`Zeroable` improvements & bindings integration
Subsystems (15)
Benno Lossin
reviewerv3 cover `Zeroable` improvements & bindings integration
This came from a discussion at [1]. And I added some more useful parts to the original idea.
I'm not sure on the impact of build times and rust-analyzer. We're adding a derive macro to every struct and union emitted by bindgen. Building using my test-config took 27.7s before and 28.2s after this change, but those are only two runs on my machine with a very reduced config (that enables all Rust code that we have at the moment).
Maybe we have to reevaluate this when more C code is scanned by bindgen.
Changelog
=========
* v3:
- removed already applied patches
- added more patches replacing usage of mem::zeroed and MaybeUninit::zeroed().assume_init() with the new pin_init::zeroed
- fix rusttest target
* v2:
- added patches with more extensive pin-init changes to Zeroable & related methods. for the upstream PR, see
https://github.com/Rust-for-Linux/pin-init/pull/56
- added patches replacing usage of mem::zeroed and MaybeUninit::zeroed().assume_init() with the new pin_init::zeroed
- fix rust-analyzer support
- use import in {uapi,bindings}/lib.rs to avoid the pin_init:: prefix
* v1: https://lore.kernel.org/all/[email protected]
Benno Lossin (11):
rust: add pin-init as a dependency to bindings and uapi
rust: derive Zeroable for all structs & unions generated by bindgen
where possible
rust: miscdevice: replace MaybeUninit::zeroed().assume_init() with
`pin_init::zeroed()`
rust: phy: replace `MaybeUninit::zeroed().assume_init()` with
`pin_init::zeroed()`
rust: block: replace `core::mem::zeroed` with `pin_init::zeroed`
rust: of: replace `core::mem::zeroed` with `pin_init::zeroed`
rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
rust: drm: replace `core::mem::zeroed` with `pin_init::zeroed`
rust: auxiliary: replace `core::mem::zeroed` with `pin_init::zeroed`
rust: acpi: replace `core::mem::zeroed` with `pin_init::zeroed`
rust: cpufreq: replace `MaybeUninit::zeroed().assume_init()` with
`pin_init::zeroed()`rust/Makefile | 14 ++++++++------ rust/bindgen_parameters | 4 ++++ rust/bindings/lib.rs | 8 ++++++++ rust/kernel/acpi.rs | 4 +--- rust/kernel/auxiliary.rs | 6 +----- rust/kernel/block/mq/gen_disk.rs | 3 +-- rust/kernel/block/mq/tag_set.rs | 4 +--- rust/kernel/cpufreq.rs | 4 +--- rust/kernel/drm/gem/mod.rs | 4 +--- rust/kernel/miscdevice.rs | 8 +++----- rust/kernel/net/phy.rs | 4 +--- rust/kernel/of.rs | 4 +--- rust/kernel/security.rs | 3 +-- rust/uapi/lib.rs | 2 ++ scripts/generate_rust_analyzer.py | 4 ++-- 15 files changed, 36 insertions(+), 40 deletions(-)
base-commit: 8f5ae30d69d7543eee0d70083daf4de8fe15d585
Signature · 1 line
2.50.1
Benno Lossin
reviewerPATCH v3 1/11 rust: add `pin-init` as a dependency to `bindings` and `uapi`
This allows bindings and uapi to implement Zeroable and use other items from pin-init.
--- rust/Makefile | 14 ++++++++------ scripts/generate_rust_analyzer.py | 4 ++-- 2 files changed, 10 insertions(+), 8 deletions(-)
rust/Makefile +8 -6
@@ -184,12 +184,12 @@ rusttestlib-kernel: $(src)/kernel/lib.rs rusttestlib-bindings rusttestlib-uapi \ $(obj)/bindings.o FORCE +$(call if_changed,rustc_test_library) -rusttestlib-bindings: private rustc_target_flags = --extern ffi -rusttestlib-bindings: $(src)/bindings/lib.rs rusttestlib-ffi FORCE +rusttestlib-bindings: private rustc_target_flags = --extern ffi --extern pin_init +rusttestlib-bindings: $(src)/bindings/lib.rs rusttestlib-ffi rusttestlib-pin_init FORCE +$(call if_changed,rustc_test_library) -rusttestlib-uapi: private rustc_target_flags = --extern ffi -rusttestlib-uapi: $(src)/uapi/lib.rs rusttestlib-ffi FORCE +rusttestlib-uapi: private rustc_target_flags = --extern ffi --extern pin_init +rusttestlib-uapi: $(src)/uapi/lib.rs rusttestlib-ffi rusttestlib-pin_init FORCE +$(call if_changed,rustc_test_library) quiet_cmd_rustdoc_test = RUSTDOC T $<
@@ -521,17 +521,19 @@ $(obj)/ffi.o: private skip_gendwarfksyms = 1 $(obj)/ffi.o: $(src)/ffi.rs $(obj)/compiler_builtins.o FORCE +$(call if_changed_rule,rustc_library) -$(obj)/bindings.o: private rustc_target_flags = --extern ffi +$(obj)/bindings.o: private rustc_target_flags = --extern ffi --extern pin_init $(obj)/bindings.o: $(src)/bindings/lib.rs \ $(obj)/ffi.o \ + $(obj)/pin_init.o \ $(obj)/bindings/bindings_generated.rs \ $(obj)/bindings/bindings_helpers_generated.rs FORCE +$(call if_changed_rule,rustc_library) -$(obj)/uapi.o: private rustc_target_flags = --extern ffi +$(obj)/uapi.o: private rustc_target_flags = --extern ffi --extern pin_init $(obj)/uapi.o: private skip_gendwarfksyms = 1 $(obj)/uapi.o: $(src)/uapi/lib.rs \ $(obj)/ffi.o \ + $(obj)/pin_init.o \ $(obj)/uapi/uapi_generated.rs FORCE +$(call if_changed_rule,rustc_library)
scripts/generate_rust_analyzer.py +2 -2
@@ -139,8 +139,8 @@ def generate_crates(srctree, objtree, sysroot_src, external_src, cfgs, core_edit "exclude_dirs": [], } - append_crate_with_generated("bindings", ["core", "ffi"]) - append_crate_with_generated("uapi", ["core", "ffi"]) + append_crate_with_generated("bindings", ["core", "ffi", "pin_init"]) + append_crate_with_generated("uapi", ["core", "ffi", "pin_init"]) append_crate_with_generated("kernel", ["core", "macros", "build_error", "pin_init", "ffi", "bindings", "uapi"]) def is_root_crate(build_file, target):
Signature · 1 line
2.50.1
Re: [PATCH v3 01/11] rust: add `pin-init` as a dependency to `bindings` and `uapi`
Quoted 8 lines · Benno Lossin
This allows
bindingsanduapito implementZeroableand use other items from pin-init.
Quoted 65 lines
rust/Makefile | 14 ++++++++------ scripts/generate_rust_analyzer.py | 4 ++-- 2 files changed, 10 insertions(+), 8 deletions(-)rust/Makefile +8
-6@@ -184,12 +184,12 @@ rusttestlib-kernel: $(src)/kernel/lib.rs rusttestlib-bindings rusttestlib-uapi \ $(obj)/bindings.o FORCE +$(call if_changed,rustc_test_library) -rusttestlib-bindings: private rustc_target_flags = --extern ffi -rusttestlib-bindings: $(src)/bindings/lib.rs rusttestlib-ffi FORCE +rusttestlib-bindings: private rustc_target_flags = --extern ffi --extern pin_init +rusttestlib-bindings: $(src)/bindings/lib.rs rusttestlib-ffi rusttestlib-pin_init FORCE +$(call if_changed,rustc_test_library) -rusttestlib-uapi: private rustc_target_flags = --extern ffi -rusttestlib-uapi: $(src)/uapi/lib.rs rusttestlib-ffi FORCE +rusttestlib-uapi: private rustc_target_flags = --extern ffi --extern pin_init +rusttestlib-uapi: $(src)/uapi/lib.rs rusttestlib-ffi rusttestlib-pin_init FORCE +$(call if_changed,rustc_test_library) quiet_cmd_rustdoc_test = RUSTDOC T $<@@ -521,17 +521,19 @@ $(obj)/ffi.o: private skip_gendwarfksyms = 1 $(obj)/ffi.o: $(src)/ffi.rs $(obj)/compiler_builtins.o FORCE +$(call if_changed_rule,rustc_library) -$(obj)/bindings.o: private rustc_target_flags = --extern ffi +$(obj)/bindings.o: private rustc_target_flags = --extern ffi --extern pin_init $(obj)/bindings.o: $(src)/bindings/lib.rs \ $(obj)/ffi.o \ + $(obj)/pin_init.o \ $(obj)/bindings/bindings_generated.rs \ $(obj)/bindings/bindings_helpers_generated.rs FORCE +$(call if_changed_rule,rustc_library) -$(obj)/uapi.o: private rustc_target_flags = --extern ffi +$(obj)/uapi.o: private rustc_target_flags = --extern ffi --extern pin_init $(obj)/uapi.o: private skip_gendwarfksyms = 1 $(obj)/uapi.o: $(src)/uapi/lib.rs \ $(obj)/ffi.o \ + $(obj)/pin_init.o \ $(obj)/uapi/uapi_generated.rs FORCE +$(call if_changed_rule,rustc_library)scripts/generate_rust_analyzer.py +2
-3@@ -139,8 +139,8 @@ def generate_crates(srctree, objtree, sysroot_src, external_src, cfgs, core_edit "exclude_dirs": [], } - append_crate_with_generated("bindings", ["core", "ffi"]) - append_crate_with_generated("uapi", ["core", "ffi"]) + append_crate_with_generated("bindings", ["core", "ffi", "pin_init"]) + append_crate_with_generated("uapi", ["core", "ffi", "pin_init"]) append_crate_with_generated("kernel", ["core", "macros", "build_error", "pin_init", "ffi", "bindings", "uapi"]) def is_root_crate(build_file, target): --2.50.1
Benno Lossin
reviewerPATCH v3 2/11 rust: derive `Zeroable` for all structs & unions generated by bindgen where possible
In reply to Benno LossinUsing the --with-derive-custom-{struct,union} option of bindgen, add #[derive(MaybeZeroable)] to every struct & union. This makes those types implement Zeroable if all their fields implement it.
Sadly bindgen doesn't add custom derives to the __BindgenBitfieldUnit struct. So manually implement Zeroable for that.
--- rust/bindgen_parameters | 4 ++++ rust/bindings/lib.rs | 8 ++++++++ rust/uapi/lib.rs | 2 ++ 3 files changed, 14 insertions(+)
rust/bindgen_parameters +4 -0
@@ -34,3 +34,7 @@ # We use const helpers to aid bindgen, to avoid conflicts when constants are # recognized, block generation of the non-helper constants. --blocklist-item ARCH_SLAB_MINALIGN + +# Structs should implement Zeroable when all of their fields do. +--with-derive-custom-struct .*=MaybeZeroable +--with-derive-custom-union .*=MaybeZeroable
rust/bindings/lib.rs +8 -0
@@ -31,11 +31,19 @@#[allow(clippy::undocumented_unsafe_blocks)] #[cfg_attr( CONFIG_RUSTC_HAS_UNNECESSARY_TRANSMUTES , allow(unnecessary_transmutes))] mod bindings_raw { +use pin_init::{MaybeZeroable ,Zeroable }; +// Manual definition for blocklisted types. type __kernel_size_t =usize ;type __kernel_ssize_t =isize ;type __kernel_ptrdiff_t =isize ; +// `bindgen` doesn't automatically do this, see +// <https://github.com/rust-lang/rust-bindgen/issues/3196> +// +// SAFETY: `__BindgenBitfieldUnit<Storage>` is a newtype around `Storage`. +unsafe impl <Storage >Zeroable for __BindgenBitfieldUnit <Storage >where Storage :Zeroable {} +// Use glob import here to expose all helpers. // Symbols defined within the module will take precedence to the glob import. pub use super ::bindings_helper::*;
rust/uapi/lib.rs +2 -0
@@ -34,4 +34,6 @@type __kernel_ssize_t =isize ;type __kernel_ptrdiff_t =isize ; +use pin_init::MaybeZeroable ; +include! (concat! (env! ("OBJTREE" ),"/rust/uapi/uapi_generated.rs" ));
Signature · 1 line
2.50.1
Re: [PATCH v3 02/11] rust: derive `Zeroable` for all structs & unions generated by bindgen where possible
Quoted 9 lines · Benno Lossin
Using the
--with-derive-custom-{struct,union}option of bindgen, add#[derive(MaybeZeroable)]to every struct & union. This makes those types implementZeroableif all their fields implement it.Sadly bindgen doesn't add custom derives to the
__BindgenBitfieldUnitstruct. So manually implementZeroablefor that.Signed-off-by Benno Lossin <[email protected]>
Benno Lossin
reviewerPATCH v3 3/11 rust: miscdevice: replace `MaybeUninit::zeroed().assume_init()` with `pin_init::zeroed()`
In reply to Benno LossinAll types in bindings implement Zeroable if they can, so use pin_init::zeroed instead of relying on unsafe code.
If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
--- rust/kernel/miscdevice.rs | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-)
rust/kernel/miscdevice.rs +3 -5
@@ -19,7 +19,7 @@ seq_file::SeqFile, types::{ForeignOwnable, Opaque}, }; -use core::{marker::PhantomData , mem::MaybeUninit , pin::Pin }; +use core::{marker::PhantomData , pin::Pin };/// Options for creating a misc device. #[derive( Copy , Clone )]
@@ -31,8 +31,7 @@ pub struct MiscDeviceOptions {impl MiscDeviceOptions {/// Create a raw `struct miscdev` ready for registration. pub const fn into_raw <T :MiscDevice >(self) -> bindings::miscdevice { -// SAFETY: All zeros is valid for this C type. -let mut result: bindings::miscdevice =unsafe {MaybeUninit ::zeroed ().assume_init () }; +let mut result: bindings::miscdevice = pin_init::zeroed (); result.minor = bindings::MISC_DYNAMIC_MINOR as ffi::c_int ; result.name = self.name.as_char_ptr (); result.fops =MiscdeviceVTable ::<T >::build ();
@@ -359,8 +358,7 @@ impl<T: MiscDevice> MiscdeviceVTable<T> { }else {None }, -// SAFETY: All zeros is a valid value for `bindings::file_operations`. - ..unsafe {MaybeUninit ::zeroed ().assume_init () } + ..pin_init::zeroed () };const fn build () -> &'static bindings::file_operations {
Signature · 1 line
2.50.1
Benno Lossin
reviewerPATCH v3 4/11 rust: phy: replace `MaybeUninit::zeroed().assume_init()` with `pin_init::zeroed()`
In reply to Benno LossinAll types in bindings implement Zeroable if they can, so use pin_init::zeroed instead of relying on unsafe code.
If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
--- rust/kernel/net/phy.rs | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)
rust/kernel/net/phy.rs +1 -3
@@ -556,9 +556,7 @@ pub const fn create_phy_driver<T: Driver>() -> DriverVTable { }else {None }, -// SAFETY: The rest is zeroed out to initialize `struct phy_driver`, -// sets `Option<&F>` to be `None`. - ..unsafe { core::mem::MaybeUninit ::<bindings::phy_driver >::zeroed ().assume_init () } + ..pin_init::zeroed () })) }
Signature · 1 line
2.50.1
Benno Lossin
reviewerPATCH v3 5/11 rust: block: replace `core::mem::zeroed` with `pin_init::zeroed`
In reply to Benno LossinAll types in bindings implement Zeroable if they can, so use pin_init::zeroed instead of relying on unsafe code.
If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
--- rust/kernel/block/mq/gen_disk.rs | 3 +-- rust/kernel/block/mq/tag_set.rs | 4 +--- 2 files changed, 2 insertions(+), 5 deletions(-)
rust/kernel/block/mq/gen_disk.rs +1 -2
@@ -93,8 +93,7 @@ pub fn build<T: Operations>( name: fmt::Arguments<'_>, tagset: Arc<TagSet<T>>, ) -> Result<GenDisk<T>> { -// SAFETY: `bindings::queue_limits` contain only fields that are valid when zeroed. -let mut lim: bindings::queue_limits =unsafe { core::mem::zeroed () }; +let mut lim: bindings::queue_limits = pin_init::zeroed (); lim.logical_block_size = self.logical_block_size; lim.physical_block_size = self.physical_block_size;
rust/kernel/block/mq/tag_set.rs +1 -3
@@ -38,9 +38,7 @@ pub fn new( num_tags: u32, num_maps: u32, ) -> impl PinInit<Self, error::Error> { -// SAFETY: `blk_mq_tag_set` only contains integers and pointers, which -// all are allowed to be 0. -let tag_set: bindings::blk_mq_tag_set =unsafe { core::mem::zeroed () }; +let tag_set: bindings::blk_mq_tag_set = pin_init::zeroed ();let tag_set:Result <_ > = core::mem::size_of ::<RequestDataWrapper >() .try_into () .map(|cmd_size| {
Signature · 1 line
2.50.1
Re: [PATCH v3 05/11] rust: block: replace `core::mem::zeroed` with `pin_init::zeroed`
Quoted 7 lines · Benno Lossin
All types in
bindingsimplementZeroableif they can, so usepin_init::zeroedinstead of relying onunsafecode.If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
Signed-off-by Benno Lossin <[email protected]>
Best regards,
Andreas Hindborg
Benno Lossin
reviewerPATCH v3 6/11 rust: of: replace `core::mem::zeroed` with `pin_init::zeroed`
In reply to Benno LossinAll types in bindings implement Zeroable if they can, so use pin_init::zeroed instead of relying on unsafe code.
If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
--- rust/kernel/of.rs | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)
rust/kernel/of.rs +1 -3
@@ -35,9 +35,7 @@ impl DeviceId {/// Create a new device id from an OF 'compatible' string. pub const fn new (compatible: &'staticCStr ) ->Self {let src = compatible.as_bytes_with_nul (); -// Replace with `bindings::of_device_id::default()` once stabilized for `const`. -// SAFETY: FFI type is valid to be zero-initialized. -let mut of: bindings::of_device_id =unsafe { core::mem::zeroed () }; +let mut of: bindings::of_device_id = pin_init::zeroed ();// TODO: Use `copy_from_slice` once stabilized for `const`. let mut i =0 ;
Signature · 1 line
2.50.1
Benno Lossin
reviewerPATCH v3 7/11 rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
In reply to Benno LossinAll types in bindings implement Zeroable if they can, so use pin_init::zeroed instead of relying on unsafe code.
If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
--- rust/kernel/security.rs | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
rust/kernel/security.rs +1 -2
@@ -25,8 +25,7 @@ impl SecurityCtx {/// Get the security context given its id. #[inline] pub fn from_secid (secid:u32 ) ->Result <Self > { -// SAFETY: `struct lsm_context` can be initialized to all zeros. -let mut ctx: bindings::lsm_context =unsafe { core::mem::zeroed () }; +let mut ctx: bindings::lsm_context = pin_init::zeroed ();// SAFETY: Just a C FFI call. The pointer is valid for writes. to_result (unsafe { bindings::security_secid_to_secctx (secid, &mut ctx) })?;
Signature · 1 line
2.50.1
Paul Moore
maintainerRe: [PATCH v3 07/11] rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
Quoted 11 lines · Benno Lossin
All types in
bindingsimplementZeroableif they can, so usepin_init::zeroedinstead of relying onunsafecode.If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
Signed-off-by Benno Lossin <[email protected]>--- rust/kernel/security.rs | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
I'm happy to take this via the LSM tree, but it would be nice to see a Reviewed-by/Acked-by from someone with a better understanding of Rust :)
Alice, you've reviewed/submitted LSM rust bits in the past, can you look at this?
Quoted 16 lines
rust/kernel/security.rs +1
-3@@ -25,8 +25,7 @@ impl SecurityCtx {/// Get the security context given its id. #[inline] pub fn from_secid (secid:u32 ) ->Result <Self > { -// SAFETY: `struct lsm_context` can be initialized to all zeros. -let mut ctx: bindings::lsm_context =unsafe { core::mem::zeroed () }; +let mut ctx: bindings::lsm_context = pin_init::zeroed ();// SAFETY: Just a C FFI call. The pointer is valid for writes. to_result (unsafe { bindings::security_secid_to_secctx (secid, &mut ctx) })?; --2.50.1
Signature · 1 line
paul-moore.com
Re: [PATCH v3 07/11] rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
Quoted 4 lines · Paul Moore
I'm happy to take this via the LSM tree, but it would be nice to see a Reviewed-by/Acked-by from someone with a better understanding of Rust :)
I think the idea is to take all these through the Rust one with Acked-bys from the maintainers (or we can skip this one and do it in a future cycle when the first patches get in).
In any case, Benno is very knowledgeable in Rust -- he wrote the library being called here -- so unless you see something out of the ordinary, it seems OK to me.
Thanks!
Cheers,
Miguel
Re: [PATCH v3 07/11] rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
Quoted 2 lines · Miguel Ojeda
I think the idea is to take all these through the Rust one with
Or through the pin-init one, which gets pulled into the Rust one.
Cheers,
Miguel
Benno Lossin
reviewerRe: [PATCH v3 07/11] rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
Quoted 6 lines · Miguel Ojeda
Quoted 2 lines · Miguel Ojeda
I think the idea is to take all these through the Rust one with
Or through the pin-init one, which gets pulled into the Rust one.
There are no pin-init changes in this patchset (in v2 there were, but I took them for v6.16), so I think it should go via the rust tree.
We can of course wait with patches 3-11 until the next cycle and let maintainers pick them individually, but since they are so small, I think it's easier to just pick them all at once via the rust tree.
---
Cheers,
Benno
Re: [PATCH v3 07/11] rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
Quoted 3 lines · Benno Lossin
There are no pin-init changes in this patchset (in v2 there were, but I took them for v6.16), so I think it should go via the rust tree.
Sounds good.
Quoted 3 lines
We can of course wait with patches 3-11 until the next cycle and let maintainers pick them individually, but since they are so small, I think it's easier to just pick them all at once via the rust tree.
Yeah, either way is fine. I would say let's just land the ones we get Acked-bys for, even if they are not all.
Cheers,
Miguel
Paul Moore
maintainerRe: [PATCH v3 07/11] rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
In reply to Miguel OjedaQuoted 9 lines · Miguel Ojeda
Quoted 4 lines · Paul Moore
I'm happy to take this via the LSM tree, but it would be nice to see a Reviewed-by/Acked-by from someone with a better understanding of Rust :)
I think the idea is to take all these through the Rust one with Acked-bys from the maintainers (or we can skip this one and do it in a future cycle when the first patches get in).
[CC'd the LSM list, as I just realized it wasn't on the original patch posting; in the future please include the LSM list on LSM related Rust patchsets/patches]
That's fine, it wasn't clear from the post that was the plan, and I vaguely recalled from past conversations with Rust devs that they preferred Rust wrappers/helpers to go in via the associated subsystem tree.
Quoted 3 lines
In any case, Benno is very knowledgeable in Rust -- he wrote the library being called here -- so unless you see something out of the ordinary, it seems OK to me.
My comment asking for additional review/ACK tags wasn't due to any distrust of Benno - thank you for your work Benno - it is just a matter of trying to make sure there are at least two sets of (knowledgeable) eyes on a patch before it is merged. If it is something I'm merging into one the trees I maintain, normally I count myself as the second set of eyes, but in this case I don't (yet) consider myself a knowledgeable Rust reviewer so I was asking for an additional explicit review tag. If someone else is going to merge this patch{set}, then it's up to them; although I would hope they would do something similar.
Signature · 1 line
paul-moore.com
Benno Lossin
reviewerRe: [PATCH v3 07/11] rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
Quoted 8 lines · Paul Moore
Quoted 6 lines · Miguel Ojeda
Quoted 4 lines · Paul Moore
I'm happy to take this via the LSM tree, but it would be nice to see a Reviewed-by/Acked-by from someone with a better understanding of Rust :)
Quoted 7 lines
Quoted 3 lines
I think the idea is to take all these through the Rust one with Acked-bys from the maintainers (or we can skip this one and do it in a future cycle when the first patches get in).
[CC'd the LSM list, as I just realized it wasn't on the original patch posting; in the future please include the LSM list on LSM related Rust patchsets/patches]
I checked and I didn't find a maintainers entry for that this file & the LSM list. I'm using scripts/get_maintainer.pl to get the people I send patches to and that also checks git commits, so I guess it added you through that (which is very good :). So can we add a maintainers entry for rust/kernel/security.rs so people don't miss this in the future? Thanks!
---
Cheers,
Benno
Paul Moore
maintainerRe: [PATCH v3 07/11] rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
Quoted 17 lines · Benno Lossin
Quoted 8 lines · Paul Moore
Quoted 6 lines · Miguel Ojeda
Quoted 4 lines · Paul Moore
I'm happy to take this via the LSM tree, but it would be nice to see a Reviewed-by/Acked-by from someone with a better understanding of Rust :)
Quoted 7 lines
Quoted 3 lines
I think the idea is to take all these through the Rust one with Acked-bys from the maintainers (or we can skip this one and do it in a future cycle when the first patches get in).
[CC'd the LSM list, as I just realized it wasn't on the original patch posting; in the future please include the LSM list on LSM related Rust patchsets/patches]
Quoted 6 lines
I checked and I didn't find a maintainers entry for that this file & the LSM list. I'm using
scripts/get_maintainer.plto get the people I send patches to and that also checks git commits, so I guess it added you through that (which is very good :). So can we add a maintainers entry forrust/kernel/security.rsso people don't miss this in the future? Thanks!
Here ya go ...
https://lore.kernel.org/linux-security-module/[email protected]/
Signature · 1 line
paul-moore.com
Re: [PATCH v3 07/11] rust: security: replace `core::mem::zeroed` with `pin_init::zeroed`
In reply to Paul MooreQuoted 5 lines · Paul Moore
That's fine, it wasn't clear from the post that was the plan, and I vaguely recalled from past conversations with Rust devs that they preferred Rust wrappers/helpers to go in via the associated subsystem tree.
Yeah, it is still the case that Rust abstractions and Rust code in general should land through the best tree possible.
However, sometimes tree-wide improvements may be easy to do with simple Acked-bys.
Quoted 2 lines
My comment asking for additional review/ACK tags wasn't due to any distrust of Benno - thank you for your work Benno - it is just a
Ah, sorry, I didn't mean that you distrusted Benno or anything like that. I was trying to give context in case it helped you evaluate the patch/risk, which also allowed me at the same time to acknowledge Benno's experience/work.
We were mainly expecting the Acked-bys to proceed with the cleanups here to remove these unsafe blocks, but we definitely want more tags if possible in all patches, as always (by the time this gets applied, we will hopefully get some more).
Cheers,
Miguel
Benno Lossin
reviewerPATCH v3 8/11 rust: drm: replace `core::mem::zeroed` with `pin_init::zeroed`
In reply to Benno LossinAll types in bindings implement Zeroable if they can, so use pin_init::zeroed instead of relying on unsafe code.
If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
--- rust/kernel/drm/gem/mod.rs | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)
rust/kernel/drm/gem/mod.rs +1 -3
@@ -308,9 +308,7 @@ impl<T: DriverObject> AllocImpl for Object<T> { }pub (super )const fn create_fops () -> bindings::file_operations { -// SAFETY: As by the type invariant, it is safe to initialize `bindings::file_operations` -// zeroed. -let mut fops: bindings::file_operations =unsafe { core::mem::zeroed () }; +let mut fops: bindings::file_operations = pin_init::zeroed (); fops.owner = core::ptr::null_mut (); fops.open =Some (bindings::drm_open);
Signature · 1 line
2.50.1
Benno Lossin
reviewerPATCH v3 9/11 rust: auxiliary: replace `core::mem::zeroed` with `pin_init::zeroed`
In reply to Benno LossinAll types in bindings implement Zeroable if they can, so use pin_init::zeroed instead of relying on unsafe code.
If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
--- rust/kernel/auxiliary.rs | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-)
rust/kernel/auxiliary.rs +1 -5
@@ -108,11 +108,7 @@ pub const fn new(modname: &'static CStr, name: &'static CStr) -> Self {let name = name.as_bytes_with_nul ();let modname = modname.as_bytes_with_nul (); -// TODO: Replace with `bindings::auxiliary_device_id::default()` once stabilized for -// `const`. -// -// SAFETY: FFI type is valid to be zero-initialized. -let mut id: bindings::auxiliary_device_id =unsafe { core::mem::zeroed () }; +let mut id: bindings::auxiliary_device_id = pin_init::zeroed ();let mut i =0 ;while i < modname.len () {
Signature · 1 line
2.50.1
Benno Lossin
reviewerPATCH v3 10/11 rust: acpi: replace `core::mem::zeroed` with `pin_init::zeroed`
In reply to Benno LossinAll types in bindings implement Zeroable if they can, so use pin_init::zeroed instead of relying on unsafe code.
If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
--- rust/kernel/acpi.rs | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)
rust/kernel/acpi.rs +1 -3
@@ -42,9 +42,7 @@ pub const fn new(id: &'static CStr) -> Self {"ID exceeds 16 bytes" );let src = id.as_bytes_with_nul (); -// Replace with `bindings::acpi_device_id::default()` once stabilized for `const`. -// SAFETY: FFI type is valid to be zero-initialized. -let mut acpi: bindings::acpi_device_id =unsafe { core::mem::zeroed () }; +let mut acpi: bindings::acpi_device_id = pin_init::zeroed ();let mut i =0 ;while i < src.len () { acpi.id[i] = src[i];
Signature · 1 line
2.50.1
Benno Lossin
reviewerPATCH v3 11/11 rust: cpufreq: replace `MaybeUninit::zeroed().assume_init()` with `pin_init::zeroed()`
In reply to Benno LossinAll types in bindings implement Zeroable if they can, so use pin_init::zeroed instead of relying on unsafe code.
If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
--- rust/kernel/cpufreq.rs | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)
rust/kernel/cpufreq.rs +1 -3
@@ -27,7 +27,6 @@use core::{ cell::UnsafeCell , marker::PhantomData , - mem::MaybeUninit , ops::{Deref ,DerefMut }, pin::Pin , ptr,
@@ -1013,8 +1012,7 @@ impl<T: Driver> Registration<T> { }else {None }, -// SAFETY: All zeros is a valid value for `bindings::cpufreq_driver`. - ..unsafe {MaybeUninit ::zeroed ().assume_init () } + ..pin_init::zeroed () };const fn copy_name (name: &'staticCStr ) -> [c_char ;CPUFREQ_NAME_LEN ] {
Signature · 1 line
2.50.1
Re: [PATCH v3 11/11] rust: cpufreq: replace `MaybeUninit::zeroed().assume_init()` with `pin_init::zeroed()`
Quoted 10 lines · Benno Lossin
All types in
bindingsimplementZeroableif they can, so usepin_init::zeroedinstead of relying onunsafecode.If this ends up not compiling in the future, something in bindgen or on the C side changed and is most likely incorrect.
Signed-off-by Benno Lossin <[email protected]>--- rust/kernel/cpufreq.rs | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)
Signature · 1 line
viresh
Alice Ryhl
maintainerQuoted 11 lines · Benno Lossin
This came from a discussion at [1]. And I added some more useful parts to the original idea.
I'm not sure on the impact of build times and rust-analyzer. We're adding a derive macro to every struct and union emitted by bindgen. Building using my test-config took 27.7s before and 28.2s after this change, but those are only two runs on my machine with a very reduced config (that enables all Rust code that we have at the moment).
Maybe we have to reevaluate this when more C code is scanned by bindgen.
Imo we should re-export Zeroable from the ffi crate and use it via kernel::ffi in most cases. I don't really want to see pin_init here:
let mut ctx: bindings::lsm_context = pin_init::zeroed();
Alice
Benno Lossin
reviewerQuoted 11 lines · Alice Ryhl
Quoted 10 lines · Benno Lossin
This came from a discussion at [1]. And I added some more useful parts to the original idea.
I'm not sure on the impact of build times and rust-analyzer. We're adding a derive macro to every struct and union emitted by bindgen. Building using my test-config took 27.7s before and 28.2s after this change, but those are only two runs on my machine with a very reduced config (that enables all Rust code that we have at the moment).
Quoted 6 lines
Quoted 1 line
Maybe we have to reevaluate this when more C code is scanned by bindgen.
Imo we should re-export Zeroable from the ffi crate and use it via
kernel::ffiin most cases. I don't really want to seepin_inithere:let mut ctx:
bindings::lsm_context=pin_init::zeroed();
Yeah sounds sensible.
---
Cheers,
Benno
Miguel Ojeda
AppliedQuoted 6 lines · Benno Lossin
rust: add
pin-initas a dependency tobindingsanduapi
rust: deriveZeroablefor all structs & unions generated by bindgen
where possible
rust: cpufreq: replaceMaybeUninit::zeroed().assume_init()with`pin_init::zeroed()`
Applied these #1, #2 and #11 to rust-next -- thanks everyone!
[ Formatted comment. - Miguel ]
The others should be able to go in independently -- I would recommend submitting them separately.
Cheers,
Miguel
Benno Lossin
reviewerQuoted 8 lines · Miguel Ojeda
Quoted 6 lines · Benno Lossin
rust: add
pin-initas a dependency tobindingsanduapi
rust: deriveZeroablefor all structs & unions generated by bindgen
where possible
rust: cpufreq: replaceMaybeUninit::zeroed().assume_init()with`pin_init::zeroed()`
Quoted 6 lines
Applied these #1, #2 and #11 to
rust-next-- thanks everyone![ Formatted comment. - Miguel ]
The others should be able to go in independently -- I would recommend submitting them separately.
I don't have the time to see all the other patches through, but I've opened an issue for users to resend the patches on my behalf:
https://github.com/Rust-for-Linux/linux/issues/1189
---
Cheers,
Benno