mbox series

[v4,00/17] slab: add struct kmem_cache_args

Message ID 20240905-work-kmem_cache_args-v4-0-ed45d5380679@kernel.org (mailing list archive)
Headers show
Series slab: add struct kmem_cache_args | expand

Message

Christian Brauner Sept. 5, 2024, 7:56 a.m. UTC
Hey,

This is v4 which allows NULL to be passed in the struct kmem_cache_args
argument of kmem_cache_create() and substitutes default parameters in
this case.

As discussed last week the various kmem_cache_*() functions should be
replaced by a unified function that is based around a struct, with only
the basic parameters passed separately.

Vlastimil already said that he would like to keep core parameters out
of the struct: name, object size, and flags. I personally don't care
much and would not object to moving everything into the struct but
that's a matter of taste and I yield that decision power to the
maintainer.

In the first version I pointed out that the choice of name is somewhat
forced as kmem_cache_create() is taken and the only way to reuse it
would be to replace all users in one go. Or to do a global
sed/kmem_cache_create()/kmem_cache_create2()/g. And then introduce
kmem_cache_setup(). That doesn't strike me as a viable option.

If we really cared about the *_create() suffix then an alternative might
be to do a sed/kmem_cache_setup()/kmem_cache_create()/g after every user
in the kernel is ported. I honestly don't think that's worth it but I
wanted to at least mention it to highlight the fact that this might lead
to a naming compromise.

However, I came up with an alternative using _Generic() to create a
compatibility layer that will call the correct variant of
kmem_cache_create() depending on whether struct kmem_cache_args is
passed or not. That compatibility layer can stay in place until we
updated all calls to be based on struct kmem_cache_args.

From a cursory grep (and not excluding Documentation mentions) we will
have to replace 44 kmem_cache_create_usercopy() calls and about 463
kmem_cache_create() calls which makes for a bit above 500 calls to port
to kmem_cache_setup(). That'll probably be good work for people getting
into kernel development.

Signed-off-by: Christian Brauner <brauner@kernel.org>
---
Changes in v4:
- Allow NULL to be passed in the struct kmem_cache_args argument of
  kmem_cache_create() and use default parameters in this case.
- Link to v3: https://lore.kernel.org/r/20240904-work-kmem_cache_args-v3-0-05db2179a8c2@kernel.org

Changes in v3:
- Reworded some commit messages.
- Picked up various RvBs.
- Added two patches to make two functions static inline.
- Link to v2: https://lore.kernel.org/r/20240903-work-kmem_cache_args-v2-0-76f97e9a4560@kernel.org

Changes in v2:
- Remove kmem_cache_setup() and add a compatibility layer built around
  _Generic() so that we can keep the kmem_cache_create() name and type
  switch on the third argument to either call __kmem_cache_create() or
  __kmem_cache_create_args().
- Link to v1: https://lore.kernel.org/r/20240902-work-kmem_cache_args-v1-0-27d05bc05128@kernel.org

---
Christian Brauner (17):
      slab: s/__kmem_cache_create/do_kmem_cache_create/g
      slab: add struct kmem_cache_args
      slab: port kmem_cache_create() to struct kmem_cache_args
      slab: port kmem_cache_create_rcu() to struct kmem_cache_args
      slab: port kmem_cache_create_usercopy() to struct kmem_cache_args
      slab: pass struct kmem_cache_args to create_cache()
      slab: pull kmem_cache_open() into do_kmem_cache_create()
      slab: pass struct kmem_cache_args to do_kmem_cache_create()
      slab: remove rcu_freeptr_offset from struct kmem_cache
      slab: port KMEM_CACHE() to struct kmem_cache_args
      slab: port KMEM_CACHE_USERCOPY() to struct kmem_cache_args
      slab: create kmem_cache_create() compatibility layer
      file: port to struct kmem_cache_args
      slab: remove kmem_cache_create_rcu()
      slab: make kmem_cache_create_usercopy() static inline
      slab: make __kmem_cache_create() static inline
      io_uring: port to struct kmem_cache_args

 fs/file_table.c      |  11 ++-
 include/linux/slab.h | 130 +++++++++++++++++++++++++++------
 io_uring/io_uring.c  |  14 ++--
 mm/slab.h            |   6 +-
 mm/slab_common.c     | 197 +++++++++++----------------------------------------
 mm/slub.c            | 162 +++++++++++++++++++++---------------------
 6 files changed, 250 insertions(+), 270 deletions(-)
---
base-commit: 6e016babce7c845ed015da25c7a097fa3482d95a
change-id: 20240902-work-kmem_cache_args-e9760972c7d4

Comments

Vlastimil Babka Sept. 5, 2024, 12:02 p.m. UTC | #1
On 9/5/24 9:56 AM, Christian Brauner wrote:
> Hey,
> 
> This is v4 which allows NULL to be passed in the struct kmem_cache_args
> argument of kmem_cache_create() and substitutes default parameters in
> this case.

Thanks, I've followed your earlier suggestion and put this series to a
branch that merges the vfs.file first:

https://git.kernel.org/pub/scm/linux/kernel/git/vbabka/slab.git/log/?h=slab/for-6.12/kmem_cache_args

and I'm including it to slab/for-next

Meanwhile I think we should look at the kerneldocs for the new
kmem_cache_create() (hopefully kerneldoc can be persuaded to document it
without throwing errors) and the kmem_cache_args, because patches 16 and
14 just removed what we had before, including the details about
freeptr_offset in patch 14.

> As discussed last week the various kmem_cache_*() functions should be
> replaced by a unified function that is based around a struct, with only
> the basic parameters passed separately.
> 
> Vlastimil already said that he would like to keep core parameters out
> of the struct: name, object size, and flags. I personally don't care
> much and would not object to moving everything into the struct but
> that's a matter of taste and I yield that decision power to the
> maintainer.
> 
> In the first version I pointed out that the choice of name is somewhat
> forced as kmem_cache_create() is taken and the only way to reuse it
> would be to replace all users in one go. Or to do a global
> sed/kmem_cache_create()/kmem_cache_create2()/g. And then introduce
> kmem_cache_setup(). That doesn't strike me as a viable option.
> 
> If we really cared about the *_create() suffix then an alternative might
> be to do a sed/kmem_cache_setup()/kmem_cache_create()/g after every user
> in the kernel is ported. I honestly don't think that's worth it but I
> wanted to at least mention it to highlight the fact that this might lead
> to a naming compromise.
> 
> However, I came up with an alternative using _Generic() to create a
> compatibility layer that will call the correct variant of
> kmem_cache_create() depending on whether struct kmem_cache_args is
> passed or not. That compatibility layer can stay in place until we
> updated all calls to be based on struct kmem_cache_args.
> 
> From a cursory grep (and not excluding Documentation mentions) we will
> have to replace 44 kmem_cache_create_usercopy() calls and about 463
> kmem_cache_create() calls which makes for a bit above 500 calls to port
> to kmem_cache_setup(). That'll probably be good work for people getting
> into kernel development.
> 
> Signed-off-by: Christian Brauner <brauner@kernel.org>
> ---
> Changes in v4:
> - Allow NULL to be passed in the struct kmem_cache_args argument of
>   kmem_cache_create() and use default parameters in this case.
> - Link to v3: https://lore.kernel.org/r/20240904-work-kmem_cache_args-v3-0-05db2179a8c2@kernel.org
> 
> Changes in v3:
> - Reworded some commit messages.
> - Picked up various RvBs.
> - Added two patches to make two functions static inline.
> - Link to v2: https://lore.kernel.org/r/20240903-work-kmem_cache_args-v2-0-76f97e9a4560@kernel.org
> 
> Changes in v2:
> - Remove kmem_cache_setup() and add a compatibility layer built around
>   _Generic() so that we can keep the kmem_cache_create() name and type
>   switch on the third argument to either call __kmem_cache_create() or
>   __kmem_cache_create_args().
> - Link to v1: https://lore.kernel.org/r/20240902-work-kmem_cache_args-v1-0-27d05bc05128@kernel.org
> 
> ---
> Christian Brauner (17):
>       slab: s/__kmem_cache_create/do_kmem_cache_create/g
>       slab: add struct kmem_cache_args
>       slab: port kmem_cache_create() to struct kmem_cache_args
>       slab: port kmem_cache_create_rcu() to struct kmem_cache_args
>       slab: port kmem_cache_create_usercopy() to struct kmem_cache_args
>       slab: pass struct kmem_cache_args to create_cache()
>       slab: pull kmem_cache_open() into do_kmem_cache_create()
>       slab: pass struct kmem_cache_args to do_kmem_cache_create()
>       slab: remove rcu_freeptr_offset from struct kmem_cache
>       slab: port KMEM_CACHE() to struct kmem_cache_args
>       slab: port KMEM_CACHE_USERCOPY() to struct kmem_cache_args
>       slab: create kmem_cache_create() compatibility layer
>       file: port to struct kmem_cache_args
>       slab: remove kmem_cache_create_rcu()
>       slab: make kmem_cache_create_usercopy() static inline
>       slab: make __kmem_cache_create() static inline
>       io_uring: port to struct kmem_cache_args
> 
>  fs/file_table.c      |  11 ++-
>  include/linux/slab.h | 130 +++++++++++++++++++++++++++------
>  io_uring/io_uring.c  |  14 ++--
>  mm/slab.h            |   6 +-
>  mm/slab_common.c     | 197 +++++++++++----------------------------------------
>  mm/slub.c            | 162 +++++++++++++++++++++---------------------
>  6 files changed, 250 insertions(+), 270 deletions(-)
> ---
> base-commit: 6e016babce7c845ed015da25c7a097fa3482d95a
> change-id: 20240902-work-kmem_cache_args-e9760972c7d4
>