[PATCH] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

[PATCH] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

[Ivy Lopez]

dev_to_node() can return NUMA_NO_NODE (-1) on systems without NUMA
topology information for the PCI device, such as single-socket
boards that don't expose device-to-node affinity. Passing -1
directly into cpumask_of_node() indexes node_to_cpumask_map[-1],
an out-of-bounds array read caught by UBSAN:

  UBSAN: array-index-out-of-bounds in arch/x86/include/asm/topology.h:72:28
  index -1 is out of range for type 'cpumask *[1024]'

Fall back to cpu_online_mask when no NUMA node is available, rather
than assuming dev_to_node() always returns a valid node index.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=221294
Signed-off-by: Ivy Lopez <skunkolee@gmail.com>
---
 drivers/scsi/mpt3sas/mpt3sas_base.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.c b/drivers/scsi/mpt3sas/mpt3sas_base.c
index 79052f2accbd..eaad6fb7f3cf 100644
--- a/drivers/scsi/mpt3sas/mpt3sas_base.c
+++ b/drivers/scsi/mpt3sas/mpt3sas_base.c
@@ -3238,7 +3238,9 @@ _base_assign_reply_queues(struct MPT3SAS_ADAPTER *ioc)
 		 * corresponding to high iops queues.
 		 */
 		if (ioc->high_iops_queues) {
-			mask = cpumask_of_node(dev_to_node(&ioc->pdev->dev));
+			int node = (dev_to_node(&ioc->pdev->dev));
+
+			mask = (node == NUMA_NO_NODE) ? cpu_online_mask : cpumask_of_node(node);
 			for (index = 0; index < ioc->high_iops_queues;
 			    index++) {
 				irq = pci_irq_vector(ioc->pdev, index);
-- 
2.55.0

Re: [PATCH] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Incorrect calculation of remaining MSI-X queues causes a Division by Zero kernel panic or severe IO imbalance.
--

--- Patch [1]: [PATCH] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues() ---
commit 2577a38cb04664cd3aa266df2234da5f8d4aff10
Author: Ivy Lopez <skunkolee@gmail.com>

scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

Fixes an out-of-bounds array read in cpumask_of_node() by checking if
dev_to_node() returns NUMA_NO_NODE. If so, it falls back to cpu_online_mask
instead of using the invalid NUMA node index.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=221294

> diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.c b/drivers/scsi/mpt3sas/mpt3sas_base.c
> index fed7aeffec58d..b354cbcd12741 100644
> --- a/drivers/scsi/mpt3sas/mpt3sas_base.c
> +++ b/drivers/scsi/mpt3sas/mpt3sas_base.c
> @@ -3238,7 +3238,9 @@ _base_assign_reply_queues(struct MPT3SAS_ADAPTER *ioc)
>  		 * corresponding to high iops queues.
>  		 */
>  		if (ioc->high_iops_queues) {
> -			mask = cpumask_of_node(dev_to_node(&ioc->pdev->dev));
> +			int node = (dev_to_node(&ioc->pdev->dev));
> +
> +			mask = (node == NUMA_NO_NODE) ? cpu_online_mask : cpumask_of_node(node);
>  			for (index = 0; index < ioc->high_iops_queues;
>  			    index++) {
>  				irq = pci_irq_vector(ioc->pdev, index);

[ ... ]

[Severity: High]
This is a pre-existing issue, but I noticed a potential division by zero
later in drivers/scsi/mpt3sas/mpt3sas_base.c:_base_assign_reply_queues():

fall_back:
	cpu = cpumask_first(cpu_online_mask);
	nr_msix -= (ioc->high_iops_queues - iopoll_q_count);
	index = 0;

	list_for_each_entry(reply_q, &ioc->reply_queue_list, list) {
		unsigned int i, group = nr_cpus / nr_msix;

Can this result in a division by zero panic?

If iopoll_q_count == 0 and all available vectors are assigned to
high_iops_queues, nr_msix becomes 0. The list loop unconditionally computes
group = nr_cpus / nr_msix before checking if the queue is a default queue,
triggering a division by zero.

Additionally, does the fallback logic incorrectly add the poll queues instead
of subtracting them when iopoll_q_count > 0? This would shrink the CPU group
size and could cause unassigned CPUs to fall back to a single queue.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260817221300.43286-1-skunkolee@gmail.com?part=1

Re: [PATCH] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

[Johannes Thumshirn]

On Mon, Aug 17, 2026 at 04:13:00PM -0600, Ivy Lopez wrote:
> dev_to_node() can return NUMA_NO_NODE (-1) on systems without NUMA
> topology information for the PCI device, such as single-socket
> boards that don't expose device-to-node affinity. Passing -1
> directly into cpumask_of_node() indexes node_to_cpumask_map[-1],
> an out-of-bounds array read caught by UBSAN:
> 
>   UBSAN: array-index-out-of-bounds in arch/x86/include/asm/topology.h:72:28
>   index -1 is out of range for type 'cpumask *[1024]'
> 
> Fall back to cpu_online_mask when no NUMA node is available, rather
> than assuming dev_to_node() always returns a valid node index.
> 
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=221294
> Signed-off-by: Ivy Lopez <skunkolee@gmail.com>
> ---
>  drivers/scsi/mpt3sas/mpt3sas_base.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.c b/drivers/scsi/mpt3sas/mpt3sas_base.c
> index 79052f2accbd..eaad6fb7f3cf 100644
> --- a/drivers/scsi/mpt3sas/mpt3sas_base.c
> +++ b/drivers/scsi/mpt3sas/mpt3sas_base.c
> @@ -3238,7 +3238,9 @@ _base_assign_reply_queues(struct MPT3SAS_ADAPTER *ioc)
>  		 * corresponding to high iops queues.
>  		 */
>  		if (ioc->high_iops_queues) {
> -			mask = cpumask_of_node(dev_to_node(&ioc->pdev->dev));
> +			int node = (dev_to_node(&ioc->pdev->dev));

Why the superfluous parenthesis?

> +
> +			mask = (node == NUMA_NO_NODE) ? cpu_online_mask : cpumask_of_node(node);

Overly long line here.

>  			for (index = 0; index < ioc->high_iops_queues;
>  			    index++) {
>  				irq = pci_irq_vector(ioc->pdev, index);
> -- 
> 2.55.0
> 
> 

Re: [PATCH] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

[John Garry]

On 17/08/2026 23:13, Ivy Lopez wrote:
> dev_to_node() can return NUMA_NO_NODE (-1) on systems without NUMA
> topology information for the PCI device, such as single-socket
> boards that don't expose device-to-node affinity. 

I think that this problem would have appeared by now for this driver (if 
it were really a problem).

However, for CONFIG_NUMA=n dev_to_node() always returns NUMA_NO_NODE.

> Passing -1
> directly into cpumask_of_node() indexes node_to_cpumask_map[-1],
> an out-of-bounds array read caught by UBSAN:
> 
>    UBSAN: array-index-out-of-bounds in arch/x86/include/asm/topology.h:72:28
>    index -1 is out of range for type 'cpumask *[1024]'

x86 seems to be only arch which has this problem now.

I have tried unsuccessfully to fix it:
https://lore.kernel.org/lkml/20260107094007.966496-5-john.g.garry@oracle.com/

[PATCH v2] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

[Ivy Lopez]

dev_to_node() can return NUMA_NO_NODE (-1) on systems without NUMA
topology information for the PCI device, such as single-socket
boards that don't expose device-to-node affinity. Passing -1
directly into cpumask_of_node() indexes node_to_cpumask_map[-1],
an out-of-bounds array read caught by UBSAN:

  UBSAN: array-index-out-of-bounds in arch/x86/include/asm/topology.h:72:28
  index -1 is out of range for type 'cpumask *[1024]'

Fall back to cpu_online_mask when no NUMA node is available, rather
than assuming dev_to_node() always returns a valid node index.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=221294
Suggested-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Fixes: 728bbc6cbff7 ("scsi: mpt3sas: Affinity high iops queues IRQs to local node")
Signed-off-by: Ivy Lopez <skunkolee@gmail.com>
---
 Changes in v2:
- Drop superfluous parentheses around dev_to_node() call.
- Split overly long ternary assignment line.
- Add Fixes: tag per checkpatch --strict suggestion.
- Link to v1: https://lore.kernel.org/r/20260817221300.43286-1-skunkolee@gmail.com

 drivers/scsi/mpt3sas/mpt3sas_base.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.c b/drivers/scsi/mpt3sas/mpt3sas_base.c
index 79052f2accbd..ce5a5882acc8 100644
--- a/drivers/scsi/mpt3sas/mpt3sas_base.c
+++ b/drivers/scsi/mpt3sas/mpt3sas_base.c
@@ -3238,7 +3238,10 @@ _base_assign_reply_queues(struct MPT3SAS_ADAPTER *ioc)
 		 * corresponding to high iops queues.
 		 */
 		if (ioc->high_iops_queues) {
-			mask = cpumask_of_node(dev_to_node(&ioc->pdev->dev));
+			int node = dev_to_node(&ioc->pdev->dev);
+
+			mask = (node == NUMA_NO_NODE) ?
+				cpu_online_mask : cpumask_of_node(node);
 			for (index = 0; index < ioc->high_iops_queues;
 			    index++) {
 				irq = pci_irq_vector(ioc->pdev, index);
-- 
2.55.0

Re: [PATCH v2] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] This problem wasn't introduced by this patch, but incorrect subtraction in the fall_back path causes unassigned CPUs in the msix table and a potential divide by zero if nr_msix evaluates to zero.
--

commit 9e90c3b985462523aebc40d2af67de713d3be849
Author: Ivy Lopez <skunkolee@gmail.com>

scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

dev_to_node() can return NUMA_NO_NODE (-1) on systems without NUMA
topology information for the PCI device, such as single-socket
boards that don't expose device-to-node affinity. Fall back to
cpu_online_mask when no NUMA node is available, rather than assuming
dev_to_node() always returns a valid node index.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=221294

> diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.c b/drivers/scsi/mpt3sas/mpt3sas_base.c
> index fed7aeffec58d..1af25a22611a1 100644
> --- a/drivers/scsi/mpt3sas/mpt3sas_base.c
> +++ b/drivers/scsi/mpt3sas/mpt3sas_base.c
> @@ -3238,7 +3238,10 @@ _base_assign_reply_queues(struct MPT3SAS_ADAPTER *ioc)
>  		 * corresponding to high iops queues.
>  		 */
>  		if (ioc->high_iops_queues) {
> -			mask = cpumask_of_node(dev_to_node(&ioc->pdev->dev));
> +			int node = dev_to_node(&ioc->pdev->dev);
> +
> +			mask = (node == NUMA_NO_NODE) ?
> +				cpu_online_mask : cpumask_of_node(node);
>  			for (index = 0; index < ioc->high_iops_queues;
>  			    index++) {
>  				irq = pci_irq_vector(ioc->pdev, index);

[Severity: High]
This isn't a bug introduced by this patch, but incorrect subtraction in the
fall_back path causes unassigned CPUs in the msix table and a potential
divide by zero.

In _base_assign_reply_queues(), the calculation evaluates incorrectly:

drivers/scsi/mpt3sas/mpt3sas_base.c:_base_assign_reply_queues() {
...
fall_back:
	cpu = cpumask_first(cpu_online_mask);
	nr_msix -= (ioc->high_iops_queues - iopoll_q_count);
	index = 0;

	list_for_each_entry(reply_q, &ioc->reply_queue_list, list) {
		unsigned int i, group = nr_cpus / nr_msix;
...
}

If iopoll_q_count > 0, the calculation nr_msix -= (ioc->high_iops_queues -
iopoll_q_count) evaluates to nr_msix - high_iops_queues + iopoll_q_count.

This mathematically adds iopoll_q_count instead of subtracting it, inflating
nr_msix. This causes group = nr_cpus / nr_msix to calculate a smaller group
size, leaving trailing CPUs mapped to vector 0.

Furthermore, if PCI MSI-X allocation limits nr_msix to exactly
high_iops_queues (e.g. system interrupt limits), nr_msix evaluates to 0.

This would trigger a hardware divide-by-zero exception when evaluating
nr_cpus / nr_msix.

Is it possible to rewrite this subtraction to avoid inflating nr_msix and
prevent a potential divide-by-zero?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260825190313.24013-1-skunkolee@gmail.com?part=1

Re: [PATCH v2] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

[John Garry]

Reviewed-by: John Garry <john.g.garry@oracle.com>

Re: [PATCH v2] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()

[Martin K. Petersen (Oracle)]

On Tue, 25 Aug 2026 13:03:13 -0600, Ivy Lopez wrote:

> dev_to_node() can return NUMA_NO_NODE (-1) on systems without NUMA
> topology information for the PCI device, such as single-socket
> boards that don't expose device-to-node affinity. Passing -1
> directly into cpumask_of_node() indexes node_to_cpumask_map[-1],
> an out-of-bounds array read caught by UBSAN:
> 
>   UBSAN: array-index-out-of-bounds in arch/x86/include/asm/topology.h:72:28
>   index -1 is out of range for type 'cpumask *[1024]'
> 
> [...]

Applied to 7.3/scsi-fixes, thanks!

[1/1] scsi: mpt3sas: avoid out-of-bounds cpumask_of_node() call in _base_assign_reply_queues()
      https://git.kernel.org/mkp/scsi/c/e0d26fe176a8

-- 
Martin K. Petersen