[Skiboot] [PATCH] opal/hmi: Fix the soft lockup issue on HMI for certain TB errors.

[Mahesh Jagannath Salgaonkar]

On 10/06/2015 02:04 AM, Benjamin Herrenschmidt wrote:
> On Mon, 2015-10-05 at 15:40 +0530, Mahesh J Salgaonkar wrote:
>> From: Mahesh Salgaonkar <mahesh at linux.vnet.ibm.com>
>>
>> Introduce volatile specifier in wait_for_subcore_threads() to force
>> re-read the fresh value from the memory. Without this, the compiler
>> optimizes the while loop not to re-fetch the data from memory pointed
>> by
>> this_cpu()->core_hmi_state_ptr. This cause CPU to spin infinitely
>> even though the other CPUs have modified the data causing soft lockup
>> in
>> kernel.
> 
> That should have a timeout, what if the other threads are stuck for
> some other reason ?
> 
> Also, I would much prefer if cpu_relax() had a memory clobber.

Do you mean we should add barrier() call inside cpu_relax() function ?

> 
> Cheers,
> Ben.
> 
>> Signed-off-by: Mahesh Salgaonkar <mahesh at linux.vnet.ibm.com>
>> ---
>>  core/hmi.c |    4 +++-
>>  1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/core/hmi.c b/core/hmi.c
>> index ee556fc..2d1ac6f 100644
>> --- a/core/hmi.c
>> +++ b/core/hmi.c
>> @@ -448,7 +448,9 @@ static int decode_malfunction(struct OpalHMIEvent
>> *hmi_evt)
>>  
>>  static void wait_for_subcore_threads(void)
>>  {
>> -	while (!(*(this_cpu()->core_hmi_state_ptr) &
>> HMI_STATE_CLEANUP_DONE))
>> +	volatile uint32_t *core_hmi_state_ptr = this_cpu()
>> ->core_hmi_state_ptr;
>> +
>> +	while (!(*(core_hmi_state_ptr) & HMI_STATE_CLEANUP_DONE))
>>  		cpu_relax();
>>  }
>>  
>>
>> _______________________________________________
>> Skiboot mailing list
>> Skiboot at lists.ozlabs.org
>> https://lists.ozlabs.org/listinfo/skiboot