Re: [PATCH] perf/x86/intel: Fix KASAN global-out-of-bounds warning

["Mi, Dapeng" <dapeng1.mi@linux.intel.com>]

On 10/24/2025 1:45 AM, Chen, Zide wrote:
>
> On 9/3/2025 6:24 PM, Dapeng Mi wrote:
>> When running "perf mem record" command on CWF, the below KASAN
>> global-out-of-bounds warning is seen.
>>
>>   196.273657] ==================================================================
>> [  196.273662] BUG: KASAN: global-out-of-bounds in cmt_latency_data+0x176/0x1b0
>> [  196.273669] Read of size 4 at addr ffffffffb721d000 by task dtlb/9850
>>
>> [  196.273676] CPU: 126 UID: 0 PID: 9850 Comm: dtlb Kdump: loaded Not tainted 6.17.0-rc3-2025-08-29-intel-next-34160-g316938187eb0 #1 PREEMPT(none)
>> [  196.273680] Hardware name: Intel Corporation AvenueCity/AvenueCity, BIOS BHSDCRB1.IPC.3544.P83.2507110208 07/11/2025
>> [  196.273682] Call Trace:
>> [  196.273683]  <NMI>
>> [  196.273684]  dump_stack_lvl+0x55/0x70
>> [  196.273689]  print_address_description.constprop.0+0x2c/0x3d0
>> [  196.273694]  ? cmt_latency_data+0x176/0x1b0
>> [  196.273696]  print_report+0xb4/0x270
>> [  196.273699]  ? kasan_addr_to_slab+0xd/0xa0
>> [  196.273702]  kasan_report+0xb8/0xf0
>> [  196.273705]  ? cmt_latency_data+0x176/0x1b0
>> [  196.273707]  cmt_latency_data+0x176/0x1b0
>> [  196.273710]  setup_arch_pebs_sample_data+0xf49/0x2560
>> [  196.273713]  intel_pmu_drain_arch_pebs+0x577/0xb00
>> [  196.273716]  ? __pfx_intel_pmu_drain_arch_pebs+0x10/0x10
>> [  196.273719]  ? perf_output_begin+0x3e4/0xa10
>> [  196.273724]  ? intel_pmu_drain_bts_buffer+0xc2/0x6a0
>> [  196.273727]  ? __pfx_intel_pmu_drain_bts_buffer+0x10/0x10
>> [  196.273730]  handle_pmi_common+0x6c4/0xc80
>> [  196.273734]  ? __pfx_handle_pmi_common+0x10/0x10
>> [  196.273738]  ? intel_bts_interrupt+0xd3/0x4d0
>> [  196.273740]  ? __pfx_intel_bts_interrupt+0x10/0x10
>> [  196.273742]  ? intel_pmu_lbr_enable_all+0x25/0x150
>> [  196.273745]  intel_pmu_handle_irq+0x388/0x700
>> [  196.273748]  perf_event_nmi_handler+0xff/0x150
>> [  196.273751]  nmi_handle.part.0+0xa8/0x2d0
>> [  196.273755]  ? perf_output_begin+0x3e9/0xa10
>> [  196.273757]  default_do_nmi+0x79/0x1a0
>> [  196.273760]  fred_exc_nmi+0x40/0x90
>> [  196.273762]  asm_fred_entrypoint_kernel+0x45/0x60
>> [  196.273765] RIP: 0010:perf_output_begin+0x3e9/0xa10
>> [  196.273768] Code: 54 24 1c 85 d2 0f 85 19 03 00 00 48 8b 44 24 18 48 c1 e8 03 42 0f b6 04 28 84 c0 74 08 3c 03 0f 8e 25 05 00 00 41 8b 44 24 18 <c1> e0 0c 48 98 48 83 e8 01 80 7c 24 2a 00 0f 85 f9 02 00 00 4c 29
>> [  196.273770] RSP: 0018:ffffc9001cf575e8 EFLAGS: 00000246
>> [  196.273774] RAX: 0000000000000080 RBX: ffff88c1a0f95028 RCX: 0000000000000004
>> [  196.273775] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff88c08c8f9408
>>
>> [  196.273777] RBP: 0000000000000028 R08: 0000000000000000 R09: ffffed18341f2a05
>> [  196.273778] R10: ffff88c1a0f9502f R11: ffff88c1a0dbe1b8 R12: ffff88c1a0f95000
>> [  196.273779] R13: dffffc0000000000 R14: 0000000000000000 R15: ffffc9001cf577e0
>> [  196.273782]  </NMI>
>>
>> The issue is caused by below code in __grt_latency_data(). The code
>> tries to access x86_hybrid_pmu structure which doesn't exist on
>> non-hybrid platform like CWF.
>>
>>         WARN_ON_ONCE(hybrid_pmu(event->pmu)->pmu_type == hybrid_big)
>>
>> So add is_hybrid() check before calling this WARN_ON_ONCE to fix the
>> global-out-of-bounds access issue.
>>
>> Reported-by: Xudong Hao <xudong.hao@intel.com>
>> Cc: stable@vger.kernel.org
>> Fixes: 090262439f66 ("perf/x86/intel: Rename model-specific pebs_latency_data functions")
>> Signed-off-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
>> ---
>>  arch/x86/events/intel/ds.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/arch/x86/events/intel/ds.c b/arch/x86/events/intel/ds.c
>> index c0b7ac1c7594..d1ac1f1ceee9 100644
>> --- a/arch/x86/events/intel/ds.c
>> +++ b/arch/x86/events/intel/ds.c
>> @@ -317,7 +317,7 @@ static u64 __grt_latency_data(struct perf_event *event, u64 status,
>>  {
>>  	u64 val;
>>  
>> -	WARN_ON_ONCE(hybrid_pmu(event->pmu)->pmu_type == hybrid_big);
>> +	WARN_ON_ONCE(is_hybrid() && hybrid_pmu(event->pmu)->pmu_type == hybrid_big);
> BTW, this line has more than 80 characters.

Ah, yes, I would fix it in v2. Thanks.


>
> Reviewed-by: Zide Chen <zide.chen@intel.com>
>
>
>>  
>>  	dse &= PERF_PEBS_DATA_SOURCE_GRT_MASK;
>>  	val = hybrid_var(event->pmu, pebs_data_source)[dse];
>>
>> base-commit: 16ed389227651330879e17bd83d43bd234006722