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: <202411041330.7A0F716E5@keescook>
Date: Mon, 4 Nov 2024 13:32:25 -0800
From: Kees Cook <kees@...nel.org>
To: Vlastimil Babka <vbabka@...e.cz>
Cc: Koichiro Den <koichiro.den@...il.com>, cl@...ux.com, penberg@...nel.org,
	rientjes@...gle.com, iamjoonsoo.kim@....com,
	akpm@...ux-foundation.org, roman.gushchin@...ux.dev,
	42.hyeyoo@...il.com, linux-mm@...ck.org,
	linux-kernel@...r.kernel.org, stable@...r.kernel.org
Subject: Re: [PATCH] mm/slab: fix warning caused by duplicate kmem_cache
 creation in kmem_buckets_create

On Mon, Nov 04, 2024 at 04:40:12PM +0100, Vlastimil Babka wrote:
> On 11/4/24 16:08, Koichiro Den wrote:
> > Commit b035f5a6d852 ("mm: slab: reduce the kmalloc() minimum alignment
> > if DMA bouncing possible") reduced ARCH_KMALLOC_MINALIGN to 8 on arm64.
> > However, with KASAN_HW_TAGS enabled, arch_slab_minalign() becomes 16.
> > This causes kmalloc_caches[*][8] to be aliased to kmalloc_caches[*][16],
> > resulting in kmem_buckets_create() attempting to create a kmem_cache for
> > size 16 twice. This duplication triggers warnings on boot:
> > 
> > [    2.325108] ------------[ cut here ]------------
> > [    2.325135] kmem_cache of name 'memdup_user-16' already exists
> > [    2.325783] WARNING: CPU: 0 PID: 1 at mm/slab_common.c:107 __kmem_cache_create_args+0xb8/0x3b0
> > [    2.327957] Modules linked in:
> > [    2.328550] CPU: 0 UID: 0 PID: 1 Comm: swapper/0 Not tainted 6.12.0-rc5mm-unstable-arm64+ #12
> > [    2.328683] Hardware name: QEMU QEMU Virtual Machine, BIOS 2024.02-2 03/11/2024
> > [    2.328790] pstate: 61000009 (nZCv daif -PAN -UAO -TCO +DIT -SSBS BTYPE=--)
> > [    2.328911] pc : __kmem_cache_create_args+0xb8/0x3b0
> > [    2.328930] lr : __kmem_cache_create_args+0xb8/0x3b0
> > [    2.328942] sp : ffff800083d6fc50
> > [    2.328961] x29: ffff800083d6fc50 x28: f2ff0000c1674410 x27: ffff8000820b0598
> > [    2.329061] x26: 000000007fffffff x25: 0000000000000010 x24: 0000000000002000
> > [    2.329101] x23: ffff800083d6fce8 x22: ffff8000832222e8 x21: ffff800083222388
> > [    2.329118] x20: f2ff0000c1674410 x19: f5ff0000c16364c0 x18: ffff800083d80030
> > [    2.329135] x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
> > [    2.329152] x14: 0000000000000000 x13: 0a73747369786520 x12: 79646165726c6120
> > [    2.329169] x11: 656820747563205b x10: 2d2d2d2d2d2d2d2d x9 : 0000000000000000
> > [    2.329194] x8 : 0000000000000000 x7 : 0000000000000000 x6 : 0000000000000000
> > [    2.329210] x5 : 0000000000000000 x4 : 0000000000000000 x3 : 0000000000000000
> > [    2.329226] x2 : 0000000000000000 x1 : 0000000000000000 x0 : 0000000000000000
> > [    2.329291] Call trace:
> > [    2.329407]  __kmem_cache_create_args+0xb8/0x3b0
> > [    2.329499]  kmem_buckets_create+0xfc/0x320
> > [    2.329526]  init_user_buckets+0x34/0x78
> > [    2.329540]  do_one_initcall+0x64/0x3c8
> > [    2.329550]  kernel_init_freeable+0x26c/0x578
> > [    2.329562]  kernel_init+0x3c/0x258
> > [    2.329574]  ret_from_fork+0x10/0x20
> > [    2.329698] ---[ end trace 0000000000000000 ]---
> > 
> > [    2.403704] ------------[ cut here ]------------
> > [    2.404716] kmem_cache of name 'msg_msg-16' already exists
> > [    2.404801] WARNING: CPU: 2 PID: 1 at mm/slab_common.c:107 __kmem_cache_create_args+0xb8/0x3b0
> > [    2.404842] Modules linked in:
> > [    2.404971] CPU: 2 UID: 0 PID: 1 Comm: swapper/0 Tainted: G        W          6.12.0-rc5mm-unstable-arm64+ #12
> > [    2.405026] Tainted: [W]=WARN
> > [    2.405043] Hardware name: QEMU QEMU Virtual Machine, BIOS 2024.02-2 03/11/2024
> > [    2.405057] pstate: 60400009 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
> > [    2.405079] pc : __kmem_cache_create_args+0xb8/0x3b0
> > [    2.405100] lr : __kmem_cache_create_args+0xb8/0x3b0
> > [    2.405111] sp : ffff800083d6fc50
> > [    2.405115] x29: ffff800083d6fc50 x28: fbff0000c1674410 x27: ffff8000820b0598
> > [    2.405135] x26: 000000000000ffd0 x25: 0000000000000010 x24: 0000000000006000
> > [    2.405153] x23: ffff800083d6fce8 x22: ffff8000832222e8 x21: ffff800083222388
> > [    2.405169] x20: fbff0000c1674410 x19: fdff0000c163d6c0 x18: ffff800083d80030
> > [    2.405185] x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
> > [    2.405201] x14: 0000000000000000 x13: 0a73747369786520 x12: 79646165726c6120
> > [    2.405217] x11: 656820747563205b x10: 2d2d2d2d2d2d2d2d x9 : 0000000000000000
> > [    2.405233] x8 : 0000000000000000 x7 : 0000000000000000 x6 : 0000000000000000
> > [    2.405248] x5 : 0000000000000000 x4 : 0000000000000000 x3 : 0000000000000000
> > [    2.405271] x2 : 0000000000000000 x1 : 0000000000000000 x0 : 0000000000000000
> > [    2.405287] Call trace:
> > [    2.405293]  __kmem_cache_create_args+0xb8/0x3b0
> > [    2.405305]  kmem_buckets_create+0xfc/0x320
> > [    2.405315]  init_msg_buckets+0x34/0x78
> > [    2.405326]  do_one_initcall+0x64/0x3c8
> > [    2.405337]  kernel_init_freeable+0x26c/0x578
> > [    2.405348]  kernel_init+0x3c/0x258
> > [    2.405360]  ret_from_fork+0x10/0x20
> > [    2.405370] ---[ end trace 0000000000000000 ]---
> > 
> > To address this, alias kmem_cache for sizes smaller than min alignment
> > to the aligned sized kmem_cache, as done with the default system kmalloc
> > bucket.
> > 
> > Cc: <stable@...r.kernel.org> # 6.11.x
> > Fixes: b32801d1255b ("mm/slab: Introduce kmem_buckets_create() and family")
> > Signed-off-by: Koichiro Den <koichiro.den@...il.com>
> > ---
> 
> Thanks for catching this.
> Wonder if we could make this a lot simpler in kmem_buckets_create() by
> starting with the current:
> 
> size = kmalloc_caches[KMALLOC_NORMAL][idx]->object_size;
> 
> and adding:
> 
> aligned_idx = __kmalloc_index(size, false);
> 
> now the rest of the loop iteration would work with aligned_idx and if it
> differs from idx, we assign the cache pointer also to idx, etc.
> 
> This should avoid duplicating the alignment calculation as we just extract
> from kmalloc_caches[] the result of what new_kmalloc_cache() already did?

Yeah, I was thinking similar stuff -- the aligned stuff is the alias for
the actual stuff.

I like the bitmask for tracking the aliases. :)

-- 
Kees Cook

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ