linux-kernel - Re: [PATCH v1] mm/huge_memory: improve split_huge_page_to_list_to

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <d97f8fdc-eb38-485e-81e2-51f748089939@redhat.com>
Date: Mon, 22 Apr 2024 21:31:26 +0200
From: David Hildenbrand <david@...hat.com>
To: John Hubbard <jhubbard@...dia.com>, linux-kernel@...r.kernel.org
Cc: linux-mm@...ck.org, Zi Yan <ziy@...dia.com>,
 Matthew Wilcox <willy@...radead.org>,
 Andrew Morton <akpm@...ux-foundation.org>
Subject: Re: [PATCH v1] mm/huge_memory: improve
 split_huge_page_to_list_to_order() return value documentation

On 19.04.24 02:15, John Hubbard wrote:
> On 4/18/24 8:18 AM, David Hildenbrand wrote:
>> The documentation is wrong and relying on it almost resulted in BUGs
>> in new callers: we return -EAGAIN on unexpected folio references, not
>> -EBUSY.
>>
>> Let's fix that and also document which other return values we can
>> currently see and why they could happen.
>>
>> Cc: John Hubbard <jhubbard@...dia.com>
>> Cc: Zi Yan <ziy@...dia.com>
>> Cc: Matthew Wilcox <willy@...radead.org>
>> Cc: Andrew Morton <akpm@...ux-foundation.org>
>> Signed-off-by: David Hildenbrand <david@...hat.com>
>> ---
>>    mm/huge_memory.c | 13 ++++++++++---
>>    1 file changed, 10 insertions(+), 3 deletions(-)
>>
>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>> index ee12726291f1b..824eff9211db8 100644
>> --- a/mm/huge_memory.c
>> +++ b/mm/huge_memory.c
>> @@ -2956,7 +2956,7 @@ bool can_split_folio(struct folio *folio, int *pextra_pins)
>>     *
>>     * 3) The folio must not be pinned. Any unexpected folio references, including
>>     *    GUP pins, will result in the folio not getting split; instead, the caller
>> - *    will receive an -EBUSY.
>> + *    will receive an -EAGAIN.
>>     *
>>     * 4) @new_order > 1, usually. Splitting to order-1 anonymous folios is not
>>     *    supported for non-file-backed folios, because folio->_deferred_list, which
>> @@ -2975,8 +2975,15 @@ bool can_split_folio(struct folio *folio, int *pextra_pins)
> 
> As an aside, the use of unconditional local_irq_disable() / local_irq_enable()
> calls in this routine almost makes me believe that we should have:
> 
> 5) Local IRQs should be enabled. Because this routine may enable them.
> 
> ...but I can't imagine a way to end up calling this with interrupts
> disabled, so it seems like documentation overkill. Just thought I'd mention
> it, though.

Yes, I think there might be more issues lurking with disabled interrupts.

anon_vma_lock_write() and i_mmap_lock_read() might even sleep ... so we
must not be in any atomic context. that's why relevant page table walkers drop the PTL.

> 
> 
>>     *
>>     * Returns 0 if the huge page was split successfully.
>>     *
>> - * Returns -EBUSY if @page's folio is pinned, or if the anon_vma disappeared
>> - * from under us.
>> + * Returns -EAGAIN if the folio has unexpected reference (e.g., GUP).
> 
> ...or if the folio was removed from the page cache before this routine
> got a chance to lock it, right? (See the "fail:" path.)

Right, that is sneaky. Let me extend to cover that case as well.

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 824eff9211db8..a7406267323ed 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -2975,7 +2975,8 @@ bool can_split_folio(struct folio *folio, int *pextra_pins)
   *
   * Returns 0 if the huge page was split successfully.
   *
- * Returns -EAGAIN if the folio has unexpected reference (e.g., GUP).
+ * Returns -EAGAIN if the folio has unexpected reference (e.g., GUP) or if
+ * the folio was concurrently removed from the page cache.
   *
   * Returns -EBUSY when trying to split the huge zeropage, if the folio is
   * under writeback, if fs-specific folio metadata cannot currently be


Naive me would assume that this happens rarely ... but not an expert :)

> 
>> + *
>> + * Returns -EBUSY when trying to split the huge zeropage, if the folio is
>> + * under writeback, if fs-specific folio metadata cannot currently be
>> + * released, or if some unexpected race happened (e.g., anon VMA disappeared,
>> + * truncation).
>> + *
>> + * Returns -EINVAL when trying to split to an order that is incompatible
>> + * with the folio. Splitting to order 0 is compatible with all folios.
>>     */
>>    int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
>>    				     unsigned int new_order)
> 
> Otherwise, looks good.

Thanks!

-- 
Cheers,

David / dhildenb