Re: mm, slab/slub: Ensure kmem_cache_alloc_bulk() is available early

[Thomas Gleixner]

On Wed, Feb 08 2023 at 10:15, Vlastimil Babka wrote:

Cc+ Willy

> On 2/7/23 19:20, Thomas Gleixner wrote:
>> On Tue, Feb 07 2023 at 15:47, Vlastimil Babka wrote:
>>> From 340d7c7b99f3e67780f6dec480ed1d27e6f325eb Mon Sep 17 00:00:00 2001
>>> From: Vlastimil Babka <vbabka@xxxxxxx>
>>> Date: Tue, 7 Feb 2023 15:34:53 +0100
>>> Subject: [PATCH] mm, slab/slub: remove notes that bulk alloc/free needs
>>>  interrupts enabled
>>>
>>> The slab functions kmem_cache_[alloc|free]_bulk() have been documented
>>> as requiring interrupts to be enabled, since their addition in 2015.
>>> It's unclear whether that was a fundamental restriction, or an attempt
>>> to save some cpu cycles by not having to save and restore the irq
>>> flags.
>> 
>> I don't think so. The restriction is rather meant to avoid huge
>> allocations in atomic context which causes latencies and also might
>> deplete the atomic reserves.
>
> Fair enough.
>
>> So I rather avoid that and enforce !ATOMIC mode despite the
>> local_irq_save/restore() change which is really only to accomodate with
>> early boot.
>
> We could add some warning then? People might use the bulk alloc unknowingly
> again e.g. via maple tree. GFP_KERNEL would warn through the existing
> warning, but e.g. GFP_ATOMIC currently not.

Correct.

> Some maple tree users could use its preallocation instead outside of the
> atomic context, when possible.

Right.

The issue is that there might be maple_tree users which depend on
GFP_ATOMIC, but call in from interrupt enabled context, which is
legitimate today.

Willy might have some insight on that.

Thanks,

        tglx