linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Asahi Lina <lina@asahilina.net>
To: Alice Ryhl <aliceryhl@google.com>
Cc: "Miguel Ojeda" <ojeda@kernel.org>,
	"Alex Gaynor" <alex.gaynor@gmail.com>,
	"Boqun Feng" <boqun.feng@gmail.com>,
	"Gary Guo" <gary@garyguo.net>,
	"Björn Roy Baron" <bjorn3_gh@protonmail.com>,
	"Benno Lossin" <benno.lossin@proton.me>,
	"Andreas Hindborg" <a.hindborg@kernel.org>,
	"Trevor Gross" <tmgross@umich.edu>,
	"Jann Horn" <jannh@google.com>,
	"Matthew Wilcox" <willy@infradead.org>,
	"Paolo Bonzini" <pbonzini@redhat.com>,
	"Danilo Krummrich" <dakr@kernel.org>,
	"Wedson Almeida Filho" <wedsonaf@gmail.com>,
	"Valentin Obst" <kernel@valentinobst.de>,
	"Andrew Morton" <akpm@linux-foundation.org>,
	linux-mm@kvack.org, airlied@redhat.com,
	"Abdiel Janulgue" <abdiel.janulgue@gmail.com>,
	rust-for-linux@vger.kernel.org, linux-kernel@vger.kernel.org,
	asahi@lists.linux.dev
Subject: Re: [PATCH 5/6] rust: page: Add physical address conversion functions
Date: Tue, 4 Feb 2025 20:43:57 +0900	[thread overview]
Message-ID: <c30d12dd-4489-4f6a-b40f-4afa9e0c807f@asahilina.net> (raw)
In-Reply-To: <CAH5fLgiRC3RHfF+2xH+7Ek+FkXDDDpZopqo-Q20VhkWhTk16Wg@mail.gmail.com>



On 2/3/25 6:35 PM, Alice Ryhl wrote:
> On Sun, Feb 2, 2025 at 2:06 PM Asahi Lina <lina@asahilina.net> wrote:
>>
>> Add methods to allow code using the Page type to obtain the physical
>> address of a page, convert to and from an (owned) physical address, and
>> borrow a Page from a physical address. Most of these operations are, as
>> you might expect, unsafe.
>>
>> These primitives are useful to implement page table structures in Rust,
>> and to implement arbitrary physical memory access (as needed to walk
>> arbitrary page tables and dereference through them). These mechanisms
>> are, of course, fraught with danger, and are only expected to be used
>> for core memory management code (in e.g. drivers with their own device
>> page table implementations) and for debug features such as crash dumps
>> of device memory.
>>
>> Signed-off-by: Asahi Lina <lina@asahilina.net>
>> ---
>>  rust/helpers/page.c | 26 +++++++++++++++++++++
>>  rust/kernel/page.rs | 65 +++++++++++++++++++++++++++++++++++++++++++++++++++++
>>  2 files changed, 91 insertions(+)
>>
>> diff --git a/rust/helpers/page.c b/rust/helpers/page.c
>> index b3f2b8fbf87fc9aa89cb1636736c52be16411301..1c3bd68818d77f7ce7806329b8f040a7d4205bb3 100644
>> --- a/rust/helpers/page.c
>> +++ b/rust/helpers/page.c
>> @@ -1,5 +1,6 @@
>>  // SPDX-License-Identifier: GPL-2.0
>>
>> +#include <linux/io.h>
>>  #include <linux/gfp.h>
>>  #include <linux/highmem.h>
>>
>> @@ -17,3 +18,28 @@ void rust_helper_kunmap_local(const void *addr)
>>  {
>>         kunmap_local(addr);
>>  }
>> +
>> +struct page *rust_helper_phys_to_page(phys_addr_t phys)
>> +{
>> +       return phys_to_page(phys);
>> +}
>> +
>> +phys_addr_t rust_helper_page_to_phys(struct page *page)
>> +{
>> +       return page_to_phys(page);
>> +}
>> +
>> +unsigned long rust_helper_phys_to_pfn(phys_addr_t phys)
>> +{
>> +       return __phys_to_pfn(phys);
>> +}
>> +
>> +struct page *rust_helper_pfn_to_page(unsigned long pfn)
>> +{
>> +       return pfn_to_page(pfn);
>> +}
>> +
>> +bool rust_helper_pfn_valid(unsigned long pfn)
>> +{
>> +       return pfn_valid(pfn);
>> +}
>> diff --git a/rust/kernel/page.rs b/rust/kernel/page.rs
>> index fe5f879f9d1a86083fd55c682fad9d52466f79a2..67cd7006fa63ab5aed4c4de2be639ed8e1fbc2ba 100644
>> --- a/rust/kernel/page.rs
>> +++ b/rust/kernel/page.rs
>> @@ -3,6 +3,7 @@
>>  //! Kernel page allocation and management.
>>
>>  use crate::{
>> +    addr::*,
>>      alloc::{AllocError, Flags},
>>      bindings,
>>      error::code::*,
>> @@ -10,6 +11,7 @@
>>      types::{Opaque, Ownable, Owned},
>>      uaccess::UserSliceReader,
>>  };
>> +use core::mem::ManuallyDrop;
>>  use core::ptr::{self, NonNull};
>>
>>  /// A bitwise shift for the page size.
>> @@ -249,6 +251,69 @@ pub unsafe fn copy_from_user_slice_raw(
>>              reader.read_raw(unsafe { core::slice::from_raw_parts_mut(dst.cast(), len) })
>>          })
>>      }
>> +
>> +    /// Returns the physical address of this page.
>> +    pub fn phys(&self) -> PhysicalAddr {
>> +        // SAFETY: `page` is valid due to the type invariants on `Page`.
>> +        unsafe { bindings::page_to_phys(self.as_ptr()) }
>> +    }
>> +
>> +    /// Converts a Rust-owned Page into its physical address.
>> +    ///
>> +    /// The caller is responsible for calling [`Page::from_phys()`] to avoid leaking memory.
>> +    pub fn into_phys(this: Owned<Self>) -> PhysicalAddr {
>> +        ManuallyDrop::new(this).phys()
>> +    }
>> +
>> +    /// Converts a physical address to a Rust-owned Page.
>> +    ///
>> +    /// # Safety
>> +    /// The caller must ensure that the physical address was previously returned by a call to
>> +    /// [`Page::into_phys()`], and that the physical address is no longer used after this call,
>> +    /// nor is [`Page::from_phys()`] called again on it.
>> +    pub unsafe fn from_phys(phys: PhysicalAddr) -> Owned<Self> {
>> +        // SAFETY: By the safety requirements, the physical address must be valid and
>> +        // have come from `into_phys()`, so phys_to_page() cannot fail and
>> +        // must return the original struct page pointer.
>> +        unsafe { Owned::from_raw(NonNull::new_unchecked(bindings::phys_to_page(phys)).cast()) }
>> +    }
>> +
>> +    /// Borrows a Page from a physical address, without taking over ownership.
>> +    ///
>> +    /// If the physical address does not have a `struct page` entry or is not
>> +    /// part of a System RAM region, returns None.
>> +    ///
>> +    /// # Safety
>> +    /// The caller must ensure that the physical address, if it is backed by a `struct page`,
>> +    /// remains available for the duration of the borrowed lifetime.
>> +    pub unsafe fn borrow_phys(phys: &PhysicalAddr) -> Option<&Self> {
>> +        // SAFETY: This is always safe, as it is just arithmetic
>> +        let pfn = unsafe { bindings::phys_to_pfn(*phys) };
>> +        // SAFETY: This function is safe to call with any pfn
>> +        if !unsafe { bindings::pfn_valid(pfn) && bindings::page_is_ram(pfn) != 0 } {
>> +            None
>> +        } else {
>> +            // SAFETY: We have just checked that the pfn is valid above, so it must
>> +            // have a corresponding struct page. By the safety requirements, we can
>> +            // return a borrowed reference to it.
>> +            Some(unsafe { &*(bindings::pfn_to_page(pfn) as *mut Self as *const Self) })
>> +        }
>> +    }
>> +
>> +    /// Borrows a Page from a physical address, without taking over ownership
>> +    /// nor checking for validity.
>> +    ///
>> +    /// # Safety
>> +    /// The caller must ensure that the physical address is backed by a `struct page` and
>> +    /// corresponds to System RAM. This is true when the address was returned by
>> +    /// [`Page::into_phys()`].
>> +    pub unsafe fn borrow_phys_unchecked(phys: &PhysicalAddr) -> &Self {
> 
> Should this be
> 
> pub unsafe fn borrow_phys_unchecked<'a>(phys: PhysicalAddr) -> &'a Self
> 
> ? That's how the signature of these raw methods usually goes, and then
> your safety requirements say that the requirements must hold for the
> duration of 'a.

The idea was to have *some* lifetime bound on an incoming variable
instead of just being able to return any arbitrary lifetime, but I don't
know if that's something worth doing / idiomatic in any way. Obviously
we can't stop a caller from doing something wrong if they really want to.

> 
>> +        // SAFETY: This is always safe, as it is just arithmetic
>> +        let pfn = unsafe { bindings::phys_to_pfn(*phys) };
>> +        // SAFETY: The caller guarantees that the pfn is valid. By the safety
>> +        // requirements, we can return a borrowed reference to it.
>> +        unsafe { &*(bindings::pfn_to_page(pfn) as *mut Self as *const Self) }
> 
> Can this just be
> 
> &*bindings::pfn_to_page(pfn).cast()

Yeah, I'll fix it for v2!

> 
> ?
> 
> Alice
> 
>> +    }
>>  }
>>
>>  // SAFETY: `Owned<Page>` objects returned by Page::alloc_page() follow the requirements of
>>
>> --
>> 2.47.1
>>
> 

~~ Lina



  reply	other threads:[~2025-02-04 11:44 UTC|newest]

Thread overview: 66+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-02-02 13:05 [PATCH 0/6] rust: page: Support borrowing `struct page` and physaddr conversion Asahi Lina
2025-02-02 13:05 ` [PATCH 1/6] rust: types: Add Ownable/Owned types Asahi Lina
2025-02-03  9:13   ` Alice Ryhl
2025-02-03 14:17     ` Asahi Lina
2025-02-03 18:17       ` Alice Ryhl
2025-02-03 19:17         ` Asahi Lina
2025-02-19  8:34           ` Andreas Hindborg
2025-02-19  8:37   ` Andreas Hindborg
2025-02-02 13:05 ` [PATCH 2/6] rust: page: Convert to Ownable Asahi Lina
2025-02-03  9:17   ` Alice Ryhl
2025-02-03  9:39   ` Fiona Behrens
2025-02-19  8:46   ` Andreas Hindborg
2025-02-02 13:05 ` [PATCH 3/6] rust: page: Make with_page_mapped() and with_pointer_into_page() public Asahi Lina
2025-02-03  9:10   ` Alice Ryhl
2025-02-03  9:43   ` Fiona Behrens
2025-02-19  8:48   ` Andreas Hindborg
2025-02-02 13:05 ` [PATCH 4/6] rust: addr: Add a module to declare core address types Asahi Lina
2025-02-03  9:09   ` Alice Ryhl
2025-02-03 15:04   ` Boqun Feng
2025-02-04 11:50     ` Asahi Lina
2025-02-04 14:50       ` Boqun Feng
2025-02-19  8:51   ` Andreas Hindborg
2025-02-02 13:05 ` [PATCH 5/6] rust: page: Add physical address conversion functions Asahi Lina
2025-02-03  9:35   ` Alice Ryhl
2025-02-04 11:43     ` Asahi Lina [this message]
2025-02-03  9:53   ` Fiona Behrens
2025-02-03 10:01     ` Alice Ryhl
2025-02-19  9:06   ` Andreas Hindborg
2025-02-02 13:05 ` [PATCH 6/6] rust: page: Make Page::as_ptr() pub(crate) Asahi Lina
2025-02-03  9:08   ` Alice Ryhl
2025-02-19  9:08   ` Andreas Hindborg
2025-02-03  9:58 ` [PATCH 0/6] rust: page: Support borrowing `struct page` and physaddr conversion Simona Vetter
2025-02-03 14:32   ` Asahi Lina
2025-02-03 21:05     ` Zi Yan
2025-02-04 10:26       ` David Hildenbrand
2025-02-04 11:41         ` Asahi Lina
2025-02-04 11:59           ` David Hildenbrand
2025-02-04 13:05             ` Asahi Lina
2025-02-04 14:38               ` David Hildenbrand
2025-02-04 17:59                 ` Asahi Lina
2025-02-04 20:10                   ` David Hildenbrand
2025-02-04 21:06                     ` Asahi Lina
2025-02-06 17:58                       ` David Hildenbrand
2025-02-06 19:18                         ` Asahi Lina
2025-02-06 19:27                           ` Asahi Lina
2025-02-12 19:06                             ` David Hildenbrand
2025-02-12 19:01                           ` David Hildenbrand
2025-02-05  7:40                 ` Simona Vetter
2025-02-12 19:07                   ` David Hildenbrand
2025-02-04 10:33   ` David Hildenbrand
2025-02-04 18:39     ` Jason Gunthorpe
2025-02-04 19:01       ` Asahi Lina
2025-02-04 20:05       ` David Hildenbrand
2025-02-04 20:26         ` Jason Gunthorpe
2025-02-04 20:41           ` David Hildenbrand
2025-02-04 20:47             ` David Hildenbrand
2025-02-04 21:18               ` Asahi Lina
2025-02-06 18:02                 ` David Hildenbrand
2025-02-04 20:49             ` Jason Gunthorpe
2025-02-05 23:17               ` Matthew Wilcox
2025-02-06 18:04                 ` David Hildenbrand
2025-02-03 10:22 ` Danilo Krummrich
2025-02-03 14:41   ` Asahi Lina
2025-02-15 19:47     ` Asahi Lina
2025-02-17  8:50       ` Abdiel Janulgue
2025-02-19  9:24         ` Andreas Hindborg

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=c30d12dd-4489-4f6a-b40f-4afa9e0c807f@asahilina.net \
    --to=lina@asahilina.net \
    --cc=a.hindborg@kernel.org \
    --cc=abdiel.janulgue@gmail.com \
    --cc=airlied@redhat.com \
    --cc=akpm@linux-foundation.org \
    --cc=alex.gaynor@gmail.com \
    --cc=aliceryhl@google.com \
    --cc=asahi@lists.linux.dev \
    --cc=benno.lossin@proton.me \
    --cc=bjorn3_gh@protonmail.com \
    --cc=boqun.feng@gmail.com \
    --cc=dakr@kernel.org \
    --cc=gary@garyguo.net \
    --cc=jannh@google.com \
    --cc=kernel@valentinobst.de \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=ojeda@kernel.org \
    --cc=pbonzini@redhat.com \
    --cc=rust-for-linux@vger.kernel.org \
    --cc=tmgross@umich.edu \
    --cc=wedsonaf@gmail.com \
    --cc=willy@infradead.org \
    /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