| Message ID | 20241107115002.413358-1-mheib@redhat.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | d9e2e290f7142d4c67c05ebbe37388d54a66c6c5 |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net] nfp: use irq_update_affinity_hint() | expand |
On Thu, Nov 07, 2024 at 01:50:02PM +0200, Mohammad Heib wrote: > irq_set_affinity_hint() is deprecated, Use irq_update_affinity_hint() > instead. This removes the side-effect of actually applying the affinity. > > The driver does not really need to worry about spreading its IRQs across > CPUs. The core code already takes care of that. when the driver applies the > affinities by itself, it breaks the users' expectations: > > 1. The user configures irqbalance with IRQBALANCE_BANNED_CPULIST in > order to prevent IRQs from being moved to certain CPUs that run a > real-time workload. > > 2. nfp device reopening will resets the affinity > in nfp_net_netdev_open(). > > 3. nfp has no idea about irqbalance's config, so it may move an IRQ to > a banned CPU. The real-time workload suffers unacceptable latency. > > Signed-off-by: Mohammad Heib <mheib@redhat.com> Reviewed-by: Simon Horman <horms@kernel.org>
On Thu, Nov 07, 2024 at 01:50:02PM +0200, Mohammad Heib wrote: > irq_set_affinity_hint() is deprecated, Use irq_update_affinity_hint() > instead. This removes the side-effect of actually applying the affinity. > > The driver does not really need to worry about spreading its IRQs across > CPUs. The core code already takes care of that. when the driver applies the > affinities by itself, it breaks the users' expectations: > > 1. The user configures irqbalance with IRQBALANCE_BANNED_CPULIST in > order to prevent IRQs from being moved to certain CPUs that run a > real-time workload. > > 2. nfp device reopening will resets the affinity > in nfp_net_netdev_open(). > > 3. nfp has no idea about irqbalance's config, so it may move an IRQ to > a banned CPU. The real-time workload suffers unacceptable latency. > > Signed-off-by: Mohammad Heib <mheib@redhat.com> > --- > drivers/net/ethernet/netronome/nfp/nfp_net_common.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c > index 6e0929af0f72..98e098c09c03 100644 > --- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c > +++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c > @@ -829,7 +829,7 @@ nfp_net_prepare_vector(struct nfp_net *nn, struct nfp_net_r_vector *r_vec, > return err; > } > > - irq_set_affinity_hint(r_vec->irq_vector, &r_vec->affinity_mask); > + irq_update_affinity_hint(r_vec->irq_vector, &r_vec->affinity_mask); > > nn_dbg(nn, "RV%02d: irq=%03d/%03d\n", idx, r_vec->irq_vector, > r_vec->irq_entry); > @@ -840,7 +840,7 @@ nfp_net_prepare_vector(struct nfp_net *nn, struct nfp_net_r_vector *r_vec, > static void > nfp_net_cleanup_vector(struct nfp_net *nn, struct nfp_net_r_vector *r_vec) > { > - irq_set_affinity_hint(r_vec->irq_vector, NULL); > + irq_update_affinity_hint(r_vec->irq_vector, NULL); > nfp_net_napi_del(&nn->dp, r_vec); > free_irq(r_vec->irq_vector, r_vec); > } Looks good to me, thank you! Reviewed-by: Louis Peens <louis.peens@corigine.com> > -- > 2.34.3 >
Hello: This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <kuba@kernel.org>: On Thu, 7 Nov 2024 13:50:02 +0200 you wrote: > irq_set_affinity_hint() is deprecated, Use irq_update_affinity_hint() > instead. This removes the side-effect of actually applying the affinity. > > The driver does not really need to worry about spreading its IRQs across > CPUs. The core code already takes care of that. when the driver applies the > affinities by itself, it breaks the users' expectations: > > [...] Here is the summary with links: - [net] nfp: use irq_update_affinity_hint() https://git.kernel.org/netdev/net-next/c/d9e2e290f714 You are awesome, thank you!
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c index 6e0929af0f72..98e098c09c03 100644 --- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c +++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c @@ -829,7 +829,7 @@ nfp_net_prepare_vector(struct nfp_net *nn, struct nfp_net_r_vector *r_vec, return err; } - irq_set_affinity_hint(r_vec->irq_vector, &r_vec->affinity_mask); + irq_update_affinity_hint(r_vec->irq_vector, &r_vec->affinity_mask); nn_dbg(nn, "RV%02d: irq=%03d/%03d\n", idx, r_vec->irq_vector, r_vec->irq_entry); @@ -840,7 +840,7 @@ nfp_net_prepare_vector(struct nfp_net *nn, struct nfp_net_r_vector *r_vec, static void nfp_net_cleanup_vector(struct nfp_net *nn, struct nfp_net_r_vector *r_vec) { - irq_set_affinity_hint(r_vec->irq_vector, NULL); + irq_update_affinity_hint(r_vec->irq_vector, NULL); nfp_net_napi_del(&nn->dp, r_vec); free_irq(r_vec->irq_vector, r_vec); }
irq_set_affinity_hint() is deprecated, Use irq_update_affinity_hint() instead. This removes the side-effect of actually applying the affinity. The driver does not really need to worry about spreading its IRQs across CPUs. The core code already takes care of that. when the driver applies the affinities by itself, it breaks the users' expectations: 1. The user configures irqbalance with IRQBALANCE_BANNED_CPULIST in order to prevent IRQs from being moved to certain CPUs that run a real-time workload. 2. nfp device reopening will resets the affinity in nfp_net_netdev_open(). 3. nfp has no idea about irqbalance's config, so it may move an IRQ to a banned CPU. The real-time workload suffers unacceptable latency. Signed-off-by: Mohammad Heib <mheib@redhat.com> --- drivers/net/ethernet/netronome/nfp/nfp_net_common.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)