linux-kernel - Re: [PATCH v3 09/11] mm/slub: Fold slab_update

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <18c33bf0-0c7e-7584-5149-33cf77b50b8a@suse.cz>
Date:   Wed, 24 May 2023 13:58:24 +0200
From:   Vlastimil Babka <vbabka@...e.cz>
To:     Peter Zijlstra <peterz@...radead.org>,
        torvalds@...ux-foundation.org
Cc:     corbet@....net, will@...nel.org, boqun.feng@...il.com,
        mark.rutland@....com, catalin.marinas@....com, dennis@...nel.org,
        tj@...nel.org, cl@...ux.com, hca@...ux.ibm.com, gor@...ux.ibm.com,
        agordeev@...ux.ibm.com, borntraeger@...ux.ibm.com,
        svens@...ux.ibm.com, tglx@...utronix.de, mingo@...hat.com,
        bp@...en8.de, dave.hansen@...ux.intel.com, x86@...nel.org,
        hpa@...or.com, joro@...tes.org, suravee.suthikulpanit@....com,
        robin.murphy@....com, dwmw2@...radead.org,
        baolu.lu@...ux.intel.com, Arnd Bergmann <arnd@...db.de>,
        Herbert Xu <herbert@...dor.apana.org.au>, davem@...emloft.net,
        penberg@...nel.org, rientjes@...gle.com, iamjoonsoo.kim@....com,
        Andrew Morton <akpm@...ux-foundation.org>,
        roman.gushchin@...ux.dev, 42.hyeyoo@...il.com,
        linux-doc@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-mm@...ck.org, linux-s390@...r.kernel.org,
        iommu@...ts.linux.dev, linux-arch@...r.kernel.org,
        linux-crypto@...r.kernel.org
Subject: Re: [PATCH v3 09/11] mm/slub: Fold slab_update_freelist()

On 5/15/23 09:57, Peter Zijlstra wrote:
> The two functions slab_update_freelist() and __slab_update_freelist()
> are nearly identical, fold and add a boolean argument and rely on
> constant propagation.
> 
> Signed-off-by: Peter Zijlstra (Intel) <peterz@...radead.org>

Something like that has been tried before and the result:
https://lore.kernel.org/all/CAHk-=wiJLqL2cUhJbvpyPQpkbVOu1rVSzgO2=S2jC55hneLtfQ@mail.gmail.com/

Your parameter is not called 'locked' but 'irq_save' which is better, but
that's just one detail.

After your refactoring in 08/11 which puts most of the code into
__update_freelist_fast() and _slow() I'd say the result is not so bad already.

BTW I have some suspicion that some SLUB code is based on assumptions that
are no longer true these days. IIRC I've seen some microbenchmark results a
while ago that showed that disabling/enabling irqs is surprisingly (to me)
very cheap today, so maybe it's not so useful to keep doing the
this_cpu_cmpxchg128 for the struct kmem_cache_cpu operations (less so for
struct slab cmpxchg128 where actually different cpus may be involved). But
it needs a closer look.

> ---
>  mm/slub.c |   80 +++++++++++++++++++++-----------------------------------------
>  1 file changed, 28 insertions(+), 52 deletions(-)
> 
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -559,53 +559,29 @@ __update_freelist_slow(struct slab *slab
>   * allocation/ free operation in hardirq context. Therefore nothing can
>   * interrupt the operation.
>   */
> -static inline bool __slab_update_freelist(struct kmem_cache *s, struct slab *slab,
> -		void *freelist_old, unsigned long counters_old,
> -		void *freelist_new, unsigned long counters_new,
> -		const char *n)
> +static __always_inline
> +bool slab_update_freelist(struct kmem_cache *s, struct slab *slab,
> +			  void *freelist_old, unsigned long counters_old,
> +			  void *freelist_new, unsigned long counters_new,
> +			  bool irq_save, const char *n)
>  {
>  	bool ret;
>  
> -	if (USE_LOCKLESS_FAST_PATH())
> +	if (!irq_save && USE_LOCKLESS_FAST_PATH())
>  		lockdep_assert_irqs_disabled();
>  
>  	if (s->flags & __CMPXCHG_DOUBLE) {
>  		ret = __update_freelist_fast(slab, freelist_old, counters_old,
>  				            freelist_new, counters_new);
>  	} else {
> -		ret = __update_freelist_slow(slab, freelist_old, counters_old,
> -				            freelist_new, counters_new);
> -	}
> -	if (likely(ret))
> -		return true;
> -
> -	cpu_relax();
> -	stat(s, CMPXCHG_DOUBLE_FAIL);
> -
> -#ifdef SLUB_DEBUG_CMPXCHG
> -	pr_info("%s %s: cmpxchg double redo ", n, s->name);
> -#endif
> -
> -	return false;
> -}
> -
> -static inline bool slab_update_freelist(struct kmem_cache *s, struct slab *slab,
> -		void *freelist_old, unsigned long counters_old,
> -		void *freelist_new, unsigned long counters_new,
> -		const char *n)
> -{
> -	bool ret;
> -
> -	if (s->flags & __CMPXCHG_DOUBLE) {
> -		ret = __update_freelist_fast(slab, freelist_old, counters_old,
> -				            freelist_new, counters_new);
> -	} else {
>  		unsigned long flags;
>  
> -		local_irq_save(flags);
> +		if (irq_save)
> +			local_irq_save(flags);
>  		ret = __update_freelist_slow(slab, freelist_old, counters_old,
>  				            freelist_new, counters_new);
> -		local_irq_restore(flags);
> +		if (irq_save)
> +			local_irq_restore(flags);
>  	}
>  	if (likely(ret))
>  		return true;
> @@ -2250,10 +2226,10 @@ static inline void *acquire_slab(struct
>  	VM_BUG_ON(new.frozen);
>  	new.frozen = 1;
>  
> -	if (!__slab_update_freelist(s, slab,
> -			freelist, counters,
> -			new.freelist, new.counters,
> -			"acquire_slab"))
> +	if (!slab_update_freelist(s, slab,
> +				  freelist, counters,
> +				  new.freelist, new.counters,
> +				  false, "acquire_slab"))
>  		return NULL;
>  
>  	remove_partial(n, slab);
> @@ -2577,9 +2553,9 @@ static void deactivate_slab(struct kmem_
>  
>  
>  	if (!slab_update_freelist(s, slab,
> -				old.freelist, old.counters,
> -				new.freelist, new.counters,
> -				"unfreezing slab")) {
> +				  old.freelist, old.counters,
> +				  new.freelist, new.counters,
> +				  true, "unfreezing slab")) {
>  		if (mode == M_PARTIAL)
>  			spin_unlock_irqrestore(&n->list_lock, flags);
>  		goto redo;
> @@ -2633,10 +2609,10 @@ static void __unfreeze_partials(struct k
>  
>  			new.frozen = 0;
>  
> -		} while (!__slab_update_freelist(s, slab,
> -				old.freelist, old.counters,
> -				new.freelist, new.counters,
> -				"unfreezing slab"));
> +		} while (!slab_update_freelist(s, slab,
> +					       old.freelist, old.counters,
> +					       new.freelist, new.counters,
> +					       false, "unfreezing slab"));
>  
>  		if (unlikely(!new.inuse && n->nr_partial >= s->min_partial)) {
>  			slab->next = slab_to_discard;
> @@ -3072,10 +3048,10 @@ static inline void *get_freelist(struct
>  		new.inuse = slab->objects;
>  		new.frozen = freelist != NULL;
>  
> -	} while (!__slab_update_freelist(s, slab,
> -		freelist, counters,
> -		NULL, new.counters,
> -		"get_freelist"));
> +	} while (!slab_update_freelist(s, slab,
> +				       freelist, counters,
> +				       NULL, new.counters,
> +				       false, "get_freelist"));
>  
>  	return freelist;
>  }
> @@ -3666,9 +3642,9 @@ static void __slab_free(struct kmem_cach
>  		}
>  
>  	} while (!slab_update_freelist(s, slab,
> -		prior, counters,
> -		head, new.counters,
> -		"__slab_free"));
> +				       prior, counters,
> +				       head, new.counters,
> +				       true, "__slab_free"));
>  
>  	if (likely(!n)) {
>  
> 
>