| 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
| ||
|
Message-ID: <35b64e41-b9b1-4fdb-af4f-1296602592ff@redhat.com>
Date: Tue, 7 Jan 2025 12:29:40 +0100
From: David Hildenbrand <david@...hat.com>
To: Alistair Popple <apopple@...dia.com>
Cc: akpm@...ux-foundation.org, dan.j.williams@...el.com, linux-mm@...ck.org,
lina@...hilina.net, zhang.lyra@...il.com, gerald.schaefer@...ux.ibm.com,
vishal.l.verma@...el.com, dave.jiang@...el.com, logang@...tatee.com,
bhelgaas@...gle.com, jack@...e.cz, jgg@...pe.ca, catalin.marinas@....com,
will@...nel.org, mpe@...erman.id.au, npiggin@...il.com,
dave.hansen@...ux.intel.com, ira.weiny@...el.com, willy@...radead.org,
djwong@...nel.org, tytso@....edu, linmiaohe@...wei.com, peterx@...hat.com,
linux-doc@...r.kernel.org, linux-kernel@...r.kernel.org,
linux-arm-kernel@...ts.infradead.org, linuxppc-dev@...ts.ozlabs.org,
nvdimm@...ts.linux.dev, linux-cxl@...r.kernel.org,
linux-fsdevel@...r.kernel.org, linux-ext4@...r.kernel.org,
linux-xfs@...r.kernel.org, jhubbard@...dia.com, hch@....de,
david@...morbit.com
Subject: Re: [PATCH v4 12/25] mm/memory: Enhance insert_page_into_pte_locked()
to create writable mappings
On 06.01.25 03:07, Alistair Popple wrote:
> On Fri, Dec 20, 2024 at 08:06:48PM +0100, David Hildenbrand wrote:
>> On 20.12.24 20:01, David Hildenbrand wrote:
>>> On 17.12.24 06:12, Alistair Popple wrote:
>>>> In preparation for using insert_page() for DAX, enhance
>>>> insert_page_into_pte_locked() to handle establishing writable
>>>> mappings. Recall that DAX returns VM_FAULT_NOPAGE after installing a
>>>> PTE which bypasses the typical set_pte_range() in finish_fault.
>>>>
>>>> Signed-off-by: Alistair Popple <apopple@...dia.com>
>>>> Suggested-by: Dan Williams <dan.j.williams@...el.com>
>>>>
>>>> ---
>>>>
>>>> Changes since v2:
>>>>
>>>> - New patch split out from "mm/memory: Add dax_insert_pfn"
>>>> ---
>>>> mm/memory.c | 45 +++++++++++++++++++++++++++++++++++++--------
>>>> 1 file changed, 37 insertions(+), 8 deletions(-)
>>>>
>>>> diff --git a/mm/memory.c b/mm/memory.c
>>>> index 06bb29e..cd82952 100644
>>>> --- a/mm/memory.c
>>>> +++ b/mm/memory.c
>>>> @@ -2126,19 +2126,47 @@ static int validate_page_before_insert(struct vm_area_struct *vma,
>>>> }
>>>> static int insert_page_into_pte_locked(struct vm_area_struct *vma, pte_t *pte,
>>>> - unsigned long addr, struct page *page, pgprot_t prot)
>>>> + unsigned long addr, struct page *page,
>>>> + pgprot_t prot, bool mkwrite)
>>>> {
>>>> struct folio *folio = page_folio(page);
>>>> + pte_t entry = ptep_get(pte);
>>>> pte_t pteval;
>>>> - if (!pte_none(ptep_get(pte)))
>>>> - return -EBUSY;
>>>> + if (!pte_none(entry)) {
>>>> + if (!mkwrite)
>>>> + return -EBUSY;
>>>> +
>>>> + /*
>>>> + * For read faults on private mappings the PFN passed in may not
>>>> + * match the PFN we have mapped if the mapped PFN is a writeable
>>>> + * COW page. In the mkwrite case we are creating a writable PTE
>>>> + * for a shared mapping and we expect the PFNs to match. If they
>>>> + * don't match, we are likely racing with block allocation and
>>>> + * mapping invalidation so just skip the update.
>>>> + */
>>>
>>> Would it make sense to instead have here
>>>
>>> /* See insert_pfn(). */
>>>
>>> But ...
>>>
>>>> + if (pte_pfn(entry) != page_to_pfn(page)) {
>>>> + WARN_ON_ONCE(!is_zero_pfn(pte_pfn(entry)));
>>>> + return -EFAULT;
>>>> + }
>>>> + entry = maybe_mkwrite(entry, vma);
>>>> + entry = pte_mkyoung(entry);
>>>> + if (ptep_set_access_flags(vma, addr, pte, entry, 1))
>>>> + update_mmu_cache(vma, addr, pte);
>>>
>>> ... I am not sure if we want the above at all. Someone inserted a page,
>>> which is refcounted + mapcounted already.
>>>
>>> Now you ignore that and do like the second insertion "worked" ?
>>>
>>> No, that feels wrong, I suspect you will run into refcount+mapcount issues.
>>>
>>> If there is already something, inserting must fail IMHO. If you want to
>>> change something to upgrade write permissions, then a different
>>> interface should be used.
>>
>> Ah, now I realize that the early exit saves you because we won't adjust the
>> refcount +mapcount.
>
> Right.
>
>> I still wonder if that really belongs in here, I would prefer to not play
>> such tricks to upgrade write permissions if possible.
>
> As you have pointed out this was all inspired (ie. mostly copied)
> from the existing insert_pfn() implementation which is used from
> vmf_insert_mixed{_mkwrite}().
>
> I agree a different interface to upgrade permissions would be nice. However
> it's tricky because in general callers of these functions (eg. FS DAX) aren't
> aware if the page is already mapped by a PTE/PMD. They only know a fault has
> occured and the faulting permissions.
>
> This wouldn't be impossible to fix - the mm does provide vm_ops->page_mkwrite()
> for permission upgrades. The difficulty is that most filesystems that support
> FS DAX (ie. ext4, XFS) don't treat a vm_ops->page_mkwrite() call any differently
> from a vm_ops->fault() call due to write fault. Therefore the FS DAX code is
> unaware of whether or not this is a permission upgrade or initial writeable
> mapping of the page in the VMA.
>
> A further issue in there is currently no vm_ops->huge_mkwrite() callback.
>
> Obviously this could all be plumbed through the MM/FS layers, but that would
> require a separate patch series. Given the current implementation has no issues
> beyond the cosmetic I'd rather not delay this series any longer, especially as
> the cosmetic defect is largely pre-existing (vmf_insert_mixed{_mkwrite}() could
> have equally had a separate upgrade interface).
Fine with me, just stumbled over it an thought "that looks odd".
--
Cheers,
David / dhildenb
Powered by blists - more mailing lists