Re: [PATCH] dma-buf: Remove unnecessary kmalloc() cast

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Am 02.07.24 um 08:40 schrieb Christoph Hellwig:
On Mon, Jul 01, 2024 at 11:26:34PM -0700, Andrew Morton wrote:
No, I do think the cast is useful:

	struct page *page = dma_fence_chain_alloc();

will presently generate a warning.  We want this.  Your change will
remove that useful warning.


Unrelatedly: there is no earthly reason why this is implemented as a
macro.  A static inline function would be so much better.  Why do we
keep doing this.
Agreed with all of the above.  Adding the dmabuf maintainers.

Thanks for adding me and I have to ask to be added on DMA-buf patches when initially sending them out.

First of all: Yes that cast is intentionally there and yes that is intentionally a define and not an inline function.

See this patch here which changed that:

commit 2c321f3f70bc284510598f712b702ce8d60c4d14
Author: Suren Baghdasaryan <surenb@xxxxxxxxxx>
Date:   Sun Apr 14 19:07:31 2024 -0700

    mm: change inlined allocation helpers to account at the call site

    Main goal of memory allocation profiling patchset is to provide accounting
    that is cheap enough to run in production.  To achieve that we inject
    counters using codetags at the allocation call sites to account every time
    allocation is made.  This injection allows us to perform accounting
    efficiently because injected counters are immediately available as opposed
    to the alternative methods, such as using _RET_IP_, which would require
    counter lookup and appropriate locking that makes accounting much more
    expensive.  This method requires all allocation functions to inject
    separate counters at their call sites so that their callers can be
    individually accounted.  Counter injection is implemented by allocation
    hooks which should wrap all allocation functions.

    Inlined functions which perform allocations but do not use allocation
    hooks are directly charged for the allocations they perform.  In most
    cases these functions are just specialized allocation wrappers used from     multiple places to allocate objects of a specific type.  It would be more     useful to do the accounting at their call sites instead. Instrument these
    helpers to do accounting at the call site.  Simple inlined allocation
    wrappers are converted directly into macros.  More complex allocators or
    allocators with documentation are converted into _noprof versions and
    allocation hooks are added.  This allows memory allocation profiling
    mechanism to charge allocations to the callers of these functions.

Regards,
Christian.



[Index of Archives]     [Linux DRI Users]     [Linux Intel Graphics]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux