Message-ID: <58758ff1-e7d7-4c47-a513-f65bdf777a6a@gmail.com>
Date: Fri, 12 Sep 2025 12:13:30 +0200
From: Andrei Lalaev <andrey.lalaev@...il.com>
To: Russ Weight <russ.weight@...ux.dev>,
 "Lalaev, Andrei" <andrei.lalaev@...on-paar.com>
Cc: "rafael@...nel.org" <rafael@...nel.org>,
 "gregkh@...uxfoundation.org" <gregkh@...uxfoundation.org>,
 "dakr@...nel.org" <dakr@...nel.org>, "mcgrof@...nel.org"
 <mcgrof@...nel.org>,
 "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
 "linux-leds@...r.kernel.org" <linux-leds@...r.kernel.org>
Subject: Re: use-after-free in firmware_fallback_sysfs

On 12.09.2025 00:35, Russ Weight wrote:
> 
> On Thu, Sep 04, 2025 at 04:58:45AM +0000, Lalaev, Andrei wrote:
>> Hi,
>>
>> We performed a long-term stability test on the LP5562 driver by repeatedly
>> downloading different LED firmwares using sysfs interface.
>> And after some time, we see the following "use-after-free" trace on kernel 6.12.11:
>>
>> [  274.759115] BUG: KASAN: slab-use-after-free in firmware_fallback_sysfs (include/linux/list.h:153 include/linux/list.h:169 drivers/base/firmware_loader/fallback.c:98 drivers/base/firmware_loader/fallback.c:162 drivers/base/firmware_loader/fallback.c:238)
>> [  274.766743] Write of size 4 at addr b31d0850 by task kworker/2:2/137
>> [  274.774680] CPU: 2 UID: 0 PID: 137 Comm: kworker/2:2 Tainted: G         C         6.12.11 #1
>> [  274.783208] Tainted: [C]=CRAP
>> [  274.786215] Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
>> [  274.792775] Workqueue: events request_firmware_work_func
>> [  274.798151] Call trace:
>> [  274.800720] dump_backtrace from show_stack (arch/arm/kernel/traps.c:259)
>> [  274.805827]  r7:600f0093 r6:600f0093 r5:b19659b4 r4:00000000
>> [  274.811510] show_stack from dump_stack_lvl (lib/dump_stack.c:122)
>> [  274.816613] dump_stack_lvl from print_report (mm/kasan/report.c:378 mm/kasan/report.c:488)
>> [  274.822073]  r7:000031d0 r6:df7dba00 r5:f13bbb20 r4:b31d0850
>> [  274.827755] print_report from kasan_report (mm/kasan/report.c:603)
>> [  274.832885]  r10:b2d56c20 r9:cdc89808 r8:00003a98 r7:b0f653ac r6:00000004 r5:00000001
>> [  274.840744]  r4:b31d0850
>> [  274.843298] kasan_report from __asan_report_store4_noabort (mm/kasan/report_generic.c:385)
>> [  274.849804]  r7:b9f72c00 r6:cdc89800 r5:b5d31e00 r4:b5d31e10
>> [  274.855488] __asan_report_store4_noabort from firmware_fallback_sysfs (include/linux/list.h:153 include/linux/list.h:169 drivers/base/firmware_loader/fallback.c:98 drivers/base/firmware_loader/fallback.c:162 drivers/base/firmware_loader/fallback.c:238)
>> [  274.863105] firmware_fallback_sysfs from _request_firmware (drivers/base/firmware_loader/main.c:941)
>> [  274.869862]  r10:f13bbda0 r9:b9f72c80 r8:b19c73d8 r7:00000004 r6:00000000 r5:b5d31e00
>> [  274.877718]  r4:f13bbd00
>> [  274.880273] _request_firmware from request_firmware_work_func (drivers/base/firmware_loader/main.c:1196)
>> [  274.887226]  r10:b9f72e98 r9:b9f72e94 r8:f13bbde0 r7:173ee5d3 r6:00000000 r5:ad2777b0
>> [  274.895097]  r4:b9f72e80
>> [  274.897661] request_firmware_work_func from process_one_work (kernel/workqueue.c:3235)
>> [  274.904613]  r10:00000000 r9:b7d7e1e0 r8:b7d7dd80 r7:d76b53c0 r6:b2013c00 r5:b9f72e80
>> [  274.912469]  r4:b2371a00
>> [  274.915024] process_one_work from worker_thread (kernel/workqueue.c:3304 (discriminator 2) kernel/workqueue.c:3391 (discriminator 2))
>> [  274.920823]  r10:b2371a2c r9:b2371a00 r8:b9f72e84 r7:b2371a2c r6:d76b53dc r5:a546e345
>> [  274.928678]  r4:00000007
>> [  274.931234] worker_thread from kthread (kernel/kthread.c:389)
>> [  274.936179]  r10:f0933ca0 r9:b2371a00 r8:b7f6b800 r7:b019ad18 r6:b2354880 r5:b7d7dd80
>> [  274.944035]  r4:b2354d00
>> [  274.946592] kthread from ret_from_fork (arch/arm/kernel/entry-common.S:138)
>> [  274.951350] Exception stack(0xf13bbfb0 to 0xf13bbff8)
>> [  274.956435] bfa0:                                     00000000 00000000 00000000 00000000
>> [  274.964647] bfc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
>> [  274.972858] bfe0: 00000000 00000000 00000000 00000000 00000013 00000000
>> [  274.979508]  r10:00000000 r9:00000000 r8:00000000 r7:00000000 r6:00000000 r5:b01ba280
>> [  274.987365]  r4:b2354d00 r3:00000000
>>
>> [  274.992472] Allocated by task 8 on cpu 0 at 273.209883s:
>> [  274.997814] kasan_save_track (mm/kasan/common.c:48 mm/kasan/common.c:68)
>> [  274.997858] kasan_save_alloc_info (mm/kasan/generic.c:566)
>> [  274.997892] __kasan_kmalloc (mm/kasan/common.c:398)
>> [  274.997926] __kmalloc_cache_noprof (mm/slub.c:4319)
>> [  274.997957] alloc_lookup_fw_priv (drivers/base/firmware_loader/main.c:131 drivers/base/firmware_loader/main.c:190)
>> [  274.997993] _request_firmware (drivers/base/firmware_loader/main.c:769 drivers/base/firmware_loader/main.c:899)
>> [  274.998023] request_firmware_work_func (drivers/base/firmware_loader/main.c:1196)
>> [  274.998055] process_one_work (kernel/workqueue.c:3235)
>> [  274.998084] worker_thread (kernel/workqueue.c:3304 (discriminator 2) kernel/workqueue.c:3391 (discriminator 2))
>> [  274.998112] kthread (kernel/kthread.c:389)
>> [  274.998149] ret_from_fork (arch/arm/kernel/entry-common.S:138)
>>
>> [  274.999686] Freed by task 8 on cpu 0 at 273.230581s:
>> [  275.004679] kasan_save_track (mm/kasan/common.c:48 mm/kasan/common.c:68)
>> [  275.004718] kasan_save_free_info (mm/kasan/generic.c:582 (discriminator 1))
>> [  275.004747] __kasan_slab_free (mm/kasan/common.c:271)
>> [  275.004782] kfree (mm/slub.c:4601 (discriminator 3) mm/slub.c:4749 (discriminator 3))
>> [  275.004810] free_fw_priv (drivers/base/firmware_loader/main.c:231)
>> [  275.004840] release_firmware (drivers/base/firmware_loader/main.c:604 drivers/base/firmware_loader/main.c:1170 drivers/base/firmware_loader/main.c:1166)
>> [  275.004871] lp55xx_firmware_loaded (drivers/leds/leds-lp55xx-common.c:549) leds_lp55xx_common
>> [  275.005074] request_firmware_work_func (drivers/base/firmware_loader/main.c:1197)
>> [  275.005112] process_one_work (kernel/workqueue.c:3235)
>> [  275.005145] worker_thread (kernel/workqueue.c:3304 (discriminator 2) kernel/workqueue.c:3391 (discriminator 2))
>> [  275.005174] kthread (kernel/kthread.c:389)
>> [  275.005212] ret_from_fork (arch/arm/kernel/entry-common.S:138)
>> [  275.006754] The buggy address belongs to the object at b31d0800 which belongs to the cache kmalloc-128 of size 128
>> [  275.018610] The buggy address is located 80 bytes inside of freed 128-byte region [b31d0800, b31d0880)
>> [  275.030938] The buggy address belongs to the physical page:
>> [  275.036533] page: refcount:1 mapcount:0 mapping:00000000 index:0x0 pfn:0x131d0
>> [  275.036568] flags: 0x0(zone=0)
>> [  275.036598] page_type: f5(slab)
>> [  275.036634] raw: 00000000 b2001400 df9544e0 00000002 00000000 00100010 f5000000 00000001
>> [  275.036659] page dumped because: kasan: bad access detected
>>
>> [  275.038189] Memory state around the buggy address:
>> [  275.043008]  b31d0700: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
>> [  275.049563]  b31d0780: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
>> [  275.056120] >b31d0800: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
>> [  275.062669]                                          ^
>> [  275.067831]  b31d0880: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
>> [  275.074386]  b31d0900: 00 00 00 00 00 00 00 00 00 00 fc fc fc fc fc fc
>> [  275.080938] ==================================================================
>>
>> I couldn't find any related patches to fix this issue in upstream, so
>> I have made the following patch to address this issue, but I have some concerns about it.
>>
>> Could someone please provide feedback or insights on this?
>>
>> ----
>>  drivers/base/firmware_loader/main.c | 1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/base/firmware_loader/main.c b/drivers/base/firmware_loader/main.c
>> index c6664a787..1a49022f4 100644
>> --- a/drivers/base/firmware_loader/main.c
>> +++ b/drivers/base/firmware_loader/main.c
>> @@ -211,6 +211,7 @@ static void __free_fw_priv(struct kref *ref)
>>  		 (unsigned int)fw_priv->size);
>>  
>>  	list_del(&fw_priv->list);
>> +	list_del(&fw_priv->pending_list);
> 
> References to &fw_priv->pending_list in this context would need to
> be bounded with "#ifdef CONFIG_FW_LOADER_USER_HELPER".
> 
> Have you repeated your long-term stability test with this change?
> 

Hi Russ,

Yes, sure. I had the test running for 2 days, and no "use-after-free" appeared.

-- 
Best regards,
Andrei Lalaev

Powered by blists - more mailing lists