On Fri, Jun 19, 2026 at 12:33:35AM -0700, Shradha Gupta wrote:
> In mana driver, the number of IRQs allocated is capped by the
> min(num_cpu + 1, queue count). In cases, where the IRQ count is greater
> than the vcpu count, we want to utilize all the vCPUs, irrespective of
> their NUMA/core bindings.
>
> This is important, especially in the envs where number of vCPUs are so
> few that the softIRQ handling overhead on two IRQs on the same vCPU is
> much more than their overheads if they were spread across sibling vCPUs.
>
> This behaviour is more evident with dynamic IRQ allocation. Since MANA
> IRQs are assigned at a later stage compared to static allocation, other
> device IRQs may already be affinitized to the vCPUs. As a result, IRQ
> weights become imbalanced, causing multiple MANA IRQs to land on the
> same vCPU, while some vCPUs have none.
>
> In such cases when many parallel TCP connections are tested, the
> throughput drops significantly.
>
> We also studied the results of setting the affinity and hint to
> NULL in these cases, and observed that, with this logic if there are
> pre existing IRQs allocated on the VM(apart from MANA), during MANA
> IRQs allocation, it leads to clustering of the MANA queue IRQs again.
> These results can be seen through case 3 in the following data.
>
> Test envs:
> =======================================================
> Case 1: without this patch
> =======================================================
> 4 vcpu(2 cores), 5 MANA IRQs (1 HWC + 4 Queue)
>
> TYPE effective vCPU aff
> =======================================================
> IRQ0: HWC 0
> IRQ1: mana_q1 0
> IRQ2: mana_q2 2
> IRQ3: mana_q3 0
> IRQ4: mana_q4 3
>
> %soft on each vCPU(mpstat -P ALL 1) on receiver
> vCPU 0 1 2 3
> =======================================================
> pass 1: 38.85 0.03 24.89 24.65
> pass 2: 39.15 0.03 24.57 25.28
> pass 3: 40.36 0.03 23.20 23.17
>
> =======================================================
> Case 2: with this patch
> =======================================================
> 4 vcpu(2 cores), 5 MANA IRQs (1 HWC + 4 Queue)
>
> TYPE effective vCPU aff
> =======================================================
> IRQ0: HWC 0
> IRQ1: mana_q1 0
> IRQ2: mana_q2 1
> IRQ3: mana_q3 2
> IRQ4: mana_q4 3
>
> %soft on each vCPU(mpstat -P ALL 1) on receiver
> vCPU 0 1 2 3
> =======================================================
> pass 1: 15.42 15.85 14.99 14.51
> pass 2: 15.53 15.94 15.81 15.93
> pass 3: 16.41 16.35 16.40 16.36
>
> =======================================================
> Case 3: with affinity set to NULL
> =======================================================
> 4 vCPU(2 cores), 5 MANA IRQs (1 HWC + 4 Queue)
>
> TYPE effective vCPU aff
> =======================================================
> IRQ0: HWC 0
> IRQ1: mana_q1 2
> IRQ2: mana_q2 3
> IRQ3: mana_q3 2
> IRQ4: mana_q4 3
>
> =======================================================
> Throughput Impact(in Gbps, same env)
> =======================================================
> TCP conn with patch w/o patch aff NULL
> 20480 15.65 7.73 5.25
> 10240 15.63 8.93 5.77
> 8192 15.64 9.69 7.16
> 6144 15.64 13.16 9.33
> 4096 15.69 15.75 13.50
> 2048 15.69 15.83 13.61
> 1024 15.71 15.28 13.60
>
> Fixes: 755391121038 ("net: mana: Allocate MSI-X vectors dynamically")
> Cc: [email protected]
> Co-developed-by: Erni Sri Satya Vennela <[email protected]>
> Signed-off-by: Erni Sri Satya Vennela <[email protected]>
> Signed-off-by: Shradha Gupta <[email protected]>
> Reviewed-by: Haiyang Zhang <[email protected]>
> Reviewed-by: Simon Horman <[email protected]>
Reviewed-by: Yury Norov <[email protected]>
> ---
> Changes in v4
> * Add mana prefix on irq_affinity_*() in mana driver
> * Corrected grammar, comment for mana_irq_setup_linear()
> * added new line as per guidelines
> * added case 3 in commit message for when affinity is NULL
> ---
> Changes in v3
> * Optimize the comments in mana_gd_setup_dyn_irqs()
> * add more details in the dev_dbg for extra IRQs
> ---
> Changes in v2
> * Removed the unused skip_first_cpu variable
> * fixed exit condition in irq_setup_linear() with len == 0
> * changed return type of irq_setup_linear() as it will always be 0
> * removed the unnecessary rcu_read_lock() in irq_setup_linear()
> * added appropriate comments to indicate expected behaviour when
> IRQs are more than or equal to num_online_cpus()
> ---
> .../net/ethernet/microsoft/mana/gdma_main.c | 78 +++++++++++++++----
> 1 file changed, 64 insertions(+), 14 deletions(-)
>
> diff --git a/drivers/net/ethernet/microsoft/mana/gdma_main.c
> b/drivers/net/ethernet/microsoft/mana/gdma_main.c
> index a0fdd052d7f1..e8b7ffb47eb9 100644
> --- a/drivers/net/ethernet/microsoft/mana/gdma_main.c
> +++ b/drivers/net/ethernet/microsoft/mana/gdma_main.c
> @@ -210,6 +210,8 @@ static int mana_gd_query_max_resources(struct pci_dev
> *pdev)
> } else {
> /* If dynamic allocation is enabled we have already allocated
> * hwc msi
> + * Also, we make sure in this case the following is always true
> + * (num_msix_usable - 1 HWC) <= num_online_cpus()
> */
> gc->num_msix_usable = min(resp.max_msix, num_online_cpus() + 1);
> }
> @@ -1909,8 +1911,8 @@ void mana_gd_free_res_map(struct gdma_resource *r)
> * do the same thing.
> */
>
> -static int irq_setup(unsigned int *irqs, unsigned int len, int node,
> - bool skip_first_cpu)
> +static int mana_irq_setup_numa_aware(unsigned int *irqs, unsigned int len,
> + int node, bool skip_first_cpu)
> {
> const struct cpumask *next, *prev = cpu_none_mask;
> cpumask_var_t cpus __free(free_cpumask_var);
> @@ -1946,11 +1948,24 @@ static int irq_setup(unsigned int *irqs, unsigned int
> len, int node,
> return 0;
> }
>
> +/* must be called with cpus_read_lock() held */
> +static void mana_irq_setup_linear(unsigned int *irqs, unsigned int len)
> +{
> + int cpu;
> +
> + for_each_online_cpu(cpu) {
> + if (len == 0)
> + break;
> +
> + irq_set_affinity_and_hint(*irqs++, cpumask_of(cpu));
> + len--;
> + }
> +}
> +
> static int mana_gd_setup_dyn_irqs(struct pci_dev *pdev, int nvec)
> {
> struct gdma_context *gc = pci_get_drvdata(pdev);
> struct gdma_irq_context *gic;
> - bool skip_first_cpu = false;
> int *irqs, err, i, msi;
>
> irqs = kmalloc_objs(int, nvec);
> @@ -1958,10 +1973,12 @@ static int mana_gd_setup_dyn_irqs(struct pci_dev
> *pdev, int nvec)
> return -ENOMEM;
>
> /*
> + * In this function, num_msix_usable = HWC IRQ + Queue IRQ.
> + * nvec is only Queue IRQ (HWC already setup).
> * While processing the next pci irq vector, we start with index 1,
> * as IRQ vector at index 0 is already processed for HWC.
> * However, the population of irqs array starts with index 0, to be
> - * further used in irq_setup()
> + * further used in mana_irq_setup_numa_aware()
> */
> for (i = 1; i <= nvec; i++) {
> msi = i;
> @@ -1975,18 +1992,51 @@ static int mana_gd_setup_dyn_irqs(struct pci_dev
> *pdev, int nvec)
> }
>
> /*
> - * When calling irq_setup() for dynamically added IRQs, if number of
> - * CPUs is more than or equal to allocated MSI-X, we need to skip the
> - * first CPU sibling group since they are already affinitized to HWC IRQ
> + * When calling mana_irq_setup_numa_aware() for dynamically added IRQs,
> + * if number of CPUs is more than or equal to allocated MSI-X, we need
> to
> + * skip the first CPU sibling group since they are already affinitized
> to
> + * HWC IRQ
> */
> cpus_read_lock();
> - if (gc->num_msix_usable <= num_online_cpus())
> - skip_first_cpu = true;
> + if (gc->num_msix_usable <= num_online_cpus()) {
> + err = mana_irq_setup_numa_aware(irqs, nvec, gc->numa_node,
> + true);
> + if (err) {
> + cpus_read_unlock();
> + goto free_irq;
> + }
> + } else {
> + /*
> + * When num_msix_usable are more than num_online_cpus, our
> + * queue IRQs should be equal to num of online vCPUs.
> + * We try to make sure queue IRQs spread across all vCPUs.
> + * In such a case NUMA or CPU core affinity does not matter.
> + * Note: in this case the total mana IRQ should always be
> + * num_online_cpus + 1. The first HWC IRQ is already handled
> + * in HWC setup calls
> + * However, if CPUs went offline since num_msix_usable was
> + * computed, queue IRQs will be more than num_online_cpus().
> + * In such cases remaining extra IRQs will retain their default
> + * affinity.
> + */
> + int first_unassigned = num_online_cpus();
>
> - err = irq_setup(irqs, nvec, gc->numa_node, skip_first_cpu);
> - if (err) {
> - cpus_read_unlock();
> - goto free_irq;
> + if (nvec > first_unassigned) {
> + char buf[32];
> +
> + if (first_unassigned == nvec - 1)
> + snprintf(buf, sizeof(buf), "%d",
> + first_unassigned);
> + else
> + snprintf(buf, sizeof(buf), "%d-%d",
> + first_unassigned, nvec - 1);
> +
> + dev_dbg(&pdev->dev,
> + "MANA IRQ indices #%s will retain the default
> CPU affinity\n",
> + buf);
> + }
> +
> + mana_irq_setup_linear(irqs, nvec);
> }
>
> cpus_read_unlock();
> @@ -2041,7 +2091,7 @@ static int mana_gd_setup_irqs(struct pci_dev *pdev, int
> nvec)
> nvec -= 1;
> }
>
> - err = irq_setup(irqs, nvec, gc->numa_node, false);
> + err = mana_irq_setup_numa_aware(irqs, nvec, gc->numa_node, false);
> if (err) {
> cpus_read_unlock();
> goto free_irq;
>
> base-commit: 96e7f9122aae0ed000ee321f324b812a447906d9
> --
> 2.34.1
