From: Anthony Yznaga <anthony.yznaga@oracle.com>
To: Jann Horn <jannh@google.com>
Cc: akpm@linux-foundation.org, willy@infradead.org,
markhemm@googlemail.com, viro@zeniv.linux.org.uk,
david@redhat.com, khalid@kernel.org, andreyknvl@gmail.com,
dave.hansen@intel.com, luto@kernel.org, brauner@kernel.org,
arnd@arndb.de, ebiederm@xmission.com, catalin.marinas@arm.com,
linux-arch@vger.kernel.org, linux-kernel@vger.kernel.org,
linux-mm@kvack.org, mhiramat@kernel.org, rostedt@goodmis.org,
vasily.averin@linux.dev, xhao@linux.alibaba.com, pcc@google.com,
neilb@suse.de, maz@kernel.org,
Lorenzo Stoakes <lorenzo.stoakes@oracle.com>,
Liam Howlett <liam.howlett@oracle.com>
Subject: Re: [PATCH v2 12/20] mm/mshare: prepare for page table sharing support
Date: Mon, 2 Jun 2025 15:02:38 -0700 [thread overview]
Message-ID: <1fe10a54-64c9-4808-9d4a-50edf5d1662a@oracle.com> (raw)
In-Reply-To: <CAG48ez0i527ibBCvZ_TF_PVt4OxfVTpS=_TYUKrk0cRQ10Bpxg@mail.gmail.com>
On 6/2/25 8:26 AM, Jann Horn wrote:
> On Fri, May 30, 2025 at 6:42 PM Anthony Yznaga
> <anthony.yznaga@oracle.com> wrote:
>> On 5/30/25 7:56 AM, Jann Horn wrote:
>>> On Fri, Apr 4, 2025 at 4:18 AM Anthony Yznaga <anthony.yznaga@oracle.com> wrote:
>>>> In preparation for enabling the handling of page faults in an mshare
>>>> region provide a way to link an mshare shared page table to a process
>>>> page table and otherwise find the actual vma in order to handle a page
>>>> fault. Modify the unmap path to ensure that page tables in mshare regions
>>>> are unlinked and kept intact when a process exits or an mshare region
>>>> is explicitly unmapped.
>>>>
>>>> Signed-off-by: Khalid Aziz <khalid@kernel.org>
>>>> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
>>>> Signed-off-by: Anthony Yznaga <anthony.yznaga@oracle.com>
>>> [...]
>>>> diff --git a/mm/memory.c b/mm/memory.c
>>>> index db558fe43088..68422b606819 100644
>>>> --- a/mm/memory.c
>>>> +++ b/mm/memory.c
>>> [...]
>>>> @@ -259,7 +260,10 @@ static inline void free_p4d_range(struct mmu_gather *tlb, pgd_t *pgd,
>>>> next = p4d_addr_end(addr, end);
>>>> if (p4d_none_or_clear_bad(p4d))
>>>> continue;
>>>> - free_pud_range(tlb, p4d, addr, next, floor, ceiling);
>>>> + if (unlikely(shared_pud))
>>>> + p4d_clear(p4d);
>>>> + else
>>>> + free_pud_range(tlb, p4d, addr, next, floor, ceiling);
>>>> } while (p4d++, addr = next, addr != end);
>>>>
>>>> start &= PGDIR_MASK;
>>> [...]
>>>> +static void mshare_vm_op_unmap_page_range(struct mmu_gather *tlb,
>>>> + struct vm_area_struct *vma,
>>>> + unsigned long addr, unsigned long end,
>>>> + struct zap_details *details)
>>>> +{
>>>> + /*
>>>> + * The msharefs vma is being unmapped. Do not unmap pages in the
>>>> + * mshare region itself.
>>>> + */
>>>> +}
>>>
>>> Unmapping a VMA has three major phases:
>>>
>>> 1. unlinking the VMA from the VMA tree
>>> 2. removing the VMA contents
>>> 3. removing unneeded page tables
>>>
>>> The MM subsystem broadly assumes that after phase 2, no stuff is
>>> mapped in the region anymore and therefore changes to the backing file
>>> don't need to TLB-flush this VMA anymore, and unlinks the mapping from
>>> rmaps and such. If munmap() of an mshare region only removes the
>>> mapping of shared page tables in step 3, as implemented here, that
>>> means things like TLB flushes won't be able to discover all
>>> currently-existing mshare mappings of a host MM through rmap walks.
>>>
>>> I think it would make more sense to remove the links to shared page
>>> tables in step 2 (meaning in mshare_vm_op_unmap_page_range), just like
>>> hugetlb does, and not modify free_pgtables().
>>
>> That makes sense. I'll make this change.
>
> Related: I think there needs to be a strategy for preventing walking
> of mshare host page tables through an mshare VMA by codepaths relying
> on MM/VMA locks, because those locks won't have an effect on the
> underlying host MM. For example, I think the only reason fork() is
> safe with your proposal is that copy_page_range() skips shared VMAs,
> and I think non-fast get_user_pages() could maybe hit use-after-free
> of page tables or such?
>
> I guess the only clean strategy for that is to ensure that all
> locking-based page table walking code does a check for "is this an
> mshare VMA?" and, if yes, either bails immediately or takes extra
> locks on the host MM (which could get messy).
Thanks. Yes, I need to audit all VMA / page table scanning. This series
already has a patch to avoid scanning mshare VMAs for numa migration,
but more issues are lurking.
next prev parent reply other threads:[~2025-06-02 22:03 UTC|newest]
Thread overview: 33+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-04-04 2:18 [PATCH v2 00/20] Add support for shared PTEs across processes Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 01/20] mm: Add msharefs filesystem Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 02/20] mm/mshare: pre-populate msharefs with information file Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 03/20] mm/mshare: make msharefs writable and support directories Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 04/20] mm/mshare: allocate an mm_struct for msharefs files Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 05/20] mm/mshare: add ways to set the size of an mshare region Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 06/20] mm/mshare: Add a vma flag to indicate " Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 07/20] mm/mshare: Add mmap support Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 08/20] mm/mshare: flush all TLBs when updating PTEs in an mshare range Anthony Yznaga
2025-05-30 14:41 ` Jann Horn
2025-05-30 16:29 ` Anthony Yznaga
2025-05-30 17:46 ` Jann Horn
2025-05-30 22:47 ` Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 09/20] sched/numa: do not scan msharefs vmas Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 10/20] mm: add mmap_read_lock_killable_nested() Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 11/20] mm: add and use unmap_page_range vm_ops hook Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 12/20] mm/mshare: prepare for page table sharing support Anthony Yznaga
2025-05-30 14:56 ` Jann Horn
2025-05-30 16:41 ` Anthony Yznaga
2025-06-02 15:26 ` Jann Horn
2025-06-02 22:02 ` Anthony Yznaga [this message]
2025-04-04 2:18 ` [PATCH v2 13/20] x86/mm: enable page table sharing Anthony Yznaga
2025-08-12 13:46 ` Yongting Lin
2025-08-12 17:12 ` Anthony Yznaga
2025-08-18 9:44 ` Yongting Lin
2025-08-20 1:32 ` Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 14/20] mm: create __do_mmap() to take an mm_struct * arg Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 15/20] mm: pass the mm in vma_munmap_struct Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 16/20] mm/mshare: Add an ioctl for mapping objects in an mshare region Anthony Yznaga
2025-04-04 2:18 ` [PATCH v2 17/20] mm/mshare: Add an ioctl for unmapping " Anthony Yznaga
2025-04-04 2:19 ` [PATCH v2 18/20] mm/mshare: provide a way to identify an mm as an mshare host mm Anthony Yznaga
2025-04-04 2:19 ` [PATCH v2 19/20] mm/mshare: get memcg from current->mm instead of mshare mm Anthony Yznaga
2025-04-04 2:19 ` [PATCH v2 20/20] mm/mshare: associate a mem cgroup with an mshare file Anthony Yznaga
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1fe10a54-64c9-4808-9d4a-50edf5d1662a@oracle.com \
--to=anthony.yznaga@oracle.com \
--cc=akpm@linux-foundation.org \
--cc=andreyknvl@gmail.com \
--cc=arnd@arndb.de \
--cc=brauner@kernel.org \
--cc=catalin.marinas@arm.com \
--cc=dave.hansen@intel.com \
--cc=david@redhat.com \
--cc=ebiederm@xmission.com \
--cc=jannh@google.com \
--cc=khalid@kernel.org \
--cc=liam.howlett@oracle.com \
--cc=linux-arch@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=lorenzo.stoakes@oracle.com \
--cc=luto@kernel.org \
--cc=markhemm@googlemail.com \
--cc=maz@kernel.org \
--cc=mhiramat@kernel.org \
--cc=neilb@suse.de \
--cc=pcc@google.com \
--cc=rostedt@goodmis.org \
--cc=vasily.averin@linux.dev \
--cc=viro@zeniv.linux.org.uk \
--cc=willy@infradead.org \
--cc=xhao@linux.alibaba.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®