendan Jackman <jackmanb@google.com>
> ---
> Changes in v2:
> - Moved it next to pgd_list instead of back to old location
> - Strip away stale bits from the comment
> - Added note about PGDs that _aren't_ in this list
> - Link to v1: https://lore.kernel.org/r/20260311-pgd_list-comment-v1-1-4ea51b41adee@google.com

Hi Dave,

I think this is ready?

(I don't see it in tip/master, apologies if I'm looking in the wrong
place).

Re: [PATCH v2] x86: update comment about pgd_list

[H. Peter Anvin]

On 2026-05-11 04:35, Brendan Jackman wrote:
> On Mon Mar 23, 2026 at 2:25 PM UTC, Brendan Jackman wrote:
>> This venerable comment got detached from its context when the code moved
>> in commit 394158559d4c ("x86: move all the pgd_list handling to one
>> place"). Put it back next to its context. It was originally on
>> pgd_list_add() but it actually describes pgd_list so put it there.
>>
>> While moving it, update it to strip away stale and superfluous info.
>> pageattr.c doesn't exist any more. pgd_list is now required
>> for all x86 architectures. Also be slightly more precise about what PGDs
>> are in this list.
>>
>> Signed-off-by: Brendan Jackman <jackmanb@google.com>
>> ---
>> Changes in v2:
>> - Moved it next to pgd_list instead of back to old location
>> - Strip away stale bits from the comment
>> - Added note about PGDs that _aren't_ in this list
>> - Link to v1: https://lore.kernel.org/r/20260311-pgd_list-comment-v1-1-4ea51b41adee@google.com
> 
> Hi Dave,
> 
> I think this is ready?
> 
> (I don't see it in tip/master, apologies if I'm looking in the wrong
> place).

The PAE code comment should still remain.

I don't know if the 32-bit PAE code is still suffering from Xen-induced drain
bramage (using a whole page to store 32 bytes), but it is nevertheless true in
a fashion -- as there are a maximum of 3 kernel PDE pages, they are considered
the de facto top level of the kernel page tables.

	-hpa

Re: [PATCH v2] x86: update comment about pgd_list

[Dave Hansen]

On 5/11/26 12:49, H. Peter Anvin wrote:
> The PAE code comment should still remain.
> 
> I don't know if the 32-bit PAE code is still suffering from Xen-induced drain
> bramage (using a whole page to store 32 bytes),

Actually, I think I recently spread the brain damage even further. The
kmalloc()'d PGDs are completely gone now. PGDs all directly out of the
page allocator. It actually simplified the 32-bit code quite a bit.

> but it is nevertheless true in a fashion -- as there are a maximum
> of 3 kernel PDE pages, they are considered the de facto top level of
> the kernel page tables.
Here's what I ended up with:

> +/*
> + * List of all process' PGD pages (excluding init_mm and some other special
> + * system-wide PGDs). Mainly used to keep the kernel portion of the address
> + * space in sync across mms.
> + */

I think what you're saying is that those (max) 3 kernel PDE pages are
static and all PGDs always have the same value in there. So, there's no
synchronizing necessary on PAE. That was definitely true for a long time.

But we actually also moved away from that:

> https://lore.kernel.org/all/20250414173232.32444FF6@davehans-spike.ostc.intel.com/

PTI forced some PMD unsharing. So I just moved 32-bit over to not share
kernel PMDs at all. It made the 32-bit code quite a bit simpler,
although it wastes memory when PTI is disabled.

[tip: x86/mm] x86: Update comment about pgd_list

[tip-bot2 for Brendan Jackman]

The following commit has been merged into the x86/mm branch of tip:

Commit-ID:     952ac097ce98901c608cdced012f89a90367401e
Gitweb:        https://git.kernel.org/tip/952ac097ce98901c608cdced012f89a90367401e
Author:        Brendan Jackman <jackmanb@google.com>
AuthorDate:    Mon, 23 Mar 2026 14:25:07 
Committer:     Dave Hansen <dave.hansen@linux.intel.com>
CommitterDate: Mon, 11 May 2026 11:26:33 -07:00

x86: Update comment about pgd_list

This venerable comment got detached from its context when the code moved
in commit 394158559d4c ("x86: move all the pgd_list handling to one
place"). Put it back next to its context. It was originally on
pgd_list_add() but it actually describes pgd_list so put it there.

While moving it, update it to strip away stale and superfluous info.
pageattr.c doesn't exist any more. pgd_list is now required
for all x86 architectures. Also be slightly more precise about what PGDs
are in this list.

[ dhansen: tweak and trim the updated comment a bit ]

Signed-off-by: Brendan Jackman <jackmanb@google.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Link: https://patch.msgid.link/20260323-pgd_list-comment-v2-1-77ccf2dc77e8@google.com
---
 arch/x86/mm/fault.c   |  5 +++++
 arch/x86/mm/pgtable.c | 11 -----------
 2 files changed, 5 insertions(+), 11 deletions(-)

diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index 63de8e8..250942e 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -169,6 +169,11 @@ is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
 }
 
 DEFINE_SPINLOCK(pgd_lock);
+/*
+ * List of all process' PGD pages (excluding init_mm and some other special
+ * system-wide PGDs). Mainly used to keep the kernel portion of the address
+ * space in sync across mms.
+ */
 LIST_HEAD(pgd_list);
 
 #ifdef CONFIG_X86_32
diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c
index da7f0a0..f32facd 100644
--- a/arch/x86/mm/pgtable.c
+++ b/arch/x86/mm/pgtable.c
@@ -99,17 +99,6 @@ static void pgd_dtor(pgd_t *pgd)
 	spin_unlock(&pgd_lock);
 }
 
-/*
- * List of all pgd's needed for non-PAE so it can invalidate entries
- * in both cached and uncached pgd's; not needed for PAE since the
- * kernel pmd is shared. If PAE were not to share the pmd a similar
- * tactic would be needed. This is essentially codepath-based locking
- * against pageattr.c; it is the unique case in which a valid change
- * of kernel pagetables can't be lazily synchronized by vmalloc faults.
- * vmalloc faults work because attached pagetables are never freed.
- * -- nyc
- */
-
 #ifdef CONFIG_X86_PAE
 /*
  * In PAE mode, we need to do a cr3 reload (=tlb flush) when