From: Xiaoyu Li xiaoyu.li@corigine.com
Before the referenced commit, when we requested a certain number of interrupts, if we could not meet the requirements, the number of interrupts supported by the hardware would be returned. But after the referenced commit, if the hardware failed to meet the requirements, the error of invalid argument would be directly returned, which caused a regression in the nfp driver preventing probing to complete.
Fixes: bab65e48cb06 ("PCI/MSI: Sanitize MSI-X checks") Cc: stable@vger.kernel.org Signed-off-by: Xiaoyu Li xiaoyu.li@corigine.com Acked-by: Simon Horman simon.horman@corigine.com Signed-off-by: Louis Peens louis.peens@corigine.com --- drivers/net/ethernet/netronome/nfp/nfp_net_common.c | 6 ++++++ 1 file changed, 6 insertions(+)
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c index 62f0bf91d1e1..0e4cab38f075 100644 --- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c +++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c @@ -370,6 +370,12 @@ nfp_net_irqs_alloc(struct pci_dev *pdev, struct msix_entry *irq_entries, { unsigned int i; int got_irqs; + int max_irqs; + + max_irqs = pci_msix_vec_count(pdev); + if (max_irqs < 0) + return max_irqs; + wanted_irqs = min_t(unsigned int, max_irqs, wanted_irqs);
for (i = 0; i < wanted_irqs; i++) irq_entries[i].entry = i;
On Wed, Mar 15, 2023 at 02:17:33PM +0200, Louis Peens wrote:
From: Xiaoyu Li xiaoyu.li@corigine.com
Before the referenced commit, when we requested a certain number of interrupts, if we could not meet the requirements, the number of interrupts supported by the hardware would be returned. But after the referenced commit, if the hardware failed to meet the requirements, the error of invalid argument would be directly returned, which caused a regression in the nfp driver preventing probing to complete.
Please don't break lines. You have upto 80 chars per-line.
Fixes: bab65e48cb06 ("PCI/MSI: Sanitize MSI-X checks") Cc: stable@vger.kernel.org Signed-off-by: Xiaoyu Li xiaoyu.li@corigine.com Acked-by: Simon Horman simon.horman@corigine.com Signed-off-by: Louis Peens louis.peens@corigine.com
drivers/net/ethernet/netronome/nfp/nfp_net_common.c | 6 ++++++ 1 file changed, 6 insertions(+)
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c index 62f0bf91d1e1..0e4cab38f075 100644 --- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c +++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c @@ -370,6 +370,12 @@ nfp_net_irqs_alloc(struct pci_dev *pdev, struct msix_entry *irq_entries, { unsigned int i; int got_irqs;
- int max_irqs;
- max_irqs = pci_msix_vec_count(pdev);
- if (max_irqs < 0)
return max_irqs;- wanted_irqs = min_t(unsigned int, max_irqs, wanted_irqs);
1. It looks like you need to fix your nfp_net_irqs_alloc() to provide valid wanted_irqs from the beginning. 2. Both wanted_irqs and min_irqs are wrong type and should be int and not unsigned int.
Thanks
for (i = 0; i < wanted_irqs; i++) irq_entries[i].entry = i; -- 2.34.1
On Thu, 16 Mar 2023 13:09:43 +0200 Leon Romanovsky wrote:
On Wed, Mar 15, 2023 at 02:17:33PM +0200, Louis Peens wrote:
From: Xiaoyu Li xiaoyu.li@corigine.com
Before the referenced commit, when we requested a certain number of interrupts, if we could not meet the requirements, the number of interrupts supported by the hardware would be returned. But after the referenced commit, if the hardware failed to meet the requirements, the error of invalid argument would be directly returned, which caused a regression in the nfp driver preventing probing to complete.
Please don't break lines. You have upto 80 chars per-line.
72 I think, git adds an indentation. Not that I personally care about "not using full lines".
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c index 62f0bf91d1e1..0e4cab38f075 100644 --- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c +++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c @@ -370,6 +370,12 @@ nfp_net_irqs_alloc(struct pci_dev *pdev, struct msix_entry *irq_entries, { unsigned int i; int got_irqs;
- int max_irqs;
- max_irqs = pci_msix_vec_count(pdev);
- if (max_irqs < 0)
return max_irqs;- wanted_irqs = min_t(unsigned int, max_irqs, wanted_irqs);
- It looks like you need to fix your nfp_net_irqs_alloc() to provide
valid wanted_irqs from the beginning.
Right, why do you have this problem in the first place? Could you provide some concrete numbers?
On Thu, Mar 16, 2023 at 02:27:10PM -0700, Jakub Kicinski wrote:
On Thu, 16 Mar 2023 13:09:43 +0200 Leon Romanovsky wrote:
On Wed, Mar 15, 2023 at 02:17:33PM +0200, Louis Peens wrote:
From: Xiaoyu Li xiaoyu.li@corigine.com
Before the referenced commit, when we requested a certain number of interrupts, if we could not meet the requirements, the number of interrupts supported by the hardware would be returned. But after the referenced commit, if the hardware failed to meet the requirements, the error of invalid argument would be directly returned, which caused a regression in the nfp driver preventing probing to complete.
Please don't break lines. You have upto 80 chars per-line.
72 I think, git adds an indentation. Not that I personally care about "not using full lines".
I care about typography.
Thanks
On Wed, 15 Mar 2023 14:17:33 +0200 Louis Peens wrote:
From: Xiaoyu Li xiaoyu.li@corigine.com
Before the referenced commit, when we requested a certain number of interrupts, if we could not meet the requirements, the number of interrupts supported by the hardware would be returned. But after the referenced commit, if the hardware failed to meet the requirements, the error of invalid argument would be directly returned, which caused a regression in the nfp driver preventing probing to complete.
Fixes: bab65e48cb06 ("PCI/MSI: Sanitize MSI-X checks") Cc: stable@vger.kernel.org Signed-off-by: Xiaoyu Li xiaoyu.li@corigine.com
Thomas, is this an expected side effect?
The commit message of bab65e48cb06 ("PCI/MSI: Sanitize MSI-X checks") makes it sound like only a harmless refactoring was intended.
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c index 62f0bf91d1e1..0e4cab38f075 100644 --- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c +++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c @@ -370,6 +370,12 @@ nfp_net_irqs_alloc(struct pci_dev *pdev, struct msix_entry *irq_entries, { unsigned int i; int got_irqs;
- int max_irqs;
- max_irqs = pci_msix_vec_count(pdev);
- if (max_irqs < 0)
return max_irqs;- wanted_irqs = min_t(unsigned int, max_irqs, wanted_irqs);
for (i = 0; i < wanted_irqs; i++) irq_entries[i].entry = i;
linux-stable-mirror@lists.linaro.org
-
Jakub Kicinski -
Leon Romanovsky -
Louis Peens