RE: [PATCH 09/22] [SCSI] mpt2sas, mpt3sas: Added a support to set cpu affinity for each MSIX vector enabled by the HBA

[Elliott, Robert (Server Storage)]

> -----Original Message-----
> From: linux-scsi-owner@xxxxxxxxxxxxxxx [mailto:linux-scsi-
> owner@xxxxxxxxxxxxxxx] On Behalf Of Sreekanth Reddy
> Sent: Tuesday, 09 December, 2014 6:17 AM
> To: martin.petersen@xxxxxxxxxx; jejb@xxxxxxxxxx; hch@xxxxxxxxxxxxx
...
> Change_set:
> 1. Added affinity_hint varable of type cpumask_var_t in adapter_reply_queue
>    structure. And allocated a memory for this varable by calling
> zalloc_cpumask_var.
> 2. Call the API irq_set_affinity_hint for each MSIx vector to affiniate it
>    with calculated cpus at driver inilization time.
> 3. While freeing the MSIX vector, call this same API to release the cpu
> affinity mask
>    for each MSIx vector by providing the NULL value in cpumask argument.
> 4. then call the free_cpumask_var API to free the memory allocated in step 2.
> 
...

> diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.c
> b/drivers/scsi/mpt3sas/mpt3sas_base.c
> index 1560115..f0f8ba0 100644
> --- a/drivers/scsi/mpt3sas/mpt3sas_base.c
> +++ b/drivers/scsi/mpt3sas/mpt3sas_base.c
...
> @@ -1609,6 +1611,10 @@ _base_request_irq(struct MPT3SAS_ADAPTER *ioc, u8
> index, u32 vector)
>  	reply_q->ioc = ioc;
>  	reply_q->msix_index = index;
>  	reply_q->vector = vector;
> +
> +	if (!zalloc_cpumask_var(&reply_q->affinity_hint, GFP_KERNEL))
> +		return -ENOMEM;

I think this will create the problem Alex Thorlton just reported
with lpfc on a system with a huge number (6144) of CPUs.

See this thread:
	[BUG] kzalloc overflow in lpfc driver on 6k core system

---
Rob Elliott    HP Server Storage



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/