[PATCH 1/1] riscv/mm: fix soft-dirty migration PMDs being treated as present

[Lance Yang]

On Mon, Oct 05, 2026 at 12:37:20PM +0200, David Hildenbrand (Arm) wrote:
>On 10/4/26 05:03, Lance Yang wrote:
>> RISC-V uses _PAGE_EXEC for swap soft-dirty tracking when
>> CONFIG_MEM_SOFT_DIRTY is enabled and Svrsw60t59b is available. That's
>> a problem for PMD migration entries, since pmd_present() also checks
>> _PAGE_LEAF (R/W/X) to recognize THPs with _PAGE_PRESENT temporarily
>> cleared during splitting.
>
>I'm curious: why do we have to set leaf indications for non-present things? The
>HW sure will ignore it, right?

Yeah, that surprised me too :) Still wrapping my head around the details
...

>Is this a sw problem? Who needs that?

IIUC, it's for software during a PMD split.

__split_huge_pmd_locked() invalidates the huge PMD and flushes the TLB
before installing the PTE table. Software still needs pmd_present() and
pmd_trans_huge() to recognize the THP in between.

RISC-V clears V but keeps the R/W/X bits for that, so the entry is invalid
to hardware but still identifiable as a THP by software.

Hopefully I didn't miss something.

>> 
>> When a soft-dirty THP is migrated, set_pmd_migration_entry() preserves
>> soft-dirty with pmd_swp_mksoft_dirty(), setting the X bit in the
>> migration PMD. Even with _PAGE_PRESENT clear, we end up treating a
>> migration PMD as a present THP! The fault handler skips
>> pmd_migration_entry_wait(), and a write fault can end up in
>> do_huge_pmd_wp_page(), where pmd_page() decodes the migration entry
>> as a mapped PFN.
>
>That sounds bad.

YES, looks a bit off ...

>> 
>> Move the swap soft-dirty bit to bit 12 and start the swap offset at
>> bit 13 when CONFIG_MEM_SOFT_DIRTY is enabled. This keeps R/W/X clear
>> in migration PMDs and lets us keep the existing pmd_present() check
>> for invalidated THPs. Leave the offset at bit 12 when soft-dirty
>> tracking is disabled.
>
>That reduces the effective swap size (and PFN we can store). Could that be a
>problem?

We don't need all 52 bits of the swap offset.

RV64 PFNs only need 44 bits for migration entries, and actual swap is
already limited to about 16 TiB per area with 4 KiB pages by
last_page (__u32) and swap_info_struct.max (unsigned int).

So there's room to reserve a bit without reducing the supported swap
size or PFN range.

CONFIG_MEM_SOFT_DIRTY is only available on RV64, so RV32 keeps its 20-bit
offset.

[...]

Cheers, Lance