diff mbox

[3/4] ACPI: IORT: Skip SMMUv3 device ID map for two steps mappings

Message ID 1506475215-2731-4-git-send-email-hanjun.guo@linaro.org (mailing list archive)
State Not Applicable, archived
Headers show

Commit Message

Hanjun Guo Sept. 27, 2017, 1:20 a.m. UTC
IORT revision C introduced SMMUv3 MSI support which adding a
device ID mapping index in SMMUv3 sub table, to get the SMMUv3
device ID mapping for the output ID (dev ID for ITS) and the
link to which ITS.

So if a platform supports SMMUv3 MSI for control interrupt,
there will be a additional single map entry under SMMU, this
will not introduce any difference for devices just use one
step map to get its output ID and parent (ITS or SMMU), such
as PCI/NC/PMCG ---> ITS or PCI/NC ---> SMMU, but we need to
do the special handling for two steps map case such as
PCI/NC--->SMMUv3--->ITS.

Take a PCI hostbridge for example,

|----------------------|
|  Root Complex Node   |
|----------------------|
|    map entry[x]      |
|----------------------|
|       id value       |
| output_reference     |
|---|------------------|
    |
    |   |----------------------|
    |-->|        SMMUv3        |
        |----------------------|
        |     SMMU dev ID      |
        |     mapping index 0  |
        |----------------------|
        |      map entry[0]    |
        |----------------------|
        |       id value       |
        | output_reference-----------> ITS 1 (SMMU MSI domain)
        |----------------------|
        |      map entry[1]    |
        |----------------------|
        |       id value       |
        | output_reference-----------> ITS 2 (PCI MSI domain)
        |----------------------|

When the SMMU dev ID mapping index is 0, there is entry[0]
to map to a ITS, we need to skip that map entry for PCI
or NC (named component), or we may get the wrong ITS parent.

Introduce iort_get_id_mapping_index() to get the index then
skip the map entry in iort_node_map_id(), also to get the
dev ID directly for iort_pmsi_get_dev_id() for ITS platform
MSI preparation.

Signed-off-by: Hanjun Guo <hanjun.guo@linaro.org>
---
 drivers/acpi/arm64/iort.c | 64 +++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 59 insertions(+), 5 deletions(-)

Comments

Lorenzo Pieralisi Sept. 27, 2017, 1:54 p.m. UTC | #1
Hi Hanjun,

On Wed, Sep 27, 2017 at 09:20:14AM +0800, Hanjun Guo wrote:
> IORT revision C introduced SMMUv3 MSI support which adding a
> device ID mapping index in SMMUv3 sub table, to get the SMMUv3
> device ID mapping for the output ID (dev ID for ITS) and the
> link to which ITS.
> 
> So if a platform supports SMMUv3 MSI for control interrupt,
> there will be a additional single map entry under SMMU, this
> will not introduce any difference for devices just use one
> step map to get its output ID and parent (ITS or SMMU), such
> as PCI/NC/PMCG ---> ITS or PCI/NC ---> SMMU, but we need to
> do the special handling for two steps map case such as
> PCI/NC--->SMMUv3--->ITS.
> 
> Take a PCI hostbridge for example,
> 
> |----------------------|
> |  Root Complex Node   |
> |----------------------|
> |    map entry[x]      |
> |----------------------|
> |       id value       |
> | output_reference     |
> |---|------------------|
>     |
>     |   |----------------------|
>     |-->|        SMMUv3        |
>         |----------------------|
>         |     SMMU dev ID      |
>         |     mapping index 0  |
>         |----------------------|
>         |      map entry[0]    |
>         |----------------------|
>         |       id value       |
>         | output_reference-----------> ITS 1 (SMMU MSI domain)
>         |----------------------|
>         |      map entry[1]    |
>         |----------------------|
>         |       id value       |
>         | output_reference-----------> ITS 2 (PCI MSI domain)
>         |----------------------|
> 
> When the SMMU dev ID mapping index is 0, there is entry[0]
> to map to a ITS, we need to skip that map entry for PCI
> or NC (named component), or we may get the wrong ITS parent.

Is this actually true ? I think that currently we would simply skip
the entry and print an error log but we can't get a wrong ITS parent.

I am rewriting this commit (I will probably split it), it is doing the
right thing but the commit log is stale (probably caused by code
reshuffling).

Thanks,
Lorenzo

> Introduce iort_get_id_mapping_index() to get the index then
> skip the map entry in iort_node_map_id(), also to get the
> dev ID directly for iort_pmsi_get_dev_id() for ITS platform
> MSI preparation.
> 
> Signed-off-by: Hanjun Guo <hanjun.guo@linaro.org>
> ---
>  drivers/acpi/arm64/iort.c | 64 +++++++++++++++++++++++++++++++++++++++++++----
>  1 file changed, 59 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/acpi/arm64/iort.c b/drivers/acpi/arm64/iort.c
> index db71d7f..1c1160e 100644
> --- a/drivers/acpi/arm64/iort.c
> +++ b/drivers/acpi/arm64/iort.c
> @@ -357,7 +357,8 @@ struct acpi_iort_node *iort_node_get_id(struct acpi_iort_node *node,
>  
>  	if (map->flags & ACPI_IORT_ID_SINGLE_MAPPING) {
>  		if (node->type == ACPI_IORT_NODE_NAMED_COMPONENT ||
> -		    node->type == ACPI_IORT_NODE_PCI_ROOT_COMPLEX) {
> +		    node->type == ACPI_IORT_NODE_PCI_ROOT_COMPLEX ||
> +		    node->type == ACPI_IORT_NODE_SMMU_V3) {
>  			*id_out = map->output_base;
>  			return parent;
>  		}
> @@ -366,6 +367,40 @@ struct acpi_iort_node *iort_node_get_id(struct acpi_iort_node *node,
>  	return NULL;
>  }
>  
> +static int iort_get_id_mapping_index(struct acpi_iort_node *node)
> +{
> +	struct acpi_iort_smmu_v3 *smmu;
> +
> +	switch (node->type) {
> +	case ACPI_IORT_NODE_SMMU_V3:
> +		/*
> +		 * SMMUv3 dev ID mapping index was introdueced in revision 1
> +		 * table, not available in revision 0
> +		 */
> +		if (node->revision < 1)
> +			return -EINVAL;
> +
> +		smmu = (struct acpi_iort_smmu_v3 *)node->node_data;
> +		/*
> +		 * ID mapping index is only ignored if all interrupts are
> +		 * GSIV based
> +		 */
> +		if (smmu->event_gsiv && smmu->pri_gsiv && smmu->gerr_gsiv
> +		    && smmu->sync_gsiv)
> +			return -EINVAL;
> +
> +		if (smmu->id_mapping_index >= node->mapping_count) {
> +			pr_err(FW_BUG "[node %p type %d] ID mapping index overflows valid mappings\n",
> +			       node, node->type);
> +			return -EINVAL;
> +		}
> +
> +		return smmu->id_mapping_index;
> +	default:
> +		return -EINVAL;
> +	}
> +}
> +
>  static struct acpi_iort_node *iort_node_map_id(struct acpi_iort_node *node,
>  					       u32 id_in, u32 *id_out,
>  					       u8 type_mask)
> @@ -375,7 +410,7 @@ static struct acpi_iort_node *iort_node_map_id(struct acpi_iort_node *node,
>  	/* Parse the ID mapping tree to find specified node type */
>  	while (node) {
>  		struct acpi_iort_id_mapping *map;
> -		int i;
> +		int i, index;
>  
>  		if (IORT_TYPE_MASK(node->type) & type_mask) {
>  			if (id_out)
> @@ -396,8 +431,19 @@ static struct acpi_iort_node *iort_node_map_id(struct acpi_iort_node *node,
>  			goto fail_map;
>  		}
>  
> +		/*
> +		 * we need to get SMMUv3 dev ID mapping index and skip its
> +		 * associated ID map for single mapping cases, error value
> +		 * returned for index will be an invalid value in practical.
> +		 */
> +		index = iort_get_id_mapping_index(node);
> +
>  		/* Do the ID translation */
>  		for (i = 0; i < node->mapping_count; i++, map++) {
> +			/* if it's a SMMUv3 device id mapping index, skip it */
> +			if (i == index)
> +				continue;
> +
>  			if (!iort_id_map(map, node->type, id, &id))
>  				break;
>  		}
> @@ -507,16 +553,24 @@ u32 iort_msi_map_rid(struct device *dev, u32 req_id)
>   */
>  int iort_pmsi_get_dev_id(struct device *dev, u32 *dev_id)
>  {
> -	int i;
> +	int i, index;
>  	struct acpi_iort_node *node;
>  
>  	node = iort_find_dev_node(dev);
>  	if (!node)
>  		return -ENODEV;
>  
> -	for (i = 0; i < node->mapping_count; i++) {
> -		if (iort_node_map_platform_id(node, dev_id, IORT_MSI_TYPE, i))
> +	index = iort_get_id_mapping_index(node);
> +	/* if there is a valid index, go get the dev_id directly */
> +	if (index >= 0) {
> +		if (iort_node_get_id(node, dev_id, index))
>  			return 0;
> +	} else {
> +		for (i = 0; i < node->mapping_count; i++) {
> +			if (iort_node_map_platform_id(node, dev_id,
> +						      IORT_MSI_TYPE, i))
> +				return 0;
> +		}
>  	}
>  
>  	return -ENODEV;
> -- 
> 1.9.1
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Hanjun Guo Oct. 10, 2017, 6:47 a.m. UTC | #2
Hi Lorenzo,

Sorry for the late reply, holidays in China for the past week.

At 2017/9/27 21:54, Lorenzo Pieralisi wrote:
> Hi Hanjun,
>
> On Wed, Sep 27, 2017 at 09:20:14AM +0800, Hanjun Guo wrote:
>> IORT revision C introduced SMMUv3 MSI support which adding a
>> device ID mapping index in SMMUv3 sub table, to get the SMMUv3
>> device ID mapping for the output ID (dev ID for ITS) and the
>> link to which ITS.
>>
>> So if a platform supports SMMUv3 MSI for control interrupt,
>> there will be a additional single map entry under SMMU, this
>> will not introduce any difference for devices just use one
>> step map to get its output ID and parent (ITS or SMMU), such
>> as PCI/NC/PMCG ---> ITS or PCI/NC ---> SMMU, but we need to
>> do the special handling for two steps map case such as
>> PCI/NC--->SMMUv3--->ITS.
>>
>> Take a PCI hostbridge for example,
>>
>> |----------------------|
>> |  Root Complex Node   |
>> |----------------------|
>> |    map entry[x]      |
>> |----------------------|
>> |       id value       |
>> | output_reference     |
>> |---|------------------|
>>      |
>>      |   |----------------------|
>>      |-->|        SMMUv3        |
>>          |----------------------|
>>          |     SMMU dev ID      |
>>          |     mapping index 0  |
>>          |----------------------|
>>          |      map entry[0]    |
>>          |----------------------|
>>          |       id value       |
>>          | output_reference-----------> ITS 1 (SMMU MSI domain)
>>          |----------------------|
>>          |      map entry[1]    |
>>          |----------------------|
>>          |       id value       |
>>          | output_reference-----------> ITS 2 (PCI MSI domain)
>>          |----------------------|
>>
>> When the SMMU dev ID mapping index is 0, there is entry[0]
>> to map to a ITS, we need to skip that map entry for PCI
>> or NC (named component), or we may get the wrong ITS parent.
>
> Is this actually true ? I think that currently we would simply skip
> the entry and print an error log but we can't get a wrong ITS parent.

So the only valid single mapping under type SMMUv3 is SMMUv3's dev id
mapping, we need to fix the IORT spec as well.

>
> I am rewriting this commit (I will probably split it), it is doing the
> right thing but the commit log is stale (probably caused by code
> reshuffling).

Do I need to resend another version, or you can help to update it?
please let me know.

By the way, this patch set was tested on Hisilicon Hip08 with MSI
for SMMU, and it works fine.

Thanks
Hanjun
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Lorenzo Pieralisi Oct. 10, 2017, 9:20 a.m. UTC | #3
On Tue, Oct 10, 2017 at 02:47:53PM +0800, Hanjun Guo wrote:
> Hi Lorenzo,
> 
> Sorry for the late reply, holidays in China for the past week.
> 
> At 2017/9/27 21:54, Lorenzo Pieralisi wrote:
> > Hi Hanjun,
> >
> > On Wed, Sep 27, 2017 at 09:20:14AM +0800, Hanjun Guo wrote:
> >> IORT revision C introduced SMMUv3 MSI support which adding a
> >> device ID mapping index in SMMUv3 sub table, to get the SMMUv3
> >> device ID mapping for the output ID (dev ID for ITS) and the
> >> link to which ITS.
> >>
> >> So if a platform supports SMMUv3 MSI for control interrupt,
> >> there will be a additional single map entry under SMMU, this
> >> will not introduce any difference for devices just use one
> >> step map to get its output ID and parent (ITS or SMMU), such
> >> as PCI/NC/PMCG ---> ITS or PCI/NC ---> SMMU, but we need to
> >> do the special handling for two steps map case such as
> >> PCI/NC--->SMMUv3--->ITS.
> >>
> >> Take a PCI hostbridge for example,
> >>
> >> |----------------------|
> >> |  Root Complex Node   |
> >> |----------------------|
> >> |    map entry[x]      |
> >> |----------------------|
> >> |       id value       |
> >> | output_reference     |
> >> |---|------------------|
> >>      |
> >>      |   |----------------------|
> >>      |-->|        SMMUv3        |
> >>          |----------------------|
> >>          |     SMMU dev ID      |
> >>          |     mapping index 0  |
> >>          |----------------------|
> >>          |      map entry[0]    |
> >>          |----------------------|
> >>          |       id value       |
> >>          | output_reference-----------> ITS 1 (SMMU MSI domain)
> >>          |----------------------|
> >>          |      map entry[1]    |
> >>          |----------------------|
> >>          |       id value       |
> >>          | output_reference-----------> ITS 2 (PCI MSI domain)
> >>          |----------------------|
> >>
> >> When the SMMU dev ID mapping index is 0, there is entry[0]
> >> to map to a ITS, we need to skip that map entry for PCI
> >> or NC (named component), or we may get the wrong ITS parent.
> >
> > Is this actually true ? I think that currently we would simply skip
> > the entry and print an error log but we can't get a wrong ITS parent.
> 
> So the only valid single mapping under type SMMUv3 is SMMUv3's dev id
> mapping, we need to fix the IORT spec as well.
> 
> >
> > I am rewriting this commit (I will probably split it), it is doing the
> > right thing but the commit log is stale (probably caused by code
> > reshuffling).
> 
> Do I need to resend another version, or you can help to update it?
> please let me know.

I reworked the patches, you can repost/retest them I made them available
in the branch below, we will have to add a guard around ACPICA smmu
struct (unfortunately I think we will have to use the ACPICA version as
a guard) or I can ask Rafael to pull the series if ACPICA goes via ACPI
tree (and your patch made it into the release - I will check ACPICA
upstream).

git://git.kernel.org/pub/scm/linux/kernel/git/lpieralisi/linux.git iort/smmu-msi-for-4.15
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Hanjun Guo Oct. 11, 2017, 5:26 a.m. UTC | #4
On 2017/10/10 17:20, Lorenzo Pieralisi wrote:
> On Tue, Oct 10, 2017 at 02:47:53PM +0800, Hanjun Guo wrote:
>> Hi Lorenzo,
>>
>> Sorry for the late reply, holidays in China for the past week.
>>
>> At 2017/9/27 21:54, Lorenzo Pieralisi wrote:
>>> Hi Hanjun,
>>>
>>> On Wed, Sep 27, 2017 at 09:20:14AM +0800, Hanjun Guo wrote:
>>>> IORT revision C introduced SMMUv3 MSI support which adding a
>>>> device ID mapping index in SMMUv3 sub table, to get the SMMUv3
>>>> device ID mapping for the output ID (dev ID for ITS) and the
>>>> link to which ITS.
>>>>
>>>> So if a platform supports SMMUv3 MSI for control interrupt,
>>>> there will be a additional single map entry under SMMU, this
>>>> will not introduce any difference for devices just use one
>>>> step map to get its output ID and parent (ITS or SMMU), such
>>>> as PCI/NC/PMCG ---> ITS or PCI/NC ---> SMMU, but we need to
>>>> do the special handling for two steps map case such as
>>>> PCI/NC--->SMMUv3--->ITS.
>>>>
>>>> Take a PCI hostbridge for example,
>>>>
>>>> |----------------------|
>>>> |  Root Complex Node   |
>>>> |----------------------|
>>>> |    map entry[x]      |
>>>> |----------------------|
>>>> |       id value       |
>>>> | output_reference     |
>>>> |---|------------------|
>>>>      |
>>>>      |   |----------------------|
>>>>      |-->|        SMMUv3        |
>>>>          |----------------------|
>>>>          |     SMMU dev ID      |
>>>>          |     mapping index 0  |
>>>>          |----------------------|
>>>>          |      map entry[0]    |
>>>>          |----------------------|
>>>>          |       id value       |
>>>>          | output_reference-----------> ITS 1 (SMMU MSI domain)
>>>>          |----------------------|
>>>>          |      map entry[1]    |
>>>>          |----------------------|
>>>>          |       id value       |
>>>>          | output_reference-----------> ITS 2 (PCI MSI domain)
>>>>          |----------------------|
>>>>
>>>> When the SMMU dev ID mapping index is 0, there is entry[0]
>>>> to map to a ITS, we need to skip that map entry for PCI
>>>> or NC (named component), or we may get the wrong ITS parent.
>>> Is this actually true ? I think that currently we would simply skip
>>> the entry and print an error log but we can't get a wrong ITS parent.
>> So the only valid single mapping under type SMMUv3 is SMMUv3's dev id
>> mapping, we need to fix the IORT spec as well.
>>
>>> I am rewriting this commit (I will probably split it), it is doing the
>>> right thing but the commit log is stale (probably caused by code
>>> reshuffling).
>> Do I need to resend another version, or you can help to update it?
>> please let me know.
> I reworked the patches, you can repost/retest them I made them available
> in the branch below, we will have to add a guard around ACPICA smmu
> struct (unfortunately I think we will have to use the ACPICA version as
> a guard) or I can ask Rafael to pull the series if ACPICA goes via ACPI
> tree (and your patch made it into the release - I will check ACPICA
> upstream).

Bob already merged my pull request yesterday, I think it will be ready for
acpica release for this month.

>
> git://git.kernel.org/pub/scm/linux/kernel/git/lpieralisi/linux.git iort/smmu-msi-for-4.15
>

Thanks! I will retest then repost.

Hanjun

--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Lorenzo Pieralisi Oct. 11, 2017, 10:24 a.m. UTC | #5
On Wed, Oct 11, 2017 at 01:26:10PM +0800, Hanjun Guo wrote:
> On 2017/10/10 17:20, Lorenzo Pieralisi wrote:
> > On Tue, Oct 10, 2017 at 02:47:53PM +0800, Hanjun Guo wrote:
> >> Hi Lorenzo,
> >>
> >> Sorry for the late reply, holidays in China for the past week.
> >>
> >> At 2017/9/27 21:54, Lorenzo Pieralisi wrote:
> >>> Hi Hanjun,
> >>>
> >>> On Wed, Sep 27, 2017 at 09:20:14AM +0800, Hanjun Guo wrote:
> >>>> IORT revision C introduced SMMUv3 MSI support which adding a
> >>>> device ID mapping index in SMMUv3 sub table, to get the SMMUv3
> >>>> device ID mapping for the output ID (dev ID for ITS) and the
> >>>> link to which ITS.
> >>>>
> >>>> So if a platform supports SMMUv3 MSI for control interrupt,
> >>>> there will be a additional single map entry under SMMU, this
> >>>> will not introduce any difference for devices just use one
> >>>> step map to get its output ID and parent (ITS or SMMU), such
> >>>> as PCI/NC/PMCG ---> ITS or PCI/NC ---> SMMU, but we need to
> >>>> do the special handling for two steps map case such as
> >>>> PCI/NC--->SMMUv3--->ITS.
> >>>>
> >>>> Take a PCI hostbridge for example,
> >>>>
> >>>> |----------------------|
> >>>> |  Root Complex Node   |
> >>>> |----------------------|
> >>>> |    map entry[x]      |
> >>>> |----------------------|
> >>>> |       id value       |
> >>>> | output_reference     |
> >>>> |---|------------------|
> >>>>      |
> >>>>      |   |----------------------|
> >>>>      |-->|        SMMUv3        |
> >>>>          |----------------------|
> >>>>          |     SMMU dev ID      |
> >>>>          |     mapping index 0  |
> >>>>          |----------------------|
> >>>>          |      map entry[0]    |
> >>>>          |----------------------|
> >>>>          |       id value       |
> >>>>          | output_reference-----------> ITS 1 (SMMU MSI domain)
> >>>>          |----------------------|
> >>>>          |      map entry[1]    |
> >>>>          |----------------------|
> >>>>          |       id value       |
> >>>>          | output_reference-----------> ITS 2 (PCI MSI domain)
> >>>>          |----------------------|
> >>>>
> >>>> When the SMMU dev ID mapping index is 0, there is entry[0]
> >>>> to map to a ITS, we need to skip that map entry for PCI
> >>>> or NC (named component), or we may get the wrong ITS parent.
> >>> Is this actually true ? I think that currently we would simply skip
> >>> the entry and print an error log but we can't get a wrong ITS parent.
> >> So the only valid single mapping under type SMMUv3 is SMMUv3's dev id
> >> mapping, we need to fix the IORT spec as well.
> >>
> >>> I am rewriting this commit (I will probably split it), it is doing the
> >>> right thing but the commit log is stale (probably caused by code
> >>> reshuffling).
> >> Do I need to resend another version, or you can help to update it?
> >> please let me know.
> > I reworked the patches, you can repost/retest them I made them available
> > in the branch below, we will have to add a guard around ACPICA smmu
> > struct (unfortunately I think we will have to use the ACPICA version as
> > a guard) or I can ask Rafael to pull the series if ACPICA goes via ACPI
> > tree (and your patch made it into the release - I will check ACPICA
> > upstream).
> 
> Bob already merged my pull request yesterday, I think it will be ready for
> acpica release for this month.

That's good, mind updating the patch series with an ACPICA guard in IORT
code in preparation for the pull request ?

> > git://git.kernel.org/pub/scm/linux/kernel/git/lpieralisi/linux.git iort/smmu-msi-for-4.15
> >
> 
> Thanks! I will retest then repost.

Thank you,
Lorenzo
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/drivers/acpi/arm64/iort.c b/drivers/acpi/arm64/iort.c
index db71d7f..1c1160e 100644
--- a/drivers/acpi/arm64/iort.c
+++ b/drivers/acpi/arm64/iort.c
@@ -357,7 +357,8 @@  struct acpi_iort_node *iort_node_get_id(struct acpi_iort_node *node,
 
 	if (map->flags & ACPI_IORT_ID_SINGLE_MAPPING) {
 		if (node->type == ACPI_IORT_NODE_NAMED_COMPONENT ||
-		    node->type == ACPI_IORT_NODE_PCI_ROOT_COMPLEX) {
+		    node->type == ACPI_IORT_NODE_PCI_ROOT_COMPLEX ||
+		    node->type == ACPI_IORT_NODE_SMMU_V3) {
 			*id_out = map->output_base;
 			return parent;
 		}
@@ -366,6 +367,40 @@  struct acpi_iort_node *iort_node_get_id(struct acpi_iort_node *node,
 	return NULL;
 }
 
+static int iort_get_id_mapping_index(struct acpi_iort_node *node)
+{
+	struct acpi_iort_smmu_v3 *smmu;
+
+	switch (node->type) {
+	case ACPI_IORT_NODE_SMMU_V3:
+		/*
+		 * SMMUv3 dev ID mapping index was introdueced in revision 1
+		 * table, not available in revision 0
+		 */
+		if (node->revision < 1)
+			return -EINVAL;
+
+		smmu = (struct acpi_iort_smmu_v3 *)node->node_data;
+		/*
+		 * ID mapping index is only ignored if all interrupts are
+		 * GSIV based
+		 */
+		if (smmu->event_gsiv && smmu->pri_gsiv && smmu->gerr_gsiv
+		    && smmu->sync_gsiv)
+			return -EINVAL;
+
+		if (smmu->id_mapping_index >= node->mapping_count) {
+			pr_err(FW_BUG "[node %p type %d] ID mapping index overflows valid mappings\n",
+			       node, node->type);
+			return -EINVAL;
+		}
+
+		return smmu->id_mapping_index;
+	default:
+		return -EINVAL;
+	}
+}
+
 static struct acpi_iort_node *iort_node_map_id(struct acpi_iort_node *node,
 					       u32 id_in, u32 *id_out,
 					       u8 type_mask)
@@ -375,7 +410,7 @@  static struct acpi_iort_node *iort_node_map_id(struct acpi_iort_node *node,
 	/* Parse the ID mapping tree to find specified node type */
 	while (node) {
 		struct acpi_iort_id_mapping *map;
-		int i;
+		int i, index;
 
 		if (IORT_TYPE_MASK(node->type) & type_mask) {
 			if (id_out)
@@ -396,8 +431,19 @@  static struct acpi_iort_node *iort_node_map_id(struct acpi_iort_node *node,
 			goto fail_map;
 		}
 
+		/*
+		 * we need to get SMMUv3 dev ID mapping index and skip its
+		 * associated ID map for single mapping cases, error value
+		 * returned for index will be an invalid value in practical.
+		 */
+		index = iort_get_id_mapping_index(node);
+
 		/* Do the ID translation */
 		for (i = 0; i < node->mapping_count; i++, map++) {
+			/* if it's a SMMUv3 device id mapping index, skip it */
+			if (i == index)
+				continue;
+
 			if (!iort_id_map(map, node->type, id, &id))
 				break;
 		}
@@ -507,16 +553,24 @@  u32 iort_msi_map_rid(struct device *dev, u32 req_id)
  */
 int iort_pmsi_get_dev_id(struct device *dev, u32 *dev_id)
 {
-	int i;
+	int i, index;
 	struct acpi_iort_node *node;
 
 	node = iort_find_dev_node(dev);
 	if (!node)
 		return -ENODEV;
 
-	for (i = 0; i < node->mapping_count; i++) {
-		if (iort_node_map_platform_id(node, dev_id, IORT_MSI_TYPE, i))
+	index = iort_get_id_mapping_index(node);
+	/* if there is a valid index, go get the dev_id directly */
+	if (index >= 0) {
+		if (iort_node_get_id(node, dev_id, index))
 			return 0;
+	} else {
+		for (i = 0; i < node->mapping_count; i++) {
+			if (iort_node_map_platform_id(node, dev_id,
+						      IORT_MSI_TYPE, i))
+				return 0;
+		}
 	}
 
 	return -ENODEV;