linux-kernel - Re: [PATCH v5 00/14] SLUB percpu sheaves

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] [day] [month] [year] [list]

Message-ID: <CAA9mObALY3L=G3Ti52pSnWYGHspDa01Xr6NeiWHwTGOX76K7nw@mail.gmail.com>
Date: Sat, 16 Aug 2025 21:28:12 -0700
From: Sudarsan Mahendran <sudarsanm@...gle.com>
To: Vlastimil Babka <vbabka@...e.cz>
Cc: Harry Yoo <harry.yoo@...cle.com>, Liam.Howlett@...cle.com, cl@...two.org, 
	howlett@...il.com, linux-kernel@...r.kernel.org, linux-mm@...ck.org, 
	maple-tree@...ts.infradead.org, rcu@...r.kernel.org, rientjes@...gle.com, 
	roman.gushchin@...ux.dev, surenb@...gle.com, urezki@...il.com, 
	Greg Thelen <gthelen@...gle.com>
Subject: Re: [PATCH v5 00/14] SLUB percpu sheaves

On Sat, Aug 16, 2025 at 11:31 AM Vlastimil Babka <vbabka@...e.cz> wrote:
>
> On 8/16/25 8:31 PM, Vlastimil Babka wrote:
> >>
> >> I assume somehow the free_to_pcs_bulk() fallback case is taken, thus
> >> calling __kmem_cache_free_bulk(), which calls free_to_pcs_bulk() ad nauseam.
> > Could it be a rebase gone wrong? Mine to 6.17-rc1 is here (but untested)

Yes Vlastimil,

You're right. It is a rebase gone wrong. Thanks for catching this.

I ported this patch series on top of v6.17-rc1 using b4 cmd

b4 am -o - 20250723-slub-percpu-caches-v5-0-b792cd830f5d@...e.cz | git
am --reject

For some reason b4 merging yielded me this:

git show 893ee67b5c75e7411e4e3c6ddaa8d0765985423e
slab: add opt-in caching layer of percpu sheaves

@@ -5252,6 +6133,15 @@ static void __kmem_cache_free_bulk(struct
kmem_cache *s, size_t size, void **p)
        if (!size)
                return;

+       /*
+        * freeing to sheaves is so incompatible with the detached freelist so
+        * once we go that way, we have to do everything differently
+        */
+       if (s && s->cpu_sheaves) {
+               free_to_pcs_bulk(s, size, p);
+               return;
+       }
+
        do {


Whereas the original patch [1] had this instead:

@@ -5033,6 +5801,15 @@ void kmem_cache_free_bulk(struct kmem_cache *s,
size_t size, void **p)
  if (!size)
  return;

+ /*
+ * freeing to sheaves is so incompatible with the detached freelist so
+ * once we go that way, we have to do everything differently
+ */
+ if (s && s->cpu_sheaves) {
+ free_to_pcs_bulk(s, size, p);
+ return;
+ }
+

I have no idea why b4 got confused between kmem_cache_free_bulk() and
__kmem_cache_free_bulk().

After I fixed this issue, I'm able to boot the kernel successfully.

[1] https://lore.kernel.org/all/20250214-slub-percpu-caches-v2-1-88592ee0966a@suse.cz/

> >
> > https://git.kernel.org/pub/scm/linux/kernel/git/vbabka/linux.git/
>
> This branch specifically
> https://git.kernel.org/pub/scm/linux/kernel/git/vbabka/linux.git/log/?h=b4/slub-percpu-sheaves