RE: [PATCH 1/2] megaraid_sas: Fix for IO failing post OCR in SRIOV environment

[Sumit Saxena <sumit.saxena@broadcom.com>]

> -----Original Message-----
> From: Tomas Henzl [mailto:thenzl@redhat.com]
> Sent: Friday, January 29, 2016 9:47 PM
> To: Sumit Saxena; jbottomley@parallels.com; hch@infradead.org;
> martin.petersen@oracle.com
> Cc: linux-scsi@vger.kernel.org; kashyap.desai@avagotech.com; Uday
Lingala
> Subject: Re: [PATCH 1/2] megaraid_sas: Fix for IO failing post OCR in
SRIOV
> environment
>
> On 28.1.2016 16:44, Sumit Saxena wrote:
> > Driver assumes that VFs always have peers present whenever they have
same
> LD IDs. But this is not the case.
> > This patch handles the above mentioned by explicitly checking for a
peer
> before making HA/non-HA path decision.
> >
> > Signed-off-by: Uday Lingala <uday.lingala@avagotech.com>
> > Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
> >
> > ---
> >  drivers/scsi/megaraid/megaraid_sas.h        | 13 ++++++---
> >  drivers/scsi/megaraid/megaraid_sas_base.c   |  6 +++--
> >  drivers/scsi/megaraid/megaraid_sas_fusion.c | 42
> > ++++++++++++++++++-----------
> >  3 files changed, 39 insertions(+), 22 deletions(-)
> >
> > diff --git a/drivers/scsi/megaraid/megaraid_sas.h
> > b/drivers/scsi/megaraid/megaraid_sas.h
> > index b6fdb48..4484e63 100644
> > --- a/drivers/scsi/megaraid/megaraid_sas.h
> > +++ b/drivers/scsi/megaraid/megaraid_sas.h
> > @@ -393,6 +393,7 @@ enum MR_EVT_ARGS {
> >
> >
> >  #define SGE_BUFFER_SIZE	4096
> > +#define MEGASAS_CLUSTER_ID_SIZE	16
> >  /*
> >   * define constants for device list query options
> >   */
> > @@ -1227,7 +1228,8 @@ struct megasas_ctrl_info {
> >  	*/
> >  	struct {
> >  #if defined(__BIG_ENDIAN_BITFIELD)
> > -		u32     reserved:26;
> > +		u32     reserved:25;
> > +		u32     passive:1;
> >  		u32     premiumFeatureMismatch:1;
> >  		u32     ctrlPropIncompatible:1;
> >  		u32     fwVersionMismatch:1;
> > @@ -1241,11 +1243,12 @@ struct megasas_ctrl_info {
> >  		u32     fwVersionMismatch:1;
> >  		u32     ctrlPropIncompatible:1;
> >  		u32     premiumFeatureMismatch:1;
> > -		u32     reserved:26;
> > +		u32     passive:1;
> > +		u32     reserved:25;
> >  #endif
> >  	} cluster;
> >
> > -	char clusterId[16];                     /*7D4h */
> > +	char clusterId[MEGASAS_CLUSTER_ID_SIZE]; /*0x7D4 */
> >  	struct {
> >  		u8  maxVFsSupported;            /*0x7E4*/
> >  		u8  numVFsEnabled;              /*0x7E5*/
> > @@ -2126,7 +2129,9 @@ struct megasas_instance {
> >  	char skip_heartbeat_timer_del;
> >  	u8 requestorId;
> >  	char PlasmaFW111;
> > -	char mpio;
> > +	char clusterId[MEGASAS_CLUSTER_ID_SIZE];
> > +	u8 peerIsPresent;
> > +	u8 passive;
> >  	u16 throttlequeuedepth;
> >  	u8 mask_interrupts;
> >  	u16 max_chain_frame_sz;
> > diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c
> > b/drivers/scsi/megaraid/megaraid_sas_base.c
> > index 54922e5..52f3419 100644
> > --- a/drivers/scsi/megaraid/megaraid_sas_base.c
> > +++ b/drivers/scsi/megaraid/megaraid_sas_base.c
> > @@ -1943,7 +1943,7 @@ void megaraid_sas_kill_hba(struct
> megasas_instance *instance)
> >  		writel(MFI_STOP_ADP, &instance->reg_set->doorbell);
> >  		/* Flush */
> >  		readl(&instance->reg_set->doorbell);
> > -		if (instance->mpio && instance->requestorId)
> > +		if (instance->requestorId && instance->peerIsPresent)
> >  			memset(instance->ld_ids, 0xff,
> MEGASAS_MAX_LD_IDS);
> >  	} else {
> >  		writel(MFI_STOP_ADP,
> > @@ -5182,7 +5182,9 @@ static int megasas_init_fw(struct
> > megasas_instance *instance)
> >
> >  	tmp_sectors = min_t(u32, max_sectors_1, max_sectors_2);
> >
> > -	instance->mpio = ctrl_info->adapterOperations2.mpio;
> > +	instance->peerIsPresent = ctrl_info->cluster.peerIsPresent;
> > +	instance->passive = ctrl_info->cluster.passive;
> > +	memcpy(instance->clusterId, ctrl_info->clusterId,
> > +sizeof(instance->clusterId));
> >  	instance->UnevenSpanSupport =
> >  		ctrl_info->adapterOperations2.supportUnevenSpans;
> >  	if (instance->UnevenSpanSupport) {
> > diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.c
> > b/drivers/scsi/megaraid/megaraid_sas_fusion.c
> > index be9c3f1..d9d0029 100644
> > --- a/drivers/scsi/megaraid/megaraid_sas_fusion.c
> > +++ b/drivers/scsi/megaraid/megaraid_sas_fusion.c
> > @@ -3325,27 +3325,37 @@ out:
> >  	return ret;
> >  }
> >
> > +/*SRIOV get other instance in cluster if any*/ struct
> > +megasas_instance *megasas_get_peer_instance(struct megasas_instance
> > +*instance) {
> > +	int i;
> > +
> > +	for (i = 0; i < MAX_MGMT_ADAPTERS; i++) {
> > +		if (megasas_mgmt_info.instance[i] &&
> > +			(megasas_mgmt_info.instance[i] != instance) &&
> > +			 megasas_mgmt_info.instance[i]->requestorId &&
> > +			 megasas_mgmt_info.instance[i]->peerIsPresent &&
> > +
(memcmp((megasas_mgmt_info.instance[i]->clusterId),
> > +			instance->clusterId, MEGASAS_CLUSTER_ID_SIZE) ==
0))
> > +			return megasas_mgmt_info.instance[i];
> > +	}
> > +	return NULL;
> > +}
> > +
> >  /* Check for a second path that is currently UP */  int
> > megasas_check_mpio_paths(struct megasas_instance *instance,
> >  	struct scsi_cmnd *scmd)
> >  {
> > -	int i, j, retval = (DID_RESET << 16);
> > -
> > -	if (instance->mpio && instance->requestorId) {
>
> Seems to me that you've removed this test (instance->requestorId) maybe
it is
> not needed, is it removed intentionally ?

Sorry for delay in response. This is removed intentionally.
>
> --tm
>
>
> > -		for (i = 0 ; i < MAX_MGMT_ADAPTERS ; i++)
> > -			for (j = 0 ; j < MAX_LOGICAL_DRIVES; j++)
> > -				if (megasas_mgmt_info.instance[i] &&
> > -				    (megasas_mgmt_info.instance[i] !=
instance)
> &&
> > -				    megasas_mgmt_info.instance[i]->mpio &&
> > -				    megasas_mgmt_info.instance[i]-
> >requestorId
> > -				    &&
> > -
(megasas_mgmt_info.instance[i]->ld_ids[j]
> > -				     == scmd->device->id)) {
> > -					    retval = (DID_NO_CONNECT <<
16);
> > -					    goto out;
> > -				}
> > +	struct megasas_instance *peer_instance = NULL;
> > +	int retval = (DID_RESET << 16);
> > +
> > +	if (instance->peerIsPresent) {
> > +		peer_instance = megasas_get_peer_instance(instance);
> > +		if ((peer_instance) &&
> > +			(atomic_read(&peer_instance->adprecovery) ==
> > +			MEGASAS_HBA_OPERATIONAL))
> > +			retval = (DID_NO_CONNECT << 16);
> >  	}
> > -out:
> >  	return retval;
> >  }
> >