Re: [PATCH v17 44/47] dept: introduce APIs to set page usage and use subclasses_evt for the usage

[Byungchul Park <byungchul@sk.com>]

On Wed, Nov 19, 2025 at 02:37:17PM +0000, Matthew Wilcox wrote:
> On Wed, Nov 19, 2025 at 07:53:12PM +0900, Byungchul Park wrote:
> > On Thu, Oct 02, 2025 at 05:12:44PM +0900, Byungchul Park wrote:
> > > False positive reports have been observed since dept works with the
> > > assumption that all the pages have the same dept class, but the class
> > > should be split since the problematic call paths are different depending
> > > on what the page is used for.
> > >
> > > At least, ones in block device's address_space and ones in regular
> > > file's address_space have exclusively different usages.
> > >
> > > Thus, define usage candidates like:
> > >
> > >    DEPT_PAGE_REGFILE_CACHE /* page in regular file's address_space */
> > >    DEPT_PAGE_BDEV_CACHE    /* page in block device's address_space */
> > >    DEPT_PAGE_DEFAULT       /* the others */
> >
> > 1. I'd like to annotate a page to DEPT_PAGE_REGFILE_CACHE when the page
> >    starts to be associated with a page cache for fs data.
> >
> > 2. And I'd like to annotate a page to DEPT_PAGE_BDEV_CACHE when the page
> >    starts to be associated with meta data of fs e.g. super block.
> >
> > 3. Lastly, I'd like to reset the annotated value if any, that has been
> >    set in the page, when the page ends the assoication with either page
> >    cache or meta block of fs e.g. freeing the page.
> >
> > Can anyone suggest good places in code for the annotation 1, 2, 3?  It'd
> > be totally appreciated. :-)
> 
> I don't think it makes sense to track lock state in the page (nor
> folio).  Partly bcause there's just so many of them, but also because
> the locking rules don't really apply to individual folios so much as
> they do to the mappings (or anon_vmas) that contain folios.

I've been trying to fully understand what you meant but maybe failed.

FWIW, dept is working based on classification, not instance by instance,
that is similar to lockdep.  This patch is for resolving issues that
might come from the fact that there is a **single class** for PG_locked,
by splitting the class to several ones according to their usages.

> If you're looking to find deadlock scenarios, I think it makes more
> sense to track all folio locks in a given mapping as the same lock
> type rather than track each folio's lock status.
> 
> For example, let's suppose we did something like this in the
> page fault path:
> 
> Look up and lock a folio (we need folios locked to insert them into
> the page tables to avoid a race with truncate)
> Try to allocate a page table
> Go into reclaim, attempt to reclaim a folio from this mapping

I think you are talking about nested lock patterns involving PG_locked.

Even though dept can do much more jobs than just tracking nested lock
patterns within a single context, of course, nested lock patterns
involving PG_locked should be handled appropriately, maybe with the
useful information you gave.  When I work on handling nested locks esp.
involving PG_locked, I will try to get you again.  Thanks.

However, I have no choice but to keep this approach for the **single
class** issue.  Feel free to ask if any.

	Byungchul

> We ought to detect that as a potential deadlock, regardless of which
> folio in the mapping we attempt to reclaim.  So can we track folio
> locking at the mapping/anon_vma level instead?
> 
> ---
> 
> My current understanding of folio locking rules:
> 
> If you hold a lock on folio A, you can take a lock on folio B if:
> 
> 1. A->mapping == B->mapping and A->index < B->index
>    (for example writeback; we take locks on all folios to be written
>     back in order)
> 2. !S_ISBLK(A->mapping->host) and S_ISBLK(B->mapping->host)
> 3. S_ISREG(A->mapping->host) and S_ISREG(B->mapping->host) with
>    inode_lock() held on both and A->index < B->index
>    (the remap_range code)